# ENVIRONMENTAL CATALYSIS AND THE CORRESPONDING CATALYTIC MECHANISM

EDITED BY : Zhimin Ao, Hongqi Sun and Andres Fullana PUBLISHED IN : Frontiers in Chemistry

### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-828-8 DOI 10.3389/978-2-88945-828-8

### 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

# ENVIRONMENTAL CATALYSIS AND THE CORRESPONDING CATALYTIC MECHANISM

Topic Editors: Zhimin Ao, Guangdong University of Technology, China Hongqi Sun, Edith Cowan University, Australia Andres Fullana, University of Alicante, Spain

Citation: Ao, Z., Sun, H., Fullana, A., eds. (2019). Environmental Catalysis and the Corresponding Catalytic Mechanism. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-828-8

# Table of Contents

*05 Editorial: Environmental Catalysis and the Corresponding Catalytic Mechanism*

Zhimin Ao, Hongqi Sun and Andres Fullana

# SECTION 1

# PHOTOCATALYSIS

*07 The Promoting Role of Different Carbon Allotropes Cocatalysts for Semiconductors in Photocatalytic Energy Generation and Pollutants Degradation*

Weiwei Han, Zhen Li, Yang Li, Xiaobin Fan, Fengbao Zhang, Guoliang Zhang and Wenchao Peng

*23 Facile Fabrication of BiOI/BiOCl Immobilized Films With Improved Visible Light Photocatalytic Performance*

Yingxian Zhong, Yuehua Liu, Shuang Wu, Yi Zhu, Hongbin Chen, Xiang Yu and Yuanming Zhang


Weitian Lin, Xiang Yu, Yi Zhu and Yuanming Zhang

*56 Adsorption and Photocatalytic Processes of Mesoporous SiO2-Coated Monoclinic BiVO4*

Duangdao Channei, Auppatham Nakaruk, Wilawan Khanitchaidecha, Panatda Jannoey and Sukon Phanichphant

*63 Catalytic and Photocatalytic Nitrate Reduction Over Pd-Cu Loaded Over Hybrid Materials of Multi-Walled Carbon Nanotubes and TiO2* Cláudia G. Silva, Manuel F. R. Pereira, José J. M. Órfão, Joaquim L. Faria and Olívia S. G. P. Soares

# SECTION 2

# CATALYTIC CONVERSION OF BIOMASS AND PPCPs


# SECTION 3

# COMPUTATIONAL CATALYSIS

*93 First Principles Study on the CO Oxidation on Mn-Embedded Divacancy Graphene*

Quanguo Jiang, Jianfeng Zhang, Zhimin Ao, Huajie Huang, Haiyan He and Yuping Wu

*104 The Activity Improvement of the TM3(hexaiminotriphenylene)2 Monolayer for Oxygen Reduction Electrocatalysis: A Density Functional Theory Study* Beibei Xiao, Hui Zhu, HouYi Liu, XiaoBao Jiang and Qing Jiang

# Editorial: Environmental Catalysis and the Corresponding Catalytic Mechanism

### Zhimin Ao<sup>1</sup> \*, Hongqi Sun<sup>2</sup> and Andres Fullana<sup>3</sup>

*<sup>1</sup> Guangzhou Key Laboratory of Environmental Catalysis and Pollution Control, Guangdong Key Laboratory of Environmental Catalysis and Health Risk Control, School of Environmental Science and Engineering, Institute of Environmental Health and Pollution Control, Guangdong University of Technology, Guangzhou, China, <sup>2</sup> School of Engineering, Edith Cowan University, Joondalup, WA, Australia, <sup>3</sup> Department of Chemical Engineering, University of Alicante, Alicante, Spain*

Keywords: environment remediation, catalysts, clean energy, photocatalysis, biomass conversion, density functional theory, catalytic mechanism

**Editorial on the Research Topic**

### **Environmental Catalysis and the Corresponding Catalytic Mechanism**

The ever growing environmental pollution has stimulated the rapid development of environmental catalysis in recent years. Environmental catalysis is a multidisciplinary research field for which more and more chemists, materials scientists, as well as environmentalists have devoted their efforts working in this field because of the bright potentials in improving human health and life quality. With the progresses in controllable materials synthesis, advanced characterizations (electron microscopy, spectroscopy, etc.), high-level analytical chemistry, together with the computational studies, catalysis continues to be the driving force for generation of clean energy, abatement of major pollutants in air and water, and meantime the theories behind the catalytic reactions are illustrated. In the current Research Topic, an elegant collection of original research and review articles reporting the synthesis of high-performance catalysts, their applications in various catalytic technologies for environment remediation, and relevant theoretical calculations for understanding the catalytic mechanisms is presented.

Photocatalysis remains to be a research focus for the environmental catalysis community due to the wide applications in carbon dioxide reduction, oxidation of volatile organic compounds (VOCs), elimination of aqueous organic pollutants and disinfection, water splitting (Chen et al., 2011; Zheng et al., 2016; Wang et al., 2017). Particularly, the synthesis of visible light and near-infrared (NIR) light responsive semiconductor photocatalysts is of immense interests to scientists because visible light and NIR light occupy around 90% of the solar light energy, in comparison to the no more than 5% for ultraviolet light (UV). To enhance the visible light absorption of the most popularly used TiO2, several routes including creating defects and element doping are developed Qin and Lin et al., introduced carbon and nitrogen elementals into TiO<sup>2</sup> simultaneously and demonstrated the improved catalytic property of such C, N-TiO<sup>2</sup> catalyst compared to anatase TiO<sup>2</sup> under simulated sunlight irradiation for degradation of 4-nitrophenol. Moreover, they also evaluated the embryonic toxicity of intermediate degradation compounds (Osin et al.). Silva et al., synthesized Pd-Cu loaded over hybrid materials of carbon nanotubes and TiO<sup>2</sup> for nitrate reduction. It reported that the Pd-Cu loaded on the hybrid materials have high photocatalytic performance for NO<sup>−</sup> 3 conversion (Silva et al.). Other than TiO2, the bismuth compounds as a new class of photocatalytic materials have been paid much attention in recently. Among them, bismuth oxychloxide (BiOCl) exhibits excellent photocatalysis behaviors driven by UV light. To expand its light absorption to visible light spectrum, efforts to fabricate graphene oxide/BiOCl nanocomposite

Edited and reviewed by: *Pascal Granger, Lille University of Science and Technology, France*

> \*Correspondence: *Zhimin Ao zhimin.ao@gdut.edu.cn*

### Specialty section:

*This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry*

Received: *17 November 2018* Accepted: *28 January 2019* Published: *13 February 2019*

### Citation:

*Ao Z, Sun H and Fullana A (2019) Editorial: Environmental Catalysis and the Corresponding Catalytic Mechanism. Front. Chem. 7:75. doi: 10.3389/fchem.2019.00075* film and BiOI/BiOCl film are reported by Zhu and Zhang et al., respectively (Lin et al.; Zhong et al.). BiVO<sup>4</sup> is found to be a visible light-activated photocatalyst due to the narrower band gap than TiO2, the involvement of graphene oxide or BiOI can further improve the photocatalytic performance for RhB degradation. However, the poor specific surface area limits its catalytic performance. To this end, Channei et al. utilized a co-precipitation method to coat SiO<sup>2</sup> onto BiVO<sup>4</sup> and obtained SiO2/BiVO<sup>4</sup> composites with larger surface area and higher photocatalytic activity and degradation efficiency towards methylene blue dye with respect to monoclinic BiVO<sup>4</sup> (Channei et al.). Furthermore, a comprehensive overview on semiconductors loaded with carbon cocatalysts as photocatalysts for water splitting and pollutant degradation was provided by Peng et al., aiming to give a clue to the rational design of low-cost photocatalysts with more efficient solar light utilization. The synthesis methods of various types of carbonsemiconductor composite photocatalysts were summarized, the contribution of different carbon allotropes like C60, carbon nanotubes, graphene to the enhanced photocatalytic activity were compared and the cocatalytic effect mechanisms were discussed (Han et al.).

Catalytic conversion of biomass to biofuels is another emerging topic in the environmental catalysis field out of the urge to transform the agricultural wastes into resources and to reduce the carbon dioxide emission from fossil fuels combustion (Wang and Xiao, 2015; Xiong et al., 2015). Heterogeneous acidic catalysts and enzyme biocatalysts are among the leading candidates for the conversion of lignocellulosic biomass to fuels and value-added chemicals. One relevant paper presented the good catalytic performance of bifunctional catalysts which were synthesized by decorating propyl/phenyl-sulfonic acid group functionalized mesoporous silica materials SBA-15 with Pt particles in the reaction of hydrodeoxygenation of bio-derived phenol to produce cyclohexane fuel (Mo et al.).

Pharmaceutical and personal care products (PPCPs) are emerging contaminants, which are widely present in

# REFERENCES


pharmaceutical and hospital wastewater, even natural water. In this Research Topic, Fe-MCM-41s were fabricated at different conditions to adsorb widely used antibiotics ciprofloxacin hydrochloride (CPX) for its removal from waste water (Wu et al.). Owing to the rise of supercomputers, computational studies become important complementary tools for elucidation of the inherent mechanisms of catalytic reactions. Currently Ab initio techniques like Density Functional Theory (DFT) are popularly adopted for theoretical calculations. Here in this Research Topic, two examples of DFT studies on the mechanisms of CO oxidation catalyzed by Mn-embedded divacancy graphene and boosted oxygen reduction reaction performance catalyzed by two-dimensional metal–organic frameworks TM3(hexaiminotriphenylene)<sup>2</sup> monolayer (where TM = Ni, Co, Fe, Pd, Rh, Ru, Pt, Ir, and Os) were shown (Jiang et al.; Xiao et al.).

We believe that this collection, Environmental Catalysis and the Corresponding Catalytic Mechanism, illustrates the advancements of catalysis for environment remediation and innovations in clean energy with diminished production of undesired by-products. At the same time, challenges and perspectives for this field are also addressed with the hope that future interests will be focused to help establish a world with clean air and water, as well as sustainable and green energy for human to live in.

# AUTHOR CONTRIBUTIONS

ZA drafted the manuscript. HS and AF made substantial improvements, and all authors approved for publication.

## ACKNOWLEDGMENTS

ZA was supported by National Natural Science Foundation of China (21777033 and 21607029), Science and Technology Program of Guangdong Province (2017B020216003), and Science and Technology Program of Guangzhou City (201707010359).

Zheng, Q., Durkin, D. P., Elenewski, J. E., Sun, Y., Banek, N. A., Hua, L., et al. (2016). Visible-light-responsive graphitic carbon nitride: rational design and photocatalytic applications for water treatment. Environ. Sci. Technol. 50, 12938–12948. doi: 10.1021/acs.est.6b02579

**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 Ao, Sun and Fullana. 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.

# The Promoting Role of Different Carbon Allotropes Cocatalysts for Semiconductors in Photocatalytic Energy Generation and Pollutants Degradation

Weiwei Han, Zhen Li, Yang Li, Xiaobin Fan, Fengbao Zhang\*, Guoliang Zhang and Wenchao Peng\*

*School of Chemical Engineering and Technology, Tianjin University, Tianjin, China*

### Edited by:

*Andres Fullana, University of Alicante, Spain*

### Reviewed by:

*Xiaoguang Duan, Curtin University, Australia Jiguang Deng, Beijing University of Technology, China*

\*Correspondence:

*Fengbao Zhang fbzhang@tju.edu.cn Wenchao Peng wenchao.peng@tju.edu.cn*

### Specialty section:

*This article was submitted to Green and Environmental Chemistry, a section of the journal Frontiers in Chemistry*

> Received: *06 September 2017* Accepted: *06 October 2017* Published: *31 October 2017*

### Citation:

*Han W, Li Z, Li Y, Fan X, Zhang F, Zhang G and Peng W (2017) The Promoting Role of Different Carbon Allotropes Cocatalysts for Semiconductors in Photocatalytic Energy Generation and Pollutants Degradation. Front. Chem. 5:84. doi: 10.3389/fchem.2017.00084* Semiconductor based photocatalytic process is of great potential for solving the fossil fuels depletion and environmental pollution. Loading cocatalysts for the modification of semiconductors could increase the separation efficiency of the photogenerated hole-electron pairs, enhance the light absorption ability of semiconductors, and thus obtain new composite photocatalysts with high activities. Kinds of carbon allotropes, such as activated carbon, carbon nanotubes, graphene, and carbon quantum dots have been used as effective cocatalysts to enhance the photocatalytic activities of semiconductors, making them widely used for photocatalytic energy generation, and pollutants degradation. This review focuses on the loading of different carbon allotropes as cocatalysts in photocatalysis, and summarizes the recent progress of carbon materials based photocatalysts, including their synthesis methods, the typical applications, and the activity enhancement mechanism. Moreover, the cocatalytic effect among these carbon cocatalysts is also compared for different applications. We believe that our work can provide enriched information to harvest the excellent special properties of carbon materials as a platform to develop more efficient photocatalysts for solar energy utilization.

Keywords: carbon allotropes, semiconductor, photocatalysis, cocatalysts, energy generation, pollutants degradation

# INTRODUCTION

Environmental pollution and fossil fuels depletion are the most serious social problems nowadays. Since the discovery of the photocatalytic splitting of water on TiO<sup>2</sup> electrodes by Fujishima and Honda in 1972, photocatalysis technology has become one of the most promising technologies for energy generation and environment remediation (Fujishima and Honda, 1972). Moreover, solar energy is clean, sustainable, and inexhaustible, which is therefore the most hopeful resource to solve the energy and environment problems (Chen et al., 2010a). Mostly, photocatalysis is a semiconductor-mediated process (Chen et al., 2010b; Wang et al., 2014; Zhang et al., 2016b). So far, kinds of semiconductor materials, including metal oxides, metal sulfides and metal containing salts have been used as photocatalysts. Some metal free materials, such as silicon, sulfur, graphic carbon nitride (g–C3N4), have also been developed as photocatalysts for the utilization of sunlight (Peng et al., 2013; Cao and Yu, 2014; Devi and ArunaKumari, 2014; He et al., 2015). However, some fundamental problems must be resolved before their real application, which are (1) low utilization efficiency of solar energy; (2) poor quantum efficiency; (3) severe photo corrosion (Zhang and Guo, 2013; Chowdhury and Balasubramanian, 2014; Han et al., 2015; Xie et al., 2015; Liu Y. et al., 2017b). To address these obstacles, modification of semiconductors with suitable cocatalysts is a frequent and effective solution (Yang J. H. et al., 2013). Metal nanoparticles and their compounds, especially noble metal based materials, are always used as cocatalysts (Bai et al., 2014; Zhang et al., 2015a; Zhong et al., 2016). Although they are effective to enhance the photocatalytic activity, the high cost and rare storage on earth limit their practical application (Ran et al., 2014). To develop cheap, highly efficient alternatives to replace noble metal based cocatalysts is still a great challenge in the photocatalysis filed.

Recently, carbon materials, including activated carbon (AC), fullerenes (C60), carbon nanotubes (CNTs), graphene (GR), and other carbon allotropes, have been widely investigated as cocatalysts for semiconductors in photocatalysis (Xiang et al., 2012; Ouzzine et al., 2014; Cao and Yu, 2016; Paulo et al., 2016; Yu et al., 2016). Specially, CNTs and GR have large specific surface areas (SSAs), excellent electric conductivity, high mechanical strength, and good thermal, and chemical stability, making them ideal substitute for noble metal cocatalysts (Zhang et al., 2012; Wang et al., 2013, 2017; Di et al., 2015; Han et al., 2016). **Figure 1** shows the structure models of the carbon materials and their photocatalytic applications as cocatalysts simply. Many kinds of carbon cocatalysts based composites have been reported for photocatalytic reactions, and the cocatalytic mechanisms have also been discussed (Woan et al., 2009; Chen et al., 2011; Lee W. J. et al., 2012; Xie et al., 2013; Shearer et al., 2014; Li et al., 2015). For example, a graphene–TiO<sup>2</sup> NPs hybrid was successfully synthesized by wrapping amorphous TiO<sup>2</sup> NPs with GO using a one-step hydrothermal method by Lee and coworkers (Lee J. S. et al., 2012). The hybrid exhibited superior photocatalytic activity for the photodegradation of MB under the visible light irradiation. Fan et al. prepared a novel 3D AgX/graphene aerogels (X = Br, Cl) structured composite, which exhibited excellent photocatalytic and cycling performance for the degradation of MO and reduction of CrVI (Fan Y. et al., 2015). They also investigated photocatalytic enhancement mechanism of the graphene aerogels in the composite, which could effectively suppress the recombination of photogenerated holes, and electrons as a capable substrate for the photocatalyst. Tian et al. reported a new CQDs/hydrogenated TiO<sup>2</sup> (H-TiO2) photocatalyst by assembling CQDs on the surface of H-TiO<sup>2</sup> (Tian et al., 2015). The photocatalytic activity of CQDs/H-TiO<sup>2</sup> was superior to P25, TiO<sup>2</sup> nanobelts, and H-TiO<sup>2</sup> nanobelts for the degradation of MO under UV-visible-NIR irradiation. The CQDs have excellent photo-induced electron transfer and reservoir properties, which could convert NIR light to visible light to be in full used by H-TiO<sup>2</sup> and effectively suppress the recombination of electron-hole pairs. Generally, loading carbon materials as cocatalysts for semiconductors, the

FIGURE 1 | Schematic illustration of the photocatalytic applications of carbon materials based semiconductor composites.

synergistic effect between them can increase the active sites, widen the absorption range of the solar light, facilitate the separation of the electron-hole pairs, and thus enhancing the photocatalytic activity.

To develop carbon materials based composite has attracted great attention for low cost and highly active photocatalysts. Lots of researches have been done on this subject, but a systematic summary about the key roles of different carbon allotropes as cocatalysts is still lacking. Herein, we aim to provide an overview on recent advances in the synthesis, multiple applications and mechanism of different carbon allotropes based composite photocatalysts. On behalf of this review, we wish more carbon based photocatalysts could be synthesized for environment remediation and energy generation.

# PHOTOCATALYSTS SYNTHESIS

The synthesis process will affect the morphologies, properties and activities of the composite photocatalyts greatly. As shown in **Table 1**, we summarized the typical synthesis methods of the recently reported carbon based photocatalysts. The semiconductors could be loaded on carbon materials by one-step grinding, stirring, ultrasonic assisted dispersing or by some complicated multi-step synthesis methods. It can be concluded that mechanical mixing, hydrothermal/solvothemal, and sol-gel process are more frequently used. In addition, photocatalytic reduction and microwave-assisted method are also reported, and they may have a great potential due to the green and sustainable synthetic processes.


*(Continued)*


TABLE 1 |

Continued

*(Continued)*


TABLE 1 |

Continued

# Hydrothermal/Solvothemal Methods

Hydrothermal or solvothermal methods are the most frequently used ways due to their mild reaction conditions, high product purity, controllable morphology, good crystallinity, and uniform distribution of obtained products (Li Q. et al., 2011). For example, Liu et al. synthesized GR–CdS nanocomposites by an one-step solvothermal method using DMSO as reductant and sulfure source (Liu et al., 2014). In the preparation procedure, GO was dispersed in DMSO to obtain the GO–DMSO dispersion, Cd(CH3COO)2·2H2O was then added. The mixture was then treated at 453 K for 12 h to obtain the final composites. The photocatalytic activity of GR–CdS nanocomposites for selective reduction of aromatic nitro compounds was dramatically enhanced compared to the pure CdS. This can be ascribed to the synergistic effect with graphene addition, the increased visible light absorption range and intensity, the improved lifetime and charge transfer ability, and the enhanced adsorption capacity of this nanocomposite toward the nitro compounds.

Han et al. synthesized 2D hexagonal α-Fe2O3/graphene nanoplate composites by a simple one-step hydrothermal method with no template (Han et al., 2014). Using hydrothermal method, not only the effective reduction of the GO to graphene was achieved, but intimate contact was also formed between the α-Fe2O<sup>3</sup> nanoplates and graphene. A significant enhancement for photocatalytic degradation of RhB could be observed after the combination with graphene cocatalyst. An et al. fabricated MWCNT–TiO<sup>2</sup> sphere composites by a facile one-step hydrothermal method using TiF<sup>4</sup> as titanium source and CNTs as structure regulator (An et al., 2012). The effects of hydrothermal temperature and hydrothermal time on the structural characteristics of MWCNT–TiO<sup>2</sup> photocatalysts were investigated. Decreasing hydrothermal temperature or prolonging the hydrothermal time could lead to the enhancement of the photocatalytic degradation efficiency of both gaseous (i.e., styrene) and aqueous (i.e., MO) phase. Decreasing the hydrothermal temperature could lead to the crystallite size decrease of TiO<sup>2</sup> (**Table 2**), while prolonging the hydrothermal time will increase the synergistic effects between TiO<sup>2</sup> and MWCNTs, thus promoting the photocatalytic performance.

As reported by Murcia-López et al. the calcination could be applied after hydrothermal process to prepare the AC/Bi2WO<sup>6</sup> and AC/TiO2/Bi2WO<sup>6</sup> photocatalysts (Murcia-Lopez et al., 2013). The introduction of optimized percentage of AC (2 wt%) could form 3D-hierarchical structures of both AC/Bi2WO<sup>6</sup> and AC/TiO2/Bi2WO6, which exhibited improved photocatalytic activities for the RhB degradation under both UV-vis and visible illumination compared to pure Bi2WO6. Here, the presence of AC could stimulate the 3D-hierarchical structure formation, and will increase the surface area and absorption ability of the catalyst at the same time.

## Mechanical Mixing Method

The loading of carbon cocatalysts can also be performed by simple mechanical mixing processes, such as magnetic stirring, ball milling, and ultrasonication (Xu et al., 2014; Guo et al., 2017). Ali et al. used C<sup>60</sup> as cocatalysts for the modification of TiO<sup>2</sup> using a simple sonication assisted mixing method (Mukthar Ali and Sandhya, 2014). The C<sup>60</sup>

TABLE 2 | Crystallite size of TiO2 in Pure TiO2 and MWCNT–TiO2 photocatalysts.


*Reprinted from An et al. (2012), Copyright 2012, with permission from American Chemical Society.*

molecules were first dispersed in β-cyclodextrin (CD), and then mixed with the suspension of TiO<sup>2</sup> with the assistance of sonication under sunlight. According to the HRTEM images of the composites, C<sup>60</sup> cocatalysts are dispersed in the composite without aggregation. They believed that the nonaggregated C<sup>60</sup> cocatylysts played a key role in increasing the amount of reactive oxygen species (ROS) and suppressing photogenerated charge recombination, thus leading to the enhanced photocatalytic activity. The photocatalytic activity of the composite shows 2 and 5 times higher than the bare TiO<sup>2</sup> for the degradation of MB and 4-CP, respectively. Gao and his coworkers successfully synthesized GO–CdS composites by a novel two-phase mixing method (Gao et al., 2013). By simply stirring for 24 h, the two phases are mixed into a homogeneous solution, and CdS nanoparticles are then uniformly deposited on GO sheets (see Scheme 1 in the original paper, Gao et al., 2013). The obtained composites show higher photocatalytic degradation and disinfection activities than CdS under visible light irradiation.

However, using the mechanical mixing method, the interaction force between semiconductors and carbon materials is a little weak without the formation of chemical bonds, resulting in a relatively lower activity enhancement compared to that from hydrothermal/solvothemal methods. For example, Hong and his coworkers reported that CdS/GO photocatalysts synthesized by in situ solvothermal method showed much higher H<sup>2</sup> evolution activity than that synthesized by mechanical loading (Hong et al., 2015).

## Sol-Gel Method

The sol-gel method is another widely applied method to get a close chemical interaction between semiconductors and carbon cocatalysts (Zarezade et al., 2011; Morales-Torres et al., 2012; Ng et al., 2012). Generally, this method need to prepare the sol

first, which is then mixed with the carbon materials uniformly. Subsequently, the gel is formed by aging followed with high temperature calcination to obtain the final composites. This method can control the crystal structure and uniformity of the supported nanoparticles, thus can fabricate photocatalysts with high activities. Li et al. used surfactant wrapping sol-gel method for the synthesis of CNT/TiO<sup>2</sup> core-shell nanocomposites (Li Z. et al., 2011). Using this method, they prepared uniform and distinct nanoscale anatase TiO<sup>2</sup> layer on the CNTs with tailored TiO<sup>2</sup> layer thickness with different Ti precursors (TEOTi, TTIP, and TBT). The CNT/TiO<sup>2</sup> composite prepared from TBT has thinner TiO<sup>2</sup> layer that provides shorter traveling distance for electron transferring to the CNT core, the activity for the degradation of MB was therefore higher than those prepared from TEOTi and TTIP. Kim et al. prepared GO–TiO<sup>2</sup> nanofibers (NFs) by using a sol-gel method and an electrospinning technique (Kim et al., 2014). They also compared the activity of GO–TiO<sup>2</sup> NF with GO(s)–TiO<sup>2</sup> NF (prepared by covering GO sheets on external surface of TiO<sup>2</sup> NF). Due to the stronger electronic coupling between GO and TiO<sup>2</sup> matrix and the reduced light shielding effect by hiding GO inside of TiO<sup>2</sup> NF, the photocatalytic H<sup>2</sup> production of GO–TiO<sup>2</sup> NF was higher than GO(s)–TiO<sup>2</sup> NF.

Although the materials prepared by sol-gel method have high purity and uniform particle size, some problems still exist, such as relatively long reaction time, large shrinkage during drying, and easy sintering with high temperature calcination. For example, Ren et al. prepared nanostructured LaFeO<sup>3</sup> nanoparticles (NPs) with rGO as a 2D template using a high temperature sol-gel method (Ren et al., 2016). Although the addition of C-support or rGO reduces the sintering degree of LaFeO3, it remains difficult to avoid sintering during the high temperature calcination for a long time.

# Other Methods

Microwave-assisted method is a green synthesis method based on the characteristics of microwave heating with tremendous advantages (Tian et al., 2016). Preparation of catalysts with special structure and high yield would be finished in a very short time using microwave heating. Thangavel et al. prepared the ZnS–rGO nanohybrids via microwave irradiation for 20 s over two cycles (Thangavel et al., 2016). Interestingly, Raman spectrum of the hybrids indicates the complete reduction of GO into rGO via the microwave treatment. After 2 h of irradiation, the ZnS–rGO showed higher degradation efficiency for MB (about 55.23%) and RhB (about 90.37%) than that of bare ZnS (about 40.79% for MB and 56.56% for RhB), respectively. They attributed the high activity to tight intermolecular binding, good interfacial contact between ZnS and rGO in the hybrid, and enhanced charge-transfer properties of rGO in nanohybrid. Zhang et al. successfully synthesized the graphene/Cu2O composites by a CVD (chemical vapor deposition) method. They also investigated the effects of the CVD growth parameters on the graphene flakes. The obtained composites were effective for the photocatalytic methyl orange degradation (Zhang et al., 2016a).

# PHOTOCATALYTIC APPLICATIONS

The photocatalytic activity of pure semiconductors can be enhanced by the addition of carbon materials as cocatalysts. The obtained composites are mainly used for the photocatalytic pollutants degradation, water splitting, CO<sup>2</sup> reduction, organic synthesis and so on (Abou Asi et al., 2013; Zhang et al., 2013a; Colmenares et al., 2016; Li K. et al., 2016; Zeng et al., 2017). In the following sections, we will focus their applications for photocatalytic hydrogen evolution and pollutants degradation.

# Photocatalytic Hydrogen Evolution

Hydrogen is considered as one of the most potential alternative energy in the twenty-first century (Zhang et al., 2015b; Zou and Zhang, 2015). Among the present hydrogen production methods, photocatalytic water splitting driven by sustainable solar energy is an ideal way to achieve clean hydrogen production (Matsuoka et al., 2007; Wang et al., 2009; Hisatomi et al., 2014). **Figure 2** describes the photocatalytic water splitting process with the presence of cocatalysts. Under the light irradiation, the electrons are photoexcited from the valence band (VB) to the conduction band (CB), while the holes are left in the VB, resulting in the separation of electrons and holes. Generally, for photocatalytic water splitting, the CB potential of semiconductor has to be more negative than hydrogen electrode potential EH+/H2, while the VB potential should be more positive than oxygen electrode potential EO2/H2O (Xu et al., 2016). Moreover, due to the impact of semiconductor band bending and presence of surface overpotential, the band gap of semiconductor should be larger than 1.23 eV to split water into H<sup>2</sup> and O<sup>2</sup> (Matsuoka et al., 2007; Moniz et al., 2015).

Carbon materials are effective H<sup>2</sup> evolution cocatalysts for the semiconductors mainly due to their large surface area and good charge mobility on their surface. Martha et al. synthesized RGO/InGaZn nanocomposites using a one-pot hydrothermal method (Martha et al., 2014). They also evaluated the effects of

hydrogen evolution.

RGO percentage on the H<sup>2</sup> evolution activity under visible-light irradiation (λ > 400 nm) (**Figure 3**). Three wt% rGO was proved to be the best loading percentage, and the H<sup>2</sup> generation rate can be as high as 435 µmol/h (**Figure 3A**). As shown in **Figure 3D**, InGaZn was uniformly dispersed on the surface of RGO, which was beneficial for the electrons moving from InGaZn to RGO. Moreover, the RGO could also provide more active adsorption sites and photocatalytic reaction centers. The stability test of RGO/InGaZn composite was also tested, and no deactivation could be found after four recycles (**Figure 3B**).

Silva et al. combined TiO<sup>2</sup> and CNTs using two different methods: hydration-dehydration labeled as (CNTox-TiO2) and one-pot oxidation (labeled as (CNT–TiO2)ox) (Silva et al., 2015). One wt% Pt was then loaded followed by calcination at 473 K and 673 K, respectively. The optimized catalyst Pt/(CNT–TiO2)ox-473 could obtain a H<sup>2</sup> evolution rate of 485 µmol/h, 2.4 times compared to the Pt/TiO2-473. According to the infrared attenuated total reflectance (ATR) spectra (see Figure 1 in the original paper, Silva et al., 2015), the bands from C = C and C– H are weaker in (CNT–TiO2)ox than in CNTox-TiO2, indicating a better dispersion of the TiO<sup>2</sup> particles at the surface of CNT in (CNT–TiO2)ox. This conclusion can be further confirmed by SEM and TEM images in. The better photocatalysis performance of (CNT–TiO2)ox might be related to the stronger interface interaction between TiO<sup>2</sup> and CNT, which is promoted by the oxidative treatment according to the ATR analysis.

Loading carbon materials as cocatayst, the bandgap of semiconductors could be narrowed to utilize the visible light with longer wavelength. Yu et al. prepared the CQDs/P25 composites with a "dyade"-like structure and applied them for photocatalytic hydrogen evolution under both UV-vis and visible light irradiation (**Figure 4**) (Yu H. et al., 2014). With methanol as the sacrificial agent, CQDs/P25–1.5 wt% showed the best photocatalytic performance under UV-vis light irradiation, and the evolution rate could reach 9.1µmol/h, 4 times higher than that of pure P25 (2.3µmol/h). While CQDs/P25–2.0 wt% was the optimized one under visible light with a H<sup>2</sup> evolution rate of 0.5µmol/h. The photocurrent response of these composites are shown in **Figures 4A,B**, which are consistent with the photocatalytic results. They believed that CQDs played dual roles to improve the photocatalytic activity of P25. CQDs could act as electron acceptors to improve the charge separation under UV-vis light irradiation. Meanwhile, they also served as a photosensitizer to sensitize P25 into a visible light response "dyade" structure for H<sup>2</sup> evolution under visible light irradiation.

Heteroatom doped carbon materials, such as nitrogen doped graphene, are proved to be better cocatalysts for semiconductor photocatalysts in recent years (Putri et al.,

2015). Yue et al. synthesized a ternary visible-light-driven photocatalyst for hydrogen evolution reaction. After decorating the CdS/Nb2O<sup>5</sup> heterojunction structure with N-doped graphene (NGR) nanosheets (Yue et al., 2017), the hybrid photocatalyst (2 wt% NGR) exhibited a high H<sup>2</sup> evolution rate of 100µmol h −1 g −1 , which was about 7.7 times than the pure CdS. Doping with nitrogen atom could change the electron density of the GR surface, thus can separation the photogenerated charges more efficiently. Jia et al. synthesized a series of nanocomposites by coupling CdS nanoparticles with NGR through calculation (Jia et al., 2011). The N-graphene/CdS was proved to be more efficient photocatalysts for hydrogen evolution compared to the CdS supported on undoped graphene. Significantly, the photocatalytic H<sup>2</sup> evolution rate of the N-graphene (2 wt %)/CdS reached 210µmol h−<sup>1</sup> without the addition of metal cocatalyst, which was much higher than graphene/CdS (99µmol h−<sup>1</sup> ) and GO/CdS (95µmol h−<sup>1</sup> ) with the same percentage of cocatalysts.

# Photocatalytic Degradation of Pollutants

Photocatalytic degradation of pollutants is another important application of photocatalysts. Photocatalysts can adsorb and degrade pollutants in water and toxic gas in air under illustration, which thus has great potential for environmental remediation. Previous studies have shown that photocatalysis technology can not only degrade organic pollutants into CO2, H2O, and inorganic salt, but also eliminate the heavy metal ions (Akpan and Hameed, 2009; Peng et al., 2014; Murgolo et al., 2015; Jing et al., 2017).

Ming et al. synthesized dandelion-like ZnS/CQDs hybrid materials using hydrothermal method with CTAB as surfactant (Ming et al., 2016). As shown in **Figures 5A–D**, some dark dots are distributed on the ZnS nanowires uniformly. Coating the optimal content of 2 wt% CQDs, the photocatalyst showed the highest degradation rate, which was about 1.67 and 2.11 times higher than bare ZnS for MB and RhB, respectively. As illustrated in **Figure 5E**, the intensity of the PL emission band decreased obviously after the loading of CQDs on ZnS. The 2 wt% CQDs/ZnS possessed the lowest intensity, suggesting the lowest recombination possibility of photoexcited holes and electrons. They also proposed the photocatalytic mechanisms on the CQDs/ZnS hybrid:

$$\text{ZnS} + \text{hv} \rightarrow \text{e}\_{\text{ZnS}}{}^- + \text{h}\_{\text{ZnS}}{}^+ \tag{1}$$

$$
\mathbf{e}\_{\text{ZnS}}^{-} \rightarrow \mathbf{e}\_{\text{CQDs}}^{-} \tag{2}
$$

$$\rm e^-\_{\rm CQDs} + \rm O\_2 \to \rm \*O\_2^- \tag{3}$$

$$\rm{h}\_{\rm ZnS}^{+} + \rm{H}\_{2}\rm{O} \rightarrow \rm{\text{\textdegree}}^{\*}\rm{OH}^{-} \tag{4}$$

$$2^\*\text{OH} \to \text{H}\_2\text{O}\_2\tag{5}$$

$$\rm{^{1}H\_{2}O\_{2} + ^{\*}O\_{2}^{-} \rightarrow OH^{-} + \rm{^{\*}OH} + O\_{2}} \tag{6}$$

$$\rm \rm \rm \rm \rm OH + dye \rightarrow \rm H\_2O + CO\_2 + intermediatees \qquad \text{(7)}$$

Qi and his co-workers prepared a series of fullerene-modified anatase TiO<sup>2</sup> (C60@a-TiO2) nanocomposites by a simple solution phase method (Qi et al., 2016). By the introduction of C60, the activity of C60@a-TiO<sup>2</sup> for photocatalytic degradation of MB

e

could be enhanced greatly under UV-A light irradiation. In order to confirm the electronic structures of C60@a-TiO2, the density functional theory (DFT) was used for a theoretical calculation toward the C60-COOH@a-TiO<sup>2</sup> (101) surface. The adsorption energy and the projected density of states (PDOS) for the C60- COOH@a-TiO<sup>2</sup> (101) surface were calculated. Strong covalent interaction between C<sup>60</sup> and the a-TiO<sup>2</sup> (101) surface was present with the calculated adsorption energy of 3.61 eV. Moreover, the introduction of C<sup>60</sup> narrows the band gap to 0.8 eV, resulting in the red shift of light absorption edge of the C60-COOH@a-TiO<sup>2</sup> heterojunctions. According to the DFT results, there is an additional doping state present between the valance band and conduction band by the incorporation of C<sup>60</sup> on the a-TiO<sup>2</sup> (101) surface. The activity of C60@a-TiO<sup>2</sup> is therefore enhanced with more efficient charge separation efficiency and increased light absorption range.

Sampaio et al. used both GO–TiO<sup>2</sup> and CNT–TiO<sup>2</sup> materials for the photocatalytic degradation of the cyanobacterial toxin, microcystin-LA (MC-LA) under simulated solar light and visible light irradiation (Sampaio et al., 2015). The GO–TiO<sup>2</sup> composite containing 4 wt% of GO exhibited the highest photocatalytic activity under both simulated solar light and visible light irradiation. The enhanced activity of GO–TiO<sup>2</sup> was attributed to the optimal assembly and interfacial coupling between TiO<sup>2</sup> nanoparticles and GO sheets, which can effectively inhibit electron-hole recombination. While the activity of CNT– TiO<sup>2</sup> for the MC-LA removal under visible light irradiation was mostly due to adsorption instead of photocatalytic degradation.

Murgolo et al. fabricated a composite photocatalyst by combining SWCNTs with nano-sized TiO<sup>2</sup> NRs (Murgolo et al., 2015). The composite showed tailored photocatalytic properties for the photocatalytic degradation of a mixture of 22 organic pollutants under both UV and simulated solar light. The experiment results showed that this composite displayed comparable degradation rates over Degussa P25 under UV irradiation. While the SWCNTs/TiO<sup>2</sup> showed slightly lower efficiency than Degussa P25 under simulated solar irradiation. The SWCNTs/TiO<sup>2</sup> can be reused easily by a mild centrifugation or a filtration. This photocatalyst has proved to be a promising candidate in photocatalytic pollutants degradation, which can also be integrated with a biological step for the enhanced removal of emerging organic pollutants.

Heteroatoms doped carbon materials are also effective cocatalysts for photocatalytic degradation reaction. Liu et al. synthesized N-CNT/mpg-C3N<sup>4</sup> composites via thermal polycondensation (Liu J. et al., 2017). N-CNT has better electronic conductivity and more defective structure than undoped CNT, which could therefore accept electrons more easily. Benefiting from the synergistic effect between N-CNT and mpg-C3N4, the composites show enhanced photo-degradation activity for rhodamine B, methyl orange and tetracycline hydrochloride under visible light irradiation. Due to the special 2D structure of graphene, which can also be combined with other layered materials to fabricate hybrid cocatalysts (Chen et al., 2017; Peng et al., 2017). Our group have used the MoS2/graphene hybrids for the modification of CdS and Ag3PO4, and the obtained composites showed improved photocatalytic activity for phenols degradation and nitroaromatic compounds detoxification (Peng et al., 2014, 2016). The photo-activity of the final composite could also be adjusted by changing the ratio of MoS<sup>2</sup> and graphene.

# COMPARISON OF CARBON ALLOTROPES AS COCATALYSTS

There have been some other relevant reviews on this subject, but as far as we are concerned, a horizontal comparison of these carbon cocatalysts in photocatalysis field is still lack. In this section, we summarized some examples which compared different carbon cocatalysts for the modification of semiconductors. Zarezade et al. used sol-gel method to synthesize TiO2/AC and TiO2/MWCNT hybrid materials (Zarezade et al., 2011). Although the surface area of TiO2/MWCNTs was smaller than that of TiO2/ACs, the activity of TiO2/MWCNTs was even higher for photocatalytic degradation of AB92. The defects of MWCNTs could be used as anchor sites for the growth of TiO<sup>2</sup> crystallites, which can lead to the uniform distribution of TiO<sup>2</sup> on the MWCNT surface. After calcination of the composite at 500◦C (**Figure 6A**), a remarkable photocatalytic performance could be achieved with a maximum degradation percentage of 86% in 2 h (**Figure 6B**).

Ye's group compared the photocatalytic behaviors of CdS– graphene (CdS–GR) and CdS–carbon nanotube (CdS–CNT) nanocomposites as photocatalysts for the hydrogen evolution and the degradation of methyl orange (MO) under visiblelight irradiation (Ye et al., 2012). **Figure 7A** reveals that both the CdS–GR and the CdS–CNT composites display enhanced photocatalytic H<sup>2</sup> evolution activities. Furthermore, the CdS–GR composite is more efficient than the CdS– CNT composite under their optimized mass ratios. The H<sup>2</sup> evolution rate over the CdS–GR composite could reach 70µmol h −1 , which is 1.3 times higher than that of the CdS–CNT (52µmol h−<sup>1</sup> ). Similarly, **Figure 7B** shows that GR is more efficient to enhance the photocatalytic performance of CdS for the degradation of MO. The degradation percentage of MO over the optimized CdS–GR (1: 0.01) is as large as 95%, 1.8 times higher than that of the optimized CdS–CNT (1: 0.03) after 60 min irradiation (**Figure 7C**). The stronger interaction and larger contact interface between CdS and GR facilitate the transfer of photogenerated electrons from CdS to GR, leading to a higher efficiency in the separation of photogenerated electron-hole pairs and a higher photocatalytic performance of the CdS–GR composite than the CdS–CNT composite.

Cherevan and coworkers hybridized both multi-walled CNTs and graphene oxide (GO) with Ta2O<sup>5</sup> semiconductor via a in situ hydrothermally assisted sol-gel method (Cherevan et al., 2014). Surprisingly, CNT–Ta2O<sup>5</sup> hybrid exhibited superior performance over GO–Ta2O<sup>5</sup> hybrid, and a maximum H<sup>2</sup> evolution rate of 1,600µmol h−<sup>1</sup> could be obtained for CNT– Ta2O5. This result is opposite to many other studies, which could be attributed to two reasons: (1) the amount of Ta2O<sup>5</sup> in the GO hybrid is much lower than in the CNT hybrid; (2) annealed CNTs are expected to possess better charge transfer properties than highly defective GO.

Jing et al. compared the degradation efficiencies of methylene blue (MB) over AgSiOx@CNT and AgSiOx@RGO nanocomposites under visible light (Jing et al., 2017). Interestingly, AgSiOx@CNT has a better photodegradation performance than AgSiOx@RGO at a small amount of CNTs, while the removal rate with AgSiOx@RGO is faster than AgSiOx@CNT at high carbon contents. This is probably because the different functional mechanism of these two carbon materials. The low content of CNT could boost the synergistic effect of the nanocomposite by reducing the electron transfer resistances and prolonging the lifetime of electron-hole pairs. However, as for AgSiOx@RGO, adsorption effect is dominant rather than photodegradation as RGO contains residual oxygen-containing groups.

Yang et al. presented a comparative study of photocatalytic selective oxidation on several carbon based photocatalysts (Yang M. Q. et al., 2013). They synthesized a series of TiO2-GR, –CNT, and –C<sup>60</sup> photocatalysts by combining sol-gel with hydrothermal methods. These three different carbon allotropes affected slightly in the morphology, crystal phase, particle size, pore volume and surface area the of the supported TiO<sup>2</sup> nanocrystals. The TiO2-carbon (GR, CNT, and C60) have similar photocatalytic activities and analogous reaction mechanisms toward selective oxidation of benzyl alcohol. Different preparation methods could obtain different structural composition and synergetic interaction between TiO<sup>2</sup> and carbon, which therefore have a greater impact on the photocatalytic performance of TiO2 carbon composites. The comparison shows that GR fails to prove its unique advantage compared to the other two carbon allotropes. Similarly, Zhang et al. investigated TiO2-Graphene as high-performance photocatalyst for the gas-phase degradation of benzene (Zhang et al., 2010). They concluded that GR was in essence the same as other carbon materials (carbon nanotube, activated carbon, and fullerene) as cocatalysts on enhancement of photocatalytic activity of TiO2, although GR has unique structural and electronic properties in comparison with other carbon allotropes.

Due to the special 2D structure and excellent physical/chemical properties, we expected the graphene will show better performance compared to other carbon allotropes (An and Yu, 2011; Zhang et al., 2011). However, it didn't show superior cocatalytic properties compared to the CNT or carbon quantum for the modification of some semiconductors (Ma et al., 2016). Researchers has tried to modify the graphene further by heteroatoms doping or activation method, which could increase its electric conductivity or surface area. The performance of the modified graphene could be then enhanced further as photocatalytic cocatalysts, thus increasing its real application potential.

# MECHANISM OF CARBON COCATALYSTS FOR PHOTOCATALYTIC ACTIVITY IMPROVEMENT

It has been proposed that the photocatalytic activity enhancement is due to the synergistic effect between semiconductor and carbon materials. Generally, carbon materials play four primary roles as cocatalysts for the activity enhancement of the semiconductors (Tan et al., 2012; Bai et al., 2016). (1) They provide a structure with larger specific surface area over which the active component can be well-dispersed, thus increasing the active sites. Activated carbon is amorphous carbon with a specific surface up to 3,000 m<sup>2</sup> g −1 (Strobel et al., 2006). Graphene, the 2-dimensional nanosheets composed of sp<sup>2</sup> -hybridized carbon atoms, possesses an extremely high specific surface area (theory value of 2,630 m<sup>2</sup> g −1 ) (Fan X. et al., 2015). While the CQDs can distribute uniformly on the surface of semiconductor materials because of its small size. (2) During the photocatalytic degradation of organic pollutants, carbon materials can be used as adsorbent to improve the adsorption capacity of semiconductors (Matos et al., 2001; Ai et al., 2015). (3) Carbon materials can be doped as a photosensitizer for bandgap narrowing, which is favorable for expanding the visible light absorption region of semiconductors. (4) By the formation of carbon materials–semiconductor heterojunction, the excellent electron transfer could be achieved, leading to the enhanced charge separation efficiency and photocatalytic activity (Guldi et al., 2006; Li X. et al., 2016; Shi et al., 2017).

# CONCLUSIONS AND FUTURE PROSPECTS

Carbon materials are important photocatalytic cocatalysts due to their low cost and high efficient. In this review, we summarized the recent development of the carbon materials based semiconductor photocatalysts, including their synthesis methods and the applications for H<sup>2</sup> evolution and pollutants degradation. Zero-dimensional C60, CQDs, one-dimensional CNTs, two-dimensional GR, and activated carbon are all involved to provide valuable information for metal free cocatalysts selection. Although much progress has been achieved, some essential issues are still unaddressed, especially for the activity and stability enhancement mechanisms. Studies about the interface between the semiconductors and the cocatalysts should be helpful for new carbon materials based photocatalysts development. Computational chemistry using DFT could also provide valuable information for the photocatalysts design. Although more in-depth studies are still needed, carbon materials based photocatalysts have great potential to address various environmental and energy-related problems.

# AUTHOR CONTRIBUTIONS

WH chose the references and edit the draft. ZL provided assistance for literature search and some revision. YL, XF, FZ, and GZ provided professional advice. WP designed the main content

# REFERENCES


and revised the manuscript. All authors read and approved the final manuscript version to be submitted.

# ACKNOWLEDGMENTS

This research was supported by the project No. 21506158 from the National Natural Science Foundation of China (NSFC).

from renewable and biodegradable resources. Green Chem. 18, 5736–5750. doi: 10.1039/C6GC02477G


photocatalytic hydrogen evolution from water under visible light irradiation. J. Phys. Chem. C 115, 11466–11473. doi: 10.1021/jp2023617


and saccharides using carbon nanotube-TiO<sup>2</sup> catalysts. Appl. Catal. B Environ. 178, 82–90. doi: 10.1016/j.apcatb.2014.10.032


**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 © 2017 Han, Li, Li, Fan, Zhang, Zhang and Peng. 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) or licensor 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.

# Facile Fabrication of BiOI/BiOCl Immobilized Films With Improved Visible Light Photocatalytic Performance

Yingxian Zhong1†, Yuehua Liu1†, Shuang Wu<sup>1</sup> , Yi Zhu<sup>1</sup> \*, Hongbin Chen<sup>1</sup> , Xiang Yu<sup>2</sup> and Yuanming Zhang<sup>1</sup> \*

*<sup>1</sup> Department of Chemistry, Jinan University, Guangzhou, China, <sup>2</sup> Analytical and Testing Center, Jinan University, Guangzhou, China*

### HIGHLIGHTS

### Edited by:

*Zhimin Ao, Guangdong University of Technology, China*

### Reviewed by:

*Fa-tang Li, Hebei University of Science and Technology, China Kaixuan Bu, Rutgers University, The State University of New Jersey, United States*

### \*Correspondence:

*Yi Zhu tzhury@jnu.edu.cn Yuanming Zhang tzhangym@jnu.edu.cn*

*† These authors have contributed equally to this work.*

### Specialty section:

*This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry*

Received: *13 December 2017* Accepted: *23 February 2018* Published: *12 March 2018*

### Citation:

*Zhong Y, Liu Y, Wu S, Zhu Y, Chen H, Yu X and Zhang Y (2018) Facile Fabrication of BiOI/BiOCl Immobilized Films With Improved Visible Light Photocatalytic Performance. Front. Chem. 6:58. doi: 10.3389/fchem.2018.00058* • A facial method was used to fabricate BiOI/BiOCl film at room temperature.


Photocatalysis has been considered to be one of the most promising ways to photodegrade organic pollutants. Herein, a series of BiOI/BiOCl films coating on FTO were fabricated through a simple method at room temperature. The photocatalytic efficiency of 30%BiOI/BiOCl could reach more than 99% aiming to degrading RhB and MB after 90 and 120 min, respectively. Compared with BiOCl, 30%BiOI/BiOCl showed 12 times higher efficiency when degrading RhB. In comparison with BiOI, 30%BiOI/BiOCl showed 5 and 6 times higher efficiency when degrading RhB and MB, respectively. These obvious enhancements were attributed to expanded visible light absorption and high separation performance of photoinduced charge. Moreover, the photocatalytic activity of 30%BiOI/BiOCl had no obvious decrease after five recycles, suggesting that it was a promising photocatalyst for the removal of MB and RhB pollutants. Finally, the possible growth process for the BiOI/BiOCl thin films and photocatalysis mechanism were investigated in details. This work would provide insight to the reasonable construction of BiOX heterojunction and the photocatalytic mechanism in degrading organic pollutants.

### Keywords: BiOI/BiOCl film, visible light, heterojunction, photodegradation, recycle

# INTRODUCTION

Recently, semiconductor photocatalysts have been potential materials in energy storage, organic pollutants degradation and so on Kisch (2013). Since TiO<sup>2</sup> had been reported to produce H<sup>2</sup> under UV light (Fujishima and Honda, 1972), transitional mental oxides have been applied as photocatalysts, such as ZnO (Soci et al., 2007), SnO<sup>2</sup> (Law et al., 2002), and WO<sup>3</sup> (Baeck et al., 2003). However, many of them have wide bandgap and are activated by UV-light (4% of solar light). To utilize more solar light, searchers pay a lot of efforts to find new photocatalysts which could maximize the utilization of solar light. Among those photocatalysts, BiOCl is considered as a new kind of promising layered material for photocatalysis due to its unique layered structure, high chemical and optical stability, corrosion resistance and nontoxicity (Li J. et al., 2014; Ding et al., 2015; Li et al., 2017). BiOCl has layered structure consisting of [Bi2O2] <sup>2</sup><sup>+</sup> sandwiched between two slabs of Cl−, which produces internal static electric fields to separate photogenerated electrons and holes (Cheng et al., 2014; Mi et al., 2016). However, the practical application of BiOCl has been hindered owing to its wide bandgap and relatively high recombination rate of photoinduced carriers (Dong et al., 2012; Xiao et al., 2012).

Aiming at solving these shortcomings, many strategies have been reported to enhance the photocatalytic efficiency of BiOCl, including: (i) impurity element doping, such as BiOClxBryI<sup>z</sup> (Sun X. et al., 2015) and BiOClxI1−<sup>x</sup> (Kim et al., 2014), (ii) surface functionalization, like inducing oxygen vacancies in BiOCl (Jiang et al., 2013), (iii) construction of the plasmonic photocatalysis system, such as Ag/BiOCl (Liu H. et al., 2012) and Ag-AgX-BiOX (X = Cl, Br, I) (Cheng et al., 2011; Xiong et al., 2011; Cao et al., 2013), (iv) construction of semiconductor heterojunctions (Jiang et al., 2011; Wang et al., 2015). Construction of semiconductor heterojunctions has been widely explored in recent years because of two advantages. First, materials with wide bandgap could match with lots of semiconductors at the energy level. In that way, it is propitious to electron and hole separation by building an interfacial electric field between different semiconductors. Cui's work showed that photodegradation efficiency of Ag3PO4/BiOI was nearly 10 times that of BiOI (Cui et al., 2013). And Cui's group found that photodegradation efficiency of BiOI/Bi2WO<sup>6</sup> was about 6.1 times higher than that of pure Bi2WO<sup>6</sup> under visible light irradiation (Li et al., 2013). Ao's work showed that Ag2MoO4/g-C3N<sup>4</sup> highly improved photocatalytic degradation performance for different organic pollutants under sunlight irradiation (Wu et al., in press). Secondly, coupled with narrow band semiconductors, BiOCl could expand visible light absorption and utilize more solar energy. Narrow bandgap materials act as the light absorber and generate photoinduced carriers with proper energy, indicating that it is a very efficient visiblelight-activated photocatalyst (Wang et al., 2017). Therefore, many BiOCl/narrow bandgap materials, such as BiOCl-C3N<sup>4</sup> (Wang et al., 2013), BiOCl/Bi24O31Cl<sup>10</sup> (Li F. et al., 2014), BiOCl/Bi2S<sup>3</sup> (Cheng et al., 2012), BiOCl/BiOI (Sun L. et al., 2015), BiOCl/BiOBr (Zhang et al., 2013), and NaBiO3/BiOCl (Chang et al., 2010), have been successfully prepared.

Based on the advantages mentioned above, BiOI is a great candidate to couple with BiOCl, which is a narrow bandgap semiconductor (1.72 eV) and has a similar layered structure (Jiang et al., 2011; Huang et al., 2015; Ning et al., 2016). Once coupled with BiOI, BiOI/BiOCl is expected to achieve the aims as followed: (i) to enhance visible light absorption, (ii) to accelerate separation efficiency of photoinduced electrons and holes (Cao et al., 2011; Xiao et al., 2012; Wang et al., 2016). Although there are a few reports about BiOI/BiOCl, most of them are powder synthesized through hydrothermal and solvothermal methods, which needs high temperature and pressure. Additionally, powder catalysts are hard to be separated and recovered because they are easily dispersed into solution when used in pollutants degradation (Zhao et al., 2015). Unlike powder catalysts, immobilized photocatalysts become more promising in practical application for easy separation and high reusability (Liu X. et al., 2012). Therefore, BiOI/BiOCl film is of great advantage in practical organic pollutants degradation.

In this work, a facial method was used to fabricate a series of BiOI/BiOCl immobilized films at room temperature. The possible growth process of BiOI/BiOCl film was investigated in detail. All BiOI/BiOCl films showed better photocatalytic performance than pristine BiOCl film. UV-vis diffusion reflectance spectra, photocurrent, fluorescence spectra (PL) and trapping experiment were used to gain insights into the reasons for remarkable enhancement of photocatalytic activity and the possible photocatalysis mechanism of BiOI/BiOCl film. Besides, recycle experiments were used to measure the stability and duration of BiOI/BiOCl film.

# EXPERIMENTAL

# Synthesis of xBiOI/BiOCl Film

In a typical procedure, 3.0 g BiCl<sup>3</sup> was mixed with 100 mL ethanol and 1 mL HCl, and stirred for 1 h to form BiCl<sup>3</sup> solution. Similarly, BiI<sup>3</sup> solution was prepared using BiI3, HI and ethanol in the same way. After that, BiCl<sup>3</sup> solution and BiI<sup>3</sup> solution were mixed with different molar ratio.1 mL of mixture solution was dropped onto FTO glass. After being dried at 100◦C for 1 h, the films were dipped into distilled water for 30 min to form BiOI/BiOCl (as shown in **Scheme 1**). Finally, the samples dried at 60◦C for 2 h. The xBiOI/BiOCl composites with molar ratios of

BiOI to BiOCl at 10, 30, and 60% were named as 10%BiOI/BiOCl, 30%BiOI/BiOCl, 60%BiOI/BiOCl, respectively.

# Characterization of Photocatalysts

The morphologies and phase structures of xBiOI/BiOCl films were observed by Field emission scanning electron microscopy (FE-SEM, Zeiss ULTRA 55), transmission electron microscopy (JEOL 2010F) and high-resolution transmission electron microscopy (JEOL 2100 F) and X-ray diffractometry (XRD, equipped with a Cu Ka X-ray source). The optical properties of as-synthesized catalysts were tested by UV–vis spectrophotometer (DRS, Hitachi- UV-3010, using BaSO<sup>4</sup> for the baseline measurement) and photoluminescence spectroscopy (PL, RF-5301PC). FT-IR spectra were recorded on an Aipha-Centuart FT-IR spectrometer.

The visible-light-driven photocatalytic efficiencies of xBiOI/BiOCl films were evaluated the degradation of Rhodamine B (RhB, 2.5 mg L−<sup>1</sup> ) and methylene blue (MB, 2.5 mg L−<sup>1</sup> ) in a reactor equipped with a 350 W Xe lamp with >420 nm filter as the light source. The as-obtained BiOI/BiOCl film was putting into a reactor, in which 100 mL dye solution were poured. Before irradiation, the solution was continuously stirring in the dark for 30 min to ensure establish adsorption-desorption equilibrium. At certain time interval, 4 mL of the suspension were sampled; the concentration of dye solution was measured by recording the absorption band maximum in the absorption spectra. For comparison, the photocatalytic activities of BiOCl and BiOI were characterized under same condition. In addition, 30%BiOI/BiOCl photocatalyst was examined by 5-cycle to characterize its stability. Before entering next cycle, samples were washed by deionized water and alcohol three times. Dried at 100◦C for 1 h and reuse in fresh dye solution.

# Electrochemical Measurements

Photocurrent of samples was studied by there-electrode system in a quartz cell, which was using Pt plate as counter electrode, Ag/AgCl as reference electrode, and the as-prepared samples as working electrode on electrochmical workstation (CHI660C.Shang-hai.). 0.1 M Na2SO<sup>4</sup> solution was used as the electrolyte. The surface area of the working electrode was 4 × 5 cm<sup>2</sup> . A 350 W Xe lamp with an filter (λ > 420 nm) was used as the visible-light source.

# RESULTS AND DISCUSSION

# XRD Patterns

**Figure 1** showed XRD patterns of the as-prepared xBiOI/BiOCl films. It could be seen that all the diffraction peaks of BiOI and BiOCl were in good agreement with the standard cards (JCPDS No. 73-2062) and (JCPDS No. 06-0249) without any impurity peaks, which indicated that they exhibited tetragonal structure and corresponded to the FT-IR results (**Figure S2**). The characteristic peaks of BiOI and BiOCl coexisted in the XRD patterns, demonstrating the formation of BiOI/BiOCl composite without the present of BiOClxI1−<sup>x</sup> solid solutions (Huang et al., 2015). With the increase of percentage of BiOI in the composites, the strength of diffraction peaks of BiOI gradually increased, on

the contrary, the intensity of BiOCl simultaneously decreased. Additional, it could see that FTO peaks in **Figure 1**, it might be due to the uneven film on glass of the sample.

# SEM, TEM, HRTEM, and EDS Images

All the samples were systematically analyzed by SEM. From **Figure 2A**, it could be observed that pristine BiOCl was composed of numerous nanosheets and its surface was very smooth. Differently, under similar preparation conditions, pristine BiOI presented hierarchical microspheres consisting of a series nanosheet in **Figure 2B**. As for xBiOI/BiOCl (**Figures 2C–E**), it could be observed that xBiOI/BiOCl showed hierarchical structure with BiOCl nanosheets adhering tightly on BiOI and the particle sizes of xBiOI/BiOCl obviously increased in comparison with pristine BiOCl. Additionally, color of sample gradually deepened compared with pristine BiOCl when percentage of BiOI increased in **Figure 2F**.

The morphology and structure of as-obtained samples were further characterized by TEM and HRTEM images. The microstructures of pristine BiOCl, pristine BiOI and 30%BiOI/BiOCl were shown in **Figure 3**. The interactions between BiOCl and BiOI were so strong that ultrasonication did not separate them during the sample preparation procedure for TEM characterization (Xiao and Zhang, 2010). **Figures 3B,D,F** indicated that the samples were highly crystallized. In **Figure 3B**, the lattice fringe with a d-spacing of 0.735 nm matched well with (001) lattice plane of BiOCl, while in **Figure 3F**, the interlayer distance of 0.280 nm responsed to the (280) plane of BiOI. **Figure 3D** showed the HRTEM of 30%BiOI/BiOCl, clear fringes with the lattice spacing of 0.264 and 0.280 nm could be indexed to (102) lattice plane of BiOCl and (110) lattice plane of BiOI, respectively. TEM results were in good consistent with XRD patterns in **Figure 1**. The results clearly confirmed the formation of heterostructure between BiOCl and BiOI. In addition, the elemental distributions of 30%BiOI/BiOCl were studied through EDS elemental mapping. The corresponding

results for 30%BiOI/BiOCl were shown in **Figures 4A–E**. It could be obviously seen that the sample consist of only I, Bi, Cl, O, elements. The results of EDS mapping confirmed the composition, structure and the high purity of 30% BiOI/BiOCl composite.

# Growth Process of xBiOI/BiOCl Film

In order to understand the growth process of BiOI/BiOCl film in this work, SEM images and XRD patterns of 30%BiOI/BiOCl film at different time stage were shown in **Figure 5**. The whole process was divided into three stages. In the first stage (0 min), as shown in **Figure 5I**, the peaks collected from 0 min sample could be indexed as a composition of BiOCl and BiI3. After the mix solution consisting of BiI<sup>3</sup> and BiCl<sup>3</sup> was dropped on FTO, there was a hydrolyzation competition between them. BiCl<sup>3</sup> was hydrolyzed to form BiOCl prior to the hydrolyzation of BiI<sup>3</sup> when ethanol volatilized, because the Ksp (BiOCl) was smaller than Ksp (BiOI). As shown in **Figure 5B**, BiI<sup>3</sup> broke down on the nanosheets structure of BiOCl to form into hierarchical structure. Besides, the diffraction peak at around 11.9◦ corresponding to the (001) plane shifted to smaller 2θ in **Figure 5I**. That might be because the ionic radius of I<sup>−</sup> (220 pm) was larger than that of Cl<sup>−</sup> (181 pm). In the second stage (1–15 min), BiI<sup>3</sup> was hydrolyzed to BiOI. In **Figure 5I**, with the increase of reaction time, diffraction peak of BiI<sup>3</sup> disappeared and the intensity of BiOI became stronger. In the meantime, the extent of hydrolyzation caused the shifting of the diffraction peak of (001) to bigger 2θ. **Figures 5C–E** showed that hierarchical BiOI and nanosheets-structure BiOCl formed a tidily hierarchical structure in the second stage. In the third stage (30 min), BiI<sup>3</sup> was hydrolyzed totally, and BiOI/BiOCl was formed.

FIGURE 5 | SEM images of (A) BiOCl and 30%BiOI/BiOCl after different hydrolysis times (B) 0 min, (C) 1 min, (D) 3 min, (E) 5 min, (F) 10 min, (G) 15 min, (H) 30 min, (I) XRD patterns of 30%BiOI/BiOCl after different hydrolysis time.

# Optical Properties

The UV-vis diffuse reflectance spectra (DRS) of xBiOI/BiOCl films were shown in **Figure 6**. BiOCl had a strong absorption edge around 360 nm, meanwhile, BiOI had a strong absorption edge around 700 nm. Compared to BiOCl, xBiOI/BiOCl showed an absorption edge shifting to larger wavelength with the increase of BiOI percentage. This shifting was in accordance with the color change of as-prepared samples (**Figure 2F**) caused by the addition of BiOI. The band gap energy of a semiconductor could be calculated from the following equation:

$$
\alpha h\upsilon = A(h\upsilon - Eg)^{n/2} \tag{1}
$$

where α, ν, Eg, and A were the absorption coefficient, light frequency, band gap energy, and a constant, respectively (Ning et al., 2016). Among them, n depended on the characteristics of the transition in a semiconductor. For example, n = 1 (direct transition) or n = 4 (indirect transition). BiOX belonged to indirection transition, thus n was estimated to be 4. The band gap of BiOI and BiOCl were 1.74 and 3.34 eV, respectively. With narrowing of band gap, xBiOI/BiOCl could exhibit enhanced visible light absorption, subsequently resulting in improved photocatalytic activity.

Besides optical absorption property, separation efficiency of photogenerated carriers played an important role in photodegradation. Photocurrent could directly indicate the capability of charge separation. The higher photocurrent density

corresponded to the greater capability of charge separation. **Figure 7** showed the photocurrent responses of the as-prepared photocatalysts under several on/off visible light irradiation cycles. BiOCl and BiOI showed a poor photocurrent response while the response of xBiOI/BiOCl increased. The photocurrent density of 30%BiOI/BiOCl was almost 6 times as high as that of pristine BiOCl and 3 times as high as that of pristine BiOI, revealing that 30%BiOI/BiOCl had superior separation efficiency of photogenerated carriers.

Photoluminescence spectra were used to characterize the photogenerated carriers' recombination rate of as-prepared samples, since the PL emission originated from free carrier's recombination. The higher PL intensity meaned the higher recombination rate in the photocatalytic procedure (Cao et al., 2011). As shown in **Figure 8**, BiOCl showed a strong emission peak with high intensity at approximate 420 and 440 nm, meanwhile, BiOI exhibited a low intensity. Decline of the PL intensity implied that adding BiOI could successful suppress recombination process during photocatalysis. In addition, 30%BiOI/BiOCl shown the lower intensity indicated the lower recombination rate, thus could promote photocatalytic activity.

### Photocatalytic Properties

The photodegradation efficiency of the xBiOI/BiOCl films were evaluated by degradating RhB and MB under visible-light irradiation. As shown in **Figure 9A**, the degradation percentage of RhB by pristine BiOCl was 48% in 90 min. It was about 70% by pristine BiOI in 90 min. Compared with pristine BiOCl and BiOI, xBiOI/BiOCl film showed a great degradation: 30% BiOI/BiOCl could degrade more than 99% of RhB in 90 min. To further illustrate the photocatalytic reaction, pseudo-first-order kinetics were fitted from the degradation process (Ning et al., 2016),

$$
\ln(\text{C}\_0/\text{C}) = k\text{t} \tag{2}
$$

where the value of rate constant k was equal to the corresponding slope of the fitting line as shown in **Figure 9C**. The rate constant

value for 30%BiOI/BiOCl was 0.07315 min−<sup>1</sup> , which was 12 times higher than BiOCl (0.00575 min−<sup>1</sup> ) and 5 times higher than BiOI (0.01303 min−<sup>1</sup> ), respectively. **Figure 9B** showed the photocatalytic performance of the xBiOI/BiOCl evaluated by degradating MB under visible-light irradiation. 48% of MB was self-degraded under visible light irradiation. Compared with selfdegradation of MB, the photocatalytic performance of BiOCl was negligible and BiOI could only degrade 60% MB. The photocatalytic performance of xBiOI/BiOCl film was much better than pristine BiOCl and BiOI. More than 99% of MB was degraded using 30%BiOI/BiOCl film in 120 min. According to **Figure 9D**, the rate constant value of 30%BiOI/BiOCl was 0.05218 min−<sup>1</sup> , which was 6 times higher than BiOI (0.00772 min−<sup>1</sup> ). This better photocatalytic performance might be due to the enhanced visible light absorption and improved separation efficiency of photoinduced carriers.

# Reusability of 30%BiOI/BiOCl Film

Efforts were made in this work to identify the stability and practicality of as prepared catalysts for dye degradation, which was a significant factor to be considered in real application. 30%BiOI/BiOCl film was reused for RhB and MB degradation in five cycles under the same condition and the result was shown in **Figures 9E,F**. It was remarkable that the efficiencies had no obvious decrease after 5 cycles, revealing its great reusability. In term of XRD patterns in **Figure S1**, there was no obvious change in phase and structure of 30%BiOI/BiOCl film after 5 cycles, demonstrating its excellent stability. The excellent reusability and stability indicated its great potential in practical application.

# Photocatalytic Mechanism

The energy band structures of BiOX were evaluated using the following equation (Xiao et al., 2016):

$$\mathbf{E\_{VB}} = \mathbf{X} - \mathbf{E^{e}} + \mathbf{0.5E\_{\mathcal{B}}} \tag{3}$$

$$\mathbf{E\_{CB}} = \mathbf{E\_{VB}} - \mathbf{E\_{\mathcal{K}}} \tag{4}$$

Where EVB was the valence band edge potentials, X was the electronegativity of BiOX, which was the geometric mean of the electronegativity of constituent atoms, E<sup>e</sup> was the energy of free electrons on the hydrogen scale (about 4.5 eV), E<sup>g</sup> was the band gap energy, ECB was the conductance band edge potentials (Xiao et al., 2016). The EVB of BiOCl and BiOI were 3.60 eV and 2.11 eV, respectively. And the ECB of BiOCl and BiOI were 0.26 and 0.37 eV, respectively.

Active species of 30%BiOI/BiOCl film was detected by typical trapping experiments. Benzoquinone (BQ) was used as superoxide radical species (•O − 2 ) scavenger, while dimethylcarbinol (IPA) was used as quencher of •OH and EDTA-2Na was used as hole scavenger (h+). In **Figures 10A,B**, IPA could significantly decrease the photocatalytic efficiency; otherwise, BQ and EDTA-2Na had less effect on it. **Figure 10** indicated that •OH, •O − 2 and h<sup>+</sup> were active species during the degradation of RhB and MB. A possible mechanism of BiOI/BiOCl film was proposed based on the above discussion. In **Figure 11**, BiOI could utilize visible-light with energy <2.95 eV (λ > 420 nm). Photoinduced electrons could be excited to a higher potential edge of BiOI (−0.84 eV) which was negative than that of BiOCl (0.26 eV). Then, photogenerated electrons could transfer to the CB of BiOCl, leaving the holes on the VB of BiOI. Thus, photogenerated carriers could be effectively

separated. The EVB of BiOI (2.11 eV) was negative than the potential of ·OH/H2O (2.27 eV), so •OH was generated by OH<sup>−</sup> (E•OH/OH<sup>−</sup> <sup>=</sup> 1.99 eV) rather than H2O (Zeng et al., 2016). Compared to the potentials of O2/•O − 2 (−0.046 eV), electrons in the BiOI/BiOCl could reduce O<sup>2</sup> to •O − 2 , followed by the generation of •OH (Wang et al., 2013; Zeng et al., 2016). In this way, h+, •O − 2 and •OH oxidized the organic compounds. which played an important role in the degradation process.

# CONCLUSIONS

In conclusion, BiOI/BiOCl films were successfully prepared using a facile method at room temperature. The growth process studies indicated that there was a hydrolyzation competition between BiOCl and BiOI in synthesis protocol. 30%BiOI/BiOCl could eliminate more than 99% of RhB within 90 min, which was 12

# REFERENCES

Baeck, S. H., Choi, K. S., Jaramillo, T. F., Stucky, G. D., and McFarland, E. W. (2003). Enhancement of photocatalytic and electrochromic properties of electrochemically fabricated mesoporous times higher than that of BiOCl. Besides RhB, 30% BiOI/BiOCl also showed a great photocatalytic performance toward MB. When degrading RhB, the efficiency of 30% BiOI/BiOCl was 5 and 12 times higher than that of pristine BiOI and BiOCl respectively. While degrading MB, 30%BiOI/BiOCl showed 6 times higher efficiency than that of pristine BiOI. These excellent enhancements were attributed to extended visible light region and high separation efficiency of charge. Five recycles indicated the as-prepared film exhibited a great reusability. In general, this work provided not only an easy and facial method to gain BiOI/BiOCl film but also insights for preparing photocatalysts which effectively utilized visible light with excellent reusability.

# AUTHOR CONTRIBUTIONS

YmZ, YZ, and HC participated in the design of this study; XY provided assistance for SEM, TEM, and data analysis; YxZ and YL carried out the experiments and drafted the manuscript. All the authors read and prove the final manuscript.

# ACKNOWLEDGMENTS

This work was supported by National Natural Science Foundation of China (21706091 and 21276104) and GuangDong Provincial Department of Science and Technology Application Research and Development Supporting Special Fund Project (2015B020235007).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2018.00058/full#supplementary-material

Figure S1 | XRD patterns of 30% BiOI/BiOCl films: fresh and after 5 cycles.

Figure S2 | FT-IR spectra of samples: BiOCl, 30% BiOI/BiOCl, BiOCl.

WO<sup>3</sup> thin films. Adv. Mater. 15, 1269-1273. doi: 10.1002/adma.200 304669

Cao, J., Xu, B., Luo, B., Lin, H., and Chen, S. (2011). Novel BiOI/BiOBr heterojunction photocatalysts with enhanced visible light photocatalytic properties. Catal. Commun. 13, 63–68. doi: 10.1016/j.catcom.2011.06.019


**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 © 2018 Zhong, Liu, Wu, Zhu, Chen, Yu and Zhang. 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 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 Degradation of 4-Nitrophenol by C, N-TiO2: Degradation Efficiency vs. Embryonic Toxicity of the Resulting Compounds

Oluwatomiwa A. Osin1,2, Tianyu Yu1,2, Xiaoming Cai <sup>3</sup> , Yue Jiang<sup>1</sup> , Guotao Peng<sup>1</sup> , Xiaomei Cheng<sup>4</sup> , Ruibin Li <sup>5</sup> , Yao Qin<sup>4</sup> \* and Sijie Lin1,2 \*

*<sup>1</sup> College of Environmental Science and Engineering, State Key Laboratory of Pollution Control and Resource Reuse, Shanghai Institute of Pollution Control and Ecological Security, Biomedical Multidisciplinary Innovation Research Institute, Shanghai East Hospital, Tongji University, Shanghai, China, <sup>2</sup> UN Environment-Tongji Institute of Environment for Sustainable Development, Tongji University, Shanghai, China, <sup>3</sup> Center for Genetic Epidemiology and Genomics, School of Public Health, Jiangsu Key Laboratory of Preventive and Translational Medicine for Geriatric Diseases, Medical College of Soochow University, Suzhou, China, <sup>4</sup> Institute for Translational Nanomedicine, Shanghai East Hospital, Institute for Biomedical Engineering & Nano Science, Tongji University School of Medicine, Shanghai, China, <sup>5</sup> School for Radiological and Interdisciplinary Sciences (RAD-X), Jiangsu Provincial Key Laboratory of Radiation Medicine and Protection, Medical College of Soochow University, Suzhou, China*

### Edited by:

*Zhimin Ao, Guangdong University of Technology, China*

### Reviewed by:

*Yuanming Zhang, Jinan University, China Haiyuan Zhang, Changchun Institute of Applied Chemistry (CAS), China*

### \*Correspondence:

*Yao Qin lilyqin@tongji.edu.cn Sijie Lin lin.sijie@tongji.edu.cn*

### Specialty section:

*This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry*

> Received: *31 March 2018* Accepted: *14 May 2018* Published: *04 June 2018*

### Citation:

*Osin OA, Yu T, Cai X, Jiang Y, Peng G, Cheng X, Li R, Qin Y and Lin S (2018) Photocatalytic Degradation of 4-Nitrophenol by C, N-TiO*2*: Degradation Efficiency vs. Embryonic Toxicity of the Resulting Compounds. Front. Chem. 6:192. doi: 10.3389/fchem.2018.00192* The photocatalytic activity of TiO<sup>2</sup> based photocatalysts can be improved by structural modification and elemental doping. In this study, through rational design, one type of carbon and nitrogen co-doped TiO<sup>2</sup> (C, N-TiO2) photocatalyst with mesoporous structure was synthesized with improved photocatalytic activity in degrading 4-nitrophenol under simulated sunlight irradiation. The photocatalytic degradation efficiency of the C, N-TiO<sup>2</sup> was much higher than the anatase TiO<sup>2</sup> (A-TiO2) based on absorbance and HPLC analyses. Moreover, using zebrafish embryos, we showed that the intermediate degradation compounds generated by photocatalytic degradation of 4-nitrophenol had higher toxicity than the parent compound. A repeated degradation process was necessary to render complete degradation and non-toxicity to the zebrafish embryos. Our results demonstrated the importance of evaluating the photocatalytic degradation efficiency in conjunction with the toxicity assessment of the degradation compounds.

Keywords: 4-nitrophenol, mesoporous C, N-TiO2 , photocatalytic degradation, intermediate compounds, embryonic toxicity

# INTRODUCTION

Photocatalytic degradation of organic pollutants is considered as an efficient, clean, and cost-effective alternative for contaminated water treatment (Wang et al., 1999; Kiros et al., 2013; Osin et al., 2017). Among various types of nano-photocatalysts, titanium dioxide (TiO2) based photocatalysts are the most popular ones owing to their low cost, simple but reliable synthesis methods, resistance to photo-corrosion, and chemical stability (Carp et al., 2004; Herrmann et al., 2007; Rajeshwar et al., 2008; Akpan and Hameed, 2009). TiO<sup>2</sup> has been used to perform photocatalytic oxidation of contaminants in both aqueous and air media (Fox and Dulay, 1993; Hoffmann et al., 1995; Linsebigler et al., 1995). However, due to their large bandgap (3.2 eV), TiO<sup>2</sup> requires UV light activation (<387 nm), which means only a small portion of the solar spectrum could be used for photocatalytic applications (Wen et al., 2011).

**34**

To increase the utility of the visible light spectrum, one common strategy to improve the efficiency of TiO<sup>2</sup> based photocatalysts is to dope or co-dope metal or non-metal elements into the crystalline structure (Asahi et al., 2001). Doping with non-metals, such as nitrogen (Asahi et al., 2001; Burda et al., 2003; Diwald et al., 2004; Cheng et al., 2012), carbon (Lettmann et al., 2001; Sakthivel and Kisch, 2003; Dong et al., 2011), fluorine (Li et al., 2005a,b), and sulfur (Ohno et al., 2004; Yu et al., 2005) has proven to be effective to narrow the bandgap. Among them, nitrogen-doped TiO<sup>2</sup> (N-TiO2) resulted a considerable increase of the photocatalytic activity in the visible light spectrum. Carbon doping can make TiO<sup>2</sup> sensitive to visible light with improved adsorption capacity for organic pollutant molecules (Park et al., 2006; Dong et al., 2011). Moreover, the simultaneous doping of two types of atoms into TiO2, such as N–F (Li et al., 2005a,b; Giannakas et al., 2013; Wang et al., 2014), N–S (Yu et al., 2006), and C–N (Wang and Lim, 2010, 2011; Hassan et al., 2014; Trevisan et al., 2014), has recently attracted considerable interests, for the possible integrated or synergistic effects by the co-doped elements.

Meanwhile, it is important to note that a complete degradation of organic pollutants by photocatalysts was difficult to achieve (Dong et al., 2015). And the intermediate compounds generated during photocatalytic degradation might render higher toxicity than their parent compound (Dong et al., 2015; Li et al., 2016). Li et al. previously reported that transformation products (TPs) of acesulfame generated by the photocatalytic processes using TiO<sup>2</sup> photocatalyst were even more toxic than the parent compound (Li et al., 2016). These results were similar to a study carried out by Sang et al. reporting the successful and substantial photocatalytic degradation of acesulfame and sucralose under simulated natural UV conditions. However, real-time observation revealed that the transformation of acesulfame led to a collection of more persistent by-products that were 500 times more toxic than the parent compound inducing significantly elevated toxicity in both marine bacteria and zebrafish embryos (Sang et al., 2014). Therefore, while evaluating the effectiveness of photocatalytic degradation, it is also important to consider the potential hazardous effects exerted by the intermediate degradation compounds (Li et al., 2017; Osin et al., 2017).

Against this background, we set out to explore the use of C and N co-doping to create a TiO<sup>2</sup> based photocatalyst with higher catalytic and degradation efficiency toward 4-nitrophenol, a representative organic pollutant. Under simulated sunlight irradiation, the as-synthesized C, N-TiO<sup>2</sup> showed much higher catalytic activity in comparison with the anatase TiO<sup>2</sup> (A-TiO2) Using model organism zebrafish embryos, we showed the intermediate compounds of 4-nitrophenol were more toxic than the parent compound. And repeated photocatalytic degradation process could render non-toxicity as a result of a complete degradation of 4-nitrophenol.

# MATERIALS AND METHODS

# Materials and Reagents

In this study, different concentrations of reagents (**Table 1**) were used to obtain the best doping element ratios and concentrations, TABLE 1 | Concentration variations of reagents.


\**Optimum concentrations which produced the best activity and selectivity of the catalyst.*

which determined the catalytic activity of the as-synthesized particles. 4-nitrophenol of 98% purity was used as a representing pollutant in this study. In the preparation of the modified TiO2, titanium tetrachloride (TiCl4, 99.9%), ammonia (NH3.H2O, 28– 30%) and citric acid (C6H8O7, 95.5%) were used as titanium (Ti), nitrogen (N) and carbon (C) source. Perchloric acid (HClO4, 70–72%) was added to serve as a pore-making agent during the synthesis process based on Equation 1:

$$24\text{HClO}\_4 \rightarrow 2\text{H}\_2\text{O} + 7\text{O}\_2 \stackrel{\sim 300^\circ \text{C}}{\uparrow} + 2\text{Cl}\_2 \stackrel{\sim 300^\circ \text{C}}{\uparrow} \tag{1}$$

All reagents were purchased from Aladdin Reagent Co., Shanghai, China and were of analytical (AR) grade, used without further purification.

# Synthesis and Physicochemical Characterizations of C, N-TiO<sup>2</sup>

The C, N-TiO<sup>2</sup> was synthesized by adding TiCl<sup>4</sup> carefully into citric acid solution and stirred for 30 min. A white suspension was obtained after ammonia (NH3) solution was added into the solution dropwise. The mixed solution was stirred for another 30 min in an ice bath. After HClO<sup>4</sup> was added, the reaction temperature was increased to 100◦C while stirring. The slurry obtained was dried in a vacuum oven for 5 h. Thereafter, the light yellow powder was collected and then annealed at 450◦C for 2 h with a ramp of 5◦C /min in air. The powder obtained was grinded and used for the further analyses.

The size and crystallinity of the as-synthesized C, N-TiO<sup>2</sup> were determined by transmission electron microscope (JEM-2100, JEOL Ltd., Japan) and X-ray diffractometer (Bruker D8 Advanced XRD, Bruker Co., Germany), respectively. Chemical compositions and oxidation states of the sample were analyzed using X-ray photoelectron spectroscopy (XPS) with monochromated Al Kα radiation (hν = 1486.6 eV). Binding energies were calibrated using C 1s peak of C-C bond set at 284.8 eV. The fitting and analysis of the spectra was performed in XPS PEAK version 4.1. N<sup>2</sup> adsorption-desorption isotherms were measured on Micromeritics ASAP 2,460 device at 77K. Before detection, the sample was degassed at 373K.

# Photocatalytic Degradation

The photocatalytic activities of the C, N-TiO<sup>2</sup> and commercial anatase-TiO<sup>2</sup> (A-TiO2) were carried out in a photoreactor coupled with a simulated sunlight source (Shanghai Deyangyibang Instruments CO., LTD, DY-D type, also shown in Figure S1). Amount of 20 mL of 4-nitrophenol at 7.0 × 10−<sup>2</sup> mM was transferred into glass tubes and placed orderly in the photo-reactor. Amount of 10 mg of the C, N-TiO<sup>2</sup> and A-TiO<sup>2</sup>

were added, and the adsorption process of all experiments was done in the dark for 2 h. A magnetic stirrer was used to achieve a satisfactory suspension of the photocatalyst and the homogeneity of the reacting mixture. Samples at 30 min intervals were withdrawn, centrifuged at 8,000 rpm for 10 min, and used for further analysis. The percent of 4-nitrophenol removal was analyzed using a UV-vis spectroscopy, measuring the absorbance at 317nm, while observed first-order rate constants (k, min−<sup>1</sup> ) were measured from Equations (2) and (3), respectively.

$$(4-nitrophend \, removal \, (\%) = \frac{co-ct}{co} \times 100 \, \tag{2}$$

$$
\ln\left(\frac{co}{ct}\right) = kt \tag{3}
$$

Where, Co and Ct are the initial and remaining 4-nitrophenol concentrations at a given time of the reaction; k is the observed first-order rate constants.

The degradation efficiency of 4-nitrophenol were also measured by high performance liquid chromatography (HPLC). HPLC analysis was performed on an Agilent 1,100 Series LC system (Agilent technologies, Santa Clara, CA) equipped with a binary pump and an ultraviolet-visible diode array detector to quantify the degradation. Samples were separated on an Eclipse Plus C18 reversed phase HPLC column (2.1 × 150 mm, 100Å, 3.5µm, Agilent technologies, Santa Clara, CA) with an isocratic gradient of methanol and 5 mM phosphoric acid solution (35:65, v/v) over 15 min at a flow rate of 200 µL/min. Injection volumes were 5 µL and the detection wavelength was 254 nm. The mineralization efficiency was measured by a total organic carbon analyzer (TOC-L-CPH Shimadzu, Japan).

# Zebrafish Embryo Toxicity Testing

The AB wild-type adult zebrafish (Danio rerio) were maintained at 28 ± 0.5◦C on a 14 h:10 h light/dark cycle in a fish breeding circulatory system (Haisheng, Shanghai, China) and were fed twice daily with live brine shrimps (Artemia salina). Two pairs of male/female fish were placed in a single mating box separated by a divider 1 day prior to spawning. Spawning was triggered by removing the divider in the morning and the embryos were collected 2 h afterwards. Using a stereomicroscope (Olympus-SZ61, Olympus Ltd., Japan), healthy and fertilized embryos at 4 h post fertilization (hpf) were selected and placed in Ubottom 96-well plates (Costar-3599, Corning, US), with one embryo per well. Each well was then filled with 200 µL of each concentration series of the test-sample suspension (4 nitrophenol, C, N-TiO2, A-TiO<sup>2</sup> and the degradation products) as well as H-buffer as negative controls. Three replicates were carried out for each treatment, each using 12 embryos. The developmental status of the zebrafish embryos was observed at 24, 48, and 72 hpf. The toxicological endpoints included hatching interference, phenotypic abnormalities and mortality (necrosis of the embryos). All experiments were carried out in accordance with the Animal Ethics Committee at Tongji University, with protocol approved by the Animal Center of Tongji University (Protocol #TJLAC-018-020).

### Statistical Analysis

All treatments were performed with at least three replicates. Data were reported as average ± standard deviations. Statistical analysis was carried out by Student's t-test to evaluate the statistical significant differences of the hatching success rate and mortality rate between the treatment groups and the negative control group. P < 0.01 was considered statistically significant between experimental and control groups.

# RESULTS AND DISCUSSION

# Composition and Chemical State Analysis of C, N-TiO<sup>2</sup>

**Figure 1A** shows the XPS survey spectrum of C, N-TiO2. It was clear that only It, O, N and C elements were detected and the corresponding atomic proportions were 29.8, 53.88, 1.25, and

15.07%, respectively. N and C were assigned to the doping species and adventitious carbon of the apparatus, respectively. The highresolution XPS spectra of Ti2p, C1s, N1s, and O1s region were shown in **Figures 1B–D** and Figure S2.

**Figure 1B** shows the nitrogen region of the XPS measured spectrum for the C, N-doped TiO2. The N1s binding energy peaks were broad and asymmetric in the range of 396–403 eV and centered around 400eV (further divided into two peaks centered at 400.1 and 400.5 eV) which can be attributed to the 1 s electron binding energy of the N atom in the environment of O-Ti-N and N-O-Ti (Chen and Burda, 2004; Cong et al., 2007; Xu et al., 2008; Bellardita et al., 2009). Therefore, the nitrogen doped into TiO<sup>2</sup> mainly existed in the form of substitutional N which resulted in the formation of new energy levels in the forbidden band of TiO<sup>2</sup> and led to the enhancement of photocatalytic activity in the visible range (Cheng et al., 2012).

Deconvolution of C1s region showed a main peak at 284.6 eV corresponding to C-H and can be assigned to the adventitious carbon contamination adsorbed from the ambient (Wu and Wang, 2013) (**Figure 1C**). In addition, the peak observed at 284.6 eV refer to the formation of C–OH and C–N bonds. Consequently, the overlap of the peaks corresponding to both C–OH and C–N bonds allows no distinction between the two ones. However, the peak at 288.6 eV indicates the formation of C = O bond that corresponds to a carbonate species present in C-doped titanium systems and that carbon may substitute some of the lattice titanium atoms forming a Ti–O–C structure (Ren et al., 2007). It has been reported that high amount of carbon led to an increased absorption in the visible region, however, it also enhanced the recombination of the charge carriers. This problem was solved by the synergistic effect from the presence of two (or more) complementary dopants (Ren et al., 2007; Li et al., 2010), leading to an improved photoactivity of TiO<sup>2</sup> in the visible light region, compared to either pure or single-doped TiO2.

In the high resolution spectra of Ti2p spectra shown in **Figure 1D**, peaks at 459.2 and 465 eV corresponds to the spin orbit coupling of Ti2P3/<sup>2</sup> and Ti2P1/2, respectively. It also shows

N-TiO2. The distance between the lattice fringes (red parallel lines) pointed by the red arrows is 0.22 nm.

that only the signals corresponding to Ti4<sup>+</sup> were detected and can be attributed to the binding energy separation between the 2p1/<sup>2</sup> and 2p3/<sup>2</sup> peaks of ∼5.8 eV (Abdullah et al., 2016). Furthermore, the Ti2p XPS peak of un-doped TiO<sup>2</sup> has been reported to appear normally at 459.5 eV (Saha and Tomkins, 1992). Therefore, the nitrogen incorporation shifted the XPS spectrum to a lower binding energy (459.2 instead of 459.5 eV). In correspondence with that of N 1s, the O 1s XPS spectra (shown in Figure S2) also shows a broadening at 531.5 eV which confirms the presence of another type of oxygen due to the more covalent nature of N-TiO<sup>2</sup> (Chi et al., 2007). This might be related to the presence of oxygen and nitrogen from the same lattice units in TiO2, which confirmed the interstitial doped form (Cong et al., 2007; Pelaez et al., 2010).

# Phase Structure, Crystal Structure and Morphologies of C, N-TiO<sup>2</sup>

The representative TEM images in **Figure 2** illustrate the crystal phase as well as the size distribution of C, N-TiO2. The product showed an obvious mesoporous structure caused by the aggregation of C, N-TiO<sup>2</sup> nanograins of 10–20 nm in size, which were also shown in the SEM image of Figure S3. HClO<sup>4</sup> was a novel pore-making agent in our recipe of preparing C, N-TiO2. As shown in Figure S4, without the addition of HClO4, the product looks amorphous with very few pores.

As expected, the porous structure showed a large surface area of 159.903 m<sup>2</sup> /g and an average pore diameter of 3.64 nm (**Figure 3B**). As previously mentioned, since the specific surface area (SSA) is one of the key factors determining the catalytic efficiency of catalysts, it is reasonable to expect that the assynthesized porous C, N-TiO<sup>2</sup> would achieve enhanced catalytic performance in the following degradation of 4-nitrophenol compared to A-TiO2, with the grain size of 20 nm and surface area of 50 m<sup>2</sup> /g.

The crystal structures and particle size of the C, N-TiO<sup>2</sup> were shown in **Figure 3A**. Only the anatase phase was detected

FIGURE 3 | (A) X-ray diffraction (XRD) patterns of C, N-TiO2. (B) Nitrogen adsorption-desorption isotherms and the corresponding pore-size distributions (the inset) of C, N-TiO2.

from the diffraction peaks of the as-synthesized C, N-TiO2. This could be attributed to the alkaline reaction conditions during the synthesis process and the N-doping as well. It has been reported that carbon introduction does not modify the crystal of the TiO<sup>2</sup> (Trevisan et al., 2014).

# Photocatalytic Activity of C, N-TiO<sup>2</sup>

The photocatalytic activity of C, N-TiO<sup>2</sup> and A-TiO<sup>2</sup> in the degradation of 4-nitrophenol from an initial concentration of 7.0 × 10−<sup>2</sup> mM under simulated sunlight is shown in **Figure 4A**. The reactivity of the photocatalysts was represented by the ratio of residual concentration to initial concentration of 4 nitrophenol, C/Co, as a function of irradiation time. C, N-TiO<sup>2</sup> and A-TiO<sup>2</sup> degraded 87 and 65% of 4-nitrophenol in 420 min under simulated sunlight irradiation, respectively. The semi-logarithmic plots of concentration data gave a straight line (**Figure 4B**), indicating that the photocatalytic degradation of 4 nitrophenol can be described by the first-order kinetic model, ln C = –kt + ln Co, with a k constant of 4.87 × 10−<sup>3</sup> min−<sup>1</sup> and 2.53 × 10−<sup>3</sup> for C, N-TiO<sup>2</sup> and A-TiO2, respectively.

However, a complete degradation was not achieved under 420 min irradiation: a challenge reported in several other studies (San et al., 2002; Hassan et al., 2014; Dong et al., 2015; Rezaei-Vahidian et al., 2017). To achieve a complete degradation, we repeated the photocatalytic degradation for multiple runs, under the same condition with freshly prepared catalysts at each repetition. A-TiO<sup>2</sup> was able to achieve almost 100% degradation after 7 times of multiple treatments, while the C, N-TiO<sup>2</sup> demonstrated a higher efficiency and achieved almost complete degradation on the 5th treatment (as shown in **Figure 4C**). Furthermore, according to the total organic carbon quantification, the photocatalytic degradation using C, N-TiO<sup>2</sup> was able to mineralize the 4-nitrophenol to about 40% after 7 h (**Figure 5A**). And after 7 repeated treatments, 4-nitrophenol was mineralized to about 60% (**Figure 5B**).

# Zebrafish Embryo Toxicity Test

To assess the toxic effects of 4-nitrophenol on zebrafish, embryos were treated with different concentrations (1, 5, 10, 15, and 20 mg/L) for 72 h. The hatching and mortality rates were assessed during the exposure period, at 24, 48, 72 hpf. Figure S5 shows that 4-nitrophenol exhibited a concentration-dependent toxicity, exerting a decreased hatching rates and increased mortality. To assess and compare the possible effects of C, N-TiO<sup>2</sup> and A-TiO2, embryos were treated with different concentrations (1, 5, 10, 50, 100, 250, and 500 mg/L) of these samples for 72 h. **Figure 6** shows that no obvious toxicity were found for both C, N-TiO2, and A-TiO<sup>2</sup> at all concentrations tested. These results coincided well with Zhu et al. who found both nano and bulk forms of TiO<sup>2</sup> were non-toxic to zebrafish (Zhu et al., 2008).

Interestingly, the analysis on the intermediate compounds produced by photocatalytic degradation showed much higher toxicity than the parent compound. Embryos treated with intermediate compounds obtained within 1 h of degradation showed a decrease in toxicity as compared to 4-nitrophenol, exhibiting a decrease in mortality rates and an increase in hatching rates of zebrafish embryos. This effect could be attributed to a decrease in concentration of 4-nitrophenol during the photocatalytic process. However, highly significant effects were observed when embryos were exposed to products obtained after 3 h of degradation, possibly due to the formation of more toxic intermediate compounds. Results showed an increase in mortality rate and decrease in hatching rate in embryos treated with degradation products obtained after 3 h irradiation time at all developmental stages (**Figure 7** and Figure S6). In particular, exposure to all intermediate products obtained after 3 h degradation led to 0% hatching rate and 100% mortality rate, respectively (as shown in **Figure 7**).

Comparative experiments showed that the hatching rate and mortality rate of embryos exposed to degradation products by A-TiO<sup>2</sup> was significantly lower than that C, N-TiO2, suggesting that higher degradation efficiency led to the formation of more toxic intermediate compounds, thus causing more severe inhibition of hatch and a higher mortality rate. Figure S7 shows embryos treated with the degradation by-products obtained at 1, 3, and 7 h irradiation time after 72 hpf. We observed five types of malformations: pericardial edema (PE), tail malfunction (TM), bent spine (BS), unhatched dead embryo (UDE) and disintegrated embryo (DE). All samples induced high percentages of these abnormalities and malformations. However, malfunctions including unhatched dead embryo and disintegrated embryo were induced by byproducts obtained at 3 and 7 h irradiation time respectively.

The mortality rates and hatching rates of zebrafish embryos exposed to intermediate products obtained after multiple degradation at 24, 48, and 72 hpf are shown in **Figures 8A,B**. There was a significant decrease in toxicity after the first run of degradation. Exposure to products obtained after multiple degradation had no significant effect on the mortality rate and

hatching rate of embryos, suggesting that all toxic intermediate compounds had been successfully degraded. HPLC analyses were consistent with this statement. As **Figure 9** showed, before degradation, 4-nitrophenol showed a distinct peak at the mark of 12-min elution time. After 3 h irradiation, a decreased of the peak at 12-min mark indicated a decrease in the concentration of 4-nitrophenol while the appearance of two small peaks at the marks of 3 and 5-min indicated the formation of intermediate compounds. With 5 repeated treatment, all peaks disappeared indicated a complete degradation of 4-nitrophenol therefore no toxicity was observed.

According to our HPLC study shown in Figure S8 as well as the analysis result reported by Luo et al. (2013), the intermediate compounds (represented by the peaks appear in the 3-min and 5-min elution time) are likely 4-nitrobenzene-1,2-diol and 4-nitrobenzene-1,3-diol. The compounds were likely the same composition but different concentrations when using A-TiO<sup>2</sup> and C, N-TiO<sup>2</sup> as the catalyst respectively.

# CONCLUSION

In summary, through C and N co-doping, we have synthesized a type of mesoporous C, N-TiO<sup>2</sup> photocatalyst by a simple solgel method combined with calcination process. The modified photocatalyst exhibited a large surface area, pore volume and simple crystal structure of anatase. C, N-TiO<sup>2</sup> showed higher photocatalytic efficiency during the degradation of 4 nitrophenol than pure A-TiO<sup>2</sup> under simulated light irradiation due to the synergistic effect of carbon and nitrogen codoping. However, a complete degradation was not achieved under 420 min and intermediate degradation products of 4 nitrophenol displayed a much higher toxicity in zebrafish embryos. A repeated degradation process was necessary to achieve complete degradation and render the compound non-toxic to the zebrafish embryos. Our results further emphasized the necessity of paying close attention to the toxicity potentials of degradation compounds while performing photocatalytic degradation to ensure comprehensive assessment of the threats and significance of these chemicals to the natural environment.

# REFERENCES


# AUTHOR CONTRIBUTIONS

OO contributed to the design and planning of the work, performed experiments, data analysis and drafted the manuscript. TY contributed in the zebrafish embryonic toxicity assessment and graphical design. XCa contributed to the HPLC analysis. YJ contributed to the photodegradation experiments. XCh contributed to the materials synthesis and data analysis. GP contributed to the toxicity assessment and data analysis. RL, YQ, and SL designed, planned and supervised the research and contributed to the manuscript writing. All authors read and approved the final manuscript.

# ACKNOWLEDGMENTS

This work was supported by the Recruitment Program of 1000plan Youth and NSFC Grant #21607115 and #21777116 awarded to SL. Additional acknowledgement to the MOFCOM Scholarship awarded to OO.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2018.00192/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 © 2018 Osin, Yu, Cai, Jiang, Peng, Cheng, Li, Qin and Lin. 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 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.

# Graphene Oxide/BiOCl Nanocomposite Films as Efficient Visible Light Photocatalysts

Weitian Lin<sup>1</sup> , Xiang Yu<sup>2</sup> , Yi Zhu<sup>1</sup> \* and Yuanming Zhang<sup>1</sup>

<sup>1</sup> Department of Chemistry, Jinan University, Guangzhou, China, <sup>2</sup> Analytical & Testing Center, Jinan University, Guangzhou, China

A novel graphene oxide/BiOCl (GO/BiOCl) nanocomposite film was prepared via a spread coating method. In visible-light photocatalytically degrading Rhodamine B (RhB) experiments, 2 wt% GO/BiOCl could degrade 99% of RhB within 1.5 h and the rate constant was 12.2 times higher than that of pure BiOCl. The degradation efficiency still kept at 80% even after 4 recycles, evidencing the relatively good recyclability. The enhancement was attributed to the improvement of visible light adsorption and charge separation. Holes and superoxide radicals·O − 2 played a major role as reactive species. The values of conduction band and valence band for GO and BiOCl were calculated and a new photocatalytic mechanism of GO/BiOCl nanocomposite was proposed.

### Edited by:

Zhimin Ao, Guangdong University of Technology, China

### Reviewed by:

Zhaohui Wang, University of Newcastle, Australia Daimei Chen, China University of Geosciences, China

> \*Correspondence: Yi Zhu tzhury@jnu.edu.cn

### Specialty section:

This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry

> Received: 15 April 2018 Accepted: 18 June 2018 Published: 24 July 2018

### Citation:

Lin W, Yu X, Zhu Y and Zhang Y (2018) Graphene Oxide/BiOCl Nanocomposite Films as Efficient Visible Light Photocatalysts. Front. Chem. 6:274. doi: 10.3389/fchem.2018.00274 Keywords: graphene, BiOCl, film, visible-light, photocatalytic activity

# INTRODUCTION

The problems of environmental pollution have been becoming a major concern with industrial development, especially water pollution, which severely impacts our lives (Maeda et al., 2006; Zhao et al., 2015; Nie et al., 2018). The photocatalysis technique is often used in dealing with water pollution because of its low cost, chemical stability, and non-toxicity (Dong et al., 2015; Meng and Zhang, 2016; Yang G. et al., 2017; Jing et al., 2018; Wu et al., 2018). The conventional UV-light-driven photocatalytic semiconducting materials, such as TiO<sup>2</sup> and ZnO, can only be inspired by UV light accounting for <5% of sunlight (Liu et al., 2009; Yang et al., 2013; Chen et al., 2014a,b). Therefore, there is a need to exploit new visible-light photocatalysts with excellent photocatalytic performance. Among various photocatalytic materials, the bismuth compounds have attracted considerable attention for its relatively high photocatalytic activity (Li et al., 2014; Chen et al., 2018). As a V-VI-VII ternary semiconductor, bismuth oxychloxide (BiOCl) with a tetragonal crystal structure, consists of a tetragonal [Bi2O2] <sup>2</sup><sup>+</sup> positive slabs interleaved by double negative slabs of Cl atoms, which provides the space large enough to polarize the related atoms and orbits (Long et al., 2015; Li M. et al., 2017) and leads to its relatively high photocatalytic activity (Cheng et al., 2014). Nevertheless, BiOCl has been prevented by low separation of the photogenerated electron–hole pairs and UV-light-driving, which make it difficult for practical applications. So it is urgent to further improve electron–hole pairs separation and its visible light adsorption to achieve a high photocatalytic activity. Methods have been applied to change this situation, such as morphology control (Zhu et al., 2010), crystal facet exposure (Wang D. H. et al., 2012), noble metal doping (Jiang et al., 2013), and so on. Among them, fabricating nanocomposites by hybridizing BiOCl with other materials is a practical way. Li et al. showed that novel BiOI/BiOCl nanocomposite demonstrated notably high photocatalytic activities over methyl orange (MO) and RhB in aqueous solution. The enhanced photocatalytic activities for BiOI/BiOCl composites

were ascribed to the matched conduction band and valence band level and effective separation of electron-hole pairs (Li et al., 2011). Wang et al. prepared the BiOCl-C3N<sup>4</sup> heterojunction photocatalyst with high specific surface areas in a solvent-thermal way, which displayed notably high photocatalytic activity in decomposing MO under visible-light (Wang X. J. et al., 2013). Zhu and his coworkers reported N-doped carbon nanotube-BiOCl using a facile solvothermal method, exhibiting enhanced photocatalytic performance compared with pristine BiOCl for decomposition of RhB under UV-light (Zhu L. et al., 2016). Wang et al. successfully fabricated polyaniline/BiOCl photocatalysts with excellent visible-light photocatalytic activity toward MO. The enhancement of photocatalytic properties could be ascribed to the synergistic effect between BiOCl and polyaniline (Wang Q. et al., 2013).

Graphene, a new type of carbon material with monolayer of sp<sup>2</sup> -hybridized carbon atoms in honeycomb structure, has aroused great attention in electronic, optical, and catalytic fields due to its good conductivity, optical and electrical properties (Berger et al., 2006; Geim, 2009; Xie et al., 2015; Xiong et al., 2015; Zhu S. et al., 2016). Recently, graphene exhibits wide applications in photocatalysis including photocatalytic water-splitting to produce hydrogen and photo-degradation for organic pollutants because of its outstanding mobility of charge carriers and much higher theoretical specific surface area (Xiang et al., 2012a,b; Cao et al., 2014; Feng J. et al., 2017). Many studies have been focused on fabricating GO/semiconductor composites to achieve enhanced visible light photocatalytic activities (Yang et al., 2015). For example, Xu et al. reported an efficient graphene hybridized with ZnO photocatalyst for the improved UV light photocatalytic activity (Xu et al., 2011). Ai and her partners fabricated BiOBrgraphene nanocomposites and investigated the excellent visiblelight photodegradation on gaseous nitrogen monoxide (Ai et al., 2011). Zhang et al. synthesized a P25-graphene nanocomposite using a facile one-step hydrothermal method and observed significant enhancement photocatalytic activities in degradation of MO (Zhang et al., 2010). Better photocatalytic performances may be obtained by combining remarkable properties of GO and BiOCl to form GO/BiOCl nanocomposites. Up to now, there have been only few investigations on graphene oxides (GO) or reduced graphene oxides (rGO)/BiOCl nanocomposites. Tian et al. prepared rGO-BiOCl hybrid materials via a facile solvothermal method. The 0.73% rGO-BiOCl hybrid showed the best photocatalytic degradation performance for RhB (Tian L. et al., 2013). Gao et al. reported chemically bonded graphene/BiOCl composites by a facile chemical-bath method, which exhibited the degradation rate twice as much as that of pure BiOCl over methylbenzene removal under UV-light (Gao et al., 2012). Kang et al. prepared size-controlled rGO-BiOCl with PVP using a hydrothermal method at low temperatures, which showed much higher visible-light photocatalytic activity toward RhB degradation, compared with pure BiOCl (Kang et al., 2015). But only GO/BiOCl nanocomposite powders were reported, which were hard to be separated and recycled. Synthesis of composite films is the most effective way to deal with these problems (Mu et al., 2012). However, there is no report on GO/BiOCl films, which are easy to be separated and recycled.

In our work, GO/BiOCl films were successfully fabricated by a spread coating method at room temperature. The prepared samples were characterized, and the photocatalytic properties were studied by degrading RhB under visible light irradiation. Accordingly, a new photocatalytic mechanism was also presented.

# EXPERIMENTAL

# Synthesis of GO/BiOCl

The GO/BiOCl was synthesized on a FTO in a facile process. 1.5 g BiCl<sup>3</sup> and 0.25 mL HCl were dissolved in 50 mL ethanol by ultrasonication for 30 min. After that, GO was added into the mixture, which was magnetically stirred at room temperature. Few milliliter of mixture was sucked and smeared homogeneously on FTO, and then dried at 80◦C for 30 min. Finally, the FTO was immerged in distilled water, and then dried at 60◦C. The approach of fabricating GO/BiOCl film was shown in **Scheme 1**. Different amounts (0.8, 2, 4, and 5 wt%) of GO were added. For comparison, BiOCl without GO was prepared in the similar way.

# Characterization

The structure and phase characterization of the as-prepared samples were recorded on an X-ray diffractometer (XD-2) at the scanning rate of 8◦min−<sup>1</sup> . The morphologies of the samples were characterized by FE-SEM (Zeiss ULTRA 55) and FE-TEM (JEOL 2010F). EDS spectra were obtained using Bruker/Quanta 200 instrument. The ESCALab250 instrument was utilized to record the XPS spectra, which could be used to analyze the surface properties and chemical statement. The BET surface areas and N<sup>2</sup> adsorption-desorption isotherms of the samples were investigated on a TriSTRA 3000 at 77.3 K. The light absorption of the samples was investigated with a UV-Vis diffuse reflectance spectrum (DRS, Hitachi UV-3010) with BaSO<sup>4</sup> take for a reference. The photoluminescence (PL) spectra were performed by a RF-5301PC fluorescence spectrophotometer.

# Photocatalytic Activity Tests

The photocatalytic properties of the as-fabricated GO/BiOCl were assessed by visible-light degradation of RhB under room temperature (a 350-W Xe lamp with a cut-off filter, λ>420 nm). In the photocatalytic test, the GO/BiOCl photocatalyst was added into a breaker with 100 mL 2.5 mg/L RhB aqueous solution. In order to establish the adsorption-desorption equilibrium, the mixture was stirred for 30 min in dark before the start of photocatalytic experiment. At given intervals, 5 mL suspension was collected and examined by the UV-vis spectrophotometer.

# Photoelectrochemical Measurements

The transient photocurrent responses and electrochemical impedance spectroscopy (EIS) measurements were measured in an electrochemical workstation (SP-150, France). The platinum wire was used as the counter electrode and the saturated Ag/AgCl electrode used as the reference electrode. The photo

electrochemical experiments were conducted in 0.1 M Na2SO<sup>4</sup> electrolyte solution.

# RESULTS AND DISCUSSION

# Characterization of the GO/BiOCl

XRD patterns of GO, pure BiOCl and the as-prepared GO/BiOCl with different GO contents was shown in **Figure 1**. In **Figure 1A**, the peak at 10.6◦ belonged to GO, which was agreed with the reported results (Shin et al., 2009). The diffraction peaks of BiOCl were identical to those of tetragonal BiOCl, which suggested the high purity (**Figure 1B**). However, peaks of GO in the GO/BiOCl samples could not be observed, which owes to the low GO content in GO/BiOCl (Du et al., 2011). It was noticed that the peaks of BiOCl for GO/BiOCl samples obviously increased, suggesting that BiOCl grown on GO adopted a better crystallinity. The similar phenomenon was also observed in the previous report (Gao et al., 2012).

The FT-IR spectra of BiOCl, GO and GO/BiOCl with different GO contents were shown in **Figure 2**. The characteristic peaks of oxygen-containing functional groups of GO were observed, such as C=O stretching vibration at 1,727 cm−<sup>1</sup> from carboxyl or carbonyl groups, C=C skeletal vibration at 1,624 cm−<sup>1</sup> from unoxidized graphitic domains, epoxide C—O—C, or phenolic C— O—H stretching at 1,226 cm−<sup>1</sup> , and C—O stretching vibration at 1,047 cm−<sup>1</sup> from epoxy groups (Szabó et al., 2006; Fu and Wang,

2011; Chen et al., 2012; Wang P. et al., 2013). The peak at 3,420 cm−<sup>1</sup> was ascribed to the absorption of water or O—H groups (Liu et al., 2013). The GO/BiOCl samples had similar spectrum of GO but with lower peak intensity, which indicated the partial reduction of GO (Chen et al., 2012). In all the GO/BiOCl samples, the prominent peaks at about 530 cm−<sup>1</sup> corresponded to Bi—O vibration (Chou et al., 2013). The appearance of the broad absorption at 1,150 cm−<sup>1</sup> referring to Bi—C vibration, suggested a chemical bonding between GO and BiOCl (Tu et al., 2012).

Raman spectroscopy was an important technique to study the structural properties of crystal and carbon materials. **Figure 3**

TABLE 1 | Summarized BET surface areas and catalytic properties of the photocatalysts.


showed the Raman spectra of GO and GO/BiOCl composites. As shown, GO displayed two characteristic peaks at around 1,340 and 1,600 cm−<sup>1</sup> , which were ascribed to the D band and G band of graphite structures. However, in the GO/BiOCl composites, the D band shifted to around 1,334 cm−<sup>1</sup> and G band moved to around 1,602 cm−<sup>1</sup> , which could be attributed to the chemical interaction between GO and BiOCl (Hu et al., 2014). ID/I<sup>G</sup> ratio was used to represent the degree of graphitization. The ID/I<sup>G</sup> intensity of GO was 0.97, but the ID/I<sup>G</sup> intensity ratio increased with the increase of GO content, which suggested the reduction of GO during the procedure (Liu et al., 2014).

The XPS spectra provided the information for chemical state and surface properties of the 2 wt% GO/BiOCl film. According to **Figure 4A**, two strong peaks centered at 159.7 eV and 165.0 eV in 2 wt% GO/BiOCl were ascribed to Bi 4f7/<sup>2</sup> and Bi 4f5/2, suggesting Bi3<sup>+</sup> in the GO/BiOCl (Ai et al., 2011). The peaks of Bi 4f in the 2 wt% GO/BiOCl shifted slightly toward higher binding energies compared with pristine BiOCl, due to the strong interaction between BiOCl and GO (Tian L. et al., 2013). The Cl 2p of XPS spectra were displayed in **Figure 4B**. The peaks located at 198.4 eV and 199.9 eV, corresponded to Cl 2p3/<sup>2</sup> and Cl 2p1/2, which were characteristics of Cl−in GO/BiOCl (Cheng et al., 2013). The XPS spectrum of C1s on GO was shown in **Figure 4C**. The peak at 284.7 eV was ascribed to C-C bond with sp<sup>2</sup> orbital. The peaks located at 286.5 eV and 288.4 eV were attributed to the C-O and C=O suggesting the existence of oxygen-containing functional groups in the GO (Liu et al., 2014). In the spectrum of 2 wt% GO/BiOCl (**Figure 4D**), the peak of C-O and C=O showed lower intensities than those of GO, suggesting the partial reduction of GO, which corresponded to the FTIR spectra (Liu et al., 2014). Besides, a new peak at 281.7 eV appeared, which was related to carburetion, and referred to the existence of Bi-C in 2 wt% GO/BiOCl. The result was in accordance with the FTIR spectrum (Akhavan and Ghaderi, 2009). The O 1s region of XPS spectra for BiOCl and 2 wt% GO/BiOCl were depicted in **Figures 4E,F**. The peak at 532 eV could be ascribed to the Bi-O bond in [Bi2O2] slabs of BiOX layered structure. The peak centered at 530.3 eV was related to

kinetics of RhB degradation under visible light irradiation by pure BiOCl, GO and GO/BiOCl composites.

TABLE 2 | Photocatalytic activities of various filmphotocatalysts on the degradation of RhB.


the hydroxyl groups or water molecules absorbed on the surface of the sample (Liu et al., 2012). The peak of hydroxyl in the 2 wt% GO/BiOCl showed higher intensities than that of BiOCl, indicating more oxygen-containing groups on the GO/BiOCl.

The morphology and structure of 2 wt% GO/BiOCl were evaluated by SEM and TEM. **Figure 5a** showed the nanosheet-like morphology of the BiOCl with the size of 400– 600 nm. **Figure 5b** showed that nanosheet-like BiOCl distribute uniformly onto the framework of GO in GO/BiOCl. **Figure 5c** depicted the TEM image of 2 wt% GO/BiOCl. As shown, the GO sheets were not very flat but displayed wrinkles. The structure of GO was shown in **Figures 5e,f**. A high-resolution TEM (HRTEM) image of BiOCl nanosheet (**Figure 5d**) exhibited the lattice spacing of 0.24 nm corresponding to (003) plane. As shown in **Figures 5g–j**, the signals of element Bi, Cl, O and C were

clearly observed respectively, evidencing that the samples was GO/BiOCl.

# Nitrogen Adsorption-Desorption Analysis

**Figure 6** presented the N<sup>2</sup> adsorption-desorption isotherm and corresponding pore size distribution (PSD) cures of BiOCl and 2 wt% GO/BiOCl. Both of them were of type IV isotherms with a hysteresis loop within the range from 0.4 to 0.9 (P/P0), confirming the mesoporous structure. The BET surface areas of the composites were exhibited in **Table 1**. The SBET of the GO/BiOCl composites were much larger than the pure BiOCl, indicating GO could improve the surface area of the GO/BiOCl composites, which was a significant factor leading to the enhanced photocatalytic activity. As shown in the inset of **Figure 6**, the pore sizes of the samples were found to be at about 2–11 nm.

# Optical Absorption Properties

**Figure 7** revealed the UV-vis DRS of pure BiOCl and GO/BiOCl composites. It was noticed that pure BiOCl showed absorption only in UV region. With the introduction of GO, the absorption intensity in the visible-light region of the GO/BiOCl samples was improved. This result indicated that GO played a major role in utilizing sunlight and worked as an electron reservoir to trap the electrons under irradiation (Wei et al., 2014), leading to better photocatalytic performance consequently. The Eg of the GO/BiOCl could be figured out according to previous report (Xie et al., 2013). As can be seen in **Figure 7B**, the band gap energy of pure BiOCl and 2 wt% GO/BiOCl were estimated to be 3.2 and 2.9 eV. It could be seen that the estimated band gap energy of the 2 wt% GO/BiOCl film was lower than that of pristine BiOCl, suggesting that hybridizing BiOCl with GO could enhance the optical absorption property of BiOCl in visible-light region.

# Photocatalytic Properties

The photocatalytic performance of the as-fabricated GO/BiOCl was studied by visible-light degradation toward RhB in an aqueous solution. As can be seen in **Figure 8A**, GO/BiOCl displayed higher photodegradation efficiency compared to pure BiOCl. Along with the increase of GO content, the degradation efficiency of GO/BiOCl increased at first, but then decreased while the GO content was larger than 2 wt%. So the highest photocatalytic activity of GO/BiOCl was gained with the optimum content of GO located at 2 wt%, which led to 99% visible-light degradation of RhB within 1.5 h. The degradation rate constants (k, h−<sup>1</sup> ) were calculated to be 0.24, 0.41, 2.93, 1.92, 1.84 and 0.52 h−<sup>1</sup> for pure BiOCl, 0.8, 2, 4 wt%, 5 wt% GO/BiOCl and GO respectively (**Figure 8B**). It could be seen that all GO/BiOCl showed much higher k-values compared to pristine BiOCl, and 2 wt% GO/BiOCl was about 12 times that of pristine BiOCl. As GO content further increased, the photocatalytic activity of RhB degradation obviously decreased, suggesting that suitable content of the GO was important to improve the photocatalytic activity.

The photocatalytic activity was also compared with other film photocatalysts (**Table 2**). The as-prepared 2 wt% GO/BiOCl film had a higher photocatalytic activity than those photocatalysts, which indicated that it had great prospect for practical applications in degrading pollutants.

The stability and regeneration of the photocatalyst are important to the practical applications. Due to immobilization, the GO/BiOCl samples could be directly separated from the aqueous solution for next recycle. **Figure 9** depicted the photocatalytic activities of 2 wt% GO/BiOCl for degradation of RhB within four cycles. It could be seen that the visiblelight photodegradation efficiencies of RhB can reach 80%. These results suggested that the GO/BiOCl had excellent stability and regeneration and could be used as an effective photocatalyst in practical application.

# Possible Photocatalytic Mechanism of GO/BiOCl

PL emission spectra have also been regarded as an efficient way to discuss the charge transportation and separation of photocatalysts. As we know, a lower PL intensity correlated with a higher separation efficiency of electron-hole pairs (Tian G. et al., 2013). As was depicted in **Figure 10A**, the emission spectra of pristine BiOCl and 2 wt% GO/BiOCl were similar, but the intensity for 2 wt% GO/BiOCl was lower than pure BiOCl, which implied that 2 wt% GO/BiOCl had a lower recombination rate of photogenerated charge carriers resulting in enhanced photocatalytic activities.

The transient photocurrent responses of BiOCl, GO and 2 wt% GO/BiOCl under visible light irradiation in on-off cycles were investigated. In **Figure 10B**, 2 wt% GO/BiOCl exhibited a significantly increased photocurrent density, which was about much stronger than that of BiOCl and GO, implying that the recombination of photogenerated carriers in 2 wt% GO/BiOCl composite was suppressed (Fan et al., 2016; Feng Y. et al., 2017). **Figure 10C** depicted the corresponding EIS of pure BiOCl, GO and 2 wt% GO/BiOCl. 2 wt% GO/BiOCl exhibited the smallest diameter of the Nyquist circle, demonstrating that the transfer efficiency of photoinduced carriers was improved in 2 wt% GO/BiOCl (Hao et al., 2017). All the PL, transient photocurrent response and EIS results suggested that the separation and transfer efficiency of electron-hole pairs were substantially improved by the addition of GO.

In order to further reveal the photocatalytic mechanism of 2 wt% GO/BiOCl in the photocatalytic degradation process, the trapping experiments of 2 wt% GO/BiOCl were conducted using different scavengers. As shown in **Figure 10D**, isopropanol (IPA), benzoquinone (BQ), and triethanolamine (TEOA) were used as the scavenger for hydroxyl radicals, superoxide radicals and holes (Li et al., 2017a,b,c). On one hand, the photodegradation efficiency of RhB was slightly decreased by adding IPA, which indicated that ·OH was not a major active species. On the other hand, with the addition of benzoquinone or triethanolamine into the system, the photodegradation of RhB decreased obviously suggesting that ·O − 2 radicals and h<sup>+</sup> played a dominant part in the visible-light degradation process. This was different from the conclusions that e<sup>−</sup> was the main active species for GO/BiOX composites in other relative reports.

**Figure 11** showed the Mott-Schottlyplots for BiOCl and GO. The positive slope of the C−<sup>2</sup> -E indicated the expected n-type BiOCl and GO of the film (Kong, 2008; Weng et al., 2013). The extrapolations of the Mott-Schottly plots provided a good approximation of the flat band potential, the value of which were about −0.26 and −0.27 eV for BiOCl and GO respectively. The measured potentials could be converted to the reversible hydrogen electrode scale via the Nernst equation (Yang G. et al., 2017; Yang J. et al., 2017), so the conduction band edges were calculated to be 0.36 and 0.34 eV. Based on the values of conduction band and the band gap, the valence band positions could be calculated to be 3.56 eV and 1.80 eV via ECB = EVB − Eg (Duo et al., 2015), in which Eg was about 3.2 eV (**Figure 7B**) and 1.46 eV (**Figure 11A**) for BiOCl and GO.

According to the above results, a new photocatalytic mechanism of GO/BiOCl nanocomposite including photodegradation and photosensitization at the same time, which was totally different from relevant reports, could be proposed as follows. (1) Holes played a dominant part in the visible-light degrading RhB process, suggesting that photodegradation was an important mechanism. Under visible light irradiation, GO was inspired and gave photo-generated electrons and holes. Electrons flew into the CB of BiOCl, which reacted with O<sup>2</sup> to generate ·O − 2 . ·O − 2 and the holes on the VB of GO degraded the contaminant (**Figure 12A**). (2) Salicylic acid was slightly degraded under visible light irradiation (Figure S1), indicating photosensitization was another important mechanism. Under visible light irradiation, RhB was excited to produce electrons, which flew into the CB of GO, leaving the activated RhB molecules (RhB<sup>∗</sup> ). And the electrons on the CB of

GO would further transfer to BiOCl, which reacted with O<sup>2</sup> to generate ·O − 2 , and degraded RhB<sup>∗</sup> directly (**Figure 12B**). During the photosensitization process, holes are not detected to be a major active species.

# CONCLUSION

In summary, GO/BiOCl composite films were successfully prepared via a facile spread coating method. The GO/BiOCl films especially 2 wt% GO/BiOCl exhibited a much higher photocatalytic activity compared with pristine BiOCl and many other film photocatalysts, which could be mainly ascribed to the improved light adsorption and separation of photoinduced electrons and holes both resulting from GO. A new photocatalytic mechanism totally different from relevant reports, was revealed to include photodegradation and photosensitization at the same time. In addition, their recycle was much easier and realizable owing to the immobilization of GO/BiOCl on FTO, and excellent recyclability under visible light was obtained. This work could provide a new way to developing novel composite film photocatalysts and studying new photocatalytic mechanisms for similar GO/BiOX nanocomposites.

# AUTHOR CONTRIBUTIONS

WL conceived and designed the experiments, performed the experiments, analyzed the data, and drafted the manuscript. WL and YiZ revised the manuscript. XY and YuZ contributed significantly to the manuscript preparation.

## REFERENCES


# ACKNOWLEDGMENTS

This work was supported by National Natural Science Foundation of China (21706091) and GuangDong Provincial Department of Science and Technology Application Research and Development Supporting Special Fund Project (2015B020235007).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2018.00274/full#supplementary-material


route and its enhanced photocatalytic activity. J. Nanoparticle Res. 15:1917. doi: 10.1007/s11051-013-1917-6


adsorption-enrichment and photocatalytic degradation. Appl. Catalysis B Environ. 205, 228–237. doi: 10.1016/j.apcatb.2016.12.035


**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 © 2018 Lin, Yu, Zhu and Zhang. 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.

# Adsorption and Photocatalytic Processes of Mesoporous SiO2-Coated Monoclinic BiVO<sup>4</sup>

Duangdao Channei 1,2 \*, Auppatham Nakaruk 3,4, Wilawan Khanitchaidecha4,5 , Panatda Jannoey <sup>6</sup> and Sukon Phanichphant <sup>7</sup>

*<sup>1</sup> Department of Chemistry, Faculty of Science, Naresuan University, Phitsanulok, Thailand, <sup>2</sup> Research Center for Academic Excellence in Petroleum, Petrochemicals and Advanced Materials, Naresuan University, Phitsanulok, Thailand, <sup>3</sup> Department of Industrial Engineering, Faculty of Engineering, Naresuan University, Phitsanulok, Thailand, <sup>4</sup> Centre of Excellence for Innovation and Technology for Water Treatment, Naresuan University, Phitsanulok, Thailand, <sup>5</sup> Department of Civil Engineering, Faculty of Engineering, Naresuan University, Phitsanulok, Thailand, <sup>6</sup> Department of Biochemistry, Faculty of Medical Science, Naresuan University, Phitsanulok, Thailand, <sup>7</sup> Materials Science Research Center, Faculty of Science, Chiang Mai University, Chiang Mai, Thailand*

### Edited by:

*Hongqi Sun, Edith Cowan University, Australia*

### Reviewed by:

*Priyabrat Mohapatra, C. V. Raman College of Engineering, India M. Nageeb Rashed, Aswan University, Egypt*

> \*Correspondence: *Duangdao Channei duangdaoc@nu.ac.th*

### Specialty section:

*This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry*

Received: *09 March 2018* Accepted: *23 August 2018* Published: *19 September 2018*

### Citation:

*Channei D, Nakaruk A, Khanitchaidecha W, Jannoey P and Phanichphant S (2018) Adsorption and Photocatalytic Processes of Mesoporous SiO*2*-Coated Monoclinic BiVO*4*. Front. Chem. 6:415. doi: 10.3389/fchem.2018.00415* The silicon dioxide (SiO2)–coated bismuth vanadate (BiVO4) composites as visible–driven–photocatalysts were successfully synthesized by the co–precipitation method. The effects of SiO<sup>2</sup> coating on the structure, optical property, morphology and surface properties of the composites were investigated by X–ray diffraction (XRD), UV–visible diffuse reflectance spectroscopy (DRS), transmission electron microscopy (TEM) and Brunauer–Emmette–Teller (BET) measurements. The photocatalytic activity of monoclinic BiVO<sup>4</sup> and BiVO4/SiO<sup>2</sup> composites were evaluated according to the degradation of methylene blue (MB) dye aqueous solution under visible light irradiation. The SiO2−coated BiVO<sup>4</sup> composites showed the enhancing photocatalytic activity approximately threefold in comparison with the single phase BiVO4.

Keywords: composite materials, photocatalysis, BiVO4 , BET isotherms, SiO2

# INTRODUCTION

Nowadays, the advanced oxidation process is known as an effective method for water purification and wastewater treatment. One of the most famous advanced oxidation process is heterogeneous photocatalysis; the contaminant (i.e., organic compounds) containing in the water and wastewater is finally degraded to carbon dioxide (Legrini et al., 1993; Mukherjee and Ray, 1999). This process can remove the organic contaminant perfectly and does not generate the second contaminant (i.e., sludge and other organic compounds) which are required the further treatment and disposal. According to the heterogeneous photocatalysis, the titanium dioxide (TiO2) has been played a role as the important catalyst to promote the photocatalytic activity. Due to its wide band gap of 3.2 eV, the photocatalyst of TiO<sup>2</sup> is typically activated under the UV light (the wavelength <390 nm is required), which accounts for 45–50% of solar radiation (Linsebigler et al., 1995; Bahnemann et al., 2007; Devipriya et al., 2012). This theoretical fact becomes the limitation and non-cost-effectiveness of actual photocatalytic system for purifying the water at the site.

Another catalyst of monoclinic bismuth vanadate (BiVO4) has been proposed to overcome the drawback of photocatalytic system using TiO<sup>2</sup> and together with enhance the photocatalytic activity during implementation. Since BiVO<sup>4</sup> has narrow band gaps of 2.4 to 2.8 eV (Kudo et al., 2001; Xie et al., 2006; Li et al., 2008), this photocatalyst can be activated by the visible light and consequences the effective use of solar energy. However, the low specific surface area and poor surface textural property are the significant disadvantages of using BiVO<sup>4</sup> as the catalyst. Its low surface area and adsorption capacity cause the low efficiency of photocatalytic system for organic contaminant removal and also the long treatment period required. Therefore, the increase in specific surface area of BiVO<sup>4</sup> catalyst is necessary prior to imply the photocatalytic system to the actual wastewater.

Recently, alternative composite materials have been synthesized by combining metal oxide and porous materials (i.e., alumina, silica, zeolites, carbon black, charcoal) (Belessi et al., 2007; Wang et al., 2012; Xing et al., 2016) with the aim of improving the specific surface area, pore structure, and photocatalytic activity of catalysts (Gan et al., 2003; Kimura et al., 2003). For example, the enhancement of Ag–doped TiO<sup>2</sup> photocatalytic activity was suggested by adding the mesoporous SiO2; the excellent efficiency of methyl orange (MO) removal was achieved by 2.5 h (Roldan et al., 2015). The increasing adsorption capacity of TiO<sup>2</sup> catalyst was observed when the catalyst was combined with SiO2; the adsorption capacity was increased (Hu et al., 2012). The SiO<sup>2</sup> addition also enhance the separation rate of electron–hole pairs under UV excitation. Further, the deposition of gold nanoparticles (Au) on the porous SiO2-WO<sup>3</sup> composite can enhance the methylene blue (MB) adsorption capacity; the adsorption capacity of Au–SiO2-WO<sup>3</sup> was greater than SiO2-WO<sup>3</sup> and WO<sup>3</sup> respectively (DePuccio et al., 2015). The complete MB removal was achieved by 300 min under visible light, and the fast kinetic of MB removal was found in Au–SiO2-WO<sup>3</sup> catalyst, following by Au–WO<sup>3</sup> and WO<sup>3</sup> catalysts.

As all the above mentions, this study aimed to improve the surface morphology and photocatalytic activity of BiVO<sup>4</sup> catalyst by coating SiO2. Various analytical techniques including X– ray diffraction (XRD), transmission electron microscopy (TEM), Brunauer–Emmett–Teller (BET) and UV–vis diffuse reflectance spectra (DRS) were used to clarify the better property of BiVO4/SiO<sup>2</sup> composite rather than BiVO<sup>4</sup> and SiO2. Further, the performance of BiVO4/SiO<sup>2</sup> composites on wastewater

treatment was preliminary studied in the batch test under visible light irradiation, and its performance was compared to the other two materials of BiVO<sup>4</sup> and SiO2.

# EXPERIMENTAL PROCEDURE

All chemicals used were of analytical grade and were used as received without any further purification. The chemicals including tetraethyl orthosilicate (TEOS), bismuth (III) nitrate pentahydrate [Bi(NO3)3·5H2O], ammonium metavanadate (NH4VO3), methylene blue powder, sodium hydroxide pellet (NaOH), ammonia solution (28%) and nitric acid (37% HNO3) were obtained from Sigma-Aldrich. All solutions were prepared with deionized water.

# Preparation of SiO<sup>2</sup> Particles

SiO<sup>2</sup> particles were prepared by the sol–gel method. Ammonia solution (28%) was added in 100 mL of a mixed solution of absolute ethanol/DI water (80: 20 v/v) and stirred under ultrasonic dispersion for 60 min. Then, 20 mL of tetraethyl

orthosilicate (TEOS) was added drop by drop to the mixed solution and stirred for 120 min at room temperature. After the reaction was homogenized, the fine particles were separated by centrifugation with typical rotating speed of 6,000 rpm for 15 min, washed by DI water and dried at 80◦C for 24 h in a hot air oven. Fine particles of SiO<sup>2</sup> were obtained as a white powder following heat treatment at 500◦C for 1 h in ambient.

# Preparation of Monoclinic BiVO<sup>4</sup> and SiO2-Coated BiVO<sup>4</sup> Composites

Monoclinic BiVO<sup>4</sup> were obtained by the co–precipitation method. Firstly, 12 mmol of bismuth (III) nitrate pentahydrate [Bi(NO3)3·5H2O] and the same volume of ammonium metavanadate (NH4VO3) were dissolved in 100 mL of 2 M nitric acid (HNO3) under vigorous stirring. The pH of the mixed solution was adjusted to 9 by adding 3 M sodium hydroxide (NaOH). The yellow precipitate was then separated by centrifugation at 6,000 rpm for 15 min, washed thoroughly with distilled water and ethanol and finally dried in a hot air oven at <sup>80</sup>◦C for 24 h. Crystalline monoclinic BiVO<sup>4</sup> was formed after calcination at 550◦C for 4 h.

BiVO4-coated SiO<sup>2</sup> composites were also prepared by the same method for comparison with an additional step of adding SiO<sup>2</sup> powder to 100 mL of 2 M HNO3.

# Photocatalytic Reaction

Photocatalytic activities of the BiVO4, SiO<sup>2</sup> and BiVO4/SiO<sup>2</sup> composites were evaluated through degradation of methylene blue (MB) dye as a model organic pollutant under visible light. A total of 0.20 g of photocatalyst was added to 100 mL MB aqueous solution (initial concentration C<sup>0</sup> = 20 ppm) under magnetic stirring in darkness for 60 min to achieve adsorption– desorption equilibrium. The system was irradiated by three 18 W halogen lamps (Essential MR, Philips, Thailand) to investigate photocatalytic degradation. Reduction of MB concentration over time (Ct) was recorded every 15 min by measuring the intensity change of the characteristic absorption peak at 664 nm using UV–vis double beam spectroscopy (UV−6100, Mapada).

# Characterisation

Crystal phase and structure of the prepared samples were characterized by powder X–ray diffraction (XRD, Philips X'Pert MPD) using Cu K<sup>α</sup> (λ = 1.54056 Å) radiation. Morphological changes in the composite materials were monitored by transmission electron microscopy (TEM, JSM−2010, JEOL). Brunauer–Emmett–Teller (BET) measurements (Adtosorb 1 MP, Quantachrome) were performed to compare the specific surface area of the BiVO<sup>4</sup> and BiVO4/SiO<sup>2</sup> composites. Measurement of UV–vis diffuse reflectance spectroscopy (DRS UV–vis, Shimadzu UV−3101PC) was carried out at room temperature to detect reflectance and absorbance spectra.

# RESULTS AND DISCUSSION

In **Figure 1**, the broad XRD peak at 2θ = 22–23◦ corresponded to the amorphous SiO2. The XRD pattern of BiVO<sup>4</sup> without SiO<sup>2</sup> was assigned to the standard monoclinic BiVO<sup>4</sup> (JCPDS no. 14–0688) (Gotic´ et al., 2005). After coating BiVO<sup>4</sup> with

Channei et al. Mesoporous SiO2–Coated Monoclinic BiVO4

SiO2, the diffraction peaks matched well with the pure phase monoclinic BiVO<sup>4</sup> and no peaks of any other phases or impurities were recorded. However, the diffraction intensity of BiVO<sup>4</sup> decreased after coating SiO2, because the amorphous substance had the negative effect on crystallinity. Alternatively, self–doped Si4<sup>+</sup> ions in the BiVO<sup>4</sup> crystal structure might cause the decreasing crystallinity of BiVO4/SiO<sup>2</sup> composites, and resulted in the broader peaks of the composite samples, which are similar to those reported by Phanichphant et al. (2016) for the binary composite CeO2/SiO<sup>2</sup> photocatalyts and Kumar et al. for TiO2/SiO<sup>2</sup> nanocomposites in solar cell applications (Arun Kumar et al., 2012).

As shown in **Figure 2a**, the BiVO<sup>4</sup> demonstrated the absorption edge of the visible region at 450 nm, corresponding to the optical band gap (Eg) of 2.60 eV which was calculated by the Kubelka–Munk function (see **Figure 2b**) (Sirita et al., 2007). Compared to BiVO4/SiO<sup>2</sup> composites, the value of the graph intercept was estimated at 2.30 eV, corresponding to the strong absorption edge in the visible region at 523 nm. The band gap energy of BiVO<sup>4</sup> decreased from 2.60 to 2.30 eV in the composite materials, due to the influence of Si4<sup>+</sup> ions doping into the lattice of BiVO<sup>4</sup> which created the abundant doping energy levels. The estimated band gap values in this study was similar to those of BiVO<sup>4</sup> reported by Jiang et al. (2012), who prepared the BiVO<sup>4</sup> photocatalysts with different morphologies using the hydrothermal method. Liu et al. (2015) observed that the band gap energy of BiVO4/SiO<sup>2</sup> catalyst estimated to be 2.32 eV, which was almost the same as that of calculate by this study.

The TEM images of SiO2, BiVO<sup>4</sup> and BiVO4/SiO<sup>2</sup> composites are presented in **Figure 3**. The SiO<sup>2</sup> image shows the aggregation of spherical–shaped particles with diameters ranging of 20–30 nm (**Figure 3a**), while **Figure 3b** shows the rod–like nanostructures of monoclinic BiVO<sup>4</sup> with the diameter of 10 nm and the length of 60 nm. Typical TEM images are used for characterizing the composite materials and proving the heterojunction formation between BiVO<sup>4</sup> and SiO2, which demonstrated that the rod–like BiVO<sup>4</sup> core was covered by the SiO<sup>2</sup> particles growing on the surface (**Figure 3c**).

The N<sup>2</sup> adsorption-desorption isotherms (**Figure 4a**) show that the N<sup>2</sup> adsorption of BiVO4/SiO<sup>2</sup> composites were relatively higher than that of the pure BiVO4, however the value was much lower than that of the SiO2. The specific surface areas of SiO2, BiVO4/SiO<sup>2</sup> composites, and BiVO<sup>4</sup> were found to be 106.9959, 37.6851, and 19.4964 m<sup>2</sup> /g, respectively. In the meanwhile, the pore size was calculated by using the BJH method, and the results were 9.0316, 11.0776, and 11.8111 nm for SiO2, BiVO4/SiO2, and BiVO<sup>4</sup> respectively (as summarized in **Table 1**). The surface area and pore size are positively related to the photocatalytic activity, therefore the photocatalytic activity of BiVO4/SiO<sup>2</sup> composites were higher than that of pure BiVO4. Even though the surface area of SiO<sup>2</sup> was higher than the BiVO4/SiO<sup>2</sup> composite, the adsorption of pollutant by SiO<sup>2</sup> with high specific surface area have only the ability to transfer pollutants to alternative phases, but not completely get rid of them. Therefore, the photocatalytic process based on using the hydroxyl radicals is required in this study.

TABLE 1 | Surface properties of the prepared samples.


Frontiers in Chemistry | www.frontiersin.org

–ln Ct /C0.

FIGURE 5 | MB concentration changes with irradiation time (A) Ct

with a small hysteresis loop at the relative pressure of 0.80–1.00. This indicated the existence of mesopores in the sample with the pore diameter ranging of 2–50 nm (Brunauer et al., 1940; Bae et al., 2010).

The information from the isotherm can be used to determine the specific surface area from the mathematical relations in Equation (1) and Equation (2) below (Itodo et al., 2010; Thommes et al., 2015)

$$\frac{\text{P}/\text{P}\_0}{\text{V}(1-\text{P}/\text{P}\_0)} = \frac{1}{\text{V}\_\text{m}\text{C}} + \frac{\text{(C}-1)}{\text{V}\_\text{m}\text{C}} \frac{\text{P}}{\text{P}\_0} \tag{1}$$

where,

P0, Initial pressure of N2; P, Equilibrium pressure of N<sup>2</sup> adsorption; Vm, Monolayer capacity; V, Amount of N<sup>2</sup> adsorbed at standard temperature and pressure (STP).

$$\text{Specific Surface area} = \frac{\text{V}\_{\text{m}}\text{N}\_{\text{a}} \times \text{A}}{\text{m} \times 22400} \tag{2}$$

where,

A, Cross-sectional area of the adsorbed N2; m, Adsorbate molecular weight; Na, Avogadro's number.

The intercept and slope of the plot in **Figure 4c** were used to calculate the maximum volume of gas adsorbed at the monolayer (Vm), it was 8.6569 cm<sup>3</sup> /g. The specific surface area was also calculated via the V<sup>m</sup> value (see Equation 2). The result showed that the surface area of BiVO4/SiO<sup>2</sup> composites was 37.6851 m2 /g.

**Figure 5a** presents the degradation efficiency of MB as a function of Ct/C<sup>0</sup> and visible irradiation time. The C<sup>0</sup> was the initial concentration of MB before irradiation and C<sup>t</sup> was the MB concentration at the interval irradiation time (t, min). For using the SiO<sup>2</sup> as catalyst, the MB was removed of 83% under the dark adsorption, and only 5% of MB was further degraded under the visible light. For using the single phase monoclinic BiVO4, the MB was removed around 10% under the dark adsorption, and 40% of MB was further degreased under the visible light irradiation. When the BiVO4/SiO<sup>2</sup> composites was used, the MB removal efficiency reached 35 and 86% under

/C0 and (B)

the dark adsorption and visible light irradiation. As above explanation, the specific surface area of BiVO4/SiO<sup>2</sup> composites were increased from BiVO4, due to the SiO<sup>2</sup> coating. The increasing specific surface area resulted in the high adsorption of MB molecules during 60 min of the darkness, and then the adsorbed MB was continuously degraded by photocatalytic activity during visible light. These results illustrated that the photocatalytic activity of BiVO<sup>4</sup> was enhanced by coating the SiO<sup>2</sup> particles.

The kinetics of MB degradation was analyzed using the pseudo–first order model, which was given in Equation (3) (Yetim and Tekin, 2017). In **Figure 5b**, the correlation of–ln Ct/C<sup>0</sup> and t were positive with linear equation; the kinetic constant (k) were 0.0073 min−<sup>1</sup> for BiVO<sup>4</sup> and 0.0207 min−<sup>1</sup> for BiVO4/SiO<sup>2</sup> composites. The kinetic constant of MB degradation using BiVO4/SiO<sup>2</sup> composites was approximately threefold higher than that using the single phase BiVO4.

$$-\ln(\mathbf{C}\_t/\mathbf{C}\_0) = \text{kt} \tag{3}$$

where k is the apparent rate constant of the pseudo–first order reaction (min−<sup>1</sup> ).

Since the photocatalytic degradation of dyes is associated with dye adsorption onto the surface of BiVO4/SiO2. Furthermore, photocatalytic degradation occurs at or near the surface of the catalyst rather than in the bulk solution. Thus the higher photocatalytic activity of BiVO4/SiO<sup>2</sup> is consistent with the higher adsorption of MB on the surface of BiVO4/SiO<sup>2</sup> photocatalyst. Mesoporous SiO<sup>2</sup> adsorbent enriches the MB molecules around the BiVO<sup>4</sup> surface as shown in **Figure 6** and the visible–light photocatalytic activity of the BiVO<sup>4</sup> interface in the composite materials is then excited to generate electrons (e–) and holes (h+). Subsequently, photoexcited electrons in the valance band and hole in the conduction band of BiVO<sup>4</sup>

## REFERENCES


react with oxygen, water and hydroxide ions to produce free superoxide radicals (O−• 2 ) and hydroxyl radicals (OH• ) as the main active oxidizing species, which then react with MB molecules during the photocatalytic process (Lin et al., 2014; Zhou et al., 2014). The final products of MB aqueous solution photocatalytic degradation are oxidized to CO2, H2O, CO2, NH<sup>+</sup> 4 , NO<sup>−</sup> 3 , and SO2<sup>−</sup> 4 (Houas et al., 2001; Luan and Hu, 2012).

# CONCLUSIONS

BiVO4/SiO<sup>2</sup> composites consisting of spherical SiO<sup>2</sup> particles coated on BiVO<sup>4</sup> nanorods were successfully prepared by co– precipitation. The composites exhibited higher photocatalytic activity compared to single monoclinic BiVO<sup>4</sup> by degrading MB under visible–light irradiation due to the greater surface area of mesoporous SiO2. Fabrication of heterogeneous semiconductors using mesoporous materials can produce promising alternative photocatalysts for wastewater treatment under light irradiation by combining adsorption and photocatalytic processes.

# AUTHOR CONTRIBUTIONS

DC designed and performed the experiments and wrote the manuscript. SP, AN advised the data analysis and edited manuscript. PJ and WK advised the data analysis. All authors reviewed the approved the manuscript.

# ACKNOWLEDGMENTS

This work has been supported by grant from the National Research Council of Thailand (NRCT) under Grant No. R2561B075. The Office of the Higher Education Commission and the Thailand Research Fund (TRF) under Grant No. MRG6080097 is also acknowledged for financial support.


**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 © 2018 Channei, Nakaruk, Khanitchaidecha, Jannoey and Phanichphant. 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.

# Catalytic and Photocatalytic Nitrate Reduction Over Pd-Cu Loaded Over Hybrid Materials of Multi-Walled Carbon Nanotubes and TiO<sup>2</sup>

Cláudia G. Silva\*, Manuel F. R. Pereira, José J. M. Órfão, Joaquim L. Faria and Olívia S. G. P. Soares

Laboratory of Separation and Reaction Engineering-Laboratory of Catalysis and Materials (LSRE-LCM), Faculdade de Engenharia, Universidade do Porto, Porto, Portugal

### Edited by:

Zhimin Ao, Guangdong University of Technology, China

### Reviewed by:

Priyabrat Mohapatra, C. V. Raman College of Engineering, India Miguel Angel Centeno, Instituto de Ciencia de Materiales de

Sevilla (ICMS), Spain

\*Correspondence: Cláudia G. Silva cgsilva@fe.up.pt

### Specialty section:

This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry

Received: 01 August 2018 Accepted: 05 December 2018 Published: 20 December 2018

### Citation:

Silva CG, Pereira MFR, Órfão JJM, Faria JL and Soares OSGP (2018) Catalytic and Photocatalytic Nitrate Reduction Over Pd-Cu Loaded Over Hybrid Materials of Multi-Walled Carbon Nanotubes and TiO2. Front. Chem. 6:632. doi: 10.3389/fchem.2018.00632 TiO<sup>2</sup> and carbon nanotube-TiO<sup>2</sup> hybrid materials synthesized by sol-gel and loaded with 1%Pd−1%Cu (%wt.) were tested in the catalytic and photocatalytic reduction of nitrate in water in the presence of CO<sup>2</sup> (buffer) and H<sup>2</sup> (reducing agent). Characterization of the catalysts was performed by UV-Vis and fluorescence spectroscopy, X-ray diffraction, temperature programed reduction, N<sup>2</sup> adsorption, and electron microscopy. The presence of light produced a positive effect in the kinetics of nitrate removal. Higher selectivity toward nitrogen formation was observed under dark condition, while the photo-activated reactions showed higher selectivity for the production of ammonium. The hybrid catalyst containing 20 %wt. of carbon nanotubes shows the best compromise between activity and selectivity. A mechanism for the photocatalytic abatement of nitrate in water in the presence of the hybrid materials was proposed, based in the action of carbon nanotubes as light harvesters, dispersing media for TiO<sup>2</sup> particles and as charge carrier facilitators.

Keywords: photocatalysis, catalytic reduction, nitrate, titanium dioxide, carbon nanotubes, palladium, copper

# INTRODUCTION

Nitrate is a naturally occurring ion that is part of the nitrogen cycle, in which nitrogen species are switched between organisms and the environment. The increasing use of inorganic nitrogenous fertilizers, the disposal of wastes (mainly from oxidation of nitrogenous waste products in human and animal excreta) and changes in land use are the main causes accounting for the progressively increasing levels of this pollutant in groundwater supplies. Nitrate ion (NO<sup>3</sup> <sup>−</sup>) is the stable form of combined nitrogen for oxygenated systems and is potentially hazardous for humans, since it can be transformed into nitrite in the human body, which may cause the blue baby syndrome, being also a precursor of carcinogenic nitrosamines (Kapoor and Viraraghavan, 1997). Moreover, it may cause the eutrophication of rivers and lakes. The maximum contaminant level in drinking water of nitrogen species as nitrate, nitrite, and ammonium is 50, 0.5, and 0.5 mg/L, respectively (EU Water Framework Directive<sup>1</sup> ).

<sup>1</sup>EU Water Framework Directive (2000). Available online at: http://ec.europa.eu/environment/water/water-framework/ index\_en.html (Accessed 2018).

The removal of nitrate from water constitutes a great challenge to safeguarding drinking water resources of suitable quality. In this context, a great effort has been put in the development of water technologies capable to address the environmental and health concerns. Conventional methods are based in physicalchemical treatment processes (reverse osmosis, ion exchange, and electrodialysis) and biological denitrification.

Heterogeneous catalytic systems like catalytic and photocatalytic reduction of nitrate have been shown to be promising processes compared to conventional treatments (Sá et al., 2009; Soares et al., 2011a, 2014; Luiz et al., 2012; Shand and Anderson, 2013). Catalytic nitrate reduction occurs by consecutive and parallel reactions where nitrate is reduced to nitrite, which is converted to ammonium, as undesired by-product, and to nitrogen, as desired product. The main issue of this process is the selectivity toward nitrogen, which is often compromised. Bimetallic catalysts composed by a noble metal (Pd, Pt, or Rh) and a promoter metal (Cu, Sn, or In) supported on different materials are the most used for catalytic hydrogenation of nitrate (Soares et al., 2009; Calvo et al., 2010; Marchesini et al., 2010; Wada et al., 2012), although monometallic catalysts also present some activity depending on the support (Barrabés et al., 2010; Anderson, 2011; Devadas et al., 2011). Pd-Cu catalysts are normally the most efficient, Pd-Sn and Pd-In also presenting good performances (Martínez et al., 2017). In the case of photocatalytic reduction, besides the type of metal catalysts, several other conditions such as the catalyst support, the pH of the solution, the irradiation source, nature of the reducing agent or electron donor (hole scavenger) also affect the performance of the process. H<sup>2</sup> and CO<sup>2</sup> are usually used in catalytic reduction reactions, in which H<sup>2</sup> serves as reductant and CO<sup>2</sup> as buffer (Prusse et al., 2000). In the case of the photo-assisted catalytic process, oxalic acid, formic acid, or methanol are the most used hole scavengers (Zhang et al., 2005; Doudrick et al., 2013).

Titanium dioxide-based materials are the most widely used catalysts for the photocatalytic reduction of nitrate, the bimetallic Pd-Cu/TiO<sup>2</sup> and Pt-Cu/TiO<sup>2</sup> being the ones showing most promising results. On the other hand, Pd-Cu catalysts supported on carbon materials and also in metal oxides are the most used for nitrate catalytic reduction. TiO2-based catalysts are normally highly active for nitrate removal, yet, due to their capacity to drive hydrogenation, low nitrogen selectivity is reached (Sá et al., 2005). It has been reported that the activity of Pd-Cu bimetallic catalysts supported on carbon materials was higher than the same Pd-Cu catalyst supported on a metal oxide, such as TiO2, Al2O3, SiO2, or ZrO2, at the same operating conditions, due to their surface chemistry and higher metals dispersion (Sakamoto et al., 2006; Soares et al., 2011a). Moreover, the coupling of carbon nanotubes with TiO<sup>2</sup> has proved to induce a positive effect in the activity and selectivity of Pd-Cu catalysts for the catalytic reduction of nitrate into N<sup>2</sup> (Soares et al., 2011a).

Following the previous findings, in the present work we explored the synergies between carbon nanotubes (CNT) and titanium dioxide in both catalytic and photocatalytic reduction of nitrate in water. For that purpose, hybrid materials of TiO<sup>2</sup> with different CNT contents, and loaded with 1%Pd and 1%Cu (wt.%) were evaluated as catalysts. Hydrogen and carbon dioxide were used as reducing and pH buffer agents, respectively. The materials were tested under the same experimental conditions, the only difference between the two processes being the introduction of a near-UV to visible light source in the case of the photocatalytic reactions.

# EXPERIMENTAL

# Catalysts Preparation

Multi-walled carbon nanotubes synthesized by catalytic decomposition of CH<sup>4</sup> were purchased from Shenzhen Nanoport Co. Ltd (purity > 95%, diameter < 10 nm; length = 5–15µm). TiO<sup>2</sup> and CNT–TiO<sup>2</sup> composite catalysts were prepared through an acid-catalyzed sol–gel procedure, as described elsewhere (Silva and Faria, 2010). Briefly, TiO<sup>2</sup> was prepared by dissolving Ti(OC3H7)<sup>4</sup> (Aldrich 97%) in ethanol. The solution was magnetically stirred for 30 min, and then nitric acid (Fluka 65%) was added.

For the composite catalysts preparation, a certain amount of CNT was added to the Ti(OC3H7)<sup>4</sup> ethanol solution. The mixture was kept stirring until a homogenous gel was formed. The gel was left aging in air for 5 days. The resulting material was then crushed into a fine powder (particle size < 100µm). The powders were calcined at 400◦C under a flow of N<sup>2</sup> for 2 h to obtain TiO<sup>2</sup> or CNT–TiO<sup>2</sup> hybrid materials. Catalysts were labeled as XCNT– TiO2, where X (5, 10, 20, 50, 70, and 90) corresponds to the weight percentage of CNT in the material.

The monometallic (Pd) and bimetallic (Pd-Cu) catalysts were prepared by incipient wetness impregnation and coimpregnation, respectively. Briefly, aqueous solutions containing the proper mass of the corresponding salts [PdCl2, Alfa Aesar 99.9%; Cu(NO3)2, Riedel-de Haen 99%] were added dropwise to TiO<sup>2</sup> and CNT-TiO<sup>2</sup> materials. In the case of the bimetallic catalyst, the materials were co-impregnated with a solution containing both precursor salts. The palladium and copper contents were fixed at 1%Pd-1%Cu and 1%Pd (weight percentages). After impregnation, the metal-loaded materials were dried in an oven at 100◦C for 24 h. Then, the catalysts were heat treated under a nitrogen flow at 200◦C for 1 h. At the end of this period, the gas stream was switched to hydrogen for 3 h to promote metals reduction. Finally, the materials were left to cool down to room temperature under a nitrogen flow.

# Catalysts Characterization

Powder X-ray Diffraction (XRD) analysis was performed on a Philips X'PertMPD diffractometer (Cu-Ka = 0.15406 nm). The Brunauer-Emmett-Teller (BET) specific surface area (SBET) was determined from N<sup>2</sup> adsorption-desorption isotherms at 196◦C, in a Quantachrome Nova 4200e apparatus. Temperature programmed reduction (TPR) was carried out in an AMI-200 (Altamira Instruments) system. The H<sup>2</sup> consumption was followed by a thermal conductivity detector (TCD) and by a mass spectrometer (Dymaxion 200 amu, Ametek). Transmission electron microscopy (TEM) micrographs were obtained using a LEO 906E microscope operating with an accelerating voltage of 120 kV. Diffuse reflectance (DR) UV-Vis spectra of the powder samples were recorded on a JASCO V-560 UV-Vis spectrophotometer, equipped with an integrating sphere attachment (JASCO ISV-469). The reflectance spectra were converted by the instrument software (JASCO) to equivalent absorption Kubelka–Munk units. Steadystate photoluminescence (PL) spectra were recorded at room temperature on a JASCO FP-8300 spectrofluorometer equipped with a 150 W Xe lamp. The morphology and elemental mapping of the materials was obtained by SEM/EDXS analysis using a FEI Quanta 400FEG ESEM/EDAX Genesis X4M instrument.

# Catalytic and Photocatalytic Nitrate Reduction Experiments

The catalytic and photocatalytic experiments were carried out in a glass cylindrical reactor. Initially, 190 mL of deionised water and 100 mg of catalyst were fed into the reactor. When used, a gas mixture of H<sup>2</sup> and CO<sup>2</sup> [1:1 flow rate = 200 cm<sup>3</sup> (STP) min−<sup>1</sup> ] was passed through the reactor to remove the dissolved oxygen; CO<sup>2</sup> acts as pH buffer (pH = 5.5). A Heraeus TQ 150 medium pressure mercury vapor lamp (λexc = 254, 313, 365, 436, and 546 nm) was used as radiation source. The lamp was located axially in the reactor and held in a quartz immersion tube. A DURAN <sup>R</sup> glass jacket was used as water circulating cooling system (temperature maintained at 25◦C) and as a filter for cutting-off low wavelength UV lines and letting pass radiation in the near-UV to visible light range (λexc ≥ 365 nm). Before turning illumination on, the solution was magnetically stirred in the dark for 15 min. After that period, 10 mL of a nitrate solution, prepared from NaNO<sup>3</sup> (Sigma-Aldrich 99%), was added to the reactor, in order to obtain an initial NO<sup>−</sup> 3 concentration of 100 mg L−<sup>1</sup> . The first sample was taken out just before the light was turned on, in order to determine the initial nitrate concentration in solution. The catalytic (dark) experiments were carried out under the same experimental conditions, but in the absence of light.

Samples were withdrawn regularly from the reactor, and centrifuged before determination of NO<sup>3</sup> <sup>−</sup>, NO<sup>2</sup> <sup>−</sup>, and NH<sup>4</sup> <sup>+</sup> concentrations. NO<sup>3</sup> <sup>−</sup> and NO<sup>2</sup> <sup>−</sup> were simultaneously determined by HPLC using a Hitachi Elite Lachrom system equipped with a diode array detector. The stationary phase was a Hamilton PRP-X100 column (150 × 4.1 mm) working at room temperature, under isocratic conditions. The mobile phase was a solution of 0.1 M NaCl:CH3OH (45:55). The concentration of NH<sup>4</sup> <sup>+</sup> was determined by potentiometry.

Palladium and copper leaching was assessed after each experiment by atomic absorption spectrometry (UNICAM 939/959), the absence of metals (within the experimental error) being confirmed for all the cases.

Reproducibility tests were performed for selected experiments, the results being in agreement with a maximum error of about 2.5%. NO<sup>3</sup> <sup>−</sup> conversion and the selectivity to NO<sup>2</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> were calculated as described elsewhere (Soares et al., 2014).

# RESULTS AND DISCUSSION

### Catalysts Characterization

XRD patterns of neat CNT, Pd-Cu/TiO2, and Pd-Cu/20CNT-TiO<sup>2</sup> are displayed in **Figure 1**. Typical (002) and (100) diffraction lines are evident in the XRD pattern of CNT. XRD

analysis of TiO<sup>2</sup> and CNT-TiO<sup>2</sup> composites revealed that only anatase phase is present in neat TiO<sup>2</sup> and composite catalysts. The XRD patterns of the Pd-Cu loaded CNT–TiO<sup>2</sup> materials are very similar to the one of Pd-Cu/TiO2, with the CNT contribution hardly been identified. The Cu and Pd phases were not detected by XRD, which must be related to the low metal percentages (1%wt.). Anatase crystallites of 8.5 nm average size were found for neat TiO2, as determined by the Scherrer equation and confirmed by TEM (**Figure 2a**). The sizes of the anatase crystallites present at the composite catalysts decreased with increasing carbon content, suggesting that CNT may act as dispersing medium for TiO<sup>2</sup> particle precursors during the crystallization process (**Table 1**).

The BET surface areas (SBET) of TiO<sup>2</sup> and CNT-TiO<sup>2</sup> materials are listed in **Table 1**. Materials with lower carbon content, namely 5CNT–TiO<sup>2</sup> and 10CNT–TiO<sup>2</sup> composites, showed surface areas lower than the ones estimated through the mass composition of the composites (SBET,calc) and even lower than for neat TiO2. These results indicate that the presence of low amounts of CNT induces the formation of big TiO<sup>2</sup> crystallite agglomerates, therefore decreasing the surface area of the composite catalyst (**Figure 2b**).

With the carbon content increasing up to 20% (20CNT– TiO2), the presence of a larger amount of CNT seems to prevent TiO<sup>2</sup> particles from agglomerating, thus increasing the surface area, which was even higher than the calculated (SBET,calc). This was confirmed by TEM (**Figure 2c**), where TiO<sup>2</sup> particles of very small dimensions can be observed surrounding the sidewalls of CNT. In the case of the composite with

FIGURE 2 | TEM micrographs of TiO<sup>2</sup> (a), 5CNT-TiO<sup>2</sup> (b), 20CNT-TiO<sup>2</sup> (c), and 70CNT-TiO<sup>2</sup> (d) loaded with 1% Cu and 1% Pd (wt. %). The arrows indicate the presence of metal nanoparticles.

similar amounts of TiO<sup>2</sup> and CNT phases (50CNT-TiO2), the values for SBET and SBET,calc were similar. Nevertheless, a further increase on the amount of CNT revealed to have a detrimental effect on the surface area of the resulting composites. A decrease in the SBET in relation to the calculated values of 15 and 40% was observed for 70CNT-TiO<sup>2</sup> and 90CNT-TiO<sup>2</sup> materials, respectively, which may be attributed to the formation of CNT bundles decreasing the accessible surface area, as could be visualized by TEM for 70CNT-TiO<sup>2</sup> (**Figure 2d**). Due to the low amount of metals used, the textural properties of the metal-loaded catalysts remained practically unchanged compared to the pristine supports. The presence of metal particles of very small dimensions (lower than 10 nm) loaded on the TiO2-based materials could be observed by TEM (**Figure 2**).

SEM-EDXS analysis of Pd-Cu/TiO<sup>2</sup> (**Figure 3a**) and Pd-Cu/20CNT-TiO<sup>2</sup> (**Figure 3b**) revealed that Pd and Cu nanoparticles are well dispersed in the catalysts, with no apparent prevalence of one of the metals.

The carbon content of the CNT-TiO<sup>2</sup> composites determined by TG analysis (CTG) agrees fairly well with the nominal percentage, indicating negligible gasification of CNT during the calcination step (**Table 1**).

TABLE 1 | Surface area, carbon content, and dimensions of the anatase crystallites of TiO2 and CNT-TiO2 composites.


<sup>a</sup>Determined by TEM; n.d., not determined.

**Figure 4** shows the TPR profiles for both TiO<sup>2</sup> and 20CNT-TiO<sup>2</sup> materials, which were obtained before heattreating the metal salt-loaded supports. Both materials show a reduction peak centered at 150◦C, assigned to the reduction of Cu oxides promoted by the presence of Pd (Soares et al., 2011a). The thermal treatment under

FIGURE 3 | SEM-EDX analysis of Pd-Cu/TiO2 (a) and Pd-Cu/20CNT-TiO2 (b) with the respective elemental mapping for Pd, Cu, Ti, and O in the selected regions (dashed rectangle).

H<sup>2</sup> should produce Pd and Cu particles in the reduced form, i.e., Pd<sup>0</sup> and Cu<sup>0</sup> (Soares et al., 2010, 2011a). Similar results were obtained for the remaining CNT-TiO<sup>2</sup> composite materials.

DR UV-Vis analysis of Pd-Cu/TiO<sup>2</sup> (**Figure 5A**) shows the TiO<sup>2</sup> characteristic absorption band at wavelength below 400 nm, a band peaking at c.a. 480 nm attributed to the presence of Pd and a broad absorption band rising from 550 nm due to the occurrence of Cu species (López et al., 2009; Wu et al., 2009; Soares et al., 2014).

As expected, the presence of CNT led to a rise of light absorption in the visible spectral region, increasing with the CNT content on the composite catalysts up to a CNT load of 50 wt.%. This behavior has been attributed not only to the capacity of CNT to absorb visible light but also to an increment of surface electronic species availability and mobility in the composite catalysts due to the introduction of CNT, as already reported in previous studies (Silva and Faria, 2010; Dai et al., 2014). A further increase in the CNT content did not produce any effect on the optical absorption of the composite materials. Moreover, the absorption peaks of the metal species could not be identified in the UV-Vis spectra of the composite materials, which may be attributed to a higher dispersion of the metal particles when supported on the composite materials.

The photoluminescence (PL) spectra were performed for having an insight in the behavior of light-generated electronic species in photocatalysts, since PL emission results from the recombination of electrons and holes. The PL emission spectra

FIGURE 5 | (A) Diffuse reflectance UV-Vis spectra of Pd-Cu-loaded TiO2 (i); 5CNT-TiO2 (ii); 10CNT-TiO2 (iii); 20CNT-TiO2 (iv); 50CNT-TiO2 (v); CNT (vi); (B) PL spectra of neat TiO2 (i), Pd-Cu/TiO2 (ii) and Pd-Cu/20CN-TiO2 (iii).

of the pure neat TiO<sup>2</sup> and Pd-Cu loaded TiO<sup>2</sup> and 20CNT-TiO<sup>2</sup> materials excited at 280 nm are shown in **Figure 5B**. The PL signal observed for neat TiO<sup>2</sup> can be attributed to the transition of electrons from the oxygen vacancies to TiO<sup>2</sup> valence band (Tahir et al., 2017). After loading TiO<sup>2</sup> with Pd and Cu, an increase in the PL intensity was observed in the range from 330 to 400 nm, which may be attributed to the higher availability of photoexcited electrons in the bimetallic catalyst. Yet, a decrease in the PL signal intensity in the 400–525 nm range is observed, meaning that electron-hole recombination was decreased by the presence of the metal nanoparticles. Pd-Cu/20CNT-TiO<sup>2</sup> shows a very significant decrease in the PL intensity as compared with bare and metal loaded TiO2, indicating highly efficient inhibition of charge carriers recombination and suggesting the existence of electronic synergies between the metals and the hybrid CNT-TiO<sup>2</sup> material (Zhang et al., 2012).

# Catalytic and Photocatalytic Nitrate Reduction

TiO<sup>2</sup> and CNT-TiO<sup>2</sup> composites loaded with 1%Pd and 1%Cu (wt.%) were used for catalytic and photocatalytic reduction of nitrate in aqueous suspensions. Monometallic Pd-TiO<sup>2</sup> was used for comparison purposes. H<sup>2</sup> and CO<sup>2</sup> were continuously added to the reaction media, acting as reducing agent and pH buffer, respectively. The materials were tested under the same experimental conditions, the only difference between the two processes being the introduction of a near-UV to visible light source in the case of the photocatalytic reactions.

The performances of the different catalysts for the (dark) catalytic reduction process are presented in **Figure 6A**. Very low nitrate conversion (4%) was achieved using the monometallic Pd-TiO<sup>2</sup> catalyst, the presence of Cu being fundamental for the reaction to occur. Moreover, as expected, it can be observed that the support has a crucial influence on the performance of the bimetallic catalysts.

In the case of the catalytic reduction process, Pd-Cu/TiO<sup>2</sup> was the most efficient catalyst in terms of the kinetics of nitrate removal, with total nitrate conversion being achieved at the end of 90 min of reaction. Similar kinetic behavior was observed for nitrate reduction reactions using the composite materials with a CNT load (Y) of 5, 10, and 20%, leading to total nitrate removal at the end of 120 min of reaction. For higher CNT loads, a progressive detrimental effect in the kinetics of nitrate removal has been observed. Nitrate conversions of 95, 79, and 27% have been obtained at the end of 240 min when using composites with Y = 50, 70, and 90%, respectively. For the Pd-Cu/CNT catalyst only 12% nitrate removal was achieved at the end of the catalytic run.

In general, the presence of light promotes a positive effect in nitrate removal (**Figure 6B**). As in the case of the (dark) catalytic process, the simultaneous presence of Pd and Cu was a sine qua non condition for nitrate conversion to occur. The composite material with the lowest CNT content (Y = 5%) produced a decrease in the efficiency of nitrate abatement compared to Pd-Cu/TiO2. Yet, a further increase in the CNT load up to Y = 20% lead to a rise in the rate of nitrate removal. For the composites with higher CNT content (Y = 50, 70, and 90%) a progressive loss in the efficiency toward NO<sup>3</sup> <sup>−</sup> conversion was observed.

The conversion of nitrate at 60 min of reaction was calculated in order to get a better understanding of the effect of CNT load in the kinetics of NO<sup>3</sup> <sup>−</sup> removal by both catalytic and photocatalytic routes (**Figure 6C**). It is notorious that the amount of CNT plays a role in the efficiency of the composite materials. As already mentioned, composite materials underperformed TiO<sup>2</sup> in catalytic nitrate reduction with a decrease in the nitrate conversion with increasing CNT load. Yet, for the photocatalytic process a positive effect was observed using 10CNT-TiO<sup>2</sup> and, in particular, 20CNT-TiO<sup>2</sup> comparing with the experiments using TiO<sup>2</sup> as support. On the other hand, the bimetallic catalyst supported on 20CNT-TiO<sup>2</sup> promotes total conversion of nitrate at 60 min of reaction. Reutilization tests were performed

all other results refer to Pd-Cu bimetallic catalysts. Nitrate conversion at 60 min (XNO<sup>3</sup> <sup>−</sup>,60 min) using Pd-Cu loaded materials (C).

using Pd-Cu/20CNT-TiO<sup>2</sup> and Pd-Cu/TiO<sup>2</sup> under catalytic and photocatalytic conditions. In both cases the results indicate that the performance of the catalysts was maintained within 5% variation over 3 consecutive runs.

**Figure 7** shows the nitrite and ammonium profiles during catalytic and photocatalytic reactions using Pd-Cu loaded TiO2, CNT, and 20CNT-TiO<sup>2</sup> catalysts. Residual amounts of NO<sup>2</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> were produced using Pd-Cu/CNT during both catalytic and photocatalytic processes. When Pd-Cu/TiO<sup>2</sup> was used as catalyst, nitrite is partially transformed into ammonia during the reaction, which is accumulated in the aqueous media. It was found that for the catalytic process, the formation of nitrite and conversion into ammonium is slower than for the photocatalytic process. Also, lower amounts of ammonia were produced in the presence of light, using Pd-Cu/TiO<sup>2</sup> and Pd-Cu/CNT, meaning a higher selectivity of the photocatalytic process toward N<sup>2</sup> formation.

For the reactions using Pd-Cu/20CNT-TiO<sup>2</sup> higher amounts of NO<sup>2</sup> <sup>−</sup> were found during the catalytic and photocatalytic reactions, which were completely depleted at the end of 180 and 60 min of reaction, respectively (**Figure 7A**). Yet, contrarious to what was observed for the reactions using Pd-Cu/TiO<sup>2</sup> and Pd-Cu/CNT, the use of the metal-loaded 20CNT-TiO<sup>2</sup> catalyst under irradiation lead to the formation of higher amounts of ammonia when compared to the catalytic process (**Figure 7B**).

# Catalytic and Photocatalytic Nitrate Reduction Mechanisms Using CNT-TiO<sup>2</sup> Catalysts

As described above, nitrate reduction over CNT-TiO<sup>2</sup> hybrid materials behave very differently in dark conditions and under irradiation. Although the photo-assisted process provide a quicker depletion of NO<sup>3</sup> <sup>−</sup>, the selectivity toward N<sup>2</sup> is greatly affected by the use of CNT-TiO<sup>2</sup> hybrid materials as Pd-Cu supports (**Figure 8**).

In the case of the catalytic reduction process, an increase in N<sup>2</sup> selectivity is observed with increasing CNT content up to 20 wt.%. A further increase in the amount of CNT led to a progressive decrease in the selectivity toward N<sup>2</sup> formation (**Figure 8A**). On the other hand, the photocatalytic reduction of NO<sup>3</sup> <sup>−</sup> using CNT-TiO<sup>2</sup> catalysts appeared to be more selective for the reduction of nitrate into ammonia (**Figure 8B**).

It is well accepted that during the (dark) catalytic reduction over bimetallic catalysts, using hydrogen as reducing agent, NO<sup>3</sup> <sup>−</sup> is converted into NO<sup>2</sup> <sup>−</sup> according to a redox reaction on the promoter metal (Cu). The role of the noble metal is to activate hydrogen, reducing the promoter metal, completing the catalytic cycle (Epron et al., 2001; Soares et al., 2011b; Zhang et al., 2013), being also active for the NO<sup>2</sup> <sup>−</sup> reduction.

In the case of the photocatalytic process using metal-loaded TiO2, the mechanism generally proposed is based on the role of metal nanoparticles as electron sinks. Since the Fermi levels of noble metals are lower than that of TiO2, the photo-excited electrons can be transferred from the conduction band of the semiconductor to the metal nanoparticles deposited on its surface, being then available for NO<sup>3</sup> <sup>−</sup> reduction (Kominami

FIGURE 7 | (A) NO2 <sup>−</sup> and (B) NH<sup>4</sup> <sup>+</sup> concentrations as a function of time during nitrate reduction over Pd–Cu loaded TiO<sup>2</sup> (● and ), 20CNT-TiO<sup>2</sup> ( and ) and CNT (▲ and △) under dark catalytic (solid symbols) and photocatalytic (open symbols) conditions.

et al., 2005; Anderson, 2012; Soares et al., 2014). Hole scavengers are generally used, acting as sacrificial electron donors, avoiding electron-hole recombination. Yet, in the present work, no scavengers were added to the reaction medium.

The role played by CNT in CNT-TiO<sup>2</sup> hybrids has been discussed in previous works reporting the use of this type of materials as photocatalysts for environmental applications (Silva and Faria, 2010; Silva et al., 2015; Zeng et al., 2015; Yang and Park, 2017). CNT may act as adsorbent, as dispersing medium for TiO<sup>2</sup> nanoparticles, it may span light absorption into the visible and may retard electron hole recombination. Although the surface area of the hybrid materials increased with CNT content, the first is not likely to be the most important effect, since no significant adsorption was observed whether the catalyst used. TEM images of CNT-TiO<sup>2</sup> hybrids show that CNT promote the dispersion of TiO<sup>2</sup> avoiding particle agglomeration (**Figure 2**). Moreover, the introduction of CNT increased the absorption of the resulting materials in the visible range, as shown by the UV-Vis spectra (**Figure 5A**), and lead to a decrease in the electron-hole recombination (**Figure 5B**).

Based in our findings, and considering the operation conditions used in this study, the following photocatalytic reaction mechanism is proposed. Since the irradiation source emits in the near UV to visible range, it is expected that TiO<sup>2</sup> and CNT could be photoexcited simultaneously (**Figure 9**, steps 1 and 1<sup>∗</sup> ).

After charge separation, electrons are transferred to Pd and Cu nanoparticles that are supported over TiO<sup>2</sup> and also over

CNTs (**Figure 9**, steps 2–4), as observed by TEM (**Figure 2**). Photogenerated electrons may reduce both nitrate and nitrite adsorbed on Cu and Pd, respectively (**Figure 9**, steps 5–7). On the other hand, positively charged holes may migrate from TiO<sup>2</sup> to the CNT phase, where, in the absence of sacrificial electron donors, water can be oxidized to H<sup>+</sup> and HO• (**Figure 9**, steps 8 and 9). Hydroxyl radicals may indirectly re-oxidize byproducts to NO<sup>3</sup> <sup>−</sup> (Tugaoen et al., 2017), while CO<sup>2</sup> •−, which can be generated from the reduction of CO<sup>2</sup> (used as pH buffer) by available electrons, may play a role as reducing mediator (Zhang et al., 2005; Sá et al., 2009). The higher selectivity toward NH<sup>4</sup> <sup>+</sup> production obtained using CNT-TiO<sup>2</sup> catalysts when irradiated may be rationalized by the excess of H<sup>+</sup> in the reaction medium, resulting from step 9 in **Figure 9**. Yet, no direct correlation between the CNT load in the hybrid materials and the selectivity toward NH<sup>4</sup> <sup>+</sup> could be found (**Figure 8B**), which may derive from the complexity and simultaneity of the reactions involved in the mechanism of the photocatalytic process.

Although the photocatalytic process appears more advantageous in terms of kinetics of nitrate removal, the (dark) catalytic reactions using CNT-TiO<sup>2</sup> hybrid materials revealed to be more selective toward N<sup>2</sup> formation. Nevertheless, the possibility of using this type of materials may be envisioned as a cleaner route for the production of ammonia, comparing with the conventional fossil fuel based process (Yamauchi et al., 2011; Hirakawa et al., 2017).

# CONCLUSION

TiO<sup>2</sup> and CNT-TiO<sup>2</sup> loaded with 1%Pd−1%Cu show high catalytic activity in the dark and under near UV to visible light irradiation, in the presence of H<sup>2</sup> and CO2. The presence of light promotes faster NO<sup>3</sup> <sup>−</sup> conversion, due to the higher availability of reducing species. Carbon nanotubes induce to a positive effect in the selectivity of the catalytic reduction process toward N<sup>2</sup> formation. In the case of the photocatalytic process, the hybrid materials lead to an increase in the yield of the reduction of NO<sup>3</sup> − to NH<sup>4</sup> <sup>+</sup>, due to the high availability of H+. The efficiency of the hybrid materials depends on the CNT load, the best performing material being that composed by 20 wt.% of carbon nanotubes. In the case of the dark catalytic process, the synergic effect observed by the introduction of CNT in the TiO<sup>2</sup> matrix is mainly ascribed to the action of the carbon phase as dispersing medium to metal oxide particles, while under irradiation, CNT produce an increase in the efficiency of charge separation and mobility in the composite material.

# AUTHOR CONTRIBUTIONS

CS and OS conceived the research work, prepared and characterized the catalysts, performed the activity tests and drafted the manuscript. JÓ, MP, and JF provided the means for the realization of this work and contributed to the interpretation of the experimental results. All authors read and approved the final manuscript.

# FUNDING

This work was financially supported by Project POCI-01-0145- FEDER-006984—Associate Laboratory LSRE-LCM funded by FEDER through COMPETE2020—Programa Operacional Competitividade e Internacionalização (POCI)—and by national funds through FCT—Fundação para a Ciência e a Tecnologia, and by the project AIProcMat@N2020—Advanced Industrial Processes and Materials for a Sustainable Northern Region of Portugal 2020, with the reference NORTE-01- 0145-FEDER-000006, supported by Norte Portugal Regional Operational Programme (NORTE 2020), under the Portugal 2020 Partnership Agreement, through the European Regional Development Fund (ERDF). CGS acknowledges the FCT Investigator Programme (IF/00514/2014) with financing from the European Social Fund and the Human Potential Operational Programme.

# REFERENCES


reactor. Catal. Today 149, 341–347. doi: 10.1016/j.cattod.2009. 05.029


**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 © 2018 Silva, Pereira, Órfão, Faria and Soares. 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.

# Hydrodeoxygenation of Bio-Derived Phenol to Cyclohexane Fuel Catalyzed by Bifunctional Mesoporous Organic–Inorganic Hybrids

### Liuye Mo<sup>1</sup> \*, Wanjin Yu<sup>2</sup> \*, Huangju Cai <sup>3</sup> , Hui Lou<sup>3</sup> and Xiaoming Zheng<sup>3</sup>

*1 Institute of Innovation & Application, Zhejiang Ocean University, Zhoushan, China, <sup>2</sup> State Key Laboratory of Fluorinated Greenhouse Gases Replacement and Control Treatment, Zhejiang Research Institute of Chemical Industry, Hangzhou, China, <sup>3</sup> Key Lab of Applied Chemistry of Zhejiang Province, Department of Chemistry, Zhejiang University, Hangzhou, China*

Edited by:

*Zhimin Ao, Guangdong University of Technology, China*

### Reviewed by:

*Yiming Tang, South China Normal University, China Jun Huang, The University of Sydney, Australia*

\*Correspondence:

*Liuye Mo liuyemo@zjou.edu.cn Wanjin Yu yuwanjin1@sinochem.com*

### Specialty section:

*This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry*

> Received: *21 March 2018* Accepted: *25 May 2018* Published: *14 June 2018*

### Citation:

*Mo L, Yu W, Cai H, Lou H and Zheng X (2018) Hydrodeoxygenation of Bio-Derived Phenol to Cyclohexane Fuel Catalyzed by Bifunctional Mesoporous Organic–Inorganic Hybrids. Front. Chem. 6:216. doi: 10.3389/fchem.2018.00216* In this work, mesoporous silica materials SBA-15 functionalized with propyl/ phenyl-sulfonic acid group were synthesized and loaded with Pt to form bifunctional catalysts. SAXRD, WAXRD, N<sup>2</sup> adsorption-desorption, TEM techniques were used to characterize the above bifunctional catalysts. These bifunctional catalysts were applied to the reaction of hydrodeoxygenation (HDO) of bio-derived phenol (PhOH) to produce cyclohexane fuel and showed very good catalytic performances. There were strong synergies between the metal sites and the acid sites on the bifunctional catalysts. This reaction of phenol HDO provides a model system for the catalytic upgrading of biomass-derived fuel.

Keywords: mesoporous silica, bifunctional catalyst, hydrodeoxygenation, bio-oil, phenol, cyclohexane

# INTRODUCTION

Due to its abundance (15–30 wt % of wood-based biomass) and remarkably lower oxygen content than polysaccharides, lignin is a favorable feedstock for the production of biofuel, which is regarded as a promising energy alternative for fossil fuels (Huber et al., 2006). Currently, two-step processes are typical strategies to utilize lignin for biofuel production. In the first step, lignin is hydrolyzed (Kudsy and Kumazawa, 1999; Shabtai et al., 1999; Hepditch and Thring, 2000; Liu et al., 2008) or fast pyrolyzed (Meier et al., 1994; Thring and Breau, 1996; Britt et al., 2000; Dobele et al., 2007; Boateng et al., 2008; Ingram et al., 2008; French and Czernik, 2010) to depolymerize into a mixture of simple aromatic compounds (mostly phenols). Unfortunately, large quantities of reactive, unstable, and corrosive oxygenate compounds are contained in crude biofuel, which cannot be used directly as a vehicle fuel (Garcia-Perez et al., 2007). Therefore, the crude mixture from the first step must be upgraded into fuels in the second step. In order to simultaneously stabilize two main kind of reactive components, aldehydes and organic acids, in the bio-oil, we have newly developed a reaction system named one-step hydrogenation-esterification (OHE) reaction as a possible approach for biofuel upgrading (Tang et al., 2008; Yu et al., 2011). However, phenols with poor combustion performance could not be converted effectively during the OHE process (Yu et al., 2011). Therefore, the conversion of phenolic compounds into hydrocarbon fuels remains a challenge (Crossley et al., 2010). Hydrodeoxygenation (HDO) is regarded as a most attractive and effective method to convert phenolic compounds to alkanes for bio-oil upgrading (Huber et al., 2006). Traditional catalysts for HDO reaction were sulfide CoMo and NiMo/γ-Al2O<sup>3</sup> catalysts, which are used in the hydrodesulfurization or hydrodenitrogenation process in petroleum refineries (Senol et al., 2005; Bui et al., 2009). However, it is well known that the fuels produced by sulfide catalysts will be contaminated as sulfur may be leached into the reaction liquids, and the catalysts will suffer from deactivation by water induction and coke accumulation (Yan et al., 2010). Lercher and Kou groups have reported a new highly efficient one-pot route for HDO of aqueous phenolic bio-oil to cycloalkanes over catalysts combining the noble catalysts with mineral acids or Brønsted acidic ionic liquids (Zhao et al., 2009, 2010; Yan et al., 2010). New progresses have been made over noble metal and transition metal catalysts combined with liquid acids or solid acids to upgrade biofuels via HDO of bio-derived phenolic compounds (Luska et al., 2015; Dongil et al., 2016; Lee et al., 2016). The good catalytic performance of the catalysts using mesoporous acidic solid materials as supports can be expected, as the bulky molecular of bio-derived phenolic compounds can easily access the active sites in the mesopores. Furthermore, the environmental benign solid acids can be recovered simply by filtration. Bifunctional catalysts of Pt/HY and Pt/HBeta were successfully used for phenolics HDO to produce hydrocarbons in fixed-bed configurations (Hong et al., 2010; Zhu et al., 2011). However, utilization of the bifunctional catalysts using mesoporous solid acid materials as supports for HDO of bio-derived phenolics was rarely reported. More recently, Xiao group has developed mesoporous zeolite ZSM-5 supported Ru to convert efficiently both small and bulky phenolic biomolecules via HDO to the corresponding alkanes owing to the open mesopores with abundant exposed acidic sites in the catalysts (Wang L. et al., 2015).

This paper focuses on the conversion of bio-derived phenol to cyclohexane via HDO reaction, a liquid fuel with good combustion properties, under mild reaction conditions over Pt bifunctional catalysts which possess functions of both hydrogenation and dehydration. The support of SBA-15 functionalized with organosulfonic acid was used, showing strong acidity and accessibility for bulky bio-derived molecules, may overcome the disadvantages both of the liquid acids and microporous solid acids.

# EXPERIMENTAL SECTION

# Catalyst Preparation

The catalysts used in this work were organic-inorganic hybrid SBA-15 materials, which were functionalized with propylsulfonic (arenesulfonic) acid groups and loaded with platinum. Organosulfonic acid-functionalized SBA-15 materials were synthesized as described elsewhere (Mbaraka and Shanks, 2006; Tang et al., 2010) with only slight modifications. Tetraethoxysilane (TEOS, 98%, Aldrich) and (3-mercaptopropyl)trimethoxysilane (MPTMS, 85 wt. %, Acros) or 2-(4-chloro-sulfonyl-phenyl) ethyl trimethoxysilane (CSPTMS, 50 wt. % in CH2Cl2, Acros) were adopted directly without further purification as the silica and the organosulfonic acid sources. Pluronic P123 (Aldrich), a triblock copolymer of polyethylene oxide-polypropylene oxide-polyethylene oxide with the molecular structure PEO20-PPO70-PEO<sup>20</sup> (Mw = 5800), was used as a structure template to synthesizethe SBA-15. H2PtCl<sup>6</sup> solution used as the platinum precursor was provided by Hangzhou Kaiming Catalyst Co., Ltd. Formaldehyde (HCHO, 37 wt.%, aqueous solution, Sinopharm Chemical reagents Co., China) was used as reducing agents. The functionalized SBA-15 hybrids with propylsulfonic or arenesulfonic acid groups were abbreviated as SBA-Pr and SBA-Ar, respectively. The nominal loading of platinum on the catalysts was fixed at 1 wt%.

## Synthesis of 1%Pt/SBA-Pr and 1%Pt/SBA-Ar

In a typical synthesis of SBA-Pr, Pluronic P123 (4.00 g, 0.69 mmol) was dissolved in HCl solution (1.9 M, 125 mL) at room temperature. Then, keeping stirring the solution was heated to 40◦C before the addition of TEOS (8.23 mL, 36.74 mmol). Before the addition of MTPMS (3.674 mmol) and H2O<sup>2</sup> (12.86 mmol), ∼45 min was allowed for prehydrolysis of TEOS. The resulting mixture was agitated for 24 h at 40◦C and then aged for another 24 h at 110◦C under static conditions. After that, the resulting solids were separated by filteration, washed with D.I. water and dried at 100◦C in an oven for 8 h. The template in the as-synthesis samples was extracted by 10% v/v HCl in ethanol and refluxing for 36 h. To ensure complete removal of the template, fresh ethanol would be introduced after every 12 h.

Pt was loaded on SBA-Pr by reduction-deposition method. 1.00 g SBA-Pr was added into the aqueous solution containing 30.0 mL of D.I. water and 0.021 g of H2PtCl6. The Formaldehyde solution (10 mL) was used as reducing agent and added dropwise into the above aqueous suspension with vigorous stirring () at 60◦C. The suspension was stirred for 24 h at 60◦C to allow the reduction of Pt4<sup>+</sup> to Pt<sup>0</sup> . After that, the suspension was filtered at room temperature and thoroughly washed with D.I. water until free of Cl<sup>−</sup> in the filtrate (tested with silver nitrate solution). The wet filter cake was re-suspended in 50 mL of 1 wt% H2SO<sup>4</sup> solution for 4 h for acidification. Finally, the solid was treated by filtering, washing, and vacuum drying at 110◦C. The synthesis procedures of SBA-Ar and **1%Pt/SBA-Ar were** the same as the synthesis of SBA-Pr and 1%Pt/SBA-Pr described above, besides CSPTMS was used instead of MPTMS.

# Catalyst Characterization

Small angle XRD (SAXRD) patterns were recorded for SBA-Pr, SBA-Ar, 1%Pt/SBA-Pr, and 1%Pt/SBA-Ar powder in order to confirm the hexagonal structure of samples. Wide angle XRD (WAXRD) was used to analyze the 1%Pt/SBA-Pr and 1%Pt/SBA-Ar. The above XRD measurements were performed on the Rigaku D/MAX-RB.

A Philips-FEI transmission electron microscope (Tecnai G2 F30 S-Twin, The Netherlands), operating at 300- kV, was used to obtain the HRTEM images of catalysts. Samples were mounted on copper grid-supported carbon films by dropping a few droplets of ultrasonically dispersed suspensions of samples in ethanol on the grids, followed by drying at ambient conditions.

CO chemisorption measurements were carried out at 25◦C on a CHEMBET-3000 pulse flow system. Prior to measurements, the catalysts were tableted formed, crushed and sieved to 40–60 mesh. To remove any oxygen from the samples, pretreatments were done at 250◦C for 1 h with a ramp of 10◦C/min from room temperature under 30 mL/min of 5 vol. % H2/He up to 250◦C. After that, the adsorbed hydrogen on samples was removed by purging with 30 mL/min helium gas for 2 h at 250◦C. Finally, the catalysts were cooled down to 25◦C under He flow and 5 vol% CO in helium were pulsed into the catalysts until the CO peak areas appeared to be identical.

The real weight percentages of Pt on the supports of SBA-Pr and SBA-Ar were determined by inductively coupled plasmamass spectroscopy (ICP-MS) on a PS1000 instrument from American LEEMAN LABS INC.

The N<sup>2</sup> adsorption/desorption method was adopted to measure the textural properties of the catalytic materials on an automated adsorption apparatus (OMNIISORP 100CX) at−196◦C. Prior to the adsorption/desorption measurement, all samples were degassed for 2 h at 195◦C in the pretreated chamber of the adsorption apparatus. The surface area, pore volume (Vp) were calculated by the Brunauer-Emmett-Teller (BET) and Barret-Joyner-Hallenda (BJH) methods. The mean pore diameter (MPD) of the samples was analyzed by the BJH method based on the adsorption branch of the N<sup>2</sup> adsorptiondesorption isotherms.

The decomposition of organic moiety and the thermal stability of the solids were examined by thermogravimetric analysis (TGA) on a Perkin-Elmer TGA7 instrument. The samples were heated under a stream of air (20 mL/min) from 50 to 700◦C with a ramp rate of 10◦C/min.

The acidic properties (strength and amount of acid sites) of the catalysts were determined by acid-base titration. The relative pKa values of the samples were estimated using the Gran plot analysis. Typically, 0.1 g of the sample was weighted and suspended in 10 ml of D.I. water, and then was titrated potentiometrically by continuous addition of 5 mM KOH aqueous solution.

# Catalytic Evaluation

All the HDO of phenol experiments were performed in a 100 mL stainless steel batch autoclave reactor. In a typical experiment, a certain mass of phenol dissolved in a certain volume of dichloromethane was added into the reactor with a certain amount of catalysts. Before each run, the reactor was purged with hydrogen for five times, 2.0 MPa at a time, to remove air from the setup. Then hydrogen was introduced into the reactor with a certain pressure ranged from 1.0 to 4.0 MPa. The reaction system would be heated to reaction temperatures (80–200◦C) within half an hour. The stirring speed and the granule size were adopted at 800 rpm and smaller than 400 mesh to ensure that the mass transfer on catalytic performances was excluded. After reactions, the reactor was put into ice water and quickly cooled down to room temperature. Subsequently, the gas in the reactor was released very slowly to a sampling gas bag that was purged in advance with H<sup>2</sup> for five times. The products in gas phase were analyzed using a gas chromatography (GC) equipped with a Porapak Q packed column and a TCD. The products in liquid phase were identified and quantified by a GC-MS and a GC (equipped with an SE-30 nonpolar capillary column and a flame ionization detector, FID), and the internal standard method was applied.

Reproducibility of experiment results was carried out by repeating each run for three times at least until the results were within acceptable limits. The conversion of PhOH (X(PhOH)), yield of i (Y(i)) and the selectivity to i (S(i)) were calculated based on the following Equations. (1–3):

$$X(PhOH) = \frac{m(PhOH, in) - m(PhOH, out)}{m(PhOH, in)} \times 100\% \quad \text{(1)}$$

$$Y(i) = \frac{\frac{\text{m}(i)}{\text{M}(i)} \times \text{M}(\text{PhOH})}{m(\text{PhOH}, i\text{m})} \times 100\% \tag{2}$$

$$S(i) = \frac{Y(i)}{X(PhOH)} \times 100\% \tag{3}$$

# RESULTS AND DISCUSSION

## Characterization Results XRD Results

**Figure 1** is small-angle XRD spectra of freshly prepared SBA-Pr, SBA-Ar, Pt/SBA-Pr, and Pt/SBA-Ar. Three well-resolved peaks were observed in each sample, which were indexed to the (100), (110), and (200) reflections of the hexagonal space group P6 mm (Zhao et al., 1998; Yue et al., 1999). All the diffraction patterns suggested that the SBA-15 structure was well preserved in the process of synthesis of SBA-Pr, SBA-Ar by co-condensation method and in the process of introduction of Pt by reductiondeposition method. The addition of precursors of MPTMS, CSPTMS and H2PtCl<sup>6</sup> did not destroy the structure of the SBA-15. Figure S1 presents wide-angle XRD spectra of Pt/SBA-Pr and Pt/SBA-Ar. Only the characteristic diffraction peaks of metallic Pt, (111), (200), (220), were detected for the two samples, demonstrating that the Pt species were primarily present in metallic form after reduction by formaldehyde. According to calculation based on the Scherrer equation, the average Pt particle size of 1%Pt/SBA-Pr was 8.9 nm, while that of 1% Pt/SBA-Ar was 9.7 nm. This illustrated that Pt particles were highly dispersed on these supports.

## DTG Characterization

The differential thermogravimetric (DTG) analysis was used to determine the thermal decomposition behaviors of the tethered organic moieties in the catalyst samples, as determined by differential thermogravimetric (DTG) analysis, (shown in Figure S2). For SBA-Pr without extraction by ethanol, three peaks centered at about 110, 250, and 460◦C were observed, which were resulted from the desorption of water, the decomposition of template and the decomposition of propylsulfonic acid, respectively. No weight loss peak appears in the vicinity of the 350◦C (which was attributed to the decomposition of the propylthiol groups, MPTMS) (Tang et al., 2010), meaning the effective oxidation of MPTMS to propylsulfonic acid by hydrogen peroxide added during the preparation of SBA-Pr. For SBA-Pr after ethanol extraction, only weight loss peak of propyl sulfonic acid (around 470◦C) was observed on DTG curves, indicating complete removal of the template P123 by ethanol extraction method. The results of DTGclearly showed that the propylsulfonic acid groups in the catalysts did not occur decomposition while using the catalyst under 350◦C owing to the higher thermal stability of tethered propylsulfonic acid.

### ICP-MS and CO Chemisorptions Results

The properties of the metal sites of two catalysts, including metal loading, and dispersion, were characterized by ICP-MS and CO chemisorption techniques. The results are listed in **Table 1**. ICP results showed that the real loading of Pt was smaller than the


theoretical loading of 1 wt.%, demonstrating loss of metal during the preparation process. When Pt was loaded on SBA-Pr, the highest load efficiency of 84% and the highest dispersion (DPt) of 53% were obtained. The metal particle sizes (dPt) showed in **Table 1** were calculated from the CO chemisorptions results according literature (Wang and Yeh, 2001). All the metal particles of Pt were highly dispersed over the SBA-Pr and the SBA-Ar with dPt < 3.0 nm. Noticeably, the dPt was as small as 2.1 nm on the Pt/SBA-Pr. Evidently, the particle sizes calculated from chemisorptions was quite smaller than the aforementioned data derived from XRD. It is well known that the XRD can only be sensitive the big crystallines but not the highly dispersed particles.

### TEM Characterization Results

TEM is a useful technique to characterize the structure of mesoporous materials. From **Figures 2**, **3**, it can be seen, the organic sulfonic acid functionalized SBA-15 materials have regular one-dimensional pore structure preserving the characteristic pore structure of SBA-15. **Figure 4** shows that after loading of Pt the mesoporous structure of the carriers were still well preserved and the Pt particles were dispersed well on the supports. It could be observed that a lot of Pt particles were as small as <2 nm located in the mesopores.

# N<sup>2</sup> Adsorption–Desorption Characterization Results

The N**2** adsorption-desorption isotherms of the samples of SBA-Pr, SBA-Ar, Pt/SBA-Pr, and Pt/SBA-Ar are shown in Figure S3. The observed isotherms of the four samples were obviously classified as IV type isotherms, which was consistent with the mesoporous structure directed by non–ionic surfactant (Buchmeiser, 2003). All the isotherms had Type-H 1 hysteresis loops, and the capillary condensation occured at higher relative pressures (0.7–0.8). This illustrated the synthesized mesoporous materials possessing regular pore arrangements and narrow pore size distributions (Schmidt et al., 1995). The nitrogen

FIGURE 4 | TEM images of 1%Pt/SBA-Pr (A) and 1%Pt/SBA-Ar (B).

sorption isotherms also revealed that no partial pore blocking has occurred upon incorporation of platinum, as the adsorption and desorption branches were almost parallel and exhibited narrow hysteresis. After loading of platinum the pore diameter, pore volume, and surface area of SBA-15 were waned showing in Table S1. After functionalization with organosulfonic acid and loading of Pt, the specific surface area and pore size of the samples were decreased compared to SBA-15, but still remained at a high level. These data also showed that the organic and inorganic functional groups did not cause remarkable clogging of pores. The structure of SBA-15 remained stable during the functionalization.

# Catalytic Testing Results Pt Based Organic-Inorganic Hybrid Materials Catalyzed HDO Reaction of Phenol

The catalytic results of phenol HDO catalyzed by bifunctional Pt/SBA-Pr, as well as the relevant control experiment results are presented in **Table 2**. From Entry 1, a good result of phenol HDO reaction was achieved over the bifunctional catalyst of 1%Pt/SBA-Pr with high conversion of phenol (X(PhOH) = 94.1%) and high selectivity to target product of cyclohexane (S(C6H12) = 98.6%). From Entry 3 and 5, it could be seen that when there is no catalyst or SBA-Pr is used alone phenol did not convert at all. Surprisingly, in Entry 2, over the catalysts of 1% Pt/SBA-15 the product of cyclohexane was also formed and the selectivity was as high as 63.8%. In entry 4, 1%Pt/SBA-15 and SBA-Pr was physically mixed to form bifunctional catalyst. It was found that both of X(PhOH) and S(C6H12) were significantly improved over the physically mixed catalyst compared to the 1%Pt/SBA-15 catalyst (entry 2). From above results, it could be considered that the metal sites of catalysts alone catalyze the HDO of phenol with the aid of negligible acidity sites of silanol groups on SBA-15 support, but the stronger acid sites would greatly enhance the HDO reaction activity of the catalyst. Compared to the physical mixed bifunctional catalyst (entry 4), the composite bifunctional catalyst (entry 1) owed the highest conversion of phenol and the best selectivity to cyclohexane. The above results exhibited that there was a synergistic effect between metal sites and acid sites for HDO reaction of phenol. Huang and Baiker group also found that the proper acid sites properties could significantly enhance the catalytic performance for the reaction of hydrogenation of acetophenone on the Pt/[Al]MCM-41 (Wang Z. et al., 2015).

### Effect of Supports on Catalyst Performances

Pt was supported on SBA-Pr and SBA-Ar to form bifunctional catalysts of 1%Pt/SBA-Pr and 1%Pt/SBA-Ar. Their catalytic results for HDO of phenol are listed in **Table 3**. 1%Pt/SBA-Pr exhibited higher conversion of phenol (94.1%) and higher selectivity to cyclohexane (98.6%), compared to SBA-Ar supported Pt catalyst with X(PhOH) = 35.0% and S(C6H12) = 67.4%. According to the characterization results (**Table 1**), though the two catalysts had the same nominal Pt loading, they actually had different Pt content. In order to exclude the influence of the metal content on the results, the TON calculated based on the target product of cyclohexane are also shown in **Table 3**. Pt/SBA-Pr shows higher TON (103.3 mol cyclohexane/mol Pt) in comparison with Pt/SBA-Ar (14.5 mol cyclohexane/mol Pt). Since the impact of metal sites on HDO

TABLE 2 | Catalytic performance for HDO of phenol.


*Reaction conditions: TR, 200*◦*C; <sup>P</sup>H*2*, 4.00 MPa; tR, 8 h, 800 rpm, 0.0125 mol PhOH in 25.0 mL CH2Cl2, 0.9756 g catalyst (0.05 mmol Pt, except for entry 3 and 5).*


*Reaction conditions: TR, 200*◦*C; <sup>P</sup>H*2*, 4.0 MPa; tR, b8 h, 800 rpm; tR, 4 h, 0.0125 mol PhOH in 25.0 mL CH2Cl2, 0.9756 g catalyst (nPt:nPhOH* = *1:250).*

reaction was excluded, only the acidities of the supports affected the HDO reaction.

The acidities of the catalysts, including the amount and strength of the acid sites, were characterized by means of acidbase titration. The results are shown in **Table 4**. In spite of the same nominal loading, the real amount of acid sites of SBA-Ar was smaller than that of SBA-15-Ar. This might be related to the fact that the molecule of arenesulfonic acid had bigger size and rigidity compared to propylsulfonic acid, which was not conducive to effective loading of arenesulfonic acid. The data also showed that the pKa value of SBA-Ar was smaller than that of SBA-Ar, i.e., SBA-Ar exhibited stronger acid sites. According to Shanks (Mbaraka and Shanks, 2006), the arenesulfonic acidfunctionalized samples had lower pKa values as the sulfonate ions in the phenyl group were more stablethan that in aliphatic carbon chain of propylsulfonic acid groups after deprotonation.

Pt/SBA-Pr had more acid sites but lower acid strength compared to Pt/SBA-Ar (**Table 4**). The catalytic activity of Pt/SBA-Pr was higher than that of Pt/SBA-Ar (**Table 3**). Therefore, it was concluded that the acid sites amount of the supports may be a dominant factor to affect the activity of a bifunctional catalyst. The more the acid sites was, the higher activity of the HDO reaction was. In addition, from Table S1, the pore volume of Pt/SBA-Pr and Pt/SBA-Ar is 0.91 and 0.78 cm<sup>3</sup> /g, respectively. Obviously, from perspective of mass diffusion the former was in favor of the reaction.

### Screening of Solvents

Solvent always plays an important role in a liquid phase reaction. **Table 5** presents the results of phenol hydrodeoxygenation in different solvents. The catalytic activity of 1%Pt/SBA-Pr was the lowest (X(PhOH)∼5.0%) as the solvent of water with high polarity was used, while the solvent of dicloromethane with low polarity was adopted both the X(PhOH) and the S(C6H12) reaches a maximum with 63.0 and 95.9%, respectively. From

TABLE 4 | Acidities of the catalysts and the supports.


TABLE 5 | HDO of phenol over 1%Pt/SBA-Pr in different solvents.


*Reaction conditions: TR, 200*◦*C; <sup>P</sup>H*2*, 3.0 MPa, 800 rpm; tR, 4 h, 0.0250 mol PhOH in 50.0 mL solvent, 0.4878 g 1%Pt/SBA-Pr (nPt:nPhOH* = *1:1000).*

the above results, solvents had significant effects for HDO of phenol, the catalytic activities increased as the polarity of a solvent lowered. The solvent of dichloromethane with the lowest polarity was the best reaction solvent among the solvents tested, including H2O, EtOH, dioxane, and dicloromethane solvents. So the optimization work was carried out in the solvent of dichloromethane.

### Optimization of Reaction Conditions

**Figure 5** shows the effects of reaction temperature (TR) on the HDO reaction of phenol over the 1%Pt/SBA-Pr catalyst. Both conversion of phenol and selectivity to cyclohexane increased with the T<sup>R</sup> increasing. In the evaluation range of temperature, at 200◦C the X(PhOH) and the S(C6H12) reaches maximum, 63.0 and 95.5%, respectively. The main byproduct was cyclohexanol.

**Figure 6** presents the effects of hydrogen pressure (PH2) on the HDO reaction of phenol over the 1% Pt/SBA-Pr catalyst. In the evaluation range the initial hydrogen pressure had little effect on the conversion of phenol, it mainly affected the selectivity to the desired product cyclohexane. The X(PhOH) increased from 56.6 to 69.7% while increasing the PH2 from 1.0 MPa to 4.0 MPa. However, the S(C6H12) increased remarkably from 60.9 to 97.4%. The higher the initial hydrogen pressure was, the larger the concentration of adsorbed active hydrogen on the catalyst surface was, and thus the more conducive to the generation of the target product of cyclohexane.

The effects of the catalyst amount on the HDO reaction of phenol was also investigated over the 1% Pt/SBA-Pr catalyst (**Figure 7**). At the certain reaction condition (PH2 = 4.00 MPa, <sup>T</sup><sup>R</sup> <sup>=</sup> <sup>200</sup>◦C, 800 rpm, t<sup>R</sup> <sup>=</sup> 4 h), the conversion of phenol increased with the nPt/nPhOH increasing from 1:1000 to 1:250, while the selectivity to the target product of cyclohexane remains almost unchanged, meaning that the selectivity to cyclohexane was independent on the amount of Pt metal sites under this reaction condition.

conditions: PH2, 3.00 MPa; 800 rpm; tR, 4 h, 0.0250 mol PhOH in 50.0 mL CH2Cl2, 0.4878 g 1%Pt/SBA-Pr (nPt:nPhOH <sup>=</sup> 1:1,000).

FIGURE 7 | Effect of nPt/nPhOH on the HDO of phenol over 1%Pt/SBA-Pr. Reaction conditions: PH2, 4.00 MPa; TR, 200◦C; 800 rpm, tR, 4 h.

The effect of reaction time on the catalytic performance of 1%Pt/SBA-Pr was also investigated. Most of the phenol was converted within the first 2 h (**Figure 8**). When increasing the reaction time from 4 h to 8 h, the conversion of phenol slightly increased from 90.2 to 94.1%, with the selectivity to cyclohexane almost unchanged.

By screening the reaction temperature, the initial hydrogen pressure, reaction time and the relative content of the catalyst and the reaction solvent, the optimal reaction condition was obtained: T<sup>R</sup> <sup>=</sup> <sup>200</sup>◦C, PH2 <sup>=</sup> 3.00 MPa, 800 rpm, t<sup>R</sup> <sup>=</sup> 4 h, 0.0125 mol PhOH in 25.0 mL CH2Cl2, 0.9756 g 1%Pt/SBA-Pr (nPt:nPhOH) = 1:250. Under these conditions, the conversion of phenol was 94.1%, and the selectivity to desired product of cyclohexane reached to 98.6%.

### Possible Reaction Mechanism

Based on the experimental results a possible reaction mechanism for HDO of phenol was proposed as **Scheme 1**. Experimental results of Kou et al (Yan et al., 2010; Zhao et al., 2010) showed that cyclohexanol can exist stably under acid catalysts below 180◦C. Once the temperature was above 180◦C under acid catalysts it dehydrated rapidly to produce cyclohexene. But in our reaction conditions, cyclohexene was never detected as, cyclohexene might be easily hydrogenated to cyclohexane in the reaction system. At lower reaction temperatures (below 180◦C), or when

# REFERENCES


there are no acidic sites on the catalyst, cyclohexanol cannot dehydrate to form cyclohexene, so hydrogenation reaction is conducted by path ①② or ⑤ to generate cyclohexanol, hydrogenolysis of which takes place to form cyclohexane follow the path of ⑥. Different from the literatures (Hong et al., 2010; Zhu et al., 2011) performing with gas phase reactants in fixed bed reactor, aromatic hydrocarbons, which were produced from hydrogenolysis (direct deoxygenation), could not be found in this work. The difference might result from the different reaction conditions and catalyst systems. Therefore, the reaction pathways might be proposed as the **Scheme 1**.

# CONCLUSIONS

In this work, organic-inorganic hybrid materials of SBA-Pr and SBA-Ar were synthesized and loaded with Pt to form bifunctional catalysts. These bifunctional catalysts had good catalytic activities for HDO of phenol to produce cyclohexane, a model reaction of biofuel upgrading. Under optimal reaction conditions, the conversion of phenol was 94.1%, with 98.6% selectivity to the target product of cyclohexane. The metal sites and the acid sites on bifunctional catalysts displayed significant synergistic effect for the HDO of PhOH. The bifunctional catalysts using in this work may be potentially applied to the upgrading of bio-fuel derived from lignin or biomass.

# AUTHOR CONTRIBUTIONS

LM was in charge of the writting and designing the experiments. WY and HC performed the experiments. XZ and HL were consultants to the project.

# FUNDING

We thank the financial support from Zhejiang Ocean University Research Startup Founding (Q1538) and Department of Education of Zhejiang Province (507300-G21513).

# ACKNOWLEDGMENTS

We thank Dr. Yang Tang's benificial discussion in the manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2018.00216/full#supplementary-material

model compounds. J. Org. Chem. 65, 1376–1389. doi: 10.1021/jo99 1479k


mixed feed. Catal. Today 143, 172–178. doi: 10.1016/j.cattod.2008. 11.024


**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 © 2018 Mo, Yu, Cai, Lou and Zheng. 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 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.

# The Correlation of Adsorption Behavior between Ciprofloxacin Hydrochloride and the Active Sites of Fe-doped MCM-41

Ying Wu<sup>1</sup> , Yiming Tang1,2,3 \*, Laisheng Li 1,2,3, Peihong Liu1,2,3, Xukai Li 1,2,3, Weirui Chen<sup>1</sup> and Ying Xue<sup>1</sup>

*<sup>1</sup> School of Chemistry and Environment, South China Normal University, Guangzhou, China, <sup>2</sup> Guangdong Provincial Engineering Technology Research Center for Drinking Water Safety, Guangzhou, China, <sup>3</sup> Guangdong Provincial Key Lab of Functional Materials for Environmental Protection, Guangzhou, China*

### HIGHLIGHTS


• Co-effect of CPX and metal cations on Fe-MCM-41 was investigated.

### Edited by:

*Zhimin Ao, Guangdong University of Technology, China*

### Reviewed by:

*Priyabrat Mohapatra, C.V. Raman College of Engineering, Bhubaneswar, India Sijie Lin, Tongji University, China*

> \*Correspondence: *Yiming Tang y.tang@m.scnu.edu.cn*

### Specialty section:

*This article was submitted to Green and Environmental Chemistry, a section of the journal Frontiers in Chemistry*

> Received: *22 September 2017* Accepted: *19 January 2018* Published: *07 February 2018*

### Citation:

*Wu Y, Tang Y, Li L, Liu P, Li X, Chen W and Xue Y (2018) The Correlation of Adsorption Behavior between Ciprofloxacin Hydrochloride and the Active Sites of Fe-doped MCM-41. Front. Chem. 6:17. doi: 10.3389/fchem.2018.00017* Fe-MCM-41s with various molar ratios of silicon to iron (20, 40, 80, and 160) were prepared to investigate adsorption properties of ciprofloxacin hydrochloride (CPX) in aqueous solutions. Fe-MCM-41s were characterized by transmission electron microscope (TEM), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), nitrogen adsorption/desorption isotherms, and infrared spectroscopy (FT-IR). Effects of silicon-iron ratio, adsorbent dosage, pH, and temperature were conducted to explore the adsorption mechanism of CPX on Fe-MCM-41. The results showed that the introduction of iron facilitated the absorption quantity for CPX from 20.04 to 83.33 mg g−<sup>1</sup> at 120 min of reaction time, which was mainly attributed to surface complexation. The promotion of hydrophobic effect, electrostatic interactions, and π-π electron donor-acceptor interaction also played coordinate roles in the adsorption process. The experimental kinetic data followed both the pseudo-second-order and intra-particle diffusion models, while the adsorption isotherm data fit well to Freundlich model at high temperature. Thermodynamic study showed that the adsorption was spontaneous. Under the effect of electrostatic interaction, pH of the solution strongly affected CPX adsorption. Five representative metal cations (Ca, Cu, Ni, Pb, and Cd) were chosen to study the effects on CPX adsorption and their complexation. The inhibiting effect of metal cations on CPX adsorption was sequenced in the order of Cu > Ni > Pb > Cd > Ca, which followed the same order as the complexation stability constants between CPX and cations. The Fe-MCM-41 adsorbent possessed excellent reusability for 4 cycles use, suggesting a potential applicability of Fe-MCM-41 to remove CPX in water.

Keywords: ciprofloxacin hydrochloride, Fe-MCM-41, adsorption capacity, pH, complexation, metal cations

# INTRODUCTION

During the last decades, pharmaceuticals and personal care products (PPCPs) among the emerging contaminants have caused more concern of environmental research than the conventional priority pollutants (Arp, 2012; Liu and Wong, 2013; Hayat and Marty, 2014; Montes-Grajales et al., 2017). With abusive use of various antibiotics, large quantity of pharmaceutical wastewater (Cardoso et al., 2014), hospital effluents (Ory et al., 2016; Verlicchi and Zambello, 2016), and excreta-urine (Zheng et al., in press) containing antibiotics have been discharged into environment and regarded as an emerging issue around the world. As a high-use second generation fluoroquinolone, ciprofloxacin hydrochloride (CPX) has the strongest antimicrobial activity and has the highest water concentration in Pearl River of Guangzhou (Bu et al., 2013). Similar to other antibiotics, CPX can transfer in natural environments either as the parent compound or its hydrolysis products, conjugates, oxides when excrete from a target organism (Sarmah et al., 2006) and bring great threats to the ecosystem and human health by inducing proliferation of drug-resistance bacteria (Johnson et al., 2015; Wang et al., 2017b). Therefore, the removal of CPX from water has become a pressing problem. Various methods have been attempted for the removal of CPX from water, such as ultrasound decomposition (Xiao et al., 2013), photocatalytic degradation (Bojer et al., 2017), membrane bioreactor (Hamjinda et al., 2017), ozonation (Gomes et al., 2017), and adsorption on bamboo-based activated carbon (Carabineiro et al., 2012; Peng et al., 2015; Wang et al., 2015), goethite (Gu et al., 2015), graphene oxide (Chen et al., 2015; Fei et al., 2016), Aluminum and Iron hydrous oxides (Gu and Karthikeyan, 2005), and palygorskite-montmorillonite (Berhane et al., 2016). Adsorption technique is widely applied to remove antibiotics from wastewater as a promising method due to its simple theoretical design, ease of operability, relatively low costs and lower amounts of toxic byproducts.

As a member of the M41S family, MCM-41 applied to adsorption area has attracted more awareness because of its hexagonal arrays of uniform channels, high surface area and pore volume and hydrothermal stability (Lee et al., 2007; Jiang et al., 2012). However, hydrophobicity of CPX limits the adsorption capacity by highly hydrophilic MCM-41. Therefore, functional modifications are desired to improve its performance. In the former research, CPX adsorption process can be influenced by metal cations which has been demonstrated that many cations have complexation ability to CPX (Turel et al., 1996). It has been reported that the environmental fate of CPX can be affected by coexisting cations such as Ca and Cu (Pei et al., 2010; Chen et al., 2013). Hence, on the one hand, the introduction of metal ingredient to MCM-41 leads to cation bridging and hydrophobic enhancement, which is beneficial to material for better absorbability. On the other hand, there are few investigations dealing with the different impacts of various free heavy metals on CPX adsorption by comparison.

The main objectives of this paper are to prepare an adsorbent which can enhance the adsorption capacity of CPX by modifying the pure MCM-41 with Fe, to evaluate the effects of various factors including silicon-iron ratio, adsorbent dosage, pH, contact time, temperature, and to study the adsorption mechanisms. Specifically, when CPX coexists with heavy metals in the aqueous phase, the relationship of the formation of Metal cations/CPX complex and the adsorption abilities of CPX is presented to estimate the optimal condition of Fe-MCM-41 adsorbent in practical use.

# EXPERIMENTAL SECTION

# Materials

These reagents were used to synthesize and modify MCM-41: cetyltrimethyl ammonium bromide (CTAB), sodium silicate (NaSiO3·9H2O) and ferric nitrate (Fe(NO3)3·9H2O) were obtained from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). For the adsorption experiments, ciprofloxacin hydrochloride (CPX) was obtained from Macklin Biochemical Co. Ltd. (Shanghai, China). Acetonitrile was chromatography grade and was purchased from Tianjin Kemiou Chemical Reagent Co. Ltd. (Tianjin, China). All the other metals were nitrate salt species and analytical grades. Deionized water was used throughout the study. pH value of the solution was adjusted with 0.1 M hydrochloric acid (HCl) and sodium hydroxide (NaOH).

# Synthesis of Adsorbent

MCM-41 was prepared via a hydrothermal treatment using Na2SiO3·9H2O as silicon source and CTAB as the structuredirecting group. Briefly, dissolved sodium silicate (0.1 mol) was stirred at 313 K for 15 min. H2SO<sup>4</sup> (2 M) was added dropwise to form a gel. After adding CTAB (7.28 g dissolved in 25 ml water at 338 K) to the gel, the mixed solution was stirring for 30 min, and then aged at 418 K for 48 h. The product gotten was filtered after natural cooling, washed with deionized water and dried. Finally, the sample was calcined for 6 h at 823 K to obtain MCM-41. Fe-MCM-41 was prepared from the same route besides adding Fe(NO3)3·9H2O as iron modification to the solution before pH adjustment. The samples were designated as Fe-MCM-41(x) (x = 20, 40, 80, and 160), where x was denoted as different molar ratios of Si to Fe for as-synthesized materials.

# Analytical Procedures

The low-angle XRD measurements (D8 ADVANCE, BRUKER, German) were carried out in the 2θ ranges of 1.0–8.0◦ . The N<sup>2</sup> adsorption/desorption isotherms (ASAP2020, Micromeritics, USA) was used to calculate the surface area. The pHpzc value was measured by potentiometric titration. The leaching iron content of the solution was determined by atomic absorption Spectrophotometer (AA-7020, EWAI, China). The FT-IR absorption spectra (6700, Nicolet, USA) was obtained over the range 4,000–400 cm−<sup>1</sup> .

The concentration of CPX was quantified using HPLC (LC10A, Shimadzu) consisting of a UV detector (SPD-10AV) at 278 nm and a Diamonsil C18 column (250 × 4.6 mm, 5µm, Dikmate technologies). The mobile phase used for detection of CPX was acetonitrile/0.01 M potassium dihydrogen phosphate (23:77, v/v) and the flow rate was 1.0 mL min−<sup>1</sup> .

# Batch Adsorption Experiments

For adsorption experiments, 0.04 g of adsorbent was mixed with 200 mL of CPX solution (20 mg L−<sup>1</sup> , initial pH = 5.40) in flasks and the mixture stirred at 303 K up to beyond the equilibrium time. The rotating speed was set to 175 rpm. Additional experiments were also performed to study the effects of equilibrium time, dosage, pH, temperature and metal cations on the adsorption properties. To study the impact of various metal cations on the adsorption behavior of CPX, predetermined amounts of Ca(II), Cu(II), Ni(II), Pb(II), Cd(II) were introduced to obtain 0.01 M ionic strength solutions at 303 K, respectively. Controls without Fe-MCM-41 were considered to explain losses from adsorption to flasks. All experiments were run in triplicate under the same conditions.

A series of adsorbent doses from 30 to 60 mg were selected to discuss the effect of dosage. The pH effects experiments were conducted by adjusting the pH between 3 and 11. To study the impact of various metal cations on the adsorption behavior of CPX, predetermined amounts of Ca(II), Cu(II), Ni(II), Pb(II), Cd(II) were added to obtain 0.01 M ionic strength solutions at 303 K, respectively. The isotherms were obtained by batch experiments performed at 293 ± 0.5, 303 ± 0.5, 313 ± 0.5 K, respectively. The initial concentrations of CPX (varying from 20 to 80 mg L−<sup>1</sup> ) were chosen based on preliminary experiments, which controlled the adsorbed amount of CPX between 30 and 90% of the initial amounts. A study of reusability was carried out by adsorption/desorption for four times.

# RESULTS AND DISCUSSION

# Characterization of Adsorbents Used in This Work

**Figure 1** showed the low angle XRD patterns of all absorbents at different molar ratios of Si to Fe and pure sample. All absorbents had three well-resolved peaks indexed to (100), (110), and (200) diffraction planes, indicating a highly ordered hexagonal mesostructure (Li et al., 2012). When the silicon-iron ratio decreased, iron content became higher, the intensity of the peaks diminished and the peak width increased. This meant the introduction of excess iron could reduce ordering of the structure. The result of the segment was consistent with the TEM exhibited in Figure S1.

All isotherms in **Figure 2** belonged to type IV isotherm according to IUPAC classification. Compared with the other four absorbents, the BET curve and the BJH pore size distribution curve of Fe-MCM-41(20) exhibited obviously different. The position closed to p/p<sup>0</sup> = 0.9 showed a distinct region of steep increase, which was corresponding to the capillary condensation in the pristine silica gel pores (Kolesnikov et al., 2017). This

ratio of Si to Fe).

was similar to the above discussion of XRD that the pores size increased after iron-doping during aging process gradually. Except for tiny amounts of macroporous structures, the vast majority of pore widths were below 5 nm. It could be seen from the structural properties of all absorbents in Table S1. The pore diameters increased from 3.15 to 4.33 nm and the BET surface area decreased from 1002.1 to 625.6 m<sup>2</sup> g −1 . Note that by replacing Si4<sup>+</sup> (0.40 Å) with Fe3<sup>+</sup> (0.63 Å) in the structure of material, the extension of radius enlarged the pore sizes and decreased surface area. Moreover, XPS was provided to clarify the chemical phase of Fe species in the synthesized Fe-MCM-41. XPS spectrum in **Figure 3** gave a description that the peaks at 712 and 725 eV for the binding energies of Fe2p3/<sup>2</sup> and Fe2p1/2, reflecting iron species existed as Fe3<sup>+</sup> in Fe-MCM-41 (Nie et al., 2015).

# Effect of Silicon-Iron Ratio

In **Figure 4**, the removal efficiency for CPX decreased with the increasing surface area of Fe-MCM-41. Hydrophobic compound was difficult to be adsorbed on the surface of mesoporous MCM-41 silica which contained large amounts of Si–OH (Hu et al., 2014), and iron doping strengthened the surface hydrophobicity. More importantly, CPX with the numerous reactive functional groups like carboxyl, keto, and piperazine could form complexes with majority of metal cations (Turel, 2002; Sun et al., 2014). As seen in **Figure 5**, the interactions between CPX and Fe-MCM-41 were investigated by comparing the FT-IR spectra of CPX, fresh and used Fe-MCM-41 in the range of 1250–1800 cm−<sup>1</sup> , which included the main characteristic peaks (Full scale FT-IR curve could be found in Figure S2 in Supplementary Information). The band positions and their corresponding band assignments were listed in **Table 1** (Gu and Karthikeyan, 2005; Trivedi and Vasudevan, 2007; Wang et al., 2011; Peng et al., 2012). The position of C=O stretching vibration of pyridine-keto upshifted slightly to a higher frequency of 1631.4 cm−<sup>1</sup> after adsorption and the 9.4 cm−<sup>1</sup> shift indicated reinforcement of the C=O bond. That was ascribed to the release

of intramolecular hydrogen bond between the ortho substituted carboxylate group and Fe-MCM-41 which was an indirect evidence for ferric-ciprofloxacin interactions. Meanwhile, the intensity of this peak was much weaker, suggesting that the fraction of pyridine-keto group interacted with Fe-MCM-41. The most significant feature of CPX adsorbed on Fe-MCM-41 was the complete disappearance of the peak at 1698.7 cm−<sup>1</sup> , which indicated that all the carboxyl groups participated in the binding reaction with Fe-MCM-41. The band shifting from 1382.8 to 1384.2 cm−<sup>1</sup> demonstrated that electrostatic interactions were occurred between the negative sites on the absorbent surface and the protonated amine group. The shift exhibited by the band between 1445.8 to 1474.6 cm−<sup>1</sup> meant the C–N bond of amine group also supported this assumption. Overall, it was almost certain that iron doping added the sites of adsorption

FIGURE 4 | Adsorption of CPX on Fe-MCM-41(x) (x = 20, 40, 80, and 160, x was denoted as molar ratio of Si to Fe; volume of CPX solution: 200 mL; rotating speed: 175 rpm/min; initial pH value: 5.40; equilibrium time: 120 min).

(see **Graphical Abstract**). The complex was present through the cation bridging of Fe(III) with the adjacent carbonyl oxygen and the oxygen from the deprotonated carboxylate group. The other absent peaks after adsorption might show signs of other adsorption affinity, for instance, hydrophobic effect and π-π interaction (discussed later). Fe-MCM-41 (20) was chosen as the adsorbent for the subsequent research.

# Influence of Adsorbent Dose and Kinetics of Adsorption

**Figure 6** illustrated the influence of adsorbent dose on the uptake of CPX. It was displayed that the removal efficiency

TABLE 1 | FT-IR band positions (cm−<sup>1</sup> ) and suggested assignments for CPX and CPX-adsorbed samples.


υ*, stretching;* δ*, bending.*

increased with adsorbent dose. This behavior could be explained by adsorption mainly occurring at active sites when the adsorbent dose was large. The high removal efficiency for CPX reflected the satisfactory textural and structural properties of the adsorbent. Therefore, 40 mg adsorbent dose was chosen for the following study.

The kinetics could be employed to evaluate the performance of the adsorbent and give the underlying adsorption mechanisms. Three available kinetic models were examined as follows:

Pseudo-first-order kinetic equation:

$$\frac{1}{q\_t} = \frac{k\_1}{q\_e} \times \frac{1}{t} + \frac{1}{q\_e} \tag{1}$$

Pseudo-second-order kinetic equation:

$$\frac{t}{q\_t} = \frac{t}{q\_e} + \frac{1}{k\_2 \times q\_e^2} \tag{2}$$

Where q<sup>e</sup> and q<sup>t</sup> were the adsorption amounts at equilibrium and time t, respectively. k<sup>1</sup> and k<sup>2</sup> were defined as reaction rate constants.

Intra-particle diffusion model:

$$q\_t = k\_i \times t^{\frac{1}{\sqrt{2}}} + c\_i \tag{3}$$

Where k<sup>i</sup> was defined as rate constant at stage i, k<sup>i</sup> could be obtained by plotting q<sup>t</sup> vs. t 1/2 .

The avidity of three kinetic models was analyzed by fitting straight lines in **Figures 7A–C**. **Table 2** listed the kinetics parameters for three reaction orders. It was obvious that the experimental data of CPX adsorbed by Fe-MCM-41 were more accurately simulated by pseudo-second-order model with higher correlation coefficients (R 2 ) value. Also, based on the second-order model, the calculated q<sup>e</sup> (qe,cal) was a good agreement with the experimental q<sup>e</sup> (qe,exp) values, besides q<sup>e</sup> (qe,exp) increased from 20.12 to 95.24 mg g−<sup>1</sup> with Fe content. Compared the various adsorption capacities of CPX with those reported adsorbents such as Graphene oxide/calcium alginate (GO/CA) (Wu et al., 2013), Carbon nanofibers (Li et al., 2015), Goethite (MacKay and Seremet, 2008) and Illite (Wang et al., 2011), H2Ti2O5·H2O (Wu et al., 2014), and Fe3O4/C (Shi et al., 2013) as showed in **Table 3**, Fe-MCM-41 showed better adsorption property which required a shorter time to reach adsorption equilibrium and possessed a higher adsorption capacity.

The models discussed above were not able to describe the phenomena of intraparticle or pore diffusion, which were often the rate-limiting steps in an adsorption process. With the intraparticle diffusion model, **Figure 7C** exhibited that the adsorption plots of adsorbents were not linear and could be divided into three linear regions, denoting that the multi stages of the intraparticle adsorption (Ghaedi et al., 2012). The first-stage portion was sharper and did not pass through zero point, which was due to the diffusion of adsorbents through the bulk solution to the external surface. It demonstrated that particle diffusion was not only the sole rate-limiting step, but also controlled by boundary layer in the initial phase of the adsorption. These results could be validated by further analyzing data with Boyd's mode which was expressed as follows:

$$F = 1 - \frac{6}{\pi^2} \exp(-Bt) \tag{4}$$

$$F = \frac{q\_t}{q\_e} \tag{5}$$

$$Bt = -0.4977 - \ln(1 - F) \tag{6}$$

model and (D) Boyd plots for CPX adsorption at different temperatures.




TABLE 3 | Adsorption capacities of CPX on various adsorbents.


Where q<sup>t</sup> and q<sup>e</sup> were the adsorption amounts at any time t and equilibrium, F given by Equation (5) was the fraction at time t, B<sup>t</sup> denoted a mathematical function of F.

The linearity of the plot was applied to differentiate the controlled rates of adsorption between external transport and intra-particle diffusion. The plot in **Figure 7D** were linear and did not move toward the zero point, demonstrating that external transfer governed adsorption process at the start and then the intra-particle diffusion.

The second-stage portion became slower than first stage due to the weak adsorption between CPX and the surface atoms of the solid. The third-stage portion tended to ease up and reached final equilibrium. It could be attributed to the smaller pores, lower concentration of CPX and enhanced electrostatic repulsion (between CPX and the adsorbents surface). What's more, comparisons of the values of k<sup>i</sup> also proved that the adsorption rate became slower during the adsorption process. It concluded that the intraparticle diffusion was present while some other mechanisms might be involved.

# Effect of the Temperature

The adsorption isotherm in **Figure 8** described the equilibrium relationships between the liquid-phase CPX concentration and the amount of CPX on the Fe-MCM-41. The adsorption capacities of the CPX increased with temperature, indicating that adsorption process on Fe-MCM-41 was favored at higher temperatures. The phenomenon could be ascribed to promote mobility of the CPX molecules in solution and increased new active sites on the Fe-MCM-41 for adsorption (Ghaedi et al., 2012). The results also suggested that it was an endothermic process in nature.

For further examining the impact of temperature on adsorption process, thermodynamic parameters was determined at three temperatures (Ao et al., 2009). As listed in **Table 4**, the adsorption parameters were fitted by the most frequently utilized adsorption isotherms models named Langmuir and Freundlich models at 293, 303, and 313 K. It was apparent that the Langmuir isotherm better represented the experimental data with higher value of the determination coefficient (R <sup>2</sup> = 0.9459) at 293 K while Freundlich model yielded the best fit at temperatures 303 K (R <sup>2</sup> = 0.9544) and 313 K (R <sup>2</sup> = 0.9675). This implied monolayer adsorption with a value of 133.33 mg g−<sup>1</sup> obtained for qmax and evenly distribution of the adsorption sites at low temperature. However, as temperature rose, the Freundlich equation became even more accurate to assess the adsorption process. It was possible to deduce that the thermal treatments activated more active sites for adsorption as described above. Molecules of CPX retained two monolayer adsorptions when the pore diameter of Fe-MCM-41 was greater than molecule diameter of CPX, and despite reversible adsorption occurred meanwhile (Song et al., 2013; Wang et al., 2017a). This non-linear adsorption also demonstrated that the adsorption process was the result of interaction of various forces. The increased values of K<sup>F</sup> with temperature demonstrated the endothermic adsorption process. The n-values were more than 1, revealing a beneficial adsorption system and heterogeneity of Fe-MCM-41 adsorption sites (Wu et al., 2017).

As listed in **Table 5**, the thermodynamic parameters, such as enthalpy change (1H), entropy change (1S), and the Gibb's free energy (1G), were evaluated to obtain insights into the changes of adsorption processes using the Van't Hoff equation (Jiang et al., 2013). The value of Kd, an equilibrium constant, was obtained by dividing q<sup>e</sup> and Ce. The thermodynamic parameters of adsorption could be determined by using Equations (7) and (8):

$$
\Delta G = \ -R \times T \times \ln K\_d \tag{7}
$$

$$
\ln K\_d = \frac{\Delta S}{R} - \frac{\Delta H}{R \times T} \tag{8}
$$

TABLE 4 | Adsorption isotherms constants and correlation coefficients.

rpm/min; initial pH-value: 5.40; equilibrium time: 120 min).

model fits to the data (volume of CPX solution: 200 mL; rotating speed: 175


TABLE 5 | Thermodynamic parameters for the adsorption under different temperature.


The negative 1G-values were obtained in the range of 293–313 K, revealing the feasible and spontaneous of adsorption process. The positive 1H-value demonstrated an endothermic process while the positive value of 1S indicated that the organization of the CPX molecular at the solid- liquid interface became more random than low temperature.

# Effect of pH on the Adsorption of CPX

The pHpzc of Fe-MCM-41(20) was estimated to be around 5.4, the surface charge of Fe-MCM-41(20) remained negative over the range of pH (5.4 < pH < 11.0) in this study. Therefore, the speciation of CPX molecules were regarded as the mainly factor to this pH range (5.4 < pH < 11.0) influence on the adsorption level. **Figure 9** indicated the pH dependency of CPX adsorption on Fe-MCM-41(20). As seen there, a bell-shaped adsorption envelop was noticed. Initially, adsorption increased with pH from 5 to 10 and reached to a maximum value. At pH values >10, adsorption decreased sharply with increasing pH. The acid dissociation constants pKa1 and pKa2 values of CPX were 6.1 and 8.7, respectively. Most of CPX molecular were in cationic form with protonated amine group in the piperazine moiety when pH was less than 6.1. Cationic CPX generated electrostatic attraction with negatively charged Fe-MCM-41(20) on the surface, accordingly promoting the adsorption. As displayed in **Figure 9B**, a great amount of Fe leached and the structure of adsorbent collapsed at strong acidic environment (pH < 4), which greatly reduced the adsorption ability. When 6.1 < pH < 8.7, CPX removal efficiency increased with its zwitterion form. The zwitterionic CPX was least soluble at pH 7.5 in water for its neutral charge (Roca Jalil et al., 2015). In other word, it could be assumed that hydrophobic effect had played an important role in the adsorption process. When pH value was above 8.7, the anionic form could induce the electrostatic repulsion between CPX and Fe-MCM-41. When 8.7 < pH < 10, another force might exist to balance the hydrophobic effect and electrostatic repulsion which suggested as π-electron-donor– acceptor (EDA) interaction between CPX and Fe-MCM-41(20). The aromatic ring could serve as π-electron acceptors due to the strong electron-withdrawing ability of fluorine group on the benzene ring of CPX. In this, the hydroxyl groups on adsorbent surface acting as electron-donors interacted strongly with the π-acceptor compound. When pH > 10, a drastic decrease of CPX adsorption amount was due to the strong electrostatic repulsion effect. Therefore, strong chemical interactions were of vital importance for CPX adsorption on Fe-MCM-41(20).

# The Effect of Metal Cations

Metal cations with relatively high concentration could impact CPX adsorption. The presence of Cu(II) facilitated the adsorption of CPX on montmorillonite at pH > 6.0 (Pei et al., 2010). Similar conclusion was also given on activated carbon at pH 3.4–6.5 (Sun et al., 2014). As depicted in **Figure 10**, all metal cations suggested a significantly decline to the adsorption property of CPX, whether it be Ca represented alkaline earth element or those metals which were common divalent metals (Cu, Ni, Pb, and Cd) in water. The influence of five metal cations on removal efficiency followed a decreasing order of Cu (52.3%) > Ni (49.0%) > Pb (28.3%) >Cd (21.9%) >Ca (18.7%). This phenomenon could be assumed that metal cations screened negative charged sites of adsorbent surface. Metal cations had a relatively strong affinity to CPX. In order to make a thorough exploration about the complexing ability of five metal ions, the complexation stability constants were listed in **Table 6** (combined with the prior results). Some conclusions can be drawn from the values. On the basis of the liberation degrees of CPX, which complexed with metal cations mainly in three forms: [M(AH)]2+, [M(AH)2] <sup>2</sup>+, [M(AH)A]+. Because of the solution under neutral conditions, [M(AH)]2<sup>+</sup> was the main form. Compared to other divalent metals, the complexation ability of CPX with Cu was much stronger, and the result agreed with the heavy metals which exhibited different removal efficiency.

### TABLE 6 | Complexation constants of CPX with metals found in the literature.


to the damage of adsorption sites. Therefore, Fe-MCM-41 could be recycled as an efficient absorbent for practical application.

# CONCLUSIONS

Fe-MCM-41s prepared from a hydrothermal process were used as adsorbents to adsorb the CPX in aqueous solutions. Experimental results described that Fe-MCM-41 with Si/Fe = 20 exhibited the best CPX removal efficiency. The XRD and TEM results confirmed that the hexagonal mesoporous structure maintained after iron doping. A combination of multiple effects including electrostatic interaction, surface complexation, hydrophobic effect and π-π interaction on the adsorption process were confirmed under different reaction conditions. The experimental kinetic data showed that the adsorption was a chemical-controlling and multi-step process, Fe-MCM-41 showed the better adsorption property. Strong acidic/alkaline environment was not conducive to the adsorption process. The adsorption performed a higher adsorption affinity at higher temperature was a spontaneous exothermic process. Moreover,

# Regeneration of the Adsorbent

It was important for an adsorbent with good reuse performance in practical application. According to the investigation of pH influence, the adsorption capacity of CPX on Fe-MCM-41 was negligible when the pH was above 10. It indicated that of CPX could desorbed from the absorbent surface using 0.1 M NaOH solution as the desorption agents possibly, then the adsorbent was filtered, washed and dried at 358 K. The regenerated adsorbent was tested by four cycles of adsorption/desorption process under the same conditions. **Figure 11** presented the removal efficiency of CPX on the recycled Fe-MCM-41, exhibiting slightly decline with a loss of 23.6% after four cycles, which might due

due to the depression of Me-CPX complex, metal cations decreased CPX adsorption on Fe-MCM-41 surface. As the same order as the complexation stability constants of CPX,

# REFERENCES


Cu impacted CPX adsorption most. Fe-MCM-41 exhibited stable performances for 4 cycles use without deterioration which could possibly applied to wastewater treatment including POPs.

# AUTHOR CONTRIBUTIONS

YW: Conducted all the experiments; YT: Supervised the project; LL: Revised the manuscript; PL and XL: Provided assistance in setting up the gears; WC and YX: Helped analysis the data.

# ACKNOWLEDGMENTS

This work was supported by the project of the Science Technology Office of Guangzhou (201607010276), China Postdoctoral Science Foundation (2016M602483) and the Ph.D. start-up Fund of Natural Science Foundation of Guangdong Province, China (2017A030310420).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2018.00017/full#supplementary-material


from pseudomorphic transformation - A minimal input data approach based on excess surface work. Micropor. Mesopor. Mater. 240, 169–177. doi: 10.1016/j.micromeso.2016.11.017


**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 © 2018 Wu, Tang, Li, Liu, Li, Chen and Xue. 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 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.

# First Principles Study on the CO Oxidation on Mn-Embedded Divacancy Graphene

### Quanguo Jiang<sup>1</sup> , Jianfeng Zhang<sup>1</sup> \*, Zhimin Ao<sup>2</sup> \*, Huajie Huang<sup>1</sup> , Haiyan He<sup>1</sup> and Yuping Wu<sup>1</sup>

*<sup>1</sup> College of Mechanics and Materials, Hohai University, Nanjing, China, <sup>2</sup> Guangzhou Key Laboratory of Environmental Catalysis and Pollution Control, Institute of Environmental Health and Pollution Control, School of Environmental Science and Engineering, Guangdong University of Technology, Guangzhou, China*

The CO oxidation mechanism on graphene with divacancy (DG) embedded with transition metal from Sc to Zn has been studied by using first principles calculations. The results indicate that O<sup>2</sup> molecule is preferentially adsorbed on Sc, Ti, V, Cr, Mn, and Fe-DG, which can avoid the CO poisoning problem that many catalysts facing and is beneficial to the CO oxidation progress. Further study indicates that Mn-DG shows the best catalytic properties for CO oxidation with consideration of both Langmuir-Hinshelwood (LH) and Eley-Rideal (ER) oxidation mechanisms. Along the ER mechanism, the reaction energy barrier for the first step (CO free + O2 pre-adsorbed → OOCO) is 0.96 eV. Along the LH mechanism, the energy barrier for the rate limiting step (CO adsorbed + O2 adsorbed → OOCO) is only 0.41 eV, indicating that the CO oxidation on Mn-DG will occur along LH mechanism. The Hirshfeld charge distributions of O<sup>2</sup> and CO molecules is tuned by the embedded Mn atom, and the charge transfer from the embedded Mn atom to the adsorbed molecules plays an important role for the CO oxidation. The result shows that the Mn-embedded divacancy graphene is a noble-metal free and efficient catalyst for CO oxidation at low temperature.

### Keywords: graphene, divacancy, Mn-embedded, CO oxidation, first principles calculations

# INTRODUCTION

Carbon monoxide (CO) is colorless, tasteless and toxic in air, while oxidation of CO is an efficient way to eliminate of the air pollutant (Xie et al., 2009). CO oxidation has important applications in atmosphere purification for hydrogen gas fuel in fuel cells as well (Qiao et al., 2015; Saavedra et al., 2016). Noble metals(Bleakley and Hu, 1999; Zhang and Hu, 2001; Liu et al., 2002; Gong et al., 2004; Dupont et al., 2006; Zhang et al., 2006; Liu, 2007) are common catalysts for the CO oxidation, where the rate limiting energy barriers are 0.46 eV (Liu et al., 2002) for Au(221), 0.91 eV (Gong et al., 2004) and 0.93 eV (Zhang and Hu, 2001) for Pd(111), 0.79 eV (Gong et al., 2004) and 0.82 eV (Dupont et al., 2006) for Pt(111), 1.17 eV (Gong et al., 2004) and 1.01 eV (Zhang et al., 2006) for Rh(111), 1.00 eV (Liu, 2007) for Rh(100). Due to the high cost and high reaction temperature of these noble metals, it is desirable to develop noble-metal-free catalysts for CO oxidation at low temperature. Noble metal clusters on supports are further studied to decrease the reaction barriers for CO oxidation (Tang et al., 2015; Ma et al., 2016; Wang et al., 2016; Ali et al., 2017; Chen et al., 2017). Furthermore, single atom catalyst decorated on appropriate matrix is attracted a lot of interests due to the excellent catalytic performance (Dvorák et al., 2016; Jones et al., 2016). The outstanding

### Edited by:

*Philippe Vernoux, UMR5256 Institut de Recherches sur la Catalyse et l'Environnement de Lyon (IRCELYON), France*

### Reviewed by:

*Xingang Li, Tianjin University, China Jose Luis Valverde, Universidad de Castilla-La Mancha, Spain*

### \*Correspondence:

*Jianfeng Zhang jfzhang@hhu.edu.cn Zhimin Ao zhimin.ao@gdut.edu.cn*

### Specialty section:

*This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry*

> Received: *06 January 2018* Accepted: *07 May 2018* Published: *29 May 2018*

### Citation:

*Jiang Q, Zhang J, Ao Z, Huang H, He H and Wu Y (2018) First Principles Study on the CO Oxidation on Mn-Embedded Divacancy Graphene. Front. Chem. 6:187. doi: 10.3389/fchem.2018.00187* physical properties (Novoselov et al., 2005; Balandin et al., 2008; Lee et al., 2008) and the large surface-to-volume ratio make graphene (Novoselov et al., 2004) to be a promising substrate to realize high performance single atom catalysis. However, the inert nature of pristine graphene usually causes clustering problems for the adsorbed metal atoms (Liu and Huang, 2017; Liu et al., 2017). The interactions between the pristine graphene and the adsorbed atoms can be enhanced by introducing different carbon vacancies. Experimentally, single carbon vacancy (SV) and double carbon vacancies (DV) are common point defects on graphene. Vacancies can be introduced into graphene by exposing it to the focussed electron beam, and the vacancy defects in graphene can be tailored by controlling the exposing time, where the DV is more often observed than SV due to the high energy of the SV under electron beam irradiation (Robertson et al., 2012). In addition, the concentration of DV changes from 0.1 to 0.5 nm<sup>2</sup> by controlling the total electron beam dose (Robertson et al., 2012). In addition, it is reported that the metal atoms resident on the SV and DV positions in graphene surfaces are stable in comparison to that on graphene edge (Robertson et al., 2013). Hence, it is important to evaluate the catalytic performance of a single atom supported on graphene with both SV and DV defects.

Many works about the CO oxidation have been reported for single atom decorated at the single carbon vacancy on graphene. Theoretically, Au- (Lu et al., 2009), Fe- (Li et al., 2010), Cu- (Song et al., 2011), Pt- (Tang et al., 2012), Si- (Zhao et al., 2012), and Al-embedded (Jiang et al., 2014) SV graphenes show high activity for the CO oxidation. Although the SV graphene with decorating metal atoms has high catalytic activity, decorating metal atom on the controllable carbon vacancies in graphene to realize different single atom catalyst is still highly desirable. The decorated metal atom is threebond coordinated on SV graphene, while it is four-bond coordinated on DV graphene, indicating that the chemical activity of the decorated metal atom on different carbon vacancies should be different due to the different chemical environment. Furthermore, as mentioned above divacancy is commonly present in graphene obtained through chemical synthesis. However, only the catalytic activity of Fe-decorated DV graphene has been studied (Tang et al., 2016; Liu et al., 2017). Therefore, further research about the catalytic performance of DV graphene decorated with metal atom for CO oxidation is needed.

For the CO oxidation, Langmuir-Hinshelwood (LH) and Eley-Rideal (ER) are two mainly mechanisms. Along the ER mechanism, the activated O<sup>2</sup> molecule directly reacts with the free CO molecule, where the activation of O<sup>2</sup> is the rate-limiting step (Lu et al., 2009). Alone the LH mechanism, the CO and O<sup>2</sup> molecules are first coadsorbed, and then react to form OOCO intermediate, which is the rate-limiting step for the oxidation progress (Lu et al., 2009). In general, the reaction energy barriers are proportional to the adsorption energy of adsorbed molecules on supported catalyst (Gong et al., 2004), indicating that the adsorption energy of adsorbed CO and O<sup>2</sup> molecules could be a benchmark for the catalytic performance of graphene for CO oxidation. In addition, a larger adsorption energy for O<sup>2</sup>

molecule than that of CO is desired during the CO oxidation progress, because the preferential adsorption of CO will block the active sites and prevent the continuous oxidation reaction (Jiang et al., 2014; Tang et al., 2016). Therefore, the adsorption energy for O<sup>2</sup> and CO on transition metals doped graphene is first calculated and the alternative decorating atom is chosen based on this rule.

In this work, by using first principles calculations, we will systematically study the CO oxidation mechanism on DV graphene decorated with transition metals from Sc to Zn, which are non-noble metals and commonly used to decorate two dimensional materials. The reaction barriers for each step are analyzed and the corresponding reaction mechanisms are discussed through analyzing the electronic property of the graphene systems.

# Calculation Methods

The density functional theory (DFT) calculations are carried out by using Dmol<sup>3</sup> package (Delley, 2000). Exchange-correlation functions are taken as generalized gradient approximation (GGA) with Perdew-Burke-Ernzerhof (PBE) (Perdew et al., 1996). The selection of exchange–correlation functional has evidential effect on the result of adsorption energies, while has much smaller effect on the reaction energy barriers (Roldán et al., 2009). DFT semicore pseudopotentials (DSPPs) core treatment is implemented for relativistic effects. Double numerical plus polarization (DNP) is employed as the basis set. The convergence tolerance of energy of 10−<sup>5</sup> Hartree is taken (1 Hartree = 27.21 eV), and the maximal allowed force and displacement are 0.002 Hartree/Å and 0.005 Å, respectively. Linear synchronous transit/quadratic synchronous transit (LST/QST) (Halgren and Lipscomb, 1977) and nudged elastic band (NEB) (Henkelman and Jonsson, 2000) tools in Dmol<sup>3</sup> module are used to investigate the minimum energy pathway for CO oxidation on graphene, which have been well validated to determine the transition state. Three-dimensional periodic boundary conditions are taken in the simulation. The simulation cell consists of a 4 × 4 graphene supercell with a vacuum width of 20 Å above the graphene layer to minimize the interlayer interaction. The kpoint is set to 5 × 5 × 1, and all atoms are allowed to relax according to previous reports (Jiang et al., 2014). After structure relaxations, the density of states (DOS) are calculated with a finer k-point grid of 15 × 15 × 1. The DFT+D method within the Grimme scheme (Grimme, 2006) is used in all calculations to consider the van der Waals forces. The electron orbits of the free and adsorbed molecules are calculated with CASTEP code (Segall et al., 2002), where the ultrasoft pseudopotentials, GGA-PBE functional, an energy cutoff of 340 eV and 5 × 5× 1 kpoint meshes are used. We have compared the total energy of the graphene system with different spin state, and choose the proper spin state for the graphene system with the smallest total energy.

The adsorption energy Ead of molecules on graphene is determined by,

$$E\_{\rm ad} = E\_{\rm molecules/graphene} - \left(E\_{\rm graphene} + E\_{\rm molecules}\right) \tag{1}$$

where Emolecules/graphene, Egraphene, and Emolecule are total energies of the adsorbed graphene system, the isolate graphene and molecules respectively.

# RESULTS AND DISCUSSION

# Adsorption of O<sup>2</sup> and CO on TM-DG

Based on the literature results in the introduction section, it is expected that transition-metal-embedded divacancy graphene (TM-DG) could also exhibit excellent catalytic behaviors for CO oxidation, similar to the cases of graphene with single vacancy systems. Herein, we present systematic DFT calculations on the adsorption energy of O<sup>2</sup> and CO molecules on the divacancy graphene embedded with transition metal (from Sc to Zn). To comprehensively understanding the adsorption behaviors of O<sup>2</sup> and CO molecules on transition metal decorated divacancy graphene, the adsorption energies of O<sup>2</sup> and CO, as well as the co-adsorption energy of O<sup>2</sup> and CO molecules on TM-DG are shown in **Figure 1**, where we can see that the adsorption energy of O<sup>2</sup> molecule is larger than that of CO molecule on TM-DG (from Sc to Fe), while the adsorption energy of CO molecule is larger than that of O<sup>2</sup> molecule on TM-DG (from Co to Zn). This result indicates that the O<sup>2</sup> molecule has priority during the adsorption progress and avoids the CO poisoning at the active sites on TM-DG (from Sc to Fe) during the CO oxidation. In addition, the co-adsorption energy of O<sup>2</sup> and CO molecules on TM-DG (from Sc to Fe) is also shown in **Figure 1**, where a local adsorption energy minimum on Mn-DG is found, which indicates that Mn-DG can facilitate the CO oxidation better, due to the fact that the energy barrier is proportional to the adsorption energy of molecules (Gong et al., 2004). Therefore, CO oxidation on Mn-DG is mainly studied in the following, and the CO oxidation on other TM-DG (Sc, Ti, V, Cr and Fe) is also studied for comparison purpose.

# Electronic Properties and Stability of Mn-DG

Before further study the catalytic properties of Mn-DG, the electronic properties and structure stability of Mn-DG are first studied in **Figure 2**. After embedding Mn atom into the divancancy graphene through substituting two carbon atoms, the reconstructed structure of graphene is shown in **Figure 2A**, where four chemical Mn-C bonds are formed in graphene with bond length lMn−<sup>C</sup> = 1.99 Å. Mn atom is out of the plane of graphene with a distance of 0.68 Å due to the larger atomic radius of 1.79 Å compared with that of C 0.91 Å. The adsorption energy of the Mn atom on the divacancy graphene is −6.81 eV. The Hirshfeld charge distributions near the dopant are also given in **Figure 2A**, where the electron-deficiency position is formed for the Mn atom with losing electrons of 0.284 e, which promotes the adsorption of O<sup>2</sup> and CO molecules. The differential charge density along C-Mn-C bonds for Mn-DG is further studied and is shown in **Figure 2B**, where the blue and red isosurfaces correspond to the increase in the number of electrons and the depletion zone, respectively. It shows that electrons accumulate near the doped Mn atom, indicating the high chemical active area.

The aggregation problems for the adsorbed metal atoms on substrate are significant for the catalytic performance, especially when the concentration of metal atom is high (Ao and Peeters, 2010a). To determine the possibility of aggregation for Mn atoms on divacancy graphene, the diffusion pathway of Mn atom to its neighboring positions is investigated based on DFT calculations (see **Figure 2C**), where the corresponding diffusion energy barrier for the decorated Mn atom is 5.91 eV. It is claimed that a surface reaction will occur when the reaction barrier is smaller than the critical value of Ecbar = 0.91 eV (Young, 2001), the decorated Mn atom on divacancy graphene is thus stable. In addition, the adsorption energy of Mn atom on divacancy graphene is −6.81 eV and it is much larger than the cohesive energy −2.92 eV/atom for Mn element (http://www. knowledgedoor.com/). Therefore, the Mn decorated divacancy graphene is quite stable without aggregation problems. Partial density of states (PDOS) are further analyzed to confirm the enhanced interactions between the Mn atom and graphene (**Figure 2D**), where the energy bands between the decorated Mn atom and carbon atoms overlap significantly. The electrondeficiency character is confirmed by the Fermi level crossing the valence band, and it is consistent with the Hirshfeld charge distributions. As shown in the insert of **Figure 2D**, the positive and negative spin density of Mn-DG is shown in green and red, respectively. Therefore, the magnetic moment of Mn-DG is 3.053 µB, which is mainly contributed by the Mn atom. The remaining unsaturated d orbital of Mn atom is reactive, which can adsorb small molecules and promote the subsequent reactions.

# Adsorption of Molecules on Mn-DG

To investigate the oxidation of CO on Mn-embedded divacancy graphene, the adsorptions of O<sup>2</sup> and CO on Mn-DG are studied carefully. **Figure 3A** shows the most stable configuration for O<sup>2</sup> molecule adsorbed on Mn-DG (Ead = −1.96 eV), where the

O-O bond is parallel to the graphene sheet, and ∼0.18 e is transferred from Mn-DG to O<sup>2</sup> molecule based on Hirshfeld charge analysis. To assess the stability of the adsorbed O<sup>2</sup> molecule, the dissociative adsorption of O<sup>2</sup> molecule on Mn-DG is then studied in **Figure 3B**. After NEB calculations, the dissociation reaction barrier for an O<sup>2</sup> molecule on the Mn-DG is 1.18 eV > Ecbar = 0.91 eV (Young, 2001), which indicates that the adsorbed O<sup>2</sup> molecule prefers to stay on Mn-DG in molecular state at room temperature. The PDOS of free O<sup>2</sup> molecule is shown in **Figure 3C**, where the 2π∗ anti-bond orbital is half filled (Honkala and Laasonen, 2000). Then, the PDOS of Mn atom and the adsorbed O<sup>2</sup> molecule is shown in **Figure 3D**, where all orbitals of the adsorbed O<sup>2</sup> molecule are also labeled to understand the interaction between the

adsorbed O<sup>2</sup> and Mn-DG. About 0.18 e is transferred from Mn-DG to the adsorbed O<sup>2</sup> molecule based on Hirshfeld method, which occupies the O2-2π∗ orbital above the Fermi level (see **Figure 3C**) and is confirmed by the new peaks for O2-2π∗ orbital below the Fermi level (see **Figure 3D**). This charge transfer elongates the O-O bond from 1.23 Å in free O<sup>2</sup> to 1.40 Å in adsorbed O<sup>2</sup> molecule. The activated O<sup>2</sup> molecule with longer O-O bond will be beneficial for the subsequent CO oxidation. The O2-2π∗ orbitals and Mn atom is strong hybridized near the Fermi level (see **Figure 3D**), which mainly responses for the chemical adsorption of O<sup>2</sup> molecule on Mn-DG.

The adsorption configuration of a CO molecule on Mn-DG is shown in **Figure 4A**, where the CO molecule is vertically

adsorbed on the top of the decorated Mn atom. CO is chemically adsorbed on Mn-DG, which is confirmed by the chemical bond between Mn atom and the carbon atom of CO molecule. The adsorption energy of a CO molecule on Mn-DG is Ead = −1.70 eV, and CO molecule loses 0.006 e to the Mn-DG. The binding energy E<sup>b</sup> of C-O bond is 11.57 eV, which is much larger than E<sup>b</sup> = 6.36 eV for O-O bond based on DFT calculations. Thus CO should be more difficult to be dissociated, which is confirmed by the fact that the dissociative energy barrier for CO molecule on Mn-DG is 5.55 eV based on DFT calculation as shown in **Figure 4B**. The PDOS of free CO molecule is shown in **Figure 4C**, where all orbitals are labeled. **Figure 4D** shows the PDOS of the adsorbed CO molecule on Mn-embedded divacancy graphene, where the orbitals of adsorbed CO molecule are labeled and displayed. The 5σ peak of the CO molecule adsorbed on Mn-DG is significantly depressed than the free CO molecule due to the charge transfer. Although the 2π∗ anti-bond orbital far above Fermi level for free CO molecule is fully empty, Mn atom transfers some electrons to CO-2π∗ orbital due to the fact that CO-2π∗ is close to the Fermi level at absorbed state, which slightly elongates the C-O bond from 1.14 Å for free CO molecule to 1.16 Å for adsorbed CO molecule. In addition, CO molecules will act as a donor with the carbon atom near the graphene surface, dur to fact that the CO-5σ orbital locates on the carbon atom (Leenaerts et al., 2008). This agrees with the above Hirshfeld analysis result. The above discussions indicate that O<sup>2</sup> and CO have strong interactions with Mn-DG (corresponding adsorption energies are −1.96 and −1.70 eV, respectively), but the adsorption of O<sup>2</sup> is much stronger. The O<sup>2</sup> molecule is activated on Mn-DG, which will facilitate the CO oxidation progress on graphene.

# CO Oxidation on Mn-DG

Two reaction mechanisms have been established for the oxidation of CO molecule: Langmuir-Hinshelwood (LH) mechanism and Eley-Rideal (ER) mechanism (Molina and Hammer, 2005; An et al., 2008; Lu et al., 2009; Li et al., 2010; Song et al., 2011; Tang et al., 2012; Zhao et al., 2012; Jiang et al., 2014). For the ER mechanism, the O<sup>2</sup> molecule is first adsorbed and activated by the Mn-DG, then a free CO molecule approaches the substrate to form an intermediate product. Along the LH mechanism, the O<sup>2</sup> and CO molecules first co-adsorb on the Mn-DG, and then form an intermediate product. Since O<sup>2</sup> has a larger adsorption energy (−1.96 eV) on Mn-DG than that of CO (−1.70 eV), the adsorption of O<sup>2</sup> on Mn-DG has higher priority, thus the ER mechanism for the CO oxidation seems to be favorable. However, the lower co-adsorption energy for O<sup>2</sup> and CO molecules (−2.06 eV) indicates that O<sup>2</sup> and CO may co-adsorb on Mn atom as discussed in literatures (Lu et al., 2009; Song et al., 2011; Tang et al., 2012; Jiang et al., 2014). Therefore, both mechanisms for CO oxidation will be discussed in the following.

In order to search the preferred reaction path for CO oxidation on Mn-DG, the first reaction step along both LH and ER mechanisms is studied in **Figure 5**. If the LH mechanism is more favorable, the free CO molecule will co-adsorb on Mn atom with pre-adsorbed O<sup>2</sup> molecule after overcoming a small reaction barrier. If the ER mechanism is more favorable, the free CO molecule will react with the O<sup>2</sup> molecule to form an intermediate (OOCO). The structure of physisorbed CO molecule on the Mn-DG with pre-adsorbed O<sup>2</sup> molecule was selected as the IS after studying all adsorption configurations as shown in **Figure 5A**. The reaction profile for the co-adsorption of CO and O<sup>2</sup> molecules on Mn atom is shown in **Figure 5A**, where the CO and O<sup>2</sup> are titled toward the graphene surface at TS state. After overcoming an energy barrier of 0.19 eV, the CO and O<sup>2</sup> coadsorb on Mn-DG with releasing energy of 0.02 eV, as shown in **Figure 5A**.

**Figure 5B** shows the reaction profile for intermediate product along the ER mechanism. When the free CO approaches the Mn atom, the O-O bond is broken (see TS in **Figure 5B**). After overcoming a relative large energy barrier of 0.96 eV, the C atom from the CO binds with two O atoms from the dissociation of the O<sup>2</sup> molecule to form a OOCO complex over the Mn atom (see FS in **Figure 5B**). This exothermic process releases an energy of 3.39 eV.

The effect of temperature on the energy barriers is considered. The free energy change (1G) between the reactant and transition state is considered as the temperature-dependent energy barrier E ′ bar and 1G = 1H - T1S, where 1H is the enthalpy change, 1S is the entropy change, and T is the room temperature (298.15 K). In addition, 1H = (1U + P1V), 1U = (1Etot + 1Evib + 1Etrans + 1Erot) and 1S = 1Svib + 1Strans + 1Srot, where 1U is the internal energy change, 1Etot is the change of total electronic energy, the vib, trans and rot indicate vibration, translation and rotation, respectively, which can be obtained through calculations of vibrational frequency. The temperaturedependent energy barriers at 298.15 K in **Figure 5** are: E ′ bar = 0.33 eV in **Figure 5A** and E ′ bar = 1.15 eV in **Figure 5B**. This indicates that the free CO molecule will desorb from substrate with increasing temperature, and the reaction is more difficult to occur. The co-adsorption of CO and O<sup>2</sup> on Mn-DG in **Figure 5A** will happen at 298.15 K due to the lower reaction barrier than the critical barrier of Ecbar = 0.91 eV (Young, 2001), thus the LH mechanism is mainly studied for the CO oxidation in the following.

For oxidation of CO on the Mn-DG along LH mechanism, several steps and also intermediate products (MS) for the oxidation procedure exist (Lu et al., 2009; Song et al., 2011; Tang et al., 2012; Jiang et al., 2014). For each step, e.g., from initial

state to intermediate state in **Figure 6**, a transition state also exists. The configuration of co-adsorbed CO and O<sup>2</sup> molecules on Mn-DG is taken as the reactant (IS in **Figure 6A**) based on the above discussions. At transition state, one O-Mn bond changes from 1.93 to 2.10 Å and a C-O bond between CO and O<sup>2</sup> is formed. After overcoming an energy barrier of 0.41 eV, an OOCO intermediate (MS in **Figure 6A**) is formed and the elongated O-Mn bond is cleaved. Then the first CO<sup>2</sup> molecule (MS2) is formed on Mn-DG after overcoming an energy barrier

for the oxidation of CO molecule on Mn-embedded graphene for ER mechanism.

of 0.15 eV (see **Figure 6A**), where the CO<sup>2</sup> is physically adsorbed on Mn-DG and its adsorption energy is −0.20 eV. The reaction for this step can release a heat of 3.81 eV, which can sufficiently overcome the adsorption energy of CO2, and the first produced CO<sup>2</sup> molecule would desorb from the Mn-DG efficiently. The O-O bond for OOCO configuration in MS is broken at TS2, and a CO<sup>2</sup> molecule with bond angle of 126.6◦ is formed. **Figure 6B** shows the formation of the second CO<sup>2</sup> molecule on Mnembedded divacancy graphene. The subsequent CO molecule

TABLE 1 | The reaction energy barriers of each step for the CO oxidation on divacancy graphene decorated with transition metal (from Sc to Fe) along the LH mechanism, where *E*bar1, *E*bar2, and *E*bar3 are the energy barriers for step 1, step 2, and step 3 during CO oxidation, respectively.


will react with the remaining O atom to produce CO<sup>2</sup> after surmounting an energy barrier of 0.17 eV (see **Figure 6B**), similar to the Au-embedded graphene (Lu et al., 2009). The reaction for this step releases a heat of 0.90 eV, which can also surmount the adsorption energy of CO<sup>2</sup> (−0.27 eV) in Mn-DG, and the second CO<sup>2</sup> molecule will desorb from Mn-DG efficiently. The reaction profile for the production of the two CO<sup>2</sup> molecules is shown in **Figure 6**, where the rate limiting step is the formation of OOCO intermediate and the energy barrier of 0.41 eV is quite small. Therefore, the CO oxidation reaction along the LH mechanism releases 0.21 eV, 3.81 eV, and 0.90 eV for step 1, step 2, and step 3, respectively, which indicates the favorable thermodynamics for the CO oxidation on Mn-DG. In addition, the small energy barrier of 0.41 eV for the rate limiting step indicates the favorable kinetics for the CO oxidation on Mn-DG.

The energy barrier of each step for the CO oxidation along LH mechanism on divacancy graphene decorated with transition metals (from Sc to Fe) are also studied as shown in **Table 1**, where the rate limiting energy barriers are 1.15, 0.95, 0.85, 1.17, 0.41, and 0.88 eV for Sc-, Ti-, V-, Cr-, Mn-, and Fe-DG, respectively. The reaction barrier is the smallest on Mn-DG, which confirms the fact that the reaction barriers of molecules are proportional to the adsorption energy on supported catalyst (Gong et al., 2004), and the co-adsorption energy of adsorbed CO and O<sup>2</sup> molecules can be a benchmark for the catalytic performance of the graphene for CO oxidation. Therefore, CO can be oxidized easily on Mn-DG at low temperature, indicating that Mn-DG is an excellent candidate during the catalysts for CO oxidation.

The energy barriers at 298.15 K for each step of CO oxidation on Mn-DG in **Figure 6** are: E ′ bar = 0.49 eV for step 1, E ′ bar = 0.10 eV for step 2, and E ′ bar = 0.25 eV for step 3, where the energy barriers for step 1 and step 3 slightly increase than those

pathway of CO oxidation along LH mechanism (D).

state is also shown in (D). Red, blue and black curves represent PDOS of adsorbed O2, CO, and Mn (or Fe) atom, respectively. The orbitals of the adsorbed O2 and CO molecules are roughly labeled. Inset is the charge density of orbitals near the Fermi level. The vertical lines indicate the Fermi level.

at 0 K. The reaction time for each step in **Figure 6** is calculated by Arrhenius equation (Pauling, 1988)

$$\pi = \frac{1}{\text{\textquotedblleft}\left(\frac{-E\_{\text{bar}}}{K\_B T}\right)}\tag{2}$$

where K<sup>B</sup> is the Boltzmann constant, ν is in order of 10<sup>12</sup> Hz, and <sup>T</sup> <sup>=</sup> 298.15 K. <sup>τ</sup><sup>1</sup> <sup>=</sup> 1.9 <sup>×</sup> <sup>10</sup>−<sup>4</sup> s for step 1, τ<sup>2</sup> = 4.9 × <sup>10</sup>−<sup>11</sup> s for step 2, and <sup>τ</sup><sup>3</sup> <sup>=</sup> 1.7 <sup>×</sup> <sup>10</sup>−<sup>8</sup> s for step 3, respectively. Therefore, the CO oxidation on Mn-DG along LH mechanism has fast kinetics.

The possibility for the reversing reaction of CO oxidation, i.e., for the dissociation of the first and the second CO<sup>2</sup> molecule into CO molecule and O atom on Mn-DG are further considered as shown in **Figure 6**. The reversing energy barriers for the step 1, step 2, and step 3 based on DFT calculations are 0.62, 3.96, and 1.07 eV, respectively. After considering the effect of temperature, the reversing energy barriers at 298.15 K are: E ′ bar = 0.69 eV for step 1, E ′ bar = 4.05 eV for step 2, E ′ bar = 1.01 eV for step 3, while the corresponding reaction time is: τ<sup>1</sup> = 0.44 s for step 1, τ<sup>2</sup> = 2.3 × 10<sup>56</sup> s for step 2, and τ<sup>3</sup> = 1.1 × 10<sup>5</sup> s for step 3. The long reaction time indicates that the reversing reactions for the CO oxidation on Mn-DG are hardly to occur and the CO oxidation can be finished thoroughly.

To consider the effect of interactions between vacancies on the oxidation of CO, the graphene with more vacancies in the 4 × 4 supercell is studied as shown in **Figure 7**. The atomic structure of graphene with two divacancies embedded with Mn atoms is shown in **Figure 7A**, and the corresponding reaction pathway of CO oxidation for LH mechanism is shown in **Figure 7B**, where the rate limiting energy barrier is 0.65 eV for the formation of OOCO complex. The atomic structure of graphene with one divacancy and one single vacancy embedded with Mn atoms is shown in **Figure 7C**, and the corresponding reaction pathway of CO oxidation for LH mechanism is shown in **Figure 7D**, where the rate limiting energy barrier is 0.88 eV for the formation of OOCO complex. Therefore, the reaction barriers on graphene with two double vacancies or one single and one double vacancies in the supercell in **Figure 7** are both large than that on graphene with one double vacancy in **Figure 6**. Considering that the distances between the nearest Mn dopants in **Figures 7A**,**B** are 4.295 Å and 5.062 Å, while that in the periodic supercell with one double vacancy in **Figure 2A** is 9.908 Å with Ebar = 0.41 eV, thus the relative larger distance between carbon vacancies is beneficial for the CO oxidation on Mn-DG.

# Origin of the High Activity of Mn-DG

To further understand the high activity of the Mn-DG for CO oxidation, the Hirshfeld charge near the adsorbed molecules along the LH reaction path is shown in **Figure 6**, and the corresponding PDOS near the adsorbed molecules is shown in **Figure 8**. O<sup>2</sup> obtains 0.115 e while CO loses 0.076 e for the initial state in **Figure 6A**. The empty component of the O2- 2π∗ and CO-2π∗ orbitals are partially filled (see **Figure 8A**) due to the electron transformation during the adsorption, which causes the elongation of the O-O bond and C-O bond to 1.34 Å and 1.15 Å, respectively. At transition state, the charge of adsorbed O<sup>2</sup> is −0.136 e while that of adsorbed CO is −0.006 e in **Figure 6A**. This indicates that the O2-2π∗ and CO-2π∗ orbitals are more filled as shown in **Figure 8B**, which is confirmed by the elongation of the O-O bond and C-O bond to 1.46 and 1.17 Å, respectively. After surmounting the energy barrier, more charge transfers from graphene to the adsorbed O<sup>2</sup> and CO, where the charge of O<sup>2</sup> is −0.154 e while that of CO is −0.058 e for MS in **Figure 6A**. The O2-2p<sup>∗</sup> and CO-2π∗ orbitals are more filled as shown in **Figure 8C**. Therefore, the Mn-DG can tune the charge distributions of the adsorbed O<sup>2</sup> and CO, and the charge transfer from Mn-DG to O<sup>2</sup> and CO molecules plays an important role for the OOCO intermediate formation.

Due to the fact that Fe is a commonly used dopant and the energy barrier for the first step on Fe-DG is 0.88 eV for CO oxidation, which is larger than that of 0.41 eV on Mn-DG, the PDOS of the TS configurations on Fe-DG is analyzed to further understood the mechanism for depressing the formation energy barrier of OOCO intermediate, as shown in **Figure 8D**. It can be seen that the overlapping area between O2-2π∗ and Mn-3d orbitals at Fermi level on Mn-DG (see **Figure 8B**) becomes much weaker compared with that on Fe-DG (see **Figure 8D**), indicating that the interaction of Mn-O bond is significant weakened on Mn-DG. It is reported that weaker interaction nearby the Fermi level generally causes smaller reaction barrier (Arellano et al., 2000; Ao and Peeters, 2010b; Jiang et al., 2016). In addition, the

### interaction from −16 to −5 eV is strengthened on Mn-DG (see **Figures 8B,D**). Therefore, the enhanced interactions in the low energy range could lead to smaller reaction barrier for the CO oxidation on Mn-DG.

# CONCLUSION

The oxidation of CO molecule on transition metals decorated graphene with divacancy (DG) has been studied by using DFT calculations. We found that Mn-DG has the best performance for the CO oxidation, while the LH mechanism is preferred, where the rate limiting energy barrier is only 0.41 eV, indicating the efficient oxidation process. The charge transfer from Mn-DG to the O2-2p<sup>∗</sup> and CO-2π∗ orbitals through the Mn atom along the LH mechanism plays a key role for depressing the energy barrier of the CO oxidation. The results indicate that Mn-DG can be a noble-metal-free and efficient catalyst for CO oxidation.

# AUTHOR CONTRIBUTIONS

QJ, JZ, and ZA designed the research and wrote the paper. QJ carried out the simulation. HJH, HYH, and YW entered the discussion. All authors commented on the manuscript.

# FUNDING

We acknowledge supports by the Fundamental Research Funds for National Natural Science Foundation of China (Grant No. 21703052), the Central Universities (Grant Nos. 2017B12914 and 2015B01914), China Postdoctoral Science Foundation (2015M571652), Natural Science Foundation of Jiangsu Province (BK20161506) and National 973 Plan Project (2015CB057803). ZA acknowledges the financial supports from National Natural Science Foundation of China (Grant No. 21607029, 21777033), Science and Technology Program of Guangdong Province (2017B020216003), Science and Technology Program of Guangzhou City (201707010359), and 1000 Plan for Young Professionals Program of China, 100 Talents Program of Guangdong University of Technology.

# REFERENCES


**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 © 2018 Jiang, Zhang, Ao, Huang, He and Wu. 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 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.

# The Activity Improvement of the TM3(hexaiminotriphenylene)<sup>2</sup> Monolayer for Oxygen Reduction Electrocatalysis: A Density Functional Theory Study

### Beibei Xiao<sup>1</sup> \*, Hui Zhu<sup>1</sup> , HouYi Liu<sup>1</sup> , XiaoBao Jiang<sup>2</sup> and Qing Jiang<sup>3</sup> \*

<sup>1</sup> School of Energy and Power Engineering, Jiangsu University of Science and Technology, Zhenjiang, China, <sup>2</sup> School of Materials Science and Engineering, Jiangsu University of Science and Technology, Zhenjiang, China, <sup>3</sup> Key Laboratory of Automobile Materials, Ministry of Education, School of Materials Science and Engineering, Jilin University, Changchun, China

### Edited by:

Zhimin Ao, Guangdong University of Technology, China

### Reviewed by:

Aijun Du, Queensland University of Technology, Australia Simonetta Antonaroli, Università degli Studi di Roma Tor Vergata, Italy

### \*Correspondence:

Beibei Xiao xiaobb11@mails.jlu.edu.cn Qing Jiang jiangq@jlu.edu.cn

### Specialty section:

This article was submitted to Green and Sustainable Chemistry, a section of the journal Frontiers in Chemistry

Received: 19 May 2018 Accepted: 25 July 2018 Published: 12 September 2018

### Citation:

Xiao B, Zhu H, Liu H, Jiang X and Jiang Q (2018) The Activity Improvement of the TM3(hexaiminotriphenylene)2 Monolayer for Oxygen Reduction Electrocatalysis: A Density Functional Theory Study. Front. Chem. 6:351. doi: 10.3389/fchem.2018.00351 Polymer electrolyte membrane fuel cells (PEMFCs) are one of the most prominent clean energy technologies designed to achieve hydrogen utilization and solve problems such as low efficiency and high pollution associated with fossil fuel combustion. In order to bring about PEMFC commercialization, especially for automobile applications, developing high-activity and -selectivity catalysts for the oxygen reduction reaction (ORR) is of critical importance. Based on the density functional theory, the catalytic activity of the conductive, two-dimensional metal–organic frameworks TM3(HITP)<sup>2</sup> monolayer (where HITP = hexaiminotriphenylene; TM = Ni, Co, Fe, Pd, Rh, Ru, Pt, Ir, and Os) for ORR has been investigated systematically. Furthermore, the classical volcano curves of the ORR activity, as a function of the OH binding, are found where the Ni, Pd, and Pt located at the weak binding side suffer from the sluggish <sup>∗</sup>OOH formation and prefer the inefficient 2e <sup>−</sup> mechanism, while for other elements belonging to the strong binding side, the reactions are hindered by the poison due to ORR intermediates. Based on the free energy profiles, the corresponding overpotentials µORR exhibit the inverted volcano curve as a function of the atomic number of the 3d/4d/5d TM active center in the same period. Based on the µORR data, ORR activity decreases in the order of Ir > Co ≈ Rh > Ni ≈ Pd > Pt ≈ Fe > Ru > Os. Herein, the Co, Rh, and Ir central atoms exhibit enhanced catalytic activity in combination with the desirable selectivity of the O<sup>2</sup> reduction to H2O. This systematic work may open new avenues for the development of high-performance non-PGM catalysts for practical applications of ORR.

Keywords: oxygen reduction reaction, activity and selectivity, 2D materials, transition metal elements, DFT calculation

# INTRODUCTION

Hydrogen is a potential candidate for future energy provision, as stated in the concept of the hydrogen economy, so as to solve the issues of the rising global energy demands, depletion of fossil fuel reserves, and associated environmental pollution issues. Due to its high efficiency, ease of operation, and low emission, the polymer electrolyte membrane fuel cell (PEMFCs) is the

**104**

most prominent technology to derive benefit from the proposed hydrogen cycle, which leads to the production of electricity from the electrochemical oxidation of hydrogen, with water as its endproduct (Colic and Bandarenka, 2016; Xia Z. et al., ´ 2016; Chen et al., 2017, 2018). However, a critical obstacle to its commercialization is the dominant voltage loss associated with the sluggish oxygen reduction reaction (ORR), even when catalyzed by the noble Pt nanoparticle (Nørskov et al., 2004). In this regard, a significant amount of Pt is required to achieve the desirable power density, making the overall cost prohibitively high (Debe, 2012). In order to overcome the economic bottleneck, design, and application of earth-abundant alternatives for ORR electrocatalysis are at the heart of PEMFC research (Xia W. et al., 2016).

Inspired by the pioneering work on cobalt phthalocyanines acting as cathode catalysts (Jasinski, 1964), a tremendous amount of research is being performed on the TM–N<sup>x</sup> carbon materials for ORR (TM denotes transition metals), especially TM–N<sup>4</sup> active motif. The intrinsic active characteristic of TM in N<sup>x</sup> carbon materials is experimentally supported, where the presence of the TM atom would boost the ORR performance compared with the metal-free counterpart (Peng et al., 2013; Yin et al., 2016; Yang et al., 2017); the corresponding activity would be suppressed by adding the SCN<sup>−</sup> ions and CO molecule (Zhang et al., 2016; Yang et al., 2017). Linear correlations between the content of TM–N<sup>x</sup> and ORR activity have been observed (Yang et al., 2017). Furthermore, ORR activity shows its dependence on the TM center atom, and this is supported by the density functional theory (DFT), a theoretical work of Rossmeisl et al., where the ORR activity of the TM–N<sup>4</sup> embedded graphene has been systematically investigated, and where the Fe, Ir, Mn, Ru, and Rh doping is identified as boosting the ORR (Calle-Vallejo et al., 2011). Additionally, the ORR mechanism is sensitive to the TM–N<sup>4</sup> active motif. Liu et al. have synthesized the Fe–N<sup>x</sup> and Co–N<sup>x</sup> doped carbon nanofiber and realized that the Fe– N<sup>4</sup> promotes 4e <sup>−</sup> ORR in comparison with the 2e <sup>−</sup> pathway of the Co–N<sup>4</sup> one (Liu et al., 2016). The same conclusion has been achieved by Kattel et al., where the O–O bond scission is presented and the efficient 4e <sup>−</sup> reduction of O<sup>2</sup> to 2H2O is preferred on the Fe–N<sup>4</sup> sites, but the reduction of O<sup>2</sup> to H2O<sup>2</sup> is found enhanced on the Co–N<sup>4</sup> motif (Kattel et al., 2013, 2014).

Despite these encouraging research works, such TM–N<sup>4</sup> carbon materials generally suffer from low activity caused by the relatively few catalytic sites as well as the experimental challenge of the well-controlled active motif (Liang et al., 2013; Peng et al., 2013; Palaniselvam et al., 2016), reducing the competition with the state-of-the-art Pt/C catalysts. Paying attention to the high TM–N<sup>4</sup> density in combination with the electronic conductivity, Miner et al. have developed the attractive Ni3(HITP)<sup>2</sup> as ORR electrocatalyst (Miner et al., 2016). However, the performance is far from satisfying expectations, with its incomplete oxidation of O<sup>2</sup> and the predominant production of H2O<sup>2</sup> under the working potential (Miner et al., 2016). In this regard, utilizing the aforementioned information, the development of the efficient TM3(HITP)<sup>2</sup> catalysts can be achieved via the variation of metal nodes (Choi et al., 2015; Zhang et al., 2015). It has naturally raised the interest to search for the optimal combinations of the TM3(HITP)2, possessing superior ORR activity as well as selectivity.

In the manuscript, DFT calculations are used within an electrochemical framework to analyze the ORR electrocatalysis over the TM3(HITP)2monolayer. In particular, the stability of the ORR intermediates is calculated, which allows to evaluate the thermodynamic ORR free energy and its overpotentials. The data provide the fundamental understanding of the mechanism of Ni3(HITP)<sup>2</sup> and further identify optimal candidates as catalysts. According the d–partial density of states, an atomistic insight of the activity origin has been provided by a thorough comparison among the considered systems. Herein, for simplification, our attention is mainly focused on the monolayer structure shown in **Figure 1** as a representative model due to the weak interaction between the interlayers of the 2D-layered materials (Sheberla et al., 2014; Chen et al., 2015; Miner et al., 2016).

# METHODS

All calculations have been performed within the DFT framework, as implemented in the DMol<sup>3</sup> code (Delley, 1990, 2000). The generalized gradient approximation with the Perdew–Burke– Ernzerhof (PBE) functional is employed to describe exchange and correlation effects (Perdew et al., 1996). The DFT semicore pseudopotentials (DSPP) core treat method is implemented for relativistic effects, which replaces core electrons by a single effective potential and introduces some degree of relativistic correction into the core (Delley, 2002). The double numerical atomic orbital augmented by a polarization function (DNP) is chosen as the basis set (Delley, 1990). A smearing of 0.005 Ha (1 Ha = 27.21 eV) to the orbital occupation is applied to achieve accurate electronic convergence. In order to ensure high-quality results, the real-space global orbital cutoff radius is set as high as 5.2 Å. In order to accurately describe the long-range electrostatic interactions of the ORR intermediates with catalysts, the PBE-D method with the TS van der Waals (vdW) correction is employed. In the geometry optimization of structures, the convergence tolerances of energy, maximum force, and displacement are

1.0×10−<sup>5</sup> Ha, 0.002 Ha/Å, and 0.005 Å, respectively. The spinunrestricted method is used for all calculations. A conductor-like screening model (COSMO) was used to simulate a H2O solvent environment for all calculations (Todorova and Delley, 2008), which is a continuum model, where the solute molecule forms a cavity within the dielectric continuum. The DMol<sup>3</sup> /COSMO method has been generalized to periodic boundary cases. The dielectric constant is set as 78.54 for H2O. Some previous results have shown that this implicit solvation model is an effective method to describe solvation (Sha et al., 2011; Zhang et al., 2015). The 15 Å-thick vacuum is added to avoid the artificial interactions between the TM3(HITP)<sup>2</sup> monolayer and its images. The proposed structure of the TM3(HITP)<sup>2</sup> monolayer is shown in **Figure 1**, where the atoms in the red square are fixed in all of the structure-optimization calculations.

The adsorption energy (Eads) of the reactant O<sup>2</sup> is calculated by the following,

$$E\_{\rm ads} \text{(O}\_2\text{)} = E\_{\rm sys} - E\_{\rm substrate} - E\_{\rm O2} \tag{1}$$

The adsorption energy (Eads) of the ORR intermediates are calculated relative to H2O and H<sup>2</sup> (Calle-Vallejo et al., 2011), through,

$$E\_{\rm ads} \text{(OOH)} = E\_{\rm sys} - E\_{\rm substrate} - \left(2E\_{\rm H2O} - 3/2E\_{\rm H2}\right) \quad \text{(2)}$$

$$E\_{\rm ads} \text{(O)} = E\_{\rm sys} - E\_{\rm substrate} - (E\_{\rm H2O} - E\_{\rm H2}) \tag{3}$$

$$E\_{\rm ads} \text{(OH)} = E\_{\rm sys} - E\_{\rm substrate} - \left(E\_{\rm H2O} - 1/2E\_{\rm H2}\right) \tag{4}$$

where Esys, Esubstrate, EH2O, and EH2 are the total energy of the adsorption systems, the TM3(HITP)<sup>2</sup> monolayer, H2O molecule, and H<sup>2</sup> molecule, respectively. The Eads < 0 corresponds to an exothermic adsorption process.

The Gibbs free energy changes (1G) of the ORR elemental steps have been calculated according to the computational hydrogen electrode (CHE) model developed by Nørskov et al., where the chemical potential of the proton/electron (H<sup>+</sup> + e <sup>−</sup>) in solution is equal to the half of the chemical potential of a gas-phase H<sup>2</sup> at the reference relative hydrogen electrode (RHE) potential (Nørskov et al., 2004). Herein, G(H++e−) = 1/2G(H2). The variance of the proton–electron pair free energy with potential is simply determined using the linear free energy dependence of the electron energy on potential, shifting the electron energy –eU, where e is the elementary positive charge and U is the electrode potential of interest on the RHE scale (Nørskov et al., 2004; Nie et al., 2014). The total chemical potential of a proton–electron pair at the potential U is written as follows:

$$G(U)\_{\text{(H+++e-)}} = 1/2G\_{\text{(H2)}} - eU \tag{5}$$

Therefore, for a general electrochemical reaction, the free energy change 1G for every elemental step can be determined as following:

$$
\Delta G = \Delta E + \Delta ZPE - T\Delta S + \Delta G\_{\rm pH} + \Delta G\_{\rm U} \tag{6}
$$

where 1E is the electronic energy difference based on DFT calculations, 1ZPE is the change in zero point energy, T is the temperature (equal to 298.15 K here), 1S is the change in the entropy, and 1GpH and 1G<sup>U</sup> are the free energy contributions due to variation in pH value and electrode potential U, respectively. 1G<sup>U</sup> = –eU, in which U is the potential related to the standard hydrogen electrode. 1GpH = –kTln10 x pH, which is the corrected free energy of H<sup>+</sup> ions depending on the concentration. According to the previous works, pH is set as 0 in acid medium and 14 in alkaline medium. The zero-point energies and entropies of the ORR intermediates are calculated from the vibrational frequencies according to standard methods. Following the suggestion of Wilcox et al. (Lim and Wilcox, 2012), in order to reduce the calculation, the TM3(HITP)<sup>2</sup> monolayer is fully constrained. The 1G < 0 corresponds to an exothermic adsorption process. The free energy G of O<sup>2</sup> is derived as G(O2) = 4.92+2G(H2O)−2G(H2) by utilizing OER equilibrium at the standard conditions; the G of H2O<sup>2</sup> is derived similarly considering that the H2O2/O<sup>2</sup> standard equilibrium potential is 0.682 V vs. SHE (Sun et al., 2014). The CHE model has been successfully applied for developing the novel electrocatalysts with prominent ORR performances, where the DFT calculations are in line with the experimental results (Nørskov et al., 2004; Greeley et al., 2009; Favaro et al., 2015; Lang et al., 2015; Jia et al., 2016; Tang et al., 2016; Liu et al., 2017; Li et al., 2018; Xu et al., 2018). Furthermore, the PBE/DNP method in Dmol<sup>3</sup> code has been widely employed for evaluating the potential of the TM-based carbon electrocatalysts (Wang et al., 2015, 2016b; Hou et al., 2016; Xiao et al., 2017). Therefore, the reliability of the mentioned approach is confirmed.

### RESULTS AND DISCUSSION

Prior to the investigation of the activity, the essential step is to determine the adsorption behavior of the ORR intermediates. For Ni3(HITP)2, the energetics of the O<sup>2</sup> adsorption is endothermic with the value of 0.13 eV, caused by the structure deformation as shown in the inset of **Figure 2A**, indicating no such adsorption has occurred on the Ni center atom. Besides the reactant, the corresponding Eads of ORR intermediates are 4.08, 3.57, and 1.17 eV for OOH, O, and OH, respectively. In comparison with the Pt(111) (3.55 eV for OOH, 1.38 eV for O, and 0.64 eV for OH) (Xiao et al., 2016), the obviously weak adsorption ability of the center Ni atom has been observed.

Considering the low affinity of the ORR intermediates to the Ni center, it is expected that the ORR activity would be boosted by selecting suitable TMs. Indeed, the change of the TM center definitely exhibits different adsorption behaviors, as shown in **Figure 2**, where the corresponding Eads data are given. From the figures, the Eads of the reactant O<sup>2</sup> and the corresponding ORR intermediates decrease monotonically in the order of Ru ≈ Os > Fe > Co > Rh ≈ Ir > Ni > Pd ≈ Pt. It is obvious that the adsorption ability of the elements is weakened from group 8 to group 10. Carefully reviewing the O<sup>2</sup> adsorption as shown in the inset of **Figure 2A**, similar to Ni, no such adsorption behaviors have occurred at the Pd and Pt active centers due to the endothermic adsorption energies. The unfavorable adsorption behavior of group 10 implies the

underlying mechanism of the OOH formation, which originates from the long-range electron transfer from the catalyst to O<sup>2</sup> molecules at the outer Helmholtz plane (Choi et al., 2014). On the contrary, group 9 binds to the reactant suitably with the corresponding end-on structure and group 8 possesses a too strong O<sup>2</sup> adsorption with the side-on adsorption configuration. Herein, the mentioned adsorption capability is in line with the general prediction of the d-band model (Hammer and Nørskov, 2000). The corresponding evidences are provided by the d– partial density of states (d–PDOS) of the central TM atom plotted in **Figure 3**. As shown, the d band moves away from the Fermi energy as the TM atom changes from group 8 to group 10. The TM atom with the higher d states possesses stronger adsorption ability. Furthermore, due to the relationship between the adsorption behavior and the d band, the linear scaling relations between Eads(O2)/Eads(OOH)/Eads(O) and Eads(OH) are observed, in accordance with the previous reports of other functionalized carbon materials (Calle-Vallejo et al., 2011; Baran et al., 2014). That is,

$$E\_{\rm ads}(\rm O\_2) = 0.93E\_{\rm ads}(\rm OH) - 1.11,\tag{7}$$

$$E\_{\rm ads} \text{(OOH)} = 0.90 E\_{\rm ads} \text{(OH)} + 3.02,\tag{8}$$

$$E\_{\text{ads}}\text{(O)} = 1.88E\_{\text{ads}}\text{(OH)} + 1.02.\tag{9}$$

To further evaluate the ORR activity of TM3(HITP)2, according to the experimental condition (Miner et al., 2016), the OOH associative mechanisms in the alkaline solution are taken into consideration, with the elemental steps R<sup>i</sup> listed as follows (Wang et al., 2016a), where asterisks denote active TM sites. Due to the small barrier of proton transfer, which can be ignored at high applied voltages, our attention is only focused on the reaction energies (Nørskov et al., 2004; Calle-Vallejo et al., 2011; Zhang et al., 2015).

$$\text{O}\_2\text{(g)} + \text{ }^\* + \text{H}\_2\text{O(l)} + e^- \rightarrow \text{OOH}^\* + \text{OH}^-, \text{(R}\_1\text{)}$$

$$\text{OOH}^\* + \text{e}^- \rightarrow \text{O}^\* + \text{OH}^-,\tag{R\_{2a}}$$

$$\mathrm{OOH}^\* + \mathrm{H}\_2\mathrm{O}(\mathrm{l}) + e^- \to \mathrm{H}\_2\mathrm{O}\_2^\* + \mathrm{OH}^-, \qquad \mathrm{(R}\_{2\mathrm{b}})$$

$$\text{O}^\* + \text{H}\_2\text{O(l)} + e^- \rightarrow \text{OH}^\* + \text{OH}^-,\qquad \text{(R}\_3\text{)}$$

$$\text{OH}^\* + e^- \rightarrow \text{OH}^-.\tag{\text{R}\_4}$$

Analyzing the free energy plots of the complete 4e <sup>−</sup> ORR pathway in **Figure 4A**, the endothermic processes of <sup>∗</sup>OOH formation (R1) and <sup>∗</sup>O formation (R2a) are observed even at <sup>U</sup> <sup>=</sup> 0 V. The corresponding values of the free energy change 1G<sup>i</sup> are 0.28 and 0.08 eV, respectively. The positive values indicate that the mentioned steps are thermodynamically unfavorable (Nørskov et al., 2004). When the ideal potential of 0.4 eV is applied, the mentioned reaction steps are even unfavorable with the increased 1G<sup>i</sup> of 0.68 and 0.48 eV, respectively. Based on the information, R<sup>1</sup> is determined as the rate-determining step (RDS), pointing to the fact that the Ni3(HITP)<sup>2</sup> monolayer suffers from insufficient O<sup>2</sup> activation (Greeley et al., 2009), in accordance with the endothermic capture of the O<sup>2</sup> molecule as shown in **Figure 2A**. To clarify the selectivity of the 2/4e <sup>−</sup> mechanism, the comparison between R2a and R2b is considered. In the alternative 2e − pathway, the 1G2b of the H2O<sup>2</sup> formation (R2b) is increased from 0.01 to 0.41 eV as the U ranged from 0 to 0.4 V, being 0.07 eV lower than 1G2a, indicating the slightly energy favorable condition of R2b. Herein, the predominant production of H2O<sup>2</sup> would be expected on the Ni3(HITP)<sup>2</sup> monolayer, which is in consistence with the experiment results of the 3D Ni3(HITP)2. In summary, the Ni3(HITP)<sup>2</sup> monolayer slightly prefers the 2e − mechanism, with the RDS located at the <sup>∗</sup>OOH formation.

Owing to the different adsorption abilities of TM3(HITP)2, the suitable TM would boost the ORR activity as well as the selectivity of the 4e <sup>−</sup> mechanism. In order to characterize the relationship between the ORR activity and the TM center atom, the reactive free energy change 1G<sup>i</sup> of the elementary steps R<sup>i</sup> at U = 0.4 V as a function of Eads(OH) is plotted in **Figure 4B**. As the weakening of the Eads(OH) occurs with the values ranging from −0.05 to 1.68 eV, 1G<sup>1</sup> and 1G2a increase from the negative to the positive values, while the opposite tendency is found for 1G<sup>3</sup> and 1G4, which decrease from the positive to the negative values. That is, the steps R<sup>1</sup> and R2a change from the exothermic to endothermic process, while the steps R<sup>3</sup> and R<sup>4</sup> become more thermodynamically favorable. Obviously, the RDS steps are identified as R<sup>4</sup> for group 8 and group 9 with the exception of Os, where the RDS step is R3, while the RDS step is located at R<sup>1</sup> for group 10.

On the basis of the assumption that the activation barrier for the RDS is equal to 1Gmax, the activity variation of TM3(HITP)<sup>2</sup> referred to as Ni3(HITP)<sup>2</sup> is estimated by the rate constant k via the simple Arrhenius equation in the following,

$$k = A \exp(-\Delta G\_{\text{max}} / k\_{\text{B}} \text{T})\tag{10}$$

$$\text{In}(k\_{\text{TM}}/k\_{\text{Ni}}) = [\Delta G\_{\text{max}}(\text{Ni}) - \Delta G\_{\text{max}}(\text{TM})]/k\_{\text{B}}\text{T}) \tag{11}$$

where A is the prefactor, k<sup>B</sup> is the Boltzmann constant, and T is the temperature (298.15 K). The classic volcano curve is observed in **Figure 5A**, where the In(kTM/kNi) as a function of the Eads(OH) is plotted. Based on our results, the adsorption ability is weakened as the atomic number increases from group 8 to group 10. The Ni/Pd/Pt active centers suffer from the too weak binding strength, leading to the energetically unfavorable process of the <sup>∗</sup>OOH formation. Meanwhile, for the Fe/Ru/Os active centers, the too strong interaction with the ORR intermediates accounts for the poisoning of the O-containing intermediates. The suitable binding strength of Co/Rh/Ir active centers indicates the balance between the O<sup>2</sup> activation and the catalyst recovery.

As discussed by the previous reports (Calle-Vallejo et al., 2011; Viswanathan et al., 2012a; Baran et al., 2014; Zhang et al., 2015), the ORR activity depends on the adsorption of the intermediates. Our results are in accordance with the previous reports in that the bond strength should be compromised in the case of the effective ORR catalysts on the basis of Sabatier principle (Greeley et al., 2009; Gao et al., 2017, 2018); this is similar with the previous studies on metal (Stephens et al., 2012), functional graphene (Calle-Vallejo et al., 2011), metal porphyrine (Baran et al., 2014) as well as the 2D MOF (Gao et al., 2017). Herein, in comparison with the Ni3(HITP)<sup>2</sup> monolayer (0.68 eV), group 9 possesses superior catalytic performance with the smaller 1Gmax of 0.50, 0.52, and 0.22 eV for the Co, Rh, and Ir center atoms, respectively. By contrast, inferior activities are observed for group 8 and group 10. The corresponding 1Gmax are 0.93, 1.23, and 1.34 eV for Fe, Ru, and Os atoms, and 0.72 and 0.90 eV for Pd and Pt atoms, respectively. Furthermore, for clarification, the data of the relative overpotentials µORR are summarized and collected for the activity comparison, depicted in **Figure 6**. As the atomic number of TM increased from Ni to Fe, the µORR reduces and then increases, presenting further evidence to the presence of the classical volcano-shaped activity (Nørskov et al., 2004; Calle-Vallejo et al., 2011; Baran et al., 2014; Zheng et al., 2017). Similar situations are found for the 4d/5d TM3(HITP)<sup>2</sup> monolayer. Based on the µORR data, ORR activity decreases in the order of Ir > Co ≈ Rh > Ni ≈ Pd > Pt ≈ Fe > Ru > Os. Furthermore, compared with the data of the TM supported on graphene and macrocyclic molecules (the minimum µORR = 0.4 eV) (Xu et al., 2018), the prominent improvement of the ORR activity on the Ir3(HITP)<sup>2</sup> monolayer is confirmed.

Besides the ORR activity, the ORR selectivity is characterized by the ratio of the corresponding reaction rate k2a (k2b) for step R2a (R2b), via the following equation,

$$\text{In}(k\_{\text{2a}}/k\_{\text{2b}}) = [\Delta G\_{\text{2b}} - \Delta G\_{\text{2a}}]/k\_{\text{B}}\text{T} \tag{12}$$

The data are plotted in **Figure 5B**. As shown, In(k2a/k2b) shows a linear relationship with Eads(OH). For the weak side, the values of 1G2a and 1G2b are 0.48 and 0.41 eV for Ni, 0.57 and 0.37 eV for Pd, and 0.20 and 0.19 eV for Pt, respectively. The corresponding

In(k2a/k2b) are −2.72, −7.78, and −0.39, indicating the energy favorability of R2b compared with that of R2a for group 10. Furthermore, considering the In(k2a/k2b) data, the 2e <sup>−</sup> reduction of O<sup>2</sup> to H2O<sup>2</sup> is prevalent at the Ni and Pd centers, while the mixing of 2/4e <sup>−</sup> ORR is reasonable for the Pt center. By contrast, R2a is preferred for other elements indicated by the positive In(k2a/k2b). That is, the 4e <sup>−</sup> mechanism is dominant for group 8 and group 9. Obviously, as the Eads(OH) is strengthened, the catalytic selectivity is changed from the 2e <sup>−</sup> to 4e <sup>−</sup> pathway. The two sides of the catalytic activity profile shown in **Figure 5A** essentially distinguish the 2e <sup>−</sup> catalysts (weak binding side) from the 4e <sup>−</sup> catalysts (strong binding side) (Viswanathan et al., 2012b; Zagal and Koper, 2016). That is, the former suffering from the insufficient O<sup>2</sup> activation favors the H2O<sup>2</sup> formation; the latter poisoned by the O–containing intermediates generally prefers the H2O formation. The formation of H2O<sup>2</sup> not only decreases the ORR efficiency, but also degrades the proton exchange membrane (Tsuneda et al., 2017). In short, the dramatic enhancements in oxygen-reduction rates and product

TABLE 1 | The Eads and 1G of ORR intermediates at the potential U = 0.4V with and without the vdW corrections.


selectivity are achieved by selecting TM centers in group 9, in line with the previous works (Gao et al., 2017; Wannakao et al., 2017).

To understand the influence of the long-term electrostatic potential on the ORR performance, the vdW correction is further applied for analyzing the optimum Ir/Co/Rh systems, where the corresponding adsorption energy and the free energy changes are listed in **Table 1**. Intuitively, the adsorption abilities of the mentioned systems are enhanced due to the presence of the longterm interaction, in consistence with our calculation data. For the Ir3(HITP)<sup>2</sup> monolayer, the Eads are slightly increased to −0.39, 3.49, 2.11, and 0.61 eV for O2, OOH, O, and OH, respectively, in comparison with the uncorrected values of −0.31, 3.59, 2.16, and 0.68 eV. The perturbation of the binding strength hinders the OH removal from the Ir active site, leading to the uphill of the 1Gmax from 0.22 to 0.30 eV. Despite the activity decrease, the 4e <sup>−</sup> reduction pathway remains, supported by the values of 1G2a with −0.23 eV and 1G2b with 1.26 eV. Herein, the same phenomenon is found for Co3(HITP)<sup>2</sup> and Rh3(HITP)<sup>2</sup> with the 1Gmax of 0.68 and 0.59 eV, respectively. The corresponding ORR activity with th 4e <sup>−</sup> reaction mechanism follows the order of Ir > Rh ≈ Co. Despite the numerical variation, the trend is roughly consistent with the results without considering vdW interaction.

Despite the presence of the Ni3(HITP)<sup>2</sup> experimentally, other 2D catalysts are only theoretical models, which need the confirmation of their synthesis. It should be noted that changing metal atoms would significantly modify the corresponding structures. Thus, by replacing the Ni central atom, the dimensionality of the Cr3(HITP)<sup>2</sup> could be transferred from 2D to 3D due to the energetic favorable insertion of the spacer linker (Foster et al., 2016). Different structures inevitably lead to distinct catalytic performances (Sun et al., 2014). In this regard, although the theoretical candidates have been rationally predicted, the ORR performance of the optimum TM3(HITP)<sup>2</sup> with the Co/Rh/Ir active sites crucially needs further experimentally verification.

# CONCLUSION

Based on the DFT, the ORR mechanisms on TM3(HITP)<sup>2</sup> monolayer have been studied. It is found that the selection of central metals affects the adsorption behaviors, tuning ORR activity and its selectivity. It is realized that the adsorption abilities are monotonously enhanced as the d band upshifts from group 10 to group 8. A classical volcano relationship

# REFERENCES


for the predicted ORR activity as a function of calculated OH adsorption energy is observed. From the calculated 1G data, the ORR activity decreases in the order of Ir > Co ≈ Rh > Ni ≈ Pd > Pt ≈ Fe > Ru > Os. That is, group 9 possesses superior activity compared with other elements. Furthermore, owing to the insufficient O<sup>2</sup> activation, the 2e <sup>−</sup> mechanism is prevalent in group 10, while the desirable 4e <sup>−</sup> mechanism is dominant for others. These results would throw insights into the nature of the ORR mechanisms of TM3(HITP)2. The materials of the Co3(HITP)2, Rh3(HITP)2, and Ir3(HITP)<sup>2</sup> have been screened out and served as the potential candidates for the ORR electrocatalysis.

# AUTHOR CONTRIBUTIONS

BX carried out the simulation and wrote the paper. QJ revised the paper. HZ, HL, and XJ entered the discussion. All authors commented on the manuscript.

# ACKNOWLEDGMENTS

We acknowledge the support from the National Natural Science Foundation of China (No. 21503097, 51631004) and the Natural Science Foundation of Jiangsu (No.BK20140518).


**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 © 2018 Xiao, Zhu, Liu, Jiang and Jiang. 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.