# MOLECULAR CATALYSTS FOR CO2 FIXATION/REDUCTION

EDITED BY : Hitoshi Ishida, Charles Machan, Marc Robert and Nobuharu Iwasawa PUBLISHED IN : Frontiers in Chemistry

### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-622-8 DOI 10.3389/978-2-88963-622-8

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# MOLECULAR CATALYSTS FOR CO2 FIXATION/REDUCTION

Topic Editors: Hitoshi Ishida, Kitasato University, Japan Charles Machan, University of Virginia, United States Marc Robert, Université Paris Diderot, France Nobuharu Iwasawa, Tokyo Institute of Technology, Japan

Citation: Ishida, H., Machan, C., Robert, M., Iwasawa, N., eds. (2020). Molecular Catalysts for CO2 Fixation/Reduction. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-622-8

# Table of Contents

	- *Diimine-Diphosphine Cu(I) Complexes as Photosensitizers* Yasuomi Yamazaki, Takayuki Onoda, Jun Ishikawa, Shota Furukawa, Chinatsu Tanaka, Tomoya Utsugi and Taro Tsubomura

Yusuke Tamaki, Kazuma Tokuda, Yasuomi Yamazaki, Daiki Saito, Yutaro Ueda and Osamu Ishitani

*38 Synthesis of a Redox-Active NNP-Type Pincer Ligand and its Application to Electrocatalytic CO2 Reduction With First-Row Transition Metal Complexes*

Kallol Talukdar, Asala Issa and Jonah W. Jurss


Hiroyuki Takeda, Yu Monma, Haruki Sugiyama, Hidehiro Uekusa and Osamu Ishitani


Hideki Ohtsu, Tsubasa Saito and Kiyoshi Tsuge

*139 An Investigation of Electrocatalytic CO2 Reduction Using a Manganese Tricarbonyl Biquinoline Complex* Meaghan McKinnon, Veronika Belkina, Ken T. Ngo, Mehmed Z. Ertem, David C. Grills and Jonathan Rochford

*153 On-Surface Modification of Copper Cathodes by Copper(I)-Catalyzed Azide Alkyne Cycloaddition and CO2 Reduction in Organic Environments* Ryota Igarashi, Ryuji Takeuchi, Kazuyuki Kubo, Tsutomu Mizuta and Shoko Kume

# Editorial: Molecular Catalysts for CO<sup>2</sup> Fixation/Reduction

### Hitoshi Ishida<sup>1</sup> \*, Charles Machan<sup>2</sup> , Marc Robert <sup>3</sup> and Nobuharu Iwasawa<sup>4</sup>

<sup>1</sup> Department of Chemistry, Graduate School of Science, Kitasato University, Sagamihara, Japan, <sup>2</sup> Department of Chemistry, University of Virginia, Charlottesville, VA, United States, <sup>3</sup> Laboratoire d'Electrochimie Moléculaire (LEM), Université de Paris, Paris, France, <sup>4</sup> School of Science and Engineering, Tokyo Institute of Technology, Tokyo, Japan

Keywords: CO<sup>2</sup> reduction, CO<sup>2</sup> fixation, electrocatalysis, photocatalysis, artificial photosynthesis

**Editorial on the Research Topic**

### **Molecular Catalysts for CO**<sup>2</sup> **Fixation/Reduction**

Conversion of CO<sup>2</sup> into useful chemicals and fuels becomes more and more important as increasing CO<sup>2</sup> emissions continue to worsen the effects of climate change. Catalytic CO<sup>2</sup> reduction and fixation are important components of artificial photosynthesis (the use of renewable energy to generate useful chemicals and fuels), which is vigorously studied as part of the solution to the environmental and energy problems that arise from its continuously elevating atmospheric concentration. However, the development of chemical processes remains a significant challenge for chemists. Molecular inorganic complexes, which can be utilized as homogeneous catalysts, have several advantages: well-defined structures that are readily characterized, catalytic mechanisms that can be elucidated with high fidelity, and the opportunity to optimize catalytic activities through molecular design. In this Research Topic, we present ten original research articles and three review articles dealing with electrochemical and photochemical catalysis for CO<sup>2</sup> reduction or fixation.

Edited and reviewed by: Luís D. Carlos, University of Aveiro, Portugal

\*Correspondence: Hitoshi Ishida ishida@sci.kitasato-u.ac.jp

### Specialty section:

This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry

Received: 06 January 2020 Accepted: 17 January 2020 Published: 25 February 2020

### Citation:

Ishida H, Machan C, Robert M and Iwasawa N (2020) Editorial: Molecular Catalysts for CO<sup>2</sup> Fixation/Reduction. Front. Chem. 8:59. doi: 10.3389/fchem.2020.00059

The metal complexes, fac-[Mn(bpy)(CO)3Br] and fac-[Re(bpy)(CO)3Cl] (bpy = 2,2′ -bipyridine), are known to be excellent catalysts for electrochemical CO<sup>2</sup> reduction. Rotundo et al. synthesized 14 new Mn(I) and Re(I) complexes with structurally diverse 2,2′ -bipyridyl derivatives and examined their catalytic activities by using cyclic voltammetry and controlled potential electrolysis. They reported that the complexes with only electron-withdrawing groups (e.g., -CF3, -CN) lost their catalytic activities. On the other hand, the complexes with the symmetric -NMe<sup>2</sup> substituted and push–pull systems (containing both -NMe<sup>2</sup> and -CF3) displayed electrocatalytic current enhancement. McKinnon et al. synthesized the manganese(I) complex fac-[Mn(bqn)(CO)3(CH3CN)]<sup>+</sup> with a 2,2′ -biquinoline (bqn) ligand, which had unique steric and electronic properties. They examined the catalytic activities for electrochemical CO<sup>2</sup> reduction and further compared them to the efficiencies of the corresponding bpy, 1,10-phenanthroline (phen), and 2,9-dimethyl-1,10-phenanthroline (dmphen) derivatives. In the reaction mechanisms, the one-electron reduced species of the bpy complex dimerized, but the bqn ligand hindered the dimer formation to favorably afford two-electron reduced species. Talukdar et al. synthesized a new NNP-type pincer ligand, 8-(diphenylphosphaneyl)-2-(pyridin-2′ -yl)quinolone, and the corresponding Fe, Co, Ni, and Zn complexes. Due to earth abundance, the development of molecular catalysts with first-row transition metals is desirable. They reported that the cobalt(II) complex showed the catalytic efficiency to produce CO in CO2-saturated DMF containing 5% water; the catalytic efficiencies for CO of the active complexes in descending order were Co2<sup>+</sup> >> Ni2<sup>+</sup> > Fe2+. The Zn(II) complex did not show catalytic activity. Nichols and Machan reviewed the secondary-sphere effects in the enzymes and the molecular catalysts for electrochemical CO<sup>2</sup> reduction. In the CO2-related enzymes (e.g., carbon monoxide dehydrogenase, formic acid dehydrogenase), some amino acid residues around the catalytic centers can act as the

pendent proton donors/shuttles, charged groups, etc. They discussed standard methodologies for estimating the effective catalytic overpotential and the maximal turnover frequency, and reviewed the development of secondary-sphere strategies for the classic molecular electrocatalysts, [Fe(tetraphenylporphyrin)]+, [Ni(cyclam)]2+, Mn(bpy)(CO)3X, and Re(bpy)(CO)3X (X = solvent or halide). Igarashi et al. reported the surface modification of copper cathodes with ethynyl and azide monomers. Metal copper cathodes are known to be effective for electrochemical CO<sup>2</sup> reduction. The introduction of organic structures onto catalytically active metal surfaces has recently received attention because it often enhances activity and selectivity for specific products. They reported that an increase in ethylene production was observed for the organic molecule-modified copper electrodes in electrochemical CO<sup>2</sup> reduction.

Photocatalytic CO<sup>2</sup> reduction can be accomplished with the same molecular catalysts employed in electrochemistry by using appropriate redox photosensitizers and sacrificial electron donors instead of an electrode. The most widely used redox photosensitizer is [Ru(bpy)3] <sup>2</sup>+, which exhibits strong absorption in the visible light region and has a long lifetime in the triplet metal-to-ligand charge transfer ( <sup>3</sup>MLCT) excited state. Tamaki et al. synthesized a new ruthenium(II) picolinate complex, [Ru(dmb)2(pic)]<sup>+</sup> (dmb = 4,4′ -dimethyl-2,2′ -bipyridine; Hpic = picolinic acid), with a wider wavelength range of visible-light absorption (λabs < 670 nm) and a strong reduction ability in the oneelectron reduced state (Ered = −1.86 V vs. Ag/AgNO3). They demonstrated that photocatalysis using [Ru(dmb)2(pic)]<sup>+</sup> as the redox photosensitizer combined with a Re(I) catalyst reduced CO<sup>2</sup> to CO under red-light irradiation (λex > 600 nm). In contrast, [Ru(dmb)3] <sup>2</sup>+, which could not absorb light at λ > 560 nm, was not capable of working as the redox photosensitizer under the same conditions. The molar extinction coefficient is also an important factor for a redox photosensitizer. A cyclometalated iridium(III) complex having 2-(pyren-1-yl)-4-methylquinoline ligands [Ir(pyr)] has a strong absorption band in the visible region (ε at 444 nm = 67,000 M−<sup>1</sup> cm−<sup>1</sup> ) but does not act as a photosensitizer for photochemical CO<sup>2</sup> reduction reactions in the presence of triethylamine as an electron donor. Kuramochi and Ishitani report that the photochemical CO<sup>2</sup> reduction catalyzed by trans(Cl)-Ru(dmb)(CO)2Cl<sup>2</sup> could proceed in a combination of [Ir(pyr)] as the photosensitizer and 1,3-dimethyl-2-(ohydroxyphenyl)-2,3-dihydro-1H-benzo[d]imidazole (BI(OH)H) as the electron donor. They further synthesized a binuclear complex in which Ir(pyr) and Ru combined via an ethylene bridge. The development of efficient redox photosensitizers based on the earth-abundant metal ions as an alternative toward noble- and/or rare-metal-based photosensitizers is of continuing interest for the field. The heteroleptic Cu(I) phenanthroline complexes, such as [Cu<sup>I</sup> (dmp)(P)2] <sup>+</sup> (dmp = 2,9-dimethyl-1,10-phenanthroline, P = phosphine ligand), are known to have long lifetimes and show strong MLCT excited state emission even in a solution at room temperature due to the Cu(I) center's d <sup>10</sup> configuration. Yamazaki et al. synthesized heteroleptic diimine-diphosphine Cu(I) complexes, [Cu<sup>I</sup> (dmpp)(diphos)]<sup>+</sup> [dmpp = 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline; diphos = methylene chains (n = 2–4) linked bis(diphenyl phosphine)]. They investigated the effects of the number of methylene chains on the photocatalytic CO<sup>2</sup> reduction by using fac-[Re(bpy)(CO)3Br] as a catalyst and 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) as an electron donor. The most effective redox photosensitizer was the 1,4-bis(diphenylphosphino)butane (dppb; n = 4) complex. They discussed the photocatalytic results in the context of the P–Cu–P angles in the photosensitizer. Takeda et al. developed visible-light responsive Cu(I)-complex photosensitizers by introducing various aromatic substituents at the 4,7-positions of a 2,9-dimethyl-1,10-phenanthroline (dmp) ligand in a heteroleptic [Cu<sup>I</sup> (dmp)(DPEphos)]+-type complexes (DPEphos = [2-(diphenylphosphino)phenyl]ether). Introduction of biphenyl groups on the dmp ligand enhanced the molar extinction coefficient of the MLCT band. They further performed the photocatalyzed CO<sup>2</sup> reduction by an earth-abundant metal complex, [Fe(dmp)2(NCS)2]. Ohtsu et al. synthesized a ruthenium(II) NAD+-type complex, [Ru(bpy)2(Me-pn)]2<sup>+</sup> (Me-pn = 2-Methyl-6-(pyridin-2-yl)- 1,5-naphthyridine). The common photosensitizers act as a one-electron reservoir. But they demonstrated the formation of the corresponding two-electron reduced ruthenium(II) complex, [Ru(bpy)2(Me-pnHH)]2<sup>+</sup> (Me-pnHH (2-methyl-6-(pyridin-2 yl)-1,4-dihydro-1,5-naphthyridine) is the NADH-type ligand), by the photo-induced hydrogenation reaction of [Ru(bpy)2(Mepn)]2+. The complex is a two-electron reservoir and will be applicable to photocatalytic CO<sup>2</sup> reduction as an efficient redox photosensitizer.

Fixation of CO<sup>2</sup> into organic molecules is important to obtain useful and valuable chemicals. Fujihara and Tsuji reviewed the transition metal-catalyzed carboxylation reactions of organic substrates with CO<sup>2</sup> via allyl metal intermediates. They summarized the carboxylation reactions via transmetalation, catalytic carboxylation reactions using allyl electrophiles and suitable reducing agents, and then the catalytic carboxylation reactions via additional reactions affording allyl metal intermediates. Murata et al. reported the visible-light-driven hydrocarboxylation of alkenes with CO<sup>2</sup> by a Rh(I) catalyst with [Ru(bpy)3] <sup>2</sup><sup>+</sup> (a redox photosensitizer) and BIH or BI(OH)H (an electron donor). The mechanistic study revealed that the catalysis required two photon-driven steps: the carboxylation of the hydrometalated Rh(I) complex on one hand, and the photo-induced generation of the Rh(III) dihydride complex on the other hand. The latter is an electron transfer process that requires the electron donor. The former is an energy transfer step, which would be sensitized by [Ru(bpy)3] <sup>2</sup>+. Interestingly, they found that the incorporation of the cyclometalated Ir(III) complex as a second photosensitizer with [Ru(bpy)3] <sup>2</sup><sup>+</sup> photosensitizer also resulted in the promotion of the reduction process. Fu et al. reviewed carbon capture and utilization (CCU) strategies. Although carbon capture and storage (CCS) strategies have been regarded as one of the promising options for controlling continuously elevating atmospheric CO<sup>2</sup> concentrations, desorption and compression of CO<sup>2</sup> need extra energy input. In the CCU strategy, the CO<sup>2</sup> absorbents trap CO<sup>2</sup> to yield the CO<sup>2</sup> adducts, which can be reacted as transcarboxylation and transformation agents. For example, CO<sup>2</sup> capture and Rh- or Ru-catalyzed CO<sup>2</sup> adducts hydrogenation was introduced.

Conversion of CO<sup>2</sup> into useful chemicals by CO<sup>2</sup> reduction/fixation continues to attract significant interest and research effort. Through the remarkable creativity demonstrated by the authors, this collection of papers (and beyond) is clear evidence that CO<sup>2</sup> fixation and molecular catalysis of CO<sup>2</sup> reduction are and will remain very active and promising fields, promoting fundamental advances. To discover more efficient catalysts, mechanistic investigations on ligand effects, as well as the modification of the second sphere in the metal complexes, will continue to expand the impact of these efforts, with the help of many various complementary in situ spectroscopic techniques, such as IR, Raman, EPR, mass, and X-ray absorption spectroscopies. Developing new hybrid systems, such as those that combine the most promising molecular catalysts with semiconductor materials, will also be important developments in the near future, merging the best attributes of homogeneous and heterogeneous catalysis. We hope that this Research Topic will contribute to stimulating the next wave of research efforts in the construction of new artificial photosynthetic systems.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# FUNDING

HI was supported by a Grant-in-Aid for Scientific Research (C) from the Ministry of Education, Culture, Sports, Science and Technology (17K05815), and a Grant-in-Aid for Scientific Research on Innovative Areas, Innovations for Light-Energy Conversion (18H05176), from the Japan Society for the Promotion of Science (JSPS).

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Ishida, Machan, Robert and Iwasawa. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Photocatalytic CO<sup>2</sup> Reduction Using Various Heteroleptic Diimine-Diphosphine Cu(I) Complexes as Photosensitizers

Yasuomi Yamazaki, Takayuki Onoda, Jun Ishikawa, Shota Furukawa, Chinatsu Tanaka, Tomoya Utsugi and Taro Tsubomura\*

*Department of Materials and Life Science, Seikei University, Musashino-shi, Japan*

The development of efficient redox-photosensitizers based on the earth-abundant metal ions as an alternative toward noble- and/or rare-metal based photosensitizers is very desirable. In recent years, heteroleptic diimine-diphosphine Cu(I) complexes have been well investigated as one of the most remarkable candidates because of their great potentials as efficient photosensitizers. Here, we investigated the effects of the structure of the diphosphine ligands on the photosensitizing abilities using a series of Cu(I) complexes bearing 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (dmpp) and various diphosphine ligands in order to explore the suitable structure for the photosensitizing reactions. The number of methylene chains between the two phosphorous atoms in the diphosphine ligands was systematically changed from two to four, and the relationship between the length of the carbon chains and the photosensitizing abilities were investigated by conducting photocatalytic CO<sup>2</sup> reduction with the Cu(I) complexes as photosensitizers. Turnover frequencies of the CO<sup>2</sup> reduction drastically increased with increasing the length of the carbon chains. The systematic study herein reported suggests that the large P-Cu-P angles should be one of the most important factors for enhancing the photosensitizing abilities.

Keywords: CO2 reduction, photocatalytic reaction, Cu(I) complexes, redox-photosensitizer, diphosphine ligands

# INTRODUCTION

In the last few decades, photocatalytic CO<sup>2</sup> reduction has been widely investigated due to emerging concerns about the serious environmental problems, e.g., global warming and depletion of carbon and energy resources. Metal complexes have played an important role both as photosensitizers and as CO<sup>2</sup> reduction catalysts in this research area, owing to their suitable photophysical and photochemical properties, high reaction selectivity, and flexibility in the molecular design (Morris et al., 2009; Windle and Perutz, 2012; Sahara and Ishitani, 2015; Yamazaki et al., 2015; Kuramochi et al., 2018). Noble- and/or rare-metal complexes in particular have shown quite high efficiency and durability. For instance, the photocatalytic systems constructed with a Re(I) tricarbonyl complex as a CO2-reduction catalyst and a ring-shaped Re(I) multinuclear complex as a photosensitizer, can trigger CO<sup>2</sup> reduction with tremendously high quantum yield (8CO = 82%) (Morimoto et al., 2013; Rohacova and Ishitani, 2016, 2017). Another example is a system using a Ru(II)-Re(I) multinuclear complexes, (Gholamkhass et al., 2005; Sato et al., 2007; Koike et al., 2009; Tamaki et al., 2012;

### Edited by:

*Hitoshi Ishida, Kitasato University, Japan*

### Reviewed by:

*Yi Du, University of Wollongong, Australia Francesco Nastasi, University of Messina, Italy*

> \*Correspondence: *Taro Tsubomura tsubomura@st.seikei.ac.jp*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *30 January 2019* Accepted: *08 April 2019* Published: *30 April 2019*

### Citation:

*Yamazaki Y, Onoda T, Ishikawa J, Furukawa S, Tanaka C, Utsugi T and Tsubomura T (2019) Photocatalytic CO*2 *Reduction Using Various Heteroleptic Diimine-Diphosphine Cu(I) Complexes as Photosensitizers. Front. Chem. 7:288. doi: 10.3389/fchem.2019.00288*

**8**

Yamazaki et al. CO2 Reduction Using Cu(I) Complexes

Tamaki et al., 2013a; Kato et al., 2015; Ohkubo et al., 2016; Tamaki and Ishitani, 2017; Yamazaki and Ishitani, 2018) which shows both high efficiency and durability, i.e., up to 50% of 8CO and more than 3,000 of the turnover number (TONCO) (Tamaki et al., 2013a; Tamaki and Ishitani, 2017; Yamazaki and Ishitani, 2018). Though the fundamental researches using noble- and/or rare-metals must be quite important in order to fully understand the reaction mechanism of the CO<sup>2</sup> reduction and to explore strategies of molecular design for efficient components in photocatalytic systems, in the future, such metals which lie under the ground in small amounts should be replaced by earthabundant elements because the amount of emission of CO<sup>2</sup> relating to consumption of fossil fuels is quite large and is increasing year by year (Takeda et al., 2016a).

In recent years, heteroleptic diimine-diphosphine Cu(I) complexes have been reported as an alternative toward nobleand/or rare-metal based photosensitizers (Takeda et al., 2016b, 2018; Heberle et al., 2017; Rosas-Hernández et al., 2017; McCullough et al., 2018; Zhang et al., 2018). Some of them show not only high efficiency but also high durability equal to or higher than those of noble- and/or rare-metal complexes, and the highest 8CO and TONCO in the photocatalytic systems for CO<sup>2</sup> reduction using Cu(I) photosensitizers were 57% and >1,300, respectively (Takeda et al., 2018). These reports clearly indicate that Cu(I) complexes should be powerful candidates not just as an alternative toward photosensitizers based on noble- and/or rare-metal ions, but as one of the most efficient photosensitizers. Therefore, further investigation using series of Cu(I) complexes to clarify the relationship between the molecular structure and photosensitizing abilities should be useful to explore the suitable molecular design for the efficient Cu(I) complex photosensitizers.

We previously reported photophysical properties of series of diimine-diphosphine Cu(I) complexes bearing 1,10 phenanthroline derivatives and various bidentate phosphine ligands (Saito et al., 2006; Tsubomura et al., 2015; Nishikawa et al., 2017). In particular, a series of the Cu(I) complexes having 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (dmpp) showed high molar extinction coefficients and emission quantum yields compared with those without phenyl groups at 4,7 positions (Tsubomura et al., 2015). The photophysical properties of the dmpp complexes were strongly affected by the structure of the bidentate phosphine ligands; wavelength of both absorption bands and emission maxima were blue-shifted and emission lifetimes became longer with increasing the length of the methylene chains between the two phosphorous atoms, likely due to the difference in the bite angles of the chelatephosphine ligands (P-Cu-P angles). The strong absorption abilities and long emission lifetimes should be useful not only as photo-luminescent materials but also as photosensitizers for photochemical reactions. Therefore, in this study, we examined the effects of diphosphine ligands in detail on the photosensitizing abilities using a series of Cu(I) complexes bearing dmpp ligands shown in **Chart 1**, i.e., dmpp complexes with 3 types of diphosphine ligands with different length of carbon chains: dppe (1,2-bis(diphenylphosphino)ethane), dppp (1,3-bis(diphenylphosphino)propane), and dppb

(1,4-bis(diphenylphosphino)butane). The dmpp complex bearing Xantphos (4,5-Bis(diphenylphosphino)-9,9 dimethylxanthene), which is often used as a diphosphine ligand of Cu(I) photosensitizers, was also investigated for comparison. The systematic study herein reported suggests that the P-Cu-P angles should be one of the most important factors which determines the photosensitizing abilities.

# RESULTS AND DISCUSSION

## Photocatalytic CO<sup>2</sup> Reduction Using Cu(I) Complexes as Photosensitizers

In a typical run of photocatalytic reactions, a mixed solution of MeCN-TEOA (TEOA = triethanolamine, 4:1 v/v) containing **Cu(dppb)** (0.5 mM) as a photosensitizer, Re(2,2'-bipyridine)(CO)3Br (**Re(bpy)**, 0.05 mM) as a CO2-reduction catalyst and 1,3-dimethyl-2-phenyl-2,3 dihydro-1H-benzo[d]imidazole (BIH, 0.1 M) as an electron donor was irradiated under a CO<sup>2</sup> atmosphere using a high-pressure mercury-lamp equipped with a UV-cut filter (>370 nm, **Figure 1**). **Re(bpy)** is well known not only as a photocatalyst for CO<sup>2</sup> reduction (Hawecker et al., 1983, 1986; Kutal et al., 1985) but also as a CO2-reduction catalyst in photocatalytic systems, which can produce CO with high selectivity; (Gholamkhass et al., 2005; Sato et al., 2007; Kiyosawa et al., 2009; Koike et al., 2009; Morris et al., 2009; Tamaki et al., 2012, 2013a,b; Kou et al., 2014; Kato et al., 2015; Ohkubo et al., 2016). Therefore, we measured the gaseous products of the photocatalysis using gas-chromatography. CO was selectively produced and no hydrogen was detected over 30-min irradiation. TONCO after 30-min irradiation was 40. Though various other CO2-reduction catalysts, i.e., Re(4,4'-dimethyl-2,2'-bipyridine)(CO)3Br, Re(1,10-phenanthroline)(CO)3Br, and Ru(6,6'-dimethyl-2,2'-bipyridine)(CO)2Cl2, were also used

instead of **Re(bpy)**, the system using **Re(bpy)** showed the highest TONCO in this reaction condition (**Figure S1**).

In the case of using a mixed solvent of DMA-TEOA (DMA = N,N-dimethylacetamide, 4:1 v/v) instead of that of MeCN-TEOA (4:1 v/v), TONCO increased and reached 580 after 25-min irradiation (**Figure 1**). The turnover frequency (TOF), which was determined from the slopes of the fitting curves of the time course in the initial stage of the photocatalysis, was relatively high (65 min−<sup>1</sup> , **Figure S2**). This result suggests that DMA should be more preferable solvent for the efficient photocatalytic reaction using **Cu(dppb)** and **Re(bpy)**. In the absence of **Cu(dppb)**, TONCO after 1-h irradiation were 20; thus, the photocatalysis by only **Re(bpy)** should be not sufficient under the condition. In addition, in the absence of **Re(bpy)** or BIH, TONCO were 0 or 7, respectively. These results clearly indicate that **Cu(dppb)** has relatively high photosensitizing ability and promoted CO<sup>2</sup> reduction on **Re(bpy)** using BIH as an electron donor. From these results, we concluded the photocatalyses using **Re(bpy)** as a CO2-reduction catalyst and a mixed solvent of DMA-TEOA as a reaction solvent are suitable for the investigation to clarify the photosensitizing abilities of the series of the Cu(I) complexes shown in **Chart 1**.

Mixed solutions of DMA-TEOA (4:1 v/v) containing **Re(bpy)** (0.05 mM), BIH (0.1 M), and various Cu(I) complexes (0.5 mM) were shined under the same condition described above. **Figure 2** shows the time-courses of the CO production and both TONCO and TOF are summarized in **Table 1** and **Figure S2**. Interestingly, both TONCO and TOF strongly depended on the structure of the diphosphine ligands. When comparing **Cu(dppb)**, **Cu(dppp),** and **Cu(dppe)**, TOF drastically increased with increasing the length of the methylene chains between the two phosphorous atoms in the diphosphine ligands. In the case

using **Cu(Xantphos)**, TOF was quite high and almost equal to that of **Cu(dppb)**. In contrast, TONCO of **Cu(Xantphos)** was low, and the time course of CO formation reached plateau within 5-min of irradiation. The absorption spectral changes of the reaction solutions after irradiation are shown in **Figure S3**. In all cases, the metal-to-ligand charge transfer (MLCT) absorption bands of the Cu(I) complexes decreased gradually; therefore, one of the main reasons for the decreasing of the reaction rate with increasing the irradiation time should be the photodecomposition of the Cu(I) complexes. Curiously, in the case using **Cu(Xantphos),** the photocatalytic CO formation stopped within 5 min even though the MLCT absorption band was obviously observed in the absorption spectrum of the reaction solution after 5-min irradiation. Though the reason for the low TONCO when using **Cu(Xantphos)** is not clear at this stage, the degradation of **Re(bpy)** possibly proceeded more rapidly than the other cases. As described later, **Cu(Xantphos)**showed quite small absorption in the longer wavelength region (>430 nm) compared to the other Cu(I) complexes. It has been recently reported that the irradiation with short wavelength light (<450 nm) drastically lowered the photo-stability of Re(I) complexes; (Lang et al., 2019) thus, it is expected that **Re(bpy)** could absorb the irradiation light in the longer wavelength region, e.g., the emission line at 436 nm from the high-pressure mercury lamp, more frequently and photo-decomposition of **Re(bpy)** proceeded rapidly. Since the analysis of the Re(I) complexes, e.g., measurement of FT-IR spectra of the reaction solutions, was difficult because of the low concentration of the Re(I) complexes, further investigation is now undergoing.

In order to precisely evaluate the efficiency of CO<sup>2</sup> reduction, the quantum yield for CO formation (8CO) was determined TABLE 1 | Photocatalytic properties using the mixed system of the Cu(I) photosensitizers and Re(bpy)<sup>a</sup> .


*<sup>a</sup>A CO*2*-saturated DMA-TEOA (4:1 v/v) mixed solution containing the Cu(I) photosensitizer (0.5 mM),* Re(bpy) *(0.05 mM), and BIH (0.1 M) was irradiated.*

*<sup>b</sup>Maximum TONCO within 30-min irradiation calculated as [product (mol)]/[added* Re(bpy) *(mol)].*

*<sup>c</sup>TOF was determined by the slopes of the fitting curves over 5 min (*Figure S2*).*

*<sup>d</sup>A mixed solution of MeCN-TEOA (4:1 v/v) was used instead.*

*<sup>e</sup>Quenching fractions of emission from Cu(I) photosensitizers by BIH determined from Stern-Volmer analyses in a mixture of MeCN-TEOA (4:1 v/v).*

*<sup>f</sup>Quenching fractions of emission from Cu(I) photosensitizers by BIH determined from Stern-Volmer analyses in a mixture of DMA-TEOA (4:1 v/v).*

*<sup>g</sup>Quantum yield of CO production calculated as [CO (mol)]/[absorbed photon (einstein)] (*λ*ex* = *430 nm, light intensity: 1.0* × *10*−<sup>8</sup> *einstein/s).*

using the system of **Cu(dppb)** as a typical example. BIH is well known as a two-electron donor, which can induce the secondelectron injection even to ground-state of photosensitizers and/or CO2-reduction catalysts owing to the quite strong reducing power of the radical species produced after oxidation by the excited state of photosensitizers and subsequent deprotonation by TEOA (Equation 1); (Tamaki et al., 2013a). Thus, the maximum 8CO of this system should be 100%. A mixture of DMA and TEOA (4:1 v/v) containing 0.5 mM **Cu(dppb)**, 0.05 mM **Re(bpy)**, and 0.1 M BIH was irradiated with 430 nm monochromic light. CO was linearly produced over 1 h irradiation, and 8CO was determined to be 37% from the slopes of the fitting curves (**Figure S4**). This value is relatively high among the photocatalytic systems for CO<sup>2</sup> reduction using Cu(I)-complex photosensitizers.

spectra of **Cu(dppb)** in various solutions, i.e., dichloromethane, DMA, a mixture of DMA-TEOA and that of MeCN-TEOA. The shapes of the spectra at around 330–450 nm were similar. Except the spectrum in dichloromethane, small absorption bands were observed at around 450–530 nm. This is probably due to the formation of [Cu(dmpp)2] <sup>+</sup>-type complex produced by the disproportionation reaction after dissolving in solvents having coordination ability (Kaeser et al., 2013). Since the molar extinction coefficient of [Cu(dmpp)2] <sup>+</sup> was large (ε<sup>484</sup> = 10,300 in a DMA-TEOA mixed solution, **Figure S5**), the amount of the [Cu(dmpp)2] <sup>+</sup> should not be large. <sup>1</sup>H NMR analyses of **Cu(dppb)** using CD2Cl<sup>2</sup> or CD3CN also revealed that the structure in CD3CN is almost identical to that in CD2Cl2, though a few mol-percent of [Cu(dmpp)2] <sup>+</sup> were observed. The

### Properties of Cu(I) Complexes Structure of Cu(I) Complexes in Solutions

As described above, the Cu(I) complexes, in particular **Cu(dppb)**, showed relatively high photosensitizing ability. In order to clarify the reason for the difference in the photosensitizing abilities of the Cu(I) complexes, we firstly investigated the molecular structure of the Cu(I) complexes in the reaction solutions in detail because it is known that Cu(I) complexes often cause structural changes in solutions due to the ligandexchange reactions. **Figure 3** illustrates the UV-vis absorption other complexes also showed similar spectrum-pattern in both absorption spectra and <sup>1</sup>H NMR spectra regardless of solvents (**Figure S6**). Therefore, the main species should be the diiminediphosphine type complexes even in the reaction solutions for the photocatalytic reactions, and the solvent molecules do not strongly affect the structure and the electronic properties of the ground state of the Cu(I) complexes.

We previously clarified that **Cu(dppb)** is dimerized by bridging with two bidentate phosphine ligands in the solid state using single crystal X-ray structure analyses. The dimerization might be a reason for the higher photosensitizing ability of **Cu(dppb)**; Thus, we measured diffusion-ordered NMR spectroscopy (DOSY-NMR) in order to clarify whether **Cu(dppb)** is dimerized even in solutions. The obtained DOSY-NMR spectra of **Cu(dppb)**, **Cu(dppp)**, **Cu(dppe)**, **Cu(Xantphos),** and [Cu(dmpp)2](PF6) measured in CD2Cl<sup>2</sup> solutions are illustrated in **Figure S7**. The diffusion coefficients of **Cu(dppb)**, **Cu(dppp)**, **Cu(dppe)**, **Cu(Xantphos),** and [Cu(dmpp)2](PF6) in the solutions were determined to be 7.5 × 10−10, 7.9 × 10−10, 8.0 × 10−10, 7.0 × 10−<sup>10</sup> and 8.4 × 10−<sup>10</sup> [m<sup>2</sup> /s], respectively. These obtained values were quite similar though they gradually decreased with increasing the molecular weight calculated as mononuclear complexes. Hence, the molecular sizes of the Cu(I) complexes might be almost the same in solutions. Since **Cu(dppp)** and **Cu(dppe)** were reported as mononuclear complexes in the solid state, **Cu(dppb)** in the reaction solutions is regarded as a mononuclear complex in this paper and the dimerization should not be the main reason for the high TOF of the system using **Cu(dppb)**.

### Absorption Abilities of the Cu(I) Complexes

**Figure 4A** shows the UV-vis absorption spectra of **Cu(dppb)** and **Re(bpy)** in a DMA-TEOA mixed solution. Both complexes showed MLCT absorption bands at similar region with similar molar extinction coefficients. However, in the reaction condition, **Cu(dppb)** can absorb visible light more efficiently than **Re(bpy)** because the used concentration of **Cu(dppb)** was 10-times higher than that of **Re(bpy)** (**Figure 4B**). The light source was a high-pressure mercury lamp equipped with a UVcut filter (>370 nm); Therefore, the main wavelength of the excitation light should be 405 and 436 nm. At 405 and 436 nm, more than 96% of absorbed photons should excite **Cu(dppb)** [Abs405(**Cu(dppb)**): Abs405(**Re(bpy)**) = 2.3: 0.10, Abs436(**Cu(dppb)**): Abs436(**Re(bpy)**) = 1.1: 0.03]. This means **Cu(dppb)** was selectively excited by the excitation light under the reaction condition.

The UV-Vis absorption spectra of the series of the Cu(I) complexes in DMA-TEOA mixed solutions are illustrated in **Figure 5**. All of the Cu(I) complexes showed MLCT absorption bands at around 350–450 nm. When comparing **Cu(dppb)**, **Cu(dppp)**, and **Cu(dppe)**, the absorption maxima of the MLCT absorption bands were gradually blue-shifted with increasing the length of the carbon chains [λabs(**Cu(dppe)**) = 417 nm, λabs(**Cu(dppp)**) = 410 nm, λabs(**Cu(dppb)**) = 391 nm]. Similar blue-shift of the absorption maxima was also observed in the reported system using dichloromethane solutions (Tsubomura et al., 2015). It is reported that the shift was strongly related to the bite angles of the diphosphine ligands, which were determined using the single crystal X-ray diffraction, i.e., the absorption energy became larger as bite angle became larger. The wavelength of the absorption maximum of **Cu(Xantphos)** was the shortest (388 nm); thus, the bite angle of Xantphos in **Cu(Xantphos)** might be larger than those in the other complexes. It should be noted that the absorbance at the irradiation wavelength (405, 436 nm) are relatively high (Abs > 0.86). Therefore, the total of the absorbed photon number of each reaction solution for the photocatalyses should be almost identical regardless of the

structure of the diphosphine ligands, and TOFs should reflect the relative quantum yields for the photocatalytic CO<sup>2</sup> reduction.

### Photophysical and Photochemical Properties

reactions ([Cu(dppb)] = 0.5 mM, [Re(bpy)] = 0.05 mM).

In order to investigate the efficiencies of the initial steps of the photosensitizing reactions, we tried to measure the photophysical properties of Cu(I) complexes in the main solvent for the photocatalyses, i.e., DMA-TEOA mixed solutions. Interestingly, however, the Cu(I) complexes showed quite low luminescence properties in DMA-TEOA mixed solutions, even though the photocatalytic reactions proceeded efficiently in them. In particular, the emission quantum yield (8em) of **Cu(dppb)** in the DMA-TEOA solutions was <0.01% and emission lifetime (τem) was shorter than 30 ns in contrast to those in other solvents (8em in CH2Cl<sup>2</sup> = 16%, τem in CH2Cl<sup>2</sup> = 10 µs). The reason for the quite low luminescence properties of **Cu(dppb)** in DMA-TEOA solutions is now under investigation, but it

might be caused by an efficient exciplex-quenching by DMA molecules (McMillin and McNett, 1998; Kuang et al., 2002). The detection of the luminescence from the DMA-TEOA solutions containing **Cu(dppb)** was difficult; therefore, we decided to compare the photophysical and photochemical properties of the series of Cu(I) complexes in the other solvents for the catalyses, i.e., a mixture of MeCN-TEOA. The emission properties of the Cu(I) complexes in a mixture of MeCN-TEOA (4:1 v/v) are summarized in **Table 2** and the emission spectra are shown in **Figure S8**. All of the complexes showed weak emission intensity and short emission lifetimes in a mixture of MeCN-TEOA compared with those in dichloromethane likely due to the exciplex-quenching by the MeCN molecules. In fact, the emission intensity of **Cu(dppb)** in MeCN solutions was almost identical to that in MeCN-TEOA solutions. Hence, the low



*<sup>a</sup>Measured in a mixture of MeCN-TEOA.*

*<sup>b</sup>The molar extinction normalized by the number of the Cu centers. The UV-Vis absorption spectra are shown in* Figure S8*.*

*<sup>c</sup>Excitation wavelength: 390 nm.*

*<sup>d</sup>Excitation wavelength: 400 nm.*

*<sup>e</sup>Excitation wavelength: 410 nm.*

*<sup>f</sup> Excitation wavelength: 337 nm.*

*<sup>g</sup>The repoted values measured in dichloromethane (Tsubomura et al., 2015).*

emission intensity observed should not be derived from the quenching by TEOA (**Figure S9**). In accordance with the blueshift of the MLCT absorption bands in the absorption spectra, the wavelength of the emission maxima of the three complexes, i.e., **Cu(dppb)**, **Cu(dppp)**, and **Cu(dppe)**, gradually decreased as the number of the methylene chains in the diphosphine ligands increased. The emission lifetimes became longer with increasing the length of the carbon chains. This is likely because the steric hindrance of the phenyl groups on the phosphorous atoms became larger with increasing the bite-angle of the diphosphine ligands, which resulted in inhibiting the attack of the solvent molecules in the excited state and decreasing non-radiative decay constants. **Cu(Xantphos)** showed the shortest wavelength of the emission maximum and the longest emission lifetime.

Since the reactions between the excited state of the Cu(I) complexes and BIH compete with the radiative and non-radiative deactivation processes, the shorter lifetime might lower the efficiencies of the quenching by BIH. Thus, we conducted quenching experiments using BIH to check whether the excited state of the Cu(I) complexes can be quenched by BIH. We measured emission intensity in the presence of various amounts of BIH. As shown in **Figure 6A**, the emission intensities became smaller with increasing the concentrations of BIH, indicating the excited state of **Cu(dppb)** was efficiently quenched by BIH. The Stern–Volmer analysis showed linear plots with an intercept at 1 (**Figure 6B**). From the slope of the fitting curve (KSV = kqτ em = 0.71 mM−<sup>1</sup> ) and the emission lifetime τ em = 240 ns, the quenching rate constant (kq) was determined to be 3.0 × 10<sup>9</sup> M−<sup>1</sup> s −1 . From these values, the quenching fraction (ηq) is estimated to be larger than 98% under the reaction condition (Equation 2) and the excited state of **Cu(dppb)** is expected to be quenched by BIH almost quantitatively, even though the emission lifetime is not so long. Moreover, in the presence of **Re(bpy)**, the emission intensity was almost identical to that in the absence of **Re(bpy)** (**Figure S10**); therefore, oxidative quenching, i.e., the electron transfer from the excited state of **Cu(dppb)** to **Re(bpy)** should be negligible. These results suggest that the photosensitizing reaction by **Cu(dppb)** mainly

proceeds via the reductive quenching by BIH.

$$
\eta\_{\rm q} = [\text{BIH}] \times \text{K}\_{\rm SV} / (1 + [\text{BIH}] \times \text{K}\_{\rm SV}) \tag{2}
$$

In the case using the other Cu(I) complexes, the emission intensity also decreased drastically by adding BIH, and emission from the Cu(I) complexes could not be detected in the presence of 0.1 M BIH. The Stern–Volmer analysis using the other Cu(I) complexes not only in a mixture of MeCN-TEOA but also in a mixture of DMA-TEOA could also be conducted owing to the larger emission intensities and the longer emission lifetimes in DMA-TEOA mixed solutions than those of **Cu(dppb)**. The Stern–Volmer analysis showed linear plots with an intercept at 1 in all cases (**Figure S11**). From the slope of the fitting curve, the quenching fractions of **Cu(dppp)**, **Cu(dppe)**, and **Cu(Xantphos)** were estimated to be larger than 99, 98, and 99%, respectively. These results clearly indicate that all of the Cu(I) complexes used in this study have relatively strong oxidizing power in the excited state and the efficiencies of the initial stage of the photosensitizing reaction, i.e., reductive quenching, should be almost identical regardless of the structure of the diphosphine ligands.

### Spectral Changes During Irradiation

As mentioned above, the excited state of the Cu(I) complexes was efficiently quenched by BIH. To clarify the produced species of the quenching, UV-vis absorption spectral changes during photo-irradiation of a DMA-TEOA mixed solution (4:1 v/v) containing **Cu(dppb)** (0.1 mM) and BIH (0.1 M) under Ar were measured (**Figure 7A**). After an induction period over a few minutes, a new absorption bands at around 350–450 nm was observed. The new band increased continuously and linearly over 30-min irradiation. After irradiation, the solution was exposed to an ambient air. The solution color was immediately changed from deep-yellow to light-yellow. In the absorption spectrum measured after exposure to air, the new absorption band was completely disappeared and the original spectrumshape was recovered (**Figure 7C**). These results clearly indicate that the quenching of the excited state of **Cu(dppb)** is a reductive quenching process, which gives one-electron reduced species (OERS) of **Cu(dppb)**. In contrast, when using **Cu(dppe)** the observed spectral changes were much smaller than **Cu(dppb)** though the shape of the new absorption band was similar to those of **Cu(dppb)** (**Figure 7B**). This difference should be derived from the difference of efficiencies for production of OERS between **Cu(dppb)** and **Cu(dppe)**. **Cu(Xantphos)** also showed rapid accumulation of OERS under the same condition, and **Cu(dppp)**, in contrast, showed slow accumulation as well as **Cu(dppe)** (**Figure S12**). Though the determination of the exact rate of production of OERS requires a spectrum obtained by electrochemical spectroscopy technics, such big difference in spectral changes and similarity of the shapes of the new absorption bands should suggest that the efficiency for production of OERS of **Cu(dppb)** and **Cu(Xantphos)** should be much larger than those of **Cu(dppp)** and **Cu(dppe)**. The tendency of the degree of the spectral changes corresponded reasonably well with that of TOF described above.

### Electrochemical Properties

**Figure 8** shows the cyclic voltammograms of **Cu(dppb)** and **Re(bpy)** measured in DMA-TEOA solutions and their redox potentials are summarized in **Table 3**. Both **Cu(dppb)** and **Re(bpy)** exhibited a quasi-reversible reduction wave as a first reduction wave at −1.96 and −1.70 V (vs. Ag/AgNO3), respectively. These waves are attributable to the reduction of the diimine ligands. The reduction potential of **Cu(dppb)** is more negative than that of **Re(bpy)**; thus, OERS of **Cu(dppb)** has relatively strong reducing power and the electron transfer process from OERS of **Cu(dppb)** to **Re(bpy)** occurs exothermically when using **Cu(dppb)** as a photosensitizer and **Re(bpy)** as a catalyst. Interestingly, when using an MeCN-TEOA solution, the first reduction wave of **Cu(dppb)** became irreversible, though

FIGURE 7 | Absorption spectral changes of DMA–TEOA (4:1 v/v) solutions containing BIH (0.1 M) and 0.1 mM of (A) Cu(dppb) or (B) Cu(dppe) under an argon atmosphere. The solution was irradiated with 430-nm monochromic light. (C) Absorption spectral changes of the reaction solution of Cu(dppb) measured before irradiation (blue solid line), just after 30-min irradiation (red solid line), and after irradiation and subsequent exposure to an ambient air (red dotted line).

the peak potential was almost identical to that in the DMA-TEOA solution. This means OERS of **Cu(dppb)** is not stable in an MeCN-TEOA solution and might be one of the reasons

FIGURE 8 | Cyclic voltammograms of Cu(dppb) (blue solid line) and Re(bpy) (black solid line) a mixture of DMA and TEOA (4:1 v/v) containing Et4NPF6 (0.1 M) as a supporting electrolyte with a Ag/AgNO3 (0.01M) reference electrode. The voltammogram of Cu(dppb) measured in a mixture of MeCN and TEOA (4:1 v/v) is also illustrated as a blue dotted line.

TABLE 3 | Electrochemical properties of the metal complexes in DMA-TEOA<sup>a</sup> .


*<sup>a</sup>Measured in a DMA-TEOA mixed solution containing the complex (0.5 mM) and Et*4*NPF*<sup>6</sup> *(0.1 M) with a scan rate of 100 mV*·*s* <sup>−</sup><sup>1</sup> *under an Ar atmosphere.*

*<sup>b</sup>A DMA solution was used instead.*

*<sup>c</sup>Peak potential.*

for the low TONCO in the case using MeCN-TEOA solutions. Even though the oxidation potentials of the complexes could not be observed due to the potential window narrowed by oxidation of TEOA, the irreversible oxidation wave was observed at +0.80 V when using a DMA solution instead of the DMA-TEOA solutions. The first reduction potential observed in the DMA solution was almost identical to that in the presence of TEOA (−1.91 V). **Cu(dppb)** showed more negative reduction potential than Ru(II) tris-diimine complexes, which are typical redox-photosensitizers in the photocatalytic systems for CO<sup>2</sup> reduction. On the other hand, the oxidation potential was similar to those of the Ru(II) complexes; (Yamazaki et al., 2015) thus, **Cu(dppb)** should have both high reducing power in the reduced state and relatively high oxidizing power in the excited state.

The other Cu(I) complexes also showed reversible or quasireversible waves in the cyclic voltammograms measured in DMA-TEOA solutions (**Figure S13**). The reduction potentials were

almost identical (−1.93 to −1.96 V); the reduction potentials were not strongly affected by the structure of the diphosphine ligands. These results are reasonable because it is well known that the lowest-unoccupied molecular orbitals (LUMO) of the Cu(I) complexes are mainly derived from the π ∗ -orbitals of diimine ligands and the series of Cu(I) complexes have the same diimine ligands. Therefore, all of the Cu(I) complexes should have high reducing power in the reduced state enough to smoothly trigger the electron transfer from OERS of the Cu(I) complexes to **Re(bpy)**.

### Expected Reaction Mechanism and Relationship Between the Molecular Structure and the Photosensitizing Abilities

From the results in the previous sections, the expected reaction mechanism of the photocatalytic systems in this study can be summarized as follows (**Figure 9**): (1) almost all of the irradiated photons are absorbed by the Cu(I) complexes; (2) the excited states of the Cu(I) complexes are reductively quenched by BIH to give OERS with high efficiencies; (3) the electron transfer from the OERS of the Cu(I) complexes to **Re(bpy)** proceeds exothermically; (4) CO<sup>2</sup> reduction takes place on **Re(bpy)** using the obtained electrons.

As described in section Photocatalytic CO<sup>2</sup> reduction using Cu(I) complexes as photosensitizers, TOFs were drastically changed by changing the structure of the diphosphine ligands. In contrast, all of the Cu(I) complexes showed quite similar absorbance, quenching fractions and reduction potentials; therefore, we could not observe a big difference in the efficiency of each step illustrated in **Figure 9**. However, only the efficiencies of the production of OERS strongly depended on the molecular structure even though the preceding process, i.e., the reductive quenching, proceeded almost quantitatively in all cases. It is reported that the efficiencies for the production of OERS strongly affected not only by the quenching fractions but also by the rate of the back electron transfer processes, i.e., the electron transfer from the OERS of the Cu(I) complexes to the one-electron oxidized species of BIH (BIH·+) produced by the reductive quenching processes (Tamaki et al., 2013a). This charge-recombination process forms the ground states of the Cu(I) complexes and BIH, resulting in wasting absorbed photons. The low efficiencies for the production of OERS in the case using **Cu(dppe)** should indicate that the back electron transfer proceeded more rapidly than the case using **Cu(dppb)** (**Figure 10**). The reason for the great difference in the rate of the back electron transfer is unclear at the moment, but the difference in the bite angles of the diphosphine ligands might be an important factor. The orientation of the four phenyl groups on the two phosphorous atoms is affected by changing the bite angles of the diphosphine ligands. Along with this, the distance between the phenyl groups on the phosphine ligands and the dmpp ligands will decrease by increasing the bite angles. The electrons obtained photochemically should mainly localize on the π <sup>∗</sup> orbital in the dmpp ligands; thus, the back electron transfer process should proceed via the attack of BIH·<sup>+</sup> to the π <sup>∗</sup> orbital. If the phenyl groups exist nearby the dmpp ligands and cause the steric hindrance, the access of BIH·<sup>+</sup> will be suppressed. As a result, the large P-Cu-P angles might suppress the charge-recombination and enhanced the efficiencies for OERS production. In fact, not only the photophysical properties

TABLE 4 | Relationship between P-Cu-P angles in the solid state of the Cu(I) complexes and photosensitizing abilities<sup>a</sup> .


*<sup>a</sup>The angles are the reported values determined by single crystal X-ray diffraction studies. TOF was determined by the slope of the fitting curves over 5 min (*Figure S2*).*

*<sup>b</sup>The values from the reported literature (Tsubomura et al., 2015).*

*<sup>c</sup>The values from the reported literature (Heberle et al., 2017). The value in [Cu(2,9 dimethyl-1,10-phenanthroline)(Xantphos)]*<sup>+</sup> *instead of* Cu(Xantphos)*.*

*<sup>d</sup> Measured in a mixture of DMA-TEOA (4:1 v/v).*

*<sup>e</sup> Measured in a mixture of MeCN-TEOA (4:1 v/v).*

but also TOFs have good correlation with the reported bite angles determined using the single crystal X-ray diffraction (**Table 4**); (Tsubomura et al., 2015; Heberle et al., 2017) in other words, not only long emission lifetimes but also high photosensitizing abilities will be obtained by increasing the bite angles of the diphosphine ligands likely due to the steric hindrance of the phenyl groups to suppress the approach of the solvent molecules to the Cu(I) center and BIH·<sup>+</sup> to the dmpp ligands.

### EXPERIMENTS

### General Procedure

NMR spectra were recorded on a JEOL ECA500 (500 MHz) spectrometer. The spectra were analyzed by Delta version 5. Transmission UV-vis (ultraviolet-visible light) absorption spectra were recorded on an Agilent 8453 spectrometer. Emission spectra were obtained in a solvent degassed by bubbling with argon. Emission spectra and emission lifetimes were recorded on a laboratory-made apparatus. The samples were excited by a monochromated xenon light source and the emission was analyzed by using a cooled CCD spectrometer. For the lifetime measurement, a nitrogen laser (USHO AN-200) was used to excite the samples, and the emission was detected by a monochromator equipped with a photomultiplier tube, and the signal was analyzed by using a digital oscilloscope. Emission quantum yields in solution were determined a combination of an integrating sphere, a monochromated xenon light source and a cooled CCD spectrometer.

### Electrochemical Measurement

Cyclic voltammograms were conducted using an electrochemical analyzer, EC-stat 100 (EC Frontier CO., Ltd), using an electrochemical cell equipped with working (glassy carbon, ϕ = 3 mm), auxiliary (platinum wire), and reference (Ag/AgNO3, 10 mM) electrodes. The solution containing a Cu(I) complex (0.5 mM) and Et4NPF<sup>6</sup> (0.1 M) was bubbled with argon before the measurements.

### Materials

**Cu(dppb)**, **Cu(dppp)**, **Cu(dppe)**, **Re(bpy)**, and BIH were prepared according to reported methods with some modifications (Hawecker et al., 1986; Tamaki et al., 2013a; Tsubomura et al., 2015). **Cu(Xantphos)** were synthesized using dmpp instead of the corresponding diimine ligand according to the method for [Cu(2,2′ -biquinoline)(Xantphos)](PF6) (McCullough et al., 2018). Other chemicals were used as purchased without further purification.

### Yamazaki et al. CO2 Reduction Using Cu(I) Complexes

### Photocatalytic Reactions

The photocatalytic reactions were performed in a 15 mL test tube (i.d. = 11 mm) containing a 2 mL DMA–TEOA (4:1 v/v) solution of **Re(bpy)** (0.05 mM), BIH (0.1 M), and a Cu(I) complex (0.5 mM) after purging with CO<sup>2</sup> for more than 20 min. The solution was irradiated using a laboratory-made merrygo-round irradiation apparatus at λ > 370 nm with a highpressure mercury lamp (SEN LIGHT Co.) combined with a UV-cut filter (Lintec Commerce. Inc.). During irradiation, the tube was cooled with a thermostatic bath (20◦C). The gaseous reaction products (CO and H2) were analyzed using a GC-TCD instrument (Shimadzu GC-8A).

### Measurement of the Absorption Spectrum of the OERS

A DMA–TEOA (4:1 v/v) solution (2 mL) containing the Cu(I) complex (0.1 mM) and BIH (0.1 M) in a 7 mL quartz cell (lightpass length: 1 cm) was purged with Ar for over 20 min. The solution was irradiated with 430-nm monochromic light derived from an LED lamp purchased from CELL System Co. UV-Vis absorption spectral changes during irradiation were recorded using a Shimadzu QYM-01 system.

### Measurement of Quantum Yields of Photocatalytic CO Formation

A DMA–TEOA (4:1 v/v) solution (2 mL) containing **Cu(dppb)** (0.5 mM), **Re(bpy)** (0.05 mM), and BIH (0.1 M) in a quartz cell (light-pass length: 1 cm, volume: 7 mL, light intensity: 1 × 10−<sup>8</sup> einstein per s) was purged with CO<sup>2</sup> for 30 min. The solution was irradiated at λex = 430 nm derived from an LED lamp purchased from CELL System Co. UV-Vis absorption spectral changes during irradiation were recorded using a Shimadzu QYM-01 system. During irradiation, the cell was cooled at 25◦C with a thermostatic bath. The gaseous reaction products (CO and H2) were analyzed using a GC-TCD instrument (Shimadzu GC-8A).

### CONCLUSION

We investigated in detail the effects of the structure of the diphosphine ligands on the photosensitizing abilities of heteroleptic diimine-diphosphine Cu(I) complexes using a series

### REFERENCES


of Cu(I) complexes bearing dmpp ligands and 4 types of diphosphine ligands by conducting photocatalytic CO<sup>2</sup> reduction using the Cu(I) complexes as photosensitizers. Turnover frequencies of the CO<sup>2</sup> reduction drastically increased up to ∼70 min−<sup>1</sup> with increase of the P-Cu-P angles, even though all of the Cu(I) complexes showed similar photophysical, photochemical and electrochemical properties. In the case using **Cu(dppb)**, rapid accumulation of OERS was observed. On the contrary, when using **Cu(dppe)**, the apparent rate of the production of the OERS was quite slow. The low efficiency for the production of the OERS should be derived from the rapid charge-recombination with BIH·+. The large bite angles make the orientation of the phenyl groups on the phosphorous atoms toward the dmpp ligands, resulting in suppression of the attack by BIH·<sup>+</sup> to the dmpp ligands. This might be one of the reasons for the high efficiencies of both the production of OERS and photocatalytic reactions when using the Cu(I) complexes connecting with the diphosphine ligands with the large bite angles.

# AUTHOR CONTRIBUTIONS

YY and TT designed the study and YY wrote the initial draft of the manuscript. TO and TU investigated the photosensitizing abilities of the Cu(I) complexes. JI contributed to analyses of the structure of the Cu(I) complexes in solutions using NMR spectroscopy. SF measured the photophysical properties of the Cu(I) complexes in detail. The electrochemical properties of the Cu(I) complexes were checked by CT. All authors approved the final version of the manuscript, and agree to be accountable for all aspects of the work.

### FUNDING

This work was supported by JSPS KAKENHI Grant Number 17K05816 and Grant from Seikei University.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00288/full#supplementary-material


10-phenanthroline and simple diphosphine ligands. Dalton Trans. 4444–4448. doi: 10.1039/b608641a


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yamazaki, Onoda, Ishikawa, Furukawa, Tanaka, Utsugi and Tsubomura. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# An Ir(III) Complex Photosensitizer With Strong Visible Light Absorption for Photocatalytic CO<sup>2</sup> Reduction

Yusuke Kuramochi\* † and Osamu Ishitani\*

*Department of Chemistry, Graduate School of Science and Engineering, Tokyo Institute of Technology, Tokyo, Japan*

A cyclometalated iridium(III) complex having 2-(pyren-1-yl)-4-methylquinoline ligands [Ir(pyr)] has a strong absorption band in the visible region (ε444nm = 67,000 M−<sup>1</sup> cm−<sup>1</sup> ) but does not act as a photosensitizer for photochemical reduction reactions in the presence of triethylamine as an electron donor. Here, 1,3-dimethyl-2-(o-hydroxyphenyl)-2,3-dihydro-1H-benzo[d]imidazole (BI(OH)H) was used instead of the amine, demonstrating that BI(OH)H efficiently quenched the excited state of Ir(pyr) and can undergo the photochemical carbon dioxide (CO2) reduction catalyzed by *trans*(Cl)-Ru(dmb)(CO)2Cl<sup>2</sup> (dmb = 4,4′ -dimethyl-2,2′ -bipyridine, Ru) to produce formate as the main product. We also synthesized a binuclear complex combining Ir(pyr) and Ru *via* an ethylene bridge and investigated its photochemical CO<sup>2</sup> reduction activity in the presence of BI(OH)H.

Keywords: strong visible-light absorption, metal complex, photocatalyst, electron donor, rhenium, CO2 reduction, photosensitizer, iridium

# INTRODUCTION

Today, the consumption of fossil resources releases a tremendous amount of carbon dioxide (CO2), which has had a serious impact on global climate change. The reduction in fossil resources in the future will induce shortages in both energy and carbon sources. To resolve these serious problems, the development of alternative energy systems that produce reduced volumes of CO<sup>2</sup> by using solar light as an energy source is desirable. To utilize a wider range of visible light from the sun, nature-inspired artificial Z-scheme systems have been developed by using semiconductors modified with metal complexes (Sato et al., 2011; Sekizawa et al., 2013; Kuriki et al., 2016, 2017; Sahara et al., 2016; Kumagai et al., 2017). Some metal complex photocatalytic systems that consist of a photosensitizer (PS) and a catalyst (CAT) can selectively induce CO<sup>2</sup> reduction and suppress hydrogen (H2) evolution. These systems require a sacrificial electron donor due to the relatively low oxidation power of the PS unit in the excited state (Yamazaki et al., 2015; Tamaki and Ishitani, 2017; Kuramochi et al., 2018a) Step-by-step excitation of both the semiconductor and the metal complex produces an electron with high reducing power and a hole with high oxidizing power, allowing for CO<sup>2</sup> reduction by weaker electron donors such as methanol (**Figure 1**; Sekizawa et al., 2013).

Ru(II) tris-diimine complexes [Ru(N∧N)3] <sup>2</sup><sup>+</sup> have been frequently used as the PS unit of supramolecular photocatalysts, which have strong absorption in the visible region, a long lifetime of the <sup>3</sup>MLCT excited state, and a stable one-electron reduced state. However, [Ru(N∧N)3] <sup>2</sup><sup>+</sup> has a problem in that one of the N∧N ligands is relatively easily released during the photocatalytic reaction to give [Ru(N∧N)2(Solvent)2] <sup>2</sup>+-type complexes that work as catalysts for CO<sup>2</sup> reduction

### Edited by:

*Hitoshi Ishida, Kitasato University, Japan*

Reviewed by: *Zhong Jin, Nanjing University, China Luca Conti, University of Florence, Italy*

### \*Correspondence:

*Yusuke Kuramochi kuramochiy@rs.tus.ac.jp Osamu Ishitani ishitani@chem.titech.ac.jp*

### †Present Address:

*Yusuke Kuramochi, Department of Chemistry, Faculty of Science Division II, Tokyo University of Science, Tokyo, Japan*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *31 January 2019* Accepted: *01 April 2019* Published: *01 May 2019*

### Citation:

*Kuramochi Y and Ishitani O (2019) An Ir(III) Complex Photosensitizer With Strong Visible Light Absorption for Photocatalytic CO*2 *Reduction. Front. Chem. 7:259. doi: 10.3389/fchem.2019.00259*

**20**

reduction, consisting of a semiconductor (TaON) and a Ru(II) binuclear complex (RuRu') (Sekizawa et al., 2013).

(Lehn and Ziessel, 1990; Yamazaki et al., 2015; Kuramochi et al., 2018a). In addition, a PS unit with stronger absorption in the visible region compared to [Ru(N∧N)3] <sup>2</sup><sup>+</sup> should be more favorable for constructing new photocatalytic systems.

In recent years, cyclometalated iridium(III) complexes, such as Ir(ppy)<sup>3</sup> (ppy = 2-phenylpyridine) and [Ir(ppy)2(N∧N)]+, have been used as PSs in various photocatalytic reactions, such as H<sup>2</sup> evolution (Goldsmith et al., 2005; Lowry and Bernhard, 2006), CO<sup>2</sup> reduction (Thoi et al., 2013; Bonin et al., 2014; Chen et al., 2015; Rao et al., 2017, 2018), and organic synthesis (**Figure 2**; Prier et al., 2013; Schultz and Yoon, 2014; Shaw et al., 2016), even though the absorption by Ir(ppy)<sup>3</sup> and [Ir(ppy)2(N∧N)]<sup>+</sup> is relatively weak in the visible region. We already reported that [Ir(piq)2(dmb)]<sup>+</sup> (piq = 1-phenylisoquinoline, dmb = 4,4′ -dimethyl-2,2′ -bipyridine), which has a stronger absorption in the visible region (ε444nm = 7,800 M−<sup>1</sup> cm−<sup>1</sup> ) than Ir(ppy)3, acts as a PS for CO<sup>2</sup> reduction without forming decomposed species that catalyze CO<sup>2</sup> reduction (**Figure 2**; Kuramochi and Ishitani, 2016). The intense absorption at the longer wavelength allowed for the selective excitation of the PS without exciting the CAT, such as fac-Re(dmb)(CO)3Br (dmb = 4,4′ -dimethyl-2,2′ -bipyridine). In addition, the supramolecular photocatalyst, where [Ir(piq)2(BL)]<sup>+</sup> (BL = bridging ligand) is connected with fac-Re(BL)(CO)3Br, works as a better photocatalyst for CO<sup>2</sup> reduction compared to the mixed system of the corresponding mononuclear complexes, i.e., [Ir(piq)2(dmb)]<sup>+</sup> and fac-Re(dmb)(CO)3Br. Although [Ir(piq)2(dmb)]<sup>+</sup> has a stronger absorption in the visible region compared to Ir(ppy)<sup>3</sup> and the advantages over [Ru(N∧N)3] 2+ as mentioned above, its absorption in the visible region is weaker than that of [Ru(N∧N)3] <sup>2</sup>+. It has been reported that several Ir(III) complexes that have stronger absorption bands in the visible region than [Ru(N∧N)3] <sup>2</sup><sup>+</sup> can act as PSs for H<sup>2</sup> evolution (Takizawa et al., 2012, 2014, 2018; Fan et al., 2014). Fan et al. reported that an Ir(III) complex with 2-(pyren-1-yl)-4-methylquinoline ligands [**Ir(pyr)**, **Figure 2**] showed a very strong absorption band in the visible region, ε(450 nm) > 60,000 M−<sup>1</sup> cm−<sup>1</sup> . Unfortunately, it could not photocatalyze H<sup>2</sup> evolution using K2PtCl<sup>4</sup> as CAT in the presence of triethylamine (TEA). This inactivity was explained by the lack of photoinduced electron transfer from TEA to the excited state of **Ir(pyr)**, as the reduction potential of the excited state of **Ir(pyr)** is less than the oxidation potential of TEA (Fan et al., 2014). We reported that the Ru(II)–Ru(II) supramolecular photocatalyst can selectively reduce CO<sup>2</sup> to formic acid (HCOOH) by using 1,3-dimethyl-2-(o-hydroxyphenyl)-2,3 dihydro-1H-benzo[d]imidazole (BI(OH)H) as an electron donor (ED) with a high turnover number (TONHCOOH) and a high quantum yield (8HCOOH; Tamaki et al., 2015). This photocatalytic reaction does not proceed in the absence of BI(OH)H even if triethanolamine (TEOA), which has a similar oxidation potential to TEA, is used. This is because BI(OH)H has a much stronger reducing power (E ◦x <sup>1</sup>/<sup>2</sup> = +0.02 V vs. Ag/AgNO3) (Hasegawa et al., 2006; Elgrishi et al., 2017; Kuramochi et al., 2018a) than TEA (E ◦x <sup>p</sup> = +0.67 V vs. Ag/AgNO3) (Yamazaki et al., 2015; Elgrishi et al., 2017).

Herein, we report the successful use of **Ir(pyr)** as a PS for CO<sup>2</sup> reduction by using BI(OH)H as ED and trans(Cl)-Ru(dmb)(CO)2Cl<sup>2</sup> (**Ru**) as CAT. We also synthesized a supramolecular photocatalyst from **Ir(pyr)** (**Ir(pyr)–Ru**; **Scheme 1**) and investigated its photocatalytic activity for CO<sup>2</sup> reduction.

### RESULTS AND DISCUSSION

### Synthesis of the Binuclear Complex, Ir(pyr)–Ru

**Ir(pyr)–Ru** was synthesized according to **Scheme 1**. The di-µchloride-iridium dimer (Fan et al., 2014) was reacted with AgPF<sup>6</sup> in acetonitrile to give the mononuclear acetonitrile–iridium complex. This complex was reacted with 1,2-bis(4′ -methyl- [2,2′ -bipyridin]-4-yl)ethane (BL), and the crude product was isolated using a silica gel column giving [Ir(pyr-mq)2(BL)](PF6) in 43% yield, based on the acetonitrile–iridium complex. This was reacted with [Ru(CO)2Cl2]n, which was pretreated by refluxing in the solvent (Kuramochi et al., 2015), giving the desired binuclear complex **Ir(pyr)–Ru** as a PF<sup>6</sup> salt in 87% yield.

# Photophysical and Electrochemical Properties

Fan et al. reported the photophysical and electrochemical properties of**Ir(pyr)** in dichloromethane or tetrahydrofuran (Fan et al., 2014). Since these solvents are less polar than the solvents suitable for CO<sup>2</sup> reduction, such as N,N-dimethylacetamide (DMA), we measured the photophysical and electrochemical properties of **Ir(pyr)** in DMA (Kuramochi et al., 2014). **Figure 3** shows the ultraviolet–visible (UV–vis) absorption spectrum of **Ir(pyr)** in DMA, which shows a much stronger absorption in the visible region (ε444nm = 67,000 M−<sup>1</sup> cm−<sup>1</sup> ) compared to [Ir(piq)2(dmb)]<sup>+</sup> and [Ru(dmb)3] <sup>2</sup>+. According to the timedependent density functional theory (TD-DFT) calculation of the UV–vis spectrum of **Ir(pyr)** in DMA (**Figure 4**, red bars), the strong absorption band at 444 nm is due to the transitions from the highest occupied molecular orbital (HOMO)-1 to the lowest unoccupied molecular orbital (LUMO)+2 and from HOMO to LUMO+1, which correspond to the π − π ∗ transitions of the pyrene moieties. The absorption at a wavelength >500 nm is assigned to the transition from HOMO-1 to LUMO and corresponds to the transition from the interligand transition from the dmb to the pyrene moieties and might include some

*N,N*-dimethylacetamide (DMA) at 298 K.

of Ir(pyr) in DMA at 298 K. (B) Frontier orbitals of Ir(pyr). Calculation method: B3LYP/LANL2DZ(Ir)/6-31G(d,p) (H, C, N) levels by using polarizable continuum model (PCM) with the default for DMA, isovalue = 0.02.

TABLE 1 | Electrochemical data in DMA (dichloromethane or THF) at 298 K<sup>a</sup> .


*<sup>a</sup>E vs. Ag/AgNO*<sup>3</sup> *(10 mM). <sup>b</sup>Estimated by DPV in acetonitrile. <sup>c</sup>Excited-state oxidation and reduction potentials of the photosensitizer were calculated from E*◦*<sup>x</sup>* 1/2 *– E*<sup>00</sup> *and Ered* 1/2 + *E*00*, respectively. <sup>d</sup>Kuramochi and Ishitani (2016). <sup>e</sup>The values were correlated by using conversion factor (*−*0.631 V) from NHE to Ag/AgNO*3*, see Fan et al. (2014) and Elgrishi et al. (2017). <sup>f</sup>Kuramochi et al. (2018b).*

contribution from the singlet–triplet transitions, as described in the literature (Fan et al., 2014). The emission spectrum of **Ir(pyr)** in DMA is shown in **Figure 4**, and the Franck–Condon line-shape analysis (**Figure S1**) gave a 0–0 band energy gap of 14,500 cm−<sup>1</sup> for **Ir(pyr)**, which is lower than that for [Ir(piq)2(dmb)]<sup>+</sup> (16,950 cm−<sup>1</sup> ) (Kuramochi and Ishitani, 2016).

The redox potentials of **Ir(pyr)** in DMA were obtained by cyclic voltammetry (CV; **Figure S2**) and differential pulse voltammetry (DPV) and are summarized in **Table 1** together with those of [Ir(piq)2(dmb)]<sup>+</sup> and **Ru** (Kuramochi and Ishitani, 2016; Kuramochi et al., 2018b). The oxidation waves of the Ir complexes were measured in acetonitrile due to its widepotential window. The CV showed three each of reversible cathodic and irreversible anodic waves, indicating that **Ir(pyr)** is stable against reduction but relatively unstable to oxidation on the CV timescale. According to the DFT calculation, the LUMO mainly distributes across the dmb ligand of **Ir(pyr)** (**Figure 4**). Thus, it is expected that the first electron is injected into the dmb ligand, which benefits the electron transfer from the one-electron reduced species of **Ir(pyr)** to the **Ru** moiety. This property has been previously observed in [Ir(piq)2(dmb)]<sup>+</sup> (Kuramochi and Ishitani, 2016).

# Emission Quenching by Electron Donors

The emission intensity of **Ir(pyr)–Ru** was similar to that of **Ir(pyr)**, suggesting that oxidative quenching of the excited state of the **Ir** unit by the **Ru** unit does not proceed in **Ir(pyr)– Ru**. This is reasonable because the oxidative quenching process is endothermic; the oxidation potential of the excited state of **Ir(pyr)** is much more positive (−0.96 V; **Table 1**) than the reduction potential of **Ru** (−1.66 V).

Emission quenching of **Ir(pyr)** by 1-benzyl-1,4 dihydronicotinamide (BNAH) was inefficient: Stern–Volmer constant (KSV) = 13 M−<sup>1</sup> in DMA. Assuming that the emission lifetime is 3.1 µs (Fan et al., 2014), the quenching rate constant (kq) was estimated to be 4.2 × 10<sup>6</sup> M−<sup>1</sup> s −1 . When a stronger electron donor, BI(OH)H (Hasegawa et al., 2006; Tamaki et al., 2015), was used, the emission of **Ir(pyr)** was more efficiently quenched (**Figure 5**); KSV reached 3,000 M−<sup>1</sup> in DMA/TEOA (5:1 v/v), and k<sup>q</sup> was 9.7 × 10<sup>8</sup> M−<sup>1</sup> s −1 , which is close to the diffusion-controlled rate constant (Tamaki et al., 2013). The KSV of **Ir(pyr)–Ru** by BI(OH)H was 2,800 M−<sup>1</sup> , which is similar to that of **Ir(pyr)**. In previous work by Fan et al. (2014), **Ir(pyr)** did not work as a PS for H<sup>2</sup> evolution because the emission from **Ir(pyr)** was not quenched by TEA. This emission is not quenched by TEOA as well because TEOA has a similar oxidation potential to TEA. Conversely,

BI(OH)H significantly quenches the emission from **Ir(pyr)**, indicating efficient electron transfer from BI(OH)H to the excited **Ir(pyr)**.

# Photocatalytic CO<sup>2</sup> Reduction

DMA-TEOA (5:1 v/v) mixed solutions containing both **Ir(pyr)** and **Ru** or **Ir(pyr)–Ru** (0.05 mM) as the photocatalysts and BI(OH)H as the ED were irradiated at λex > 480 nm under a CO<sup>2</sup> atmosphere. In both cases, HCOOH was mainly detected with small amounts of CO and H2. **Figure 6** shows the time profiles of product formation during the photocatalytic reaction. Blank experiments in the absence of the **Ru** catalyst produced trace amounts of CO (3.9 µmol) and formate (7.6 µmol) after 24 h. From the Stern–Volmer constants, the quenching efficiencies of the excited states of **Ir(pyr)** and **Ir(pyr)–Ru** by 0.1 M BI(OH)H were estimated as η<sup>q</sup> > 99%, indicating that the excited states of **Ir(pyr)** and **Ir(pyr)–Ru** were almost completely quenched by BI(OH)H under these reaction conditions. **Figure 6** shows that **Ir(pyr)** does work as a PS for CO<sup>2</sup> reduction when using BI(OH)H. The time profiles for the mixture of **Ir(pyr)** and **Ru** showed a linear increase reaching a TONHCOOH = ∼2,000 during 24-h irradiation, indicating that **Ir(pyr)** has a high durability during photocatalytic CO<sup>2</sup> reduction. Although **Ir(pyr)–Ru** also worked as a photocatalyst for CO<sup>2</sup> reduction, it showed a lower activity compared to the mixture of **Ir(pyr)** and **Ru**. While the initial formation rates of the products were similar, the reaction stopped after just 5 h of irradiation in the case of **Ir(pyr)–Ru**. Because the reaction solution of **Ir(pyr)–Ru** was decolorized during the photocatalytic reaction, the low activity of **Ir(pyr)– Ru** would result from its low durability. The decoloration was also observed in irradiation experiments of [Ir(piq)2(dmb)]<sup>+</sup> and ED without the CAT because of hydrogenation of the ligands in [Ir(piq)2(dmb)]<sup>+</sup> (Kuramochi and Ishitani, 2016). Thus, it is also expected that the decoloration of **Ir(pyr)–Ru** is caused by hydrogenation of the Ir unit. In **Ir(pyr)–Ru**, the accumulated electron(s) might be stabilized because of electron hopping between the Ir and Ru units, which might be enhanced by the hydrogenation of the Ir unit.

**Figure 7** illustrates the time profiles of the products during the irradiation of CO2-saturated DMA/TEOA (5:1 v/v, 2.0 ml) solutions containing **Ir(pyr)** or [Ru(dmb)3] <sup>2</sup><sup>+</sup> as PS in the presence of trans-Ru(bpy)(CO)2Cl<sup>2</sup> (bpy = 2,2-bipyridine) and BI(OH)H as CAT and ED, respectively. The initial formation rate of HCOOH in the system using **Ir(pyr)** (**Figure 7A**) is slower than that using [Ru(dmb)3] <sup>2</sup><sup>+</sup> (**Figure 7B**), although **Ir(pyr)** has more intense absorption band at >480 nm than [Ru(dmb)3] 2+ (**Figure 3**). In **Figure 7**, trans-Ru(bpy)(CO)2Cl<sup>2</sup> is used instead of **Ru**. Although trans-Ru(bpy)(CO)2Cl<sup>2</sup> has a much less negative reduction potential (−1.51 V vs. Ag/AgNO3; Kuramochi et al., 2015) than **Ru** (−1.66 V vs. Ag/AgNO3), the initial formation rate of HCOOH in the system using trans-Ru(bpy)(CO)2Cl<sup>2</sup> (**Figure 7A**) is similar to that using **Ru** (**Figure 6A**), suggesting that the electron transfer process from the one-electron reduced **Ir(pyr)** to CAT is not the rate-determining step and does not significantly affect the reaction rate. Considering that the emission of **Ir(pyr)** is almost completely quenched, the slow initial formation rate of HCOOH in **Ir(pyr)** would result from the competitive back-electron transfer process soon after the electron transfer from BI(OH)H to the excited state of **Ir(pyr)** in the solvent cage (Kavarnos, 1993; Nakada et al., 2015).

# CONCLUSION

Photocatalytic CO<sup>2</sup> reduction using **Ir(pyr)** as PS, which has a strong absorption in the visible region, proceeded efficiently for more than 1 day when a suitable electron donor, BI(OH)H, and **Ru** were used as CAT. A new supramolecular photocatalyst, **Ir(pyr)–Ru**, was successfully synthesized, which exhibited a similar reaction rate during the initial stage of CO<sup>2</sup> reduction to that of the mixed system, but the durability of **Ir(pyr)– Ru** was lower than that of the mixed system. While Ir(pyr) showed high durability in the mixed system, the initial formation rate of HCOOH tended to be slower than that of the catalytic system using [Ru(dmb)3] <sup>2</sup><sup>+</sup> as PS, which is possibly due to the faster back-electron transfer from the reduced Ir(pyr) to the oxidized BI(OH)H.

## EXPERIMENTAL SECTION

### General Procedure

All chemicals and solvents were of commercial reagent quality and were used without further purification unless otherwise stated. DMA was dried over molecular sieves of size 4 Å and distilled under reduced pressure. TEOA was distilled under reduced pressure. Tetraethylammonium tetrafluoroborate was dried in vacuo at 100◦C overnight before use. [Ir(piq)2(BL)](PF6) (Kuramochi and Ishitani, 2016), BNAH (Mauzerall and Westheimer, 1955), BI(OH)H (Hasegawa et al.,

FIGURE 6 | Time dependences of the products during the photo-irradiation of CO2-saturated DMA/TEOA (5:1 v/v, 2.0 ml) solutions containing (A) a mixed system of Ir(pyr) (0.05 mM) and Ru (0.05 mM) or (B) Ir(pyr)–Ru (0.05 mM) in the presence of BI(OH)H (0.1 M): CO (◦), HCOOH () and H<sup>2</sup> (1). A 500-W high-pressure Hg lamp was used for the irradiation (λ > 480 nm).

2005; Zhu et al., 2008), **Ru** (Anderson et al., 1995), and BL (Sun et al., 1997) were synthesized according to literature procedures. <sup>1</sup>H NMR spectra were recorded on an AL400 NMR spectrometer. IR spectra were measured in dichloromethane on a JASCO FT/IR-610 spectrometer. Electrospray ionization–mass spectroscopy (ESI-MS) was undertaken using a SHIMADZU LCMS-2010A system with acetonitrile as a mobile phase. UV– vis absorption spectra were recorded with a JASCO V-670 instrument. Emission spectra were measured at 25◦C under an Ar atmosphere using a JASCO FP-8600 spectrofluorometer with correlation for the detector sensitivity. Emission quenching experiments were performed in DMA or DMA/TEOA (5:1 v/v) solutions containing a complex and several different concentrations of BNAH or BI(OH)H.

### Emission Spectral Fitting

Double-mode Franck–Condon band shape analysis was used to fit the emission spectra. The spectral fittings were carried out according to the following equation (Caspar et al., 1984) using the Wavemetrics Igor software.

$$I(\tilde{\nu}) = \sum\_{n\_1=0}^{5} \sum\_{n\_2=0}^{5} \left( \frac{E\_{00} - n\_1 \tilde{\nu}\_1 - n\_2 \tilde{\nu}\_2}{E\_{00}} \right)^4 \left( \frac{S\_1^{\eta\_1}}{n\_1!} \right) \left( \frac{S\_2^{\eta\_2}}{n\_2!} \right) \tag{1}$$

$$\exp \left[ -4 \log 2 \left( \frac{\tilde{\nu} - E\_{00} + n\_1 \tilde{\nu}\_1 + n\_2 \tilde{\nu}\_2}{\tilde{\nu}\_{1/2}} \right)^2 \right]$$

I(ν) is the relative emission intensity at frequency ν. E<sup>00</sup> is the energy gap between the zeroth vibrational levels in the ground and excited states, n<sup>1</sup> and n<sup>2</sup> are the vibrational quantum numbers of the high- and low-frequency vibrational modes, respectively, S<sup>1</sup> and S<sup>2</sup> are the Huang–Rhys factors, and ν1/<sup>2</sup> is the half-width at half-maximum (fwhm) of the individual vibronic band. The 0–0 band energy gaps between the lowest excited state and the ground state were obtained from the emission spectral fitting (**Figure S1**).

# [Ir(pyr-mq)2(CH3CN)2](PF6)

[(pyr-mq)2Ir-µ-Cl]<sup>2</sup> (100 mg, 5.4 × 10−<sup>5</sup> mol), AgPF<sup>4</sup> (31 mg, 1.2 × 10−<sup>4</sup> mol), and acetonitrile (10 ml) were placed in a 50 ml flask. The mixture was stirred for 4 h at room temperature. The resulting suspension was filtered through Celite pad to remove AgCl. The filtrate was concentrated to ca. 1 ml, and the product was precipitated by the addition of diethyl ether (10 ml). After cooling the suspension for 30 min at 0◦C, the product was collected by filtration, giving 96 mg (80%) of the titled compound as a dark orange solid: <sup>1</sup>H NMR (CDCl3, 400 MHz) δ 9.18 (brs, 2H), 8.73 (d, J = 8.8 Hz, 2H), 8.56 (s, 2H), 8.23 (d, J = 7.2 Hz, 2H), 8.15–7.95 (m, 8H), 7.85–7.70 (m, 6H), 7.27 (m, 2H), 6.40 (brs, 2H), 3.23 (s, 6H), 2.22 (s, 6H, CH3CN).

# [Ir(pyr-mq)2(BL)](PF6)

The bridging ligand (BL, 80 mg, 2.2 × 10−<sup>4</sup> mol), dichloromethane (40 ml), and methanol (20 ml) were placed in a 100-ml flask, and the system was purged with argon gas. A solution of [Ir(pyr-mq)2(CH3CN)2](PF6) (80 mg, 7.3 × 10−<sup>5</sup> mol) in dichloromethane (10 ml) and methanol (5 ml) was then added dropwise at room temperature. The reaction mixture was stirred at room temperature. The reaction progress was monitored with ESI-MS. After stirring for 48 h, the resulting solution was evaporated. The residue was purified with a silica gel column (1 to 2 vol% methanol in dichloromethane). The second red band eluted with 2 vol% methanol was collected and evaporated to dryness, giving 42 mg (43% based on the iridium precursor) of the titled complex as a dark red solid: ESI-MS m/z: 1,243 ([M–PF<sup>−</sup> 6 ] <sup>+</sup>). <sup>1</sup>H NMR (CDCl3, 400 MHz) δ 8.94 (brs, 1H), 8.92 (brs, 1H), 8.60 (d, J = 2.8 Hz, 2H), 8.51 (d, J = 5.2 Hz, 1H), 8.45 (d, J = 5.2 Hz, 1H), 8.24–8.12 (m, 8H), 8.02 (s, 1H), 8.00 (s, 1H), 7.91–7.76 (m, 8H), 7.58 (s, 2H), 7.52–7.38 (m, 5H), 7.35 (m, 1H), 7.25–7.21 (m, 1H), 7.09 (d, J = 4.4 Hz, 1H), 6.99–6.88 (m, 3H), 6.77 (m, 1H), 3.13–3.00 (m, 4H), 2.97 (s, 3H), 2.95 (s, 3H), 2.45 (s, 3H), 2.42 (s, 3H).

### Ir(pyr)–Ru

An acetone/ethenol (1:2 v/v) mixed solution (6 ml) containing [Ru(CO)2Cl2]<sup>n</sup> (9.9 mg, 4.3 × 10−<sup>5</sup> /n mol) was refluxed for 1 h, and then [Ir(pyr-mq)2(BL)](PF6) (30 mg, 2.2 × 10−<sup>5</sup> mol) was added to it. The reaction mixture was heated to 60◦C and stirred for 4.5 h under Ar atmosphere. As the reaction proceeded, the starting red solution became a red suspension. The resulting solid was filtered and washed with ethanol. The solid was dissolved in dichloromethane (ca. 2 ml) and filtered to remove insoluble materials. The solution was evaporated to afford 30 mg (87%) of the titiled compound as a dark red solid. ESI-MS m/z: 1,471 ([M–PF<sup>−</sup> 6 ] <sup>+</sup>). FT-IR νCO/cm−<sup>1</sup> : 1,992, 2,058. Anal. calcd (%) for C78H54Cl2F6IrN6O2PRu· 3H2O: C, 56.08; N, 5.03; H, 3.62. Found (%): C, 55.88; N, 4.81; H, 3.21. <sup>1</sup>H NMR (CDCl3, 400

### REFERENCES

Anderson, P. A., Deacon, G. B., Haarmann, K. H., Keene, F. R., Meyer, T. J., Reitsma, D. A., et al. (1995). Designed synthesis of mononuclear MHz) δ 8.97–8.88 (m, 4H), 8.74 (brs, 1H), 8.65–8.62 (m, 3H), 8.42 (brs, 1H), 8.30 (brs, 1H), 8.25–8.21 (m, 2H), 8.18–8.15 (m, 2H), 8.03–8.01 (m, 2H), 7.96–7.79 (m, 8H), 7.61 (s, 1H), 7.56 (s, 1H), 7.53–7.49 (m, 2H), 7.45–7.40 (m, 4H), 7.36 (m, 1H), 7.28–7.24 (m, 1H), 7.09 (d, J = 5.6 Hz, 1H), 6.98–6.88 (m, 3H), 3.05–2.94 (m, 4H), 2.99 (s, 3H), 2.98 (s, 3H), 2.61 (s, 3H), 2.47 (s, 3H).

# Photocatalytic CO<sup>2</sup> Reduction

DMA-TEOA (2 ml; 5:1 v/v) solutions containing a mixture of PS and CAT or the supramolecular **Ir(pyr)–Ru** complex and BI(OH)H were bubbled with CO<sup>2</sup> for 30 min. Photo-irradiations were carried out in 11-mL Pyrex tubes (i.d. = 8 mm) with light at λ > 480 nm using a 500-W high-pressure Hg lamp combined with a K2CrO<sup>4</sup> solution filter (30% w/w, optical path length: 1 cm) using a merry-go-round apparatus. The reaction temperature was maintained at 25◦C using an IWAKI constant-temperature system (CTS-134A). The gaseous reaction products (CO and H2) were quantified with GC-TCD (GL Science GC323), and the product (formate) in the solutions was analyzed with a capillary electrophoresis system (Otuka Electronics Co.CAPI-3300I).

### Computational Methods

DFT calculations were carried out using the Gaussian 09 package of programs (Frisch et al., 2009). Each structure was fully optimized using the B3LYP functional using the 6-31G(d,p) basis set for all atoms except Ir and the standard doubleζ type LANL2DZ basis set with the effective core potential of Hay–Wadt for Ir. The calculation was carried out by using the polarizable continuum model (PCM) with default parameter for DMA. The stationary points were verified using the vibrational analysis.

# AUTHOR CONTRIBUTIONS

YK concieved the research and conducted experiments. OI directed the project and co-wrote the paper.

### FUNDING

This work was supported by JST CREST (grant number JPMJCR13L1) and the Strategic International Collaborative Research Program (SICORP) of JST.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00259/full#supplementary-material

Tris(heteroleptic) ruthenium complexes containing bidentate polypyridyl ligands. Inorg. Chem. 34, 6145–6157. doi: 10.1021/ic00128a028

Bonin, J., Robert, M., and Routier, M. (2014). Selective and efficient photocatalytic CO<sup>2</sup> reduction to CO using visible light and an iron-based homogeneous catalyst. J. Am. Chem. Soc. 136, 16768–16771. doi: 10.1021/ ja510290t


visible light in Aqueous solution. Angew. Chem. Int. Ed. 56, 4867–4871. doi: 10.1002/anie.201701627


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Kuramochi and Ishitani. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Ruthenium Picolinate Complex as a Redox Photosensitizer With Wide-Band Absorption

Yusuke Tamaki, Kazuma Tokuda, Yasuomi Yamazaki † , Daiki Saito, Yutaro Ueda and Osamu Ishitani\*

Department of Chemistry, Tokyo Institute of Technology, Tokyo, Japan

### Edited by: Hitoshi Ishida,

Kitasato University, Japan

### Reviewed by:

Wee Han Ang, École Polytechnique Fédérale de Lausanne, Switzerland Mahmut Özacar, Sakarya University, Turkey

> \*Correspondence: Osamu Ishitani ishitani@chem.titech.ac.jp

### †Present Address:

Department of Materials and Life Science, Seikei University, Tokyo, Japan

### Specialty section:

This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry

Received: 30 January 2019 Accepted: 24 April 2019 Published: 14 May 2019

### Citation:

Tamaki Y, Tokuda K, Yamazaki Y, Saito D, Ueda Y and Ishitani O (2019) Ruthenium Picolinate Complex as a Redox Photosensitizer With Wide-Band Absorption. Front. Chem. 7:327. doi: 10.3389/fchem.2019.00327

Ruthenium(II) picolinate complex, [Ru(dmb)2(pic)]<sup>+</sup> (Ru(pic); dmb = 4,4′ -dimethyl-2,2′ -bipyridine; Hpic = picolinic acid) was newly synthesized as a potential redox photosensitizer with a wider wavelength range of visible-light absorption compared with [Ru(N∧N)3] <sup>2</sup><sup>+</sup> (N∧N = diimine ligand), which is the most widely used redox photosensitizer. Based on our investigation of its photophysical and electrochemical properties, Ru(pic) was found to display certain advantageous characteristics of wideband absorption of visible light (λabs < 670 nm) and stronger reduction ability in a oneelectron reduced state (E red <sup>1</sup>/<sup>2</sup> = −1.86 V vs. Ag/AgNO3), which should function favorably in photon-absorption and electron transfer to the catalyst, respectively. Performing photocatalysis using Ru(pic) as a redox photosensitizer combined with a Re(I) catalyst reduced CO<sup>2</sup> to CO under red-light irradiation (λex > 600 nm). TONCO reached 235 and 8CO was 8.0%. Under these conditions, [Ru(dmb)3] <sup>2</sup><sup>+</sup> (Ru(dmb)) is not capable of working as a redox photosensitizer because it does not absorb light at λ > 560 nm. Even in irradiation conditions where both Ru(pic) and Ru(dmb) absorb light (λex > 500 nm), using Ru(pic) demonstrated faster CO formation (TOFCO = 6.7 min−<sup>1</sup> ) and larger TONCO (2347) than Ru(dmb) (TOFCO = 3.6 min−<sup>1</sup> ; TONCO = 2100). These results indicate that Ru(pic) is a superior redox photosensitizer over a wider wavelength range of visible-light absorption.

Keywords: redox photosensitizer, CO2 reduction, photocatalyst, Ruthenium(II) complex, wide-band absorption

# INTRODUCTION

Redox photosensitizers, which absorb visible light and facilitate the electron transfer process, play a key role in various photochemical reactions, such as CO<sup>2</sup> reduction (Takeda et al., 2017; Tamaki and Ishitani, 2017), water oxidation (Fukuzumi et al., 2016), hydrogen evolution (Schulz et al., 2012), and organic synthesis (Prier et al., 2013). Effective photosensitizers should be endowed with three important properties, including (1) visible-light absorption, (2) a long lifetime in the excited state to initiate the electron transfer process, and (3) reducing and/or oxidizing power that is strong enough to donate electrons or holes to the catalyst. In particular, the utilization of visible-light over a wider range of wavelengths is important both to utilize sunlight efficiently and avoid the internal filter effect and side reactions that are commonly caused by the light-absorption of catalysts and/or electron donor/acceptor. Ru(II) complexes coordinated with three diimine ligands, [Ru(N∧N)3] 2+ (N∧N = diimine ligand) are the most widely used redox photosensitizers in various photochemical redox reactions because these types of complexes exhibit strong absorption in the visible-light region and have a long lifetime in their triplet metal-to-ligand charge-transfer (3MLCT) excited states (Juris et al., 1988; Thompson et al., 2013).

However, one of the disadvantages of [Ru(N∧N)3] <sup>2</sup>+-type photosensitizers is the limited access to the wavelength region of visible light, e.g., λabs < 560 nm in the cases of N∧N = 2,2′ -bipyridine (bpy) and 4,4′ -dimethyl-2,2′ -bipyridine (dmb), and these complexes cannot utilize visible light having lower energy (λ > 560 nm). To overcome this, ligand-modified Ru(II) photosensitizers have been reported. For example, Ru(II) complexes have an extended π-system for photodynamic therapy (Zhang et al., 2017) and multinuclear Ru(II) complexes by conjugated bridging ligand are used for hydrogen evolution (Tsuji et al., 2018). However, these modifications lower the reducing power of photosensitizers and limit the choice of catalyst especially for the reduction of CO2. On the other hand, we have reported an osmium(II) analog, i.e., [Os(N∧N)3] 2+, which could function as a redox photosensitizer utilizing a much wider wavelength range of visible light (λabs < 700 nm) due to its singlet-to-triplet direct excitation (S-T absorption) and drive photocatalytic CO<sup>2</sup> reduction by red-light irradiation (λex > 620 nm) in the combination with rhenium(I) catalyst unit (Tamaki et al., 2013b), whereas the high toxicity of OsVIIIO<sup>4</sup> inhibits the wider application of osmium complexes.

Therefore, we developed a novel ruthenium(II) redox photosensitizer that can utilize a wider wavelength range of visible light than [Ru(N∧N)3] <sup>2</sup>+. In the photocatalytic system for CO<sup>2</sup> reduction, a photosensitizer mediates an electron from a sacrificial electron donor to a catalyst. Since the positive shift of the LUMO level of redox photosensitizer should limit the choice of a catalyst for reducing CO2, for the expansion of the useable wavelength range, we try to decrease the energygap between HOMO and LUMO by the negative shift of the HOMO level, while maintaining the LUMO level. We introduced anionic electron-donating picolinate instead of a diimine ligand into a ruthenium complex (Norrby et al., 1997; Couchman et al., 1998). [Ru(dmb)2(pic)]<sup>+</sup> (**Ru(pic)**; Hpic = picolinic acid) was synthesized, and we investigated its photophysical properties and functions as a redox photosensitizer using [Ru(dmb)3] <sup>2</sup><sup>+</sup> (**Ru(dmb)**) as a reference redox photosensitizer and Re(dmb)(CO)3Br (**Re**) as a catalyst for the reduction of CO<sup>2</sup> (Hawecker et al., 1983; Gholamkhass et al., 2005; Tamaki et al., 2016). **Chart 1** shows structures and abbreviations of the metal complexes used.

# RESULTS AND DISCUSSION

**Figure 1** displays UV-vis absorption spectra of **Ru(pic)**, **Ru(dmb),** and **Re** measured in N,N-dimethylacetamide (DMA). **Ru(pic)** exhibited a broad singlet MLCT absorption band at λabs = 450–640 nm, with molar absorptivity at an absorption maximum (λmax = 498 nm) of 1.04 × 10<sup>4</sup> M−<sup>1</sup> cm−<sup>1</sup> , which was red-shifted in wavelength compared to that of **Ru(dmb)** (λabs = 420–550 nm). The absorption band attributed to the π-π ∗ transition of dmb ligands was observed at 294 nm. According to this result, **Ru(pic)** have the potential to utilize visible light over a wider range of wavelengths (λabs < 670 nm) than **Ru(dmb)** (λabs < 560 nm). This expected red-shift of the MLCT band should be induced by the stronger electron-donating ability of the picolinate ligand to negatively shift the energy level of HOMO.

**Ru(pic)** exhibited phosphorescence from its <sup>3</sup>MLCT excited state (**Figure 2**) with a quantum yield of 8em = 0.8% and a lifetime of τem = 66 ns. Emission spectrum of **Ru(pic)** (λem = 734 nm) was also red-shifted compared to that of **Ru(dmb)** (λem = 638 nm). The quantum yield and lifetime of **Ru(pic)** were smaller and shorter than those of **Ru(dmb)** (8em = 9.1%, τem = 741 ns) due to the 12-times faster non-radiative deactivation process (**Ru(pic)**: knr = 1.5 × 10<sup>7</sup> s −1 ; **Ru(dmb)**: knr = 1.2 × 10<sup>6</sup> s −1 ), which is a reasonable behavior from energy-gap law. **Table 1** summarizes photophysical properties of **Ru(pic)** along with those of **Ru(dmb)** and **Re**.

**Figure 3** shows the cyclic voltammograms of **Ru(pic)** and **Ru(dmb)** and their redox potentials are summarized in **Table 2** along with that of **Re**. **Ru(pic)** displayed two reversible reduction waves and a reversible oxidation wave, which are attributable to the subsequent reduction of two dmb ligands and the oxidation couple of RuIII/II, respectively. Both the first reduction (E red <sup>1</sup>/<sup>2</sup> = −1.86 V vs. Ag/AgNO3) and oxidation (E 0x <sup>1</sup>/<sup>2</sup> = 0.41 V) waves were observed at more negative potentials than those of **Ru(dmb)** (E red <sup>1</sup>/<sup>2</sup> = −1.74 V and E 0x <sup>1</sup>/<sup>2</sup> = 0.77 V), which should be induced by the stronger electron-donating ability of the picolinate ligand. The stronger reducing power of one-electron reduced species

Chart 1 | Structures and abbreviations of complexes used.

(OERS) of **Ru(pic)** (E red <sup>1</sup>/<sup>2</sup> = −1.86 V) facilitates an increase in the number of choices of applicable catalyst because the electron transfer from OERS of **Ru(pic)** to a catalyst must occur during photocatalysis in the case of reductive quenching mechanisms. When using **Ru(pic)** as a photosensitizer and **Re** as a catalyst, the electron transfer process from OERS of **Ru(pic)** to **Re** (E red <sup>1</sup>/<sup>2</sup> = −1.76 V) occurs exothermically.

These results indicated that **Ru(pic)** had some advantages with respect to its function as a redox photosensitizer compared with **Ru(dmb)**, including its wider wavelength range of visiblelight absorption and stronger reducing power of OERS, which is effective in the electron transfer to the catalyst. However, certain unfavorable properties were also observed, i.e., a shorter lifetime (τem = 66 ns) and weaker oxidizing power in its excited state (1E = E(**Ru(dmb)**<sup>∗</sup> /**Ru(dmb)**−)–E(**Ru(pic)**<sup>∗</sup> /**Ru(pic)**−) = 0.28–(−0.11) = 0.39 V). In the reductive quenching process, an excited photosensitizer accepts an electron from a sacrificial electron donor. Weaker oxidation power in the excited state of a photosensitizer should decrease the driving force of this electron transfer process. In addition, since this process competes with the radiative and non-radiative deactivation processes from the excited state of a photosensitizer by itself, the shorter lifetime results in less opportunity of the reductive quenching process to occur. To evaluate whether reductive quenching occurs, the emission intensity from **Ru(pic)** was compared in the presence of five different concentrations of a sacrificial electron donor, 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) (Tamaki et al., 2013a; Hasegawa et al., 2015) in DMA-triethanoamine (TEOA; 5:1 v/v). As shown in **Figure 4**, the emission intensities from the <sup>3</sup>MLCT excited state of **Ru(pic)** decreased at higher concentrations of BIH, which indicated that the excited **Ru(pic)** was quenched by BIH. The quenching rate constant was determined to be k<sup>q</sup> = 1.7 × 10<sup>8</sup> M−<sup>1</sup> s −1 from the Stern-Volmer plot (**Figure S1**) and the lifetime of the emission (τem = 66 ns), which was 8-times slower than that of **Ru(dmb)** (k<sup>q</sup> = 1.4 × 10<sup>9</sup> M−<sup>1</sup> s −1 ) as expected from the weaker oxidizing power in the <sup>3</sup>MLCT excited state of **Ru(pic)**. In the photocatalytic reaction condition, i.e., [BIH] = 0.2 M, 69% of the excited **Ru(pic)** was estimated to be quenched by BIH, which should be enough to initiate a photocatalytic reaction.

To clarify the produced species as a result of the quenching of excited **Ru(pic)** by BIH, UV-vis absorption spectral change was observed during photo-irradiation of **Ru(pic)** in the presence of BIH (**Figure 5**). Irradiation by light at λex = 480 nm caused spectral changes and new absorption bands appeared at λabs = 420 and 547 nm. The shape of differential absorption spectra before and after irradiation (**Figure 5B**) were quite similar to that of OERS of **Ru(pic)** obtained by electrochemical spectroscopy (**Figure S2**). These results indicate that the reductive quenching

FIGURE 3 | Cyclic voltammograms of Ru(pic) and Ru(dmb) measured in a DMA solution containing Et4NBF4 (0.1 M) as a supporting electrolyte with a Ag/AgNO3 (10 mM) reference electrode.


TABLE 1 | Photophysical properties of Ru(pic), Ru(dmb), and Re. a .

<sup>a</sup>Measured in DMA. <sup>b</sup>Excitation wavelength: 480 nm. <sup>c</sup>Excitation wavelength: 510 nm. <sup>d</sup>Rate constants for radiative deactivation calculated as k<sup>r</sup> = 8em/τem. <sup>e</sup>Rate constants for non-radiative deactivation calculated as knr = (1–8em)/τem. <sup>f</sup> Energy for 0-0 transition obtained from Franck-Condon analyses of the emission spectra.



<sup>a</sup>Measured in a DMA solution containing the complex (0.5 mM) and Et4NBF<sup>4</sup> (0.1 M) with a scan rate of 200 mV·s <sup>−</sup><sup>1</sup> under an Ar atmosphere. <sup>b</sup>Redox potentials of the photosensitizers (PS) in their excited states were calculated from E0<sup>x</sup> 1/2 -E<sup>00</sup> and Ered <sup>1</sup>/<sup>2</sup> + E00, respectively.

of the <sup>3</sup>MLCT excited state of **Ru(pic)** by BIH proceeded successfully to give OERS of **Ru(pic)** (Equation 1) and **Ru(pic)** can be expected to function as a redox photosensitizer over the wide-range absorption of visible light.

$$\mathsf{Ru(pic)} \xrightarrow[\mathsf{RMod-CT}]{\mathsf{RMod-CT}} \mathsf{Ru(pic)} \xrightarrow[\mathsf{Roll}]{\mathsf{RMod-CT}} \mathsf{Ru(pic)}\tag{1}$$

The results of photocatalytic reactions for the reduction of CO<sup>2</sup> are summarized in **Table 3**. In a typical run of photocatalytic reactions, a mixed solution of DMA-TEOA (5:1 v/v) containing **Ru(pic)** (50µM), **Re** (50µM), and BIH (0.2 M) as a sacrificial electron donor was irradiated under a CO<sup>2</sup> atmosphere using light at λex > 620 nm. CO production proceeded linearly and selectively and the turnover number for CO production (TONCO) was 235 after 36 h of irradiation (**Figure 6A**). The quantum yield for CO formation (8CO) was determined to be 8CO = 8% using λex = 600-nm light (light intensity: 6.0 × 10−<sup>9</sup> einstein·s −1 ). By contrast, when using **Ru(dmb)** as a redox photosensitizer instead of **Ru(pic)**, no photocatalysis proceeded (**Figure 6A**) because **Ru(dmb)** does not absorb lower-energy light at λex > 620 nm (**Figure 1**). To compare the function as a redox photosensitizer, the photocatalytic reactions were also conducted under photo-irradiation condition, where both **Ru(pic)** and **Ru(dmb)** absorb incident light (λex > 480 nm). In this condition, both systems photocatalytically produced CO with high selectivity. **Figure 6B** shows the time course of photocatalytic CO production using light at λex > 500 nm, and the system using **Ru(pic)** formed CO faster (TOFCO = 6.7 min−<sup>1</sup> ) than **Ru(dmb)** (TOFCO = 3.6 min−<sup>1</sup> ) in the initial stage of photocatalysis. TONCO reached 2347 and 2100 after 36 h of irradiation using **Ru(pic)** and **Ru(dmb)**, respectively. The values of 8CO using light at λex = 480 nm (light intensity: 6.0 × 10−<sup>9</sup> einstein·s −1 ) were 10% and 44% in the cases using **Ru(pic)** and **Ru(dmb)**, respectively. The **Ru(pic)** system demonstrated similar 8CO values in both irradiation conditions (λex = 600 and 480 nm). These results indicated that **Ru(pic)** has a clear advantage of a wider wavelength range of utilizable visible light compared to **Ru(dmb)**, even for the photocatalytic condition of λex > 480 nm. Since **Ru(pic)** displays larger molar absorptivity in the λabs > 480-nm region and a wider wavelength range than **Ru(dmb)** (**Figure 1**), **Ru(pic)** absorbs a much larger number of photons at λex > 480-nm, which leads to a faster TOFCO and larger TONCO, even though the quantum yields for CO production were lower.

The quantitative analyses of BIH and its oxidized compound during photocatalysis were conducted in the system using 0.1 M of BIH to simplify the HPLC analyses. As the only oxidized compound of BIH, two-electron oxidized and deprotonated BIH (BI+) was observed (Equation 2).

$$\otimes\_{\mathsf{X}}^{\mathsf{T}} \otimes \xleftarrow{\colon\_{\mathsf{X}}} \otimes \xleftarrow{\colon\_{\mathsf{X}}} \otimes \xleftarrow{\colon\_{\mathsf{X}}} \tag{2}$$

**Figure 7** shows the change in the amounts of both BIH and BI<sup>+</sup> during photocatalytic reaction along with the amount of CO produced. The amount of produced BI<sup>+</sup> was fairly similar to that of CO. For example, after 20 h of irradiation, 205 µmol of BI<sup>+</sup> and 203 µmol of CO formed. CO is the two-electron reduced compound of CO2, and BIH supplies two electrons per molecule to give BI<sup>+</sup> as a oxidized form. These results clearly indicate that BIH acted as a two-electron donor in the photocatalytic reactions using **Ru(pic)** as a redox photosensitizer (Equation 3).

$$\text{CO}\_2 + \text{BlH} \xrightarrow[\text{DMM-TEOM}]{\text{Ru(pic)} + \text{Be} \wedge \text{hv}} \text{CO} + \text{Bl}^+ + \text{[O}^{2-}] + \text{H}^+ \tag{3}$$

The reaction mechanisms of the photocatalytic reactions using **Ru(pic)** and **Re** were investigated. Since **Re** does not absorb

FIGURE 5 | UV-vis (A) absorption and (B) differential absorption spectral change of a DMA-TEOA (5:1 v/v, 4 mL) solution containing Ru(pic) (0.1 mM) and BIH (0.2 M) during irradiation using light at <sup>λ</sup>ex <sup>=</sup> 480 nm (0–8 min at 1-min intervals). The incident light intensity was 5.0 <sup>×</sup> <sup>10</sup>−<sup>9</sup> einstein·<sup>s</sup> −1 . Blue and red lines represent spectra at 0 and 8-min irradiation, respectively.



<sup>a</sup>A CO2-saturated DMA-TEOA (5:1 v/v) mixed solution containing the photosensitizer (50 µM), Re (50 µM), and BIH (0.2 M) was irradiated. <sup>b</sup>Turnover number for the reaction products after 36 h of irradiation calculated as [product (mol)]/[added Re (mol)]. <sup>c</sup>λex > 620 nm. <sup>d</sup>λex > 500 nm. <sup>e</sup> Irradiation for 12 h. <sup>f</sup>Quantum yield of CO production calculated as [CO (mol)]/[absorbed photon (einstein)]. <sup>g</sup>λex = 600 nm (light intensity: 6.0 × 10−<sup>9</sup> einstein·s −1 ). <sup>h</sup>λex = 480 nm (light intensity: 6.0 × 10−<sup>9</sup> einstein·s −1 ). <sup>i</sup>Quenching rate constants for emission from Ru(II) photosensitizers by BIH obtained from the slopes of Stern-Volmer plots and lifetimes of excited states. <sup>j</sup>Quenching fractions of emission from Ru(II) photosensitizers by 0.2 M of BIH calculated as 0.2kqτem/(1 + 0.2kqτem). <sup>k</sup>Quantum yield for one-electron reduction of the photosensitizer using light at λex = 480 nm (light intensity: 5.0 × 10−<sup>9</sup> einstein·s −1 ).

light at λex > 460 nm, as shown in **Figure 1**, **Ru(pic)** should absorb the irradiated photon selectively under photocatalytic reaction conditions, i.e., λex > 600 nm or > 480 nm. The photon absorption by **Ru(pic)** gives its OERS via the reductive quenching process of its <sup>3</sup>MLCT excited state by BIH, as described above (Equation 1). The reducing power of OERS of **Ru(pic)** (E red <sup>1</sup>/<sup>2</sup> = −1.86 V) is strong enough to trigger electron transfer to **Re** (E red 1/2 = −1.76 V), which functions as a catalyst for the reduction of CO2. The process of two-electron supply using BIH has already been reported in the photocatalytic reaction system using a Ru(II)-Re(I) supramolecular photocatalyst (Tamaki et al., 2013a). The initial process of the photocatalysis is also a photoinduced electron transfer from BIH to the Ru(II) tris-diimine type photosensitizer unit, forming OERS of the photosensitizer unit and one-electron oxidized BIH (BIH· <sup>+</sup>). BIH· <sup>+</sup> is rapidly deprotonated by TEOA to give BI·. TEOA functioned only as a base, but not as a sacrificial electron donor to quench the excited photosensitizer unit. BI· has a strong reducing power (E red <sup>1</sup>/<sup>2</sup> = −1.95 V) (Zhu et al., 2008) enough to provide one more electron to the supramolecular photocatalyst to be converted to BI+. In other words, BIH works as a two-electron donor by one-photon excitation of the photocatalyst via the ECE mechanism. Similar processes should also proceed in the photocatalytic system using **Ru(pic)** and **Re** because both **Ru(pic)** (E red <sup>1</sup>/<sup>2</sup> = −1.86 V) and **Re** (E red <sup>1</sup>/<sup>2</sup> = −1.76 V) have a lower reduction potential than BI· (E red <sup>1</sup>/<sup>2</sup> = −1.95 V). Based on this investigation, the electronsupply processes of BIH are presumed, as depicted in Equation 4.

Photocatalysis using **Ru(pic)** displayed an advantages of a wider wavelength region of visible-light absorption, which achieved both red-light driven CO<sup>2</sup> reduction (λex > 620 nm) and faster CO production than the system using **Ru(dmb)** (λex > 500 nm), whereas the quantum yield for CO formation using **Ru(pic)** (8CO = 10%) was 1/4 the value when **Ru(dmb)** (8CO = 44%) was used. The main reason for smaller 8CO should be the smaller quantum yield of one-electron reduction (8OERS) of **Ru(pic)**. 8OERS of **Ru(pic)** using light at λex = 480 nm (light intensity: 5.0 × 10−<sup>9</sup> einstein·s −1 ) was determined to be 8.3%, which was 1/8 that of **Ru(dmb)** (8OERS = 66%). The elementary processes of one-electron reduction of **Ru(pic)** is displayed in **Scheme 1**. The reductive quenching of the <sup>3</sup>MLCT excited state of **Ru(pic)** by BIH gives an ion pair, [**Ru(pic)**−···BIH· <sup>+</sup>]. If the ion pair dissociate, free OERS and BIH· <sup>+</sup> are obtained. The charge-recombination processes from the ion pair or by the re-collision of OERS of **Ru(pic)** and BIH· + should form **Ru(pic)** and BIH. The differences in properties between **Ru(pic)** and **Ru(dmb)**, i.e., the cationic valence and the reducing power of OERS, should affect each elementary process and consequently the quantum yield for one-electron reduction. Since OERS of **Ru(dmb)** is a monovalent cation, the ion pair with BIH· <sup>+</sup> involves cationic repulsion, which should accelerate the dissociation process. On the other hand, OERS of **Ru(pic)** is zero-valent, which provides no repulsion between BIH· <sup>+</sup>, and therefore, the dissociation process should become slower when using **Ru(pic)** (smaller kesc). In addition, since the reducing power of OERS of **Ru(pic)** (E red <sup>1</sup>/<sup>2</sup> = −1.86 V) is stronger than that of **Ru(dmb)** (E red <sup>1</sup>/<sup>2</sup> = −1.74 V), the driving forces for the charge-recombination processes become larger when **Ru(pic)** is used (larger krec1, krec2). Consequently, the smaller 8OERS using **Ru(pic)** should be induced by the slower dissociation process of the ion pair and the faster charge-recombination processes. The quantitative analyses of the factors controlling 8OERS of photosensitizing complexes are in progress and will be reported elsewhere.

In the photocatalytic reaction conditions, the electronconsuming process for CO<sup>2</sup> reduction via the electron transfer to **Re** (the broken box in Scheme 1) will compete against the charge-recombination by the re-collision of OERS and BIH· +. Therefore, since 8OERSs were determined in the absence of **Re**, 8CO (10%) was larger than the expected value from half of 8OERS (8.3/2 = 4.2%), which was derived from the fact that the reduction of CO<sup>2</sup> to CO is a two-electron reduction process. Higher reduction potential of **Ru(pic)** should operate in favor of the electron transfer to **Re**. Therefore, the ratio of quantum yields for CO<sup>2</sup> reduction between using **Ru(pic)** and **Ru(dmb)**, i.e., 8CO(**Ru(pic)**)/8CO(**Ru(dmb)**) = 10/44 = 0.23, became larger than that for one-electron reduction (8OERS(**Ru(pic)**)/8OERS(**Ru(dmb)**) = 8.3/66 = 0.13). In other words, **Ru(pic)** has another advantage of faster electron transfer to **Re** in the photocatalysis.

### EXPERIMENTS

### General Procedures

<sup>1</sup>H NMR spectra were measured using a JEOL ECA400II (400 MHz) system in solutions of acetone-d6. The residual

protons of acetone-d<sup>6</sup> were used as an internal standard for measurements. Electrospray ionization-mass spectroscopy (ESI-MS) was performed using a Shimadzu LCMS-2010A system with acetonitrile as the mobile phase. UV-vis absorption spectra were measured with a JASCO V-565 spectrophotometer. Emission spectra were measured using a Horiba Fluorolog-3-21 spectrofluorometer equipped with a NIR-PMT R5509-43 near infrared detector. A Horiba FluoroCube time-correlated single-photon counting system was used to obtain emission lifetimes. The excitation light source was a NanoLED-515L pulse lamp (510 nm). A HAMAMATSU absolute PL quantum yield spectrometer C9920-02 was used to determine emission quantum yields. The samples were degassed by Ar-bubbling of solutions for 30 min prior to measuring emissions. Emission quenching experiments were performed on solutions containing the complexes and five different concentrations of BIH. The quenching rate constants k<sup>q</sup> were calculated from linear Stern-Volmer plots for the emission from the <sup>3</sup>MLCT excited state of the photosensitizing complexes and their lifetimes. The redox potentials of the complexes were measured in an Ar-saturated DMA solution containing Et4NBF<sup>4</sup> (0.1 M) as a supporting electrolyte using cyclic voltammetric techniques performed with an ALS CHI-720Dx electrochemical analyzer with a glassy carbon disk working electrode (3 mm diameter), a Ag/AgNO<sup>3</sup> (10 mM) reference electrode, and a Pt counter electrode. The supporting electrolyte was dried under vacuum at 100◦C for 1 day prior to use. The scan rate was 200 mV·s −1 .

### Photocatalytic Reactions

Photocatalytic reactions were performed in DMA–TEOA (5:1 v/v) solutions containing the photosensitizer (50µM), **Re** (50µM), and BIH (0.2 M). After the solution was purged with CO<sup>2</sup> for 20 min, the solution was irradiated. For TON measurements, the mixed solution (2 mL) in an 11 mL test tube (i.d. 8 mm) was irradiated in a merry-go-round apparatus using λex > 620 nm light from a halogen lamp equipped with a Rhodamin B (0.2% w/v, d = 1 cm) solution filter or λex > 500 nm light from a high-pressure Hg lamp equipped with a uranyl glass and a K2CrO<sup>4</sup> (30% w/w, d = 1 cm) solution filter. During irradiation, the temperature of the solution was maintained at 25◦C using an EYELA CTP-1000 constant-temperature system. For quantum yield measurements, the mixed solution in a quartz cubic cell (11 mL, light pass length: 1 cm) was irradiated in a Shimadzu photoreaction quantum yield evaluation system QYM-01 using 600 nm or 480 nm light from a 300 W Xe lamp equipped with a 600 nm or 480 nm (FWHM: 10 nm) bandpass filters. The temperature of the solution was controlled during irradiation at 25 ± 0.1◦C using an IWAKI CTS-134A constanttemperature system. The gaseous products of photocatalysis, i.e., CO and H2, were analyzed by GC-TCD (GL science GC323). A capillary electrophoresis system (Agilent 7100) was used to analyze HCOOH. HPLC analyses for BIH and BI<sup>+</sup> were conducted using a JASCO 880-PU pump, a Develosil ODS-UG-5 column (250 × 4.6 mm), a JASCO 880–51 degasser, and a JASCO UV-2070 detector. The column temperature was maintained at 30◦C using a JASCO 860-CO oven. The mobile phase was a 6:4 (v/v) mixture of acetonitrile and a NaOH– KH2PO<sup>4</sup> buffer solution (50 mM, pH 7) with a flow rate of 0.5 mL·min−<sup>1</sup> .

### Electrochemical Spectroscopy

Electrochemical spectroscopy to determine the molar absorptivity of OERS was performed using a JASCO PU-980 pump and an EC Frontier flow-type electrolysis cell VF-2 equipped with a carbon felt working electrode (18 mm diameter), a Ag/AgNO<sup>3</sup> (10 mM) reference electrode, and a Pt wire counter electrode in an Ar-saturated acetonitrile solution of **Ru(pic)** (0.5 mM) and Et4NBF<sup>4</sup> (0.1 M) as a supporting electrolyte. Applied potential was controlled using an ALS CHI-720Dx electrochemical analyzer and UV-vis absorption spectra were measured using a Photal MCPD-9800 spectrometer (Otsuka Electronics) and a flow-type transmission cell (light pass length: 1.5 mm) (Ishitani et al., 1994).

# Quantum Yields for One-Electron Reduction of Photosensitizers

A 4-mL DMA–TEOA (5:1 v/v) solution of the photosesnsitizer (0.1 mM) and BIH (0.2 M) in a quartz cubic cell (light pass length: 1 cm) was purged with Ar for 20 min, and then irradiated with the 500-W Xe lamp combined with a 480-nm (FWHM = 10 nm) bandpass filter (Asahi Spectra Co.), ND filter, and a 5-cm-long H2O solution filter. UV–vis absorption spectral changes during irradiation were measured using a Photal MCPD-9800 spectrometer (Otsuka Electronics). The light intensity was determined as 5.0 × 10−<sup>9</sup> einstein·s −1 using a K3Fe(C2O4)<sup>3</sup> actinometer.(Hatchard and Parker, 1956) The amount of OERS of **Ru(pic)** was calculated using the molar absorption coefficient of OERS (500–700 nm) obtained by electrochemical spectroscopy.

### MATERIALS

DMA was dried over molecular sieves 4A, distilled under reduced pressure (∼10 mmHg) and used in a week. TEOA was distilled under reduced pressure (<1 mmHg) and used in a month. Both solvents were kept under Ar in the dark. All other reagents were of reagent-grade quality and used without further purification.

### Synthesis

**Ru(dmb)** (Sullivan et al., 1978), **Re** (Morimoto et al., 2013), and BIH (Hasegawa et al., 2005; Zhu et al., 2008) were prepared according to the methods reported in the literatures. **Ru(pic)** was synthesized using a method similar to the synthesis of [Ru(bpy)2(pic)](PF6) (bpy = 2,2′ -bipyridine) (Norrby et al., 1997; Couchman et al., 1998), except for using dmb instead of bpy. [Ru(dmb)2(pic)](PF6) (**Ru(pic)**): <sup>1</sup>H NMR (acetone-d6) δ/ppm: 8.81 (d, J = 5.6 Hz, 1H), 8.65 (s, 1H), 8.63 (s, 1H) 8.60 (s, 1H), 8.55 (s, 1H), 8.14 (dd, J = 5.6, 0.8 Hz, 1H), 8.03 (dd, J = 6.4, 2.4 Hz, 1H), 7.94 (d, J = 5.6 Hz, 1H), 7.91 (d, J = 5.6 Hz, 1H), 7.77 (d, J = 5.6 Hz, 1H), 7.70 (dd, J = 5.6, 0.8 Hz, 1H), 7.64 (d, J = 5.6 Hz, 1H), 7.50 (dd, J = 6.4, 2.4 Hz, 1H), 7.43 (dd, J = 5.6, 1.2 Hz, 1H), 7.26 (dd, J = 5.6, 1.2 Hz, 1H), 7.21 (dd, J = 5.6, 1.2 Hz, 1H), 2.67 (s, 3H), 2.58 (s, 3H), 2.55 (s, 3H), 2.49 (s, 3H). ESI-MS (in acetonitrile) m/z: 592 ([M–PF<sup>−</sup> 6 ] <sup>+</sup>). Anal. calcd for

# REFERENCES


C30H28F6N5O2PRu·H2O: C, 47.75; H, 4.01; N, 9.28. Found: C, 47.72; H, 3.75; N, 9.40.

# CONCLUSION

Ruthenium(II) picolinate complex, **Ru(pic)**, successfully functioned as a redox photosensitizer with a much wider wavelength range of visible-light absorption (λabs < 670 nm) compared with a fairly typical **Ru(dmb)** (λabs < 560 nm). The system using **Ru(pic)** as a photosensitizer and **Re** as a catalyst photocatalyzed the reduction of CO<sup>2</sup> to CO by red-light irradiation (λex > 620 nm). TONCO reached 235 and 8CO was 8.0%. Even in the irradiation conditions where **Ru(dmb)** also absorbed light, i.e., λex > 500 nm, the system using **Ru(pic)** demonstrated faster CO formation (TOFCO = 6.7 min−<sup>1</sup> ) and larger TONCO (2347) than that using **Ru(dmb)** (TOFCO = 3.6 min−<sup>1</sup> , TONCO = 2100).

# AUTHOR CONTRIBUTIONS

KT, DS, YY, and YT performed all experiments. YU and OI designed this project. YT wrote the manuscript.

# FUNDING

This work was supported by JSPS KAKENHI Grant Numbers JP17H06440 in Scientific Research on Innovative Areas Innovations for Light-Energy Conversion (I4LEC), JP18K14238, JP17K14526.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00327/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Tamaki, Tokuda, Yamazaki, Saito, Ueda and Ishitani. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Synthesis of a Redox-Active NNP-Type Pincer Ligand and Its Application to Electrocatalytic CO<sup>2</sup> Reduction With First-Row Transition Metal Complexes

Kallol Talukdar, Asala Issa and Jonah W. Jurss\*

*Department of Chemistry and Biochemistry, University of Mississippi, University, MS, United States*

We report the synthesis of a rigid phosphine-substituted, redox-active pincer ligand and its application to electrocatalytic CO<sup>2</sup> reduction with first-row transition metal complexes. The tridentate ligand was prepared by Stille coupling of 2,8-dibromoquinoline and 2-(tributylstannyl)pyridine, followed by a palladium-catalyzed cross-coupling with HPPh2. Complexes were synthesized from a variety of metal precursors and characterized by NMR, high-resolution mass spectrometry, elemental analysis, and cyclic voltammetry. Formation of bis-chelated metal complexes, rather than mono-chelated complexes, was favored in all synthetic conditions explored. The complexes were assessed for their ability to mediate electrocatalytic CO<sup>2</sup> reduction, where the cobalt complex was found to have the best activity for CO2-to-CO conversion in the presence of water as an added proton source.

Keywords: molecular catalyst, CO2 reduction, electrocatalysis, pincer ligand, first-row metals

# INTRODUCTION

Carbon dioxide (CO2) is a greenhouse gas which is produced in large quantities from fossil fuel combustion. From the Industrial Revolution forward, anthropogenic CO<sup>2</sup> emissions have increased at an alarming rate, along with associated concerns including climate change, rising sea levels, and ocean acidification (Hansen et al., 2008; Jiang and Guan, 2016; Blunden and Arndt, 2017). These environmental issues can be mitigated by effective technologies capable of converting CO<sup>2</sup> into sustainable fuels (Lim et al., 2014). In this context, electrocatalytic CO<sup>2</sup> reduction to generate carbon-neutral fuels and value-added commodity chemicals is a promising strategy that allows CO<sup>2</sup> to serve as a cheap and abundant C<sup>1</sup> feedstock.

However, reducing CO<sup>2</sup> is an energetically uphill process, and more efficient and selective catalysts are needed to facilitate its conversion. Many homogeneous metal complexes have been developed for CO<sup>2</sup> reduction, but these systems often rely on precious metals and/or are limited to privileged supporting ligands such as tetraazamacrocycles and bipyridines (Qiao et al., 2013; Francke et al., 2018). Tridentate "pincer" ligands represent an underexplored class of ligands, which are of particular interest given their strong chelate effect, preorganized geometry, high tunability, and potential for ligand-based redox activity (van der Vlugt and Reek, 2009; Younus et al., 2014; Peris and Crabtree, 2018). Indeed, Earth-abundant first-row transition metals supported by well-designed pincer ligands have been successfully employed in a variety of catalytic applications

*Edited by:*

*Hitoshi Ishida, Kitasato University, Japan*

### *Reviewed by:*

*Karthik Ramasamy, Los Alamos National Laboratory (DOE), United States Andrew Nattestad, University of Wollongong, Australia*

> *\*Correspondence: Jonah W. Jurss jwjurss@olemiss.edu*

### *Specialty section:*

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

*Received: 30 January 2019 Accepted: 24 April 2019 Published: 21 May 2019*

### *Citation:*

*Talukdar K, Issa A and Jurss JW (2019) Synthesis of a Redox-Active NNP-Type Pincer Ligand and Its Application to Electrocatalytic CO2 Reduction With First-Row Transition Metal Complexes Front. Chem. 7:330. doi: 10.3389/fchem.2019.00330*

**38**

(Benito-Garagorri and Kirchner, 2008; van der Vlugt and Reek, 2009; van der Vlugt, 2012; Chakraborty et al., 2015; Bauer and Hu, 2016). Yet, only a limited number of pincer complexes have been reported for electrocatalytic CO<sup>2</sup> reduction (Arana et al., 1992; Chiericato et al., 2000; Elgrishi et al., 2014, 2015; Sheng et al., 2015; Rao et al., 2016; Cope et al., 2017; Liu et al., 2018; Therrien et al., 2018; Myren et al., 2019). Early examples from Arana et al. (1992) and Chiericato et al. (2000) were reported with redox-active terpyridine and bis(imino)pyridinebased ligands (**Figure 1**). The metal ions decorated with these NNN-type ligands showed catalytic activity under reducing potentials in CO2-saturated solutions. Detailed product analysis was not reported, but formic acid was a notable reduced carbon product obtained with a cobalt bis(imino)pyridine complex. Recently, Liu et al. (2018) prepared a related pincer ligand, based in part on bis(imino)pyridine, in which steric bulk was added and one arm was replaced with pyridine to access a bipyridyl unit, producing a cobalt catalyst that is selective for CO<sup>2</sup> reduction to formate. Rao et al. (2016) reported a PNPligated manganese-tricarbonyl complex which can reduce CO<sup>2</sup> to CO without any added Brønsted acid. Mixed-donor Nheterocyclic carbene-based pincer ligands have also proven to be effective supporting frameworks for transition metal catalysts for CO<sup>2</sup> conversion (Sheng et al., 2015; Cope et al., 2017; Therrien et al., 2018; Myren et al., 2019).

Given these results, we sought to design a new redoxactive mixed-donor ligand with extended conjugation for CO<sup>2</sup> reduction employing first-row transition metals. The ligand design aims to exploit the well-established redox activity of bipyridine at a lower overpotential by extending its conjugation. In addition, the triphenylphosphine donor fragment serves as a good π-acceptor to facilitate metal-based reductions at lower potentials. We note that closely-related NNP ligands have been recently reported (Basu et al., 2018; Kamitani et al., 2018) and applied to the synthesis of new Fe and Co complexes for catalytic hydrosilylation and dehydrogenative silylation. Herein, firstrow metal complexes of a rigid NNP pincer ligand and their electrocatalytic activity toward CO<sup>2</sup> reduction are reported.

### EXPERIMENTAL

### Materials and Methods

All synthetic manipulations were carried out using standard Schlenk techniques or in an MBraun glovebox under nitrogen atmosphere. Freshly distilled acetonitrile (CH3CN) and N,N-dimethylformamide (DMF) were used in synthesis and electrochemistry. Tetrahydrofuran, toluene, and diethyl ether were dried with a Pure Process Technology solvent purification system. 1,1′ -Bis(diisopropylphosphino)-ferrocene (DiPPF), diphenylphosphine, iron(II) trifluoromethanesulfonate, and copper(II) perchlorate hexahydrate were purchased from Strem Chemicals. Zinc(II) trifluoromethanesulfonate and sodium tert-butoxide were purchased from Acros Organics. Nickel perchlorate hexahydrate and palladium(II) acetate were purchased from Alfa Aesar and Chem-Impex International, respectively. All other chemicals were reagent or ACS grade, purchased from commercial vendors, and used without further

purification. <sup>1</sup>H, <sup>31</sup>P, and <sup>13</sup>C NMR spectra were obtained using Bruker spectrometers operating at 500 MHz (1H), 167 MHz ( <sup>31</sup>P), or 126 MHz (13C). Spectra were calibrated vs. observed solvent peaks. Chemical shifts are reported in parts per million (ppm). Solution magnetic susceptibilities were determined by NMR using the Evans method. High-resolution electrospray ionization mass spectra (HR-ESI-MS) were obtained with a Waters SYNAPT HDMS Q-TOF mass spectrometer and elemental analyses of carbon, hydrogen, and nitrogen were conducted by Atlantic Microlab, Inc., Norcross, Georgia. Gas samples were analyzed by a custom Agilent 7890B Gas Chromatograph (Agilent PorapakQ column, 6 ft, 1/8 in. OD) with a dual detector system (TCD and FID). Calibration curves for the observed gases were prepared from commercial standards of known concentration obtained from buycalgas.com.

### Electrochemical Measurements

Cyclic voltammetry was performed with a typical three-electrode setup using a CH Instruments 600E Series potentiostat. The electrochemical cell was equipped with a glassy carbon disk working electrode (3 mm diameter), a platinum wire counter electrode, and a silver wire quasi-reference electrode. Acetonitrile or DMF solutions containing 0.1 M Bu4NPF<sup>6</sup> as the supporting electrolyte were used in all studies as specified. Ferrocene was added at the end of experiments and served as an internal standard to reference the potential. Electrolysis solutions and electrochemical cells were thoroughly degassed with nitrogen or carbon dioxide for 10–20 min prior to each experiment. Freshly made solutions were used for each experiment. Controlled potential electrolyses (CPE) were conducted in an airtight twocompartment cell with a glassy carbon rod working electrode (type 2, Alfa Aesar, 2 mm diameter), a silver wire quasi-reference electrode, and a high-surface area platinum mesh counter electrode. The platinum counter electrode was positioned inside a small isolation chamber with a fine frit. The isolation chamber contained the same solution, but without catalyst. The electrolysis solution in the working electrode compartment was continuously stirred during the experiments. Evolved gases were quantified by taking aliquots (0.30 mL) from the headspace using a sealable gastight syringe for injection into the gas chromatograph for analysis.

## X-Ray Crystallography

A single crystal coated with Paratone-N hydrocarbon oil was mounted on the tip of a MiTeGen micromount. The temperature was maintained at 200 K with an Oxford Cryostream 700 during data collection at the University of Mississippi, Department of Chemistry and Biochemistry, X-ray Crystallography Facility. Samples were irradiated with Mo-Kα radiation with λ = 0.71073 Å using a Bruker Smart APEX II diffractometer equipped with a fine-focus sealed tube source and APEX-II detector. The Bruker APEX2 v. 2009.1 software package was used to integrate raw data which were corrected for Lorentz and polarization effects (APEX2, 2009). A semi-empirical absorption correction (SADABS) was applied (Sheldrick, 2000). The space group was identified based on systematic absences, E-statistics, and successive refinement of the structure. The structure was solved using direct methods and refined by least-squares refinement on F2 and standard difference Fourier techniques using SHELXL (Sheldrick, 1990, 2008, 2014). Thermal parameters for all non-hydrogen atoms were refined anisotropically, and hydrogen atoms were included at ideal positions. Crystallographic data in CIF format was deposited into The Cambridge Crystallographic Data Center (CCDC); see deposition number CCDC 1909986 (**NiL2**<sup>+</sup> **2** ) and **Table S1**.

## Synthesis of the Ligand

Ligand precursor 8-bromo-2-(pyridin-2′ -yl)quinoline (**2**) was synthesized as previously reported (Mao et al., 2005; Wickramasinghe et al., 2015). New ligand (8- (diphenylphosphaneyl)-2-(pyridin-2′ -yl)quinoline) was prepared using a modified procedure (Murata and Buchwald, 2004; Zhang et al., 2017).

### 8-(diphenylphosphaneyl)-2-(pyridin-2′ -yl)quinoline, L

Inside the glovebox, to an oven-dried pressure flask with a Teflon screw cap was added 8-bromo-2-(pyridin-2′ -yl)quinoline, **2** (0.80 g, 2.81 mmol), Pd(OAc)<sup>2</sup> (0.032 g, 0.140 mmol), 1,1′ bis(diisopropylphosphino)ferrocene (0.071 g, 0.168 mmol), and NaOtBu (0.324 g, 3.37 mmol). Then 5 mL anhydrous toluene was added and the mixture was stirred for 1 h. Diphenylphosphine (0.41 mL, 2.34 mmol) was added to the flask via a micropipette and the flask was re-sealed and heated at 120◦C for 36 h. After cooling to room temperature, the reaction mixture was concentrated and purified by a deactivated silica column eluting with hexanes:ethyl acetate (5:1). The product was obtained as a light yellow solid, which is air stable in the solid phase (0.667 g, 73 %). <sup>1</sup>H NMR (CD3CN, 500 MHz): δ 8.64 (dq, J = 0.7 Hz, J = 4.8 Hz, 1H), 8.54 (d, J = 8.6 Hz, 1H), 8.40 (d, J = 8.6 Hz, 1H), 8.04 (d, J = 8.0 Hz, 1H), 7.95 (d, J = 8.2 Hz, 1H), 7.75 (td, J = 1.7 Hz, J = 7.8 Hz, 1H), 7.50 (td, J = 0.8 Hz, J = 8.0 Hz, 1H), 7.43–7.34 (m, 11H), 7.13 (qd, J = 1.3 Hz, J = 3.8 Hz, 1H). <sup>13</sup>C{1H} NMR (CDCl3, 126 MHz): δ 156.18 (s), 154.94 (s), 148.86 (s), 148.63 (d), 139.26 (d), 137.62 (d), 137.13 (s), 136.98 (s), 134.54 (s), 134.38 (s), 134.06 (s), 128.74 (s), 128.56 - 128.43 (m), 127.93 (s), 126.96 (s), 124.11 (s), 122.37 (s), 119.00 (s). <sup>31</sup>P{1H} NMR (CDCl3, 167 MHz) δ−13.35. HR-ESI-MS (M+) m/z calc. for [L+Cs+] +, 523.0340, Found, 523.0327.

### Synthesis of the Metal Complexes Common Synthetic Procedure

An acetonitrile or methanol solution (5 mL) of the NNP ligand **L** (0.050 g, 0.128 mmol) was prepared in a two-neck round bottom flask equipped with a reflux condenser before 0.5 equivalents of the corresponding metal precursor was added to the solution. The reaction mixture was refluxed for 6 h under nitrogen. After cooling to room temperature, the solvent was evaporated under reduced pressure and the residue was washed with Et2O.

### FeL2<sup>+</sup> 2 , [FeL2](OTf)<sup>2</sup>

The complex was prepared from Fe(OTf)<sup>2</sup> (0.023 g, 0.064 mmol) and 2 equivalents of **L** (0.050 g, 0.128 mmol) in acetonitrile. A short size-exclusion column (Sephadex **<sup>R</sup>** LH-20) was run, eluting with methanol, and a single dark band was collected and dried under vacuum. The complex was further purified by crystallization from a concentrated solution of methanol by slow diffusion of diethyl ether. Yield = 0.065 g (90%). <sup>1</sup>H NMR (CD3CN, 500 MHz): δ 9.04 (d, J = 8.7 Hz, 2H). 8.79 (d, J = 8.0 Hz, 2H), 8.62 (d, J = 8.7 Hz, 2H), 8.33 (br t, 2H), 8.25 (t, J = 7.6 Hz, 2H), 8.14 (d, J = 8.0 Hz, 2H), 7.65 (t, J = 7.65 Hz, 2H), 7.11–7.06 (m, 6H), 6.91 (t, J = 6.35 Hz, 2H), 6.80 (t, J = 7.35 Hz, 4H), 6.67 (t, J = 7.3 Hz, 4H), 6.31 (br t, 4H), 6.11 (br t, 4H). <sup>13</sup>C{1H} NMR (acetone-d6, 126 MHz): δ 161.36 (s), 157.83 (t), 157.55 (s), 125.27 (s), 141.26 (s), 139.20 (s), 138.62 (s), 134.92 (s), 131.79 (t), 131.49 (t), 130.97–130.89 (m), 130.56–130.48 (m), 129.77 (t), 129.51– 129.38 (m), 129.22 (s), 129.03 (s), 128.87 (s), 128.71 (s), 128.54 (s), 128.44 (s), 128.29 (s), 128.12 (d), 127.97 (s), 126.30 (s), 123.42 (s), 121.81 (s), 120.87 (s). <sup>31</sup>P{1H} NMR (CDCl3, 167 MHz) δ 58.94. HR-ESI-MS (M+) m/z calc. for [FeL2(OTf)]+, 985.1441, Found, 985.1454.

### [FeL2](Br)<sup>2</sup>

This complex was made in analogous fashion to [FeL2](OTf)<sup>2</sup> as detailed above using FeBr<sup>2</sup> (0.014 g, 0.064 mmol). Yield = 0.058 g (91%). <sup>1</sup>H NMR (DMSO-d6, 500 MHz): δ 9.29 (d, J = 8.8 Hz, 2 H), 9.03 (d, J = 8.85 Hz, 2H), 8.95 (d, J = 8.1 Hz, 2H), 8.54 (d, J = 8.05 Hz, 2H), 8.46 (br t, 2H), 8.32 (t, J = 7.6 Hz, 2H), 7.79 (t, J = 7.8 Hz, 2H), 7.26 (d, J = 5.55 Hz, 2H), 7.13 (t, J = 7.45 Hz, 2H), 7.08 (t, J = 7.35 Hz, 2H), 7.02 (t, J = 6.25 Hz, 2H), 6.85 (t, J = 7.55 Hz, 4H), 6.68 (t, J = 7.45 Hz, 4H), 6.27 (br t, 4H), 6.03 (br t, 4H). <sup>13</sup>C{1H} NMR (DMSO-d6, 126 MHz): δ 160.19 (s), 156.43 (t), 156.26 (s), 151.31 (s), 140.07 (s), 138.35 (s), 137.65 (s), 133.98 (s), 130.81 (s), 130.25 (t), 129.72–129.59 (m), 129.33 (s), 129.15 (t), 128.53 (t), 128.37 (t), 128.14 (s), 127.95 (s), 127.79 (s), 127.67– 127.59 (m), 126.77 (s), 126.61 (s), 126.46 (s), 125.49 (s), 121.16 (s). <sup>31</sup>P{1H} NMR (DMSO-d6, 167 MHz) δ 59.07. HR-ESI-MS (M+) m/z calc. for [FeL2Br]+, 915.1104, Found, 915.1077.

### CoL2<sup>+</sup> 2 , [CoL2](OTf)2]

The complex was prepared from Co(CH3CN)2(OTf)<sup>2</sup> (0.028 g, 0.064 mmol) and 2 equivalents of **L** in acetonitrile. A short size-exclusion column (Sephadex <sup>R</sup> LH-20) was run, eluting with methanol, and a single dark band was collected and dried under vacuum. Yield = 0.067 g (92%). HR-ESI-MS (M+) m/z calc. for [CoL2(OTf)]+, 988.1424, Found, 988.1389. The complex was fully characterized as a Co(III) complex, prepared by stirring a methanol solution of the Co(II) species overnight under air. Crystals of the Co(III) complex were grown from methanol by slow diffusion of diethyl ether. Elem. Anal. calc. for C52H38CoN4P2(CF3SO3)3·(H2O): C, 50.62; H, 3.09; N, 4.29. Found: C, 50.19; H, 3.05; N, 4.12. HR-ESI-MS (M+) m/z calc. for [CoL2(OTf)2] <sup>+</sup>, 1137.0944. Found, 1137.0947.

### NiL2<sup>+</sup> 2 , [NiL2](ClO4)<sup>2</sup>

The complex was synthesized by reacting Ni(ClO4)2·6H2O (0.023 g, 0.064 mmol) and **L** (0.050 g, 0.128 mmol) in methanol. The complex was purified by recrystallization from hot methanol. Yield = 0.057 g (86%). Elem. Anal. calc. for C52H38Cl2N4NiO8P2·(H2O)2: C, 58.13; H, 3.94; N, 5.21. Found: C, 58.33; H, 3.86; N, 5.20. HR-ESI-MS (M+) m/z calc. for [NiL2(ClO4)]+, 937.1410. Found, 937.1398.

### CuL2<sup>+</sup> 2 , [CuL2](ClO4)<sup>2</sup>

The complex was prepared from Cu(ClO4)2·6H2O (0.024 g, 0.064 mmol) and two equivalents of **L** in acetonitrile. The complex was found to be unstable in solutions and could not be fully purified. Yield = 0.061 g (92%). ESI-MS (M+) m/z calc. for [CuL2(ClO4)]+, 942.14. Found, 942.13.

### ZnL2<sup>+</sup> 2 , [ZnL2](OTf)<sup>2</sup>

Treatment of Zn(OTf)<sup>2</sup> (0.023 g, 0.064 mmol) with 2 equivalents of **L** in methanol gave the product as a light green powder. Yield = 0.065 g (89%). <sup>1</sup>H NMR (CD3CN, 500 MHz): δ 9.13 (d, J = 8.75 Hz, 2H), 8.80 (d, J = 8.75 Hz, 2H), 8.56 (t, J = 8.55 Hz, 4H), 8.01–7.94 (m, 6H), 7.44 (br d, 2H), 7.22 (t, J = 7.35 Hz, 2H), 7.13 (t, J = 7.15 Hz, 2H), 6.94–6.89 (m, 10H), 6.70 (br m, 4H), 6.38 (br d, 4H). <sup>13</sup>C{1H} NMR (acetone-d6, 126 MHz): δ 152.88 (s), 148.71 (s), 148.56 (s), 148.07 (t), 144.93 (s), 142.76 (s), 142.35 (s), 141.95 (s), 134.96 (t), 134.49 (s), 132.37 (s), 132.18 (t), 131.02 (t), 130.67 (d), 130.11 (t), 129.92 (t), 128.92 (s), 128.30 (t), 127.41 (t), 125.63 (s), 122.07 (s). <sup>31</sup>P{1H} NMR (CDCl3, 167 MHz) δ −34.69. HR-ESI-MS (M+) m/z calc. for [ZnL2(OTf)]+, 993.1383, Found, 993.1382.

# RESULTS AND DISCUSSION

# Synthesis of the Ligand and the Metal Complexes

Ligand precursor 2,8-dibromoquinoline **(1)** was prepared via a three-step synthetic route starting with 2-bromoaniline and cinnamoyl chloride (Wickramasinghe et al., 2015). A CombiFlash <sup>R</sup> Rf<sup>+</sup> system was utilized to obtain high purity product from the crude reaction mixture. As shown in **Figure 2**, Stille coupling of 2-(tributylstannyl)pyridine and **1** afforded 8 bromo-2-(pyridin-2′ -yl)quinoline **(2)** (Wickramasinghe et al.,

2015). The final ligand, 8-(diphenylphosphaneyl)-2-(pyridin-2′ yl)quinoline **(L)**, was prepared from **2** and diphenylphosphine using a modified cross-coupling reaction (Murata and Buchwald, 2004; Zhang et al., 2017). A deactivated silica gel column allowed purification of the product, in which 2-(pyridin-2′ -yl)quinoline was found to be the only major side-product.

The metal complexes were synthesized by refluxing the appropriate metal(II) salts with 2 equivalents of the ligand **L** in acetonitrile or methanol. Metal complexes in powder form were obtained after evaporation of the solvent and washing with diethyl ether. The Fe and Co complexes were initially purified by a Sephadex <sup>R</sup> column and obtained in high yields of ∼90%. Further purification of all complexes was done by recrystallization from concentrated methanol or acetonitrile solutions, generally at −20◦C or by slow diffusion of diethyl ether into the solution. The complexes are stable in the solid state to ambient light, air, and moisture.

In all cases, formation of the bis-chelated complexes was favored in the reaction conditions employed here. For reactions involving Co(II), Ni(II), Cu(II), and Zn(II), treatment of the appropriate metal precursor with 1 equivalent of **L** in polar solvents initially led to the formation of a mixture of mono- and bis-chelated metal complexes. Stirring the reaction mixture for prolonged periods led to clean formation of the bis-chelated complexes. In the case of Fe(II), the reaction of **L** and Fe(OTf)<sup>2</sup> in a 1:1 ratio immediately formed bischelated complex as the only product. A similar result was observed when FeBr<sup>2</sup> was used instead of Fe(OTf)2. This observation with iron complexes is analogous to the [(Ph-PNN)2Fe]2<sup>+</sup> complexes reported by Zell et al. (2013). In their studies, DFT calculations indicated that formation of the bischelated complexes is thermodynamically favored relative to the mono-chelated compounds in all conditions. They could tune this coordination behavior by altering the substituents of the phosphine donor where bulky tert-butyl groups allowed formation of the mono-chelated complex only and isopropyl substituents gave access to both the mono- and bis-chelated complexes depending on the initial ratio of metal precursor and ligand during metalation. Basu et al. (2018) followed the same principle to avoid bis-chelation in their report.

Synthesis of the mono-chelated Co(II) complex was attempted by treating a tetrahydrofuran solution of CoCl<sup>2</sup> with 1 equivalent of **L**. A green precipitate appeared immediately which was identified by mass spectroscopy to be mono-chelated **L**-CoCl<sup>2</sup> complex. However, due to limited solubility in traditional noncoordinating solvents, it could not be further characterized. Dissolving the complex in polar solvents such as acetonitrile, methanol, or DMF led to a gradual color change to a reddishbrown solution, indicating formation of the bis-chelated **CoL2+ 2** complex. This behavior is similar to the observation made by Harris et al. (1969) for Co(tpy) complexes.

Stirring 2:1 mixtures of the ligand and the appropriate metal precursors in methanol or acetonitrile for ∼12 h led to quantitative formation of the bis-chelated complexes, ML2<sup>+</sup> 2 . Elevated temperatures significantly reduced the reaction time. Each of the ML2<sup>+</sup> 2 complexes are stable in the solid state, and apart from the Cu(II) complex, they are also stable in solution. **CuL2+ 2** slowly decomposes in solution and could not be purified or fully characterized. We note that the cobalt(II) complex is easily oxidized to the cobalt(III) complex in solution when exposed to air.

### Characterization of Metal Complexes

Elemental analyses and mass spectrometry of each ML2<sup>+</sup> 2 complex show the 2:1 ratio of ligand to metal. The Co(II), Ni(II), and Cu(II) complexes are paramagnetic. As expected of a d<sup>10</sup> metal center, the Zn(II) complex is diamagnetic. The diamagnetic nature of the d<sup>6</sup> Fe(II) complex indicates a low-spin octahedral electronic state. Two distinct sets of <sup>1</sup>H signals were observed for the phenyl groups in these complexes, which suggests that no plane of symmetry passes through the metal center. Notably, only one peak is observed in the <sup>31</sup>P NMR spectra of **FeL2**<sup>+</sup> **2** and **ZnL2**<sup>+</sup> **2** signifying that the phosphorus atoms are chemically equivalent, consistent with κ 3 coordination modes for both NNP pincer ligands and octahedral complexes.

Indeed, a crystal structure of the octahedral nickel complex was obtained as shown in **Figure 3**. Single crystals of **NiL2**<sup>+</sup> **<sup>2</sup>** were grown from a concentrated solution of hot methanol that was allowed to cool to room temperature. The nickel-nitrogen bond distances range from 2.054(4) to 2.093(4) Å, with pyridine donors trans to the phosphine donor having an average Ni-N bond distance of 2.090 Å while those that are trans to another pyridine have an average Ni-N bond distance of 2.058 Å. Significantly longer bond distances are observed for the phosphine donors at 2.4323(13) for Ni-P(1) and 2.4183(13) for Ni-P(2) in the solid state.

For the three reported paramagnetic complexes, solution magnetic susceptibilities (µeff) were determined by the Evans method (Evans, 1959). At room temperature, the values are 2.6, 2.8, and 2.0 for **CoL2**<sup>+</sup> **2** , **NiL2**<sup>+</sup> **2** , and **CuL2**<sup>+</sup> **2** , respectively. The experimental values are close to previously reported values of similar octahedral complexes and characteristic of 1, 2, and 1 unpaired electron(s) for the Co(II), Ni(II), and Cu(II) complexes,

respectively. The relatively high µeff value for cobalt (2.6) is typical of cobalt polypyridyl complexes, and indicates significant orbital contribution (Chen et al., 2018).

### Electrochemistry

Cyclic voltammetry was performed on the series of metal complexes to assess and compare the redox potentials in acetonitrile/0.1 M Bu4NPF<sup>6</sup> solutions under inert atmosphere. The cyclic voltammograms are shown in **Figure 4** and E1/<sup>2</sup> values (peak potentials for the irreversible redox features) are summarized in **Table S2**. All potentials are reported against the ferrocenium/ferrocene couple (V vs. Fc+/<sup>0</sup> ).

The cyclic voltammogram (CV) of **ZnL2+ 2** shows four redox couples at −1.36, −1.54, −2.16, and −2.35 V. Since this complex has a redox-inactive metal center, these processes are assigned to ligand-based redox events. The zinc complex is helpful in identifying metal- and ligand-based redox processes in the remaining complexes. Notably, **ZnL2+ 2** is unstable under repeated scans. The redox couples at −2.16 and −2.35 V become irreversible and a visible change in peak intensities is observed presumably due to passivation of the working electrode.

The reduction waves at <sup>E</sup>p,<sup>c</sup> = −2.22 and <sup>−</sup>2.36 V of **CuL**2<sup>+</sup> 2 can be safely assigned to ligand-based reductions. A sharp irreversible oxidation peak at −0.76 V is also observed, indicative of adsorption on the electrode surface. The redox features of the copper complex also lose their reversibility and peak shape under repeated scans.

Reversible metal-based MIII/II redox couples occur at E1/<sup>2</sup> = 0.80, −0.46, and 0.66 V for **FeL2+ 2** , **CoL2+ 2** , and **NiL2+ 2** , respectively. In addition, three reversible redox events are

observed at negative potentials with the iron and cobalt complexes in the potential range studied here. The nickel complex shows slightly different behavior with two reversible reductions at −1.28 and −1.53 V, and a quasireversible reduction at −2.25 V. The quasireversible reduction has two small return oxidations at −2.15 and −1.83 V. The most negative features of these complexes are assigned to ligand-based reductions by comparison to the zinc analog.

# Electrochemical Reduction of CO<sup>2</sup>

Next, the complexes were analyzed by cyclic voltammetry to probe their reactivity toward CO<sup>2</sup> reduction. We focused on **FeL2**<sup>+</sup> **2** , **CoL2**<sup>+</sup> **2** , and **NiL2**<sup>+</sup> **2** as the copper and zinc complexes were not stable under electrochemical reduction. DMF solutions containing 0.1 M Bu4NPF<sup>6</sup> were employed for electrocatalytic studies for two reasons. First, using acetonitrile resulted in a significant amount of precipitation during controlled potential electrolysis (CPE) experiments. This is consistent with previously reported observations with 3d transition metal terpyridine complexes and ruthenium polypyridyl complexes. It was concluded that the build-up of HCO<sup>−</sup> 3 /CO2<sup>−</sup> 3 during bulk electrolyses resulted in precipitation of the complexes as HCO<sup>−</sup> 3 or CO2<sup>−</sup> 3 salts (Chen et al., 2011; Elgrishi et al., 2014). We confirmed the formation of carbonate in CPE experiments with **CoL2**<sup>+</sup> **2** by barium triflate titration (Yang et al., 2018), consistent with the reductive disproportionation of CO<sup>2</sup> to CO and CO2<sup>−</sup> 3 . The use of DMF solved this solubility issue as no precipitate was observed during electrolyses over extended periods of time. Second, using DMF gave us the opportunity to directly compare the performance of these complexes with the previously reported terpyridine systems (Elgrishi et al., 2014, 2015).

Under N<sup>2</sup> atmosphere, CVs of **FeL2**<sup>+</sup> **2** , **CoL2**<sup>+</sup> **2** , **NiL2**<sup>+</sup> **2** , and **ZnL2**<sup>+</sup> **2** in DMF (shown in **Figure S1**) exhibit similar behavior to



*Glassy carbon working electrode, platinum wire counter electrode, and silver wire quasireference electrode.*

that in acetonitrile, but with significantly lower stability for the nickel and zinc compounds. Redox potentials of the complexes in DMF are summarized in **Table 1**. Comparing the CV of the cobalt complex to its zinc counterpart, we assign the redox couples (E1/2) at −1.06 and −1.74 V to a metal-based CoII/<sup>I</sup> process and to a ligand-based process, respectively. The open circuit potential of the **CoL2**<sup>+</sup> **<sup>2</sup>** was measured at −0.61 V, which is consistent with these assignments. A reversible ligand-based wave was also observed at E1/<sup>2</sup> = −2.07 V. Adding free ligand to the solution increases the intensity of this peak (**Figure S2**), which further supports the assignment of this feature as a ligandbased reduction. Scanning past this ligand-based reduction gives rise to an irreversible ligand-based reduction at −2.65 V, which was found to be unstable under repeated scans (**Figure S3**).

For **FeL2**<sup>+</sup> **2** , reversible waves with E1/<sup>2</sup> values at −1.40, −1.64, and −2.27 V are attributed to ligand-based reductions by comparison with **ZnL2**<sup>+</sup> **2** . Different behavior was observed with **NiL2**<sup>+</sup> **2** relative to the Fe and Co complexes with one reversible redox feature at E1/<sup>2</sup> = −1.27 V, followed by one irreversible reduction at Ep,<sup>c</sup> = −1.51 V and two quasireversible reductions at Ep,<sup>c</sup> = −1.83 and −2.25 V. The open circuit potential was determined to be −0.68 V. Thus, the reversible redox process at −1.27 V is expected to be a metal-centered NiII/<sup>I</sup> couple, while the remaining waves are consistent with ligand-based redox events. Irreversible reductions are presumably due to dissociation of one of the tridentate ligands from the complex. This redox behavior is different from the related Co(tpy)2<sup>+</sup> 2 and Ni(tpy)2<sup>+</sup> 2 systems (Hamacher et al., 2009; Elgrishi et al., 2014). CVs of **FeL2**<sup>+</sup> **2** , **CoL2**<sup>+</sup> **2** , and **NiL2**<sup>+</sup> **2** at scan rates ranging from 10 to 1,000 mVs−<sup>1</sup> show that the peak currents change linearly with the square root of the scan rate (**Figures S5**–**S7**), consistent with diffusion-controlled homogeneous species. The diffusion coefficients were calculated from the slopes of the linear fits of the first reduction peaks using the Randles-Sevcik equation (Bard and Faulkner, 2001) and are presented in **Table 1**.

Water was added as a proton source to investigate the proton-coupled reduction of CO<sup>2</sup> with these complexes. The concentration of added water was kept fixed at 5% in these studies. No significant current enhancement was observed in the CVs when water was introduced to the systems under N<sup>2</sup> (**Figure 5**). Likewise, there was little change to the CV of **CoL2+ 2** with added H2O. The third (most negative) reduction peak of **FeL2**<sup>+</sup> **2** and **NiL2**<sup>+</sup> **2** becomes irreversible in the presence of H2O.

FIGURE 5 | Cyclic voltammograms of FeL2<sup>+</sup> 2 , CoL2<sup>+</sup> 2 , and NiL2<sup>+</sup> 2 at 0.5 mM concentrations in DMF/0.1 M Bu4NPF6 under N2 (black), CO2 (red), N2 with 5% H2O (green), and CO<sup>2</sup> with 5% H2O (blue). Scan rate = 100 mV/s; glassy carbon disk working electrode.

In CO2-saturated anhydrous solutions, the third reduction of all three complexes loses reversibility and shows current enhancement, suggesting catalytic reactivity toward CO<sup>2</sup> (**Figure 5**). A 2.4-fold current increase was observed at the third reduction peak of **FeL2**<sup>+</sup> **<sup>2</sup>** with the appearance of a new peak at Ep,<sup>c</sup> = −2.23 V. The new peak showed current enhancement of 1.9-fold. In **CoL2**<sup>+</sup> **2** , the most negative reduction peak was catalytically enhanced with an icat/i<sup>p</sup> value of 2.9 (where icat is the limiting catalytic current obtained under CO<sup>2</sup> and i<sup>p</sup> is the reductive peak current of the catalyst in the absence of substrate). CVs of **NiL2**<sup>+</sup> **2** showed current enhancement of 3.5- and 1.7-fold in the third and fourth reduction waves, respectively.

Additional current enhancement at the terminal reduction peaks of the **FeL2**<sup>+</sup> **2** and **CoL2**<sup>+</sup> **2** systems was observed when 5% water was added to CO2-saturated solutions, indicating that the complexes are capable of catalyzing the proton–coupled reduction of CO<sup>2</sup> (**Figure S10**). The CV of **NiL2**<sup>+</sup> **2** under the same conditions showed a decrease in the initial catalytic waves, but an additional and larger catalytic wave appeared at Ep,<sup>c</sup> = −2.40 V (**Figure 5**).

To better understand the reactivity of the complexes and to quantify the products, a series of controlled potential electrolyses (CPEs) were performed at different applied potentials with and without added H2O. Gaseous products were analyzed by periodic sampling of the headspace of the airtight electrochemical cell. **Table 2** summarizes the product distribution and Faradaic efficiencies (FE) of each complex over the first 2 h. Representative charge vs. time plots are provided in **Figures S11**–**S13**, **S16**. Carbon monoxide and hydrogen were the major products detected in all cases. A trace amount of methane was also found in the electrolyses of **CoL2**<sup>+</sup> **2** and **FeL2**<sup>+</sup> **2** but corresponds to <1% of the total charge passed.

The maximum overall FE was found to be 58% for **CoL2**<sup>+</sup> **2** at an applied potential of −2.11 V in the presence of 5% water (**Figure 6**). A maximum FE of 23% for evolved CO was recorded at Eappl = −2.06 V. The production of H<sup>2</sup> varied with the applied potential from 7 to 42%. When methanol—a more acidic proton source—was used, the overall FE decreased, but the ratio for H<sup>2</sup> production increased (**Table 2**). No evolved H<sup>2</sup> was observed when electrolyses were performed in anhydrous solutions. However, Faradaic efficiencies for CO were lower. Rinse tests were performed on the glassy carbon rod after CPEs to investigate possible adsorption or deposition of a heterogeneous catalyst on the electrode surface. There was no visible deposition and the glassy carbon rod did not exhibit any catalytic activity in fresh CO2-saturated DMF/0.1 M Bu4NPF<sup>6</sup> solution, which suggests catalysis is governed by a molecular cobalt species.

The turnover frequency (TOF) of **CoL2**<sup>+</sup> **<sup>2</sup>** was calculated as previously described (Narayanan et al., 2016; Yang et al., 2018) from scan rate-dependent linear sweep voltammograms performed in N2- and CO2-saturated solutions containing 5% H2O. A plot of the icat/i<sup>p</sup> current ratio at different scan rates as a function of potential is shown in **Figure S8**. Although a scan rateindependent TOF was not completely reached at scan rates up to 1 V/s, an estimated TOF of 12.4 s−<sup>1</sup> was determined (**Figure S9**).

CPEs of the **NiL2**<sup>+</sup> **2** complex show that the complex is selective for CO<sup>2</sup> reduction over proton reduction, albeit with very low FEs. The maximum FE observed with the nickel complex was



*For the CPE with ZnL2+ 2 , a 2 mM concentration of the complex was used. All values are the average of at least two runs.*

8%. The total charge passed was also very low compared to its iron and cobalt counterparts. The **FeL2**<sup>+</sup> **2** complex primarily catalyzed proton reduction in the presence of H2O. Indeed, the FE for CO production was ∼1% in anhydrous DMF solutions or with 5% added H2O. Under anhydrous conditions, proton reduction was suppressed, but CO production remained low. Rinse tests were also performed following electrolyses of both the nickel and iron complexes. For **FeL2**<sup>+</sup> **2** , the electrode did not show any sign of deposition. On the other hand, visible signs of deposition on the glassy carbon rod were observed with **NiL2**<sup>+</sup> **2** and the electrode showed catalytic activity in fresh solutions that signals decomposition of the complex occurs to form an active heterogeneous material.

### DISCUSSION

In general, to be an effective catalyst for CO<sup>2</sup> or H<sup>+</sup> reduction, a metal complex should be able to activate the substrate molecule via an open coordination site at the metal center. Thus, molecular catalysts are typically designed with at least one open or labile coordination site (Chiericato et al., 2000; Benson et al., 2009; Thoi et al., 2013; Su et al., 2018). In contrast, the complexes investigated here form an active catalyst in situ by ligand loss or dissociation of a donor moiety during electrolysis. Given the rigidity of the ligand, we hypothesize that one of the ligands completely dissociates. Thus, the bis-chelated **ML2**<sup>+</sup> **2** complex acts as a precatalyst, which under reducing conditions loses a ligand to generate a mono-chelated, catalytically active species (bearing a single pincer ligand) with available coordination sites for substrate activation and conversion. Our attempts to isolate the mono-chelated active species were unsuccessful. However, controlled potential electrolysis of **CoL2**<sup>+</sup> **<sup>2</sup>** was performed under inert atmosphere at an applied voltage of −2.65 V, and the electrolysis solution was analyzed by mass spectrometry, which showed evidence of free ligand that was not present in preelectrolysis solutions (**Figure S4**). In addition, CVs taken before and after electrolysis of **CoL2**<sup>+</sup> **2** show the formation of new redox features, indicating that a new electroactive species is formed under reducing conditions (**Figure S14**). Together, these experiments indicate that a mono-chelated complex is formed when the pre-catalyst is subjected to electrochemical reduction, and it is this species that is responsible for the catalytic activity observed in the presence of substrates (CO<sup>2</sup> and H+).

Given the structural similarities, **CoL2**<sup>+</sup> **2** likely operates by the mechanism proposed by Elgrishi et al. (2014) and Elgrishi et al. (2015) for Co(tpy)2<sup>+</sup> 2 complexes. The low Faradaic efficiencies are also consistent with their findings, which showed that the reduced mono-chelated active species can dimerize to give a deactivated species that does not reduce CO2. Moreover, the redox-active ligand liberated during creation of the active catalyst can be reduced further, contributing to the low overall Faradaic efficiencies (**Table 2**). Proposed mechanisms for CO<sup>2</sup> reduction as well as the competing hydrogen evolution reaction are shown in **Figure S18** for the cobalt(II) complex.

The **FeL2**<sup>+</sup> **2** and **NiL2**<sup>+</sup> **2** complexes behave differently than the Fe and Ni bis(terpyridine) complexes reported by Elgrishi et al. (2014). In their studies, Fe(tpy)2<sup>+</sup> 2 did not show any current enhancement in CO2-saturated solutions containing a proton source. Conversely, the Ni(tpy)2<sup>+</sup> 2 complex showed considerable catalytic activity under comparable conditions to those employed here and selectively produced CO in bulk electrolyses. The redox potentials of the bis(tpy) complexes and the present systems are also significantly different under <sup>N</sup><sup>2</sup> atmosphere. It was concluded from CVs of Fe(tpy)2<sup>+</sup> 2 , Ni(tpy)2<sup>+</sup> 2 , and Zn(tpy)2<sup>+</sup> 2 that they undergo ligand-based reductions exclusively in the potential window studied. The current enhancement observed in CVs of the complexes studied here under anhydrous conditions may be a result of CO<sup>2</sup> binding to the reduced active species to form a stable intermediate, which is not effective for catalysis and does not liberate a reduced carbon product. Related observations have been reported for molecular nickel complexes that have poor Faradaic efficiencies for CO<sup>2</sup> reduction despite encouraging results from cyclic voltammetry (Narayanan et al., 2016; Lieske et al., 2018). The irreversible reductions and dissimilarity between the anodic and cathodic scans of **NiL2**<sup>+</sup> **2** under N<sup>2</sup>

indicates formation of a new species. Consistent with the rinse test, ligand dissociation likely occurs from the nickel complex during electrolysis, which is followed by formation of a heterogeneous material on the electrode surface. Decomposition of the molecular species following reduction may account for the low Faradaic efficiencies.

**ZnL2**<sup>+</sup> **<sup>2</sup>** was electrochemically and spectroscopically analyzed to better understand the underlying mechanism and stability of the catalysts. The zinc complex did not show any notable current increase in CO2-saturated solutions by cyclic voltammetry (**Figure S15**). It also performed poorly in CPEs with an overall FE of <sup>&</sup>lt;1%. However, the diamagnetic nature of **ZnL2**<sup>+</sup> **<sup>2</sup>** was used to obtain insight into the decomposition pathway of the reported complexes. We conducted an electrochemical experiment as described by Elgrishi et al. (2014) and Fuchs et al. (1981), in which **ZnL2**<sup>+</sup> **<sup>2</sup>** was subjected to bulk electrolysis and the electrolyzed product was studied by <sup>1</sup>H NMR. NMRs were taken of the solution before and after electrolysis, and after adding iodomethane as an alkylating agent (**Figure S17**). Free ligand peaks were visible in the <sup>1</sup>H NMR spectrum of the electrolyzed solution after a 3-h CPE, consistent with ligand dissociation. Changes in the aromatic region of the NMR spectrum and appearance of new peaks in the aliphatic region after alkylation suggest possible dearomatization and carboxylation of the ligand (Fuchs et al., 1981; Elgrishi et al., 2014).

While our cobalt system suffers from low overall Faradaic efficiency and poor selectivity, the ligand structure is amenable to electronic and steric modifications. Elgrishi et al. (2015) were able to tune the selectivity of Co(tpy)2<sup>+</sup> 2 complexes by introducing different substituents on the tpy ligand. They improved selectivity toward CO production by disfavoring the hydrogen evolution reaction (HER), but overall Faradaic efficiencies remained low. The best derivative featuring bulky tert-butyl substituents on the tpy ligand led to the highest overall FE at 41%, presumably by reducing catalyst deactivation via dimerization. However, these results indicate that ligand modifications are largely ineffective for improving catalyst stability and thus they were not pursued with the NNP ligand investigated here.

An early example of electrochemical CO<sup>2</sup> reduction was reported by Arana et al. (1992) involving bis-chelated Fe, Co, and Ni complexes of redox-active NNN-type ligands. These complexes showed CO<sup>2</sup> reduction activity to various extents based on their current response in cyclic voltammetry studies. Among them, Co(dapa)2<sup>+</sup> 2 and Ni(dapa)2<sup>+</sup> 2 complexes (dapa = 2,6-bis-[1-(phenylimino)ethyl]pyridine) showed the best activity in CVs. Under inert atmosphere, both complexes showed one metal-based and two ligand-based reductions, similar to the complexes reported here. Catalytic current was observed at the second reduction for these systems, whereas catalysis begins in earnest at the third reduction with our system. However, the stability and substrate selectivity of the catalysts were not reported. Only a limited product analysis was done for the cobalt system, which had a ∼60% FE for formic acid production. This clearly suggests that this system follows a different catalytic mechanism than ours and the Fontecave systems where CO was the sole carbon-containing product.

As discussed previously, introducing steric bulk in pincer ligands can enforce mono-chelation, thereby avoiding the ligand dissociation step needed to access available coordination sites for substrate binding to the metal core. In addition, they can obstruct dimerization of reduced intermediates. Recently, electrocatalytic CO<sup>2</sup> reduction with cobalt and nickel complexes featuring modified dapa ligands were reported by Kang and Rochford, respectively. Mono-chelation was achieved by introducing diisopropyl groups in the ligand framework (Narayanan et al., 2016; Liu et al., 2018). The Co complex showed minimal reactivity toward CO<sup>2</sup> and was not studied further, whereas the Ni complex showed a good current response in CVs under CO2. However, during bulk electrolysis, the Ni complex produced only a small amount of CO with nearly quantitative formation of H2. Liu et al. (2018) hypothesized that catalysis could be improved by replacing one imino group with a pyridyl moiety. Their new NNN ligand has a redox-active bipyridine unit with extended π-conjugation and a sterically bulky imino group on one side. With this ligand, the corresponding cobalt complex shows metaland ligand-based redox events under argon. Adding water to the system gave rise to new waves corresponding to a Co-hydride species, which was found to mediate formate production in CO2 saturated solution. Consistent with DFT calculations, the reduced species in the presence of substrate can access two different pathways to generate both CO and formate, where formate is produced via a CoII-H intermediate. Despite similarities between this ligand and the NNP ligand reported here, our **CoL2**<sup>+</sup> **2** complex has different product selectivity and CO was found to be the only 2e<sup>−</sup> reduced C<sup>1</sup> product. This is likely due to a difference in hydricity of the CoII-H intermediate formed in our case, in which CO<sup>2</sup> insertion into the metal-hydride to make a formato intermediate is not competitive with protonation to form H<sup>2</sup> (Thoi et al., 2013; Kang et al., 2015).

Examples of phosphine-based pincer-ligated 3d metal complexes are still scarce among CO<sup>2</sup> reduction electrocatalysts. Recently, a manganese tri-carbonyl complex with a symmetric PNP-pincer ligand was reported by the Richeson group. This complex shows an unusual coordination environment with a mer-M(CO)<sup>3</sup> motif rather than the usual fac-M(CO)<sup>3</sup> configuration (Rao et al., 2016). Moreover, this complex also triggers CO<sup>2</sup> reduction at its first reduction. Addition of H2O as a weak proton source increased the current and lowered the catalytic overpotential, but also reduced substrate selectivity as HER activity was enhanced.

Pincer ligands with N-heterocylic carbene (NHC) donors have also been employed in CO<sup>2</sup> reduction. A nickel-based catalyst for CO<sup>2</sup> reduction supported by a mixed donor CNCtype ligand was reported by Sheng et al. (2015). This distorted square planner complex has a labile acetonitrile ligand at the fourth position and shows three irreversible reductions under N2. Among them, the first reduction at −1.19 V (vs Fc+/<sup>0</sup> ) was assigned to the NiII/<sup>I</sup> couple. Catalysis occurs at the most negative reduction in CO2-saturated CH3CN. Added water was found to enhance catalysis without compromising product selectivity. Myren et al. (2019) recently reported two manganese tricarbonyl complexes with similar CNC pincer ligands which can reduce CO<sup>2</sup> to CO at the first reduction potential. Though the overpotential is large, these catalysts also exhibit high selectivity in the presence of a proton source.

Cope et al. (2017) replaced the pyridyl group of the Sun catalyst with an aryl donor to prepare a Ni complex ligated by an anionic CCC-pincer ligand. This more electron-rich ligand on nickel results in faster catalysis (30x TOF), at the expense of a higher overpotential, and produces a mixture of CO and formate. Electrocatalysis occurs at the first reduction with this catalyst and added water enhances CO<sup>2</sup> reduction and the rate of catalysis. Bulk electrolysis experiments confirmed that the catalyst is selective, producing CO and formate with an optimal FE of 81% for reduced carbon products. However, catalysis slows down after ∼1 h of electrolysis for unspecified reasons.

## CONCLUSIONS

A rigid NNP-type pincer ligand and its mid-to-late firstrow transition metal complexes have been synthesized. The ML2<sup>+</sup> 2 complexes reported here are the first examples of an asymmetric phosphine-substituted pincer ligand employed for electrocatalytic CO<sup>2</sup> reduction. Despite the steric bulk of the phenyl substituents on this ligand, the metal ions prefer to coordinate with two ligands to form pseudo-octahedral bischelated complexes. Electrochemical studies confirm that the cobalt complex is capable of catalyzing the electrochemical reduction of CO<sup>2</sup> to CO. However, catalytic activity is greatly limited by deleterious side reactions. We reason that the rigid NNP framework inhibits ligand dissociation from the reduced

### REFERENCES


complex and, thus, impedes formation of the catalytically active species.

# AUTHOR CONTRIBUTIONS

KT synthesized and characterized the catalysts, collected and analyzed data, and wrote the original draft. AI synthesized and characterized the catalysts and collected data. JJ conceptualized the experiments, analyzed data, supervised research activities, acquired funding, and revised the manuscript.

### FUNDING

We thank the National Science Foundation for financial support through seed grant funding (OIA-1539035) and a CAREER Award (CHE-1848478). We are also grateful for start-up support from the University of Mississippi (UM).

### ACKNOWLEDGMENTS

We thank Prof. Jared Delcamp (UM) for allowing us to use his gas chromatograph and automated flash chromatography system. We also thank Senthil Kumar Eswaramoorthy and Prof. Amala Dass (UM) for their support with high-resolution mass spectrometry.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00330/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Talukdar, Issa and Jurss. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

**48**

# Improved Conditions for the Visible-Light Driven Hydrocarboxylation by Rh(I) and Photoredox Dual Catalysts Based on the Mechanistic Analyses

Kei Murata† , Nobutsugu Numasawa, Katsuya Shimomaki, Jun Takaya and Nobuharu Iwasawa\*

*Department of Chemistry, Tokyo Institute of Technology, Tokyo, Japan*

### Edited by:

*Soumyajit Roy, Indian Institute of Science Education and Research Kolkata, India*

### Reviewed by:

*Munetaka Akita, Tokyo Institute of Technology, Japan Yang Yang, University of Central Florida, United States*

### \*Correspondence:

*Nobuharu Iwasawa niwasawa@chem.titech.ac.jp*

### †Present Address:

*Kei Murata, Institute of Industrial Science, The University of Tokyo, Tokyo, Japan*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *07 March 2019* Accepted: *06 May 2019* Published: *22 May 2019*

### Citation:

*Murata K, Numasawa N, Shimomaki K, Takaya J and Iwasawa N (2019) Improved Conditions for the Visible-Light Driven Hydrocarboxylation by Rh(I) and Photoredox Dual Catalysts Based on the Mechanistic Analyses. Front. Chem. 7:371. doi: 10.3389/fchem.2019.00371* The improved catalytic conditions and detailed reaction mechanism of the visible-light driven hydrocarboxylation of alkenes with CO<sup>2</sup> by the Rh(I) and photoredox dual catalysts were investigated. The use of the benzimidazoline derivative, BI(OH)H, as a sacrificial electron donor was found to increase the yield of the hydrocarboxylated product by accelerating the reduction process. In addition, the incorporation of the cyclometalated Ir(III) complex as a second photosensitizer with [Ru(bpy)3] <sup>2</sup><sup>+</sup> photosensitizer also resulted in the promotion of the reduction process, supporting that the catalytic cycle includes two photochemical elementary processes: photoinduced electron and energy transfers.

Keywords: carboxylation, CO2 fixation, photoredox catalyst, rhodium catalyst, visible light

# INTRODUCTION

Catalytic hydrocarboxylation of unsaturated hydrocarbons with CO<sup>2</sup> is one of the promising methods for the CO<sup>2</sup> fixation (Luan and Ye, 2018; Yan et al., 2018; for recent reviews, see: Tortajada et al., 2018). The most common strategy to accomplish the hydrocarboxylation is to utilize a metal hydride complex as an active species. However, in these reactions, more than a stoichiometric amount of highly active, metallic reductants such as ZnEt2, AlEt3, or hydrosilanes are usually required to promote the reduction process in the catalytic cycle (Takaya and Iwasawa, 2008; Williams et al., 2008; Fujihara et al., 2011; Li et al., 2011; Hayashi et al., 2015; Wang et al., 2015; Zhu et al., 2015; Kawashima et al., 2016). In order to realize a more efficient and environmentallyfriendly system, the reaction which necessitates just a catalytic amount of metallic reagents is highly desirable. Meanwhile, the photochemical reduction process has been widely employed in the field of artificial photosynthesis, such as photocatalytic hydrogen generations (for review, see Esswein and Nocera, 2007) and CO<sup>2</sup> reductions (for reviews, see Morris et al., 2009; Doherty et al., 2010; Takeda and Ishitani, 2010) in homogeneous systems. In these reactions, transition-metal catalysts are combined with redox photosensitizers and sacrificial electron donors to drive the multielectron transfer processes under visible-light irradiation. When the electron transfer is accompanied by the

**Abbreviations:** bpy, 2,2′ -bipyridyl; dtbbpy, 4,4′ -di-tert-butyl-2,2′ -bipyridyl; bpz, 2,2′ -bipyrazine; ppy, 2-phenylpyridiine; dF(CF3)ppy, 3,5-difluoro-2-[5-(trifluoromethyl)-2-pyridinyl]phenyl; SCE, saturated calomel electrode; BIH, 1,3 dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole; BI(OH)H, 1,3-dimethyl-2-(o-hydroxyphenyl)-2,3-dihydro-1Hbenzo[d]imidazole; SED, sacrificial electron donor; PC, photoredox catalyst; SCE, saturated calomel electrode; vt, virtual coupling. For the abbreviations of the rhodium complexes, see **Supplementary Table 1**.

proton transfer, metal hydrides can act as an active species in the catalytic cycle (for reviews, see Stoll et al., 2015; Adams et al., 2018). Although transition-metal/photoredox dual catalysis has been actively studied in the field of organic synthesis (for recent reviews, see Fabry and Rueping, 2016; Skubi et al., 2016; Twilton et al., 2017), few examples have been reported for catalytic organic transformations driven by photochemically-generated metal hydrides (Ghosh et al., 2015).

We recently developed the visible-light driven hydrocarboxylation of alkenes with CO<sup>2</sup> for the first time by means of the photochemical generation of Rh(I) hydride species (Murata et al., 2017). 4-Cyanostyrene was transformed to the branched hydrocarboxylated product by using a Rh(I) hydride or chloride complex as a carboxylation catalyst, [Ru(bpy)3] 2+ as a photoredox catalyst, <sup>i</sup>Pr2NEt as a sacrificial electron donor, with visible-light irradiation under CO<sup>2</sup> atmosphere at room temperature (**Figure 1**). The photoredox catalysis made it possible to take electrons from tertiary amines and drive the reduction process without using a metallic reductant. Since then, several photoredox-catalyzed hydrocarboxylation reactions of unsaturated hydrocarbons with CO<sup>2</sup> have been reported by other groups (Seo et al., 2017b; Hou et al., 2018b; Meng et al., 2018). Concomitantly, difunctionalizations of alkenes such as thiocarboxylation (Ye et al., 2017), carbocarboxylation and silylcarboxylation (Yatham et al., 2017; Hou et al., 2018a) have also been developed by incorporating an appropriate radical precursor with CO2. Furthermore, in addition to unsaturated hydrocarbons, various substrates such as aryl and alkyl halides (Meng et al., 2017; Shimomaki et al., 2017), amines (Seo et al., 2017a), imines and enamides (Fan et al., 2018; Ju et al., 2018) have been carboxylated with CO<sup>2</sup> by photoredox catalysis so far (for review, see Yeung, 2019). These examples demonstrated wide applicability of the photoinduced electron transfer to carboxylation reactions.

On the basis of our previous experiments, the reaction mechanism of the hydrocarboxylation by Rh(I) and photoredox dual catalysts was proposed as shown in **Figure 2**. Initially, the hydrometallation of a styrene derivative by Rh(I) hydride species **A** gave the Rh(I) benzyl species **B** (i), and the visible-light promoted nucleophilic addition to CO<sup>2</sup> afforded the Rh(I) carboxylate species **C** (ii). Then, the reductive quenching cycle of [Ru(bpy)3] <sup>2</sup><sup>+</sup> with <sup>i</sup>Pr2NEt mediated 2-electron, 2 proton transfers afforded the Rh(III) dihydride carboxylate species **D** (iii), followed by the base-promoted liberation of the carboxylated product to regenerate the active species **A** (iv). Although this reaction demonstrated fundamental aspects of the application of photochemical reduction processes to catalytic carboxylation reactions, there still has been room for improvement from the viewpoint of the applicability in organic synthesis: (i) The efficiency of the reaction was not very high. Good yield was obtained with 4-cyanostyrene and moderate yields were obtained with several other substrates. (ii) A large excess amount of a tertiary amine and long reaction time (>24 h) were necessary for completion of the reaction even for the reactive substrates, (iii) A significant amount of the hydrogenated product was produced as a byproduct. In order to resolve these problems, further screenings of the catalytic conditions were desired.

Herein, we explored the improved conditions of the visible-light driven hydrocarboxylation, and the catalytic efficiency was analyzed based on the detailed mechanistic study with a series of stoichiometric reactions of the rhodium intermediates. Through the investigation, the hydrocarboxylation was successfully improved by the alteration of the sacrificial electron donor or the incorporation of the second photosensitizer. The mechanistic study suggested that the promotion of the photochemical reduction process was crucial for the enhancement of the catalytic reaction.

# RESULTS AND DISCUSSION

### Screening of Reaction Conditions

On the basis of our previous experiments in terms of the screening of catalytic conditions and the observation of the reaction intermediates under the catalytic conditions, the followings were demonstrated: (i) As a carboxylation catalyst, Rh(I) hydride or chloro complexes with triarylphosphines were applicable. In particular, the µ-chloro bridged Rh(I) dimer [Rh(P(4-CF3C6H4)3)2Cl]<sup>2</sup> (**4**) was the most effective catalyst. (ii) When **4** was employed under the catalytic conditions, the resting state was the corresponding Rh(I) carboxylate complex, indicating that the rate-determining step was its transformation to the Rh(I) hydride species. This result suggested that the promotion of the reduction process was crucial for the improvement of the catalytic reaction. According to these considerations, the reaction conditions were screened in terms of the photoredox catalyst and sacrificial electron donor,

which would have taken an important part in the reductive quenching cycle.

### Photoredox Catalyst

Photoredox catalysts were initially screened by performing the reaction of 4-cyanostyrene (**1a**) in a mixture of 2.0 mol% of a photosensitizer, 3.5 mol% of **4** and 4.0 equiv. of <sup>i</sup>Pr2NEt under a CO<sup>2</sup> atmosphere at room temperature (**Table 1**). In the case of [Ru(bpy)3](PF6)<sup>2</sup> (E II∗ /I <sup>1</sup>/<sup>2</sup> = +0.77 V, E II/I <sup>1</sup>/<sup>2</sup> = −1.33 V vs. SCE) (Kalyanasundaram, 1982) as a photoredox catalyst, the hydrocarboxylated (**2a**) and hydrogenated (**3a**) products were obtained in 54 and 25% yields, respectively, after visiblelight irradiation for 24 h (**Table 1**, entry 1). A small amount of polymerized product of **1a** was also produced as byproduct. Though no other photosensitizers overcame this activity, the yield of **2a** was found to be strongly dependent on the photoredox catalyst. For instance, when [Ru(bpz)3](PF6)<sup>2</sup> (E II∗ /I <sup>1</sup>/<sup>2</sup> = +1.45 V, E II/I <sup>1</sup>/<sup>2</sup> = −0.80 V vs. SCE) (Crutchley and Lever, 1980) or fac-Ir(ppy)<sup>3</sup> (E III<sup>∗</sup> /II <sup>1</sup>/<sup>2</sup> = +0.31 V, E III/II <sup>1</sup>/<sup>2</sup> = −2.19 V vs. SCE) (Flamigni et al., 2007) was employed, the yields substantially decreased compared to [Ru(bpy)3](PF6)<sup>2</sup> (**Table 1**, entries 3, 6). On the other hand, when using [Ir(dF(CF3)ppy)2(dtbbpy)](PF6) (E III<sup>∗</sup> /II 1/2 = +1.21 V, E III/II <sup>1</sup>/<sup>2</sup> = −1.37 V vs. SCE) or [Ir(ppy)2(dtbbpy)](PF6) (E III<sup>∗</sup> /II <sup>1</sup>/<sup>2</sup> = +0.66 V, E III/II <sup>1</sup>/<sup>2</sup> = −1.51 V vs. SCE) (Lowry et al., 2005), moderate yields were obtained (**Table 1**, entry 4, 5). These results indicated that both sufficient oxidizing ability of the excited state and reducing ability of the one-electron reduced species are at least necessary for the photosensitizer. However, the detailed dependency was not simple, as other factors such as absorption properties, excited-state energies and photochemical stability of the photosensitizer could also affect the catalytic performances. Meanwhile, the screenings of additives with [Ru(bpy)3] <sup>2</sup><sup>+</sup> photosensitizer demonstrated that the addition of Cs2CO<sup>3</sup> as an inorganic base significantly improved the yield of the hydrocarboxylated product by suppressing the formation of the hydrogenated byproduct: the yield of **2a** increased to 67% while the yield of **3a** decreased to 1% (**Table 1**, entry 11).

### Sacrificial Electron Donor and Additives

Sacrificial electron donors were then screened in the presence of an excess amount of Cs2CO3. In order to highlight the reactivity, a less reactive alkene, 3,5-bis(trifluoromethyl)styrene (**1b**), was used as a substrate. The reactions of **1b** were performed in a mixture of 2.0 mol% of [Ru(bpy)3](PF6)2, 3.5 mol% of **4**, 4.0 equiv. of sacrificial electron donor and 1.2 equiv. of Cs2CO<sup>3</sup> under a CO<sup>2</sup> atmosphere at room temperature (**Table 2**). When <sup>i</sup>Pr2NEt was employed as a sacrificial electron donor, 32% yield of the hydrocarboxylated product (**2b**) and a trace amount of the hydrogenated product (**3b**) were obtained after visiblelight irradiation for 12 h (**Table 2**, entry 1). Although the use of TEOA (triethanolamine) slightly increased the yield of **2b**, the formation of **3b** became pronounced probably due to the increase of proton concentration (**Table 2**, entry 3). On the other hand, the use of BI(OH)H (1,3-dimethyl-2-(o-hydroxyphenyl)- 2,3-dihydro-1H-benzo[d]imidazole) successfully accelerated the hydrocarboxylation and increased the yield of **2b** considerably with maintaining the low yield of **3b** (**Table 2**, entry 5). Furthermore, the incorporation of BI(OH)H made it possible to reduce the amounts of the photoredox catalyst and the sacrificial electron donor: the use of only 1.0 mol% of [Ru(bpy)3](PF6)<sup>2</sup> and 2.0 equiv. of BI(OH)H gave 70% yield of **2b** (**Table 2**, entry 7). When the reaction was performed under an Ar atmosphere in the presence of Cs2CO3, no hydrocarboxylated product **2b** was obtained. This result confirmed that the carbonate did not work as a source of CO<sup>2</sup> in the present reaction (**Table 2**, entry 9). BI(OH)H has been known to work as a 2-electron, 2-proton

TABLE 1 | Optimization of the reaction conditions of the hydrocarboxylation of 4-cyanostyrene.


*<sup>a</sup>NMR yield.*

*<sup>c</sup>1 mol% of [Ru(bpy)3](PF6)2.*

*<sup>d</sup>5 mol% of [Ru(bpy)3](PF6)2.*

donor with high reducing ability in redox photosensitizing reactions (Hasegawa et al., 2005, 2006; Tamaki et al., 2015). Since a tertiary amine contributed to the reductive quenching cycle of [Ru(bpy)3] <sup>2</sup>+, the increase in the yield of **2b** was attributed to the promotion of the reduction process of the Rh(I) carboxylate species, which was the rate-determining step in the hydrocarboxylation. These results indicated that the redox property of a sacrificial electron donor is one of the crucial factors for the efficient promotion of the reaction.

### Generality of the Hydrocarboxylation Under the Improved Conditions

Based on the improved conditions using **1b** as discussed above, the generality of the hydrocarboxylation was examined using various alkene substrates. 1.0 mol% of [Ru(bpy)3](PF6)2, 3.5 mol% of **4**, 2.0 equiv. of BI(OH)H and 1.2 equiv. of Cs2CO<sup>3</sup> were employed for the hydrocarboxylation (**Table 3**). In the cases of using styrenes with an electron-withdrawing group such as **1a** and 4-methoxycarbonyl styrene (**1d**), the reaction was almost completed after irradiation for 12 h, and the yields of the corresponding hydrocarboxylated products were significantly improved compared with those obtained in the previous conditions where 2.0 mol% of [Ru(bpy)3](PF6)<sup>2</sup> and 4.0 equiv. of <sup>i</sup>Pr2NEt were employed. Moreover, 4-trifluoromethyl styrene (**1c**) and non-substituted styrene (**1e**), which exhibited quite low reactivities in the previous conditions, did react to afford significant amounts TABLE 2 | Optimization of the reaction conditions of the hydrocarboxylation of 3,5-bis(trifluoromethyl)styrene.


*<sup>a</sup>NMR yield.*

*<sup>b</sup>GC yield. <sup>c</sup>1 mol% of [Ru(bpy)3](PF6)<sup>2</sup> was used. <sup>d</sup>Without CO2, under an Ar atmosphere.*

*e Isolated yield was 63%.*

of the corresponding hydrocarboxylated products though the yields were still not sufficiently high. The yields of the hydrocarboxylated products were also improved in the case of alkyl acrylates (**1f** and **1g**). Consequently, the introduction of BI(OH)H electron donor with Cs2CO<sup>3</sup> base successfully resulted in the increase in the yields of the present hydrocarboxylation reaction.

### Mechanistic Study

In order to reveal the reaction mechanism of the photocatalytic hydrocarboxylation, the stoichiometric reactions of the possible rhodium intermediates, which corresponded to each elementary step in the proposed catalytic cycle, were examined.

### Rh Hydride Formation

Initially, the Rh(I) hydride formation step was investigated using Rh(PPh3)2(OAc) (**5**) as a model complex of the Rh(I) carboxylate intermediate. The DMA solution of **5** was irradiated by visible-light in the presence of a catalytic amount of [Ru(bpy)3](PF6)2, an excess amount of <sup>i</sup>Pr2NEt and 1.5 equivalent of PPh3. After visible-light irradiation for 6 h, the Rh(I) monohydride complex, Rh(PPh)3H (**6**) was successfully obtained in 57% yield with the recovery of ca. 30% of **5**

*<sup>b</sup>GC yield.*

*<sup>a</sup>NMR yield.*

*b Isolated yield after methyl esterification: 57% (*2a*), 59% (*2d*).*

*<sup>c</sup>With 2 mol% of [Ru(bpy)*3*](PF*6*)*2.

(based on <sup>1</sup>H NMR using an internal standard) (**Figure 3A**; **Supplementary Figure 1**). Control experiments demonstrated that [Ru(bpy)3](PF6)2, <sup>i</sup>Pr2NEt, and visible-light were all essential for the transformation. Since hydrogen evolution was not evident during the reaction, the contribution of gaseous hydrogen was excluded. Thus, this transformation was considered to proceed via (i) stepwise 2-electron, 2-proton transfers from the tertiary amine by the photoredox catalysis to give the Rh(III) dihydride carboxylate (**7**), and (ii) the baseassisted elimination of the carboxylic acid to give **6**. In terms of step (i), the similar mechanisms have been proposed in photocatalytic hydrogen generation systems by a Rh(I) catalyst (Stoll et al., 2015). The initial single electron transfer to the protonated form of **5** would give the Rh(II) carboxylate monohydride, and the following electron and proton transfers or disproportionation of the two Rh(II) hydride species would give **7** (**Figure 4**). The presence of the Rh(III) dihydride intermediate was also supported by the fact that Rh(PCy3)2(OAc) (**5** ′ ) was transformed to Rh(PCy3)2(OAc)(H)<sup>2</sup> (**7** ′ ) almost quantitatively under the similar conditions although the reaction was relatively slow (**Figure 3B**; **Supplementary Figure 2**). In this case, PCy<sup>3</sup> ligands with strong σ-donation were considered to stabilize the Rh(III) dihydride intermediate to inhibit the following elimination reaction.

In order to confirm the carboxylic acid elimination step (ii), the reactivity of **7** was investigated in the presence of base. **7** was alternatively synthesized by the hydrogenation of **5** with H2, and was treated with an excess amount of <sup>i</sup>Pr2NEt in the presence of PPh<sup>3</sup> in C6D6. The reaction readily gave a mixture of **7** and **6** with liberation of [iPr2NHEt]+[CH3COO]−. Furthermore, addition of a small excess amount of [iPr2NHEt]+[CH3COO]<sup>−</sup> to the C6D<sup>6</sup> solution of **6** resulted in the quantitative formation of **7**. These results demonstrated that **7** was in equilibrium with **6** in the presence of <sup>i</sup>Pr2NEt and PPh<sup>3</sup> (**Figure 5**). When the treatment of **7** with <sup>i</sup>Pr2NEt was similarly conducted in DMA, **6** was detected as a sole rhodium species in the reaction mixture, indicating that the equilibrium was almost completely shifted to the product side owing to the solvent effect of DMA.

The possible mechanisms for generation of the hydrogenated product were (i) 2-electron, 2-proton transfers to the Rh(I) benzyl intermediate by the photoredox catalysis to give Rh(III) benzyl dihydride intermediate, which would undergo reductive elimination of the hydrogenated product, and (ii) the alkene insertion to the Rh(III) dihydride intermediate and the successive reductive elimination. Both pathways could be inhibited by lowering proton concentrations, as proton transfers would become inefficient in the former, and the competing carboxylic acid elimination from the dihydride complex would be promoted in the latter. Therefore, the inhibition of the hydrogenated product formation by the addition of Cs2CO<sup>3</sup> was attributed to the decrease of the proton concentration in the catalytic system.

The photochemical formation of Rh(I) monohydride species was also feasible by using Rh(I) chloride complex as a Rh(I) source. It was demonstrated by the fact that Wilkinson's type complex Rh(PPh3)3Cl was converted to Rh(PPh)3H (**6**) by visible-light irradiation in the presence of a catalytic amount of [Ru(bpy)3](PF6)<sup>2</sup> and an excess amount of <sup>i</sup>Pr2NEt. Therefore, the Rh(I) chloride complex was confirmed to work as a precursor of the Rh(I) hydride active species.

### Hydrometalation and Carboxylation

Since the Rh(I) monohydride species was successfully generated from the Rh(I) carboxylate species by photoredox catalysis, the hydrometallation and subsequent carboxylation processes were then investigated to complete the catalytic cycle. Treatment of Rh(PPh)3H (**6**) with an excess amount of **1a** at room temperature readily formed the Rh(I) benzyl species, Rh(PPh3)2(η 3 -CHCH3(4-CNC6H4)) (**8**), almost quantitatively with the liberation of a PPh<sup>3</sup> ligand (**Figure 6**i). The benzyl ligand in **8** was found to possess η 3 -coordination to the Rh(I) center based on NMR spectroscopic data (Werner et al., 1994). However, the attempt for isolation of **8** was not successful due to the presence of an equilibrium with **6**. Therefore, in situ generated **8** was directly used for the carboxylation step.

To investigate the carboxylation process with CO2, a DMA solution of a 1:1 mixture of in-situ generated **8** and PPh<sup>3</sup> was exposed to the atmospheric pressure of CO<sup>2</sup> under various conditions. The carboxylation did not proceed under dark even by heating, which was against our expectations based on the general reactivity of organorhodium(I) complexes with CO<sup>2</sup> (Ukai et al., 2006; Mizuno et al., 2011; Suga et al., 2014; Kawashima et al., 2016). Quite interestingly, when the mixture was irradiated by visible-light for 30 min in the presence of 30 mol% of [Ru(bpy)3](PF6)2, **8** was successfully converted to

*<sup>d</sup>Yield obtained by the previous conditions with [Ru(bpy)*3*](PF*6*)*<sup>2</sup> *(2 mol%),* 4 *(3.5 mol%). <sup>i</sup>Pr*2*NEt (4 equiv.) and irradiation for 24 h.*

the Rh(I) carboxylate complex, Rh(PPh3)3(η 1 -O2CCHCH3(4- CNC6H4)) (**9**), almost quantitatively (**Figure 6**ii). <sup>31</sup>P{1H} NMR spectroscopy confirmed the clean formation of **9**: a pair of the doublet of doublet signals attributed to **8** completely disappeared with the PPh<sup>3</sup> signal, and the doublet of doublet and doublet of triplet signals attributed to **9** appeared in 2: 1 ratio by visible-light irradiation (**Figure 7**). The control experiments demonstrated that CO2, [Ru(bpy)3](PF6)<sup>2</sup> and visible-light were all essential for the carboxylation, suggesting that the nucleophilic addition of **8** to CO<sup>2</sup> was facilitated by the photosensitization of [Ru(bpy)3] 2+. The luminescence quenching experiment demonstrated that the excited state of [Ru(bpy)3] <sup>2</sup><sup>+</sup> was effectively quenched by **8** (**Figure 8**). The quenching constant was determined to be K<sup>q</sup> = 2.07 × 10<sup>3</sup> , which was much larger than that by <sup>i</sup>Pr2NEt (K<sup>q</sup> =

1.56 × 10<sup>2</sup> , **Supplementary Figure 3A**). This result indicates that either photoinduced electron transfer or triplet-triplet energy transfer to **8** contributed to the quenching (Campagna et al., 2007; Arias-Rotondo and McCusker, 2016; Strieth-Kalthoff et al., 2018). However, the photoinduced electron transfer mechanism was unlikely in this case since (i) the carboxylation of **8** proceeded with a catalytic amount of [Ru(bpy)3](PF6)<sup>2</sup> even in the absence of the sacrificial electron donor, and (ii) the cyclic voltammogram of **8** showed no significant redox peak within the window where the oxidative quenching of [Ru(bpy)3] <sup>2</sup><sup>+</sup> was possible. Therefore, the photoinduced triplet-triplet energy transfer from the excited [Ru(bpy)3] <sup>2</sup><sup>+</sup> to **8** was considered to be the most likely process in this carboxylation process. Although not very common, several examples on the photocatalytic organic transformations mediated by the triplet-triplet energy transfer were previously reported (Ikezawa et al., 1986; Osawa et al., 2001; Islangulov and Castellano, 2006; Lu and Yoon, 2012; Farney and Yoon, 2014).

In order to investigate the detailed effect of the photoinduced energy transfer, the electronic structure analyses were performed in terms of the ground (S0) and the lowest excited triplet (T1) states of **8** based on DFT/TD-DFT methods. The calculated energy level of the T<sup>1</sup> state of **8** (1.07 eV, based on the comparison between the S<sup>0</sup> and T<sup>1</sup> optimized geometries) was much lower than that of [Ru(bpy)3] <sup>2</sup><sup>+</sup> (2.17 eV), indicating that the triplet-triplet energy transfer from the excited [Ru(bpy)3] 2+ to **8** was feasible. In terms of the optimized structures, a notable difference was found on the coordination manner of the benzyl ligand between the S<sup>0</sup> and T<sup>1</sup> geometries. In the S<sup>0</sup> optimized structure, the η 3 -coodination of the benzyl ligand was represented by the similar three Rh-C distances, which coincided with the results of the <sup>1</sup>H NMR observation (**Figure 9A**; **Table 4**). On the other hand, in the T<sup>1</sup> optimized structure, while the Rh-C1 (benzyl carbon) distance remained unchanged, the Rh-C2/C3 distances significantly elongated compared with those of the S<sup>0</sup> structure. These results indicated that the benzyl ligand changed its coordination-mode from η 3 -type (αbenzyl) to η 1 -type (σ-benzyl) in the T<sup>1</sup> state. According to the analysis on the electronic transition characters, the T<sup>1</sup> state was mainly contributed by the transitions of HOMO→LUMO (88%) and HOMO→LUMO+16 (4%) (**Supplementary Table 2**). The molecular orbital distribution indicated that LUMO and LUMO+16 mainly localized on the Rh (dπ) and benzyl ligand (π ∗ ) while the HOMO localized on the Rh (dσ) center (**Figure 9B**). As these LUMOs partially possessed the antibonding character on the Rh-C2/C3 bonds, the photoexcitation induced the dissociation of these Rh-C bonds, which resulted in the isomerization to the σ-benzyl species.

Concerning the acceleration of the carboxylation step, one possibility is the generation of the coordination site by taking σ-benzyl structure in the T<sup>1</sup> state, which would promote the following carboxylation by facilitating coordination of CO<sup>2</sup> to Rh center. Indeed, a similar thermal process has been proposed as a plausible mechanism for the carboxylation of organorhodium(I)

TABLE 4 | Selected bond lengths and NBO natural charges in the optimized geometries of 8(S0) and 8(T1).


complexes (Darensbourg et al., 1987). Another possibility is the direct nucleophilic addition of the benzyl carbon to CO<sup>2</sup> in the T<sup>1</sup> state. The NBO analysis demonstrated that the natural charge on the C1 atom significantly shifted to the negative side while that on the Rh atom shifted to the positive side in the T<sup>1</sup> state. The increase of the electron density on the C1 atom in the T<sup>1</sup> state would result in the acceleration of the nucleophilic addition to CO2. Therefore, the structural and/or electronic factors associated with the transition to the T<sup>1</sup> state are thought to contribute to the carboxylation of **8**.

The effect of the photoactivation of the Rh(I) π-benzyl complex was also supported by the reactivity of the Rh(I) σ-alkyl complex with CO2. When a mixture of **6** and an excess amount of methyl acrylate (**1f**) was subjected to a CO<sup>2</sup> atmosphere for 3 h even under dark, the quantitative formation of Rh(PPh3)3(η 1 - O2CCHCH3(CO2CH3)) (**10**) was indicated by <sup>31</sup>P{1H} NMR spectroscopy. The carboxylation of **1f** was confirmed by the fact that the corresponding hydrocarboxylated product (**2f**) was obtained from the reaction mixture. This result indicated that the photosensitization by [Ru(bpy)3] <sup>2</sup><sup>+</sup> was not essential in this case. Thus, the major role of the excitation was thought to be the transformation from π-benzyl to σ-benzyl complexes to generate a coordination site and to make them more nucleophilic.

### Addition of the Second Photosensitizer

The above mechanistic study revealed that a photosensitizer played two key roles in the hydrocarboxylation cycle: one is a "photoredox catalyst" to reduce the Rh(I) carboxylate species, and the other is a "triplet photosensitizer" to promote carboxylation of the Rh(I) benzyl species. With a single photosensitizer, the excited state of the photosensitizer was quenched by either a tertiary amine for the electron transfer or a Rh(I) benzyl species for the energy transfer, and these two processes competed during the reaction. Since the former was related to the rate-determining step when using <sup>i</sup>Pr2NEt as a sacrificial electron donor, the incorporation of the second photosensitizer possessing suitable redox properties for the reductive quenching cycle was expected to facilitate the catalytic reaction.

On the basis of the idea, 2.0 mol% of a cyclometalated Ir(III) complex was added as a second photosensitizer to a mixture of **1b**, 3.5 mol% of **4**, 2.0 mol% of [Ru(bpy)3](PF6)<sup>2</sup> and 4.0 equiv. of <sup>i</sup>Pr2NEt, and the solution was irradiated under CO<sup>2</sup> atmosphere at room temperature (**Table 5**). To excite both photosensitizers, a wide range of UV-visible-light (380–750 nm) was applied to the reactions. As expected, the addition of the second photosensitizer was found to be effective. For instance, when [Ir(ppy)2(dtbbpy)](PF6) was added, the reaction was completed after irradiation for only 6 h, and the yield of **2b** was increased more than five-fold compared to that of the reaction without the second photosensitizer (**Table 5**, entry 2). According to the redox properties of [Ir(ppy)2(dtbbpy)](PF6), the acceleration of the reaction was thought to be attributed mainly to the


TABLE 5 | Screening of the reaction conditions of the hydrocarboxylation of 3,5-bis(trifluoromethyl)styrene with a second photosensitizer.

*<sup>a</sup>NMR yield.*

*<sup>b</sup>GC yield. <sup>c</sup>BI(OH)H (1.2 equiv.) was used instead of <sup>i</sup>Pr2NEt (4.0 equiv.).*

high reducing ability of the one-electron reduced species to promote the reduction process. The yield of **2b** further increased when incorporating [Ir(dF(CF3)ppy)2(dtbbpy)](PF6) as a second photosensitizer (**Table 5**, entry 3), and its concentration could be reduced to 1.0 mol% without lowering the yield (**Table 5**, entry 4). This result was assumed to be due to the high oxidizing ability of the excited state in addition to the sufficient reducing ability of the one-electron reduced species. The excited state of [Ir(dF(CF3)ppy)2(dtbbpy)](PF6) was found to be able to work as an energy transfer agent of **8** based on the luminescence quenching experiment (K<sup>q</sup> = 2.76 × 10<sup>4</sup> , **Supplementary Figure 3B**). However, it is considered to contribute to the reaction mainly as an electron transfer agent under the catalytic conditions owing to the efficient quenching by the sacrificial electron donor. On the other hand, the addition of fac-Ir(ppy)<sup>3</sup> resulted in only a small acceleration, which was probably attributable to the inferior oxidizing ability in the excited state (**Table 5**, entry 6). These results demonstrate that photosensitizers possessing both high oxidizing ability of the excited state and high reducing ability of the one-electron reduced species are advantageous as a second photosensitizer. The positive result on the addition of the two appropriate photosensitizers reflected the fact that the catalytic cycle was composed of the multiple photochemical processes, and the acceleration of the reduction process led to the enhancement of the catalytic activity when using <sup>i</sup>Pr2NEt as a sacrificial electron donor.

### Rate-Determining Step in the Hydrocarboxylation With BI(OH)H

The previous experiments demonstrated that the ratedetermining step of the catalytic cycle was the reduction process of the Rh(I) carboxylate species **9** when employing <sup>i</sup>Pr2NEt as a sacrificial electron donor. In order to investigate the contribution of BI(OH)H to the catalytic cycle, the similar examination was carried out using BI(OH)H as a sacrificial electron donor instead of <sup>i</sup>Pr2NEt. Interestingly, the resting-state was found to be Rh(I) π-benzyl intermediate **8** when a mixture of **1a**, catalytic amounts of **6** and [Ru(bpy)3](PF6)2, and 1.2 equiv. of BI(OH)H was irradiated by visible-light under CO<sup>2</sup> atmosphere (**Supplementary Figure 4**). In that case, **9** was not detectable even after prolonged irradiation, indicating that the rate-determining step obviously altered from the reduction process to the carboxylation process by changing the sacrificial electron donor. This result was also supported by the fact that the acceleration of the reaction by the addition of a second photosensitizer was not observed in the case of the reaction using BI(OH)H as a sacrificial electron donor (**Table 5**, entry 8). These observations suggested that the use of BI(OH)H strongly accelerated the reductive quenching cycle of [Ru(bpy)3] <sup>2</sup><sup>+</sup> to promote the reduction process.

### CONCLUSION

In this study, the improved catalytic conditions of the visiblelight driven hydrocarboxylation by Rh(I) and [Ru(bpy)3] 2+ catalysts were explored, and the detailed reaction mechanism was investigated. On the basis of the stoichiometric reactions of the possible rhodium intermediates, the proposed catalytic cycle was confirmed to be composed of (i) the hydrometallation of alkenes by Rh(I) monohydride species, (ii) the photochemical carboxylation of the Rh(I) benzyl species with CO2, (iii) the photoinduced 2-electron, 2-proton transfers to the Rh(I) carboxylate species, and (iv) the base-assisted carboxylic acid elimination. One strategy for the enhancement of the catalytic reaction was to employ BI(OH)H possessing superior reducing ability as a sacrificial electron donor instead of <sup>i</sup>Pr2NEt. It successfully improved the efficiency of the reaction, which had been major challenges in the previous catalytic conditions. The alteration of the resting-state by changing the sacrificial electron donor indicated that the addition of BI(OH)H significantly promoted the reduction process of the Rh(I) carboxylate species through the enhancement of the reductive quenching efficiency of [Ru(bpy)3] <sup>2</sup>+. Another strategy for the enhancement of the efficiency was to add the second photosensitizer in charge of the reductive quenching cycle. The acceleration of the catalytic reaction by the addition of the appropriate cyclometalated Ir(III) complex together with [Ru(bpy)3] <sup>2</sup><sup>+</sup> supported this hypothesis. These two effective strategies suggested that the promotion of the reduction processes was a key to enhance the catalytic activity in the present system. In addition to expand the versatility of the present hydrocarboxylation, this study would provide fundamental insights into the catalytic organic transformations by transitionmetal/photoredox dual catalysis.

### EXPERIMENTAL

### General

All operations were carried out under an argon atmosphere unless otherwise noted. <sup>1</sup>H, <sup>13</sup>C, and <sup>31</sup>P NMR spectra were recorded on Bruker DRX-500, JEOL ECZ-500, ECX-400, and ECS-400 spectrometers. <sup>31</sup>P and <sup>19</sup>F NMR chemical shifts were calibrated using external 85% H3PO<sup>4</sup> (δ: 0.0 ppm) and neat C6F<sup>6</sup> (δ: −164.9 ppm), respectively. IR spectra were recorded on an SC-100-VIR with an ATR PRO450-S accessory (JASCO Co., Ltd.). Emission spectra were recorded on an FP-6500 spectrofluorometer (JASCO Co., Ltd.). FAB-MS and FD-MS spectra were recorded on a JEOL JMS-700 spectrometer and a JMS-T100 spectrometer, respectively. Gas chromatography (GC-FID / TCD) was recorded on a Shimadzu GC-2010 spectrometer. Analytical thin-layer chromatography (TLC) was performed with a glass plate coated with silica gel (Wakogel B-5F). Visiblelight irradiation was performed with a Relyon Twin LED Light (3W × 2, λirr. = 425 ± 15 nm), and UV-visible-light irradiation was performed with an USHIO Optical Modulex OPM2-502XQ (500 W Xe lamp) with a super cold filter ZSC0750 (ASAHI Spectra Inc.).

THF, toluene, pentane and diethyl ether were purified by a solvent purification system by Glass Contour. Dehydrated dimethylacetamide (DMA) and dimethylformamide (DMF) were purchased from Kanto Chemical Co., Inc., degassed by argon bubbling and stored in a glovebox. Tertiary amines were distilled, degassed three times by freeze-pump-thaw method and stored under N2. Solvents for NMR measurements were dried over molecular sieves, degassed three times by freeze-pump-thaw method and stored under N2. All other solvents were distilled, degassed by argon bubbling and stored in a glovebox. CO<sup>2</sup> and H<sup>2</sup> gases were purchased from Taiyo Nippon Sanso Corporation. [Rh(coe)2Cl]<sup>2</sup> (Van der Ent et al., 1990), P(4-CF3C6H4)<sup>3</sup> (Suomalainen et al., 2001), Rh(PPh3)2(OAc) (**5**) (Grushin et al., 1995), [Rh(cod)(OAc)]<sup>2</sup> (Chatt and Venanzi, 1957), Rh(PPh3)3H (**6**) (Annibale and Song, 2014), [Ru(bpy)3](PF6)<sup>2</sup> (Damrauer et al., 1997), [Ru(dmbpy)3](PF6)<sup>2</sup> (Damrauer et al., 1997), [Ru(bpz)3](PF6)<sup>2</sup> (Schultz et al., 2015), Ir(ppy)2(dtbbpy)(PF6) (Tellis et al., 2014), [Ir(dF(CF3)ppy)2(dtbbpy)](PF6) (Slinker et al., 2004), fac-Ir(ppy)<sup>3</sup> (Tamayo et al., 2003), and diisopropylethylammonium acetate (Anouti et al., 2008) were prepared according to the published methods. 4-Cyanostyrene (**1a**) was prepared by Wittig reaction of 4-cyanobenzaldehyde (Falk et al., 2013). Other chemicals were purchased and used as received.

### Photocatalytic Reactions

For screening conditions with alkenes (**1a**–**1g**), a DMA solution (0.6 mL) of an alkene (0.060 mmol), [Rh(P(4-CF3C6H4)3)2Cl]<sup>2</sup> (**4**, 4.5 mg, 0.0021 mmol), photoredox catalyst(s), sacrificial electron donor and inorganic base (defined amounts) was prepared in a glass tube (ϕ 2.0 cm, 18 cm) under an argon atmosphere. Then the headspace gas was replaced by an atmospheric pressure of CO2, and the reaction vessel was put in a water bath placed at a distance of 10 mm from light sources. The mixture was irradiated with visible-light from blue LED lamp (λirr. = 425 nm, two sockets) or UV-visible-light from Xe lamp (λirr. = 380–800 nm) for defined time in the closed system. The product mixture was analyzed by <sup>1</sup>H NMR and GC to determine the NMR yield of the hydrocarboxylated product (**2a**–**2g**) and the GC yield of the hydrogenated product (**3a**, **3b**), respectively (internal standard: 1,1,2,2-tetrachloroethane).

For isolation of the methyl esters of the hydrocarboxylated products (**2a**, **2b**, **2d**), a DMA solution (1.2 mL) of a styrene (0.12 mmol), **4** (9.0 mg, 0.0042 mmol), [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol), BI(OH)H (58 mg, 0.24 mmol), and Cs2CO<sup>3</sup> (47 mg, 0.14 mmol) was prepared in a glass tube (ϕ 2.0 cm, 18 cm), and irradiated with visible-light from blue LED lamp (λirr. = 425 nm, three sockets) for defined time after replacement of the headspace gas by an atmospheric pressure of CO2. After irradiation, the reaction mixture was diluted with diethyl ether and extracted with H2O three times. The combined aqueous layer was acidified by 1N HCl aq., and then extracted with diethyl ether three times. The combined organic layer was dried over MgSO4, filtered and evaporated under reduced pressure to give the hydrocarboxylated product. Then, the product was dissolved in Et2O-MeOH, and TMSCHN<sup>2</sup> (excess) was added at 0 ◦C. The mixture was stirred at 0◦C for 30 min and the solvent was removed under reduced pressure. The crude product was purified by preparative TLC (AcOEt/n-hexane = 1/5) to give the corresponding methyl-esterified product.

# Preparations of Rhodium Complexes and Their Stoichiometric Reactions

### Preparation of [Rh(P(4-CF3C6H4)3)2Cl]<sup>2</sup> (4)

A solution of P(4-CF3C6H4)<sup>3</sup> (200 mg, 0.429 mmol) in toluene (2 mL) was added dropwise to a solution of [Rh(coe)2Cl]<sup>2</sup> (77 mg, 0.107 mmol) in toluene (2 mL) and the mixture was stirred at room temperature overnight. After removal of solvent under reduced pressure, the crude product was dissolved in THF and then pentane was added to induce precipitation. The precipitates were collected to give the target product (218 mg, 0.102 mmol, 95% yield). <sup>1</sup>H NMR (500 MHz, THFd8, r.t., δ/ppm): δ 7.73–7.67 (m, 24 H, PAr3), 7.39 (d, J = 8 Hz, 24 H, PAr3). <sup>13</sup>C{1H} NMR (125 MHz, r.t., THF-d8, δ/ppm): δ 139.3 (vt, N = 22 Hz, PAr3), 135.7 (s, PAr3), 132.1 (q, JC−<sup>F</sup> = 33 Hz, PAr3), 125.0 (s, PAr3), 124.6 (q, JC−<sup>F</sup> = 272 Hz, -CF3). <sup>31</sup>P{1H} NMR (202 MHz, THF-d8, r.t., δ/ppm): δ 53.4 (d, J = 194 Hz). <sup>19</sup>F NMR (471 MHz, THF-d8, r.t., δ/ppm): δ −60.6 (s). ESI-MS: m/z = 1035 [M/2 − Cl]+. Anal. Found (calcd for C84H48Cl2F36P4Rh2): C, 46.94 (47.11); H, 2.28 (2.26).

### Preparation of Rh(PCy3)2(OAc) (5′ )

[Rh(cod)(OAc)]<sup>2</sup> (60 mg, 0.111 mmol) and PCy<sup>3</sup> (125 mg, 0.444 mmol) were suspended in DMA (4 mL), and the mixture was irradiated with UV-visible-light (500 W Xe lamp, λirr. = 380– 800 nm) at room temperature with vigorous stirring for 30 h. The precipitate was filtered, washed with DMA and cold pentane (−35◦C), and then recrystallized from a minimum volume of pentane at −35◦C to yield the target compound (98 mg, 0.135 mmol, 61%). IR (KBr): ν(OCOas) = 1,528, ν(OCOsym) = 1,445 cm−<sup>1</sup> . <sup>1</sup>H NMR (500 MHz, C6D6, r.t., δ/ppm): δ 2.35 – 1.24 (m, 69 H, PCy<sup>3</sup> and O2CCH3). <sup>13</sup>C{1H} NMR (125 MHz, r.t., C6D6, δ/ppm): δ 188.2 (s, O2CCH3), 35.9 (vt, N = 10 Hz, PCy3), 31.1 (s, PCy3), 28.4 (vt, N = 5 Hz, PCy3), 27.3 (s, PCy3), 24.9 (s, O2CCH3). <sup>31</sup>P{1H} NMR (202 MHz, C6D6, r.t., δ/ppm): δ 59.1 (d, JP−Rh = 198 Hz). HR-MS (FAB): m/z = 722.3817 [M]<sup>+</sup> (calcd for [C38H69O2P2Rh]+: 722.3828).

### Preparation of an Authentic Sample of Rh(PCy3)2(OAc)(H)<sup>2</sup> (7′ )

[Rh(cod)(OAc)]<sup>2</sup> (60 mg, 0.111 mmol) and PCy<sup>3</sup> (125 mg, 0.444 mmol) were dissolved in THF (3 mL), and the mixture was stirred under H<sup>2</sup> (1 atm) at room temperature overnight. After removal of solvent, the crude product was dissolved in toluene and filtered through Celite <sup>R</sup> . The resultant solid after evaporation was washed with cold diethyl ether (−35◦C) to give the target compound (113 mg, 0.156 mmol, 70%). IR (KBr): ν(RhH) = 2,143, ν(OCOas) = 1,551, ν(OCOsym) = 1,436 cm−<sup>1</sup> . <sup>1</sup>H NMR (500 MHz, C6D6, r.t., δ/ppm): δ 2.20 – 1.22 (m, 69 H, PCy<sup>3</sup> and O2CCH3), −23.7 (dt, JH−Rh = 24, JH−<sup>P</sup> = 15 Hz, 2 H, Rh-H) <sup>13</sup>C{1H} NMR (125 MHz, r.t., C6D6, δ/ppm): δ 180.2 (s, O2CCH3), 35.8 (vt, JC−<sup>P</sup> = 10 Hz, PCy3), 30.5 (s, PCy3), 28.3 (vt, N = 5 Hz, PCy3), 27.1 (s, PCy3), 24.8 (s, O2CCH3). <sup>31</sup>P{1H} NMR (202 MHz, C6D6, r.t., δ/ppm): δ 51.1 (d, JP−Rh = 115 Hz). HR-MS (FD): m/z = 724.3975 (calcd for [C38H71O2P2Rh]+: 724.3984).

### Preparation of Rh(PPh3)2(η 3 -CHCH3(4-CNC6H4)) (8)

To a solution of Rh(PPh3)3H (**6**) (5.4 mg, 0.0060 mmol) in THFd<sup>8</sup> (0.6 mL) in a J. Young NMR tube was added 4-cyanostyrene (**1a**) (1.6 µL, 0.012 mmol) at room temperature. Rh(PPh3)2(η 3 - CHCH3(4-CNC6H4)) (**8**) formed almost quantitatively. <sup>1</sup>H NMR (500 MHz, THF-d8, −10◦C, δ/ppm): δ 7.47 – 7.00 (m, 30 H, PPh3), 6.81 (brd, J = 8 Hz, 1 H, Ar), 6.57 (brd, J = 7 Hz, 1 H, Ar), 6.10 (brd, J = 8 Hz, 1 H, Ar), 4.89 (brd, J = 7 Hz, 1 H, Ar), 2.47 – 2.40 (m, 1 H, -CHCH3), 0.92 – 0.87 (m, 3 H, - CHCH3). <sup>31</sup>P{1H} NMR (202 MHz, THF-d8, −10◦C, δ/ppm): δ 46.2 (dd, JP−Rh = 263 Hz, JP−<sup>P</sup> = 31 Hz), 39.6 (dd, JP−Rh = 178 Hz, JP−<sup>P</sup> = 31 Hz). HR-MS (FAB): m/z = 757.1539 (calcd for [C45H38NP2Rh]+: 757.1535).

### Preparation of Rh(PPh3)3(η 1 -O2CCHCH<sup>3</sup> (4-CNC6H4)) (9)

To a solution of Rh(PPh3)3H (**6**) (60 mg, 0.090 mmol) in toluene (3 mL) was added dropwise a solution 2-(4 cyanophenyl)propionic acid (16 mg, 0.090 mmol) in toluene (2 mL), and the mixture was stirred at room temperature for 4 h. After removal of solvent under reduced pressure, the crude product was dissolved in toluene and then pentane was added to induce precipitation. The precipitates were collected to give the target product as a 9: 1 mixture with Rh(PPh3)2(η 2 - O2CCHCH3(4-CNC6H4)) which was formed by dissociation of PPh<sup>3</sup> from **9** (59 mg). **9** IR (KBr): ν(CN) = 2,224, ν(OCOas) = 1,604, ν(OCOsym) = 1,342 cm−<sup>1</sup> . <sup>1</sup>H NMR (500 MHz, THF-d8, −60◦C, δ/ppm): δ 7.56 – 6.79 (m, 47 H, PPh3, O2CCHCH3Ar), 6.30 (brd, J = 8 Hz, 2 H, O2CCHCH3Ar), 1.65 (brq, J = 7 Hz, 1 H, O2CCHCH3Ar), 0.25 (brd, J = 7 Hz, 3 H, O2CCHCH3Ar). <sup>13</sup>C{1H} NMR (125 MHz, r.t., THF-d8, −30◦C, δ/ppm): δ 176.6 (s, O2CCHCH3Ar), 150.8 (s, O2CCHCH3Ar), 135.7 (vt, N = 6 Hz, PPh3), 131.1 (s, O2CCHCH3Ar), 129.6 (s, O2CCHCH3Ar), 129.2 (s, PPh3), 127.8 (s, PPh3), 127.5 (s, PPh3), 127.4 (s, PPh3), 119.8 (s, -CN), 108.9 (s, O2CCHCH3Ar), 47.8 (s, O2CCHCH3Ar), 18.1 (s, O2CCHCH3Ar). <sup>31</sup>P{1H} NMR (202 MHz, THF-d8, −30◦C, δ/ppm): δ 51.5 (dt, JP−Rh = 174 Hz, JP−<sup>P</sup> = 41 Hz), 34.9 (dd, JP−Rh = 153 Hz, JP−<sup>P</sup> = 41 Hz). HR-MS (FAB): m/z = 801.1453 [M−(PPh3)]<sup>+</sup> (calcd for [C46H38NO2P2Rh]+: 801.1433). Rh(PPh3)2(η 2 -O2CCHCH3(4- CNC6H4)) <sup>31</sup>P{1H} NMR (202 MHz, THF-d8, −30◦C, δ / ppm): δ 57.6 (brd, JP−Rh = 210 Hz).

### Redox-Photosensitized Reaction of Rh(PPh3)2(OAc) (5)

Rh(PPh3)2(OAc) (**5**) (2.7 mg, 0.0040 mmol), [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol), PPh<sup>3</sup> (1.5 mg, 0.0060 mmol), <sup>i</sup>Pr2NEt (41 µL, 0.24 mmol) and DMA (0.6 mL) were added in a glass tube with a magnetic stirrer. The mixture was irradiated with visible-light (λirr. = 425 nm) at room temperature for 6 h. The solvent was removed under reduced pressure, and the resulting solid was analyzed by <sup>1</sup>H and <sup>31</sup>P NMR spectroscopies (THF-d8, internal standard: mesitylene). The Rh(I) monohydride species corresponding to Rh(PPh3)3H (**6**) formed in 57% yield (based on the Rh-hydride signal) and ca. 30% of the starting material **5** remained in the product mixture. **6** was highly fluxional in the reaction solution at room temperature in the presence of triphenylphosphine. <sup>1</sup>H NMR (500 MHz, THF-d8, δ/ppm, r.t.): −8.45 (brd, JH−Rh = 13 Hz, Rh-H). <sup>31</sup>P NMR (202 MHz, THF-d8, δ/ppm, r.t.): 41 (br). The intensity of the Rh-hydride signal was significantly increased at room temperature when Rh(PPh3)3H (**6**) synthesized alternatively was added to the reaction mixture, also supporting the formation of **6**. In addition, when the solution was cooled to −90◦C, the signals attributed to Rh(PPh3)4H were observed instead, indicating **6** was converted to Rh(PPh3)4H at low temperature. <sup>1</sup>H NMR (500 MHz, THF-d8, δ/ppm, −90◦C): −13.5 (dq JH−Rh = 118, JH−<sup>P</sup> = 15 Hz, Rh-H). <sup>31</sup>P NMR (202 MHz, THF-d8, δ/ppm, −90◦C): 33.4 (dm, JP−Rh = 112), 30.2 (dd, JP−Rh = 162, JP−<sup>P</sup> = 32 Hz). The spectrosopic data was analogous to the values reported previously (Dewhirst et al., 1968; Strauss and Shriver, 1978).

### Redox-Photosensitized Reaction of Rh(PCy3)2(OAc) (5′ )

Rh(PCy3)2(OAc) (**5** ′ ) (2.9 mg, 0.0040 mmol), [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol), <sup>i</sup>Pr2NEt (41 µL, 0.24 mmol) and DMA (0.6 mL) were added in a glass tube with a magnetic stirrer. The mixture was irradiated with visible-light (λirr. = 425 nm) at room temperature for 12 h. The solvent was removed under reduced pressure, and the resulting solid was analyzed by <sup>1</sup>H and <sup>31</sup>P NMR spectroscopies (C6D6, internal standard: mesitylene). Rh(PCy3)(OAc)(H)<sup>2</sup> (**7** ′ ) was formed in 94% yield (based on the rhodium hydride signal). The spectroscopic feature well agreed with the complex **7** ′ synthesized alternatively (vide supra). No other Rh hydride signal was observed even by addition of 1 equiv. of PCy<sup>3</sup> to the reaction mixture.

### Reaction of Rh(PPh3)2(OAc)(H)<sup>2</sup> (7) and <sup>i</sup>Pr2NEt

A solution of Rh(PPh3)2(OAc) (**5**) (2.7 mg, 0.0040 mmol) in C6D<sup>6</sup> (0.6 mL) was put in a J. Young NMR tube under hydrogen atmosphere at room temperature. After 2 h, Rh(PPh3)2(OAc)(H)<sup>2</sup> (**7**) formed in situ. <sup>1</sup>H NMR (500 MHz, C6D6, δ/ppm, r.t.): 7.86 – 7.00 (m, PPh3), −20.8 (dt, JH−Rh = 22 Hz, JH−<sup>P</sup> = 17 Hz, Rh-H). <sup>31</sup>P{1H} NMR (202 MHz, C6D6, δ/ppm, r.t.): 41.4 (d, JP−Rh = 121 Hz). Then, the reaction solution was degassed by freeze-pump-thaw method 3 times to remove the hydrogen gas. Addition of 1 equiv. of PPh<sup>3</sup> (1.5 mg, 0.0060 mmol) and 80 equiv. of <sup>i</sup>Pr2NEt (41 µL, 0.24 mmol) gave Rh(PPh3)3H (**6**) in 58% yield (based on the rhodium hydride signal). Ca. 20% of **7** remained in the reaction mixture. The spectroscopic data of **6** was identical with that of the sample prepared by the photosensitizing reaction of Rh(PPh3)2(OAc) (**5**). The same reaction was also performed in DMA. In this case, **7** converted fully after addition of PPh<sup>3</sup> and <sup>i</sup>Pr2NEt, and **6** was obtained as a major product.

### Reaction of Rh(PPh3)3H (6) and [iPr2NHEt]<sup>+</sup> [CH3COO]<sup>−</sup>

To a solution of Rh(PPh3)3H (**6**) (3.6 mg, 0.0040 mmol) in C6D<sup>6</sup> (0.6 mL) was added diisopropylethylammonium acetate (1.1 mg, 0.0060 mmol) and the mixture was stirred at room temperature for 10 min. Rh(PPh3)2(OAc)(H)<sup>2</sup> (**7**) formed quantitatively. The spectroscopic data of the product in C6D<sup>6</sup> was identical with those of the sample prepared by the hydrogenation of Rh(PPh3)2(OAc) (**5**).

### Redox-Photosensitized Reaction of Rh(PPh3)3Cl

Rh(PPh3)3Cl (3.7 mg, 0.0040 mmol), [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol), <sup>i</sup>Pr2NEt (41 µL, 0.24 mmol) and DMF-d<sup>7</sup> (0.6 mL) were added in a J. Young NMR tube. The mixture was irradiated with visible-light (λirr. = 425 nm) at room temperature for 6 h. Rh(PPh3)3H (**6**) formed in 52% yield (based on the Rh-hydride signal) and ca. 40% of the starting material remained in the reaction mixture. <sup>1</sup>H NMR (500 MHz, δ/ppm, −50◦C): −8.48 (ddt, JH−Rh = 101 Hz, JH−<sup>P</sup> = 18 Hz, JH−<sup>P</sup> = 15 Hz, Rh-H). <sup>31</sup>P NMR (202 MHz, δ/ppm, −50◦C): 43.7 (dd, JP−Rh = 170 Hz, JP−<sup>P</sup> = 25 Hz), 38.9 (dm, JP−Rh = 149 Hz). The spectrosopic data was analogous to the values reported previously (Dewhirst et al., 1968; Strauss and Shriver, 1978).

### Reaction of Rh(PPh3)3H (6) With 4-Cyanostyrene (1a) and CO<sup>2</sup>

To a solution of Rh(PPh3)3H (**6**) (3.6 mg, 0.0040 mmol) in DMA (0.6 mL) was added 4-cyanostyrene (**1a**) (2.6 µL, 0.020 mmol) in a J. Young NMR tube at room temperature. Rh(PPh3)2(η 3 -CHCH3(4-CNC6H4)) (**8**) was generated in situ almost quantitatively. Then, [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol) was added, and the mixture was irradiated with visible-light (λirr. = 425 nm) under a CO<sup>2</sup> atmosphere (1 atm) at room temperature for 30 min. Rh(PPh3)3(η 1 -O2CCHCH3(4- CNC6H4)) (**9**) was formed almost quantitatively, as the spectroscopic data well agreed with the complex synthesized alternatively.

### Reaction of Rh(PPh3)3H (6) With Methyl Acrylate (1f) and CO<sup>2</sup>

To a solution of Rh(PPh3)3H (**6**) (3.6 mg, 0.0040 mmol) and [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol) in DMA (0.6 mL) was added methyl acrylate (**1f**) (1.8 µL, 0.020 mmol) in a J. Young NMR tube at room temperature. The mixture was exposed to a CO<sup>2</sup> atmosphere (1 atm) under dark at room temperature for 3 h. The observed <sup>31</sup>P{1H} NMR spectral feature was analogous to that of **9**, indicating the formation of Rh(PPh3)3(η 1 -O2CCHCH3(CO2CH3)) (**10**). <sup>31</sup>P{1H} NMR (162 MHz, DMA, −15◦C, δ/ppm): δ 50.6 (dt, JP−Rh = 174 Hz, JP−<sup>P</sup> = 44 Hz), 33.5 (dd, JP−Rh = 153 Hz, JP−<sup>P</sup> = 44 Hz). The reaction solution was treated with NaHCO<sup>3</sup> aq., and then acidified with 1N HCl aq. The organic layer was extracted with diethyl ether three times, dried over MgSO4, filtered and evaporated under reduced pressure to give methyl methylmalonate in 54% yield (based on **6**).

### Observation of the Reaction Intermediates Under Catalytic Conditions

Rh(PPh3)3H (**6**) (3.6 mg, 0.0040 mmol), [Ru(bpy)3](PF6)<sup>2</sup> (1.0 mg, 0.0012 mmol), BI(OH)H (17.3 mg, 0.072 mmol), 4-cyanostyrene (**1a**) (7.7 µL, 0.060 mmol) and DMA (0.6 mL) were added in a J. Young NMR tube. The mixture was irradiated with visible-light (λirr. = 425 nm) under CO<sup>2</sup> atmosphere (1 atm) at room temperature for 30 min. The <sup>31</sup>P{1H} NMR spectra were observed at −15◦C before and after irradiation, and only Rh(PPh3)2(η 3 - CHCH3(4-CNC6H4)) (**8**) was observed as a resting-state in both spectra.

### Theoretical Study

Theoretical calculations were performed at the DFT level with the Gaussian 09 package. The geometry optimizations were performed using the mPW1PW91 functional (Adamo and Barone, 1998). The LanL2DZ basis set was used for all atoms and extended by a polarization function (except for H) (Dunning and Hay, 1976; Wadt and Hay, 1985a,b). To address solvation effects, the conductor-like polarizable continuum model (CPCM, N,N-Dimethylacetamide) (Tomasi et al., 2005) was used for the ground and excited states. For validation, vibrational frequencies were calculated for the ground and excited states. The orbital plots as well as the graphical representations were performed using Molekel (Varetto, 2009). Natural bond orbital (NBO) analysis was used to predict and interpret the computational results (Glendening et al., 2001). Total ZPE energies and cartesian coordinates of computed structures are given in **Supplementary Table 3**.

### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

# AUTHOR CONTRIBUTIONS

NN and KM performed the experiments. KS instructed the experiments of NN. KM performed the theoretical calculations. NI and JT supervised the project. KM and NI wrote the paper.

# FUNDING

The authors are grateful for the financial supports by an ACT-C program JPMJCR12Y3 from JST and JSPS KAKENHI Grant no. 15H05800, 17H06143, and 24245019.

# ACKNOWLEDGMENTS

The authors also thank Prof. Osamu Ishitani for valuable discussions on the use of photosensitizers and sacrificial electron donors.

### REFERENCES


# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00371/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer MA declared a shared affiliation, with no collaboration, with the authors NN, KS, JT, NI, to the handling editor at the time of review.

Copyright © 2019 Murata, Numasawa, Shimomaki, Takaya and Iwasawa. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Electronic Effects of Substituents on fac-M(bpy-R)(CO)<sup>3</sup> (M = Mn, Re) Complexes for Homogeneous CO<sup>2</sup> Electroreduction

Laura Rotundo1,2,3, Emanuele Azzi <sup>1</sup> , Annamaria Deagostino<sup>1</sup> , Claudio Garino1,2,3 , Luca Nencini <sup>1</sup> , Emanuele Priola1,2, Pierluigi Quagliotto1,2, Riccardo Rocca1,2,3 , Roberto Gobetto1,2,3 \* and Carlo Nervi 1,2,3 \*

### Edited by:

Hitoshi Ishida, Kitasato University, Japan

### Reviewed by:

Baojian Shen, China University of Petroleum, Beijing, China Yusuke Tamaki, Tokyo Institute of Technology, Japan

### \*Correspondence:

Roberto Gobetto roberto.gobetto@unito.it Carlo Nervi carlo.nervi@unito.it

### Specialty section:

This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry

Received: 29 January 2019 Accepted: 21 May 2019 Published: 05 June 2019

### Citation:

Rotundo L, Azzi E, Deagostino A, Garino C, Nencini L, Priola E, Quagliotto P, Rocca R, Gobetto R and Nervi C (2019) Electronic Effects of Substituents on fac-M(bpy-R)(CO)3 (M = Mn, Re) Complexes for Homogeneous CO2 Electroreduction. Front. Chem. 7:417. doi: 10.3389/fchem.2019.00417 <sup>1</sup> Department of Chemistry, Università degli Studi di Torino, Turin, Italy, <sup>2</sup> NIS Interdepartmental Centre, Università degli Studi di Torino, Turin, Italy, <sup>3</sup> Consorzio Interuniversitario Reattività Chimica e Catalisi (CIRCC), Bari, Italy

Synthesis and characterization of 14 new 2,2′ -bipyridine metal complexes fac-M(bpy-R)(CO)3X (where M = Mn, X = Br or M = Re, X = Cl and R = -CF3, -CN, -Ph, -PhOH, -NMe2) are reported. The complexes have been characterized by NMR, IR spectroscopy and elemental analysis. Single crystal X-Ray diffraction structures have been solved for Re(dpbpy)(CO)3Cl (dpbpy = 4,6-diphenyl-2,2′ -bipyridine) and Re(hpbpy)(CO)3Cl (hpbpy = 4-(2-hydroxy-phenyl)-6-phenyl-2,2′ -bipyridine). Electrochemical behaviors of the complexes in acetonitrile under Ar and their catalytic performances for CO<sup>2</sup> reduction with added water and MeOH have been investigated by cyclic voltammetry and controlled potential electrolysis. The role of the substituents on the electrochemical properties and the related over potentials required for CO<sup>2</sup> transformation have been analyzed. The complexes carrying only electron withdrawing groups like -CF3, -CN totally lose their catalytic activities toward CO<sup>2</sup> reduction, whereas the symmetric -NMe<sup>2</sup> substituted and push-pull systems (containing both -NMe<sup>2</sup> and -CF3) still display electrocatalytic current enhancement under CO<sup>2</sup> atmosphere. The complexes carrying a phenyl or a phenol group in position 4 show catalytic behaviors similar to those of simple M-bpy systems. The only detected reduction product by GC analysis is CO: for example, fac-Re (bpy-4,4′ -NMe2)(CO)3Cl gives CO with high faradic efficiency and a TON of 18 and 31, in absence of external proton source and with 5% MeOH, respectively. DFT calculations were carried out to highlight the electronic properties of the complexes; results are in agreement with experimental electrochemical data.

Keywords: CO2 electroreduction, Mn complexes, Re complexes, DFT calculations, bipy ligands, homogeneous catalysis, electron-withdrawing, electron-donating

# INTRODUCTION

Nowadays the CO<sup>2</sup> concentration in the atmosphere is continuously increasing alongside with the overall world energy demand. Converting carbon dioxide via electrochemical reduction into useful chemicals and fuels for energy storage is an attractive and promising approach. CO<sup>2</sup> reduction is a competition between its thermodynamic and kinetic: the one electron reduction in water occurs at a very negative potential (−1.90 V vs. SCE at pH 7) (Hammouche et al., 1988; Saveant, 2008) because it requires a drastic change in the geometry, from the linear CO<sup>2</sup> molecule to the bent CO−• 2 radical anion. The reason of such high negative overpotential is due to the slow kinetics of the electron transfer, which is associated to the different geometries of the neutral and reduced species, respectively. However, the reduction reactions involving multiple electron transfers coupled with proton transfers provide a significant lowering of the thermodynamic barrier. To avoid the CO−• 2 as intermediate and to lower the energy cost of the reduction process, key catalytic strategies have been developed with the aim of obtaining the various products selectively. The best electrocatalysts currently studied work at a potential 100 mV negative with respect to E<sup>0</sup> CO2/P (where P generically indicates the reduction products, CO, HCOOH, HCHO, CH3OH, CH4) (Francke et al., 2018; Franco et al., 2018). Despite the numerous advantages of heterogeneous electrocatalysis (Sun et al., 2016; Rotundo et al., 2019a), clever integration with the homogeneous counterpart allows a rational design of the catalysts, by tuning both the metal center and/or the ligand. One of the greatest challenge of the homogeneous approach lies in the search of stability, durability and improved turnover number (TON) efficiencies (Grice, 2017; Takeda et al., 2017; Wang et al., 2019). Bipyridine transition metal complexes represent one of the most studied classes of molecular electrocatalysts since the 1980s (Stanbury et al., 2018). Bipyridine has been extensively studied as ligand in the field of electro and photo-catalysis because of the capability to store electrons and subsequently delocalize electronic density on its π orbitals (Vlcek, 2002; Elgrishi et al., ˇ 2017). Both fac-Re(bpy)(CO)3Cl and fac-Mn(bpy)(CO)3Br are capable of reducing CO<sup>2</sup> to CO with high faradaic efficiency (Hawecker et al., 1984; Bourrez et al., 2011). Comparing these two complexes, Mn exhibits a catalytic peak that is shifted more anodically (around 300 mV, E<sup>p</sup> = −1.51 V vs. Ag/AgCl in MeCN) in comparison with the second reduction of Re analogous (E<sup>p</sup> = −1.8 V vs. Ag/AgCl in MeCN). Another important difference is that usually the Mn-bpy complexes show their catalytic activities only in the presence of external proton sources. To better shed light on this uniqueness, our research group synthesized Mn(pdbpy)(CO)3Br (Franco et al., 2014, 2017), in which two pendant phenolic groups act as local proton source (Costentin et al., 2012), capable of reducing CO<sup>2</sup> even in anhydrous acetonitrile. In this case a considerable amount of HCOOH was also detected. Conversely the analogous complex in which the OH groups were replaced by methoxy groups did not show any catalytic activity without the addition of Brønsted acid.

The electrochemical behavior of the two well-known Re and Mn-bpy complexes could be reasonably altered by varying the bipyridine moiety, i.e., introducing electron-withdrawing and electron-donating groups (Machan et al., 2014; Walsh et al., 2015; Stanbury et al., 2018). Kubiak and his group investigated the effect of 4,4′ -di-tertbutyl-2,2′ -bipyridine (tBu2-bpy) firstly on rhenium carbonyl complexes, secondarily on manganese (Smieja and Kubiak, 2010; Smieja et al., 2013). In other works they studied the role of the modification in the 6,6′ position of the bipyridine (Sampson et al., 2014; Sampson and Kubiak, 2016). A similar approach has already been applied to Mo and W-bpy complexes (Franco et al., 2017; Rotundo et al., 2018), to both Mn and Rebpy complexes by some of us (Franco et al., 2017; Rotundo et al., 2019b) and in the current work.

The target of the modification is the reduction potential of the catalyst, namely E 0 cat: usually a less negative E 0 cat corresponds to a decreased rate of CO<sup>2</sup> conversion (Francke et al., 2018). Electronic properties of organic groups are commonly described by the inductive (±I) and mesomeric (±M) effects. Substituents with electron withdrawing groups like -CF<sup>3</sup> (strong –I effect) and -CN (weaker –I effect) and with electron donating groups, like - N(Me)<sup>2</sup> (strong +M effect), -Ph and -PhOH (weaker +M effect), were placed in the 4,4′ , 4,6, and 5,5′ positions of the bipyridine ligand coordinated to the metals. Combining both push and pull effects in the so called "push-pull" system, an electronic gradient is forced through the bipyridine. In this paper we explore the electrocatalytic properties of novel Mn and Re bpy-type complexes, bearing 7 differently substituted ligands (**Scheme 1**). More generally electron-donating groups are expected to convey greater nucleophilicity to the metal-center, although catalysis should require higher overpotentials. DFT calculations have been used as complementary tool to better correlate experimental electrochemical data, whereas Controlled Potential Electrolysis (CPE) experiments are useful to elucidate the stability and durability of the catalysts in acetonitrile solutions.

### RESULTS

# Synthesis and Structure

Fourteen new 2,2′ -bipyridine metal complexes fac-M(bpy-R)(CO)3X where M = Mn, X = Br or M = Re, X = Cl are reported, namely Mn(bpy-4,4′ -CF3)(CO)3Br (**1a**), Mn(bpy-5,5′ -CF3)(CO)3Br (**1b**), Mn(bpy-4,4′ -CN)(CO)3Br (**1c**), Mn(bpy-4,4′ -CF3-NMe2)(CO)3Br (**1d**), Mn(dpbpy)(CO)3Br (**1e**), Mn(hpbpy)CO)3Br (**1f**), Mn(bpy-4,4′ -NMe2)(CO)3Br (**1g**), Re(bpy-4,4′ -CF3)(CO)3Cl (**2a**), Re(bpy-5,5′ -CF3)(CO)3Cl (**2b**), Re(bpy-4,4′ -CN)(CO)3Cl (**2c**), Re(bpy-4,4′ -CF3-NMe2)(CO)3Cl (**2d**), Re(dpbpy)(CO)3Cl (**2e**), Re(hpbpy)(CO)3Cl (**2f**), Re(bpy-4,4′ -NMe2)(CO)3Cl (**2g**).

The ligands and the corresponding Mn and Re complexes have been synthesized according to the procedure reported in the experimental section. The complexes have been characterized by NMR, IR spectroscopy and elemental analysis. Single crystal X-Ray diffraction structures have been solved for **2e** and **2f** (for XRD data see **Tables S1**–**S6**).

The complex Re(dpbpy)(CO)3Cl (**2e**) crystallized from both acetonitrile and benzene solutions by slow evaporation, forming prismatic orange platelets of a phase of the pure molecular product (structure A, **Figure S1a**) and a solvate with two benzene

molecules (structure B, **Figure S1b**). The first presents the monoclinic centrosymmetric P21/a space group (**Figure S1c**) while the second the monoclinic centrosymmetric P2/n space group (**Figure S1d**). The complex Re(hpbpy)(CO)3Cl (**2f**) crystallized in dark by slow evaporation of solutions of toluene and benzene as yellow platelets and from ethyl acetate as orange prisms, both stable to air (**Figure S2a**). The crystal structure has been obtained from a platelet obtained from toluene solution and has monoclinic P21/n space group type (**Figure S2b**). The structures of both **2e** and **2f** present a quite distorted octahedral geometry around the rhenium center, as can be seen by the values in **Table 1**. The coordination bond of N1 is 0.1 Å longer than that of N2 (see **Figure 1** for numeration) and a similar asymmetry can be observed in the coordination of CO in trans position to the nitrogens. This asymmetric coordination, also present in other terpyridine derivatives (Anderson et al., 1990; Civitello et al., 1993; Wang et al., 2013; Klemens et al., 2016) is completely different from the very symmetrically bonded bpy derivatives (with N-Re average distances equal in the two coordinating pyridyl rings and long 2.17 Å). At the same time, while most of the 2,2′ -bipyridine derivatives of fac-Re(CO)3Cl complex unit are almost planar respect to the basal OC–Re–CO plane (Kurz et al., 2006; Smieja and Kubiak, 2010; Bullock et al., 2012; Machan et al., 2014; Manbeck et al., 2015), in the case of **2e** and **2f**, the ligand is distorted outside this plane (see **Table 1**). This behavior can be detected in all the terpyridines, in which the third pyridine ring is not coordinated to the metal center and is equivalent to a phenyl ring, and in 6-phenyl substituted 2,2′ -bipyridine derivatives. The reason of such distortion is clarified by considering the steric hindrance of the vicinal phenyl ring (the distance between the nearest CO group and the centroid of the phenyl ring in ortho to N1 is about 3.2 Å) that pushes up all the framework of the organic ligand, modifying the Re environment. This phenyl group is rotated to follow the shape of carbonyls with a torsion angle respect to the central pyridine ring of about 130◦ .

The distortion effects observed in the solid state are predicted also by molecular DFT calculations. Indeed, the optimized geometries of **2e** and **2f** perfectly overlap with the experimentally determined structures (**Figure 1D** and **Figure S3**), confirming that these anomalies originate from the coordination of 6-phenyl substituted 2,2′ -bipyridine ligands to carbonyl complexes and are not induced by crystal packing contributions.



The crystal packing of **2e** (both structure A and structure B) and **2f** is dominated by weak C–H···Cl, C–H···O and π···π stacking interactions. In the case of **2f**, the presence of - OH group on the organic ligand interacting with the chloride induces the formation of hydrogen-bonded molecular chains (see **Figure 1C** and **Figure S2c**). For a more detailed analysis on the crystal packing consults the Supplementary Material (see **Figures S4–S9**).

### Cyclic Voltammetry Under Ar

Cyclic Voltammetries (CVs) of all manganese complexes are reported in **Figure 2**. The CV of the Mn(bpy)(CO)3Br under our experimental conditions (in black) is included for comparison. This complex undergoes two successive irreversible reduction reactions and two reoxidation peaks (Bourrez et al., 2011). The first (Ep1 = −1.29 V vs. Ag/AgCl) and the second (Ep2 = −1.51 V vs. Ag/AgCl) reduction processes lead to the formation of the dimer and the mononuclear pentacoordinated anion species, respectively. Reoxidations of the pentacoordinated anion and of the dimer are located at −1.09 and −0.21 V vs. Ag/AgCl, respectively. The new synthesized complexes are supposed to display similar electrochemical behavior. **Table 2** reports the peak potentials of the first and second reductions. As expected, the presence of –CF<sup>3</sup> (**1a** and **1b**) and –CN (**1c**) shifts first and second potentials toward more positive values, when compared to Mn(bpy)(CO)3Br. In a recent paper (Rawat et al., 2019) DFT calculations suggested that electron-withdrawing substituents like -CF<sup>3</sup> stabilize the radical anion. Furthermore,

the formation of all Mn-Mn dimers was indicated as unfavorable. However, the electrochemical mechanism outlined above is commonly accepted, and the Mn dimer is strongly favored. We experimentally found that all complexes undergo a first and second chemically irreversible reductions. For example, **1a** shows two chemical irreversible processes followed by the reoxidations of the pentacoordinated radical anion and that of the dimer (**Figure S10**), even at high scan rates (1 V/s, **Figure S11**), thus confirming the general mechanism. CV of **1g** confirms that the strong electron-donating properties of dimethyl amino group result in more negative reduction potentials with respect to Mn(bpy)(CO)3Br. The push-pull system **1d**, **1e** and **1f** display similar potential values to the unsubstituted bpy-complex. In some complexes decomposition processes occurring after reduction generate small peaks (i.e., for **1e** and **1f** E<sup>p</sup> = −1.65 and −1.63 V vs. Ag/AgCl are observed, respectively). The very negative peak around −3V vs. Ag/AgCl, is commonly assigned to a ligand-centered reduction.

CVs of all rhenium complexes are reported in **Figure 3**. CV of the reference Re(bpy)(CO)3Cl under our experimental conditions is included for comparison. This complex exhibits a first reversible reduction, due to the formation of the radical anion, which is more stable than the analogous with manganese, thus resulting in electrochemical reversibility and no presence of the reoxidation peak of the dimer (Hawecker et al., 1984). The second reduction leads to the pentacoordinated anion. Intrinsically, rhenium complexes require slightly higher overpotentials with respect to the corresponding manganese ones. The first reversible reduction of Re(bpy)(CO)3Cl is located at E1/<sup>2</sup> = −1.35 V vs. Ag/AgCl, whereas the second chemically irreversible reduction is at Ep2 = −1.80 V vs. Ag/AgCl. Similarly to manganese, the new synthesized rhenium complexes show no significant difference in the electrochemical pathway under Ar when compared to the reference Re(bpy)(CO)3Cl (**Table 2**). Electron withdrawing groups (**2a**, **2b,** and **2c**) shift the reduction processes toward less negative values; electron donating groups like in **2g** essentially merge the first and second reductions into a single peak. The push-pull system **2d**, in analogy with the case of **1d**, do not significantly alter the positioning of the reduction potentials. The third peak, around −2.5V, also for these complexes, is generally attributed to a bpy-centered reduction.

The reduction potentials estimated by DFT calculation are in excellent agreement with the experimental values obtained for the first reduction potential of all rhenium complexes (**Figure 4**). This confirms the reversible nature of the first reduction process, leading to stable radical anions characterized by a very small increase (about 0.05 Å) in the Re–Cl bond length (**Figure S12**). Conversely, in the case of manganese complexes, DFT calculations, which compute thermodynamic reduction potentials, are not suitable to estimate the irreversible electrochemical behavior of the compounds (**Figure S13**). Indeed, the high instability of the radical anion leads to the weakening of the Mn–Br bond, as clearly evidenced by the significant increase of the Mn–Br bond length in the anion structures (**Figure S14**). Our DFT calculations that include weak interactions, are in agreement with the chemically irreversibility of the first reduction and dimer production, even in the case of electron-withdrawing substituents. For example, the formation of the dimer [Mn(bpy)(CO)3]<sup>2</sup> from its radical anion is favored by 143.5 kJ/mol, and in the case of **1a** is favored by 102.6 kJ/mol.

### Cyclic Voltammetry Under CO<sup>2</sup>

The electrochemical behavior of manganese complexes under CO<sup>2</sup> and with H2O (5%v) is reported in **Figure 5** for **1d** to

**1g** and in **Figure S15** for **1a**, **1b,** and **1c**, (these complexes are catalytically inactive toward CO<sup>2</sup> reduction). CVs of **1a-1g** under CO<sup>2</sup> with 5%v MeOH are included for comparison in **Figure S15** too. All complexes do not show significant current increases switching from Ar to CO<sup>2</sup> atmosphere. TABLE 2 | Reduction peak potentials [V] from the CV of Mn and Re complexes.


While this is expected for manganese complexes, **1g** shows a current increase, though limited, even in absence of Brønsted acid addition.

The electrochemical behavior of rhenium complexes under CO<sup>2</sup> and with 5%v MeOH is reported in **Figure 6** for **2d** to **2g** and in **Figure S16** for **2a**, **2b,** and **2c**. CVs of **2a-2g** under CO<sup>2</sup> with 5%v H2O are included for comparison in **Figure S16** too. All complexes exhibit catalytic current switching from Ar to CO<sup>2</sup> atmosphere, even in absence of Brønsted acid, as expected for rhenium complexes.

### Controlled Potential Electrolysis Experiments of the Complexes

Bulk electrolysis experiments of all manganese and rhenium complexes series under CO<sup>2</sup> were performed upon setting the potential at values slightly negative to the second reductions, with and without external added Brønsted acids (water and methanol, 5%). A CO<sup>2</sup> flow of 50 mL min−<sup>1</sup> was kept constant during the experiments, gaseous products were determined by gas chromatography, and formate, if present, was assessed by NMR spectroscopy at the end of the experiments. **Table 3** summarizes the results obtained during these CPE experiments. A general trend can be outlined: addition of water results in increased TONs for Mn complexes (**Table 3** and **Figure S17**), differently from methanol, which drops them to lower values (**Table S7**). On the other hand, for the case of Re complexes, addition of methanol (**Table 3** and **Figure S18**) seems to enhance the catalytic activity with respect to water, except for complex **2d**, which is catalytically inactive, despite the promising current increase in presence of MeOH under CO2.

# CONCLUSIONS

In summary, a systematic study of the effect of the electronic properties of the substituents on 2,2′ -bipyridine Mn and Re complexes was conducted. Electron-withdrawing substituents shift the reduction potentials to more positive values, and

eventually inhibit the catalytic activities of the corresponding Mn and Re complexes toward CO<sup>2</sup> reduction. In the case of electron-donating substituents the opposite trend is observed. These observations are in agreement with the induced electron density localized on the metal, strongly influencing the change in the reactivity with the weak electrophile CO2. Increasing or decreasing the electron density on the metal should facilitate or prevent the formation of intermediate in which the CO<sup>2</sup> is coordinated to the metal. Another interesting effect in varying the electron properties of the substituents is the merging of the first and second reduction processes, observed in some Re and Mn complexes. A judicious selection of bpy substituents provides an alternative way to the use of bulky substituents to prevent dimer formation (Sampson et al., 2014), with the aim of transforming two 1e reduction into a single **2e** reduction process (CO<sup>2</sup> to CO reduction requires **2e**).

It is interesting to note that in spite of the presence of a - OH group in **1f**, no formic acid is detected, probably because in contrast with Mn carrying local proton sources (Franco et al., 2014, 2017) the hydroxyl group is located far from the metal center, thus the generation of the metal-hydride, commonly considered catalyst for formate production, is no longer entropically favored. It is interesting to note how CVs under CO<sup>2</sup> of the push-pull systems **1d** and **2d** show enhanced catalytic currents; during CPE the Mn derivative **1d** displays the higher TON value, while the corresponding Re derivative **2d**, albeit from CV appears to be a potentially highly active catalyst, undergoes decomposition. While Mn catalysts suffer from the presence of MeOH, they appear to work better in water, which seems to react promptly with reduced Mn. In fact, even if the reduction potential of **1g** is rather negative (**Table 2**), no hydrogen is produced. This is in line with the high TON values observed for the CO<sup>2</sup> electrochemical reduction in pure water by Mn electrocatalysts supported on electrode surface (Walsh et al., 2015; Reuillard et al., 2017; Rotundo et al., 2019b). DFT calculations performed with dispersion correction agree with

the experimental data for both Mn and Re complexes. The Mn–Br bonds computed for the Mn radical anions undergo a significant elongation, around 0.2 Å, which indicate nonnegligible weakening of the Mn–Br bonds. The release of halogen is also very probably favored by the polar and coordinating solvent MeCN; indeed Br substitution by MeCN has been experimentally observed not only in the radical anion, but also in neutral Mn species (Franco et al., 2017). We demonstrated here how the appropriate choice of the electron properties of the ligands is of critical importance in the design of more effective bipyridine Mn and Re electrocatalysts for CO<sup>2</sup> reduction.

# MATERIALS AND METHODS

### General Considerations

CV and CPE experiments were performed using a Metrohm Autolab 302n potentiostat. CO and H<sup>2</sup> as CO<sup>2</sup> reduction products were detected and quantified by an Agilent 490 Micro GC. NMR spectra were recorded on a JEOL ECP 400 FT-NMR spectrometer (1H operating frequency 400 MHz) or on a JEOL ECZR 600 FT-NMR spectrometer (1H operating frequency 600 MHz) at 298 K. <sup>1</sup>H and <sup>13</sup>C chemical shifts are reported relative to TMS (δ = 0) and referenced against solvent residual peaks. IR-ATR spectra were collected on a Fourier transform Equinox 55 (Bruker) spectrophotometer equipped with an ATR device; resolution was set at 2 cm−<sup>1</sup> for all spectra. A spectral range of 400–4,000 cm−<sup>1</sup> was scanned, using KBr as a beam splitter. GC-MS spectra were obtained on a mass selective detector Agilent 5,970 B operating at an ionizing voltage of 70 eV connected to a HP 5,890 GC equipped with a HP-1 MS capillary column (25 m length, 0.25 mm I.D., 0.33µm film thickness. Elemental analyses (C, H, N) were performed on a Fisons Instruments EA-1108 CHNS-O Elemental Analyzer. ESI-MS spectra were recorded with a Thermo Advantage Max Spectrometer equipped with an ion trap analyzer and an ESI ion source.

# Synthesis and Characterization of Ligands

Flasks and all equipment used for the generation and reaction of moisture-sensitive compounds were dried by electric heat gun under N2. All commercially obtained reagents and solvents were used as received. Anhydrous DMF was purchased by Sigma Aldrich. Products were purified by preparative column chromatography on Macherey Nagel silica-gel for flash chromatography, 0.04–0.063 mm/230–400 mesh. Reactions were monitored by TLC using silica-gel on TLC-PET foils Fluka, 2–25µm, layer thickness 0.2 mm, medium pore diameter

60 Å. hpbpy (4-(2-hydroxy-phenyl)-6-phenyl-2,2′ -bipyridine) was synthesized by using the Kröhnke (1976) reaction and subsequently coupling the pyridinium iodide salt (**1**) with the corresponding hydroxy chalcone (**2**). dpbpy (4,6-diphenyl-2,2' bipyridine) was synthesized according to reported procedure (Franco et al., 2015).

Synthesis of 4,4′ -bis(trifluoromethyl)-2,2′ -bipyridine: the method of O'Donnell et al. (2016) was slightly modified to synthesize 4,4′ -bis(trifluoromethyl)-2,2′ -bipyridine first reported by Furue et al. (1992). In a degassed 20 mL screw cap vial a solution of 2-bromo-4-(trifluoromethyl)pyridine (500 mg, 2.2 mmol) in 7.5 mL of anhydrous DMF was poured and degassed for 5 min. Pd(OAc)<sup>2</sup> (25 mg, 0.11 mmol) was added later and the mixture degassed for additional 5 min. TBAI (815 mg 2.2 mmol), anhydrous K2CO<sup>3</sup> (460 mg, 3.3 mmol), and i-PrOH (0.35 mL, 4.4 mmol) were then added to the mixture, subsequently heated at 100◦C for 20 h. The heating was suspended, and the reaction mixture was filtered through a pad of celite. The filtrate was diluted in 25 mL of DCM and was washed with deionized water (3 × 25 mL). The organic phase was collected, and the water layer further extracted with 10 mL of DCM. The collected organic phase was dried with anhydrous MgSO4, filtered, and the solvent removed under reduced pressure to afford the crude solid that was purified by column chromatography on silica gel (eluent PE/AcOEt 10/1) to afford 220 mg of white solid as product. (Yield = 80%).

Synthesis of [2,2′ -bipyridine]-4,4′ -dicarbonitrile: the procedure reported by Losse et al. (2008) was followed with slight modifications. 4-Cyanopyridine (700 mg, 6.7 mmol) and 10% Pd/C (50 mg) were added into a 25 mL round-bottomed flask and five vacuum/nitrogen cycle were operated to minimize the amount of oxygen, then the flask was connected to a reflux condenser under nitrogen atmosphere. The mixture was heated to 230◦C in a sand bath in order to reflux the 4-Cyanopyridine. After 24 h the mixture was cooled to room temperature, CHCl<sup>3</sup> (15 mL) was added and the black suspension filtered through a frit. Solvent was removed under reduced pressure from the paleyellow solution obtained until the product started to crystallize. Pentane (15 mL) was then added, and the concentrated solution cooled at 4–6◦C for 16 h. The precipitate was filtered on Buchner funnel and washed with cold EtOH and dried to yield 100 mg of the product as a yellow-orange solid. (Yield 15%).

General procedures for synthesis of 2-substitued N,Ndimethylpyridin-4-amine derivatives. The previously reported procedure of Cuperly et al. (2002) was adapted according to the used electrophile. 2-(dimethylamino)-ethanol (0.8 mL, 8.0 mmol) was added to a three-necked round bottomed flask and dissolved in hexane (10 mL) under a N<sup>2</sup> atmosphere. The


TABLE 3 | TON and faradic efficiencies (η) upon CPE (applied potential E in V vs. Ag/AgCl) of solutions of manganese and rhenium complexes (0.5 mM) in 0.1 M TBAPF6/MeCN in the presence of Brønsted acids (5%v).

solution was cooled at −5 ◦C, BuLi (2.5 M, 6.4 mL, 16.0 mmol) was added dropwise for 10 min and the resulting mixture stirred for 40 min at −5 ◦C. 4-DMAP (488 mg, 4.0 mmol) was then added and stirring continued for additional 60 min at 0◦C, then the reaction medium was cooled at −78◦C and the solution of the appropriate electrophile (10.0 mmol) was added dropwise by mean of a dropping funnel with pressure balance. Once the addition of the electrophile was completed the temperature was allowed to raise to 0◦C (1.5 h) and the reaction quenched with deionized water at this temperature.

Synthesis of N,N-dimethyl-2-(tributylstannyl)pyridin-4-amine. 2-(dimethylamino)ethanol (0.8 mL, 8.0 mmol) in hexane (10 mL), BuLi (2.5 M, 6.4 mL, 16.0 mmol), and 4-DMAP (0.488 g, 4.0 mmol) were reacted as previously described then a solution of Bu3SnCl (3.255 g, 10.0 mmol) in 15 mL of hexane was added dropwise for 20 min. The reaction was quenched with deionized water (15 mL) then the aqueous phase was extracted with DCM (2 × 20 mL) and AcOEt (2 × 20 mL). The collected organic phase was dried with anhydrous Na2SO4, filtered, and the solvent removed under reduced pressure to afford a crude orange oil, that was used without further purification in the cross-coupling reactions (NMR-calculated yield = 69%).

Synthesis of 2-iodo-N,N-dimethylpyridin-4-amine: 2- (dimethylamino)ethanol (1.6 mL, 16.0 mmol) in hexane (25 mL), BuLi (2.5 M, 12.8 mL, 32.0 mmol), and 4-DMAP (0.980 g, 8.0 mmol) were reacted as previously described, then a solution of resublimed I<sup>2</sup> (5.080 g, 20.0 mmol) in 50 mL of freshly distilled Et2O was added dropwise for 35 min. The reaction was quenched with a saturated solution of Na2S2O<sup>3</sup> (25 mL) and stirred for additional 20 min at 0◦C. The organic phase was separated and washed again with Na2S2O<sup>3</sup> solution (2×15 mL) and brine (2 × 15 mL). The collected organic phase was dried with anhydrous Na2SO4, filtered, and the solvent removed under reduced pressure to afford a crude brown solid that was purified by column chromatography on silica gel (eluent: PE/AcOEt 5/5) to afford 1.683 g of white solid as product (Yield = 86%).

General Procedure for Stille cross-coupling of N,N-dimethyl-2-(tributylstannyl)pyridin-4-amine with 2-halopyridine. In a degassed 50 ml screw cap vial freshly distilled toluene (25 mL) was added and degassed for 10 min then Pd(OAc)<sup>2</sup> (22 mg, 0.1 mmol) and PPh<sup>3</sup> (52 mg, 0.2 mmol) were added in one portion. The resulting mixture was stirred and degassed until toning of the solution to red. Subsequently, N,N-dimethyl-2- (tributylstannyl)pyridin-4-amine (370 mg, 0.9 mmol), LiI (40 mg, 0.3 mmol), and CuI (57 mg, 0.3 mmol) were added to the mixture, then degassed for additional 5 min. Finally, the appropriate 2 halopyridine (1.1 mmol) was poured in the reaction medium that was kept under N<sup>2</sup> atmosphere and heated to reflux for 16 h. After being cooled at room temperature, the resulting mixture was diluted with EtOAc (30 mL) and washed with a NH4OH solution (10 M) until the water layer did not turn blue any more, indicating that all the copper had been extracted. The collected organic phase was filtered on a pad of celite, diluted with DCM (30 mL) and dried with anhydrous Na2SO4. Then the solvents were removed under reduced pressure to afford the crude solid subsequently purified to yield the desired product.

Synthesis of N<sup>4</sup> ,N<sup>4</sup> ,N<sup>4</sup> ',N<sup>4</sup> '-tetramethyl-[2,2′ -bipyridine]-4,4′ diamine: Following the general procedure**,** N,N-dimethyl-2- (tributylstannyl)pyridin-4-amine (370 mg, 0.9 mmol) and 2 iodo-N,N-dimethylpyridin-4-amine (270 mg, 1.1 mmol) in 25 mL of freshly distilled toluene were reacted to afford a crude yellowish solid that was suspended in 25 mL of Et2O and vigorously stirred for 30 min. The desired product was separated by filtration to yield 115 mg of a brown powder (Yield = 53 %).

Synthesis of N,N-dimethyl-4′ -(trifluoromethyl)-[2,2′ bipyridin]-4-amine: following the general procedure, N,N-dimethyl-2-(tributylstannyl)pyridin-4-amine (370 mg, 0.9 mmol) and commercially available 2-bromo-4- (trifluoromethyl)pyridine (250 mg, 1.1 mmol) in 25 mL of freshly distilled toluene were reacted to afford a crude brown solid that was purified by column chromatography on silica gel (eluent: MeOH/DCM 5/95) to yield 169 mg of the desired product as a brown powder (Yield = 70%).

Synthesis of N-(2-pyridylacetyl)-pyridinium iodide (**1**): in a three-necked flask 82.5 mmol of 2-acetylpyridine and 90.5 mmol of iodine are dissolved in 100 ml of pyridine. The solution is refluxed for 3 h. The shiny black precipitate is then filtered and washed with diethyl ether (20 ml). The solid is recrystallized in hot ethanol: fine-scaled golden crystals are obtained (overall yield: 54%).

Synthesis of (E)-3-(2-hydroxyphenyl)-1-phenyl-2-propen-1-one (**2**): is synthesized by Claisen-Schmidt condensation of salicylaldehyde (50 mmol) with acetophenone (40 mmol) dissolved in 50 mL of ethanol (Tatsuzaki et al., 2006; Yin et al., 2012). The solution is vigorously stirred and 10 mL of KOH 40% (w/w) is dropwise added. The mixture is then heated to 60◦C for 2–4 h until the disappearance of the starting reagents, monitored by TLC (EtAc: hexane 1:4). Once the reaction is completed, the suspension is poured into cold distilled water and acidified with HCl 2 M (final pH 2–3). The resulting precipitate is collected and washed once with water and then with cyclohexane (yield: 70%).

Synthesis of 4-(2-hydroxy-phenyl)-6-phenyl-2,2′ -bipyridine (hpbpy): precursors **1** (4.8 mmol) and **2** (4.8 mmol) are added to a three-necked flask with an excess of ammonium acetate (48 mmol) and glacial acetic acid (4 mL). The solution is refluxed for 3 h. The glacial acetic acid is removed, and the residue dissolved in methanol. The yellowish product crystallizes on cooling.

Synthesis of 4,6-diphenyl-2,2′ -bipyridine (dpbpy): is already reported in literature; NMR data are in agreement with previously published data (Cave and Raston, 2001).

# Synthesis of Mn Complexes

[Mn(CO)5Br] (0.100 mmol, 1 equiv) and the corresponding bipyridyl ligand (0.101 mmol, 1,01 equiv) were dissolved in sealed flasks containing 5 mL of diethyl ether (anhydrous for complex **1f**), and heated in a Biotage Initiator<sup>+</sup> microwave reactor. When the microwave vial has been inserted into the microwave cavity and the cavity lid has been closed, high-frequency microwaves (2.45 GHz), generated by the magnetron, heat the reaction mixture at a constant temperature of 75◦C for 15 min. The reaction mixtures were cooled to room temperature and the products centrifuged and washed once with diethyl ether. Yields of reactions: (**1a:** 75%, **1b:** 84%**, 1c:** 70%**, 1d**: 81%, **1e:** 68%, **1f:** 65%, **1g:** 78%).

## Synthesis of Re Complexes

[Re(CO)5Cl] (0.100 mmol, 1 equiv) and the corresponding bipyridyl ligand (0.101 mmol, 1.01 equiv) were dissolved in sealed flasks containing 5 mL of toluene (anhydrous for complex **2f**), and heated in the Biotage microwave reactor at a constant temperature of 130◦C for an hour. After cooling of the reaction mixtures to room temperature, petroleum ether was added to precipitate the products, which were then centrifuged, filtered and washed once with cold diethyl ether. Yields of reactions: (**2a:** 70%, **2b:** 82%**, 2c:** 67%**, 2d**: 85%, **2e:** 65%, **2f:** 65%, **2g:** 75%).

### Elemental Analysis of the Complexes

The samples for microanalyses were dried in vacuum to constant weight (20◦C, ca. 0.1 Torr). Elemental analysis (C, H, N) was performed in-house with a Fisons Instruments 1108 CHNS-O Elemental Analyzer.

Anal. Calcd.(%) for **1a (C15H6BrF6N2O3Mn)**: C 35.25; H, 1.18; N, 5.48. Found: C 35.49, H 1.25, N, 5.36.

Anal. Calcd.(%) for **1b (C15H6BrF6N2O3Mn)**: C, 35.25; H, 1.18; N, 5.48. Found: C 35.55, H 1.09, N 5.57.

Anal. Calcd.(%) for **1c (C15H6BrN4O3Mn)**: C, 42.38; H, 1.42; N, 13.18. Found: C 42.62, H, 1.50, N, 13.07.

Anal. Calcd.(%) for **1d (C16H12BrF3N3O3Mn)**: C, 39.53; H, 2.49; N, 8.64. Found: C, 39.41, H, 2.61, N, 8.75.

Anal. Calcd.(%) for **1e (C25H16BrN2O3Mn)**: C, 56.95; H, 3.06; N, 5.31. Found: C, 57.13, H, 2.96, N, 5.45.

Anal. Calcd.(%) for **1f (C25H16BrN2O4Mn)**: C, 55.27; H, 2.97; N, 5.16. Found: C, 55.15, H, 3.14, N, 4.95.

Anal. Calcd.(%) for **1g (C17H17BrN4O3Mn)**: C, 44.37; H, 3.72; N, 12.17. Found: C, 44.51, H, 3.82, N, 12.04

Anal. Calcd.(%) for **2a (C15H6ClF6N2O3Re)**: C, 30.13; H, 1.01, N, 4.69. Found: C, 30.20, H, 1.12, N, 4.35.

Anal. Calcd.(%) for **2b (C15H6ClF6N2O3Re)**: C, 30.13; H, 1.01, N, 4.69. Found: C, 29.95, H, 1.16, N, 4.78.

Anal. Calcd.(%) for **2c (C15H6ClN4O3Re)**: C, 35.20; H, 1.18; N, 10.95. Found: C, 35.34, H, 1.26, N, 10.77.

Anal. Calcd.(%) for **2d (C16H12ClF3N3O3Re)**: C, 33.54; H, 2.11; N, 7.33. Found: C, 33.43, H, 2.26, N, 7.48.

Anal. Calcd.(%) for **2e (C25H16ClN2O3Re)**: C, 48.90; H, 2.63; N, 4.56. Found: C, 49.06, H, 2.75, N, 4.43.

Anal. Calcd.(%) for **2f (C25H16ClN2O4Re)**: C, 47.66; H, 2.56; N, 4.45. Found: C, 47.53, H, 2.78, N, 4.36.

Anal. Calcd.(%) for **2g (C17H17ClN4O3Re)**: C, 37.33; H, 3.13; N, 10.24. Found: C, 37.25, H, 3.22, N, 10.16.

### Single-Crystal X-Ray Diffraction

The single-crystal data were collected with a Gemini R Ultra diffractometer with graphite-monochromated Mo-Kα radiation (λ = 0.71073) by the ω-scan method. The cell parameters were retrieved with the CrysAlisPro (Agilent, 2015) software, and the same program was used to perform data reduction with corrections for Lorenz and polarizing effects. Scaling and absorption corrections were applied through the CrysAlisPro1 multiscan technique. The structures of complex **2e** (both structure A and B) were solved with direct methods, while in the case of **2f** a meaningful initial guess for electron density was obtained only with Patterson Function by using SHELXS-14 (Sheldrick, 2008, 2015). All the structures were refined with fullmatrix least-squares techniques on F<sup>2</sup> with SHELXL-14 (Macrae et al., 2006) using the program Olex2 (Dolomanov et al., 2009). All non-hydrogen atoms were refined anisotropically. Hydrogen atoms were calculated and riding on the corresponding bonded atoms. The graphic of the crystal structures was generated using Mercury 3.9 (Macrae et al., 2006). CCDC codes 1891407– 1891409 contain the supplementary crystallographic data for **2e** (structure A), **2e** (structure B) and **2f**. These data can be obtained free of charge via https://www.ccdc.cam.ac.uk/conts/retrieving. html, or from the Cambridge Crystallographic Data Center, 12 Rotundo et al. Electronic Effects of Substituents

Union Road, Cambridge CB2 1EZ, UK; fax: (+44) 1223-336-033; or e-mail: deposit@ccdc.cam.ac.uk.

### CV and CPE Experiments

Acetonitrile used for the experiments was freshly distilled over calcium hydride and purged with Ar before use. 0.5–1 mM solutions of the complexes were prepared with tetrabutylammonium hexafluorophosphate (TBAPF6, Sigma-Aldrich, 98%) as supporting electrolyte (0.1 M). A single-compartment cell was employed for CV measurements, equipped with working a glassy carbon electrode (GCE, Ø = 1 mm), alongside a Pt counter electrode and a Ag/AgCl (KCl 3 M) reference electrode. The Ar- and CO2-saturated conditions were achieved by purging gases for 5 min before each potential sweep. A double-compartment H-type cell was used for CPE measurements, thus allowing to separate through a glass frit the anodic compartment from the cathodic one (a Pt wire is placed as counter electrode). A glassy carbon rod is employed as working electrode jointly with the Ag/AgCl reference electrode. A controlled and constant flow of CO<sup>2</sup> (50 mL min−<sup>1</sup> ) was maintained during the CPE measurements by means of a Smart Trak 100 (Sierra) flow controller. Under these experimental conditions, the redox couple (Fc+/Fc) is located at E1/<sup>2</sup> = 0.35 V.

### Quantitative Analysis of CO<sup>2</sup> Reduction Products

µGC measurements were used to detect CO and H2. Two modules equipped with CP-Molsieve 5 Å columns were kept at 105◦C and at a pressure of 30 and 28 psi, with a thermal conductivity detector. The carrier gases were Ar for H<sup>2</sup> and He for CO detection, respectively. The backflush vent option time was set to 7 s. The gas inside the measurement cell was sampled for 30 s every 3 min to fill the Micro GC 10 µL sample loop, and eventually 500 nL was injected into the column for the analyses. Instrument calibration was carried out measuring two different certified standards of CO and H<sup>2</sup> in Ar matrix (Rivoira). Formate production was assessed by NMR spectroscopy.

### Computational Details

All the calculations were performed by the Gaussian 16 Revision B.01 (G16) program package (Frisch et al., 2016), employing density functional theory (DFT). Calculations were run using

### REFERENCES

Agilent (2015). CrysAlis PRO. Yarnton: Agilent Technologies Ltd.


the Becke three-parameter hybrid functional (Becke, 1993), and the Lee–Yang–Parr gradient-corrected correlation functional (B3LYP) (Lee et al., 1988). Dispersion effects were added as semiempirical corrections with Becke–Johnson damping approach (GD3BJ) (Grimme et al., 2010, 2011). The solvent effect was included using the conductor-like polarizable continuum model (CPCM) with acetonitrile as solvent (Miertus et al., 1981). The def2TZVP basis set and effective core potential were used for the Mn, Br, and Cl atoms and the def2-SVP basis set was used for all the other atoms (Weigend and Ahlrichs, 2005). Unrestricted open-shell calculations were performed on the radical anions. Geometry optimizations were carried out without any symmetry constraints. The nature of the stationary points in the potential energy hypersurface was characterized by using harmonic vibrational frequency calculations. No imaginary frequencies were found, thus indicating we had located the minima on the potential-energy surfaces. Molecular-graphic images were produced by using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (Pettersen et al., 2004).

### AUTHOR CONTRIBUTIONS

RG and CN as corresponding authors wrote and revised the manuscript. EP performed Crystal X-ray structures. EA, AD, and PQ made the synthesis of the ligands. LR, RR, and LN synthesized the organometallic catalysts and performed the electrochemical and GC measurements. CG and CN made the DFT calculations.

### ACKNOWLEDGMENTS

We gratefully thank R. Buscaino (Department of Chemistry, Università degli studi di Torino) for the ESI-MS spectra of the complexes. CRT (Fondazione CRT, ref. 2017.0812) project is acknowledged.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00417/full#supplementary-material


Eur. J. Inorg. Chem. 2008, 4448–4452. doi: 10.1002/ejic.2007 01304


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Rotundo, Azzi, Deagostino, Garino, Nencini, Priola, Quagliotto, Rocca, Gobetto and Nervi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Development of Visible-Light Driven Cu(I) Complex Photosensitizers for Photocatalytic CO<sup>2</sup> Reduction

Hiroyuki Takeda\* † , Yu Monma, Haruki Sugiyama† , Hidehiro Uekusa and Osamu Ishitani

Department of Chemistry, School of Science, Tokyo Institute of Technology, Tokyo, Japan

### Edited by:

Hitoshi Ishida, Kitasato University, Japan

### Reviewed by:

Helene Serier-Brault, UMR6502 Institut des Matériaux Jean Rouxel (IMN), France Yang Yang, University of Central Florida, United States

### \*Correspondence:

Hiroyuki Takeda takedah@chem.titech.ac.jp; takedah@recap.osaka-cu.ac.jp

### †Present Address:

Hiroyuki Takeda, Research Center for Artificial Photosynthesis, Osaka City University, Osaka, Japan Haruki Sugiyama, Research and Education Center for Natural Sciences, Keio University, Yokohama, Japan

### Specialty section:

This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry

Received: 08 March 2019 Accepted: 21 May 2019 Published: 06 June 2019

### Citation:

Takeda H, Monma Y, Sugiyama H, Uekusa H and Ishitani O (2019) Development of Visible-Light Driven Cu(I) Complex Photosensitizers for Photocatalytic CO2 Reduction. Front. Chem. 7:418. doi: 10.3389/fchem.2019.00418 The visible-light responsive Cu(I)-complex photosensitizers were developed by introducing various aromatic substituents at the 4,7-positions of a 2,9-dimethyl-1,10-phenanthroline (dmp) ligand in a heteroleptic Cu<sup>I</sup> (dmp)(DPEphos)+-type complexes (DPEphos = [2-(diphenylphosphino)phenyl]ether) for photocatalytic CO<sup>2</sup> reduction. Introducing biphenyl groups (Bp-) on the dmp ligand enhanced the molar extinction coefficient (ε) of the metal-to-ligand charge transfer (MLCT) band in the visible region (ε = 7,500 M−1cm−<sup>1</sup> ) compared to that of the phenyl (Ph-)-containing analog (ε = 5,700 M−1cm−<sup>1</sup> at λmax = 388 nm). However, introducing 4-R-Ph- groups (R = the electron-withdrawing groups NC-, or NO2-) led to a red shift in the band to λmax = 390, 400, and 401 nm, respectively. Single-crystal X-ray analysis showed the Ph- groups were twisted because of the steric repulsion between the 2,6-protons of the Ph- groups and 5,6-protons of the dmp ligand. The result strongly indicated that the π-conjugation effect of the 4-R-Ph- groups is so weak that the lowering of the energy of the dmp π <sup>∗</sup> orbitals is small. However, when 4-R-ph- was substituted by a 5-membered heterorings, there was a larger red shift, leading to an increase in the ε value of the MLCT absorption band. Thus, the substitution to 2-benzofuranyl- groups resulted in visible-light absorption up to 500 nm and a shoulder peak at around 420 nm (ε = 12,300 M−1cm−<sup>1</sup> ) due to the expansion of π-conjugation over the substituted dmp ligand. The photocatalytic reaction for CO<sup>2</sup> reduction was tested using the obtained Cu<sup>I</sup> complexes as photosensitizers in the presence of a Fe(dmp)2(NCS)<sup>2</sup> catalyst and 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole as a sacrificial reductant, which showed improved CO generation.

Keywords: Cu(I) diimine complex, CO2 reduction photocatalyst, redox photosensitizer, visible-light absorption, emission

# INTRODUCTION

As one of the most important components of artificial photosynthesis, photocatalytic CO<sup>2</sup> reduction is attracting much attention. Although metal complexes are key players in CO<sup>2</sup> reduction due to their promising photosensitizing and/or catalytic ability, the metals used are often limited to low abundance ones, such as Ru, Re, Os, or Ir (Yamazaki et al., 2015). Thus, substituting these metals with more abundant elements is now being widely focused on in many groups, using the first row transition metals such as Mn, Fe, Co, and Ni to produce a multi-electron catalyst for CO<sup>2</sup> reduction (Takeda et al., 2017). However, such the attempts, especially to produce a redox

**77**

photosensitizer (which functions as a light absorber to transfer an excited electron in its excited state to the catalyst), are limited; the excited state deactivates quickly because of low lying d-d excited state.

Thus, the heteroleptic Cu<sup>I</sup> phenanthroline complexes, such as [Cu<sup>I</sup> (dmp)(P)2] <sup>+</sup> (dmp = 2,9-dimethyl-1,10-phenanthroline, P = phosphine ligand), are gaining popularity, because of their long lifetimes, showing strong metal-to-ligand (MLCT) excitedstate emission even in a solution at room temperature due to the Cu<sup>I</sup> center's d <sup>10</sup> configuration (Ruthkosky et al., 1998). Although homoleptic-type [Cu<sup>I</sup> (dmp)2] <sup>+</sup> has the advantage as a photosensitizer in terms of visible-light utilization because it can absorb longer wavelength light (Khnayzer et al., 2013), in the excited state, its oxidation power is less and lifetime shorter than those of a heteroleptic complex. Thus, heteroleptic-type of Cu complexes are now being used as redox photosensitizers to construct many photocatalytic systems (Lazorski and Castellano, 2014). Since these Cu complexes had previously been reported to have excellent emissive properties by McMillin et al. (Breddels et al., 1982), recent developments by Beller et al. on the catalytic H<sup>2</sup> evolution reaction using such complexes (Luo et al., 2013; Mejía et al., 2013; Rosas-Hernández et al., 2017) have opened up the way in this area.

However, when using heteroleptic Cu complexes as redox photosensitizers in photocatalytic reactions, some issues remain, such as their stability and light-absorption in the visible region. The first issue, stability, has previously been tackled by our group, where the stability of the Cu<sup>I</sup> complex photosensitizer was improved by connecting phenanthroline phosphine ligands together using –C4H8– alkyl groups (Takeda et al., 2016). Because the tetrahedral coordination around the Cu<sup>I</sup> center was maintained in this molecule, the complex has a long lifetime in the excited state, exhibiting strong emission. Thus, the resulting Cu<sup>I</sup> dimer complexes were much more stable, not only against thermal ligand exchanges in the ground state (Kaeser et al., 2013; Lennox et al., 2016) but also against "exciplex" deactivation of the excited state, even when CH3CN was used as a coordinating solvent (McMillin et al., 1985) and a Cu<sup>0</sup> metal particle formed via ligand dissociation in the one-electron reduced state (Eggleston et al., 1997). Thus, utilizing this Cu<sup>I</sup> complex as a redox photosensitizer in the photocatalytic CO<sup>2</sup> reduction, we could obtain the best photocatalytic performances for CO<sup>2</sup> reduction and clarified the photosensitizing scheme, which forms the corresponding one-electron reduced species through reductive quenching by a reductant such as BIH (1,3 dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole) in the excited state, donating the added electrons to CO<sup>2</sup> reduction catalyst (Takeda et al., 2018).

So far, many groups have attempted to red shift the <sup>1</sup>MLCT absorption band by expanding the dmp ligand's π-conjugation. These attempts are reasonable because the lowest excited state of the Cu<sup>I</sup> complexes can be mainly attributed to the MLCT state, in which a charge on the Cu<sup>I</sup> center transfers to the phenanthroline ligand's low-lying π <sup>∗</sup> orbitals. Tsubomura et al. introduced phenyl groups at the 4,7-positions of dmp, which is known neocuproine, to produce a ligand called bathocuproine (bcp), enhancing the molar extinction coefficient of the resulting Cu<sup>I</sup> complex, thus causing a red shift in the absorption maximum of the MLCT band (Tsubomura et al., 2015). The Cu<sup>I</sup> complexes having a pyrazine- or phenazine-fused dmp ligand have been also reported. However, the resulting Cu complexes showed absorption maxima at around 380 nm (Xu et al., 2015; Heberle et al., 2017). The same group also reported introducing (thiophen-2-yl)vinyl groups at the 2,9-positions of the phenanthroline, but the MLCT absorption band maximum of the Cu complex was still only observed at 388 nm (Chen et al., 2017). Although introducing 4-benzoic acid or thiophene-3 carboxylic acid at the 4,7-positions shifts the absorption band to a longer wavelength, the absorption maxima are limited to 394 and 392 nm, respectively (Chen et al., 2017). Introducing sulfonate groups at the 5,6-positions of bcp also limits the absorption maximum to 390 nm (Rockstroh et al., 2014). Recently, Giereth et al. reported a Cu complex having anthracene-fused dmp, which shows absorption at a longer wavelength at around 430 nm (Giereth et al., 2019). McCullough et al. also reported a unique Cu complex incorporating 2,2′ -biquinoline as an α-diimine and applied it as a photocatalytic system for H<sup>2</sup> evolution (McCullough et al., 2018). This type of complex showed a surprising red shift in its maximum to 455 nm (Zhang et al., 2007). However, the first-reduction potential of these Cu complexes underwent a positive shift from −2 V for the dmp complex to −1.19 V vs. Fc+/Fc in CH3CN for the anthracenefused dmp complex and −1.69 V vs. Fc+/Fc in CH2Cl<sup>2</sup> for the biquinoline complex, that are more positive than that of the common redox-photosensitizer Ru(dmb)2<sup>+</sup> 3 (dmb = 4,4′ dimethyl-2,2′ -bipyridine). Thus, further study of these types of Cu<sup>I</sup> complexes is required so that they can be used as visible-light absorbing redox-photosensitizers.

In this study, Cu<sup>I</sup> complexes are synthesized with new bcp-based phenanthroline ligands, with electron-withdrawing groups on the phenyl groups, or various different aromatic groups such as heterorings, instead of the phenyl groups of the bcp ligand. The aim is improving these systems' visiblelight absorption. The Cu<sup>I</sup> complexes' UV-Vis absorption spectra, photophysical properties, and electrochemical properties were examined, providing evidence for a significant red shift in the MLCT absorption band. The redox-photosensitizing ability for photocatalytic CO<sup>2</sup> reduction of these Cu<sup>I</sup> complexes was also tested using Fe(dmp)2(NCS)<sup>2</sup> as a catalyst in the presence of a reductant.

### RESULTS AND DISCUSSION

The structures of newly synthesized Cu<sup>I</sup> complexes' structures are shown in **Figure 1** alongside the reported Cu<sup>I</sup> complexes **Cu(H)** (Cuttell et al., 2002) and **Cu(ph)** (Luo et al., 2013). Two types of substituents were introduced at the 4,7-positions of the dmp ligand, (a) 4-substituted phenyl groups (**Figure 1A**) and (b) heteroring aryl groups (**Figure 1B**). In the type (a) substituents, **Cu(Bph)** is an expanded form of **Cu(ph)** with an increased number of the phenyl groups, while **Cu(NCph)** and **Cu(NO2ph)** have π-electron withdrawing cyano and nitro substituents on the phenyl groups, respectively. In the (b) substituents, the

ligands have aryl groups that are smaller in size, i.e., 5-membered heterorings containing O or S atoms at 2- or 3-position, than the phenyl groups adjacent to the dmp ligand. Because S and O atoms have different electronegativities of 3.44 and 2.58, respectively, and van der Waals radii (Bondi, 1964) of 1.52 and 1.80 Å, respectively, the electronic properties of the Cu<sup>I</sup> complexes should be different.

The Cu<sup>I</sup> complexes' UV-Vis absorption spectra in a noncoordinating solvent, such as CH2Cl2, are shown in **Figure 2**, and the data are summarized in **Table 1**. As a representative system, **Cu(ph)** exhibited a <sup>1</sup>MLCT transition band at 350–500 nm that was moderate in intensity (molar extinction coefficient (ε) ∼ 6000 M−<sup>1</sup> s −1 ) and an intraligand π-π ∗ transition band in the shorter-wavelength range with stronger intensity (ε ∼50000 M−<sup>1</sup> s −1 ), as shown in **Figures 2A–C** (dotted line). Because the <sup>1</sup>MLCT transition is the lowest in energy from the spectrum, the lowest excited state of this complex is strongly indicated to be the MLCT state. Although the π-π ∗ excited state which is higher in energy can proceed thermal relaxation to populate the lowest state, direct excitation of the lowest MLCT transition in the visible region can be assumed to be the most significant in a redox-photosensitizing reaction. Thus, we focused <sup>1</sup>MLCT band of the Cu<sup>I</sup> complexes, referring the π-π <sup>∗</sup> band as an indicator of π-system expansion over the aryl-substituted diimine ligands.

In the case of **Cu(Bph)** (**Figure 2A**, red), the intensities of both the <sup>1</sup>MLCT and π-π <sup>∗</sup> bands were enhanced to 7500 and ∼70000 M−<sup>1</sup> cm−<sup>1</sup> , respectively, with no shift observed for the <sup>1</sup>MLCT band and a red-shift observed for the π-π <sup>∗</sup> band. Reports indicate that, for the similar types of heteroleptic Cu<sup>I</sup> complexes, adding phenyl groups at the 4,7-positions of the dmp ligand, as in **Cu(ph)**, enhances the <sup>1</sup>MLCT absorption band intensity. This enhancement results in a red shifts of the band to λabs = 388 nm (ε = 5700 M−<sup>1</sup> cm−<sup>1</sup> ) from that of **Cu(H)** (λabs = 380 nm, ε = 2700 M−<sup>1</sup> cm−<sup>1</sup> ) due to the enhancement in the corresponding transition dipole moment and stabilization of the π <sup>∗</sup> orbitals in the dmp ligand (Tsubomura et al., 2015). In our case, **Cu(Bph)** also retained this tendency, where the phenyl groups were modified from those in **Cu(ph)**, but to a much lesser degree.

Introducing NC- groups or NO2- groups at the 4-position of the two phenyl groups, as in **Cu(NCph)** and **Cu(NO2ph)** (**Figure 2A**, blue and purple), respectively, resulted in significant red shift in the <sup>1</sup>MLCT absorption bands to 400 and 401 nm, with almost no changes observed in their intensities. Because these substituents are known as π-electron-withdrawing substituents, these red shifts indicate that the π ∗ -orbitals on the dmp ligand were stabilized by these substituents through the π-systems containing the phenyl groups.

Such the red shifts of the <sup>1</sup>MLCT bands were also observed for the Cu<sup>I</sup> complexes featuring aryl substituents (**Figure 2B**). In the case of **Cu(3fu)** and **Cu(3th)** (**Figure 2B**, black and red, respectively), both <sup>1</sup>MLCT bands underwent slight red shifts to 390 and 391 nm, respectively, while retaining their intensities, compared to **Cu(ph)**. This result indicates the different electronegativities of the O and S atoms made only negligible effect on the absorption spectra. **Cu(3Bzth)** also underwent a further red shift of the <sup>1</sup>MLCT band than the other complexes in the form of a shoulder at 395 nm (**Figure 2B**,

yellow). In this case, the π-π ∗ absorption band was also red shifted to 330 nm, indicating destabilized π orbitals over the substituted dmp ligand due to the expansion of the π-systems.

Compared to these aryl groups, substituting the 2-positions of the 5-membered rings with heteroatoms resulted in much larger red shifts in both the <sup>1</sup>MLCT and π-π ∗ absorption bands. For **Cu(2th)** (**Figure 2C**, green), **Cu(2Bzth)**, and **Cu(Bz2fu)** (**Figure 2C**, blue and purple, respectively) gradual shifts in the <sup>1</sup>MLCT bands of these complexes to longer wavelengths (400, 410, and 420 nm, respectively) occurred alongside an increase in the ε values to 6800, 8200, and 12300 M−<sup>1</sup> cm−<sup>1</sup> , respectively. Additionally, the π-π <sup>∗</sup> bands were also red shifted largely up to 350 nm. Thus, the π-conjugation over the substituted dmp ligand was effectively observed, especially for the complex featuring 2 benzofuryl substituents [**Cu(Bzfu)**]. These trends were almost similar in CH3CN solutions (**Figure S1** and **Table 2**), although undesired generation of the corresponding homoleptic-type Cu<sup>I</sup> complexes with <sup>1</sup>MLCT absorption bands at over 450 nm obscured the original heteroleptic complexes.

Then, the single-crystal structures of these Cu<sup>I</sup> complexes were determined to obtain information about the planarity of the substituted dmp ligand, which determines the overlap of the π <sup>∗</sup> orbitals between the aryl substituents and dmp. Here, we could obtain four types of single crystals of Cu<sup>I</sup> complexes, **Cu(ph)**, **Cu(NCph)**, **Cu(2Bzth)**, and **Cu(2Bzfu)**. The structures of these are shown in **Figure 3** and full analytical details are in the **Supplementary Information** (**Figures S44–47** and **Tables S2–6**). All of these complexes have tetrahedral Cu<sup>I</sup> center with an almost right angle between NCuN and PCuP planes (88.6, 89.0, 88.6, and 86.8◦ , respectively), a configuration required for these types of Cu<sup>I</sup> complexes to exhibit strong emission. There are no remarkable differences in the complexes' residual parts, other than in the substituted dmp ligand, as follows.

The twisted angles between the phen and ph planes in the **Cu(ph)** crystal were found to be 60.2 and 53.9◦ . Because the H atoms at the 5,6-positions of the phen ligand and the 2,6 positions of the ph groups are in close contact, these ph groups cannot lie in the same plane as that of the phen ligand. Actually, the shortest distances between these H atoms were measured as 2.49 and 2.54 Å. These values are closing to the limitation of van der Waals radius of the H atom, at 1.2 Å (Bondi, 1964).

TABLE 1 | Photophysical Properties of the Cu<sup>I</sup> Complexes in CH2Cl2..


<sup>a</sup>λabs, <sup>1</sup>MLCT absorption maxima;

b ε, molar extinction coefficients;

<sup>c</sup>λem, emission maxima;

<sup>d</sup>Φem, emission quantum yield;

e τem, emission lifetimes;

<sup>f</sup> out of range for accurate determination;

c−e samples degassed using the freeze–pump–thaw method.

Thus, the steric repulsion between these H atoms hinders the planarity of the ph-phen plane, resulting in a small expansion of the π-conjugation over the substituted dmp ligand. Even when π-electron withdrawing NC- groups were introduced at the 4-position of ph as for **Cu(NCph)**, the twist angles and shortest distances between the H atoms were 49.4 and 53.9◦ , and 2.50 and 2.41 Å, respectively. Thus, the steric hindrance of the H atoms should still play a role in hindering the NCph-phen bonds' planarity.

Complexes with 5-membered, heteroring substituents exhibited decreased twist angles. In the case of **Cu(2Bzth)**, the angles were 51.1 and 46.3◦ , maintaining the distances between the H atoms of 2.45 and 2.32 Å. However, the angles in **Cu(2Bzfu)** drastically decreased to 32.2 and 7.1◦ , shortening the H–H distance to 2.13 Å and changing the direction of O atom in another 2-benzofuryl substituent to neighbor to the 5-H atom of phen with the shorter distance of 2.17 Å. This means the substituents were able to lie in the same plane as the phen moiety. Because the van der Waals radius of the O atom is 1.52 Å, which is smaller than the 1.80 Å radius of the S atom (Bondi, 1964), the steric hindrance of the planarity over the substituted dmp ligand is not strong enough to result in a planner structure. Thus, the π-systems over the substituted dmp should be the strongest for **Cu(2Bzfu)** among the Cu<sup>I</sup> complexes in this work. This structural accessibility of a planner structure should lower the π <sup>∗</sup> orbitals over the substituted dmp ligand, resulting in a red shift in the <sup>1</sup>MLCT absorption band (**Figure 2C**, purple).

**Figure 4** shows the emission spectra and time-dependences of the emission intensity, and the data are summarized in **Table 1**. All of the Cu<sup>I</sup> complexes showed strong emission, apart from **Cu(NO2ph)**. In the case of **Cu(Bph)** (**Figures 4A,B**, red), the spectrum was almost the same as that of **Cu(ph)**, with a maximum (λem) at 577 nm; however, its quantum yield (8r) of 0.60 was greater than that of **Cu(ph)**, which was 0.52. This is TABLE 2 | Photophysical Properties of the Cu<sup>I</sup> Complexes in CH3CN.


<sup>a</sup>λabs, <sup>1</sup>MLCT absorption maxima;

b ε, molar extinction coefficients;

<sup>c</sup>λem, emission maxima;

<sup>d</sup>Φem, emission quantum yield;

e τem, emission lifetimes; <sup>f</sup> minimum values considering partial decomposition to the corresponding homoleptic-type Cu<sup>I</sup> complexes (see the spectra in Figure S1);

c−e samples degassed using the freeze–pump–thaw method.

reasonable when considering the increased oscillator strength for the MLCT transition, as observed in the increased ε value of the <sup>1</sup>MLCT absorption band in the UV-Vis spectrum. Thus, the radiative transition rate (kr) of the excited state should be enhanced, increasing the 8<sup>r</sup> because this value is defined in the context of k<sup>r</sup> in all of the deactivation paths of the excited state (k<sup>r</sup> + knr in this case). The increased, excitedstate lifetime [τ em = 1/(k<sup>r</sup> + knr)] for **Cu(Bph)** of 22.5 µs from that of **Cu(ph)** (19.4 µs) indicates that the knr value of **Cu(Bph)** is lower than that of **Cu(ph)**. This should be due to the increased mixing of the π-π ∗ excited state in the lowest MLCT state, which is indicated by a red shift in the intraligand <sup>1</sup>π-π ∗ absorption band close to the <sup>1</sup>MLCT band. Because these Cu<sup>I</sup> complexes exhibit delayed fluorescence, the actual deactivation rates of k<sup>r</sup> and knr could not be obtained in these measurements.

In the cases of **Cu(NCph)** and **Cu(NO2ph)** (**Figures 4A,B**, blue and purple, respectively), the emission maxima were red shifted to 609 and 640 nm, respectively, manifesting as red shifts in the corresponding <sup>1</sup>MLCT absorption bands. In both cases, the 8<sup>r</sup> and τ em values were significantly decreased [8<sup>r</sup> = 0.16 and τ em = 5.6 µs for **Cu(NCph)**], and **Cu(NO2ph)** showed almost no emission (8<sup>r</sup> < 0.00, τ em = 0.01 µs). These can be reasonably explained by the energy-gap law similarly to the substituent effects on the emission behaviors of α-diimine complexes of Ru and Re (Cook et al., 1984; Worl et al., 1991). The lowered energy of the MLCT transition resulted in a lower energy gap

between the excited state and the ground state, thus leading to an increased knr value. **Cu(NO2ph)** exhibited almost no emissions, which can be attributed to charge localization on the strong, electron-withdrawing NO<sup>2</sup> groups in the excited state, leading to a charge-separated radical-like state (Prei et al., 2015). This, in turn, should lead to an increase in the knr value.

The Cu<sup>I</sup> complexes, substituted with smaller aryl groups, showed two different sets of behavior. The emission spectra of the **Cu(3fu)**, **Cu(3th)**, and **Cu(3Bzth)** complexes containing 5-membered ring systems with heteroatoms at 3-position (**Figures 4C,D**, black, red, and yellow, respectively), showed almost the same features as those found in the **Cu(ph)** spectrum. All of these complexes showed strong emissions with 8<sup>r</sup> values of 0.24–0.45 and τ em values of ca. 20 µs, values that are almost the same as those of **Cu(ph)**. In these cases, changing the heteroatoms from O to S in the aryl rings appeared to have no great effect on the complexes' transition energies, but 8<sup>r</sup> value increased from 0.24 for **Cu(3fu)** to 0.35 for **Cu(3th)**, which further increased in **Cu(3Bzth)** (8<sup>r</sup> = 0.45).

However, in systems where the heteroatoms featured at the 2-position of the rings, as in **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)** (**Figures 4E,F**, green, blue, and purple for **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)**, respectively), the emissions were drastically different from those of the other complexes. The emission spectra were observed to gradually red shift in the complexes in the aforementioned order and their shapes changed to vibronic shapes, particularly for **Cu(2Bzfu)**. Because the excitation spectrum from which this emission was monitored was in accordance with the UV-Vis absorption spectrum of **Cu(2Bzfu)** (**Figure S3A**), this emission could be confirmed to arise from the excited state of **Cu(2Bzfu)**.

This emission of **Cu(2Bzfu)** was found to be strongly air sensitive. If the emission spectrum was recorded for a sample prepared under an Ar atmosphere instead of a vacuumdegassed sample, the spectrum was observed to drastically change, resulting in a mismatch between the excitation and absorption spectra of **Cu(2Bzfu)** in CH2Cl<sup>2</sup> (**Figure S3B**). However, this was not the case for the data recorded in a CH3CN solution, which showed no remarkable changes in the spectral features, even for a sample prepared under an Ar atmosphere (**Figure S12C**) compared to under vacuum conditions (**Figure S3C**).

Interestingly, these Cu<sup>I</sup> complexes' τ em values increased to over 100 µs, with a corresponding decrease in the 8<sup>r</sup> values. These results strongly indicate that the emissive lowest excited state changed from the MLCT state to another state, possibly the intraligand π-π ∗ excited state. This is a reasonable assumption, considering the red shift in the π-π ∗ transition band, as observed in the UV-Vis absorption spectra, from the π-expansion of the dmp ligand. Thus, in these complexes, the emission from the

spectra and (F) time-dependence data of the intensity of Cu(2th) (green), Cu(2Bzth) (blue), and Cu(2Bzfu) (purple). Left panels on (B, D, and F) were observed from a shorter range of each right panel. The samples were degassed using the freeze–pump–thaw method. Table S1 summarizes the excitation and monitored wavelengths.

lowest excited state should arise from the delayed fluorescence of the intraligand <sup>3</sup>π-π ∗ state.

The spectral feature depended strongly on the heteroatoms on the 2-positions of the aryl substituents. In the case of **Cu(2th)** and **Cu(2Bzth)**, the emission spectra were much broadened than that of **Cu(2Bzfu)**, although the lifetimes were still long. These results strongly indicate that the accessibility of the planar structure over the aryl substituted dmp enhances the πconjugation over the dmp ligand. The similar phenomena about the long excited-state lifetime caused by the planarity between aryl substituents and α-diimine ligand has been discussed for emissive Ru complexes, in which the delocalization of negative charge over the substituted bpy (bpy = 2,2′ -bipyridine) in the <sup>3</sup>MLCT excited state lowers their structural changes from the ground state, leading to decreasing knr (Damrauer et al., 1997; Majewski et al., 2016). It is also reported that the Cu<sup>I</sup> (dmp) complex has the close-lying <sup>3</sup>π-π ∗ state near the lowest <sup>3</sup>MLCT excited state as for Cu<sup>I</sup> (dmp)(PPh3) + 2 , which showed the vibronic emission spectrum at 77 K as a long-lifetime component as τ em ∼ 10 ms (Rader et al., 1981). Therefore, the long lifetime emission of **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)** can be reasonably attributed to the participation of <sup>3</sup>π-π ∗ excited state.

All of these tendencies were also valid even in CH3CN as a photocatalytic reaction solvent, except that the emission was significantly quenched compared to that observed in CH2Cl<sup>2</sup> solutions due to "exciplex" formation, in which a solvent molecule coordinates to the Cu center as a fifth ligand in the excited-state Cu<sup>I</sup> complexes (**Figure S2** and **Table 2**). This type of quenching is characteristic of tetrahedral Cu<sup>I</sup> complexes as the d <sup>10</sup> Cu<sup>I</sup> center decreases its charges upon closing to a d 9 configuration in the MLCT excited state, changing

the structure to planer one because of the Jahn-Teller effect (McMillin et al., 1985).

Thus, the quenching all of the Cu<sup>I</sup> complexes studied in CH3CN indicates that they have MLCT character in their excited state to some extent, even for **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)**, in which the excited state was mainly the π-π ∗ state. The extents of the MLCT character should increase especially in CH3CN because this state has CT in character, that are stabilized in a polarized solvent such as CH3CN more strongly than the intraligand π-π ∗ excited state. Actually, such the stabilizations were seen as red-shift of emission maxima in CH3CN compared to those in CH2Cl2. In the case of **Cu(NO2ph)**, the emission behavior completely changed in CH3CN, possibly as a result of photodecomposition, manifesting in a different excitation spectrum than **Cu(NO2ph)**'s original absorption spectrum.

Photocatalytic CO<sup>2</sup> reduction reactions were performed using these newly designed Cu<sup>I</sup> complexes as redox photosensitizers mixed with FeII(dmp)2(NCS)<sup>2</sup> (0.05 mM) as a catalyst and BIH (10 mM) as a reductant under a 436 nm monochromic light in a mixed solution of CH3CN–triethanolamine [TEOA, 5:1 (v/v)]. The Cu photosensitizer's concentration was fixed at 0.5 mM, showing almost no transparency at 436 nm in reaction cells with a 1-cm optical path length. **Figure 5** shows the photocatalytic reaction results. The products were found to be only CO and H2, without the formation of any HCOOH as reported in a previous study performed under similar conditions (Takeda et al., 2016), and these compounds were linearly generated against the photoirradiation time for up to 2 h.

The photocatalytic properties were highly dependent on the Cu complex photosensitizer used (**Table 3**). When **Cu(Bph)** was used, the efficiencies were enhanced to 8<sup>r</sup> = 3.9 with H<sup>2</sup> formation (8<sup>r</sup> = 1.4). Take into account the low results for **Cu(H)** [8<sup>r</sup> = 1.1 (Takeda et al., 2016)] and **Cu(ph)** (8<sup>r</sup> = 2.3), increasing the number of the ph groups at the 4,7-positions of dmp tended to enhance the photosensitizing ability of these complexes for CO<sup>2</sup> reduction.

Such the enhancement in the photosensitizing ability was also confirmed in the cases of complexes featuring 5-membered ring systems with heteroatoms at the 3-position, such as **Cu(3fu)**, **Cu(3th)**, and **Cu(3Bzth)**. The most efficient case was **Cu(3Bzth)**, which showed 8<sup>r</sup> values of 4.1 and 2.0 for CO and H<sup>2</sup> evolution, respectively. Even in the cases of **Cu(3fu)** and **Cu(3th)**, the efficiencies for the CO<sup>2</sup> reduction were improved to 8<sup>r</sup> (CO) = 3.5 and 2.9, respectively, over those of **Cu(ph)**.

However, the Cu<sup>I</sup> complexes containing 5-membered ring systems with heteroatoms at 2-position, such as **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)**, showed no any photosensitizing ability in this reaction system. These complexes may have lower oxidative power in their excited state because their highest occupied molecular orbital (HOMO) energies should now increase since the π orbitals in the substituted dmp, as well as the d orbitals on the Cu<sup>I</sup> center, are destabilized. Actually, the first oxidation wave in the cyclic voltammetry (CV) measurements, an indicator of the HOMO energy, showed positive shifts for **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)** (shoulder waves around 0.9 V vs. Ag/AgNO3) compared to that of **Cu(ph)** (E<sup>p</sup> = 0.92 V vs. Ag/AgNO3) (**Figure S4**).

The excited states of these Cu<sup>I</sup> complexes were effectively quenched in the presence of BIH, except for that of **Cu(NO2ph)**, which means that all of these complexes might generate the corresponding one-electron reduced state through the reductive quenching of their excited states by BIH during photoirradiation (Tamaki et al., 2013; Takeda et al., 2018). The quenching rate constants (kq) were estimated to be ∼10<sup>10</sup> M−<sup>1</sup> s −1 , corresponding to quantitative quenching (ηq) under photocatalytic conditions with 10 mM of BIH (**Table 3**). Because the quenching behavior in terms of the emission lifetimes was in good agreement with the emission intensity changes, and no changes were observed in the UV-vis and emission spectra of the Cu<sup>I</sup> complexes, even in the presence of BIH (**Figures S5–S12**). These quenching reactions can be attributed to a dynamic quenching process, in which the excited state reacts with BIH bimolecularly.



<sup>a</sup>photocatalytic system consisting of Fe(dmp)2(NCS)<sup>2</sup> (0.05 mM) as a catalyst and BIH (10 mM) as a reductant in CH3CN–TEOA (5:1 v/v) (see Figure 4);

<sup>b</sup>Potentials [V] vs. Ag/AgNO<sup>3</sup> (0.01 M);

<sup>c</sup>ηq, quenching fraction of the excited-state Cu<sup>I</sup> complexes by BIH (10 mM), estimated by η<sup>q</sup> = KSV [BIH]/(1 + KSV [BIH]), in which the KSV [M-1] values are the slopes of the Stern-Volmer plots in Figures S5–S12;

d from Takeda et al. (2016);

<sup>e</sup>not determined due to the low product amount;

<sup>f</sup> not detected due to weak emission.

Then, in situ UV-Vis absorption spectral changes in the reaction solutions during the photocatalytic reactions were investigated and compared to those of the complexes recorded under an Ar atmosphere for up to 1 h of photoirradiation. In the cases of **Cu(3fu)**, **Cu(3th)**, and **Cu(3Bzth)** (**Figures S15–17**), which were good photosensitizers for photocatalytic CO<sup>2</sup> reduction, almost no changes occurred in their spectra. This indicates that there was almost no decomposition of the Cu complexes in the photosensitizing cycles, donating electrons to the Fe catalyst, for the oxidation of BIH and CO<sup>2</sup> reduction. Under an Ar atmosphere, the MLCT absorption bands of the original Cu<sup>I</sup> complexes decreased just after the photoirradiation caused broad absorption up to 950 nm. These changes clearly indicate the original Cu<sup>I</sup> complexes' photodecomposition. Because the one-electron reduced state produced by the reductive quenching of the excited state cannot donate to the CO<sup>2</sup> reduction cycle of the Fe catalyst, the reduced species of the Cu<sup>I</sup> complexes accumulate in the reaction solutions, decomposing due to their anionic radical character if CO<sup>2</sup> is not present.

On the other hand, non-photosensitizing Cu<sup>I</sup> complexes, such as **Cu(NCph)**, **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)** (**Figures S13**, **S18–20**, respectively) showed signs of drastic decomposition even in the presence of CO2. This means that the Cu complexes cannot donate their electrons generated through the reductive quenching by BIH in the excited state for the CO<sup>2</sup> reduction. Thus, these complexes showed no photocatalytic CO<sup>2</sup> reduction properties. However, the lack of the photosensitizing ability of **Cu(NO2ph)** should be due to its unstable excited state because no changes were observed, neither under a CO<sup>2</sup> nor an Ar atmosphere, meaning that no any redox cycles occurred under photoirradiation.

The Cu<sup>I</sup> complexes' electrochemical properties were then examined, and the results are shown in **Figure 6**. All of the Cu<sup>I</sup> complexes, apart from **Cu(H)**, showed a reversible redox couple in the CV as the first reduction wave, indicating that the corresponding one-electron reduced species was stabilized via the introduction of the aryl substituents on the dmp ligand, as observed for **Cu(ph)**. This is quite reasonable considering that the first reduction occurs in the dmp ligand's low-lying π ∗ orbital. However, for the CV of **Cu(NO2ph)**, the redox couple current was much higher without any sharpening. Thus, the first reduction wave in the CV of **Cu(NO2ph)** arises as a result of two sequential local electron reductions in the two NO2ph parts of one **Cu(NO2ph)**. The redox potentials (E1/2) of these first reduction waves are summarized in **Table 3**. Apparently, introducing electron-withdrawing groups, such as NCph and NO2ph, resulted in positive shifts in the E1/<sup>2</sup> values up to −1.75 and −1.31 V vs. Ag/AgNO<sup>3</sup> for **Cu(NCph)** and **Cu(NO2ph)**, respectively, indicating that the electron-donating ability of the corresponding reduced species was lowered. Those of the **Cu(Bph)**, **Cu(3fu)**, **Cu(3th)**, and **Cu(3Bzth)** complexes that have photosensitizing ability in this photocatalytic system maintained values more negative than −1.9 V, although these E1/<sup>2</sup> values were positively shifted from those of **Cu(H)** and **Cu(ph)** (E1/<sup>2</sup> = −2.03 and −1.96 V vs. Ag/AgNO3, respectively). For **Cu(2th)**, **Cu(2Bzth)**, and **Cu(2Bzfu)**, in which the π-system largely extends to the aryl substituents, exhibited more positive first reduction potentials of −1.83, −1.74, and −1.70 V, respectively, due to the stabilization of the π <sup>∗</sup> orbital over the substituted dmp ligand. Thus, the lowering energy of the π <sup>∗</sup> orbital on the dmp ligand via the introduction of phenyl rings with πelectron-withdrawing groups or 5-membered aryl groups was clearly confirmed.

Compared to the first reduction wave potential of Fe(dmp)2(NCS)<sup>2</sup> in CH3CN–TEOA (5:1 v/v) solution (−1.61 V vs. Ag/AgNO3) under a CO<sup>2</sup> atmosphere (Takeda et al., 2016), the reducing power of these Cu complexes is enough for them to donate their first single electron on the corresponding reduced species to the Fe catalyst. However, the decomposition behavior during the photoirradiation process indicates that the electron transfer to the Fe catalyst is inefficient, resulting in an accumulation of the reduced species of the Cu complexes and decomposition over longer time scales than those of the CV measurements. We are now continuing the photocatalytic reaction using the other catalysts for the CO<sup>2</sup> reduction that function at a more positive potential.

### CONCLUSION

This study found the heteroleptic Cu<sup>I</sup> complexes' visible-light absorption properties could be improved via expanding the dmp ligand's π-systems. Simple expansion using biphenyl substitution at the 4,7-popsitions of dmp enhanced ε without red shift

of the <sup>1</sup>MLCT band compared to that of the ph substituted one. Although π-electron accepting substituents such as 4- NCph or 4-NO2ph shifted the band, the λmax values were still around 400 nm. The X-ray single crystal analysis showed the presence of protons at the 2,6-positions of the ph substituents slows the π-conjugation between the ph groups and dmp due to the steric hindrance of the 5,6-protons in the dmp ligand. Thus, substitution with 5-membered heterorings, with heteroatoms and no protons at the 2-positions, was enough to make improvements to the properties of the complexes. Thus, the complexes containing 2-benzofuryl groups having an O atom which size is smaller than a S atom at each 2 position showed a planner structure over the substituted dmp ligand, resulting in a red shift in the <sup>1</sup>MLCT absorption bands up to 500 nm. Because the 3-thienyl and 3-furyl substitutions showed almost no effect on the light absorbing property, this red shift in **Cu(2Bzfu)** was not from a difference of electronegativities of the S and O atoms. Photocatalytic CO<sup>2</sup> reduction reactions were carried out with the complexes in the presence of a Fe catalyst, showing quantum yields of up to 4% for the maximum amount of CO generated. The knowledge obtained in this study should help not only for further expansion of visible-light absorbing but also for designing the efficient redox-photosensitizer using the Cu<sup>I</sup> complex to construct efficient earth-abundant system for the photocatalytic CO<sup>2</sup> reduction.

# EXPERIMENTAL

### General Measurements

<sup>1</sup>H NMR spectra were recorded on a JEOL ECA-400II NMR spectrometer. The chemical shifts and coupling values arising from the second order effects of the protons in the phenanthroline derivatives were assigned by analyzing the obtained <sup>1</sup>H NMR spectra using the iNMR (ver. 6.1.8) and WINDNMR (ver. 7.1.14) software (Reich, 1995). UV-vis absorption spectra were recorded on a JASCO V-565 or V-670 spectrometers. To measure the molar extinction coefficiencies (ε [M−<sup>1</sup> cm−<sup>1</sup> ]), a 10 mL solution containing quantitatively weighted each Cu<sup>I</sup> complexes (∼2 mg) was prepared by dissolving in CH2Cl<sup>2</sup> or CH3CN as a stock solution. An 1 mL of the stock solution was further diluted to 5 mL by each solvent and UV-vis absorption spectra of these two solutions were recorded using a cuvette cell having an 1-cm path length. After confirmation of no concentration effect on the spectral features indicating no dimer formation by comparison these two spectra, the absorbance was divided by each concentration to obtain ε value. The spectrum of the stock solution was used for assessment of the <sup>1</sup>MLCT band and the latter was used for π-π ∗ absorption band because the detection range of absorbance (0.1−1.0) is suitable. Emission and excitation spectra were recorded on a JASCO FP8600 fluorescence spectrometer. Emission quantum yields of the Cu<sup>I</sup> complexes in CH3CN were determined by

Takeda et al. Cu(I) Complex Photosensitizer

a relative method, using an air-saturated CH3CN solution containing Ru(bpy)3(PF6) as a standard (8 = 0.018) (Suzuki et al., 2009; Ishida and Beeby, 2016). Emission quantum yields in CH2Cl<sup>2</sup> were collected using a Hamamatsu Photonics C9920– 02G absolute photoluminescence quantum yield measurement system, consisting of a calibrated integrating sphere and a multichannel spectrometer. Excited-state lifetimes were recorded on a HORIBA FluoroCube 1000U-S single-photon counting system. The sample solutions were degassed, via freeze–pump–thaw cycles, before the emission and lifetime measurements. For the quenching experiments, changes in the emission intensities and excited-state lifetimes of the Cu<sup>I</sup> complexes in CH3CN, with or without various concentrations of BIH, were monitored under an Ar atmosphere. CVs were measured using Et4NBF<sup>4</sup> (0.1 M) as a supporting electrolyte under an Ar atmosphere using a BAS CHI620EX or CHI760Es electrochemical analyzer with a glassy carbon working electrode (diameter, 3 mm), an Ag/AgNO<sup>3</sup> (0.01 M) as a reference electrode, and a Pt counter electrode.

### Photocatalytic Reactions

A 4-mL CH3CN–TEOA (5:1 v/v) solution containing Fe(dmp)2(NCS)<sup>2</sup> (0.05 mM), and the Cu photosensitizer (0.5 mM), and BIH (10 mM) as a reductant in a quartz cubic cell (1 cm path length; 11.0 mL volume) was bubbled with CO<sup>2</sup> for 20 min. For the quantum yield determinations, a SHIMADZU QYM-01 photoreaction, quantum yield, evaluation system was adopted. The sample solution was irradiated using an Asahi Spectra Co. MAX-303 300-W Xe lamp with a 436 nm band-pass filter (FWHM: 10 nm), with the irradiated light intensity set at 3 × 10−<sup>8</sup> einstein s−<sup>1</sup> and the temperature of the reaction solution was maintained at 25 ± 0.1 ◦C using an IWAKI CTS-134A cooling thermo pump during the irradiation. The gaseous products, i.e., CO and H2, were analyzed using GC-TCD (GL science GC323) with an active carbon column. The amount of formic acid in the reaction solution with diluted water was analyzed using a capillary electrophoresis system (Otsuka Electronics Co. 7100L) with a buffer solution (pH 5.9) consisting of quinolinic acid, hexadecyltrimethylammonium hydroxide, and 2-amino-2-hydroxymethyl-1,3-propanediol as the electrolyte. To observe in situ UV-Vis spectral changes of the photocatalytic-reaction solution, the solution was irradiated, using an Ushio Optical Module high-pressure Hg lamp (BA-H500), with a 436 nm band-pass filter (FWHM = 10 nm),

# REFERENCES


purchased from Asahi Spectra Co., and a CuSO<sup>4</sup> solution (250 g L −1 , 5 cm path length) filter with a 436 nm band-pass filter (FWHM: 10 nm) purchased from Asahi Spectra Co. Neutral density glass filters reduced the light intensity. The spectral changes were recorded on a Photal MCPD-6800 photodiode array detector equipped with a MC-2530 light source.

# DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

# AUTHOR CONTRIBUTIONS

HT and OI planned the all experiments. HT and YM synthesized the samples and performed the photophysical, electrochemical, and photocatalytic measurements. HT, HS, and HU analyzed the single crystal structures. HT wrote the manuscript and all authors have read and approved it.

# FUNDING

This work was supported by JST CREST (Grant Number JPMJCR13L1). HT has received funding from JSPS KAKENHI (Grant Number JP16K17876). HU has received funding from JSPS KAKENHI (Grant Number JP17K05745 and JP18H04504).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00418/full#supplementary-material

Data Sheet 1 | The Cu<sup>I</sup> complexes' photophysical properties in CH3CN, emission and excitation spectra of Cu(2Bzfu) in CH2Cl2 under Ar or degassed atmospheres, positive scans of the cyclic voltammograms of the Cu<sup>I</sup> complexes, quenching experiments of the excited state of the Cu<sup>I</sup> complexes by BIH, in situ UV-Vis spectral changes during the photoirradiation reactions, experimental details such as materials, synthesis, <sup>1</sup>H NMR spectra, peak analysis of the <sup>1</sup>H NMR spectra of the newly synthesized ligands, and details of the single-crystal X-ray crystallography of Cu(ph), Cu(NCph), Cu(2Bzth), and Cu(2Bzfu).

Data Sheet 2 | Crystallographic data for Cu(ph), Cu(NCph), Cu(2Bzth), and Cu(2Bzfu) (CCDC numbers: 1895526-1895529) embedding RES and structure factors.

2,2'-Bipyridyls with Ruthenium(II). J. Chem. Soc. Perkin Trans. 2, 1303–1307. doi: 10.1039/P29840001303


copper(I) photosensitizers used in hydrogen evolving systems. ChemPhysChem 15, 3709–3713. doi: 10.1002/cphc.201402585


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Takeda, Monma, Sugiyama, Uekusa and Ishitani. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Secondary-Sphere Effects in Molecular Electrocatalytic CO<sup>2</sup> Reduction

Asa W. Nichols and Charles W. Machan\*

*Department of Chemistry, University of Virginia, Charlottesville, VA, United States*

The generation of fuels and value-added chemicals from carbon dioxide (CO2) using electrocatalysis is a promising approach to the eventual large-scale utilization of intermittent renewable energy sources. To mediate kinetically and thermodynamically challenging transformations of CO2, early reports of molecular catalysts focused primarily on precious metal centers. However, through careful ligand design, earth-abundant first-row transition metals have also demonstrated activity and selectivity for electrocatalytic CO<sup>2</sup> reduction. A particularly effective and promising approach for enhancement of reaction rates and efficiencies of molecular electrocatalysts for CO<sup>2</sup> reduction is the modulation of the secondary coordination sphere of the active site. In practice, this has been achieved through the mimicry of enzyme structures: incorporating pendent Brønsted acid/base sites, charged residues, sterically hindered environments, and bimetallic active sites have all proved to be valid strategies for iterative optimization. Herein, the development of secondary-sphere strategies to facilitate rapid and selective CO<sup>2</sup> reduction is reviewed with an in-depth examination of the classic [Fe(tetraphenylporphyrin)]+, [Ni(cyclam)]2+, Mn(bpy)(CO)3X, and Re(bpy)(CO)3X (X = solvent or halide) systems, including relevant highlights from other recently developed ligand platforms.

### Edited by:

*Soumyajit Roy, Indian Institute of Science Education and Research Kolkata, India*

### Reviewed by:

*Zhong Jin, Nanjing University, China Kyle Grice, DePaul University, United States*

### \*Correspondence:

*Charles W. Machan machan@virginia.edu orcid.org/0000-0002-5182-1138*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *30 January 2019* Accepted: *17 May 2019* Published: *13 June 2019*

### Citation:

*Nichols AW and Machan CW (2019) Secondary-Sphere Effects in Molecular Electrocatalytic CO2 Reduction. Front. Chem. 7:397. doi: 10.3389/fchem.2019.00397* Keywords: inorganic, electrocatalysis, CO2 , secondary-sphere, molecular

# INTRODUCTION

The development of scalable and cost-effective processes to store electrical energy in chemical bonds using CO<sup>2</sup> as a primary feedstock remains a significant challenge for energy research (Centi and Perathoner, 2009; Senftle and Carter, 2017). Of particular interest are homogeneous catalysts for CO<sup>2</sup> reduction, as their well-defined structures give chemists an opportunity to effectively characterize intermediates relevant to the operating mechanism and further optimize active catalyst families through iterative ligand design (Benson et al., 2009; Rountree et al., 2014). While the influence of different ligand types in the primary coordination sphere can be observed in the catalytic activity and selectivity of transition metals across the d-block towards CO<sup>2</sup> reduction, more subtle effects—including rate enhancement and lowered overpotentials—can be obtained through modulation of the secondary coordination sphere within specific classes of metal complexes (Jiang et al., 2019). This has been particularly successful in the design of highly active and selective catalysts for CO<sup>2</sup> reduction, directing a paradigmatic shift in the general understanding of "design principles" (Costentin et al., 2012a, 2014a,b; Sampson et al., 2014; Azcarate et al., 2016; Ngo et al., 2017).

While this review focuses on molecular catalysts which were specifically examined for electrochemical CO<sup>2</sup> reduction, secondary-sphere effects have been successfully harnessed in related catalytic processes, including thermal CO<sup>2</sup> hydrogenation (Himeda et al., 2004, 2005, 2007; Hull et al., 2012; Wang et al., 2012, 2013, 2014; Manaka et al., 2014; Suna et al., 2014; Cammarota et al., 2017), hydrogen evolution (Curtis et al., 2003; Henry et al., 2005, 2006; Wilson et al., 2006; Fraze et al., 2007; Jacobsen et al., 2007a,b; DuBois and DuBois, 2009; Gloaguen and Rauchfuss, 2009; Helm et al., 2011; Reback et al., 2013; Ginovska-Pangovska et al., 2014), hydrogen oxidation (Curtis et al., 2003; Henry et al., 2005, 2006; Wilson et al., 2006; Fraze et al., 2007; Jacobsen et al., 2007a,b; Dutta et al., 2013, 2014; Ginovska-Pangovska et al., 2014), formate oxidation (Galan et al., 2011, 2013; Seu et al., 2012), and oxygen reduction (Collman, 1977; Collman et al., 2004; Lewis and Tolman, 2004; Mirica et al., 2004; Fukuzumi, 2013; Ray et al., 2014; Fukuzumi et al., 2015; Nam, 2015; Sahu and Goldberg, 2016; Elwell et al., 2017; Hong et al., 2017; Sinha et al., 2019) reactions. In this review, we focus on how the mechanism of CO<sup>2</sup> reduction relates to the type of secondary-sphere effects employed in molecular systems. Strategies discussed here which have been shown to increase catalytic activity and selectivity include pendent proton donors/shuttles in the form of Brønsted sites, charged moieties, sterically demanding functional groups, bimetallic active species, and stoichiometric participation of alkali and alkaline earth metal cations (**Graphical Abstract**).

To contextualize the motivation and principles, examples of secondary-sphere effects in enzymes which catalyze the interconversion of CO<sup>2</sup> with either CO or formic acid are discussed. This overview is followed by a careful examination of secondary-sphere effects in several abiotic molecular electrocatalyst examples, beginning with the [Fe(tetraphenylporphyrin)]<sup>+</sup> [Fe(TPP)]<sup>+</sup> systems pioneered by Savéant, Robert, and Constentin, including a discussion of the effects of pendent proton source placement and the distance dependence of through-space effects induced by charged residues (Costentin et al., 2014a,b; Manbeck et al., 2015; Mohamed et al., 2015; Azcarate et al., 2016; Zahran et al., 2016; Margarit et al., 2018; Nichols et al., 2018b; Sinha and Warren, 2018). Next, M(bpy)(CO)3X catalysts (M = Mn or Re; X = solvent molecule or halide) (Wong et al., 1998; Bourrez et al., 2011; Smieja et al., 2013; Chabolla et al., 2014, 2017; Franco et al., 2014; Machan et al., 2014a, 2015, 2016; Riplinger et al., 2014; Agarwal et al., 2015; Manbeck et al., 2015; Riplinger and Carter, 2015; Machan and Kubiak, 2016; Ngo et al., 2017; Sahu et al., 2017; Sinha et al., 2019) in which steric parameters, pendent Lewis acid effects, and charged residues have been shown to be effective will be discussed. Finally, [Ni(cyclam)]2<sup>+</sup> (cyclam = 1,4,8,11-tetraazacyclotetradecane), which contains pendent proton donors on the coordinating N atoms of the macrocycle, is discussed (Beley et al., 1984; Barefield et al., 1986; Balazs and Anson, 1993; Kelly et al., 1999; Froehlich and Kubiak, 2012; Song et al., 2014; Nichols and Chang, 2018). These systems are among the most highly studied in the field and representative of the progress that has been made in understanding how secondarysphere coordination effects enhance molecular electrocatalysis. Additional discussions on emerging systems for CO<sup>2</sup> reduction which utilize secondary-sphere effects are included to summarize some of the current work in the field.

# ENZYMES FOR THE INTERCONVERSION OF CO<sup>2</sup> AND CO OR FORMIC ACID

### Cu,Mo-Containing Carbon Monoxide Dehydrogenase (Cu,Mo-CODH)

The structure and function of Cu,Mo-CODH enzymes have been previously reviewed in great detail (Appel et al., 2013; Kroneck and Torres, 2014). Of this class of enzyme, the Cu,Mo-CODH in O. carboxydovorans has been the most extensively studied (**Figure 1A**) (Dobbek et al., 2001a, 2002; Hille, 2013). The active site of this enzyme contains Mo and Cu ions, which are bridged by a µ2-sulfido ligand. Mo is coordinated in a distorted square pyramidal fashion by an enedithiolate moiety (from the pyran ring found in the pyranopterin cofactor), a µ2-sulfido ligand, and an oxo/hydroxo ligand in the equatorial plane. The apical ligand is an oxo, with a glutamine residue within hydrogen-bonding distance and a glutamate residue in a trans position. A unique structural feature of Cu,Mo-CODH in comparison to other Mocontaining hydrogenases is its covalent linkage through a cysteine residue to the Cu ion, connecting the heterobimetallic active site to the large subunit of the CODH enzyme.

In the catalytic cycle for transformation of CO to CO<sup>2</sup> by Cu,Mo-CODH, CO is proposed to initially bind to the Cu(I) ion **2**, followed by nucleophilic attack by the equatorial oxo ligand to generate the bridging µ2-η 2 -CO<sup>2</sup> adduct **3** (**Figure 1B**) (Dobbek et al., 2002). This five membered metallacycle results from a redox-neutral reaction which maintains the Mo(VI) state (**Figure 1B**). Following rearrangement and net oxidation of CO, CO<sup>2</sup> release occurs with hydroxide binding to produce the Mo(IV) species **4**. Subsequent 2e<sup>−</sup> oxidation returns the Mo active site to the initial Mo(VI) oxidation state. Notably, the glutamine residue, which is in contact with the equatorial

oxo ligand (**Figure 1B**), can act as a Brønsted base (proton acceptor) to regenerate the more nucleophilic oxo ligand when transitioning from **4** to **1**. If the reverse of this catalytic cycle is imagined, it is clear that a proton-donating residue would be beneficial for CO<sup>2</sup> reduction and C–O bond cleavage to produce CO.

### Monofunctional Ni,Fe-Containing CODHs

The active site of Ni,Fe-containing CODHs, known as cluster C, is proposed to cycle between three separate redox states during catalytic CO oxidation (Cred1, Cint, and Cred2; **Figure 2**) (Jeoung and Dobbek, 2007). The behavior and activity of cluster C from Carboxydothermus hydrogenoformans have been experimentally interrogated in these three different redox states utilizing chemical-reducing agents (Jeoung and Dobbek, 2007). The cofactor contains an Fe4Ni cluster bridged by sulfide ligands with a single Fe and Ni atom in the active site, which also contains histidine and lysine residues in close proximity to interact with active site-bound substrate molecules. In Cred1, the active site contains Fe2<sup>+</sup> and Ni2<sup>+</sup> ions with an Fe-bound hydroxo ligand that is within hydrogen-bonding distance of the proximal lysine residue. Upon exposure to CO, the Fe-bound hydroxide is deprotonated, and the resultant oxo species can form a new C–O bond to generate the µ2-η 2 -CO<sup>2</sup> adduct Cred2-CO2, where CO<sup>2</sup> is bound through C by Ni and O by Fe (**Figure 2**) (Jeoung and Dobbek, 2007). This µ2-η 2 -CO<sup>2</sup> binding mode is stabilized by hydrogen-bonding interactions with the pendent histidine and lysine residues (Jeoung and Dobbek, 2007). A 2e<sup>−</sup> oxidation with concomitant binding of an aquo ligand reduces the Ni2<sup>+</sup> to Ni<sup>0</sup> in Cred1, triggering release of CO2. Oxidation of the reduced cofactor by 2e<sup>−</sup> regenerates the Cred1 catalytic resting state.

### Formic Acid Dehydrogenase

Formic acid dehydrogenase (FADH) enzymes catalyze the reversible 2e−/1H<sup>+</sup> interconversion of CO<sup>2</sup> and formate, another reaction of interest to CO<sup>2</sup> utilization (Sawers, 1994). The active site of the FADH enzyme in Escherichia coli contains a single Mo(VI) atom coordinated to four S atoms from two cis-dithiolene moieties originating from the bis(molybdopterin guanine dinucleotide) cofactor, a hydroxyl ligand, and a SeCysteine (Boyington et al., 1997). The active site also contains histidine and arginine residues in a position to interact with activated Mo-bound intermediates (**Figure 2B**). The catalytic oxidation of formate begins by displacement of the hydroxyl ligand with an equivalent of formate, which binds as an η 1 -OCHO ligand; the η 1 -O binding mode is stabilized through hydrogen-bonding interactions between the arginine and histidine residues and the unbound H and O atoms of formate (**Figure 2B**). Subsequent oxidation and transfer of 2e<sup>−</sup> from formate to the Mo center occurs with the release of CO<sup>2</sup> and proton transfer to the SeCysteine residue. The Mo(IV) center can then be returned to the resting state through the loss of 2e<sup>−</sup> as the histidine deprotonates the Mo-bound SeCysteine. Abiotic structural motifs similar to these three examples are discussed in the subsequent sections where either a) bimetallic active sites or b) secondary-sphere moieties are used to mediate electrochemical CO<sup>2</sup> reduction.

# Benchmarking Molecular Electrocatalysts for CO<sup>2</sup> Reduction

Electrochemical techniques can facilitate the determination of kinetic and thermodynamic parameters for both Faradaic and catalytic reactions. Cyclic voltammetry (CV), a nondestructive potentiostatic technique, is particularly informative. Indeed, the breadth and importance of rigorous CV analysis have been reviewed in great detail (Savéant, 2008, 2018; Costentin et al., 2012b; Costentin and Savéant, 2014, 2018a; Rountree et al., 2014). Experimental determination of fundamental reaction parameters

is essential for comparing the effects of pendent functional groups on the activity of electrocatalysts.

The effective catalytic overpotential (η) is the difference between the standard potential of CO<sup>2</sup> reduction (E 0 CO2/CO) and the potential at half catalytic current height (Ecat/2) as described in Equation (1) and describes a thermodynamic quantity: the electrochemical energy beyond the standard potential which is required to drive a reaction of interest at an appreciable rate. We note that some prefer to define the overpotential term as ECO2 − E1/2. We distinguish between these through the use of "effective catalytic overpotential" to describe the overpotential calculated utilizing Ecat/<sup>2</sup> (Appel and Helm, 2014).

$$
\eta = E\_{\text{CO2/CO}}^0 - E\_{\text{cat}/2} \tag{1}
$$

Another reaction parameter commonly measured through CV is the maximal turnover frequency (TOFmax), which can also be described as the observed catalytic rate constant (kobs) with units of s−<sup>1</sup> (Costentin and Savéant, 2014). In a Nernstian electrocatalytic reaction, TOF is related to overpotential by the catalytic Tafel equation, Equation (2). TOF is the turnover frequency at the applied potential, F is Faraday's constant, R is the ideal gas constant, T is the temperature, E1/<sup>2</sup> is the catalyst standard reduction potential, and ηapp is the difference between E 0 CO2/CO and the applied potential. It is worth noting, however, that the limitations of the molecular catalytic reactions, including diffusion and side phenomena, can cause 'peaks' in catalytic CV waves and truncate the region where the Tafel relationship results in increased activity as the applied potential becomes more negative.

$$\text{TOF} = \frac{\text{TOF}\_{\text{max}}}{1 + \exp\left[\frac{F}{RT}\left(E\_{\text{CO2/CO}}^0 - E\_{1/2} - \eta\_{\text{app}}\right)\right]} \tag{2}$$

To benchmark the kinetic and thermodynamic parameters of different electrocatalysts, Equation (2) can be used to generate catalytic Tafel plots (nota bene, these explicitly include the limitations imposed on molecular Tafel behavior mentioned above and enable comparison across different experimental conditions). **Figure 3** shows an example of a catalytic Tafel plot for a generic molecular catalyst: "better" catalysts are located at the top left of catalytic Tafel plots, where overpotential is low and TOFmax is large and "worse" catalysts are located at the bottom right of a catalytic Tafel plot where the trends in overpotential and TOFmax are reversed. **Figure 4** shows a Tafel plot comparing reported catalysts.

Finally, simple plots of log(TOFmax) vs. Ecat/<sup>2</sup> can be used to compare catalysts in the same family to look for secondarysphere effects (Pegis et al., 2017; Costentin and Savéant, 2018b; Nichols et al., 2018b). If there is no secondary-sphere effect, then

a linear scaling relationship based on electrochemical driving force should occur; however, if a secondary-sphere effect is present, a deviation from the linearity achieved by comparing inductive substitution effects can be observed. One should utilize caution using this method for comparing catalysts, however, as changes to the reaction mechanism that alter cosubstrate identity, concentration dependence, solvent, and cosolvent effects can greatly impact the catalytic activity of a series of complexes through scaling relationships that are unrelated to the secondarysphere interactions of interest. Likewise, the effects of Nernstian changes in the experimentally observed potential based on changes in the reaction equilibrium, Keq, can obscure details if not properly accounted for (Costentin and Savéant, 2018b).

# Secondary-Sphere Effects in [Fe(TPP)]<sup>+</sup>

For [Fe(TPP)]<sup>+</sup> complexes, increased activity for the electrocatalytic reduction of CO<sup>2</sup> to CO is observed with the addition of both Lewis and Brønsted acids (Bhugun et al., 1994, 1996; Gennaro et al., 1996). As a result, [Fe(TPP)]<sup>+</sup> complexes bearing secondary-sphere functionalities exploiting classical Brønsted acid/base push–pull reactions have been the subject of extensive studies (Costentin et al., 2012a, 2014a,b; Sinha and Warren, 2018). In particular, iron 5, 10, 15, 20 tetrakis-(2,6-dihydroxyphenyl)-porphyrin ([Fe((OH)8TPP)]+) was identified as a promising candidate for electrochemical CO<sup>2</sup> reduction because it positioned proton donors oriented toward the active site in close enough proximity to interact with bound substrate. Experimentally, the pendent –OH moieties were observed to cause a large catalytic current enhancement and a decrease in overpotential (Costentin et al., 2012a). Control experiments using [Fe((OMe)8TPP)]+, where the –OH groups were converted to –OMe ether moieties, also showed an increased TOFmax relative to [Fe(TPP)]+; however, the overpotential was much larger than both [Fe(TPP)]<sup>+</sup> and [Fe((OH)8TPP)]+. These differences relate to the mechanism of CO<sup>2</sup> reduction by [Fe(TPP)]<sup>+</sup> and how key steps are affected by the presence of the pendent –OH moieties (Bonin et al., 2017).

The proposed mechanism for the reduction of CO<sup>2</sup> by [Fe(TPP)]<sup>+</sup> requires the electrochemical generation of [Fe(0)TPP]2<sup>−</sup> at the electrode surface, followed by CO<sup>2</sup> binding and activation to generate [Fe(I)(TPP)(η 1 -CO•− 2 )]2−. The sacrificial proton donor, AH, stabilizes the binding of the CO•− 2 radical anion through hydrogen bonding. Concerted protonation and electron transfer from the metal center cause C–O bond cleavage, leading to the formation of [Fe(II)(TPP)(CO)]<sup>0</sup> . Release of CO is facilitated by a comproportionation reaction with a second equivalent of [Fe(0)TPP]2<sup>−</sup> to generate two equivalents of [Fe(I)TPP]<sup>−</sup> and one equivalent of CO. We note that a second mechanism has been proposed for Fe(TPP), wherein upon CO<sup>2</sup> binding, the metal center is oxidized by 2e<sup>−</sup> from Fe(0) to Fe(II) in the η 1 -CO<sup>2</sup> adduct before interaction with the proton donor, rather than generating an Fe(I) species (Fukuzumi et al., 2018).

The introduction of a pendent proton source minimally alters the mechanism for CO<sup>2</sup> reduction but causes catalytic rate increases by favoring several steps of the reaction. Upon the generation of [Fe(0)((OH)8TPP)]2−, CO<sup>2</sup> also binds in η 1 fashion and is activated to the CO•− 2 radical anion, but stabilization occurs through hydrogen-bonding interactions with the pendent proton donors (Costentin et al., 2012a). Experimentally, this is observed as a prewave to the catalytic feature in CV experiments, which can be more easily examined through additional modulation of electron density at the metal center with the related partially fluorinated derivative [Fe((F)10(OH)8TPP)]<sup>+</sup> (Costentin et al., 2014a,b). Subsequent intramolecular protonation balanced by the exogenous proton source can occur, generating η 1 - •CO2H at an Fe(I) center (Costentin et al., 2014a). The cleavage of the C–OH bond is induced upon further reduction and a concerted intramolecular

[Fe(*o*-2-amide-TPP)]+. Reprinted with permission from Nichols et al. (2018b)*—*published by the Royal Society of Chemistry.

protonation reaction (again balanced by proton transfer from the sacrificial donor) to regenerate the resting state, Fe(I), of the electrocatalytic cycle and CO product. The enhanced catalytic activity at more negative potentials of the ether-containing control complex may also be explained through an enhancement of this mechanism (the ether groups function as a Brønsted base to orient and enhance the proton activity of the exogeneous proton donor), but this has a relatively lower enhancement effect on the CO<sup>2</sup> reduction reaction. A comparison of the catalytic activity increases from these effects can be seen through the catalytic Tafel analysis shown in **Figure 4**.

Another area of study for secondary-sphere effects focuses on mono-functionalized porphyrins, which also orient functional groups toward the metal active site. Both the positioning of these functional groups relative to the active site and the pK<sup>a</sup> of the pendent proton source are of importance to the electrocatalytic activity for CO<sup>2</sup> reduction. In a study by Chang and co-workers the positional effects of pendent amide groups were investigated. It was found that orientation toward the active site and positioning above the active site were both important for catalytic enhancement, as these stabilized catalytic intermediates through hydrogen bonding and facilitated efficient proton transfer (Nichols et al., 2018b). This was quantified experimentally through equilibrium binding constants for CO<sup>2</sup> determined by rapid-scan CV techniques. When a pendent amide group was attached to the meso-phenyl of the porphyrin in the ortho position, [(Fe(o-2-amide-TPP)]+, the highest catalytic activity was observed. In comparison with the other reported derivatives, a "Goldilocks" relationship was observed, where positioning the amide group either closer or farther did not result in comparable catalytic current enhancement (**Figure 5**).

Nocera and co-workers found the pK<sup>a</sup> of the pendent proton source was important for the stabilization of CO<sup>2</sup> binding in the active site of related "hangman-type" porphyrin architectures (**Figure 6**) (Margarit et al., 2018). Pendent phenoland guanidinium-based hangman functional groups were predicted by DFT to cause a 2.1–6.6 kcal/mol stabilization of bound CO<sup>2</sup> within the hangman pocket. Conversely, a sulfonic acid derivative was found to function as a proton donor under the experimental conditions in the absence of applied potential, which was attributed to an estimated pK<sup>a</sup> of ∼3 under experimental conditions. The resultant anionic sulfonate derivative showed diminished catalytic activity relative to the porphyrins functionalized with phenol and guanidinium. The anionic charge of the conjugate base, combined with the overall steric bulk of the sulfonate, was proposed to prevent CO<sup>2</sup> binding within the hangman cleft and contribute minimally to the stabilization of the CO<sup>2</sup> adduct (Margarit et al., 2018). The pendent phenol-based "hangman-type" architecture was predicted to have the greatest stabilization of CO<sup>2</sup> binding by DFT calculations and was observed to have the fastest catalytic rate constant by CV methods.

Others have examined the relationship between electrocatalytic activity and the H-bonding ability of pendent residues on triazole-based picket-fence-type [Fe(TPP)]<sup>+</sup> derivatives with pendent amide functional groups (Sen et al.,

2019). In this report, a correlation between log(TOFmax) and the pK<sup>a</sup> of pendent proton donors was established for a picket fence amide with an encapsulated water molecule in comparison to previously reported [Fe(TPP)]<sup>+</sup> derivatives. The pendent groups are proposed to contribute to the stabilization of the Fe-bound CO<sup>2</sup> molecule through hydrogen-bonding interactions and facilitate proton transfer to mediate the rate-determining C–O bond cleavage step.

The studies discussed above have shown that pendent protons can be beneficial for [Fe(TPP)]+-based electrocatalysts, but also that careful consideration of steric constraints and the pK<sup>a</sup> of the chosen pendent proton donor under experimental conditions is necessary. Installation of pendent proton groups in porphyrin ligands can have a detrimental effect if their pK<sup>a</sup> and spatial orientation are not chosen carefully (e.g., sulfonic acid, vide supra). A recent study has also shown that the choice of solvent can largely alter the catalytic response of [Fe(TPP)]<sup>+</sup> with an asymmetrically substituted pendent proton moiety (Sinha and Warren, 2018). Using a porphyrin containing a single pendent hydroxy functionality in the 2-position of a single meso-phenyl ring, it was demonstrated that the electrocatalytic activity of this system can be greatly hindered by utilizing solvents with strong hydrogen-bonding properties through a comparison of activity in MeCN, DMF, and DMSO. Strong hydrogen bond acceptor solvents like DMF and DMSO lead to a decrease in activity of the catalyst, while in MeCN, the activity approached that reported for [Fe((ortho-TMA)TPP)]5+, which is the fastest molecular electrocatalyst for CO<sup>2</sup> reduction reported to date. The authors postulated that the interference of competitive hydrogen bonding between solvent molecules leads to slowed electrochemical kinetic parameters (Sinha and Warren, 2018). This suggests that more rigorous interrogations of functional group cooperativity (specifically the number and type of functional groups) and the interactions of functional groups with solvent and proton donors will offer additional insight into these mechanisms. Indeed, Costentin and Savéant described the origin of catalytic enhancement by pendent proton relays (Costentin and Savéant, 2018b). Boosting of electrocatalytic activity by pendent proton donors occurs when the forward rate constant of proton transfer from the pendent group to the active site is competitive with that of exogeneous proton donor directly to the active site (under the assumption that reprotonation of the pendent donor is extremely rapid). Should these forward rate constants not be well-matched, inefficiency in the proton relay mechanism will limit the ultimate catalytic current enhancement.

Further studies on this class of catalysts have modulated the potential of the FeI/<sup>0</sup> reduction where the catalytic response occurs by examining electron-withdrawing perfluorophenylsubstituted tetraphenylporphyrins ([Fe(F5TPP)]+, [Fe(F10TPP)]+, and [Fe(F20TPP)]+, where F<sup>5</sup> corresponds to a single perfluorinated phenyl ring, etc.) (Azcarate et al., 2016). This is a purely electronic effect, perturbing the continuum of electronic distribution in the ligand–metal manifold. The inclusion of proximal ionic charges has also been explored with trimethylanilinium- ([Fe(-ortho-TMA-TPP)]5<sup>+</sup> and [Fe( para-TMA-TPP)]5+) and sulfonato-functionalized porphyrins ([Fe(-para-sulfonato-TPP)]3−) to understand the role of through-space electrostatic interactions (Azcarate et al., 2016). The perfluorinated derivatives demonstrated a relationship between the catalytic response and the FeI/<sup>0</sup> potential: as the potential of that redox couple becomes more positive, log(TOFmax) decreases in a linear fashion, corresponding to a relative decrease in added electron density at the metal center upon electrochemical reduction (Azcarate et al., 2016; Costentin et al., 2018). However, for [Fe(TPP)]<sup>+</sup> substituted with a charged functional group (para–SO2<sup>−</sup> 3 , para–NMe<sup>+</sup> 3 , and ortho–NMe<sup>+</sup> 3 ), log(TOFmax) increases linearly as the redox potential of the FeI/<sup>0</sup> couple shifts to more positive potentials. [Fe-ortho-TMA-TPP]5<sup>+</sup> has the most positive FeI/<sup>0</sup> redox potential of any [Fe(TPP)]+-based electrocatalyst reported to date, as well as largest log(TOFmax) value. To understand this, one can once again imagine the CO<sup>2</sup> binding mode wherein a single electron generates radical CO•− 2 anion, which is stabilized by the charged groups close to the active site. It is compelling that such dramatic enhancements should be observed, but this could suggest that the reaction pathway might also be significantly altered from other [Fe(TPP)]<sup>+</sup> derivatives, as the effect is an inverse scaling relationship to that predicted for purely inductive reasons (Azcarate et al., 2016).

The data points from charge-based functionalization are relatively limited; the proposal of design principles for new systems for molecular CO<sup>2</sup> reduction systems requires additional information on these effects. For instance, the synthesis of further positional isomers of charged systems could give more evidence that a scaling relationship exists based on the distance between the charged moiety and the active site. Ultimately, it is clear that the inclusion of charge should seek to explore deliberate manipulation of the known mechanism to achieve the greatest enhancement effect.

### Re(bpy)(CO)3X

Re(bpy)(CO)3X (where bpy is a 2,2′ -bipyridine, often additionally functionalized in a symmetric fashion, and X is a halide anion or solvent molecule) is active for the electrocatalytic reduction of CO<sup>2</sup> to CO in near quantitative fashion (Hawecker et al., 1984; Grice and Kubiak, 2013). Under Faradaic conditions, this complex is proposed to undergo reduction according

to **Figure 7**. Initial reduction is localized at the bpy ligand, followed by loss of Cl<sup>−</sup> from [Re(bpy)(CO)3Cl]<sup>−</sup> via an overall EC mechanism to generate the neutral five-coordinate species [Re(bpy)(CO)3] 0 . At this stage, two separate mechanisms for reduction can occur: 1) a single reduction assigned to the ligand framework can occur, generating the catalytically active monoanionic species [Re(bpy)(CO)3] <sup>−</sup>. Commonly, the [Re(bpy)(CO)3Cl]<sup>−</sup> species is stable long enough on the CV timescale that [Re(bpy)(CO)3] <sup>−</sup> forms instead at the second reduction with Cl<sup>−</sup> loss. 2) Following initial single-electron reduction and Cl<sup>−</sup> loss, a Re–Re bond between two equivalents of [Re(bpy)(CO)3] 0 can form to generate Re(bpy)(CO)3]2. Dimer formation requires two sequential reductions to cleave the Re–Re bond and form the [Re(bpy)(CO)3] <sup>−</sup> active species. The formation of the metal–metal bond is slow under most conditions because of the persistence of [Re(bpy)(CO)3Cl]<sup>−</sup> on the CV timescale and is most often outcompeted by the unimolecular pathway to [Re(bpy)(CO)3] <sup>−</sup> (Grills et al., 2014).

There are also two possible mechanisms for CO<sup>2</sup> reduction under electrocatalytic conditions. First, a relatively slower bimolecular process involving two equivalents of [Re(bpy)(CO)3] 0 can occur, resulting in a net reductive disproportionation of two equivalents of CO<sup>2</sup> into one each of CO and CO2<sup>−</sup> 3 (Sullivan et al., 1985). Second, a unimolecular catalytic process with [Re(bpy)(CO)3] <sup>−</sup> as the resting state, where the two-electron reduction of CO<sup>2</sup> occurs at a single Re center (Sullivan et al., 1985; Keith et al., 2013). As is observed with the [Fe(TPP)]+-based electrocatalysts above, there are observable enhancements in TOFmax and catalyst stability for CO<sup>2</sup> reduction upon the addition of a sacrificial proton donor (Wong et al., 1998).

In the proposed mechanism for electrocatalytic CO<sup>2</sup> reduction, a monoanionic five-coordinate species [Re(bpy)(CO)3] <sup>−</sup> is invoked as the active species which binds CO<sup>2</sup> (**Figure 8)** (Keith et al., 2013). This has been validated experimentally through direct synthesis of [Re(bpy)(CO)3] <sup>−</sup> using chemical-reducing agents, with subsequent spectrochemical studies demonstrating a kinetic preference for CO<sup>2</sup> over H<sup>+</sup> (Smieja et al., 2012; Sampson et al., 2013). Re(bpy)(CO)3(η 1 -CO2H) is initially formed upon the binding and activation of CO<sup>2</sup> by [Re(bpy)(CO)3] <sup>−</sup> with a proton donor present. Further reduction generates an anionic species, [Re(bpy)(CO)3(η 1 -CO2H)]−, at which point C–O bond cleavage is facilitated by an exogenous proton source to generate water and the neutral species, [Re(bpy)(CO)4] 0 . CO release from the 19e<sup>−</sup> complex [Re(bpy)(CO)4] 0 is facile, and additional reduction regenerates the resting [Re(bpy)(CO)3] − state. In the Re(bpy)(CO)3X catalyst family, most work has focused on modulating the steric and electronic properties of the bipyridine ligand in attempts to modulate the reducing power and activity of Re. Generally, the use of electron-rich bpy ligands like 4,4′ -di-tert-butylbpy enhances the activity of the catalyst by creating a more nucleophilic Re center upon reduction, albeit at larger overpotentials (Clark et al., 2018).

Kubiak and co-workers have demonstrated, however, that steric considerations can supersede electronic effects in this class of catalysts. Re(bpy)(CO)3X-type complexes were synthesized with 3,3′ - and 5,5′ -dimethyl bpy (Chabolla et al., 2014). The methyl groups in those ring positions have identical Hammett parameters, so for purely electronic reasons, their respective electrocatalytic activities with a [Re(CO)3] <sup>+</sup> core would be expected to be similar. When compared to the unsubstituted parent compound, however, the 5,5′ -dimethyl bpy complex shows increased catalytic current at slightly more negative potentials, while the 3,3′ -dimethyl bpy shows decreased catalytic current at nearly the same potential. This is rationalized through the steric hindrance from the methyl groups at the 3,3′ -positions, which is observed in the non-planarity of the bpy ligand in the crystal structure of the precatalyst. Previous experimental and computational studies have postulated that a key mechanistic

component is the participation of π <sup>∗</sup> orbitals in the bpy ligand in the redox mechanism, suggesting that an inability to achieve a higher degree of planarity upon reduction diminishes catalytic activity for 3,3′ -dimethyl bpy-based Re catalysts (Smieja et al., 2012; Benson et al., 2013; Keith et al., 2013; Chabolla et al., 2014). Since the distribution of added electron density between the bpy ligand and the Re metal center has been identified as important to the selectivity for CO<sup>2</sup> over H<sup>+</sup> as the electrophile of choice; it is noteworthy that an steric perturbation of the HOMO of the active species should have this effect.

The effect of secondary-sphere modifications on the Re(bpy)CO3X system was explored through the incorporation of –OH groups at the 4,4′ - and 6,6′ -positions of the bpy, with the goal of using them both as a pendent proton source and to contribute electron density to the bpy ligand (Manbeck et al., 2015). In this case, the authors found atypical behavior during electrocatalysis: the O–H bonds of the bipyridine ligand are cleaved by stepwise electrochemical reduction, which they propose leads to dearomatization of the doubly reduced bpy ligand. By isolating the deprotonated species in a chemical fashion, the authors were able to give both UV-vis and IR spectroscopic evidence—which aligned well with infrared spectroelectrochemistry (IR-SEC), UV-vis spectroelectrochemistry (UV-vis-SEC), and DFT calculations—supporting this hypothesis. Interestingly, only the 4,4′ -dihydroxybpy complex is active as an electrocatalyst for CO<sup>2</sup>

reduction, while the 6,6′ -dihydroxybpy complex completes only ∼1 TON for CO, despite having nearly identical electrochemical properties under Faradaic conditions by CV. This result suggests slow CO release and decomposition of intermediates at applied potentials limit activity for the 6,6′ -dihydroxybpy derivative.

Further work with monomeric Re(bpy)(CO)3X electrocatalysts showed promising results with the incorporation of peptide linkages of varying lengths containing proton relays and hydrogen-bonding groups on the bpy backbone (Chabolla et al., 2017). Through electrochemical experiments, 2D NMR spectroscopy, and molecular dynamics modeling, the study showed that an optimal chain length of five peptides allows for the peptide backbone to adopt conformations which allow for intramolecular interactions on the NMR timescale. Importantly, this study showed the Re(bpy)(CO)3X system to be stable under peptide synthesis conditions, allowing for insertion of the complex into peptide linkages at any desired point (Chabolla et al., 2017).

The asymmetric incorporation of a thiourea tether into Re(bpy)(CO)3X systems has been shown to be a successful strategy for enhancing the CO<sup>2</sup> reduction activity (Haviv et al., 2018). Interestingly, the sulfur atom of the thiourea tether was shown to bind CO<sup>2</sup> prior to reduction of the Re catalyst, which is expected to lower the reorganization energy penalty expected for the reduction of the linear CO<sup>2</sup> molecule. This interaction was shown to work synergistically with the reduced Re state, as the

thiourea moiety also enhanced the catalytic response by acting as a pendent proton donor capable of beneficial hydrogen-bonding interactions and facilitating C–O bond cleavage. Added Brønsted acids inhibited catalytic activity, likely the result of competitive interactions with the pendent thiourea tether. In a separate study on the asymmetric incorporation of phenolic pendent proton sources to Re(bpy)(CO)3X, catalysis was 'turned on' at lower overpotentials, specifically the first reduction potential by CV in the two complexes studied (Rotundo et al., 2019). These complexes did, however, suffer from low Faradaic efficiencies for CO when no external proton source was present. Each of these catalysts also had issues with electrode absorption phenomena, a deleterious reaction with inhibitory consequences for nominally molecular electrocatalysts.

A Re(bpy)-based CO<sup>2</sup> reduction catalyst which features an imidazolium group as a charged residue in the secondary sphere reported by Nippe and co-workers was observed to cause changes in redox properties and mechanism compared to unfunctionalized Re(bpy)(CO)3X (Sung et al., 2017). It was proposed that the C2-H carbon of the imidazolium moiety was important for the catalytic enhancement through an alteration of mechanism: theoretical methods suggested that hydrogen bond-like or electrostatic C2-H—X (X <sup>=</sup> Cl−, CO<sup>−</sup> 2 , or H2O) interactions change the ground state energies of intermediates relevant to the catalytic cycle. These assignments were supported experimentally through testing of a control complex where the imidazolium C2-H was replaced by C2-CH3, and the non-linear dependence of the catalytic activity of each complex on [H2O], which is anomalous to the archetypal electrocatalytic response of Re(bpy)(CO)3X. It was postulated that a reduction-first mechanism for CO<sup>2</sup> reduction was occurring, where reduction of the Re(I)(η 1 -CO2H) adduct preceded protonation and C– O bond cleavage, as is classically seen in the unfunctionalized complex (**Figure 8**).

The effects of charge on the electrocatalytic activity of Re(bpy)(CO)3X were studied in a series of polymeric frameworks using a series of charged monomers (Sahu et al., 2017). Three norbornenyl-based polymers containing either positively charged quaternary ammonium, neutral phenyl, or negatively charged trifluoroborate moieties were generated through ROMP and covalently end-labeled with a Re(bpy)(CO)3X-based terminating reagent. Electrochemical studies in acetonitrile indicated that the polymers containing quaternary ammonium salts exhibited catalytic behavior at a significantly more positive potential (∼300 mV) than the neutral polymer, which behaved consistently with unfunctionalized Re(bpy)(CO)3X. The incorporation of negatively charged groups caused a shift to more reducing potentials, and catalytic activity was not observed in the solvent window. The incorporation of known catalysts onto a polymeric framework with the ability to tune reduction potential is a possible precursor to highly ordered structures such as thin films, abiotic metalloproteins, porous catalytic membranes, and cationic nanoparticles for use in devices.

# Mn(bpy)(CO)3X

Unpublished results referenced by Johnson et al. in 1996 stated that Mn(bpy)(CO)3X, unlike its third-row congener Re, was inactive for electrocatalytic CO<sup>2</sup> reduction under aprotic conditions (Johnson et al., 1996). A more recent examination by Deronzier and co-workers in 2011 repeated this result, but also showed that the addition of a weak proton donor facilitated a significant and selective electrocatalytic response for CO<sup>2</sup> reduction to CO (Bourrez et al., 2011). One important mechanistic detail about the Mn(bpy)(CO)3X system is that upon reduction by a single electron and subsequent loss of X, a Mn–Mn dimer [Mn(bpy)(CO)3]<sup>2</sup> can rapidly form with rates approaching the diffusion limit (**Figure 9**) (Grills et al., 2014). The formation of this dimer has two detrimental effects: (1) it increases the electrochemical driving force required to generate the catalytically active monoanionic five-coordinate [Mn(bpy)(CO)3] <sup>−</sup> species required for CO<sup>2</sup> reduction (the Mn– Mn bond requires more reducing potentials to cleave), and (2) it reduces the activity of the complex toward CO<sup>2</sup> reduction as a non-catalytic competing pathway (Grice and Kubiak, 2013; Smieja et al., 2013).

In attempts to combat these detrimental effects from dimerization, a bulky bpy analog was designed (6,6′ dimesitylbpy) (Sampson et al., 2014; Sampson and Kubiak, 2016). This ligand framework proved to be effective in eliminating the dimerization reaction. Rather than two irreversible 1e<sup>−</sup> waves on the reductive sweep as seen in Mn(bpy)(CO)3X, a single reversible 2e<sup>−</sup> wave was observed (Sampson et al., 2014). This leads to the formation of monomeric, anionic [Mn(mesbpy)(CO)3] <sup>−</sup> at 300 mV more positive potentials than in the case of the original Mn(bpy)(CO)3X complexes (Sampson et al., 2014). In the presence of CO<sup>2</sup> and a proton source, the 2e<sup>−</sup> reversible feature becomes irreversible and shifts toward positive potentials, which is indicative of CO<sup>2</sup>

binding (Sampson et al., 2014). This mechanistic difference was confirmed through control experiments, IR-SEC (Ashley and Pons, 1988; Zavarine and Kubiak, 2001; Best et al., 2008; Kaim and Fiedler, 2009; Machan et al., 2014b), and the direct synthesis of the active species. A significant catalytic response does not occur in the presence of Brønsted acids until potentials are similar to those that are catalytic for the unfunctionalized parent complex (Sampson et al., 2014). Rapid C–O bond cleavage in the hydroxycarbonyl intermediate [Mn(mesbpy)(CO)3(η 1 - CO2H)] does not occur until "overreduction" to generate [Mn(mesbpy)(CO)3(η 1 -CO2H)]<sup>−</sup> at potentials 400 mV more negative than the 2e<sup>−</sup> reversible feature (**Figure 10**) (Riplinger et al., 2014). To take advantage of the initial CO<sup>2</sup> binding event by [Mn(mesbpy)(CO)3] <sup>−</sup> at more positive potentials, a subsequent report used Lewis acid additives (Sampson and Kubiak, 2016). This strategy proved successful, as the addition of Mg2<sup>+</sup> ions as cosubstrate to solution aided in C–O bond cleavage at the potential where CO<sup>2</sup> binding occurs, greatly reducing the overpotential required for the generation of CO from CO<sup>2</sup> (Sampson and Kubiak, 2016).

Initial attempts to incorporate pendent proton relays into the Mn(bpy)(CO)3X framework involved asymmetric attachment of phenol moieties onto the bpy ligand (Franco et al., 2014; Agarwal et al., 2015). This led to enhanced activity in comparison to the parent Mn(bpy)(CO)3X complex, including activity in the absence of a proton source; however, the competitive dimerization reaction was still apparent in each case (Franco et al., 2014; Agarwal et al., 2015). Converting the pendent –OH groups to ethers with methoxy groups showed insignificant or no catalytic activity in control studies for these derivatives.

To extend these studies to symmetrically functionalized systems, Rochford and co-workers synthesized a methoxy ether-containing analog of the bulky bpy ligand, 6,6′ -(2,6 dimethoxyphenyl)bpy [(MeO)2Ph]2bpy (Ngo et al., 2017). A combined experimental and theoretical analysis showed that through the introduction of sufficiently acidic exogeneous proton sources, the slower "protonation-first" pathway seen in the original bulky bpy study could be "turned on" at lower overpotentials in comparison to the "reduction-first" pathway at more negative potentials (Ngo et al., 2017). In the protonation-first pathway, a bound hydroxycarbonyl is formed, and further protonation by a sufficiently strong acid can lead to the facilitation of C–O bond cleavage prior to reduction and release of CO at more positive potentials than the reductionfirst pathway (**Figure 10**). In the reduction-first pathway, a bound hydroxycarbonyl is again formed; however, without a sufficiently

strong acid present, reduction of the hydroxycarbonyl must occur at more negative potentials prior to protonation and C–O bond cleavage.

Mn((OMe)2bpy)(CO)3X. Adapted with permission from Ngo et al. (2017). Copyright American Chemical Society 2017.

The acid dependence of the "protonation-first" and "reduction-first" pathways for Mn([(MeO)2Ph]2bpy)(CO)3X was rationalized through a hydrogen-bonding interaction between the Mn-bound η 1 -CO2H hydroxycarbonyl species and the pendent Lewis base groups located on the ligand (**Figure 11**). This hydrogen bond donor–acceptor interaction was also proposed to facilitate the subsequent transfer of a second equivalent of a sufficiently strong Brønsted acid additive to protonate and facilitate C–O bond cleavage, ultimately producing H2O and Mn-bound CO (Ngo et al., 2017). In the presence of a proton source of insufficient acidity, the reduction of the bound η 1 -CO2H hydroxycarbonyl species is instead required before protonation and C–O bond cleavage can occur, as is observed with the "overreduction" of the Mn catalyst with the "bulky" bpy ligand (Ngo et al., 2017). Importantly, no metal–metal dimerization reaction was observed for this ligand.

Recently, a charged imidazolium functionality was also introduced asymmetrically into the Mn(bpy)(CO)3X system (Sung et al., 2019). A series of derivatives examining hydrogenbonding effects showed activity for CO<sup>2</sup> reduction at potentials of only −1.4 V vs. Fc+/Fc in the presence of water. This was postulated to originate from a synergistic effect between the pendent imidazolium functionality and a network of water molecules in the solvation shell that facilitated CO<sup>2</sup> reduction. The combined use of charge and hydrogen-bonding donors shows promise for lowering the catalytic overpotential, but further work is needed to enhance Faradaic efficiencies to match other catalyst platforms in this family.

# Ni(cyclam)]2<sup>+</sup>

The electrocatalytic activity of a [Ni(cyclam)]2<sup>+</sup> (cyclam = 1,4,8,11-tetraazacyclotetradecane) derivative was first reported in 1980 by Fisher and Eisenberg (1980). It has been the subject of continuous study because of its tolerance for a wide range of acid strengths and solvent systems without a

loss of catalytic activity (Beley et al., 1984; Barefield et al., 1986; Balazs and Anson, 1993; Kelly et al., 1999; Schneider et al., 2012). The activity of Ni(cyclam)]2<sup>+</sup> for electrocatalytic reduction of CO<sup>2</sup> to CO in the presence of water as a Brønsted acid source was attributed specifically to the Trans I isomer (one of six possible isomers based on the orientation of the H atoms on the four metal-coordinated secondary amines in the macrocycle; all four H atoms are cofacial). This isomer is the most favorable for high CO<sup>2</sup> reduction activity due to the hydrogen bond donor–acceptor interactions between the ligand NH groups and a Ni-bound CO<sup>2</sup> molecule (Froehlich and Kubiak, 2012). Furthermore, DFT calculations have indicated that the trans I isomer has a more stable CO<sup>2</sup> adduct than the trans III isomer (two H atoms are cofacial) by approximately 3 kcal/mol (Song et al., 2014).

Further evidence for the importance of the amine protons and these hydrogen-bonding interactions is found through a comparison with the N-alkylated derivatives [Ni(dimethylcyclam)]2<sup>+</sup> and [Ni(tetramethylcyclam)]2+, where greatly diminished activity for electrocatalytic CO<sup>2</sup> reduction is observed (Froehlich and Kubiak, 2012). The absence of hydrogen bond-induced stabilization effects and increased steric parameters make CO<sup>2</sup> binding less facile in the N-alkylated derivatives (**Figure 12**). Pendent proton donor effects have been shown with other functional groups as well: the introduction of a carboxylic acid on the carbon backbone of Ni(cyclam)]2<sup>+</sup> improves its activity by making it stable and selective for CO<sup>2</sup> reduction down to pH 2 at similar overpotentials to other reported water-soluble CO<sup>2</sup> reduction catalysts (Neri et al., 2016).

The use of an exogenous urea additive with Ni(cyclam)]2<sup>+</sup> was shown to improve catalytic activity for CO<sup>2</sup> reduction for similar reasons (Nichols and Chang, 2018). This study demonstrated that the urea additive acted as a cocatalyst for the system, and it was proposed that its unique structure allows for it to promote CO<sup>2</sup> reduction through the formation of multipoint hydrogen bonds with the bound CO<sup>2</sup> adduct [Ni(cyclam)(η 1 -CO2)]<sup>+</sup> (Nichols and Chang, 2018). This conclusion was supported through the introduction of multiple cationic and neutral additives with similar pKas. In each case, the cocatalytic response observed with urea was not present (Nichols and Chang, 2018). Although the urea additive was not tethered to cyclam, it is conspicuous that it should have a cocatalytic role with both Ni(cyclam)]2+- and Re(bpy)-based catalysts for CO<sup>2</sup> reduction, vide supra.

### Multimetallic Systems

Multimetallic systems are important examples of secondarysphere interactions, because they are common to the enzymes which catalyze CO<sup>2</sup> reduction chemistry (Dobbek et al., 2001b, 2002). In many cases, these multimetallic systems work cooperatively to both store excess charge and to activate the molecule of interest using "push–pull" donor–acceptor effects. Homobimetallic cofacial [Fe(TPP)]<sup>+</sup> systems connected by a phenylene bridge were synthesized to generate a bimetallic species from the well-known [Fe(TPP)]<sup>+</sup> system discussed above (Mohamed et al., 2015). By tuning the Fe–Fe distance through synthetic modification, CO<sup>2</sup> binding could be induced at the Fe3+/2<sup>+</sup> wave rather than the Fe1+/<sup>0</sup> as observed in most [Fe(TPP)]<sup>+</sup> systems (Mohamed et al., 2015; Bonin et al., 2017). This was rationalized through the expected Fe–Fe distance of 3.2– 4.0 Å in the ortho-bridged system, which would be suitable for binding the linear CO<sup>2</sup> molecule. In contrast, the meta-bridged system was expected to have a significantly shorter separation and showed diminished activity similar to monomeric [Fe(TPP)]<sup>+</sup> (Mohamed et al., 2015). The ortho-bridged system has significant catalytic activity (TOFmax = 4300 s−<sup>1</sup> ) and high Faradaic efficiency for CO (95%) at an overpotential of ∼0.7 V in the presence of 10% H2O in DMF, which is a significant improvement over monomeric [Fe(TPP)]+. The overpotential for this class of homobimetallic catalyst could be further tuned using electronwithdrawing and -donating substituents on the phenyl rings of each [Fe(TPP)]<sup>+</sup> unit (**Figure 13**) (Zahran et al., 2016). When the overpotential was synthetically tuned to ∼0.4 V with electronwithdrawing groups, a ∼3-fold decrease in activity in comparison to the parent phenyl-functionalized dimer was observed (Zahran et al., 2016). Chang and co-workers have recently demonstrated that these beneficial multimetallic effects extend beyond bimetallic systems using a porous organic cage containing six

[Fe(TPP)]<sup>+</sup> units, which was active for CO<sup>2</sup> reduction in aqueous solutions (Smith et al., 2018).

A multimetallic Fe carbonyl cluster system with an interstitial main group element [Fe4X(CO)12−nLn] <sup>−</sup> (X = N or C, L = another ligand, n = 1 or O) has been developed by the Berben group for the reduction of CO<sup>2</sup> to formate under both aqueous and non-aqueous conditions (Rail and Berben, 2011; Taheri et al., 2015; Taheri and Berben, 2016; Loewen et al., 2017). This cluster is proposed to generate formate via an intermediate bridging hydride. The bridging hydride on the cluster motif has the ideal hydricity and pK<sup>a</sup> to selectively generate formate in the presence of CO2, preventing the competitive formation of H<sup>2</sup> (Loewen et al., 2017). Pendent proton groups in this system alter the selectivity of this catalyst family from nearly quantitative generation of formate to the generation of H2, which highlights the importance of controlling the kinetics of substrate delivery in determining selectivity for competitive CO<sup>2</sup> and H<sup>+</sup> reduction reactions (Loewen et al., 2016).

The development of bimetallic Re complexes based on Re(bpy)(CO)3X and its derivatives has been of recent interest for both photocatalytic (Bruckmeier et al., 2012) and electrocatalytic processes due to the concentration-dependent formation of binuclear intermediates as part of the catalytic cycle (Machan et al., 2014a, 2016; Wilting et al., 2017; Yang et al., 2018). In electrocatalytic systems, the first bimolecular Re system studied utilized acetoamidomethyl modified bpy to generate a supramolecular catalyst system in situ. This system operated at more positive potentials (∼250 mV) in MeCN than the 4,4′ dimethylbpy-based control complex as the result of a hydrogenbonded dimer active state, albeit with a lower TOF and FE than the unimolecular process (Machan et al., 2014a). To further probe this hydrogen bond-based dimer system, a subsequent

report focused on a heterobimetallic Re–Mn construct using a 1:1 mixture of acetoamidomethyl-modified Re(bpy)(CO)3X with acetoamidomethyl-modified Mn(bpy)(CO)3X (Machan and Kubiak, 2016). Results from this study indicated a cooperative heterobimetallic pathway was operative, and it was proposed that the Mn center was activating CO<sup>2</sup> followed by protonation to generate a Mn-bound hydroxycarbonyl species within the heterobimetallic dimer. Enhanced reduction-first pathway kinetics were initiated in this case by electron transfer from the reduced Re species present in the dimer.

Further modifications to this supramolecular system replaced the acetoamidomethyl unit with an amide-linked PhOHcontaining tyrosine functional group (Machan et al., 2016). This modification led to an increased TOFmax in comparison to the initial complex and near quantitative Faradaic efficiency; mechanistic studies where the PhOH unit was substituted for a phenyl ring showed the pendent –OH functionality was essential for improving catalytic activity in the bimetallic mechanism (Machan et al., 2016). The "soft" non-covalent linkages used in this strategy are reminiscent of biological active sites: the catalytic system can adopt a variety of conformations on the potential energy surface facilitated by weaker interactions instead of more rigid systems reliant on distance or conformational tuning through purely synthetic means. The success of this approach is dependent on how well the weak interactions overcome the added diffusional component of the bimolecular reaction mechanism.

Work to generate rigid homobimetallic Re(bpy)(CO)3X systems where the metal centers are in close proximity has been achieved using an anthracene linker by Jurss and coworkers (Yang et al., 2018). This complex can be isolated as cis or trans isomers through chromatography, which alters the relative positioning of the Re centers (Yang et al., 2018). The cis isomer, in which the Re centers are in close proximity to one another, outperformed both a monometallic anthracene control complex and the trans isomer of the homobimetallic species in catalytic studies. Another covalently linked homobimetallic system with an imidazole–pyridine bridge was examined by Siewert and co-workers (Wilting et al., 2017). The observed catalytic activity of the homobimetallic species outperformed the mononuclear control complex, which was inactive for CO<sup>2</sup> reduction (Sinha et al., 2017). The introduction of the phenol linker between the two imidazole–pyridine arms to generate a proton relay in close proximity to the active site of the two metal centers was also examined (Wilting et al., 2017). The pendent proton source enhanced the activity of the bimetallic Re complex, with Faradaic efficiencies for CO of ∼60% (Wilting et al., 2017).

### EMERGING SYSTEMS

Several relatively new systems have also been reported where pendent functional group interactions are essential to the overall mechanism (**Figure 14**). A recent report on a series of cobalt(cyclopentadienyl)(P2N2) complexes by Artero and coworkers showed that pendent tertiary amines enabled selective formate generation from CO<sup>2</sup> (Roy et al., 2017). DFT calculations suggest that a reaction-defining transition state occurs where one of the pendent amines forms a hydrogen-bonding interaction with water to align it with CO<sup>2</sup> as simultaneous hydride transfer occurs from the Co center, lowering the overall transition state energy and enhancing activity (Roy et al., 2017). Work in our own group has identified a Schiff base-type ligand based on bpy as a promising new direction: the Fe(III) derivative is active for the reduction of CO<sup>2</sup> to formate with PhOH as a proton source. Mechanistic investigations suggest the Febound oxygen atoms act as a site for protonation upon initial reduction of the complex, generating a pendent proton source for the reaction at applied potential (Nichols et al., 2018a). A macrocyclic, aminopolypyridyl Co complex from Marinescu and co-workers generated CO in a near quantitative fashion in the presence of MeOH as a proton source (Chapovetsky et al., 2016, 2018). Upon alkylation of the pendent amine functionalities, a two-fold decrease in activity occurred, suggesting that the pendent protons on the amines linking the pyridyl groups play an important role as hydrogen bond donors during CO<sup>2</sup> reduction (Chapovetsky et al., 2016, 2018). The activity and selectivity of all these platforms are promising for future studies on optimizing secondary-sphere effects.

### CONCLUSIONS AND OUTLOOK

Molecular electrocatalysts for CO<sup>2</sup> reduction are of continuing interest for their possible utility in storing renewable energy in chemical bonds. Through careful mechanistic observations and design principles inspired by nature, researchers have managed to improve many of the known catalyst systems for CO<sup>2</sup> reduction. This iterative optimization of catalysts has demonstrated that the incorporation of pendent Brønsted acids/bases, charged groups, sterically bulky groups, Lewis basic sites, and the use of multimetallic sites with careful solvent choice can lead to improved catalytic activity and even new mechanisms through secondary-sphere effects reminiscent of biological systems.

With these successes in mind, it is useful to acknowledge that there is still much work to do: CODHs can reversibly interconvert CO2/CO in aqueous systems with a TOFmax of 0.5 s−<sup>1</sup> (CO<sup>2</sup> to CO) and 31,000 s−<sup>1</sup> (CO to CO2) with minimal overpotentials (Maynard and Lindahl, 2001). However, the use of secondary-sphere effects in abiotic molecular electrocatalysts has already been shown to be important for enhancing selectivity and activity. Further development in this area may lead to the development of catalysts capable of reducing products beyond CO and HCO2H, which are a burgeoning area for molecular electrocatalysts. The study of immobilized molecular Cu(TPP) systems has shown to be a successful strategy in generating higher ordered products like methane and ethylene in an electrocatalytic fashion (Weng et al., 2016), which is interesting, since crystalline copper electrodes are known to do this chemistry as heterogenous electrode materials (Kuhl et al., 2012). Bringing molecular design principles to materials seems to be another viable strategy for using secondary-sphere effects in electrocatalysis: Gong et al. (2017) immobilized porphyrin cages on Cu electrodes to tune activity and selectivity for carbon–carbon coupling products from CO<sup>2</sup> reduction through supramolecular effects. Recent reports have also described photocatalysts which convert CO<sup>2</sup> to methane from molecular catalysts related to those described here, which

### REFERENCES


is promising for developing eventual electrocatalytic behavior (Rao et al., 2017; Shirley et al., 2019).

Running CO<sup>2</sup> reduction reactions reversibly with abiotic systems could close the "loop" on the energy cycle, enabling the development of new fuel cell technologies beyond H<sup>2</sup> (Matsumoto et al., 2011, 2013; Kadirov et al., 2018). New catalyst systems which approach these biological efficiencies are unlikely to be rapidly found through purely synthetic routes due to the inherent depth and synthetic difficulty of the parameter space. Rather, the most efficient approach to future catalyst development should utilize a multidisciplinary approach which combines statistical, computational, and experimental methods to assist in the search for new CO<sup>2</sup> reduction catalysts unique from the current catalytic systems by improving predictive power.

### AUTHOR CONTRIBUTIONS

The manuscript was written through contributions of all the authors. All authors have given approval to the final version of the manuscript.

### FUNDING

Funding was obtained from the University of Virginia.

### ACKNOWLEDGMENTS

The authors thank the University of Virginia for the generous funding.


interfaces for electrocatalytic CO reduction. ACS Cent. Sci. 3, 1032–1040. doi: 10.1021/acscentsci.7b00316


CF3SO3) as catalyst precursors: infrared spectroelectrochemical investigation. Organometallics 15, 3374–3387. doi: 10.1021/om960044+


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Nichols and Machan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Carboxylation Reactions Using Carbon Dioxide as the C1 Source via Catalytically Generated Allyl Metal Intermediates

### Tetsuaki Fujihara\* and Yasushi Tsuji

*Department of Energy and Hydrocarbon Chemistry, Graduate School of Engineering, Kyoto University, Kyoto, Japan*

The use of carbon dioxide (CO2) is an important issue with regard to current climate research and the Earth's environment. Transition metal-catalyzed carboxylation reactions using CO<sup>2</sup> are highly attractive. This review summarizes the transition metal-catalyzed carboxylation reactions of organic substrates with CO<sup>2</sup> via allyl metal intermediates. First, carboxylation reactions via transmetalation are reviewed. Second, catalytic carboxylation reactions using allyl electrophiles and suitable reducing agents are summarized. The last section discusses the catalytic carboxylation reactions via addition reactions, affording allyl metal intermediates.

### Edited by:

*Hitoshi Ishida, Kitasato University, Japan*

### Reviewed by:

*Annaluisa Mariconda, University of Basilicata, Italy Tian-Sheng Mei, Shanghai Institute of Organic Chemistry (CAS), China*

> \*Correspondence: *Tetsuaki Fujihara tfuji@scl.kyoto-u.ac.jp*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *05 February 2019* Accepted: *27 May 2019* Published: *27 June 2019*

### Citation:

*Fujihara T and Tsuji Y (2019) Carboxylation Reactions Using Carbon Dioxide as the C1 Source via Catalytically Generated Allyl Metal Intermediates. Front. Chem. 7:430. doi: 10.3389/fchem.2019.00430* Keywords: carbon dioxide, carboxylic acids, allyl metals, transmetalation, oxidative addition, addition reactions

# INTRODUCTION

The development of fixation methods of carbon dioxide (CO2) is an important task for chemists (Aresta et al., 2014; Artz et al., 2018). However, CO<sup>2</sup> is a kinetically and thermodynamically stable material. Therefore, using CO<sup>2</sup> as a substrate poses several difficulties such as high-pressure or high-temperature reaction conditions. Transition metal-catalyzed carboxylation reactions of organic substrates with CO<sup>2</sup> can allow the diverse production of carboxylic acids and their derivatives under mild reaction conditions via carbon-carbon bond formation. During the last decade, diverse transition metal-catalyzed carboxylation reactions have been reported and good reviews have also been published (Huang et al., 2011; Tsuji and Fujihara, 2012; Cai and Xie, 2013; Zhang and Hou, 2013; Liu et al., 2015; Yu et al., 2015; Borjesson et al., 2016; Sekine and Yamada, 2016; Wang et al., 2016; Hazari and Heimann, 2017; Chen et al., 2018; Luan and Ye, 2018; Tortajada et al., 2018).

Allyl metal reagents are known for being excellent reagents during organic synthesis. The characteristic feature of allyl metal species is its isomerization via π-σ-π intermediates. σ-Allyl metal reagents act as nucleophiles, whereas π-allyl metals serve as electrophiles. When CO<sup>2</sup> is used as electrophiles for transition metal-catalyzed reactions, a method to generate a catalytically active nucleophilic allyl metal intermediate is highly reliable. **Scheme 1** shows three methods to access these allyl metal species from several organic substrates. The first method involves the transmetalation between allyl metal precursors such as allyl boranes or allylstannanes and transition metals (**Scheme 1A**). The second method involves the oxidative addition of allyl halides or allyl acetates using low-valent transition metals (**Scheme 1B**). In this case, suitable reducing agents must be used to regenerate low-valent active species. The third method involves the addition of metal species to conjugated dienes (**Scheme 1C**). In this reaction, the key is a regioselective addition that yields an ally metal intermediate selectively.

This paper overviews the transition metal-catalyzed carboxylation reactions of organic substrates with CO<sup>2</sup> via allyl metal intermediates. First, Pd- and Cu-catalyzed carboxylation reactions via transmetalation are summarized. Second, Pd- and Ni-catalyzed carboxylation reactions using ally electrophiles such as allyl acetates are reviewed. Finally, the catalytic carboxylation reactions via addition reactions are discussed.

# RESULTS AND DISCUSSION

# Transmetalation

Transmetalation is a straightforward method that allows accessing catalytically active allyl metal intermediates during a carboxylation event. A pioneering work has been done by Nicholas who carried out the reaction of allylstannane using Pd(PPh3)<sup>4</sup> as the catalyst in tetrahydrofurane (THF) at 70◦C under 33 atm of CO<sup>2</sup> (**Scheme 2A**, Shi and Nicholas, 1997). The reaction proceeds via allylstannane transmetalation, followed by a carboxylation reaction with CO2. Wendt observed an allyl Pd complex bearing a pincer ligand reacted with CO<sup>2</sup> under mild reaction conditions (**Scheme 2B**, Johansson and Wendt, 2007). In 2011, Hazari reported the Pd-catalyzed carboxylation of allylstannanes and allyl boranes, which were used as substrates (**Scheme 3**, Wu and Hazari, 2011). The reactions proceeded using a Pd complex with an N-heterocyclic (NHC) ligand as the catalyst under 1 atm of CO2. Under the same reaction conditions, several allylboronates were also converted to the corresponding carboxylated products in good yields.

Duong reported the Cu-catalyzed carboxylation of allylboronic esters with CO<sup>2</sup> (**Scheme 4**, Duong et al., 2013). The desired reactions proceeded in the presence of a Cu catalyst with IPr as the ligand in THF under 1 atm of CO2. Various allylboronic esters were subjected to the reaction, which generated corresponding products in moderate-to-high yields.

### Oxidative Addition

Known as an elemental reaction during a Tsuji–Trost reaction, the oxidative addition of ally halides or allyl esters with lowvalent metal species yields an ally metal intermediate (Trost and van Vranken, 1996; Trost and Crawley, 2003). Torii et al.

reported the Pd-catalyzed carboxylation of allyl acetates when exposed to electroreductive reductions (**Scheme 5**, Torii et al., 1986). The reaction of cinnamyl acetate with CO<sup>2</sup> in the presence of PdCl2(PPh3)<sup>2</sup> and PPh<sup>3</sup> afforded a mixture of regioisomers in 76% total yield. A branched substrate also afforded a mixture of regioisomers, indicating that the reaction proceeded via a π-allyl Pd intermediate. Dunach et al. found the Ni-catalyzed electrochemical carboxylation of allyl acetate with CO2. In the reaction, a mixture of two regioisomers was obtained (Medeiros et al., 2011). Mei et al. also found the Pd-catalyzed electrochemical carboxylation of allyl acetate, giving branched product selectively (Jiao et al., 2018).

Martin et al. reported the Ni-catalyzed ligand-controlled regiodivergent carboxylation of allyl acetates using manganese as the reducing agent (**Scheme 6**, Moragas et al., 2014). When **L1** was used as the ligand, the linear carboxylated products were obtained. In contrast, the reactions using **L2** produced branched carboxylated products in good-to-high yields. The oxidative addition of a low-valent Ni species affords an allylic Ni intermediate. The reaction with CO<sup>2</sup> yields Ni carboxylate intermediates; this is a regio-determining step for carboxylation.

Allyl alcohols are excellent reagents that allow the generation of allyl metal intermediates. However, in general, the hydroxyl group must be converted to the corresponding halides or esters before their use. Thus, the direct use of alcohols as the substrate for carboxylation reactions is important for both atom- and step-economical points. Mita and Sato found the Pdcatalyzed carboxylation of allyl alcohols in the presence of Et2Zn (**Scheme 7**, Mita et al., 2015). This reaction proceeded in the presence of a Pd catalyst with a PPh<sup>3</sup> ligand in a THF/hexane solution at room temperature under 1 atm of CO2. Diverse allyl alcohols were converted to the corresponding branched

allyl acetates with CO2.

carboxylic acids in good to high yields with high regioselectivity. Noteworthy is that a wide range of functional groups such as an ester and a free hydroxyl group was tolerated in the reactions. Regarding a reaction mechanism, allylic alcohols would be activated by ZnEt<sup>2</sup> followed by oxidative addition, leading to a π-allylpalladium intermediate (step 1). Then, nucleophilic σ-allylpalladium can capture CO<sup>2</sup> to generate a carboxylato Pd intermediate (step 3). Finally, transmetalation followed by reduction can regenerate Pd(0) active species (step 4).

Mei also reported that a Ni complex catalyzed the carboxylation of allyl alcohols using manganese as a reducing agent (**Scheme 8**, Chen et al., 2017). This reaction proceeded in the presence of an Ni complex with a neocuproine ligand in N.N-dimethylformamide (DMF) at room temperature under 1 atm of CO2. In this reaction, tetrabutylammonium acetate

SCHEME 8 | Ni-catalyzed carboxylation of allyl alcohols with CO2 using a manganese as reducing agent.

(TBAOAc) was essential; without the additive, the desired product could not be obtained at all. The reactions using both linear and branched allyl alcohols produced linear carboxylic acids in good-to-high yields with high regioselectivity. The reactions tolerated several functionality such as cyano and ester groups.

Bao found the Pd-catalyzed carboxylative coupling of benzyl chlorides and an allylstannane under CO<sup>2</sup> atmosphere (**Scheme 9**, Feng et al., 2013). Diverse benzyl chloride derivatives took part in the reactions, which produced the corresponding carboxylative coupling products in good-to-high yields. A key intermediate, the π-benzyl-π-allyl Pd complex, can be obtained by the oxidative addition of Pd(0) to benzyl chloride followed by transmetalation.

Transition metal-catalyzed carboxylation reaction via direct carbon-hydrogen (C-H) activation is a central research issue. Thus, the development of less reactive C-H carboxylation reactions is important. Mita and Sato observed a unique C-H carboxylation reaction with allylbenzene derivatives (**Scheme 10**, Michigami et al., 2017). The reaction of 4-phenyl allylbenzene was carried out in the presence of AlMe<sup>3</sup> (3 equivalent) using Co(acac)<sup>2</sup> and xantphos as the ligand in DMA at 60◦C under 1 atm of CO<sup>2</sup> in the presence of CsF as an additive. With these reaction conditions, a carboxylated product was obtained in 68% yield. Reactions were performed with diverse substrates, which produced corresponding carboxylated products in good-to-high yields. Notably, the reactions tolerated substrates having ester and ketone moieties, which are generally more reactive with nucleophiles than CO2. The reaction starts by generating a low-valent CH3-Co(I) species (**A**). The C-C double bond of the substrate coordinates to the cobalt center, and subsequent cleavage of the adjacent allylic C–H bond affords η 3 -allyl-Co(III) species **B** (step 1). Then, the reductive elimination of CH<sup>4</sup> from **B** affords a low-valent allyl-Co(I) species **C** (step 2). Next, C–C bond formation at the γ-position occurs via a reaction with CO2, giving a carboxylato Co species **D** (step 3). Finally, a linear carboxylated product is obtained by the transmetalation between **D** and Al(CH3)3, with the concomitant regeneration of **A** (step 4).

### Addition Reactions

Addition reactions are an important method to generate nucleophilic metal intermediates using unsaturated hydrocarbons. Among them, 1,2-dienes as well as 1,3-dienes are candidates accessing allyl metal intermediates. However, the regioselectivity of addition reactions must be controlled. Iwasawa et al. found hydrocarboxylation of 1,2-dienes (**Scheme 11**, Takaya and Iwasawa, 2008). These reactions proceeded via the hydropalladation of Pd–H to 1,2-dienes (step 1) followed by carboxylation (step 2). A P–Si–P coordinated pincer ligand was crucial for the reaction and AlEt<sup>3</sup> was employed as a reducing agent. With a similar Pd catalyst system, one-to-one coupling between 1,3-dienes and CO<sup>2</sup> also occurred (Takaya et al., 2011). The reaction between a Pd precursor and triethylaluminum reagents followed by β-hydrogen elimination produces a Pd–H intermediate. Then, hydropalladation of conjugated dienes yields allyl palladium intermediates, which can regioselectively react with CO2.

For Cu catalysts, we reported Cu-catalyzed hydrohydroxymethylation using 1,2-dienes as the substrate (**Scheme 12**, Tani et al., 2015). The reaction of 1,2-dienes with CO<sup>2</sup> (1 atm) and a hydrosilane was examined in the presence of a copper catalyst bearing xantphos derivatives as the ligand. Using 1.5 mmol of hydrosilane, which corresponds

to 2.0 equivalent for the required amount to yield the homoallyl alcohols, three H atoms were incorporated into the product. The corresponding homoallyl alcohols were obtained regioselectively, and the resulting terminal olefin moieties were preserved without hydrosilylation in the presence of an excess amount of hydrosilane. Various ester functionalities on the 1,2-dienes remained intact without reducing to their corresponding diols. Addition of a copper hydride **A** across a terminal double bond of allene **6** with the Cu atom at the less hindered site generates allylcopper intermediate **B** (step 1), which reacts with CO<sup>2</sup> regioselectively at the γ -position via a six-membered transition state to afford a copper carboxylate **C** (step 2). Then, carboxylate **C** could be further reduced by the hydrosilane to copper alkoxide **E** via a silyl copper ketal derivative **D** (steps 3 and 4). Finally, σ-bond metathesis

the product completely switched from homoallyl alcohols to β,γ unsaturated carboxylic acids (**Scheme 13**, Tani et al., 2015). A stoichiometric amount of base is essential to induce selective carboxylic acid formation. Thus, products at two different oxidation levels (alcohols and carboxylic acids) were selectively prepared by simply adjusting the base equivalence and ligands in the Cu-catalyzed reactions. Yu reported enantioselective hydrohydroxymethylation of 1,3-dienes using a copper catalyst (**Scheme 14**, Gui et al., 2017). The reaction was carried out using 1-aryl-1,3-butadiens as the substrates and a hydrosilane as the reducing agent in cyclohexane

regenerates (step 5).

between **E** and the hydrosilane provides silyl ether and **A**

Using a stoichiometric amount of base and ClIPr as the ligand,

1,3-dienes using a hydrosilane and CO2.

at room temperature in the presence of a copper catalyst bearing 3,5-di-tert-butyl-4-methoxyphenyl (DTBM)-segphos as the ligand. Diverse 1-aryl-1,3-butadienes were used in the

### REFERENCES


reactions, yielding the corresponding products in good-to-high yields with high enantioselectivity. The reactions tolerated several functional groups such as the chloro and amino groups.

Silylcupration to carbon–carbon multiple bonds is a reliable and powerful process, which produces both C–Si and C–Cu bonds. We found the Cu-catalyzed silacarboxylation of 1,2 dienes (**Scheme 15**, Tani et al., 2014). The reaction was carried out by using 1,2-dienes and PhMe2Si–B(pin) under CO<sup>2</sup> (1 atm) with a catalytic amount of Cu(OAc)2·H2O and rac-Me-DuPhos as the ligand. With these reaction conditions, diverse carboxylated vinylsilanes were regioselectively obtained in good yields. The functionalities, such as alkenyl and ester groups were tolerated under the reaction conditions. In the reactions, a β-silyl allyl copper intermediate (**B**) that was generated by the addition of silylcopper (**A**) across an allene (step 1) captured CO<sup>2</sup> (step 2). Then, a carboxylate copper (**C**) reacts with a silylborane to afford **A** (step 3). The choice of an appropriate ligand can efficiently control the regioselectivity. Carboxylated allylsilanes were also obtained using CuCl/AcONa and PCy<sup>3</sup> as the catalyst system.

### Conclusions

In conclusion, this short review summarized transition metalcatalyzed carboxylation reactions via allyl metal intermediates. Allylstannanes and allylboronates act as good reagents for Pdor Cu-catalyzed carboxylation reactions via transmetalation. Transition metal-catalyzed allyl electrophiles via oxidative addition are disclosed herein. Ni complexes successfully produce ligand-controlled regiodivergent carboxylation. Allyl alcohols are possible substrates for Pd- and Ni-catalyzed carboxylation. Substrates, such as 1,2-dienes or 1,3-dienes, are usable for the generation of allyl metal intermediates via the addition of metal reagents. These processes involve hydrocarboxylation, hydrohydroxymethylation, and silacarboxylation. These methods allow efficient syntheses of unsaturated carboxylic acids using CO<sup>2</sup> as the C1 source. In the future, these methodologies will be applied to synthesizing numerous valuable molecules in shorter steps using CO2.

### AUTHOR CONTRIBUTIONS

TF arranged and wrote the manuscript. YT discussed the contents and commented on the manuscript.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Fujihara and Tsuji. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# CO<sup>2</sup> Capture and in situ Catalytic Transformation

### Hong-Chen Fu1,2, Fei You<sup>2</sup> , Hong-Ru Li 1,2 \* and Liang-Nian He<sup>2</sup> \*

*<sup>1</sup> College of Pharmacy, Nankai University, Tianjin, China, <sup>2</sup> State Key Laboratory and Institute of Elemento-Organic Chemistry, College of Chemistry, Nankai University, Tianjin, China*

The escalating rate of fossil fuel combustion contributes to excessive CO<sup>2</sup> emission and the resulting global climate change has drawn considerable attention. Therefore, tremendous efforts have been devoted to mitigate the CO<sup>2</sup> accumulation in the atmosphere. Carbon capture and storage (CCS) strategy has been regarded as one of the promising options for controlling CO<sup>2</sup> build-up. However, desorption and compression of CO<sup>2</sup> need extra energy input. To circumvent this energy issue, carbon capture and utilization (CCU) strategy has been proposed whereby CO<sup>2</sup> can be captured and *in situ* activated simultaneously to participate in the subsequent conversion under mild conditions, offering valuable compounds. As an alternative to CCS, the CCU has attracted much concern. Although various absorbents have been developed for the CCU strategy, the direct, *in situ* chemical conversion of the captured CO<sup>2</sup> into valuable chemicals remains in its infancies compared with the gaseous CO<sup>2</sup> conversion. This review summarizes the recent progress on CO<sup>2</sup> capture and *in situ* catalytic transformation. The contents are introduced according to the absorbent types, in which different reaction type is involved and the transformation mechanism of the captured CO<sup>2</sup> and the role of the absorbent in the conversion are especially elucidated. We hope this review can shed light on the transformation of the captured CO<sup>2</sup> and arouse broad concern on the CCU strategy.

### Keywords: CO<sup>2</sup> capture, activation, conversion, in situ catalysis, green chemistry

# INTRODUCTION

The demand for energy of the rapid industrialization results in large-scale combustion of fossil fuel, which causes excessive emissions of carbon dioxide. As the detrimental environmental impacts of CO<sup>2</sup> have drawn considerable attention, various strategies have been developed to mitigate CO<sup>2</sup> accumulation in the atmosphere, among which carbon capture and storage/sequestration (CCS) is considered as a promising CO<sup>2</sup> reducing option (Alexander et al., 2015). Nowadays, a plethora of CO<sup>2</sup> absorbents have been developed to facilitate CO<sup>2</sup> capture and desorption. Nevertheless, the extensive energy needed in the absorbent regeneration and CO<sup>2</sup> separation is not conducive to the implementation of CCS strategy.

In contrast to carbon sequestration, converting CO<sup>2</sup> into valuable chemicals could be a sustainable option, which has been proposed by Ciamician as early as 1912 (Ciamician, 1912). In recent decades, CO<sup>2</sup> conversion has attracted considerable concern and been intensively investigated (Rahman et al., 2017). However, in most processes for CO<sup>2</sup> conversion, pure or high pressure CO<sup>2</sup> is needed, implying that the CO<sup>2</sup> from the atmosphere or industrial exhaust cannot be used as C<sup>1</sup> source directly and thus the energy issue in CO<sup>2</sup> capture and separation still remains.

### Edited by:

*Hitoshi Ishida, Kitasato University, Japan*

# Reviewed by:

*Salah S. Massoud, University of Louisiana at Lafayette, United States Marc Robert, Paris Diderot University, France Jinquan Wang, Institute of Bioengineering and Nanotechnology (A*∗*STAR), Singapore*

### \*Correspondence:

*Hong-Ru Li lihongru@nankai.edu.cn Liang-Nian He heln@nankai.edu.cn*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *11 January 2019* Accepted: *09 July 2019* Published: *24 July 2019*

### Citation:

*Fu H-C, You F, Li H-R and He L-N (2019) CO*2 *Capture and in situ Catalytic Transformation. Front. Chem. 7:525. doi: 10.3389/fchem.2019.00525*

**116**

To address the energy penalties associated with CCS strategy and realize the direct fixation of CO<sup>2</sup> from the atmosphere or industrial exhaust, the CO<sup>2</sup> capture and utilization (CCU) strategy, whereby the captured CO<sup>2</sup> is used as a non-toxic, abundant, and sustainable feedstock to produce valuable organic compounds via chemical, electrochemical or photochemical reactions, was proposed and now is flourishing (**Scheme 1**). By now, both organic compounds and functional materials containing the bridging-carbonato metal complexes can be obtained from atmospheric CO<sup>2</sup> using the CCU strategy (Yang et al., 2011; Liu et al., 2012; Massoud et al., 2015). Although realizing the attractive prospect of the CO<sup>2</sup> capture and in situ conversion in the industry scale remains a challenge (Zhang and Lim, 2015), the emergence of efficient absorbents and the development of CO<sup>2</sup> transformation will cast light on it. In continuation of our work on the conversion of the captured CO<sup>2</sup> into value-added organic chemicals, this review summarized the recent progress on CO<sup>2</sup> capture and in situ conversion into organic products.

To realize the carbon capture and in situ conversion strategy, effective absorbents are always necessary. Ideally, the absorbents for CCU strategy should not only capture CO2, but also activate CO<sup>2</sup> and even the substrate. Thus, the chemical transformation can proceed under mild conditions. Up to now, organic and inorganic bases, N-heterocyclic carbenes (NHCs) and Nheterocyclic olefins (NHOs), ionic liquids (ILs) and frustrated Lewis pairs (FLPs) have already been applied to CO<sup>2</sup> capture and in situ conversion. A plethora of valuable organic chemicals have been obtained through the CCU strategy as shown in **Scheme 2**.

### INORGANIC/ORGANIC BASES

Due to the electrophilicity of carbon atom in CO2, the organic and inorganic bases containing strong nucleophilic atom have been widely used in CO<sup>2</sup> trapping, where the base can interact with CO<sup>2</sup> directly or function as a proton acceptor. The resulting CO<sup>2</sup> capture products i.e., CO<sup>2</sup> adducts have been employed for subsequent synthesis of various valuable chemicals.

Considering the transformations of the captured CO<sup>2</sup> derived from primary and secondary amines and amino alcohols to isocyanates, carbamates, ureas, and oxazolidinones have been concerned by several excellent review papers (Hampe and Rudkevich, 2003; Chaturvedi and Ray, 2006; Yang et al., 2012; Tamura et al., 2014; Wang et al., 2017a,b), here we focus on the transcarboxylation effect and other transformations of the captured CO2, namely CO<sup>2</sup> derivatives.

### Synthesis of Carbamates and Ureas

In the synthesis of carbamates, the aprotic organic bases can function as CO<sup>2</sup> absorbents and transcarboxylation agents. The initial attempt was made by Rossi group, in which CO<sup>2</sup> is trapped by a methanol solution of commercially available tetraethylammonium hydroxide. The resulting tetraethylammonium hydrogen carbonate can be used as a surrogate of CO<sup>2</sup> in the synthesis of carbamate. Meanwhile, the presence of tetraethylammonium ion as counterion increases the nucleophilicity of carbamate anion (Inesi et al., 1998).

Soon after, Franco group has successfully identified the DBU-CO<sup>2</sup> complex via reacting CO<sup>2</sup> with DBU (1,8- Diazabicyclo[5.4.0]undec-7-ene) in anhydrous acetonitrile, implying that DBU can be used as CO<sup>2</sup> trap reagent (Pérez et al., 2002). Moreover, the resulting reactive DBU-CO<sup>2</sup> adduct can be utilized as transcarboxylating reagent for synthesis of N-alkyl carbamates. Later, the same group revealed the activation capacity of CO<sup>2</sup> by other bicyclic amidines and observed the inverse relation between the thermal stability and the transcarboxylating activity for the amidine-CO<sup>2</sup> adducts (**Scheme 3**) (Pérez et al., 2004), which is the first time to investigate the activation ability of organic bases to CO2.

The combination of organic base and alcohol is an efficient CO<sup>2</sup> capture system and the absorbed CO<sup>2</sup> can be in situ transformed. The prototypical example is the polyethylene glycol (PEG)/superbase system developed by our group in 2011 (Yang et al., 2011). In the capture step, the superbase is used as a proton acceptor and almost equimolar CO<sup>2</sup> per mole superbase can be absorbed (**Scheme 4**). The resulting liquid amidinium carbonate can directly react with n-butylamine at 110◦C to afford dibutyl urea in almost quantitative yield (96%) without any other additives. This protocol can be used in the synthesis of other symmetrical urea derivatives.

In the above examples, the captured CO<sup>2</sup> in the transcarboxylating agents can be regarded as the activated CO<sup>2</sup> because the linear structure of CO<sup>2</sup> is converted to bent structure, which is more liable to nucleophilic attack.

### Synthesis of Oxazolidinones

The "CO<sup>2</sup> absorption and subsequent transcarboxylation" triggers the research on CO<sup>2</sup> capture and in situ transformation. Several years later, M. Yoshida and coworkers use DBU to enrich and activate CO<sup>2</sup> in air and perform the first example of directly transforming atmospheric CO<sup>2</sup> into the substituted 5-vinylideneoxazolidin-2-ones using propargylic substrate 4- (benzylamino)-2-butynyl carbonates or benzoates as a substrate (**Scheme 4**) (Yoshida et al., 2008). In their follow-up work, they further improve the reaction efficiency by utilizing AgNO<sup>3</sup> as catalyst and propargylic amines as substrates (**Scheme 4**) (Yoshida et al., 2012).

Inspired by these works, our group designs a series of novel CO<sup>2</sup> capture and activation systems. For example, by employing ammonium iodide as catalyst, the cycloaddition reaction of various aziridines with the captured CO<sup>2</sup> by NH2PEG150NH<sup>2</sup> gives rise to oxazolidinones at 40◦C in >94% yield and selectivity (**Scheme 4**) (Yang et al., 2011).

Soon after, we report the first example of steric-hindrancecontrolled CO<sup>2</sup> absorption, where the sodium N-alkylglycinates and N-alkylalaninates dissolved in PEG<sup>150</sup> are used to capture CO2, generating the carbamic acid rather than the ammonium carbamate (Liu et al., 2012). N-isopropylglycinate is found to be the best absorbent for the rapid and reversible capture of almost equimolar CO2. Crucially, the captured CO<sup>2</sup> can be activated simultaneously and the resulting carbamic acid can react with either aziridine or propargyl amine to afford oxazolidinones in the presence of NH4I and AgOAc as a catalyst, respectively (**Scheme 4**).

Motivated by these results, we further develop potassium phthalimide as absorbent to realize equimolar CO<sup>2</sup> capture in PEG150. Moreover, the obtained product can be used as in situ transcarboxylating reagent to synthesize oxazolidinone derivatives (**Scheme 4**) (Zhang et al., 2014).

Recently, Hu group subtly designs a CCU example (Yu et al., 2016), in which carbamate salts generated from CO<sup>2</sup> and primary amines are used as substrates. The captured CO<sup>2</sup> not only acts as a reactant but also acts as a protecting reagent for the amine to avoid poisoning of the copper catalyst. By using 5 mol% of CuI as catalyst, carbamate salts can react with aromatic aldehydes and aromatic terminal alkynes, affording the important oxazolidin-2 ones (**Scheme 4**).

### Synthesis of β-Oxopropylcarbamates

Based on these inspiring results, our group uses ammonium carbamates as surrogates of carbon dioxide and secondary amines in the three-component synthesis of β-oxopropylcarbamates from propargylic alcohols, secondary amines, and CO2. Catalyzed by silver (I) catalyst, ammonium carbamates can react with propargylic alcohols to generate β-oxopropylcarbamates under atmospheric pressure (**Scheme 5**) (Song et al., 2015). In this example, the substitution of pure CO<sup>2</sup> with the captured CO<sup>2</sup> can facilitate the reaction running at atmospheric pressures with a broad substrate and reaction application scope. Furthermore, the solid ammonium carbamates are easier to handle and quantify than the volatile amines and gaseous CO2.

The transcarboxylation is an important transformation strategy for the CO<sup>2</sup> adducts to valuable chemicals. However, for the CO<sup>2</sup> adducts formed by base and CO2, the transcarboxylation is still limited to the substrates including amines, propargylamines and aziridines. Therefore, novel CO<sup>2</sup> absorbents and extended substrates are expected to facilitate the application of CO<sup>2</sup> adducts as transcarboxylation reagent in CCU strategy. Besides transcarboxylation, the integral transformation of CO<sup>2</sup> capture products is another attractive option in CCU strategy, wherein the ammonium carbamates derived from CO<sup>2</sup> and amines is a promising raw material. Nevertheless, the integral transformation of ammonium carbamates to valuable chemicals remains sporadic and underexplored. Hopefully, more conversion protocols of ammonium carbamates can be designed based on the reactivity of amine and CO2.

# CO<sup>2</sup> Capture and in situ Hydrogenation

CO<sup>2</sup> hydrogenation is widely investigated in the CCU strategy because the basic absorbent can react with the resulting formic acid to form formate, thus overcomes the thermodynamic limitation in the hydrogenation of CO2. In the researches on CO<sup>2</sup> capture and in situ hydrogenation, both metal-based homogeneous catalysts (containing Rh-, Ru-, and Fe-based catalysts) and heterogeneous catalysts have been investigated.

### Hydrogenation Using Rh-Based Catalysts

The first example of CO<sup>2</sup> capture and in situ hydrogenation is reported by our group in 2013, in which polyethyleneimine 600 (PEI600) or the combination of PEI<sup>600</sup> and ethylene glycol is developed to absorb gaseous CO2, affording the PEI-CO<sup>2</sup> or ethylene glycol-CO<sup>2</sup> adducts. With RhCl3·3H2O/CyPPh<sup>2</sup> as catalyst, the captured CO<sup>2</sup> can be in situ transformed to formate (**Scheme 6**) (Li et al., 2013). Furthermore, direct hydrogenation of ammonium carbamate derived from CO2, e.g., DETA+CO<sup>2</sup> <sup>−</sup> (DETA = diethylenetriamine), ammonium carbonates such as [DBNH] [OCO2(C2H4O)3H], [DBNH] [OCO2CH2OH] (DBN=1,5-diaza bicycle[4.3.0]non-5-ene), is also successfully performed facilitated by this Rh-based catalyst. Notably, a higher reaction rate and better results can be achieved when using the captured CO<sup>2</sup> in the form of ammonium carbonates as feedstock than using equivalent free gaseous CO<sup>2</sup> or ammonium carbamate, implying CO<sup>2</sup> activation upon capture with DBN/PEI and glycol.

After that, we further design a tunable ethoxyl-functionalized amidine to absorb CO<sup>2</sup> in order to avoid the use of volatile proton donor (Li et al., 2014). As an activated form of CO2, the captured CO<sup>2</sup> in the form of zwitterionic amidinium carbonate is further hydrogenated to formate employing RhCl3/DPEphos as the catalyst (**Scheme 6**). In the same time, we find that the CO<sup>2</sup> capture product of potassium phthalimide in PEG150, can also be in situ hydrogenated to formic acid catalyzed by RhCl3·3H2O/CyPPh<sup>2</sup> (**Scheme 6**) (Zhang et al., 2014).

### Hydrogenation Using Ru- and Fe-Based Catalysts

Ru-based catalysts are also promising candidates for the hydrogenation of captured CO2. In the study of Yadav et al., CO<sup>2</sup> is captured by DBU and an alcohol to form the alkyl carbonate ionic liquid and the resulting alkyl carbonate is hydrogenated into [DBUH+] formate and methyl formate facilitated by RuCl2(PPh3)<sup>3</sup> (**Scheme 7**) (Yadav et al., 2014). Although the reactive species (i.e., alkyl carbonates or CO2) cannot be identified at this stage, this result indicates that alkyl carbonates may be a substrate for hydrogenation, in addition to free CO2.

In 2015, Sanford group combined CO<sup>2</sup> capture to form a carbamate salt with hydrogenation to generate CH3OH (Rezayee et al., 2015). In their study, NHMe<sup>2</sup> is used to capture CO<sup>2</sup> and a homogeneous Ru-based catalyst is used to facilitate the hydrogenation of the captured CO<sup>2</sup> to a mixture of DMF and CH3OH (**Scheme 7**). Although the formation of carbamate salt can decrease the eletrophilicity of CO2, causing the captured CO<sup>2</sup> difficult to hydrogenate, the existence of the equilibrium between DMC (Dimethylammonium dimethylcarbamate) and

CO<sup>2</sup> allows the release of CO<sup>2</sup> possible, thus promoting the CO<sup>2</sup> hydrogenation.

By employing pentaethylenehexamine (PEHA) as CO<sup>2</sup> absorbent and Ru-based complexes as catalyst, Olah and Surya Prakash group develops a process that combines CO<sup>2</sup> capture and the following hydrogenation in an ethereal solvent for the production of MeOH (Kothandaraman et al., 2016a). CO<sup>2</sup> from air can be captured by an aqueous solution of PEHA and up to 61% yield of MeOH can be obtained in the triglyme/H2O mixtures at 155◦C in the following hydrogenation (**Scheme 7**). The resulting MeOH can be easily separated by simple distillation from the reaction mixture.

Later, the same group captures CO<sup>2</sup> with aqueous amine solution and then in situ hydrogenates the resulting ammonium bicarbonate/carbonate utilizing Ru- and Fe-based pincer complexes in a biphasic solvent system (water/Me-THF) (Kothandaraman et al., 2016b). The superbases (DABCO, TMG, and DBU) shows to be efficient for both CO<sup>2</sup> capture and hydrogenation with more than 90% yield of formate under moderate reaction conditions (50 bar H<sup>2</sup> at 55◦C) (**Scheme 7**). The biphasic system features easy separation of product and catalyst and the catalyst can be reused for at least five cycles.

### Hydrogenation Using Heterogeneous Catalysts

Besides homogeneous catalysts, the heterogeneous catalysts were also used in the hydrogenation of CO<sup>2</sup> capture products. For example, H. Lin group applies Pd/AC catalyst to the hydrogenation of CO<sup>2</sup> capture products originated from ammonia. In the hydrogenation step, the dependence of the activity of CO<sup>2</sup> capture products on the solvent is observed (Su et al., 2015a,b). For example, the ammonium bicarbonate in water and ammonium carbamate in 70 wt% ethanol-water solution can offer more than 90% yield of formate under high H<sup>2</sup> pressure (5.52 and 2.75 MPa, respectively) at 20◦C. The ammonium carbonate presents similar activity with ammonium carbamate. Identification of the species in the reactant solutions suggests the bicarbonate ion and ethyl carbonate ion, instead of the carbamate ion, are the activation forms of CO<sup>2</sup> in the hydrogenation (**Scheme 8**). Coincidently, Enthaler finds that sodium bicarbonate in methanol can be hydrogenated to sodium formate catalyzed by the nickel hydride complex while CO<sup>2</sup> cannot be hydrogenated in the identical conditions, which further confirms the activity of the captured CO<sup>2</sup> in hydrogenation (Enthaler et al., 2015).

Mertens and coworkers report the in situ hydrogenation of the captured CO<sup>2</sup> using Cu/ZnO-Al2O<sup>3</sup> as catalyst under retrieval of the CO<sup>2</sup> capture reagent N,N-diethylethanolamine (DEEA) (Reller et al., 2014). In the reaction, DEEA can also function as a trapping reagent for the resulting formic acid and drives the hydrogenation forward. The authors find that the generation of the products 2-diethylaminoethylformate and methanol can be regulated by the reaction temperature (**Scheme 9**). The combination of CO<sup>2</sup> capture and hydrogenation realizes the energy integration by using the reaction heat of CO<sup>2</sup> hydrogenation in the energy demanding CO<sup>2</sup> stripping process.

In addition to the liquid absorption system, the alkali metal and alkali earth metal based solid CO<sup>2</sup> adsorbents are also developed (Li et al., 2010, 2011; Lee et al., 2011) and applied in the CCU strategy recently. Duyar et al. design a series of novel dual function materials (DFM) consisting of the catalyst and adsorbent components to couple the endothermic CO<sup>2</sup> desorption step with the exothermic hydrogenation of CO<sup>2</sup> (Duyara et al., 2016). The results show that DFM with the composition of 5% Ru 10% K2CO3/Al2O<sup>3</sup> and 5% Ru

Frontiers in Chemistry | www.frontiersin.org

10% Na2CO3/Al2O<sup>3</sup> have a methanation capacity of 0.91 and 1.05 g-mol/kg DFM, respectively. Similarly, A. Urakawa group develops the catalyst consisting of earth-abundant chemical elements (FeCrCu/K/MgO–Al2O3), which can trap CO<sup>2</sup> from fuel gas in the form of surface carbonates and subsequently hydrogenated the adsorbed CO<sup>2</sup> to CO (Bobadilla et al., 2016). Accordingly, these DFMs are identified as promising candidates for CO<sup>2</sup> capture and direct utilizations.

Hydrogenation of captured CO<sup>2</sup> to energy chemicals can facilitate turning hydrogen gas to liquid fuel as well as realize carbon cycling. Albeit the hydrogenation of captured CO<sup>2</sup> has been extensively investigated and various capture reagents and catalysts have been developed, the identification of CO<sup>2</sup> activation forms is still controversial. For example, the CO<sup>2</sup> capture products alkyl carbonate ammonium salts are considered as the activated CO<sup>2</sup> species (Li et al., 2013, 2014; Su et al.,

2015a,b). However, the results of Jessop group show that [DBUH][OC(O)OMe] salt is less active than free CO<sup>2</sup> when using RuCl(O2CMe)(PMe3)<sup>4</sup> as catalyst in MeOH solution (Munshi et al., 2002). Thus, the relationship between the activity of CO<sup>2</sup> capture products and the catalyst is still underdeveloped.

### N-HETEROCYCLIC CARBENES AND N-HETEROCYCLIC OLEFINS

It has been verified that N-heterocyclic carbenes and Nheterocyclic olefins can react with CO2, forming the CO<sup>2</sup> adduct which can be used as "all-in-one" carboxylating agent (Zhou et al., 2008; Kelemen et al., 2014; Dong et al., 2015; Talapaneni et al., 2015; Finger et al., 2016; Saptal and Bhanage, 2016). For example, Tommasi group shows the CO<sup>2</sup> adduct 1-butyl-3-methylimidazolium-2-carboxylate and 1,3-dimethylimidazolium-2-carboxylate behaves as active CO2 carriers and reacted with CH3OH and acetophenone for the synthesis of methylcarbonate and benzoylacetate. The other organic compounds with active hydrogen (acetone, cyclohexanone, benzylcyanide, and propargyl alcohols) can also be carboxylated with these CO2-transfer agents for the synthesis of carboxylates of pharmaceutical interest (Tommasi and Sorrentino, 2005, 2006, 2009) (**Scheme 10**). Similarly, the transcarboxylation of IPrCO<sup>2</sup> (1,3-bis(2,6-diisopropylphenyl) imidazolum-2-carboxylate) to acetophenone with NaBPh<sup>4</sup> to yield sodium benzoylacetate and direct dicarboxylation of MeCN using ItBuCO<sup>2</sup> (1,3-bis(tert-butyl)-imidazolium-2-carboxylate) are also reported (Van Ausdall et al., 2011) (**Scheme 10**).

The transcarboxylation capacity of NHOs-CO<sup>2</sup> adducts has also been verified by 1-ethyl-3-methyl-imidazolium-2 methylenecarboxylate through realizing the C-C coupling of CO<sup>2</sup> and MeCN (**Scheme 10**) (Finger et al., 2016). In this transcarboxylation process, the basicity of NHOs should be strong enough to abstract proton from the CH acid.

As highly efficient carboxylating agents, the NHC-CO<sup>2</sup> and NHO-CO<sup>2</sup> complexes can be easily obtained by reacting NHCs or NHOs with atmospheric CO2. However, instead of serving as absorbent, NHCs and NHOs are usually used as catalysts to promote the conversion of pure CO<sup>2</sup> by forming transient NHC-CO<sup>2</sup> and NHO-CO<sup>2</sup> complexes (Kayaki et al., 2009; Zhou

et al., 2017). The reason is that NHCs and NHOs are sensitive to air and moisture thus they cannot be used as absorbents for CO<sup>2</sup> in air and industry exhaust. Nowadays, it is found that the imidazolium ionic liquids containing basic anion can absorb CO2, producing imidazolium carboxylates (Gurau et al., 2011; Wang and Wang, 2016). Considering the imidazolium ionic liquids are stable to air and moisture, it opens a new way for the utilization of NHC-CO<sup>2</sup> and NHO-CO<sup>2</sup> complexes in CCU strategy.

# IONIC LIQUIDS (ILS)

Ionic liquids (ILs) offer a new opportunity for developing novel CO<sup>2</sup> capture reagents (Huang and Rüther, 2009; Gurkan et al., 2010; Wang et al., 2011; Yang and He, 2014). Especially, the active site-containing ionic liquids can trap and activate CO<sup>2</sup> through chemical absorption. Besides, IL can also function as catalyst in CO<sup>2</sup> transformation (Lang et al., 2016; Zhang et al., 2017; Xia et al., 2018). Therefore, it is promising to combine the multiple roles of ILs in CCU strategy. Up to now, cyclocarbonates, oxazolidinones and quinazoline-2,4-(1H,3H)-diones have been synthesized using ILs as CO<sup>2</sup> absorbents and catalysts.

Wang group performs a series of investigation on ILsbased CO<sup>2</sup> capture and conversion. For example, they design bifunctionalized ionic liquids to capture and simultaneously fix CO<sup>2</sup> in the simulation of fuel gas to cyclic carbonates (**Scheme 11**) (Luo et al., 2016). The cation can capture CO<sup>2</sup> and the anion I <sup>−</sup> can activate the substrate to facilitate CO<sup>2</sup> insertion. In the presence of a small amount of water, the yield of product can be improved, making this reaction more applicable to industrial exhaust.

Later, the same group finds that the basicity of anion of ILs is very important for CO<sup>2</sup> capture and transformation. A hydroxyl functionalized aprotic ionic liquid shows high efficiency in synthesis of quinazoline-2,4(1H,3H) -diones from atmospheric CO2. The captured CO<sup>2</sup> instead of atmospheric CO<sup>2</sup> is also used and only 13% yield is obtained, being ascribed to the strong interaction between [Im]<sup>−</sup> and CO<sup>2</sup> (**Scheme 11**) (Shi et al., 2018). They also demonstrates the feasibility of using captured CO<sup>2</sup> as starting material in their another report, where the CO<sup>2</sup> captured by azole-type anion [DEIm]<sup>−</sup> renders a high

yield of alkylidene carbonates in the carboxylative cyclization of propargyl alcohol due to the weak interaction between CO<sup>2</sup> and the anion (**Scheme 11**) (Chen et al., 2016).

Liu group reports that azole-anion-based ILs with the [Bu4P]<sup>+</sup> cation can capture CO2. With appropriate substrates, the forming carbamate intermediates can be transformed into the αalkylidenecyclic carbonate, quinazoline-2,4(1H,3H)-diones and benzimidazolone without other catalysts (**Scheme 11**) (Zhao et al., 2016). Later, the same group reveals that a series of tetrabutylphosphonium ([Bu4P]+)-based ILs with multiple-site for CO<sup>2</sup> capture and activation in their anions can be used in CO<sup>2</sup> capture and conversion, wherein the IL [Bu4P]3[2,4- OPym-5-Ac] shows the optimal performance in preparation of α-alkylidene cyclic carbonates from propargylic alcohol substrate

(Wu et al., 2017). The resulting polycarbonates derived from CO<sup>2</sup> and the anion is proved to be the key intermediate in this reaction (**Scheme 11**).

The ILs [HDBU][MIm] and [HDBU][TFE] can capture CO<sup>2</sup> and also show catalytic activity to the reaction of CO<sup>2</sup> and propargylic amines for the synthesis of 2-oxazolidinones and the reaction of CO<sup>2</sup> with 2-aminobenzonitrile derivatives to synthesis quinazoline-2,4-(1H,3H)-diones. By enhancing the mass transfer with gas-liquid laminar flow continuous-flow microreactor, the simultaneous capture and fixation CO<sup>2</sup> to 2 oxazolidinones and quinazoline-2,4-(1H,3H)-diones is realized (**Scheme 11**) (Vishwakarma et al., 2017).

Due to the dual function as CO<sup>2</sup> absorbents and conversion catalysts, ionic liquids can realize the transformation of captured CO<sup>2</sup> with several kinds of substrates. Furthermore, the non-volatility characteristic of ionic liquids can facilitate product separation. Thus, the ionic liquids are considered as promising absorbents for the CCU process.

### FRUSTRATED LEWIS PAIRS (FLPS)

The CO<sup>2</sup> capture capacity of FLPs has been reported soon after the FLPs concept was in 2006 proposed (Welch et al., 2006; Momming et al., 2009; Travis et al., 2013; Weicker and Stephan, 2015; Wolff et al., 2016). As early as 2010, Stephan group revealed that 1:2 mixtures of PMes3/AlX<sup>3</sup> (X = Cl or Br) in bromobenzene can react with CO2, forming CO<sup>2</sup> adduct which can be converted to CH3OH with ammonia borane as reductant (**Scheme 12**) (Ménard and Stephan, 2010). In the same year, Piers group found that CO<sup>2</sup> captured by FLP consisting of 2,2,6,6-tetramethylpiperidine (TMP) and B(C6F5)<sup>3</sup> can be reduced to methane with triethylsilane (**Scheme 12**) (Berkefeld et al., 2010). Soon, it is found that the FLP composed by bis-borane 1,2-C6H4(BCl2)<sup>2</sup> and PtBu<sup>3</sup> can capture CO<sup>2</sup> and the forming capture product can be reduced to methanol by reductant such as amineborane Me2NHBH<sup>3</sup> or [C5H6Me4NH2][HB(C6F5)2(C7H11)] (**Scheme 12**) (Sgro et al., 2012). Wang group reports the FLP comprising of bis(2,4,6-tris(trifluoromethyl)phenyl)borane and a secondary amine (such as HNiPr<sup>2</sup> or HNEt2) readily reacts with CO<sup>2</sup> at 80◦C, affording carbamate boryl esters which can function as an intramolecular FLP to activate H2, affording ammonium borylformate salt and formamide adducts (**Scheme 12**) (Lu et al., 2013).

Recently, Yan group incorporates FLP acceptor and donor into the styrene-based monomers, respectively to prepare two diblock copolymers consisting of the complementary FLP blocks and common polystyrene block. These two diblock copolymers can bind CO2, forming nanoparticle. The nanoparticle is then used as CO<sup>2</sup> reservoir and catalyst to facilitate the formylation of amines with phenylsilane (**Scheme 12**) (Chen et al., 2018).

The CO<sup>2</sup> capture and H<sup>2</sup> activation capacity makes FLPs attractive for CCU strategy, especially for the hydrogenation of captured CO2. However, in the current study, the FLPs promoted CO<sup>2</sup> hydrogenation encounters difficulty in FLPs regeneration (Ashley et al., 2009). By now, the FLPs are merely used as CO<sup>2</sup> capture reagents and reductants are still needed. Thus, the hydrogen activation ability of FLPs hasn't

been utilized. Recently, the breakthrough is made by Jazzar and Bertrand group. By combining the copper catalyst and Lewis pair, hydrogenation of carbon dioxide into formate is realized (Romero et al., 2018). Latter, X. Hu and Y. Wu group reports the first catalytic hydrogenation process of CO<sup>2</sup> to formate using transition metal free catalyst (B(C6F5)3/M2CO3, M = Na, K, and Cs) (Zhao et al., 2019). These results open new vistas in the field of FLPs facilitated CO<sup>2</sup> capture and hydrogenation.

# CATALYST DESIGN FOR DIRECT CONVERSION OF DILUTED CO<sup>2</sup>

In addition to the CO<sup>2</sup> capture and transformation strategy, there are also examples that the CO<sup>2</sup> from waste streams can be directly converted by designing catalysts that tolerate to the contaminants in the waste streams such as exogenous water, nitrogen, SO2, amine etc. For example, Williams et al. reports the synthesis of poly(cyclohexylene carbonate) using the power station generated CO<sup>2</sup> facilitated by the homogeneous dinuclear Zn or Mg catalysts, which is stable in the presence of contaminants from gas streams (Chapman et al., 2015) (**Scheme 13**). D'Elia and Basset group develops the combination of early transition metal halides (Y, Sc, Zr) and TBAB to quantitatively convert CO<sup>2</sup> from diluted streams and produce cyclic organic carbonates (Barthel et al., 2016). The features of metal-organic frameworks (MOFs) to selectively capture and catalyze CO<sup>2</sup> conversion make them a new type of platform for diluted CO<sup>2</sup> transformation. Recently, Hong group design and synthesize an acid-base resistant Cu(II)-MOF which can convert CO<sup>2</sup> from simulated post-combustion flue gas into corresponding cyclic carbonates (Liang et al., 2017). In direct conversion of diluted CO2, the stability of the catalysts to the contaminants in the gas streams is crucial to the success of the process.

### CONCLUSION AND OUTLOOK

The past 10 years have witnessed great advances in CO<sup>2</sup> capture and in situ conversion. Different CO<sup>2</sup> absorbents including inorganic and organic bases, NHCs, and NHOs, ILs, FLPs and polymeric functional materials have been employed in this field. As an emerging field, much effort is still desired to explore the potential conversion of the captured CO2.

Considering amines and CO<sup>2</sup> can be involved in diverse reactions, it is hoped that the conversion of ammonia carbamate can be further extended to other valuable products besides isocyanates, carbamates and ureas.

On the other hand, although the hydrogenation of CO<sup>2</sup> captured by inorganic/organic bases and the combination of base and alcohol has been investigated, the hydrogenation of CO<sup>2</sup> captured by ILs and/or FLPs remains sporadic and underexplored. In light of the successful electrochemical reduction of IL-captured CO<sup>2</sup> and hydrogenation of CO<sup>2</sup> in ILs (Zhang et al., 2008, 2009; Wesselbaum et al., 2012; Scott et al., 2017), it is reasonable to conclude that the captured CO<sup>2</sup> by IL can be hydrogenated by designing appropriate catalyst. Besides, the breakthrough in FLP-mediated CO<sup>2</sup> hydrogenation opens new possibilities for FLP-based CO<sup>2</sup> capture and transformation (Romero et al., 2018; Zhao et al., 2019).

Another important transformation pathway for CO<sup>2</sup> adducts derived from reacting with amines, NHCs, NHOs and ILs is transcarboxylation reaction in the synthesis of cyclocarbonates, oxazolidinones and quinazoline-2,4-(1H,3H)-diones. It is foreseeable that the application of these transcarboxylating reagents will be investigated continuously with the emergence of new CO<sup>2</sup> conversion reactions.

Although a plethora of CO<sup>2</sup> absorbents has been developed and valuable products can be obtained from the resulting CO<sup>2</sup> adducts by different strategy, however, the cost and feasibility must be considered from the viewpoint of industrial application. With this in mind, amines and ionic liquids are considered as promising absorbents for the CCU strategy. For amine absorbents, their features of high absorbing capacity and comerical availability are attractive for industrial application. Besides, the CO<sup>2</sup> adducts derived from amines and CO<sup>2</sup> are non-volatile liquids or solids, which can be used as starting materials instead of the volatile amines and CO<sup>2</sup> gas to develop the gas free progress. For ionic liquid- based CO<sup>2</sup> absorbents, the acceptable cost, tunable structure as well as their high CO<sup>2</sup> capture capacity make them competent to the commercial CCU process. Moreover, the ionic liquids are nonvolatile and the corrosion of equipment can be avoided by subtly design the structure of ionic liquids, all of which can facilitate the industry operability.

In summary, the CO<sup>2</sup> capture and in situ catalytic transformation is still in its infancy. We hope this review can inspire the extensive research on CO<sup>2</sup> capture and in situ transformation, which will benefit for the design of efficient CO<sup>2</sup> capture and utilization system (CCU) and realize the valorization of diluted CO<sup>2</sup> in waste gas streams or directly from the atmosphere.

### AUTHOR CONTRIBUTIONS

All authors contributed for the writing of the manuscript. L-NH designed this proposal and determined the contents. H-RL wrote the Abstract, Introduction, Conclusion, and Outlook parts. H-CF wrote the Inorganic/organic bases and Ionic liquids parts. FY wrote N-heterocyclic carbenes and N-heterocyclic olefins and frustrated lewis pairs parts. H-RL and L-NH revised the manuscript.

### REFERENCES


### FUNDING

This research was funded by National Key Research and Development Program (2016YFA0602900), the National Natural Science Foundation of China (21672119).

### ACKNOWLEDGMENTS

We thank the Ministry of Science and Technology of China (National Key Research and Development Program) and the National Natural Science Foundation of China.

in 1,3-dialkylimidazolium acetate ionic liquids. Angew. Chem. Int. Ed. 50, 12024–12026. doi: 10.1002/anie.201105198


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Fu, You, Li and He. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Novel Photo-Driven Hydrogenation Reaction of an NAD+-Type Complex Toward Artificial Photosynthesis

Hideki Ohtsu\*, Tsubasa Saito and Kiyoshi Tsuge

*Graduate School of Science and Engineering, University of Toyama, Toyama, Japan*

The photocatalytic reduction of carbon dioxide (CO2) to value-added chemicals is an attractive strategy to utilize CO<sup>2</sup> as a feedstock for storing renewable energy, such as solar energy, in chemical bonds. Inspired by the biological function of the nicotinamide adenine dinucleotide redox couple (NAD+/NADH), we have been developing transition-metal complexes containing NAD+/NADH-functionalized ligands to create electro- and/or photochemically renewable hydride donors for the conversion of CO<sup>2</sup> into value-added chemicals. Our previous findings have provided insights for the development of photocatalytic organic hydride reduction reactions for CO2, however, further examples, as well as investigation, of these photo-driven NAD+/NADH-type hydrogenation and organic hydride transfer reactions are required not only to explore the mechanism in detail but also to develop a highly efficient catalyst for artificial photosynthesis. In this paper, we report the synthesis, characterization, and photo-induced NAD+/NADH conversion properties of a new ruthenium(II) complex, [Ru(bpy)2(Me-pn)](PF6)<sup>2</sup> (1), which contains a new NAD+-type ligand, Me-pn (2-methyl-6-(pyridin-2-yl)-1,5-naphthyridine). In addition, we have succeeded in the isolation of the corresponding two-electron reduced ruthenium(II) complex containing the NADH-type ligand Me-pnHH (2-methyl-6-(pyridin-2-yl)-1,4-dihydro-1,5-naphthyridine), i.e., [Ru(bpy)2(Me-pnHH)](PF6)<sup>2</sup> (1HH), by the photo-induced hydrogenation reaction of 1. Thus, in this study, a new photo-driven NAD+/NADH-type hydrogenation reaction for possible CO<sup>2</sup> reduction using the NAD+/NADH redox function has been constructed.

### Edited by:

*Hitoshi Ishida, Kitasato University, Japan*

### Reviewed by:

*M. Jasim Uddin, University of Texas Rio Grande Valley Edinburg, United States Koji Tanaka, Kyoto University, Japan*

> \*Correspondence: *Hideki Ohtsu ohtsu@sci.u-toyama.ac.jp*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *21 February 2019* Accepted: *02 August 2019* Published: *20 August 2019*

### Citation:

*Ohtsu H, Saito T and Tsuge K (2019) A Novel Photo-Driven Hydrogenation Reaction of an NAD*+*-Type Complex Toward Artificial Photosynthesis. Front. Chem. 7:580. doi: 10.3389/fchem.2019.00580* Keywords: CO2 reduction, photo-induced hydrogenation, NAD+, NADH, organic hydride, ruthenium complex

# INTRODUCTION

Currently, the design and development of novel visible-light photoredox catalysts for carbon dioxide (CO2) reduction are considered to be crucial challenges. In particular, it is vital not only to clarify how to employ solar energy, which is a sustainable clean energy source, but also to develop methods to convert CO2, which is a major greenhouse gas and is harmful for the environment, into chemical fuels and feedstocks in the viewpoint of the current global energy and environmental crisis (Fox and Dulay, 1993; Kisch, 2013). Many research efforts have so far been devoted to the search for new and efficient catalysts for CO<sup>2</sup> conversion into value-added chemicals such as carbon monoxide (CO), formic acid (HCO2H), and methanol (CH3OH) (Meyer, 1989; Alstrum-Acevedo et al., 2005; Wang et al., 2015).

**131**

However, there are several problems facing photochemical CO<sup>2</sup> conversion, particularly related to the side reactions and low selectivity toward specific reduction products, as well as poor energy efficiency (Leitner, 1995; Jones et al., 2014). To solve these problems, the use of transition-metal coordination compounds as photocatalysts for CO<sup>2</sup> reduction has drawn significant attention because some of these compounds exhibit significant photocatalytic activity for CO<sup>2</sup> reduction (Morris et al., 2009; Berardi et al., 2014), and they play an essential role in artificial photosynthesis(Fukuzumi et al., 2018), as well as natural photosynthesis (Silva et al., 2015).

In particular, ruthenium complexes are promising candidates for photocatalysts because most Ru complexes show excellent photophysical properties with relatively long excited-state lifetimes and visible light absorption bands originating from metal-to-ligand charge transfer (MLCT) (Medlycott and Hanan, 2005; Sun et al., 2015; Dongare et al., 2017). In the past several decades, a number of Ru complexes having photocatalytic ability toward CO<sup>2</sup> reduction have been reported (Jessop et al., 1995; Tanaka and Ooyama, 2002; Kuramochi et al., 2018). In this context, we have focused on Ru complexes having NAD+/NADH-functionalized ligands because the biological function of the NAD+/NADH redox couple is as a generator and reservoir of hydride ions (H−), which are equivalent to two electrons and one proton (Eisner and Kuthan, 1972; Walsh, 1980; Stout and Meyers, 1982; Gebicki et al., 2004; Bilan et al., 2015), and is of great interest for the development of photorenewable hydride reagents.

Previously, we have successfully demonstrated that a ruthenium(II) complex containing an NADH-type ligand, pbnHH [Ru(bpy)2(pbnHH)]2<sup>+</sup> (bpy = 2,2′ -bipyridine, pbnHH = 2-(pyridin-2-yl)-5,10-dihydrobenzo[b][1,5]naphthyridine), which is photochemically converted from the corresponding NAD+-type complex [Ru(bpy)2(pbn)]2<sup>+</sup> (pbn = 2-(pyridin-2-yl)benzo[b][1,5]naphthyridine) (Koizumi and Tanaka, 2005; Polyansky et al., 2007; Fukushima et al., 2009, 2010; Ohtsu and Tanaka, 2012a), can reduce CO<sup>2</sup> to HCO<sup>−</sup> 2 via organic hydride transfer involving C–H bond dissociation of the NADH-type ligand in [Ru(bpy)2(pbnHH)]2<sup>+</sup> driven by the association of the bases (Ohtsu and Tanaka, 2012b; Ohtsu et al., 2015). These findings stimulated us to explore a new type of NAD+/NADHfunctionalized complex, and the expansion of the scope of our previous work on NAD+/NADH model chemistry potentially opens new perspectives for the design of more efficient molecular photocatalysts for CO<sup>2</sup> reduction. As part of our ongoing research into Ru NAD+/NADH-functionalized complexes, we have synthesized and characterized the photo-induced NAD+/NADH-type hydrogenation reaction properties of a ruthenium(II) complex bearing a new NAD+-type ligand, Me-pn (Me-pn = 2-methyl-6-(pyridin-2-yl)-1,5-naphthyridine), i.e., [Ru(bpy)2(Me-pn)](PF6)<sup>2</sup> (**1**). In addition, we have also successfully isolated a ruthenium(II) complex having the corresponding NADH-type ligand, Me-pnHH (Me-pnHH = 2-methyl-6-(pyridin-2-yl)-1,4-dihydro-1,5-naphthyridine), i.e., [Ru(bpy)2(Me-pnHH)](PF6)<sup>2</sup> (**1HH**), by the photochemical reduction of **1** under light irradiation (λ > 420 nm) in the presence of a sacrificial reagent.

# MATERIALS AND METHODS

### Materials

All chemicals used for the synthesis of the ligands and complexes were commercial products of the highest available purity and were used without further purification unless otherwise indicated. Solvents were purified by standard methods before use (Armarego and Chai, 2009).

# Synthesis

All ligands and complexes used in this study were prepared according to the following procedures and the structures of the products were confirmed from analytical data (vide infra).

# 2-Methyl-6-(pyridin-2-yl)-1,5-naphthyridine (Me-pn)

This ligand was prepared in the same manner as that for the synthesis of 2-(pyridin-2-yl)-1,5-naphthyridine (Singh and Thummel, 2009) using 2-chloro-6-methyl-1,5-naphthyridine (Gellibert et al., 2004) instead of 2-chloro-1,5-naphthyridine. Anal. Calcd for C14H11N3: C, 76.00; H, 5.01; N, 18.99. Found: C, 75.84; H, 4.88; N, 18.81. <sup>1</sup>H NMR (300 MHz, CDCl3): δ 8.73 ∼ 8.79 (m, 2H), 8.60 (dt, J = 8.0, 1.1 Hz, 1H), 8.44 (d, J = 8.8 Hz, 1H), 8.35 (d, J = 8.6 Hz, 1H), 7.88 (td, J = 7.8, 2.0 Hz, 1H), 7.54 (d, J = 8.6 Hz, 1H), 7.38 (ddd, J = 7.4, 4.8, 1.3 Hz, 1H), 2.81 (s, 3H).

# [Ru(bpy)2(Me-pn)](PF6)<sup>2</sup> (1)

To a 2-methoxyethanol solution (20 mL) of [Ru(bpy)2Cl2] (363 mg, 0.75 mmol) was added AgPF<sup>6</sup> (381 mg, 1.5 mmol) in 2-methoxyethanol, and the mixture was stirred for 12 h at 70◦C. The resulting mixture was cooled to room temperature, and insoluble material was removed by filtration through celite. After the addition of the ligand Me-pn (754 mg, 0.75 mmol) to the filtrate, the reaction mixture was stirred for 24 h at 70◦C. The solution was concentrated to ca. 1 mL and poured into aqueous NH4PF<sup>6</sup> solution. The solid formed was collected and dried in vacuo. Recrystallization from acetonitrile/diethylether gave **1** as reddish purple crystals (389 mg, 56.1%). ESI-MS: m/z = 780 [M– PF6] <sup>+</sup>. Anal. Calcd for C34F12H29N7OP2Ru: C, 43.32; H, 3.10; N, 10.40. Found: C, 43.48; H, 3.25; N, 10.30. <sup>1</sup>H NMR(300 MHz, CD3CN): δ 8.73 (d, J = 9.1 Hz, 1H), 8.72 (d, J = 8.2 Hz, 1H), 8.51 ∼ 8.61 (m, 3H), 8.38 (d, J = 8.2 Hz, 1H), 8.30 (d, J = 8.1 Hz, 1H), 8.18 ∼ 8.32 (m, 5H), 7.89 ∼ 7.99 (m, 2H),7.67 (d, J = 5.6 Hz, 1H), 7.58 (d, J = 5.7 Hz, 1H), 7.38 ∼ 7.53 (m, 4H), 7.27 ∼ 7.38 (m, 3H), 7.11 (d, J = 9.1 Hz, 1H), 2.62 (s, 3H).

# [Ru(bpy)2(Me-pnHH)](PF6)<sup>2</sup> (1HH)

The NADH-type two-electron reduced complex was prepared photochemically: An anaerobic CH3CN solution (2.35 mL) of **1** (30.7 mg, 0.033 mmol) containing TEOA (0.208 g, 1.4 mmol) was irradiated with light through a longpass filter (HOYA W-Y455) and a super cold filter (ASAHI SPECTRA SC0751) using a 150 W Xenon lamp for 0.5 h. The resulting solution was added to aqueous NH4PF<sup>6</sup> solution, and the orange solid formed was collected by filtration and dried in vacuo. Yield: 24.2 mg (78.5%). ESI-MS: m/z = 782 [M–PF6] <sup>+</sup>. Anal. Calcd for C34F12H29N7P2Ru: C, 44.07; H, 3.15; N, 10.58. Found: C, 43.95; H, 3.21; N, 10.36. <sup>1</sup>H NMR(300 MHz, CD3CN): δ 8.52 (d, J = 8.3 Hz, 2H), 8.44 (t, J = 7.1 Hz, 2H), 8.18 ∼ 8.26 (m, 2H), 8.06 ∼ 8.15 (m, 3H), 7.89 ∼ 8.05 (m, 3H), 7.64 ∼ 7.72 (m, 2H), 7.52 (td, J = 6.4, 1.4 Hz, 1H), 7.42 ∼ 7.48 (m, 2H), 7.24 ∼ 7.37 (m, 3H), 7.16 (td, J = 6.6, 1.3 Hz, 1H), 7.10 (d, J = 8.7 Hz, 1H), 6.68 (br, s, 1H), 3.85 (br, m, 1H), 3.36 (ddd, J = 21.5, 3.9, 1.6, Hz, 1H), 2.35 (ddd, J = 21.5, 3.4, 1.8 Hz, 1H), 1.62 (br, m, 3H).

### Physical Measurements

All manipulations were carried out under an argon atmosphere using standard Schlenk techniques or in a glovebox. The absorption spectra were obtained using a Hewlett–Packard 8453 diode array spectrophotometer or a Shimadzu UV-3100PC scanning spectrophotometer. The emission spectra were recorded on a JASCO FP-8500 spectrofluorometer. The emission lifetimes were performed with a UNISOKU LSP-1000N-PX spectrometer. The electrospray (ESI)-mass spectrometry (MS) data were obtained with a Shimadzu LCMS-2020. Elemental analyses were performed using a Yanaco CHN Corder MT-5 (A Rabbit Science Japan Co., Ltd.). NMR measurements were measured with a JEOL JMN-ECX 300 (300 MHz) NMR spectrometer. Electron spin resonance (ESR) spectra were taken with a JEOL JES-X310 equipped with a variable temperature (VT) apparatus and were recorded under non-saturating microwave power conditions. The magnitude of the modulation was chosen to optimize the resolution and the signal-to-noise ratio of the observed spectra. The g values were calibrated with a Mn(II) marker used as a reference. Cyclic voltammetry measurements were performed on an ALS/Chi model 733D electrochemical analyzer in a deaerated solvent containing 0.1 M tetra-nbutylammonium hexafluorophosphate (TBAPF6) as a supporting electrolyte. A conventional three-electrode cell was used with a glassy-carbon working electrode and a platinum wire as the counter electrode. The glassy-carbon working electrode was routinely polished with a BAS polishing alumina suspension and rinsed with acetone before use. The reversibility of the electrochemical processes was evaluated by standard procedures, and all potentials were recorded against an Ag/AgCl reference electrode. All electrochemical measurements were carried out under an argon atmosphere.

### X-ray Crystal Structure Determination

The single crystal X-ray diffraction data of **1**•CH3CN•Et2O were collected on a Rigaku VariMax RAPID-DW/NAT with Mo-Kα radiation (λ = 0.71075 Å) at 173 K and processed using RapidAuto program (Rigaku). An empirical absorption correction resulted in acceptable transmission factors. The data were corrected for Lorentz and polarization factors. The structure was solved by direct methods using SHELXT-2018/2 (Sheldrick, 2015) and refined by full-matrix least-squares techniques on F <sup>2</sup> using SHELXL-2018/3 (Sheldrick, 2015). All non-hydrogen atoms were refined anisotropically and all hydrogen atoms were included in the refinement process as per the riding model. CCDC-1894384 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge via www.ccdc.cam.ac.uk/conts/retrieving.html (or from the Cambridge Crystallographic Data Center, 12, Union Road, Cambridge CB21EZ, UK; fax: (+44)1223-336-033; or deposit@ccdc.cam.ac.uk).

## RESULTS AND DISCUSSION

The ruthenium(II) complex, [Ru(bpy)2(Me-pn)](PF6)<sup>2</sup> (**1**), which bears our newly designed NAD+-type ligand, 2-methyl-6-(pyridin-2-yl)-1,5-naphthyridine (Me-pn), was successfully

complex [Ru(bpy)2(Me-pn)](PF6)2•CH3CN•Et2O (1•CH3CN•Et2O) (50% probability ellipsoids. orange: ruthenium, blue: nitrogen, gray: carbon). The counter anion (PF<sup>−</sup> 6 ), hydrogen atoms, and CH3CN and Et2O molecules are omitted for clarity.

prepared by mixing [Ru(bpy)2Cl2] and AgPF<sup>6</sup> in a 1:2 ratio in 2-methoxyethanol, followed by the addition of 1 equivalent of Me-pn at 343 K, and the composition of **1** was determined by elemental analysis, ESI-MS, and <sup>1</sup>H-NMR (see Material and Methods).

Well-shaped single crystals of **1** suitable for X-ray structure determination were obtained by recrystallization from an acetonitrile (CH3CN) solution using the slow vapor diffusion of diethylether (Et2O). This compound crystallizes in the monoclinic P21/n space group with four molecules in the unit cell. The structure also contains solvated CH3CN and Et2O for each complex. The molecular structure of **1** is shown in **Figure 1**. The ruthenium(II) ion in **1** has a hexacoordinate octahedral structure formed of four N atoms of the two bpy ligands and two N atoms of the Me-pn ligand. The coordination environment of **1** is almost the same as that of a previously reported ruthenium(II) NAD+-type complex [Ru(bpy)2(pbn)](PF6)<sup>2</sup> (Koizumi and Tanaka, 2005), and the bond lengths between the ruthenium(II) ion and the six N atoms of the ligands for **1** (2.046(4) to 2.119(4) Å) are not only the same as those of [Ru(bpy)2(pbn)](PF6)<sup>2</sup> (2.038(5) to 2.116(5) Å) (Koizumi and Tanaka, 2005) but also within the typical range of those of reported ruthenium(II) complexes having a similar coordination environment (Fukushima et al., 2009, 2010; Ohtsu and Tanaka, 2012a; Kobayashi et al., 2016). The ESI-MS results for **1** in CH3CN show a dominant signal at m/z = 780, as shown in **Figure S1**, and the observed mass value, as well as the isotopic pattern, agree well with those of the simulated pattern of [Ru(bpy)2(pbn)](PF6)]+. In addition, the <sup>1</sup>H-NMR spectrum of **1** in CD3CN (**Figure S2**) displays a spectral pattern reflecting the Cs symmetry of **1**, including 24 aromatic proton signals between 7.11 and 8.73 ppm and three proton singlet signal assigned to the methyl group at 2.62 ppm. These results are in complete agreement with the single-crystal X-ray diffraction analysis described above.

The absorption and emission spectra of **1** in CH3CN at 298 K are shown in **Figure 2**. Characteristic absorption bands centered around 440 and 480 nm can be assigned to MLCT transitions from the ruthenium(II) centers to the bpy ligand (Caspar and Meyer, 1983) and the Me-pn ligand, respectively. Complex **1** also exhibits a <sup>3</sup>MLCT emission band around 685 nm upon excitation at 480 nm, as shown in **Figure 2**. Additionally, the emission lifetime has been evaluated from the resulting single exponential emission decay (368 ns, see **Figure S3**), and this emission lifetime is larger than that of a previously reported ruthenium(II) NAD+ type complex [Ru(bpy)2(pbn)](PF6)<sup>2</sup> (140 ns) (Polyansky et al., 2007). Furthermore, the emission quantum yield (λex = 480 nm) was estimated at ø = 0.37% using a relative method (Eaton, 1988) with [Ru(bpy)3](PF6)<sup>2</sup> as a reference (0.095) (Suzuki et al., 2009; Ishida et al., 2010). This value is more than 500 times larger than that of [Ru(bpy)2(pbn)](PF6)<sup>2</sup> (0.00071%) (Ohtsu and Tanaka, 2012a) reported previously.

**Figure 3A** shows the cyclic voltammogram (CV) of **1** (1 mM) in dimethylformamide (DMF) at 298 K. The CV shows one reversible redox wave at E1/2(a) = 1.42 V (vs. Ag/AgCl) in the range above the rest potential (Erest = 0.00 V) and three reversible redox waves at E1/2(b) = −0.88 V, E1/2(c) = −1.31 V, and E1/2(d)

= −1.55 V in the range below Erest = 0.00 V. The E1/2(a) redox wave can be assigned to the Ru(II)/Ru(III) couple on the basis of the rest potential. To clarify the assignment of the other three redox waves, the CV of **1** upon the addition of an excess amount of a proton source such as acetic acid (AcOH) was measured in accordance with the previously reported procedure (Kobayashi et al., 2016). The CV changes observed upon the addition of 10 equivalents of AcOH into a solution of **1** in DMF are shown in **Figure 3B**. As shown, the reversibility of the E1/2(b) redox wave changes drastically to an irreversible cathodic peak at E ′ pc(b) = −0.92 V, and a new anodic peak at E ′ pa(b) = 0.49 V coupled with an irreversible E ′ pc(b) peak appears, whereas the other two E1/2(c) and E1/2(d) redox waves are barely changed from the redox waves at E ′ <sup>1</sup>/2(c) = −1.19 V and E ′ <sup>1</sup>/2(d) = −1.41 V, respectively, maintaining redox reversibility. The lack of reversibility of the E1/2(b) redox wave in the presence of the proton source is most likely caused by the protonation of the non-bonded N atom of the reduced Me-pn ligand; thus, the redox wave at E1/2(b) = −0.88 V can be assigned to the Mepn/Me-pn•− couple. The large separation between the cathodic and anodic peak potentials of the Me-pn localized redox reaction under protic conditions may result from significant structural changes, such as the hydrogenation of the NAD+-type Me-pn ligand to form the NADH-type Me-pnHH ligand through a two electron and two proton process. In the case of the other two redox waves (E1/2(c) = −1.31 V and E1/2(d) = −1.55 V), it is safe to conclude that the E1/2(c) and E1/2(d) redox waves

correspond to the redox couples of the two bpy ligands since almost no changes in the reversibility and the position of these redox waves are observed upon the addition of 10 equivalents of AcOH on the basis of CV measurements described above. In addition, the pK<sup>a</sup> of the Me-pn ligand in **1** has been determined by spectrophotometric acid–base titration from pH 0.29 to pH 4.65, as shown in **Figure S4**. The estimated pK<sup>a</sup> value of the protonated species [Ru(bpy)2(Me-pnH)]3<sup>+</sup> is 1.6, the value of which is almost the same as previously reported in the case of [Ru(bpy)2(pbn)](PF6)<sup>2</sup> (pK<sup>a</sup> = 1.7) (Polyansky et al., 2007).

To estimate the potential photochemical reactivity of **1**, the excited state reduction potential must be known. The excited state reduction potential of **1** (E ◦ ([Ru(bpy)2(Mepn)]2+<sup>∗</sup> /+)) is correlated with E1/2(Me-pn/Me-pn•−) and Eem(0–0) and is calculated to be 1.00 V using equation (1), where E1/2(Me-pn/Me-pn•−) = −0.88 V and Eem(0–0) = 1.88 eV, which was estimated from the <sup>3</sup>MLCT emission band (λex = 658 nm at 77 K in n-butyronitrile). The fact that the excited state reduction potential (E ◦ ([Ru(bpy)2(Me-pn)]2+<sup>∗</sup> /+) = 1.00 V) is located at a more positive potential than the oxidation potential of triethanolamine (TEOA, Eox = 0.80 V vs. SCE) (Kalyanasundaram, 1986) stimulated us to research the photochemical reactivity of **1**, such as the photo-driven NAD+/NADH-type hydrogenation reaction reported previously (Polyansky et al., 2007) in the presence of TEOA as a sacrificial reagent.

$$E^{\circ}(\text{[Ru(bpy)}\_{2}\text{(Me-pn)}]^{2+\text{\*}/+})$$

$$=E\_{1/2}(\text{Me-pn/Me-pn}^{\bullet-}) + E\_{\text{em}}(0-0)\tag{1}$$

In the absence of TEOA, complex **1** in CH3CN was quite stable even when exposed to visible light (150-W Xe lamp, λ > 420 nm) (see **Figure S5**). However, absorption spectral changes were observed upon the irradiation with visible light to a solution of **1** in the presence of TEOA (CH3CN:TEOA = 4:1 v/v) at 293 K as shown in **Figure 4**. This reaction consists of two distinct steps. First, the characteristic absorption band located at 480 nm arising from **1** (red line) rapidly changes, and a spectrum (purple line) exhibiting a new broad band around 1,000 nm with an accompanying isosbestic point at 448 nm from 0 to 20 s is obtained; in addition, the intensity of the spectrum

FIGURE 4 | Absorption spectral changes observed upon the irradiation with visible light (λ > 420 nm) to a solution (CH3CN:TEOA = 4:1 v/v) of 1 at 293 K from 0 to 20 s (A) and from 20 to 900 s (B), respectively. Inset: time course of the absorption changes at 420, 480, and 1,000 nm.

around 1,000 nm subsequently decreased. This is accompanied by an increase in the absorption band located at 420 nm (orange line) with an isosbestic point at 477 nm from 20 to 900 s. When triethylamine (TEA, Eox = 0.66 V vs. SCE) (Yanagida et al., 1997), which is also a sacrificial reagent, was used instead of TEOA, almost the same absorption spectral changes indicating a twostep process have been observed (**Figure S6**), and a dominant signal (m/z = 782), which agrees well with the simulated pattern corresponding to the NADH-type complex, [Ru(bpy)2(MepnHH)](PF6)]+, has been detected by ESI-MS measurements of the final reaction mixture, as shown in **Figure S7**. Furthermore, attempts to isolate the NADH-type complex [Ru(bpy)2(MepnHH)](PF6)<sup>2</sup> (**1HH**) by photochemical reduction using TEOA were successful based on the ESI-MS at m/z = 782 and the observed mass value, as well as the isotopic pattern, which agrees well with the simulated pattern of [Ru(bpy)2(MepnHH)](PF6)]+. In addition, 22 characteristic proton signals in the aromatic region, a proton N–H signal at 6.68 ppm, and two proton signals at 3.36 and 2.35 ppm, which can be assigned to the methylene protons of the 1,4-dihydropyridine frameworks in the Me-pnHH ligand (see Material and Methods and **Figure S8**), were observed. As reported for the previous ruthenium(II) NAD+-type complex [Ru(bpy)2(pbn)]2+, [Ru(bpy)2(pbn)]2<sup>+</sup> is efficiently reduced to the corresponding NADH-type complex ([Ru(bpy)2(pbnHH)]2+) by photochemical reduction in CH3CN in the presence of a sacrificial electron donor such as TEOA or TEA. The absorption spectral changes for the photo-driven NAD+/NADH-type hydrogenation reaction in [Ru(bpy)2(pbn)]2<sup>+</sup> system were observed in a one-step process (Ohtsu and Tanaka, 2012a) despite the use of the same experimental conditions as those of **1** (150-W Xe lamp(λ > 420 nm), CH3CN:TEOA = 4:1 v/v, 293 K). Such a difference in the reactivity may be ascribed to the steric effects of the methyl group of the Me-pn ligand in **1**, which result in weaker π-π interactions between two [Ru(bpy)2(Me-pnH• )]2<sup>+</sup> (**1** •−**H**) radical intermediate species compared to the case of the nonsubstituted pbn ligand in [Ru(bpy)2(pbn)]2+; this is based on a possible mechanism for the photo-driven NAD+/NADH-type hydrogenation reaction discussed later.

To gain further mechanistic insight into the photoinduced NAD+/NADH-type hydrogenation reaction of **1**, ESR measurements after photoirradiation for 20 and 900 s under the same conditions as those used to observe the absorption spectral changes shown in **Figure 4** were carried out (**Figure 5**). The ESR spectrum of **1** in CH3CN:TEOA = 4:1 v/v solution after photoirradiation for 20 s at 173 K displays a pseudo-isotropic signal at g = 2.000 with a line width of 1H = 19 G, as shown in **Figure 5A**. In contrast to the case of photoirradiation for 900 s, no ESR signal is observed (**Figure 5B**). The observed ESR signal in the case of photoirradiation for 20 s (**Figure 5A**) is similar to that of a one electron reduced [Ru(bpy)3] 2+ complex, namely [Ru(bpy)2(bpy•−)]<sup>+</sup> (g = 1.996 and 1H = 26 G at 173 K) (Motten et al., 1981), which can be assigned to a ligand-centered radical species [Ru(bpy)2(Me-pnH• )]2<sup>+</sup> (**1** •−**H**) based on a previous mechanistic study of the conversion of [Ru(bpy)2(pbn)]2<sup>+</sup> into [Ru(bpy)2(pbnHH)]2<sup>+</sup> (Polyansky et al., 2008). Complex **1HH**, which is the photoreduction product of **1** in the presence of a sacrificial reagent, is diamagnetic, which is consistent with the results of the ESR measurements in the case of photoirradiation for 900 s.

Based on the various experimental results described above and the previously reported mechanism for the photodriven NAD+/NADH-type hydrogenation reaction using [Ru(bpy)2(pbn)]2<sup>+</sup> (Polyansky et al., 2008), a photo-driven hydrogenation reaction pathway of **1** into **1HH** in the presence of a sacrificial reagent under photoirradiation can be proposed,

as shown in **Scheme 1**. First, the NAD+-type complex **1** is photochemically reduced to the Me-pn radical anion species [Ru(bpy)2(Me-pn•−)]<sup>+</sup> (**1** •−) by the reductive quenching of the photoexcited **1** <sup>∗</sup> by a sacrificial donor such as TEOA or TEA (step I). Secondly, the protonation of the non-coordinating N atom of the Me-pn•− in **1** •− produces [Ru(bpy)2(Me-pnH• )]2<sup>+</sup> (**1** •−**H**) radical species (step II). Then, each **1** •−**H** dimerizes via the π-π stacking of the neutral Me-pnH• moieties to afford {[Ru(bpy)2(Me-pnH• )]2+}<sup>2</sup> ((**1** •−**H)2**) (step III). Finally, intramolecular proton-coupled electron transfer from one Me-pnH• in (**1** •−**H)<sup>2</sup>** to another leads to disproportionation, thus generating the **1HH** NADH-type complex and regenerating the **1** NAD+-type complex. At present, we have yet to clarify the reason why there is a significant difference in the photo-driven NAD+/NADH-type hydrogenation reaction between **1** and the previously reported [Ru(bpy)2(pbn)]2<sup>+</sup> system. However, the steric hindrance of the methyl groups in Me-pn complex may be the reason why we can see the radical intermediate species such as **1** •−**H** in the ESR measurements more easily than for the previously reported [Ru(bpy)2(pbn)]2+.

### CONCLUSIONS

In conclusion, the present study has demonstrated the synthesis, characterization, and photoinduced NAD+/NADH conversion properties of a new ruthenium(II) complex, [Ru(bpy)2(Mepn)](PF6)<sup>2</sup> (**1**), as well as the isolated NADH-type complex [Ru(bpy)2(Me-pnHH)](PF6)<sup>2</sup> (**1HH**) under photochemical reduction conditions. The findings of the present study suggest a new type of photo-driven NAD+/NADH-type hydrogenation reaction that makes possible the development of photoinduced CO<sup>2</sup> reduction reactions utilizing the NAD+/NADH redox function. These results provide valuable information for the further development of functional transition-metal NAD+/NADH-type complexes and shed new light on the applications of the NADH-type complexes for future catalysts and energy storage materials in the area

### REFERENCES


of artificial photosynthesis. Further investigations concerning photochemical reduction of CO<sup>2</sup> and other substrates by using the NAD+/NADH-functionalized complexes are now in progress.

### DATA AVAILABILITY

The datasets generated for this study can be found in Cambridge Crystallographic Data Center, CCDC 1894384.

### AUTHOR CONTRIBUTIONS

HO directed the project, conceived and designed the experiments, supervised the progress of this work, and wrote the manuscript. TS carried out the syntheses of ligands and complexes and performed most of the experiments. KT contributed X-ray structural analyses. All authors listed have discussed the results, drawn conclusions, and given the approval to the final version of the manuscript.

### FUNDING

This work was supported by the Ministry of Education, Culture, Sports, Science, and Technology of Japan (MEXT) and the Japan Society for the Promotion of Science (JSPS) through Grantsin-Aid for Scientific Research C (JP17K05806, HO). HO also appreciates the financial support from Yashima Environment Technology Foundation and Yamaguchi Educational and Scholarship Foundation. TS gratefully acknowledges the financial support from The Japan Science Society by the Sasakawa Scientific Research Grant.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00580/full#supplementary-material


a novel series of potent and selective TGF-ß type I receptor inhibitors. J. Med. Chem. 47, 4494–4506. doi: 10.1021/jm0400247


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ohtsu, Saito and Tsuge. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# An Investigation of Electrocatalytic CO<sup>2</sup> Reduction Using a Manganese Tricarbonyl Biquinoline Complex

Meaghan McKinnon<sup>1</sup> , Veronika Belkina<sup>1</sup> , Ken T. Ngo1†, Mehmed Z. Ertem<sup>2</sup> \*, David C. Grills <sup>2</sup> \* and Jonathan Rochford<sup>1</sup> \*

<sup>1</sup> Department of Chemistry, University of Massachusetts Boston, Boston, MA, United States, <sup>2</sup> Chemistry Division, Brookhaven National Laboratory, Upton, NY, United States

The subject of this study [fac-Mn(bqn)(CO)3(CH3CN)]<sup>+</sup> (bqn = 2,2′ -biquinoline), is of particular interest because the bqn ligand exhibits both steric and electronic influence over the fundamental redox properties of the complex and, consequently, its related catalytic properties with respect to the activation of CO2. While not a particularly efficient catalyst for CO<sup>2</sup> to CO conversion, in-situ generation and activity measurements of the [fac-Mn(bqn)(CO)3] <sup>−</sup> active catalyst allows for a better understanding of ligand design at the Mn center. By making direct comparisons to the related 2,2′ -bipyridyl (bpy), 1,10-phenanthroline (phen), and 2,9-dimethyl-1,10-phenanthroline (dmphen) ligands via a combination of voltammetry, infrared spectroelectrochemistry, controlled potential electrolysis and computational analysis, the role of steric vs. electronic influences on the nucleophilicity of Mn-based CO<sup>2</sup> reduction electrocatalysts is discussed.

Keywords: carbon dioxide reduction, electrocatalysis, manganese, carbon monoxide, hydrogen evolution, computational modeling

### INTRODUCTION

The catalytic reduction of CO<sup>2</sup> into useful C-1 chemical feedstocks offers one potential strategy to develop a carbon-neutral alternative to our current dependence on fossil fuels. A major challenge for the catalysis community is to develop molecular catalysts capable of coupling electrochemical reduction with protonation of the CO<sup>2</sup> substrate toward useful C-1 products. One such example is the proton-coupled two-electron reduction of CO<sup>2</sup> to CO, which comes at a cost of just −0.52 V vs. SHE in water at pH 7 (Equation 1) (Arakawa et al., 2001), while in dry acetonitrile at pH 0 (Equation 2), the standard electrode potential for this reaction has been estimated, using two different thermodynamic cycles, as −0.13 V vs. Fc+/<sup>0</sup> (Matsubara et al., 2015) or −0.12 V vs. Fc+/<sup>0</sup> (Pegis et al., 2015).

$$\text{CO}\_{2(g)} + 2\text{H}^+\text{(aq)} + 2\text{e}^- \leftrightharpoons \text{CO}\_{(g)} + \text{H}\_2\text{O}\_{(l)}\text{ }E^0 = -0.52\text{ V vs. SHE (pH.7)}\tag{1}$$

$$\text{CO}\_{2(g)} + 2\text{H}^+\text{(CH3CN)} + 2\text{e}^- \leftrightharpoons \text{CO}\_{(g)} + \text{H}\_2\text{O}\_{\text{(CH3CN)}}\text{ }E^0 = -0.13\text{ V vs. }\text{Fe}^{+/0}\text{ (pH 0)}\tag{2}$$

This seemingly simple reaction is limited by slow kinetic parameters, however, which demands an appropriate catalyst to access a kinetically efficient pathway for CO<sup>2</sup> utilization – albeit at the cost of an electrochemical overpotential (η) (Appel and Helm, 2014). Of the homogeneous transition metal-based molecular catalysts used in this field (Francke et al., 2018; Sinopoli et al., 2018; Stanbury et al., 2018), manganese(I) polypyridyl tricarbonyl catalysts of the type [fac-Mn<sup>I</sup> (N∧N)(CO)3X]<sup>n</sup> (where N∧N = polypyridyl ligand, X = Br<sup>−</sup> (n = 0) or CH3CN (n = +1)), and their analogs, have been of keen interest due to their high selectivity, low

### Edited by:

Federico Cesano, University of Turin, Italy

### Reviewed by:

Julio Lloret-Fillol, Institut Català d'Investigació Química, Spain Charles Machan, University of Virginia, United States

### \*Correspondence:

Mehmed Z. Ertem mzertem@bnl.gov David C. Grills dcgrills@bnl.gov Jonathan Rochford jonathan.rochford@umb.edu

### †Present address:

Ken T. Ngo, National Renewable Energy Laboratory (DOE), Golden, CO, United States

### Specialty section:

This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry

Received: 01 March 2019 Accepted: 02 September 2019 Published: 24 September 2019

### Citation:

McKinnon M, Belkina V, Ngo KT, Ertem MZ, Grills DC and Rochford J (2019) An Investigation of Electrocatalytic CO2 Reduction Using a Manganese Tricarbonyl Biquinoline Complex. Front. Chem. 7:628. doi: 10.3389/fchem.2019.00628 cost, and low overpotential (Grills et al., 2018). Recently we have reported on tuning both the inner coordination sphere (McKinnon et al., 2019) and second coordination sphere (Ngo et al., 2017) of the ligand to optimize catalytic efficiency and selectivity for CO formation, notably against the competitive two-electron two-proton coupled redox transformations of CO<sup>2</sup> to HCO2H and H<sup>+</sup> to H2. More subtle modification on the periphery of the bpy ligand is also known to strongly influence the reduction potentials observed for these catalysts. This is evident, for example, with the inductive electron donating influence of the dtbpy (dtbpy = 4,4′ <sup>t</sup>Bu2-bpy) ligand, which shifts the reduction potential 0.11 V more negative relative to the simple bpy analog (Smieja et al., 2013). Similarly, while maintaining the same inner coordination sphere of [fac-Mn(N∧N)(CO)3L]<sup>n</sup> catalysts but introducing redox non-innocence on the ligand backbone, a recent report on [fac-MnBr(phen-dione)(CO)3] (phen-dione = 1,10-phenanthroline-5,6-dione) demonstrated vastly different electrochemistry, maintaining a high selectivity for CO evolution, compared to the analogous bpy and phen (phen = 1,10-phenanthroline) complexes, notwithstanding significant ligand-based redox activity at the dione functional group to generate a bis-carboxylate phenanthroline intermediate species (Stanbury et al., 2017).

In the current study we have focused on a rather simple modification of the polypyridyl ligand, but one which allows us to directly investigate both steric and electronic influences in the [fac-Mn(bqn)(CO)3(CH3CN)]+**([4-CH3CN]**+**)** pre-catalyst, where bqn = 2,2′ -biquinoline (**Figure 1**). While fundamental in approach, a simple systematic extension of the π-conjugated system of the polypyridyl ligand has not yet been reported for Mn(I) CO<sup>2</sup> reduction electrocatalysts. Four pre-catalysts of the general structure [fac-Mn(N∧N)(CO)3(CH3CN)]<sup>+</sup> are here investigated using the benchmark 2,2′ -bipyridyl (bpy) ligand **([1-CH3CN]**+**)** alongside the 1,10-phenanthroline (phen) **([2-CH3CN]**+**)**, 2,9-dimethyl-1,10-phenanthroline (dmphen) **([3-CH3CN]**+**)**, and aforementioned bqn ligands. Although the phen ligand does not have as extensive a π-conjugation as the bqn ligand, it serves a critical role in this study by bridging the gap between the sterically related dmphen and bqn ligands. Indeed, a direct comparison of the sterically and electronically expanded bqn ligand to the sterically related dmphen ligand probes the question of whether steric or electronic effects dominate in determining the redox, and ultimately catalytic, properties of this class of CO<sup>2</sup> reduction electrocatalyst. As demonstrated below using a combination of IR-SEC and computational studies, the combined electronic and steric influences of the π-extended bqn ligand hinder formation of a **[4–4]<sup>0</sup>** dimer, facilitating a concerted twoelectron ECE mechanism for the generation of **[4]**−. However, this comes at the cost of a change in product selectivity for the **[4-CH3CN]**<sup>+</sup> pre-catalyst to favor H<sup>2</sup> evolution in the presence of excess TFE. Selective CO<sup>2</sup> activation is observed for all catalysts here studied upon in-situ generation of a three-electron reduced [fac-Mn(−1)(N∧N•−)(CO)3] 2− active catalyst.

# RESULTS AND DISCUSSION

## Synthesis and Structural Characterization

Initial metathesis of [MnBr(CO)5] with silver triflate was completed as a first step to produce the [Mn(CO)5(OTf)] intermediate (Scheiring et al., 2000), allowing for the subsequent straightforward isolation of pure [fac-Mn<sup>I</sup> (N∧N)(CO)3(OTf)] products. Microwave reflux of [Mn(CO)5(OTf)] with one equivalent of the appropriate ligand in tetrahydrofuran afforded the pale yellow [fac-Mn(bpy)(CO)3(OTf)] (**1-OTf**), [fac-Mn(phen)(CO)3(OTf)] (**2-OTf**), and [fac-Mn(dmphen)(CO)3(OTf)] (**3-OTf**) products, and the red-orange [fac-Mn(bqn)(CO)3(OTf)] (**4-OTf**) solid, in quantitative yield following precipitation in excess diethyl ether. Each product was satisfactorily characterized by <sup>1</sup>H NMR and FTIR spectroscopies as well as elemental analysis. The facial (fac) arrangement of the Mn(CO)<sup>3</sup> core structure in each complex was confirmed by FTIR spectroscopy, where characteristic ν(CO) vibrational stretching modes for each of the solvated **[1-CH3CN]**+, **[2-CH3CN]**+, **[3-CH3CN]**<sup>+</sup> and **[4-CH3CN]**<sup>+</sup> complexes in neat acetonitrile provides a very useful comparison of structural and electronic properties of all four complexes using a basic knowledge of point-group symmetry and Mn(dπ)-CO(π ∗ ) back-bonding, respectively (**Table 2**). The bpy and phen complexes, **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**+, each exhibit pseudo-C3v point group symmetry with identical FTIR spectra composed of a sharp, symmetric ν(CO) stretching mode at 2,050 cm−<sup>1</sup> and a second broad, lower frequency asymmetric stretching mode at 1,958 cm−<sup>1</sup> . Consistent with voltammetry and computational analysis presented below, the identical ν(CO) stretching modes illustrates the negligible electronic influence of the π-extended phen ligand in comparison to the bpy system. This observation is consistent with related reports of the analogous bromide complexes (Kurtz et al., 2015; Stanbury et al., 2017; Tignor et al., 2018). In contrast, the dmphen complex **[3-CH3CN]**<sup>+</sup> exhibits a descent in symmetry to pseudo-C<sup>s</sup> , evident in a breaking of degeneracy for its lower frequency asymmetric ν(CO) stretching modes, which occur at 1,959 and 1,944(sh) cm−<sup>1</sup> . This suggests a sterically induced distortion of the fac-Mn(CO)<sup>3</sup> core, likely by the 2,10-dimethyl substituents of the dmphen ligand. Furthermore, the higher frequency symmetric ν(CO) stretching mode of **[3-CH3CN]**<sup>+</sup> exhibits a 4 cm−<sup>1</sup> shift to lower frequency at 2,046 cm−<sup>1</sup> , consistent with an inductive electron-donating influence of the two methyl substituents at dmphen, likely due to increased Mn(dπ)→CO(π ∗ ) back-bonding. Interestingly, the bqn complex, **[4-CH3CN]**<sup>+</sup> exhibits an FTIR spectral profile with ν(CO) = 2,047 and 1,959 cm−<sup>1</sup> , indicative again of pseudo-C3v symmetry, similar to **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**+. However, with the high frequency symmetric ν(CO) stretching mode occurring at 2,047 cm−<sup>1</sup> , this suggests that the inductive electron-donating influence of the bqn ligand is similar to that of the dmphen ligand in **[3-CH3CN]**+. Experimental FTIR spectral profiles are consistent with calculated spectra, as illustrated in **Figure 2**.

# Cyclic Voltammetry Under Non-catalytic Conditions

Electrochemical characterization of each complex was carried out under inert conditions (1 atmosphere of argon), in the absence of auxiliary Brønsted acid, prior to screening for catalytic activity. Each complex displays an irreversible Mn(II/I) oxidation event within 20 mV of each other in the range of +1.02 to +1.04 V vs. the ferricenium/ferrocence (Fc+/<sup>0</sup> ) pseudoreference (**Figure 3**, **Table 1**). This suggests that there is little

difference in electron density between each Mn-center upon varying the polypyridyl ligand across the four complexes. The very subtle differences observed in the v(CO) stretching modes by FTIR spectroscopy are unlikely to be reproduced by cyclic voltammetry. Furthermore, voltammetry data is more complex due to the in-situ generation of the Mn(II) oxidation state. We will focus on the reduction properties of these complexes forthwith due to their greater relevance toward our subsequent catalytic studies. The electrochemical properties of **[1-CH3CN]**<sup>+</sup> have been recently reported (Grills et al., 2018; McKinnon et al., 2019). Three sequential one-electron cathodic peaks are observed at Epc = −1.48 V, −1.83 V and −2.94 V with only the third reduction exhibiting quasi-reversible behavior (E1/<sup>2</sup> = −2.90 V, 1E<sup>p</sup> = 81 mV at υ = 0.1 V s−<sup>1</sup> ) vs. Fc+/<sup>0</sup> . The phen complex, **[2-CH3CN]**<sup>+</sup> exhibits a first one-electron irreversible reduction at Epc = −1.48 V vs. Fc+/<sup>0</sup> , identical to that observed for TABLE 1 | Redox potentials recorded by cyclic voltammetry for [1-CH3CN]+, [2-CH3CN]+, [3-CH3CN]+, and [4-CH3CN]<sup>+</sup> reported vs. the ferricenium/ferrocene (Fc+/<sup>0</sup> ) pseudo reference.


<sup>a</sup>Epa, irreversible one-electron oxidation.

<sup>b</sup>Epa, Mn<sup>0</sup> -Mn<sup>0</sup> dimer oxidation.

<sup>c</sup>Epc, irreversible one-electron reduction.

<sup>d</sup>E1/2, quasi-reversible one-electron reduction.

<sup>e</sup>E1/2, reversible concerted two-electron reduction.

Conditions: 1 mM sample concentration; 0.1 M [Bu4N][PF6] acetonitrile supporting electrolyte; 3 mm diameter glassy carbon working electrode; Pt wire counter electrode; Ag/AgPF<sup>6</sup> acetonitrile non-aqueous reference electrode; υ = 0.1 V s−<sup>1</sup> .

**[1-CH3CN]**+, further supporting the electronic similarity of bpy and phen at least in the [fac-Mn(N∧N)(CO)3L]<sup>n</sup> class of complexes (Tignor et al., 2018). This first reduction has been previously established (Grills et al., 2018) as ligand-based according to an electrochemical-chemical (EC) reaction scheme whereby, upon one-electron reduction of the bpy π <sup>∗</sup> orbital, rapid CH3CN dissociation occurs with a concurrent shift in radical character from the ligand to form the neutral five-coordinate 17-valence electron intermediate, [fac-Mn<sup>0</sup> (bpy)(CO)3] (**[1]<sup>0</sup>** ). This metastable Mn(0) complex, in the absence of any steric hindrance (Sampson et al., 2014), rapidly forms the Mn<sup>0</sup> -Mn<sup>0</sup> bound 18-valence electron [fac-Mn<sup>0</sup> (bpy)(CO)3]<sup>2</sup> dimer, **[1–1]<sup>0</sup>** . While unequivocal evidence of [fac-Mn<sup>0</sup> (N∧N)(CO)3] dimer formation is presented below via infrared spectroelectrochemical (IR-SEC) studies, its oxidation is often evident in the reverse anodic scan in cyclic voltammetry, observed here at −0.61 V and −0.58 V for **[1–1]<sup>0</sup>** and **[2–2]<sup>0</sup>** , respectively. Thus, the second reduction event for [fac-Mn(N∧N)(CO)3L]<sup>n</sup> complexes is often attributed to an irreversible two-electron EC event whereby the dimer is reductively cleaved to generate two equivalents of the two-electron reduced [fac-Mn<sup>0</sup> (N∧N•−)(CO)3] <sup>−</sup> anion, now established as the principal active catalyst for CO<sup>2</sup> activation by this class of complex via the most common reduction-first pathway (Riplinger et al., 2014). Reduction of the phen based dimer, **[2–2]<sup>0</sup>** is observed at Epc = −1.86 V, suggesting that **[2]**<sup>−</sup> is more nucleophilic than the benchmark bpy analog, **[1]**<sup>−</sup> generated at the slightly more positive potential of Epc = −1.83 V. Similar to its bpy analog, the native **[2-CH3CN]**<sup>+</sup> precursor exhibits a quasi-reversible third reduction event at Epc = −2.72 V (E1/<sup>2</sup> = −2.66 V, 1E<sup>p</sup> = 114 mV at υ = 0.1 V s−<sup>1</sup> ). This third quasi-reversible reduction has rarely been discussed in the literature but computational analysis here suggests that it is a predominantly ligand based reduction giving rise to a 19-electron [fac-Mn<sup>0</sup> (N∧N2−)(CO)3] <sup>2</sup><sup>−</sup> dianion. The related dmphen complex, **[3-CH3CN]**<sup>+</sup> exhibits unexceptionally similar voltammetry in comparison to both its bpy and phen analogs. Its first reduction again involves a ligand based EC pathway leading to formation of the **[3–3]<sup>0</sup>** dimer. This EC reaction is shifted negatively by 10 mV to Epc = −1.49 V, relative to formation of both **[1–1]<sup>0</sup>** and **[2–2]<sup>0</sup>** , consistent with the inductive donating character of the 2,9-dimethyl substituents observed by FTIR studies. Although there is weak evidence of dimer oxidation in the full scan voltammogram presented in **Figure 3**, reversing the scan after just one-electron reduction clearly demonstrates a significant dimer oxidation peak at Epa = −0.65 V (**Figure S5**). Reductive cleavage of the **[3–3]<sup>0</sup>** dimer to form the two-electron reduced five-coordinate [fac-Mn<sup>0</sup> (dmbpy•−)(CO)3] <sup>−</sup> anion (**[3]**−), however, occurs at a more positive potential of −1.78 V. This observation suggests that the steric influence of dmphen slightly hinders electronic coupling between the Mn(0) and dmphen•− radical centers in **[3]**−, rendering it slightly less nucleophilic as a result. The third reduction wave for **[3]**−/2<sup>−</sup> is observed at −2.86 V and, in contrast to its bpy and phen analogs, is completely irreversible.

Noticeably in **Figure 3**, the bqn complex, **[4-CH3CN]**<sup>+</sup> exhibits a single, reversible, concerted two-electron reduction event at Epc = −1.30 V (E1/<sup>2</sup> = −1.28 V, 1E<sup>p</sup> = 53 mV at υ = 0.1 V s−<sup>1</sup> ) followed by a single additional quasi-reversible oneelectron reduction event at Epc = −2.51 V (E1/<sup>2</sup> = −2.48 V, 1E<sup>p</sup> = 71 mV at υ = 0.1 V s−<sup>1</sup> ) vs. Fc+/<sup>0</sup> . Equally important here is the lack of any evidence for dimer oxidation in the reverse anodic scan of **[4-CH3CN]**+. This electrochemical behavior, at least in its peak profile, is uncannily similar to previously reported bulky 6,6′ -substituted bpy ligands investigated at the identical [fac-Mn<sup>0</sup> (N∧N)(CO)3(CH3CN)]<sup>+</sup> center (Sampson et al., 2014; Ngo et al., 2017). Mn<sup>0</sup> -Mn<sup>0</sup> formation is known to favor a staggered structure of both bpy ligands in **[1–1]<sup>0</sup>** (Machan et al., 2014), and thus far only the inclusion of steric bulk orthogonal to the plane of the polypyridyl ligand has been demonstrated to prevent dimer formation (Sampson et al., 2014; Ngo et al., 2017). Thus, it is appropriate here to question whether the redox behavior exhibited by **[4-CH3CN]**<sup>+</sup> is in fact due to an electronic influence of this relatively electron deficient ligand, or rather, alternatively is the result of a steric influence of the bqn ligand. The bqn ligand is, after all, π-extended from the same 6,6′ -positions as previously reported 6,6′ -orthoganol-sterically bulky bpy stystems, albeit here non-orthogonal to, but in the same plane as, the core bpy structure. This question is addressed explicitly via computational analysis presented below.

### Infrared Spectroelectrochemistry (IR-SEC)

To gain structural insight into electrochemical activation of these pre-catalysts, IR-SEC was carried out in 0.1 M [Bu4N][PF6] acetonitrile supporting electrolyte under 1 atmosphere of argon (**Table 2**). IR-SEC is a powerful investigative tool that takes advantage of the structural specificity of IR spectroscopy for transition metal carbonyl complexes and the ability to prepare in-situ catalytic intermediates simply by gradually stepping the potential of the working electrode (Kaim and Fiedler, 2009; Machan et al., 2014). At open-circuit potential, i.e., resting potential of the native pre-catalyst, **[1-CH3CN]**<sup>+</sup> and **[2- CH3CN]**<sup>+</sup> both display identical ν(CO) stretching modes as presented earlier (**Figure 2**), confirming negligible influence of the 0.1 M [Bu4N][PF6] electrolyte on their FTIR spectra. When the potential was biased beyond the first-reduction at −1.60 V

### TABLE 2 | <sup>ν</sup>(CO) infrared stretches for [1-CH3CN]+, [2-CH3CN]+, [3-CH3CN]+, and [4-CH3CN]<sup>+</sup> precatalysts, and their one-electron reduced, dimeric, and two-electron reduced derivatives, obtained by solution phase FTIR and IR-SEC spectroscopy.


<sup>a</sup>Recorded in neat CH3CN.

<sup>b</sup>Recorded in 0.1 M [Bu4N][PF6] acetonitrile electrolyte.

vs. Fc+/<sup>0</sup> for both **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**+, a progression was observed where the native ν(CO) stretching modes diminished while four unique but related ν(CO) stretching modes grew in concurrently at 1,857, 1,879, 1,933, and 1,976 cm−<sup>1</sup> from **[1-CH3CN]**<sup>+</sup> (**Figure S10**) and at 1,857, 1,880, 1,935, and 1,977 cm−<sup>1</sup> from **[2-CH3CN]**<sup>+</sup> (**Figure S11**). These data are consistent with an earlier report of the **[1–1]<sup>0</sup>** dimer (Hartl et al., 1995), and the peak profile associated with related Mn<sup>0</sup> - Mn<sup>0</sup> dimers previously observed via IR-SEC (Grills et al., 2018), and they are also consistent with our computational frequency analysis. A similar assignment is made for **[2–2]<sup>0</sup>** which exhibits almost identical ν(CO) stretching frequencies (vide supra). Upon biasing the working electrode potential further negative (−2.00 V vs. Fc+/<sup>0</sup> ) beyond the second Mn<sup>0</sup> -Mn<sup>0</sup> /Mn<sup>−</sup> based reduction for both **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**+, loss of the four dimer stretches was observed (**Figures S10, S11**) with concurrent growth of two new ν(CO) stretching modes at lower wavenumber, corresponding to the five-coordinate two-electron reduced [fac-Mn<sup>0</sup> (N∧N•−)(CO)3] <sup>−</sup> active catalysts, **[1]**<sup>−</sup> [ν(CO) = 1,811 and 1,911 cm−<sup>1</sup> ] and **[2]**<sup>−</sup> [ν(CO) = 1,813 and 1,923 cm−<sup>1</sup> ]. Consistent with its comparable voltammetry behavior, the dmphen based complex, **[3-CH3CN]**<sup>+</sup> exhibits a similar IR-SEC transition to the two-electron reduced active catalyst, **[3]**<sup>−</sup> [ν(CO) = 1,799 and 1,897 cm−<sup>1</sup> ] via the **[3–3]<sup>0</sup>** dimer intermediate [ν(CO) = 1,849(sh), 1,860, 1,923, and 1,968 cm−<sup>1</sup> ] as illustrated in **Figure 4**. Although clean transformation of **[3- CH3CN]**<sup>+</sup> to **[3]**<sup>−</sup> was observed upon direct electrolysis at −2.00 V, it is worth noting that selective electrolysis to the intermediate **[3–3]<sup>0</sup>** dimer species did exhibit evidence of minor decomposition to an unidentified side-product [ν(CO) = 1,919 and 2,026 cm−<sup>1</sup> ] with concurrent growth of a broad weak infrared absorption consistent with free CO (**Figure S13**). This is consistent with an earlier report of poor stability of the simpler phen-based dimer, **[2–2]<sup>0</sup>** (Stanbury et al., 2017).

As a concerted two-electron reduction event occurs for **[4- CH3CN]**+, the potential was set at −1.65 V vs. Fc+/<sup>0</sup> for

FIGURE 4 | IR-SEC spectra recorded on [3-CH3CN]<sup>+</sup> at the resting potential (blue), upon one-electron reduction (green) and upon two-electron reduction (red).

quantitative in-situ formation of the two-electron reduced product. Indeed, no evidence for an intermediate one-electron reduced monomer or the Mn<sup>0</sup> -Mn<sup>0</sup> dimer was observed. Instead, consistent with the observed voltammetry (vide supra) only a transition to two lower-frequency stretching modes at ν(CO) = 1,828 and 1,925 cm−<sup>1</sup> was observed, which are attributed to the two-electron reduced [fac-Mn<sup>0</sup> (bqn•−)(CO)3] − species, **[4]**<sup>−</sup> (**Figure 5**).

# Voltammetry Under 1 Atm CO<sup>2</sup> in the Presence of 0.3 % H2O

Prior studies of Mn(I) polypyridyl based electrocatalysts for CO<sup>2</sup> reduction have only reported catalytic current upon addition of excess weak Brønsted acid with, for example, 2.71 M (5%) addition of H2O (Bourrez et al., 2011). In contrast to their Re counterparts, which have been established for some time to promote CO<sup>2</sup> reduction in the absence of a proton donor, Mn(I) polypyridyl based electrocatalysts typically require the presence of a proton donor to promote intermediate metallocarboxylic acid formation due to the poorer nucleophilicity of Mn vs. Re catalysts (Riplinger and Carter, 2015). Cyclic voltammograms for pre-catalysts **[1-CH3CN]**+, **[2-CH3CN]**+, **[3-CH3CN]**<sup>+</sup> and **[4-CH3CN]**<sup>+</sup> recorded under 1 atm CO<sup>2</sup> at 0.1 V s−<sup>1</sup> in 0.1 M [Bu4N][PF6] acetonitrile electrolyte exhibit clear evidence of catalytic behavior for each complex (**Figure 6**). Although excess Brønsted acid was not added in these experiments, the reagent grade acetonitrile that was used still contains 0.17 M (0.3 %) residual H2O, which was recently demonstrated to be sufficient to facilitate proton-coupled CO<sup>2</sup> reduction to form CO (McKinnon et al., 2019). This observation is logical due to the relatively larger overpotential applied, especially as the active catalyst under these experimental conditions is predicted by computation (vide infra) to be the more nucleophilic three-electron reduced dianion.

Without knowledge of the in-situ pH of the electrolyte solution, the half-wave potential of the catalytic wave (Ecat/2) can be used to make a relative comparison of overpotentials under these experimental conditions. As expected, π-extension with the bqn ligand in pre-catalyst **[4-CH3CN]**<sup>+</sup> gives rise to a significant positive shift of catalytic current with Ecat/<sup>2</sup> = −2.42 V, whereas the bpy, phen and dmphen pre-catalysts exhibit Ecat/<sup>2</sup> = −2.85, −2.64, and −2.74 V, respectively. One difficulty encountered under these experimental conditions was in attempting to establish steady-state catalytic conditions with respect to the rate-limiting consumption of CO<sup>2</sup> within the electrochemical double-layer. As discussed later, this became less problematic once excess Brønsted acid was added. As such, TOFmax could not be satisfactorily determined under these experimental conditions. Thus, the TOF value is reported for each catalyst in **Table 3** using Equation 3 and the icat/i<sup>p</sup> ratio determined at a scan rate of 0.1 V s−<sup>1</sup> , to aid their side-byside comparison,

$$\text{TOF} = 0.1992 \left( \frac{F \upsilon}{RT} \right) \left( \frac{n\_p^3}{n\_{cat}^2} \right) \left( \frac{i\_{cat}}{i\_p} \right)^2 \tag{3}$$

where F is the Faraday constant (96,485 s A mol−<sup>1</sup> ), υ is the scan rate (V s−<sup>1</sup> ), R is the universal gas constant (8.3145 V A s K−<sup>1</sup> mol−<sup>1</sup> ), T is the temperature (K), n<sup>p</sup> is the number of electrons involved in the non-catalytic Faradaic current response (responsible for the non-catalytic Faradaic current, ip, as described by the Randles-Sevcik equation (Bard and Faulkner, 2001), and ncat is the number of electrons required for catalysis (two electrons for the reduction of CO<sup>2</sup> to CO as shown in Equations 1 and 2). In our calculations of TOF, the first one-electron reduction wave (n<sup>p</sup> = 1) was used for reference to determine the non-catalytic Faradaic current (ip) for **[1-CH3CN]**+, **[2-CH3CN]**<sup>+</sup> and **[3-CH3CN]**+. In contrast, the first reduction wave of **[4-CH3CN]**<sup>+</sup> exhibits a concerted twoelectron event under non-catalytic conditions such that n<sup>p</sup> = 2 at Epc = −1.30 V. The significance of this distinction is important when determining TOF from cyclic voltammetry analysis using

[1-CH3CN]+, [2-CH3CN]+, [3-CH3CN]<sup>+</sup> and [4-CH3CN]<sup>+</sup> under 1 atm CO<sup>2</sup> (red) at <sup>υ</sup> <sup>=</sup> 0.1 V s−<sup>1</sup> with 0.1 M [Bu4N][PF6] acetonitrile supporting electrolyte containing residual 0.17 M (0.3%) H2O as a Brønsted acid source. Cyclic voltammograms recorded under 1 atm of argon are also shown (blue). The current (y-axis) data are normalized with respect to the non-catalytic Faradaic response (ip).

Equation 3 as the ratio of <sup>n</sup> 3 p n 2 cat can differ by a factor of 8 (1/4 vs. 8/4 for n<sup>p</sup> = 1 or 2, respectively). However, this should, with ideal Randles-Sevcik behavior (i<sup>p</sup> ∝ n 3/2 p ) of the twoelectron reduction event, be completely offset by the 8× smaller (icat/ip) 2 ratio when n<sup>p</sup> = 2. Thus, in theory, kinetic analysis of **[4-CH3CN]**<sup>+</sup> can be conducted using either the concerted two-electron reduction event at Epc = −1.30 V (n<sup>p</sup> = 2) or the subsequent one-electron reduction event at Epc = −2.51 V (n<sup>p</sup> = 1) for reference in determining icat/ip. Using data recorded at υ = 0.1 V s−<sup>1</sup> , Equation 3 results in different TOF values



<sup>a</sup>All potentials reported vs. the ferricenium/ferrocene pseudo reference recorded at υ = 0.1 V s−<sup>1</sup> .

<sup>b</sup>Calculated at υ = 0.1 V s−<sup>1</sup> .

<sup>c</sup>TOF calculated using n<sup>p</sup> = 1 electron.

<sup>d</sup>Steady-state conditions not achieved, calculated at υ = 0.1 V s−<sup>1</sup> .

<sup>e</sup>Average TOFmax determined over a range of scan rates at steady-state conditions (see Supporting Information).

of 25 s−<sup>1</sup> (icat/i<sup>p</sup> = 4.0, n<sup>p</sup> = 2) or 14 s−<sup>1</sup> (icat/i<sup>p</sup> = 8.5, n<sup>p</sup> = 1), indicating non-ideal behavior of the concerted twoelectron reduction event, which is unsurprising considering this is an ECE mechanism and not a pure two-electron concerted EE mechanism. Scan rate analysis of this two-electron reduction event for **[4-CH3CN]**<sup>+</sup> clearly demonstrates an increase in peak separation between its cathodic and anodic waves (**Figure S9**). More importantly, the 2.83-fold increase in current for a twoelectron event, predicted by the Randles-Sevcik equation (i<sup>p</sup> ∝ n 3/2 p ) assuming similar diffusion coefficients for each redox state, is in fact found to be significantly smaller (2.13-fold) when we compare the i<sup>p</sup> values of **[4-CH3CN]**<sup>+</sup> at Epc = −1.30 (n<sup>p</sup> = 2) and Epc = −2.51 V (n<sup>p</sup> = 1). The TOF value of 25 s−<sup>1</sup> (icat/i<sup>p</sup> = 4.0, n<sup>p</sup> = 2) is therefore an overestimate of this catalyst's efficiency, and the value of TOF = 14 s−<sup>1</sup> is henceforth quoted for **[4-CH3CN]**<sup>+</sup> (icat/i<sup>p</sup> = 8.5, n<sup>p</sup> = 1) recorded at υ = 0.1 V s−<sup>1</sup> . In summary, the benchmark bpy-based pre-catalyst, **[1-CH3CN]**<sup>+</sup> (icat/i<sup>p</sup> = 11.3) exhibits the highest TOF in this study of 25 s−<sup>1</sup> at υ = 0.1 V s−<sup>1</sup> , followed closely by **[3-CH3CN]**<sup>+</sup> at 22 s−<sup>1</sup> (icat/i<sup>p</sup> = 10.7), the bqn-based system **[4-CH3CN]**<sup>+</sup> at 14 s−<sup>1</sup> (icat/i<sup>p</sup> = 8.5) and finally **[2-CH3CN]**<sup>+</sup> at just 11 s−<sup>1</sup> (icat/i<sup>p</sup> = 7.4).

## Voltammetry Under 1 Atm CO<sup>2</sup> in the Presence of Trifluoroethanol

To further probe the catalytic activity of **[1-CH3CN]**+, **[2- CH3CN]**+, **[3-CH3CN]**<sup>+</sup> and **[4-CH3CN]**+, the voltammetry conditions were altered by adding incremental amounts of a non-aqueous proton source. For the purpose of this study, 2,2,2-trifluoroethanol (TFE, pKa(CH3CN) = 35.4(est.) Lam et al., 2015) was added as a Brønsted acid to observe and optimize proton-coupled catalytic CO<sup>2</sup> reduction. The addition of at least 1.5 M TFE results in a plateau of catalytic current for **[1- CH3CN]**+, **[2-CH3CN]**<sup>+</sup> and **[3-CH3CN]**<sup>+</sup> when monitored at υ = 0.1 V s−<sup>1</sup> (**Figure 7**, **Figures S16–S19**). Notably, each of their catalytic waves exhibit a significant positive shift compared to experiments in the absence of excess Brønsted acid, with Ecat/<sup>2</sup> = −1.94 V for **[1-CH3CN]**<sup>+</sup> and Ecat/<sup>2</sup> = −1.97 V for both **[2- CH3CN]**<sup>+</sup> and **[3-CH3CN]**+. The predominant catalytic peak observed for **[1-CH3CN]**+, **[2-CH3CN]**+, and **[3-CH3CN]**<sup>+</sup> is assigned, with the aid of computations, to the reduction-first

current (y-axis) data are normalized with respect to the non-catalytic Faradaic response (ip). The inset plot of 'TOF vs. TFE concentration' demonstrates that zero-order conditions with respect to TFE concentration were achieved at 2.0 M TFE.

pathway with the five-coordinate, two-electron reduced [fac-Mn<sup>0</sup> (N∧N•−)(CO)3] <sup>−</sup> monoanion being the active catalyst.

Interestingly, a weak grow-in of catalytic current can also be observed at Ecat/<sup>2</sup> = −1.48 V for both **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**<sup>+</sup> (a similar maximum is less discernable for **[3- CH3CN]**+). This lower overpotential weak catalytic current occurs directly from the one-electron reduced species and is thus attributed to catalytic activity of the Mn<sup>0</sup> -Mn<sup>0</sup> dimer intermediate. Indirect support of this hypothesis is the fact that such weak catalytic current is not observed with the bqn precatalyst, **[4-CH3CN]**<sup>+</sup> which we have already established, via IR-SEC studies, does not form a Mn<sup>0</sup> -Mn<sup>0</sup> dimer intermediate. While appearance of this lower energy catalytic pathway is promising, the very weak current observed, and thus the minimal TOF (<1 s−<sup>1</sup> ) prompted us to focus on the more efficient reduction-first pathway for these complexes. Furthermore, this mechanism has already been the focus of other independent studies (Bourrez et al., 2014; Neri et al., 2019).

In contrast to the above mentioned studies in the absence of excess Brønsted acid, the presence of optimal TFE concentrations benefitted the pursuit of steady-state catalytic conditions, thus allowing for an estimation of a maximum turnover frequency (TOFmax), at least in the case of **[1-CH3CN]**<sup>+</sup> and **[2- CH3CN]**+. A collection of scan-rate dependent voltammetry data is provided in the **Supporting Information** for all precatalysts (**Figures S20–S23**) with kinetic analysis summarized in **Table 3**. Pre-catalysts **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**<sup>+</sup> perform comparably with TOFmax values estimated at 75 s−<sup>1</sup> under steady-state catalytic conditions. Pre-catalyst **[3-CH3CN]**<sup>+</sup> still exhibits a scan-rate dependent current upon increasing scan rate >0.1 V s−<sup>1</sup> , thus preventing pure kinetic steady-state conditions, and hence TOFmax, to be obtained. Thus, an estimated TOF value of 67 s−<sup>1</sup> is reported for **[3-CH3CN]**<sup>+</sup> recorded at a scan rate of υ = 0.1 V s−<sup>1</sup> (**Figure S22**). Results from the addition of TFE to **[4-CH3CN]**<sup>+</sup> were rather complicated due to the growin of an additional pre-wave at approx. −2.42 V which prevented accurate determination of Ecat/<sup>2</sup> or TOFmax (**Figure S23**). At least from a qualitative perspective, it can be stated that TFE addition does give rise to a modest increase in the observed catalytic current maximum (**Figure 8**). Unfortunately, this phenomenon of multiple catalytic waves raises concerns about product selectivity and competitive side reactions, which are borne true following controlled potential electrolysis experiments, discussed below, which conclusively confirm hydrogen evolution as not just a competitive but a dominant process under these experimental conditions with **[4-CH3CN]**+. Mechanistic details of competitive hydrogen evolution vs. CO<sup>2</sup> reduction are discussed in more detail below via computational analysis.

### Controlled Potential Electrolysis

Product selectivity of all four pre-catalysts in the absence and presence of TFE under 1 atmosphere of CO<sup>2</sup> was investigated by controlled potential electrolysis (CPE) with in-situ gas chromatography analysis used for CO and H<sup>2</sup> quantification over a time period of 4 h. The potential bias applied in each case corresponded to Ecat/<sup>2</sup> for specific experimental conditions as summarized in **Table 3** for each catalyst. The tabulated Faradaic efficiency (FE) data in **Table 4** represent the peak CO selectivity observed at a single time interval over the duration of the experiment, whereas the turnover number (TON) tabulated for both CO and H<sup>2</sup> represents the total TON over the entire course of the experiment. Plots summarizing all electrolysis data are provided in **Figures S28–S35**. Ultimately, all pre-catalysts yielded CO product regardless of the conditions employed. Each of the pre-catalysts, **[1-CH3CN]**+, **[2-CH3CN]**+, and **[3- CH3CN]**<sup>+</sup> exhibited a clear selectivity for CO production in the absence or presence of TFE with only **[4-CH3CN]**<sup>+</sup> yielding significant H<sup>2</sup> in the presence of 1.5 M TFE. Unfortunately, however, in most cases there is FE that is unaccounted for. This is especially true for pre-catalysts **[2-CH3CN]**<sup>+</sup> and **[3- CH3CN]**<sup>+</sup> which is likely due to rapid decomposition, consistent with their poor TONs for CO or H<sup>2</sup> evolution. Pre-catalyst **[1- CH3CN]**<sup>+</sup> performed significantly better, exhibiting percentage FE's of 62:2 CO:H<sup>2</sup> in the absence of TFE, which increased to 84:2 in the presence of 2.0 M TFE. Efforts to quantify any formate (HCO<sup>−</sup> 2 ) production to account for the full FE of all catalysts were in vain. This is likely due to the low turnover numbers of these catalysts, which as a whole performed quite poorly under the experimental conditions employed; a recognized problem for the [fac-Mn(N∧N)(CO)3] <sup>−</sup> class of electrocatalysts under homogeneous conditions (Grills et al., 2018). Indeed, two prior CPE studies of the [fac-MnBr(phen)(CO)3] pre-catalyst have each highlighted the poor performance of this system with the FECO ranging from 18 to 57% (Stanbury et al., 2017; Tignor et al., 2018). Interestingly, the bqn-derived pre-catalyst, **[4-CH3CN]**<sup>+</sup> exhibited the greatest FECO of 98% in the absence of TFE, albeit with a very low TON of just 3. However, with such a low TON we cannot rule out catalyst decomposition as a contributing factor. As anticipated from its irregular voltammetry behavior,

at least in comparison to **[1-CH3CN]**+, **[2-CH3CN]**+, and **[3- CH3CN]**+, upon the introduction of TFE, a dramatic shift in product selectivity was observed, with FE's of 14:69 CO:H2. As discussed below in the computational section, this shift in product selectivity is possibly a consequence of the reduced nucleophilicity of the two-electron reduced **[4]**<sup>−</sup> active catalyst, in large part due to the lower lying π <sup>∗</sup> orbitals of the bqn ligand.

### Computational Analysis

Density functional theory (DFT) calculations at the M06 level of theory (Zhao and Truhlar, 2008a,b, 2010) in conjunction with the SMD continuum solvation model (Marenich et al., 2009) for acetonitrile were performed to examine the catalyst activation and electrocatalytic CO<sup>2</sup> reduction mechanisms of **[1-CH3CN]**+, **[2-CH3CN]**+, and **[4-CH3CN]**+. Mechanistic calculations on **[3-CH3CN]**<sup>+</sup> were omitted for the sake of brevity due to its similarity in behavior to **[2-CH3CN]**+. The results are summarized in **Schemes 1** and **2** along with tabulated energetics (**Tables 5**, **6**). Further details on the electronic structures of selected reaction intermediates are provided in **Supporting Information**.

### Catalyst Activation and Dimer Formation

The activation of each catalyst starts with one-electron reduction of the solvent-coordinated **[Mn-CH3CN]**<sup>+</sup> species. For all considered pathways in the present study, **[1-CH3CN]**<sup>+</sup> and **[2- CH3CN]**<sup>+</sup> exhibit similar energetics (**Scheme 1**), so henceforth we will focus on the activation of **[1-CH3CN]**<sup>+</sup> to highlight the distinct behavior of **[4-CH3CN]**+. The one-electron reduction of **[Mn-CH3CN]**<sup>+</sup> (E1) results in the formation of a ligand-based radical anion for all of the complexes investigated (**Figure S36**).

TABLE 4 | Summary of controlled potential electrolysis data in the absence of TFE and at optimum TFE concentrations<sup>a</sup> .


<sup>a</sup>Applied potential was equal to Ecat/<sup>2</sup> as summarized in Table 3 for each catalyst. <sup>b</sup>Peak CO:H<sup>2</sup> FE ratio observed over a 4 h duration.

<sup>c</sup>Totaled over a 4 h duration.

Typically observed as an EC mechanism in voltammetry (E11G<sup>1</sup> in **Scheme 1**), subsequent dissociation of the acetonitrile ligand (1G1) leads to localization of the unpaired spin on the Mn center, generating a formally Mn<sup>0</sup> species (**[Mn]<sup>0</sup>** ) (**Figure S36**). The extensive π-conjugation of **[4-CH3CN]**<sup>+</sup> introduces an ∼+0.5 V anodic shift (E<sup>1</sup> = −1.23 V) compared to that of **[1-CH3CN]**<sup>+</sup> (E<sup>1</sup> = −1.74 V), in line with the assumption of an initial ligand-based reduction. The computed energetics indicate that acetonitrile dissociation is very favorable upon reduction of **[1- CH3CN]**<sup>+</sup> (1G<sup>1</sup> = −6.6 kcal/mol) to generate pentacoordinate **[1]<sup>0</sup>** , further reduction of which (E<sup>5</sup> = −1.79 V) will form **[1]**−. In contrast, theory predicts that **[4-CH3CN]<sup>0</sup>** and **[4]<sup>0</sup>** will coexist based on acetonitrile dissociation being nearly isoergic (1G<sup>1</sup> = −0.4 kcal/mol). This distinct behavior is responsible for the experimentally observed concerted two-electron reduction of **[4-CH3CN]**<sup>+</sup> to **[4]**<sup>−</sup> (E<sup>3</sup> = −1.36 V), which requires a lower potential than the sequential one-electron reduction pathway, i.e., **[4-CH3CN]<sup>0</sup>** to **[4]**<sup>−</sup> conversion (E<sup>4</sup> = −1.48 V) in contrast to **[1-CH3CN]**<sup>+</sup> (E<sup>3</sup> = −1.62 V and E<sup>4</sup> = −1.50 V).

Another intriguing difference is in the energetics of dimerization of **[Mn]<sup>0</sup>** to **[Mn-Mn]<sup>0</sup>** , which is quite favorable for **[1]<sup>0</sup>** (1G<sup>2</sup> = −11.5 kcal/mol), in contrast to a nearly isoergic driving force in the case of **[4]<sup>0</sup>** (1G<sup>2</sup> = −1.2 kcal/mol). Closer inspection of the dimer geometries indicates that staggered conformations are more favorable and Mn–Mn distances are predicted as 2.96 Å and 3.16 Å for **[1–1]<sup>0</sup>** and **[4–4]<sup>0</sup>** , respectively (**Figure S37**). This difference is partly attributed to ligand-induced steric effects that lead the Mn center to be outof-plane with respect to the ligand in the case of **[4]<sup>0</sup>** (for **[4]<sup>0</sup>** , Cbridge-N-Mn-N is 23.2◦ , compared to 1.1◦ for **[1]<sup>0</sup>** ). It should be noted that dimerization of **[4]<sup>0</sup>** is even further suppressed due to the predicted equilibrium between hexacoordinate **[4-CH3CN]<sup>0</sup>** and pentacoordinate **[4]<sup>0</sup>** , even if the latter forms during electrocatalysis. As a result of the computational analysis described above, **[1-CH3CN]**<sup>+</sup> and **[2-CH3CN]**<sup>+</sup> are predicted to form **[1]<sup>0</sup>** and **[2]<sup>0</sup>** , respectively, via an EC process followed by fast dimerization to form **[1–1]<sup>0</sup>** and **[2–2]<sup>0</sup>** , respectively, which can then be reduced via sequential one-electron processes, or a concerted two-electron reduction process, to generate

TABLE 5 | Computed reduction potentials (V vs. Fc+/<sup>0</sup> ) and free energy changes (1G, kcal/mol) of catalyst activation and dimerization pathways for Mn<sup>I</sup> polypyridyl complexes [1-CH3CN]+, [2-CH3CN]+, and [4-CH3CN]+.


the catalytically active **[1]**<sup>−</sup> and **[2]**<sup>−</sup> species. We should note that two-electron reduction of the **[Mn-Mn]<sup>0</sup>** dimer practically proceeds via two sequential one-electron reduction steps as the dimer is expected to decompose into **[Mn]<sup>0</sup>** and **[Mn]**<sup>−</sup> upon the first reduction, and the computed reduction potential of **[Mn]<sup>0</sup>** is more anodic than that of **[Mn-Mn]<sup>0</sup>** (e.g., E<sup>5</sup> = −1.79 V for **[1]<sup>0</sup>** vs. E<sup>6</sup> = −2.29 V for **[1–1]<sup>0</sup>** ).

### CO<sup>2</sup> Binding and CO Evolution

Next, we turn our attention to CO<sup>2</sup> binding to the twoelectron reduced **[Mn]**<sup>−</sup> catalyst and subsequent steps of electrocatalytic CO<sup>2</sup> reduction (**Scheme 2**). Similar to earlier reports (Riplinger et al., 2014), CO<sup>2</sup> binding to **[1]**<sup>−</sup> is computed to be uphill (1G<sup>3</sup> = 8.6 kcal/mol) and is driven by proton transfer from a Brønsted acid to generate **[1-CO2H]<sup>0</sup>** (pKa1 = 24.8). Interestingly, we could not locate an optimized structure of CO2-bound **[4-CO2]** <sup>−</sup>, indicating that since the potential to generate **[4]**<sup>−</sup> is nearly +0.5 V more positive compared to that of **[1]**−, the pentacoordinate **[4]**<sup>−</sup> does not possess enough reducing power to activate CO2. On the other hand, further reduction to **[4]2**<sup>−</sup> (**Figure S38**), results in increased reactivity toward CO<sup>2</sup> to generate **[4-CO2] <sup>2</sup>**<sup>−</sup> (1G<sup>4</sup> = 10.8 kcal/mol), and subsequent protonation yields **[4-CO2H]**<sup>−</sup> (pKa3 = 29.4) (**Scheme 2**). It should be noted that the spin density (**Figure S38**) and total electron density difference (**Figure S39**) plots indicate that the third reduction is predominantly ligand centered in **[4]2**<sup>−</sup> and best characterized as [Mn<sup>0</sup> -bqn2−] 2−, although the additional negative charge is shared between the metal center and the ligand in the case of **[1]2**<sup>−</sup> and **[2]2**<sup>−</sup> such that the electronic structure exhibits a resonance between [Mn−-(N∧N)•−] <sup>2</sup><sup>−</sup> and [Mn<sup>0</sup> -(N∧N)2−] <sup>2</sup>−. **[Mn-CO2H]**<sup>−</sup> is predicted to be the common intermediate for both **[1-CH3CN]**<sup>+</sup> and **[4-CH3CN]**<sup>+</sup> electrocatalysts, as C–OH bond cleavage is expected to proceed predominantly via the reduction-first pathway (**Scheme 2**). Scission of the C–OH bond in **[Mn-CO2H]**<sup>−</sup> assisted by TFE leads to **[Mn-CO]<sup>0</sup>** . CO evolution may occur spontaneously from **[Mn-CO]<sup>0</sup>** prior to one-electron reduction (Grice et al., 2013) and generation of **[Mn]**−. However, this mechanism is yet to be experimentally verified for this class of Mn catalyst. In contrast to **[1-CH3CN]**+, **[4-CH3CN]**<sup>+</sup> produces H<sup>2</sup> as the dominant product in the presence of TFE as a Brønsted acid, which is attributed to less favorable interaction of **[4]**<sup>−</sup>

(1G) are in units of kcal/mol and reduction potentials are in units of volts (V) vs. Fc+/<sup>0</sup>

.

TABLE 6 | Computed reduction potentials (V vs. Fc+/<sup>0</sup> ), free energy changes (1G, kcal/mol) and activation free energies (1G‡ , kcal/mol), and pKa's relevant for electrocatalytic reduction of CO<sup>2</sup> to CO for Mn<sup>I</sup> polypyridyl complexes

[1-CH3CN]+, [2-CH3CN]+, and [4-CH3CN]+, shown in Scheme 2, comparing the protonation-first and reduction-first pathways vs. the three-electron reduction pathway.


with CO<sup>2</sup> compared to **[1]**−. However, we should also note that the pKa's of **[1]**<sup>−</sup> (pK<sup>a</sup> = 24.8) and **[1]2**<sup>−</sup> (pK<sup>a</sup> = 36.0) are significantly higher than their counterparts, **[4]**<sup>−</sup> (pK<sup>a</sup> = 17.3) and **[4]2**<sup>−</sup> (pK<sup>a</sup> = 29.0), indicating that hydride formation is not as favorable in the latter either. In the absence of a Brønsted acid as a proton source, all the catalysts are predicted to form the three-electron reduced **[Mn]2**<sup>−</sup> active catalyst before binding CO<sup>2</sup> and subsequently producing CO and CO2<sup>−</sup> 3 via interaction of a second CO<sup>2</sup> molecule with **[Mn-CO2] <sup>2</sup>**−.

### CONCLUSIONS

Through a systematic variation of the polypyridyl ligand from bpy to phen to dmphen and finally bqn, both stericand electronic-based ligand influences on the activity of a [fac-Mn(N∧N)(CO)3(CH3CN)]<sup>+</sup> class of CO<sup>2</sup> reduction precatalysts has been established, providing critical insight into the manipulation of CO<sup>2</sup> binding affinities and resulting product selectivity for these catalysts. Through a combination of IR-SEC and computational studies, the combined electronic and steric influences of the π-extended bqn ligand have been probed in **[4- CH3CN]**+, where formation of the **[4–4]<sup>0</sup>** dimer is hindered and a concerted two-electron ECE mechanism for the generation of **[4]**<sup>−</sup> is favored. Computations have revealed how the lower lying π ∗ -orbitals of the bqn ligand have rendered this pentacoordinate **[4]**<sup>−</sup> intermediate inactive with respect to CO2, even in the presence of a Brønsted acid. In contrast to the bpy, phen, and dmphen derived catalysts, this has resulted in a shift in product selectivity for the **[4-CH3CN]**<sup>+</sup> pre-catalyst to favor H<sup>2</sup> evolution in the presence of excess TFE. However, electrochemical and computational investigations have established successful CO<sup>2</sup> activation following in-situ generation of a three-electron reduced [fac-Mn(−1)(N∧N•−)(CO)3] <sup>2</sup><sup>−</sup> active catalyst, a first for any [fac-Mn(N∧N)(CO)3(CH3CN)]<sup>+</sup> pre-catalyst. Although at the cost of additional overpotential, this three-electron pathway results in increased reactivity toward CO<sup>2</sup> to generate the previously established [fac-Mn(I)(CO2H)(N∧N•−)(CO)3] − intermediate, which subsequently propagates the catalytic cycle via the standard reduction-first pathway involving ratedetermining, proton-coupled C–OH bond cleavage.

# MATERIALS AND METHODS

Acetonitrile (ACS reagent grade, 99.5%), bromopentacarbonylmanganese(I) (98%), potassium carbonate (>99%), silver trifluoromethanesulfonate (>99%), tetrahydrofuran (anhydrous, 99.9%) and 2,2,2-trifluoroethanol (>99%) were purchased from Sigma Aldrich and used as received. Dichloromethane (ACS reagent grade, >99.9%) and diethyl ether were purchased from Pharmco-Aaper (ACS reagent grade, >99.9%) and used as received. The water content in ACS reagent grade acetonitrile was confirmed by Karl-Fisher titration to be 0.17 M (0.3%). Tetrabutylammonium hexafluorophosphate (99%, Sigma Aldrich) was recrystallized thrice from ethanol and dried under vacuum prior to electrolyte preparation. Steady-state FTIR spectra were recorded on a Thermo Nicolet 670 FTIR spectrophotometer using a liquid cell with CaF<sup>2</sup> windows in spectrophotometric grade acetonitrile (99.5%, Sigma Aldrich) solvent. NMR spectra were recorded on an Agilent spectrometer operated at 399.80 MHz for <sup>1</sup>H nuclei. CD3CN was used as received from Sigma Aldrich and its residual <sup>1</sup>H solvent signal used as an internal reference for reporting the chemical shift (δ = 1.96 ppm). Voltammetry and bulk electrolysis were carried out on a CH Instruments 620E potentiostat. A custom threeelectrode cell was used for both voltammetry and bulk electrolysis experiments allowing airtight introduction of working, counter and reference electrodes as well as septa for gas purging. For cyclic voltammetry, glassy carbon (3 mm diameter) and Pt wire were used as working and counter electrodes, respectively, with 0.1 M [Bu4N][PF6] in ACS reagent grade acetonitrile as the supporting electrolyte. A non-aqueous reference electrode was used to minimize ohmic potential drop at the solvent interface. This consisted of a Ag wire in 0.10 M [Bu4N][PF6] acetonitrile supporting electrolyte isolated by a vycor frit and was calibrated in-situ using the ferricenium/ferrocene redox couple as an internal reference. Redox potentials (E) were determined from cyclic voltammetry as (Epa + Epc)/2, where Epa and Epc are the anodic and cathodic peak potentials, respectively. Where E could not be calculated due to irreversible behavior, Epc or Epa are reported accordingly. For CO<sup>2</sup> concentration dependent studies, gas cylinders were ordered from Airgas containing pre-mixed ratios of Ar:CO<sup>2</sup> (100:0, 80:20, 60:40, 50:50, 40:60, 20:80, 0:100). For controlled potential bulk electrolysis experiments a vitreous carbon (Duocell) working electrode soldered to a copper wire was used. A Pt gauze counter electrode was used, isolated from the main compartment by a fine porosity vycor tube+frit to minimize mass transfer resistance. Gas chromatography data were recorded on a custom Shimadzu GC-2014 instrument where a Ni "methanizer" catalyst was used to convert CO to CH<sup>4</sup> prior to quantification of CH<sup>4</sup> by the thermal conductivity detector. H<sup>2</sup> was simultaneously monitored by a flame ionization detector during the same injection. The GC was pre-calibrated for CO and H<sup>2</sup> quantification by mimicking bulk electrolysis conditions (i.e., 5 mL supporting electrolyte in the same cell, with electrodes, under 1 atm CO2). Standard curves for H<sup>2</sup> and CO were generated using this cell where known volumes of the analyte gas (H<sup>2</sup> or CO) were injected and the solution stirred for 30 min to allow equilibration of the analyte between the electrolyte and headspace prior to GC injection.

# General Synthesis of [fac-Mn(N∧N)(CO)3(OTf)] Complexes 1-4

Bromopentacarbonyl manganese(I) (100 mg, 0.36 mmol) and silver trifluoromethanesulfonate (93.5 mg, 0.36 mmol) were charged to a 50-mL round bottomed flask with approximately 25 mL of degassed dichloromethane under 1 atmosphere of argon. The solution was allowed to mix in the dark for 1 h at room temperature after which the AgBr precipitate was removed by filtration through celite. The dichloromethane solvent was subsequently removed by a rotary evaporator. After confirming quantitative transformation to the [fac-Mn(OTf)(CO)5] intermediate by FTIR spectroscopy, the solid was transferred to a 10-mL microwave vial with 3 mL of tetrahydrofuran. To this solution was added a slight deficit of the appropriate ligand (0.33 mol). The vial was then sealed and reacted in a CEM Discovery microwave reactor at 70◦C for 10 min. Each reaction afforded a bright yellow-orange solution. The tetrahydrofuran was reduced in volume to roughly 0.5 mL on a rotary evaporator at which point the pure product was immediately precipitated by addition of excess diethyl ether. The solid was isolated via vacuum filtration, rinsed with diethyl ether and dried under vacuum. No further purification was necessary.

**[fac-Mn(OTf)(bpy)(CO)3] (1)** FTIR (CH3CN) ν(CO): 2,050, 1,958 cm−<sup>1</sup> . <sup>1</sup>H-NMR (CD3CN) δ: 7.70–7.73 (2H, m), 8.21–8.25 (2H, m), 8.40 (2H, d, J = 8.0 Hz), 9.14 (2H, d, J = 5.6 Hz) ppm. Anal. Calcd. for C14H8F3MnN2O6S: C, 37.85; H, 1.82; N, 6.31. Found: C, 38.27; H, 1.99; N, 6.02.

**[fac-Mn(OTf)(phen)(CO)3] (2)** FTIR (CH3CN) ν(CO): 2,050, 1,958 cm−<sup>1</sup> . <sup>1</sup>H-NMR (CD3CN) δ: 8.04 (2H, dd, J<sup>1</sup> = 4.0, J<sup>2</sup> = 5.2 Hz), 8.20 (2H, s), 8.78 (2H, dd, J<sup>1</sup> = 1.2, J<sup>2</sup> = 8.0 Hz), 9.49 (2H, dd, J<sup>1</sup> = 1.2, J<sup>2</sup> = 4.8 Hz) ppm. Anal. Calcd. for C16H8F3MnN2O6S: C, 41.04; H, 1.72; N, 5.98. Found: C, 41.36; H, 1.88; N, 5.42.

**[fac-Mn(OTf)(dmphen)(CO)3] (3)** FTIR (CH3CN) ν(CO): 2,046, 1,959, 1,944(sh) cm−<sup>1</sup> . <sup>1</sup>H-NMR (CD3CN) δ: 3.30 (6H, s), 7.90 (2H, d, J = 8.0 Hz), 8.04 (2H, s), 8.57 (2H, d, J = 8.0 Hz). Anal. Calcd. for C18H12F3MnN2O6S: C, 43.56; H, 2.44; N, 5.64. Found: C, 44.05; H, 2.69; N, 5.18.

**[fac-Mn(OTf)(bqn)(CO)3] (4)** FTIR (CH3CN) ν(CO): 2,047, 1,959 cm−<sup>1</sup> . <sup>1</sup>H-NMR (CD3CN) δ: 7.90 (2H, dd, J<sup>1</sup> = J<sup>2</sup> = 8.0 Hz), 8.12 (2H, dd, J<sup>1</sup> = J<sup>2</sup> = 8.0 Hz), 8.21 (2H, d, J = 8.0 Hz), 8.61 (2H, d, J = 8.0 Hz), 8.83 (2H, d, J = 8.0 Hz), 8.88 (2H, d, J = 8.0 Hz) ppm. Anal. Calcd. for C22H12F3MnN2O6S: C, 48.54; H, 2.22; N, 5.15. Found: C, 49.10; H, 2.60; N, 4.91.

### Computational Methods Density Functional Theory

All geometries were fully optimized at the M06 level of density functional theory (Zhao and Truhlar, 2008a,b, 2010) with the SMD continuum solvation model (Marenich et al., 2009) for acetonitrile as solvent using the Stuttgart [8s7p6d2f | 6s5p3d1f] ECP10MDF contracted pseudopotential basis set (Dolg et al., 1987) on Mn and the 6-31G(d) basis set on all other atoms (Hehre et al., 1986). Non-analytical integrals were evaluated using the integral = grid = ultrafine option as implemented in the Gaussian 16 software package (Frisch et al., 2016). The nature of all stationary points was verified by analytic computation of vibrational frequencies, which were also used for the computation of zero-point vibrational energies, molecular partition functions, and for determining the reactants and products associated with each transition-state structure (by following the normal modes associated with imaginary frequencies). Partition functions were used in the computation of 298 K thermal contributions to the free energy employing the usual ideal-gas, rigid-rotator, harmonic oscillator approximation (Cramer, 2004). Free-energy contributions were added to singlepoint, SMD-solvated M06 electronic energies computed at the optimized geometries obtained with the initial basis with the SDD basis set on Mn and the larger 6-311+G(2df,p) basis set on all other atoms to arrive at final, composite free energies.

### Solvation and Standard Reduction Potentials

As mentioned above, solvation effects for acetonitrile were accounted for by using the SMD continuum solvation model. A 1 M standard state was used for all species in solution (except for acetonitrile as solvent for which the standard state was assigned as 19.14 M). Thus, the free energy in solution is computed as the 1 atm gas-phase free energy, plus an adjustment for the 1 atm to 1 M standard-state concentration change of RT ln (24.5), or 1.9 kcal/mol, plus the 1 M to 1 M transfer (solvation) free energy computed from the SMD model. Standard reduction potentials were calculated for various possible redox couples to assess the energetic accessibility of different intermediates at various oxidation states. For a redox reaction of the form

$$\begin{array}{rcl} O\_{\text{(soln)}} & + & n \, e^-\_{\text{(g)}} \to & R\_{\text{(soln)}} \end{array} \tag{4}$$

where O and R denote the oxidized and reduced states of the redox couple, respectively, and n is the number of electrons involved in redox reaction, the reduction potential E o O|R relative to SCE was computed as

$$E\_{O|R}^{\circ} = -\frac{\Delta G\_{O|R}^{\circ}}{nF} - \Delta E\_{ref}^{\circ} \tag{5}$$

where 1G o O|R is the free energy change associated with Equation 1 (using Boltzmann statistics for the electron) and 1E o ref

is taken as 0.141 V (Keith et al., 2013), which is required for the conversion of calculated E o O|R vs. normal hydrogen electrode (NHE) in aqueous solution (ENHE = −4.281 V) (Kelly et al., 2006) to E o O|R vs. the saturated calomel electrode (SCE) in acetonitrile (ESCE = −4.422 V) (Isse and Gennaro, 2010). We obtained reduction potentials referenced to the ferricenium/ferrocene couple by using a shift of −0.384 V from E o O|R vs. SCE.

### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the manuscript/**Supplementary Files**.

### AUTHOR CONTRIBUTIONS

MM, VB, and JR contributed to the synthesis, spectroscopic, and electrochemical characterization of all complexes studied. KN and DG contributed IR-SEC studies. ME contributed all computational studies. MM, ME, DG, and JR contributed equally to manuscript preparation.

### REFERENCES


### FUNDING

JR thanks the National Science Foundation for support under Grant No. CHE-1800062. This work at BNL (DG and ME) was supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences & Biosciences, under Contract No. DE-SC0012704. KN was grateful to the DOE for an Office of Science Graduate Student Research (SCGSR) award.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00628/full#supplementary-material

Scan rate dependent voltammetry studies under inert and catalytic conditions; experimental and computational IR-SEC spectra; catalyst, acid and CO<sup>2</sup> concentration dependence; controlled potential electrolysis plots; computational spindensity plots, total electron density difference plots, and cartesian coordinates. These materials are provided in the supporting information document.

five-coordinated anions [Mn(CO)3(α-diimine)]<sup>−</sup> from fac-[Mn(X)(CO)3(αdiimine)] (X = halide) at low temperatures. Recl. Trav. Chim. 114, 565–570. doi: 10.1002/recl.19951141123


in Mn and Re tricarbonyl complexes. Organometallics 38, 1317–1329. doi: 10.1021/acs.organomet.8b00584


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 McKinnon, Belkina, Ngo, Ertem, Grills and Rochford. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# On-Surface Modification of Copper Cathodes by Copper(I)-Catalyzed Azide Alkyne Cycloaddition and CO<sup>2</sup> Reduction in Organic Environments

Ryota Igarashi, Ryuji Takeuchi, Kazuyuki Kubo, Tsutomu Mizuta and Shoko Kume\*

*Department of Chemistry, Graduate School of Science, Hiroshima University, Higashihiroshima, Japan*

In this study, organic structures were introduced onto copper cathodes to induce changes in their electrocatalytic CO<sup>2</sup> reduction activity. Poorly soluble organic polymers were distributed onto the copper surface as a thin layer by polymerizing monomeric precursors *via* a copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC) activated by anodization of the copper substrate. The resulting structure possesses copper surface atoms that are available to participate in the CO<sup>2</sup> reduction reaction—comparable to close-contact organic structures—and stabilize the adsorption of organic layers through the CO<sup>2</sup> reduction process. The CO<sup>2</sup> reduction performance of the on-surface modified copper cathode exhibited improved CO<sup>2</sup> reduction over H<sup>2</sup> evolution compared with traditional cast modification systems. Preventing organic moieties from forming densely packed assemblies on the metal surface appears to be important to promote the CO<sup>2</sup> reduction process on the copper atoms. The suppression of H<sup>2</sup> evolution, a high methane/ethylene ratio, and the influence of stirring demonstrate that the improved CO<sup>2</sup> reduction activity is not only a result of the copper atom reorganization accompanied by repeating anodization for modification; the organic layer also apparently plays an important role in proton transfer and CO<sup>2</sup> accumulation onto the copper surface.

### Edited by:

*Hitoshi Ishida, Kitasato University, Japan*

### Reviewed by:

*Enrico Andreoli, Swansea University, United Kingdom Tomohiko Inomata, Nagoya Institute of Technology, Japan*

> \*Correspondence: *Shoko Kume skume@hiroshima-u.ac.jp*

### Specialty section:

*This article was submitted to Inorganic Chemistry, a section of the journal Frontiers in Chemistry*

Received: *30 January 2019* Accepted: *26 November 2019* Published: *17 December 2019*

### Citation:

*Igarashi R, Takeuchi R, Kubo K, Mizuta T and Kume S (2019) On-Surface Modification of Copper Cathodes by Copper(I)-Catalyzed Azide Alkyne Cycloaddition and CO*<sup>2</sup> *Reduction in Organic Environments. Front. Chem. 7:860. doi: 10.3389/fchem.2019.00860* Keywords: CO<sup>2</sup> reduction, copper cathode, CuAAC, organic modification, hydrocarbons

# INTRODUCTION

The reduction of CO<sup>2</sup> on metal copper cathodes has been of interest since it was first reported by Hori in 1985 (Hori et al., 1985, 1986). The specific character of copper allows highly reduced hydrocarbons to be obtained at relatively large negative potentials, unlike other catalysts, which mainly afford formate and CO (Gattrell et al., 2006; Peterson and Nørskov, 2012; Zhang et al., 2014; Feaster et al., 2017).

Since common polycrystalline copper surfaces can yield various types of hydrocarbons, product distribution is difficult to control, especially for the selective formation of valuable C2-C3 products such as ethylene and ethanol. Product distribution is known to be highly influenced by the applied potential (Hori et al., 2003; Gattrell et al., 2006; Kuhl et al., 2012), Cu crystal facet (Hori et al., 2003; Gupta et al., 2006; Schouten et al., 2012; Huang et al., 2017; Qiu et al., 2017), and proton transfer conditions (Hori et al., 1997; Singh et al., 2016; Varela et al., 2016; Ooka et al., 2017). Oxide-derived Cu nanostructures are of current interest (Kas et al., 2014; Ren et al., 2015; Dutta et al., 2016; Handoko et al., 2016; Mistry et al., 2016; Huang et al., 2017; Mandal et al., 2018); however, their high C2-C3 selectivity has not been fully elucidated, as these materials typically include multiple Cu facets endowing various activities. Furthermore, their nano-scale morphology influences local pH, and the remaining oxygen atoms are considered to influence the electronic nature of the surface. Other nanostructured Cu materials synthesized via various preparation methods (Tang et al., 2011; Li and Kanan, 2012; Reske et al., 2014; Ma et al., 2015; Kim et al., 2017; Zhao et al., 2017; Jeon et al., 2018; Luna et al., 2018) and alloys comprising copper and other elements are reported to show excellent performance in C2-C3 product formation (Long et al., 2017; Ma et al., 2017; Zhang et al., 2017).

The introduction of organic structures onto catalytically active metal surfaces has recently received attention, particularly the preparation of self-assembled monolayers (SAMs). Traditionally, the adsorption of organic molecules has been used to deactivate pristine metal surfaces and is often exploited to achieve higher selectivity, as demonstrated by the Lindlar catalyst. Recent developments have shown that organic molecules have a positive impact on enhancing selectivity and activity (Schoenbaum et al., 2014). Regarding CO<sup>2</sup> reduction, several groups have introduced organic molecules onto copper surfaces as SAMs to improve the reaction selectivity of CO and CO<sup>2</sup> reduction (Xie et al., 2016; Gong et al., 2017; Ahn et al., 2018). In these studies, phase-separated and densely packed structures of organic molecules on metal surfaces are carefully avoided, and the design of open-surface metal centers with neighboring organic structures to allow their cooperation seems to be a prerequisite for a productive reaction environment. To obtain the right reaction environment, similar to coordination catalysts, in which vacant metal centers can cooperate with organic ligands, requires that specific strategies be adopted in relation to contactsurface preparation.

Herein, we have developed a new method to modify metallic copper cathodes with organic layers, in contrast to SAM adsorption routes. To place open metal surface atoms in the neighborhood of the organic structure and to enhance stability, we adopted a rigid organic polymer structure. These polymers are poorly soluble in common solvents and are difficult to distribute homogeneously on the surface without aggregation. To distribute the polymer structure across the surface as a thin layer, the monomeric precursors were polymerized on the copper electrode by exploiting the Cu(I)-catalyzed azide-alkyne cycloaddition (CuAAC) catalytic activity response to surface oxidation (**Figure 1**). In this report, the copper modification achieved via the on-surface CuAAC approach was studied, and an efficient CO<sup>2</sup> reduction activity, specific to the polymermodified electrode, was observed.

## LAYER PREPARATION AND CHARACTERIZATION

### Preparation

Electrochemical on-surface modification was performed in an electrolyte solution containing ethynyl and azide monomers (**Figure 2**). The redox activity of the surface copper was observed by repetitive anodic scanning of the polycrystalline copper electrode. In the absence of organic additives, the dissolution of oxidized copper from the surface proceeded continuously with an onset potential of ca. −0.6 V (**Figure 3A**). The first anodic scan exposed a fresh surface, making the current double afterward, and the anodic current profile was constant without decrease in the subsequent scans. When **1E** or **3E** was present in the solution, the onset potential of the copper anodization shifted positively, indicating that the ethynyl moieties promoted adsorption on the copper surface and thus inhibited copper stripping **(Figures 3B–E)**. The appearance of a negatively shifted cathodic peak at ca. −1 V indicates coordination of the dissolved copper ion species. The ethynyl moiety likely formed insoluble coordination polymers with copper(I) through sigma- and picoordination bonds (Abrantes et al., 1984), and the anodic scanning appeared to first dissolve the copper(I) species into the solution (Ahrland, 1982), with subsequent coordination of ethynyl moieties. When azide precursors coexisted, these peaks were less recognizable (**Figures 3B,D**), implying that ethynylcopper(I) was consumed in the subsequent CuAAC

reaction. Upon scanning, the anodic current gradually decreased, and an anti-corrosive growth layer was observed on the surface. The current decrease was prominent in **3E**+**3A**, and to a smaller degree, when the solution contained **3E**, **1E** alone or **1E**+**1A**. The covalent polymers composed of **3E**+**3A**, via the CuAAC reaction, appear to induce strong adsorption compared with **3E** without covalent bond formation. The anodic scanning of **1E**+**1A** is expected to afford monomeric 1,2,3 triazoles because they contain only one reaction point for each, resulting in very weak adsorption. These features show that the anticorrosive layer efficiently covered the Cu surface by anodization with **3E**+**3A**.

### X-Ray Photoelectron Spectroscopy (XPS)

XPS analysis of the **Cu\_3E**+**3A**-modified Cu surface showed that a fair amount of nitrogen was introduced onto the surface (**Figure 4**). The binding energy associated with the introduced N species does not indicate an azide moiety, as this exhibited a peak located at ∼405 eV (Collman et al., 2006; Chisholm et al., 2016). The majority of the nitrogen, when affixed to the surface, appeared in the form of a triazole. Conversely, in addition to the Cu 2p peaks, **Cu\_1E**+**1A** also exhibited a small nitrogen peak upon modification (**Figure S1**). These nitrogen features correspond to the difference in adsorption of polymeric and monomeric structures, as discussed in the on-surface preparation section, showing a significant and robust organic modification in polymeric **Cu\_3E**+**3A**.

### Scanning Electron Microscopy (SEM)

During the on-surface preparation, a significant amount of grain structures grew from an initially smooth polished Cu polycrystal surface (**Figure 5**). Additionally, the structures of **Cu\_3E**+**3A**, **Cu\_1E**+**1A**, and **Cu\_1E** were similar. Anodic scans of the copper surface result in dissolution and re-deposition of Cu<sup>+</sup> in acetonitrile in the presence of phenylacetylene, and insoluble copper phenylacetylide is introduced as a stable intermediate (Ahrland, 1982). The grain structures seemed to grow through an oxidation–coordination–reduction process in repetitive anodization cycles, similar to the preparation of copper nanostructures derived from copper oxide CO2 reduction catalysts. When the number of scan cycles was reduced to seven, the surface was covered by a smaller structure in the early stage of growth (**Figure S2**). XPS showed that **Cu\_3E**+**3A** contained a higher amount of organic groups on its surface than **Cu\_1E**+**1A**. However, these structures had very similar appearances, implying that the organic moieties were uniformly dispersed on copper grains. The thickness of the rough surface structure is estimated to have been hundreds of nanometers from the cross-sectional SEM image (**Figure S3**). When molecular triazole was cast on the copper surface (**Figure 5E**), platelet crystals were formed on the surface.

Additionally, the basal copper itself was observed to be slightly roughened by casting (**Figure S4**), as discussed in the roughness factor section.

## Roughness Factor

The roughness of the electrodes was calculated from the doublelayer capacitance (**Figure S5**, **Table S1**). **Table 1** details the roughness factors of the Cu electrodes, with the unmodified electrode as the reference. **Cu\_1E**+**1A** and **Cu\_1E** exhibited a high degree of roughness, which is consistent with the apparent surface topology observed in the SEM micrographs. Despite the apparent granular morphology, the roughness of **Cu\_3E**+**3A** was similar to that of the smooth unmodified electrode. The surface of the Cu electrode may have been partially insulated by the presence of a thick organic layer film. When the triazole molecule was cast onto the surface, the surface area slightly increased, although no anodization of the copper was achieved. The adsorption of densely packed triazole, and the subsequent cathodization, may result in the re-construction of the surface copper (Gunathunge et al., 2017), as observed in the SEM micrographs.

# Underpotential Deposition (UPD)

The amount of exposed Cu atoms on the electrode surface was estimated by UPD analysis (**Figure 6**). There were fewer exposed Cu atoms in the **Cu\_3E**+**3A** electrode than estimated on the basis of the SEM micrograph and roughness factor. The surface of **Cu\_3E**+**3A** is considered to have been largely covered by the organic layer, which inhibited the approach and deposition of Pb2+; however, the system maintained its function as an electrode, as the layer could almost be considered as a monolayer in terms of thickness (Bandyopadhyay et al., 1998; Feng et al., 2017). The size of the redox wave was significantly larger in **Cu\_1E**+**1A**, as the monomeric triazole was not able to strongly adsorb onto the surface. Furthermore, the majority of the grain structure was composed of copper. In **Cu\_1E**+**1A\_cast,** the organic moieties seemed to be aggregated to form massive crystals that inhibited the Pb deposition process on the electrode surface.

# CO<sup>2</sup> REDUCTION

The formation of carbon products by CO<sup>2</sup> reduction at −1.4 V (vs. reversible hydrogen electrode, RHE, solution resistance uncorrected, **Figure 7**) showed that **Cu\_3E**+**3A** had the highest amount of carbonaceous product formation. The products demonstrate two remarkable effects of modification: (i) ethylene formation was enhanced by anodic scanning in the presence of the ethynyl precursors; (ii) only the **Cu\_3E**+**3A** electrode exhibited improved methane formation as a result of modification, while the other electrodes appeared to compensate for ethylene formation with methane formation.

Comparison of the Faradaic efficiency (**Figure 8**) reveals that the high CO<sup>2</sup> reduction performance of **Cu\_3E**+**3A** did not result simply from an increase in total current with increasing roughness of the electrode surface. Besides the increase in methane and ethylene formation, **Cu\_3E**+**3A** exhibited remarkably low hydrogen production, which is undesirable for the proton-consuming side reaction. Also, the control experiment under argon only bore hydrogen from **Cu\_3E**+**3A** at the potential range, showing that the modified organics were not the source of hydrocarbons (**Figure S6**).

The linear sweep voltammetry (LSV) curves (**Figure 9**) of the electrodes provide additional information on reaction selectivity. At low current densities (−0.4 to −0.8 V), the current ratio appears to largely depend on the roughness factor. Further reducing the negative potential to under −1.0 V resulted in similar current densities for all electrodes (ca. −2 mAcm−<sup>2</sup> ),

despite observed differences in the roughness factor. At high current densities, the thickness of the CO<sup>2</sup> and proton diffusion layer increased beyond the scale of the thickness of the deposited surface structures. Therefore, the reaction is limited by the transportation properties. The polarization involves multiple electrochemical processes. By incrementally increasing the potential, the onset of CO<sup>2</sup> reduction to CO and hydrogen evolution was observed to first emerge at approx. −0.4 V, and the reduction of adsorbed CO into hydrocarbons was further enhanced when the potential was under −1.0 V. The current in **Cu\_3E**+**3A** displayed slow increments despite of a relatively positive current onset, indicating that significant CO adsorption results in a current drop until the potential reaches −1.0 V, as CO<sup>2</sup> reduction is significantly more efficient than H<sup>2</sup> evolution. Thereafter, the current rapidly increased below −1.0 V as the adsorbed CO was removed through further TABLE 1 | Roughness factors of copper electrodes.


*<sup>a</sup>Electrolysis was carried out for 50 min from* −*1.0 to* −*1.4 V vs. RHE with a stepwise increment of* −*0.1 V in every 10 min.*

reduction into hydrocarbons followed by dissociation, consistent with the potential dependence of Faradaic efficiency for H<sup>2</sup> and CO (**Figure 8**).

The effect on CO<sup>2</sup> reduction is influenced by how the organic moieties contact with the copper surface. In repeated cast modifications of the polished Cu electrode, the total current density was not observed to decrease; however, the formation of carbonaceous products decreased depending on the number of times the copper electrode was subjected to the cast process and was even surpassed by hydrogen production (**Figure 10**). The casting of molecular triazole typically forms densely packed crystalline structures that inhibit the adsorption of CO<sup>2</sup> on the surface. Conversely, this is not the case for H<sup>2</sup> production, as H<sup>2</sup> evolution requires significantly smaller surface vacancies to adsorb H atoms, and it may even be possible on the adsorbed organic molecules.

An ethylene increase was observed for the electrodes that were subjected to repetitive anodic scanning in the presence of the ethynyl precursor. The existence of the insoluble copper(I) acetylide intermediate on the surface is important for copper reorganization. This observed trend is similar to the high ethylene formation efficiency observed for nanostructured CO2 reduction catalysts that are prepared by the formation of an insoluble copper oxide layer on the surface followed by reduction to Cu(0) (Kas et al., 2014; Ren et al., 2015; Dutta

et al., 2016; Handoko et al., 2016; Mistry et al., 2016; Mandal et al., 2018). During anodization in the absence of the ethynyl precursor ("scanned" in **Figure 7**), CO<sup>2</sup> reduction resulted in a decrease in the efficiency of ethylene formation. Continuous dissolution of copper into acetonitrile, without the precursor, may remove the copper atoms at the defect site of the copper crystal facet.

In the majority of Cu electrocatalytic processes, there is a trade-off in the efficiency of methane and ethylene selectivity. Additionally, the pH of the electrolyte influences the ratio of methane and ethylene, as the reduction process is related to competition of the dimerization of the adsorbed CO and the proton-coupled reduction (Schouten et al., 2011, 2013; Huang et al., 2017). The potential dependence of the ethylene/methane ratio also shows that methane formation is dominant with increased current density, as the reduction rate surpasses CO dimerization. Further increasing the current density results in an increase in pH on the surface due to H<sup>+</sup> consumption, resulting in dominant H<sup>2</sup> production through H2O reduction because CO<sup>2</sup> is not stable in alkaline media (Singh et al., 2016). This trend appears to hold true in the case of **Cu\_1E**+**1A** and **Cu\_1E**, as the total CO<sup>2</sup> reduction efficiency remained at a similar level to that of the unmodified electrode. In **Cu\_3E**+**3A**, both methane and ethylene increased with remarkably low hydrogen production. Hence, the organic layers play a role in increasing the total CO<sup>2</sup> reduction efficiency.

Stirring the electrolyte solution also changes the transport conditions. When stirring of the solution was stopped (**Figure 11**), the amount of dissolved components in the electrolyte (most likely formate) decreased for both the unmodified and **Cu\_3E**+**3A** electrodes. Hydrogen was observed to increase for the unmodified electrode, but with the **Cu\_3E**+**3A** electrode, the decrease of the soluble product content appeared to be compensated for by hydrocarbon formation, especially methane. Under a high current-flow condition without stirring, a pH increase at the electrode surface was observed to occur, as every electron transfer was accompanied by a proton transfer. Furthermore, the depletion of CO<sup>2</sup> was more severe for two reasons: (i) CO<sup>2</sup> possesses a lower diffusion coefficient than <sup>H</sup>+, and (ii) CO<sup>2</sup> transforms to HCO<sup>−</sup> 3 in basic media. The suppression of hydrogen evolution by modification of the electrode appears to result from the preference of CO<sup>2</sup> reduction and adsorption over the reduction of H2O. CO<sup>2</sup> selectivity is affected by the competition between CO<sup>2</sup> reduction and proton reduction forming surface-adsorbed H, and the latter process tends to lead to the formation of H<sup>2</sup> and formate. Several recent reports have demonstrated that hydrophobic modification of the surface improves the supply of CO<sup>2</sup> relative to H+, which leads to better selectivity to CO and hydrocarbons (Buckley et al., 2019; Li et al., 2019). Moreover, the three-way triazole moiety was reported to have specific affinity in MOF CO<sup>2</sup> storage (Wang et al., 2013). In our case, the introduction of the organic layer made the surface more hydrophobic (**Figure S7**). The presence of the organic layer seems to create a CO2-rich environment on the electrode surface, possibly by improved proton transfer and better CO<sup>2</sup> affinity with the layer than with the bare electrode surface.

# CONCLUSIONS

A novel method has been designed to modify the surface of a metal electrode to enhance surface catalytic activity by anodization. The metal-organic contact structure formed offers unique properties over those of traditional cathodes prepared via the cast method by inducing a thin layer of organic moieties that are accessible to surface copper atoms. The influence of the modified electrode on CO<sup>2</sup> reduction reveals that the contact surface has a preference toward CO<sup>2</sup> reduction over H<sup>2</sup> evolution, contrary to the observed results from surfaces modified by the cast method. The organic layer appears to enhance the environment to achieve improved CO<sup>2</sup> affinity and proton transport ability at the surfaces of the copper atoms. The re-organization of copper atoms upon modification did influence CO<sup>2</sup> reduction selectivity, as in copper oxide-derived electrocatalysis. However, this property is not the sole factor governing the observed CO<sup>2</sup> reduction selectivity, because a large total CO<sup>2</sup> reduction efficiency and high methane formation are observed, even at low potentials. The method herein

can readily introduce various types of organic substituents to the organic-open metal contact surface. Studies are underway to investigate controlling the CO<sup>2</sup> reduction based on the discussion in this report. Furthermore, the introduction of an organic functional center, such as ligands and dyes, would allow for CO<sup>2</sup> reduction with cooperated activation or lightdriven functionality.

### METHODS

### Electrode Modification

A disc copper electrode (3 mmϕ) was made from the cross-section of a polycrystalline rod embedded in a glass

from cast-modified electrodes with the number cast processes. Error bars are placed at the top of each component.

tube and polished with Bicalox0.05CR (0.05µm alumina abrasive). Modification was performed in 5 mL electrolyte (0.1 M <sup>n</sup>Bu4NPF6-acetonitrile) containing ethynyl (6.6 mM) and azide (6.6 mM) precursors under an argon atmosphere. A potential

scan was repeated 15 times (from −0.2 to −1.0 V vs. Ag/Ag+, 0.1 Vs−<sup>1</sup> ) using an ALS 650D potentiostat, followed by washing with acetonitrile. For cast modification, 0.7 µL of acetonitrile solution containing 1-benzyl-4-phenyl-1,2,3-triazole (4.3 mM) was dropped on the copper disc and air-dried. Modified copper substrates for XPS and SEM measurements were prepared using 5 mm square copper foil instead of copper discs.

### Characterization of Surface Structure

SEM images were collected with a Hitachi FE-SEM S-5200. XPS data were collected using a KRATOS ESCA 3400 with an Al-Kα source. Double-layer capacitance was measured in CO<sup>2</sup> saturated 0.1M KHCO3aq on the basis of the current dependence upon the potential scan rate in the region of −0.6 V to −0.7 V vs. Ag/AgCl. UPD was achieved by immersing the modified electrode in electrolyte-containing lead salt (0.5 mM PbCl2, 0.5 mM HCl/10 mM HClO4aq, scan rate; 10 mVs−<sup>1</sup> ).

# CO<sup>2</sup> Reduction

The CO<sup>2</sup> reduction activity on copper electrodes was investigated in CO2-saturated 0.1 M KHCO3aq (pH 6.8). The three-electrode

# REFERENCES

Abrantes, L., Fleischmann, M., Hill, I., Peter, L., Mengoli, M., and Zotti, G. (1984). An investigation of Cu(I) acetylide films on Cu electrodes: Part I. Electrochemical and Raman spectroscopic setup was connected to a potentiostat (ALS 650D). Ag/AgCl was used as a reference electrode, and Pt mesh was used as the counter electrode. The applied potential was converted to RHE according to the equation ERHE = EAg/AgCl + 0.197 V + 0.0591 × pH.

The experiments were performed in a two-compartment cell. A copper working electrode and Ag/AgCl electrode were placed in the gas-tight cathodic compartment. It was separated from the open anodic compartment, which was equipped with a Pt mesh counter electrode, by a 10 mmϕ ion-exchange membrane (SelemionTM AMV). Electrolysis was performed for 10 min with stirring (ca. 600 rpm), without additional supply of CO2. Depletion of CO<sup>2</sup> by electrolysis did not exceed a few % of the amount of CO<sup>2</sup> in solution (34 mM, 6.2 mL), which is estimated from the charge flow (typically 1 C). For the quantification of gaseous products, 0.1 mL of gas product was collected from the head space (9.9 mL) of the gas-tight compartment and introduced in a gas chromatograph (Shimazu-2010) equipped with a 2.0 m × 1.0 mm ID column packed with SHINCARBON ST and a BID-2010 detector. Copper electrodes were first cathodized at −1.6 V vs. Ag/AgCl for 1 h to remove weakly physisorbed materials and reduce oxidized copper species prior to the collection of CO2RR products. The current density was calculated according to the geometric area of the Cu electrode (3 mm ϕ). CO<sup>2</sup> reduction data are provided as averages from at least three separately prepared electrodes.

# AUTHOR CONTRIBUTIONS

SK, KK, and TM contributed the overall concept of the study. SK and RT developed the experimental apparatus for electrolysis. RT and RI performed the experiments and analysis of the collected data. All authors contributed to manuscript revision and read and approved the submitted version.

# FUNDING

This work was supported by JSPS KAKENHI Grant Number: JP16K05723.

# ACKNOWLEDGMENTS

The authors acknowledge Dr. M. Maeda for support in SEM image collection and Dr. H. Dote for assistance in XPS measurements. We thank the Edanz Group (www.edanzediting. com/ac) for editing a draft of this manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00860/full#supplementary-material

analysis of the film formation. J. Electroanal. Chem. Interfacial Electrochem. 164, 177–187. doi: 10.1016/S0022-0728(84)8 0238-7

Ahn, S., Klyukin, K., Wakeham, R. J., Rudd, J. A., Lewis, A. R., Alexander, S., et al. (2018). Poly-amide modified copper foam electrodes for enhanced electrochemical reduction of carbon dioxide. ACS Catal. 8, 4132–4142. doi: 10.1021/acscatal.7b04347


coverage on the selective formation of ethylene. ACS Catal. 7, 1749–1756. doi: 10.1021/acscatal.6b03147


electrochemical reduction of carbon dioxide on copper electrodes. Chem. Sci. 2, 1902–1909. doi: 10.1039/c1sc00277e


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Igarashi, Takeuchi, Kubo, Mizuta and Kume. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.