# NEXT-GENERATION ELECTROCHEMICAL ENERGY STORAGE DEVICES

EDITED BY : Yuxin Tang, Oleksandr Malyi, Wenping Sun and Jianqing Zhao PUBLISHED IN : Frontiers in Chemistry

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-569-6 DOI 10.3389/978-2-88963-569-6

#### 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

## NEXT-GENERATION ELECTROCHEMICAL ENERGY STORAGE DEVICES

Topic Editors: Yuxin Tang, University of Macau, China Oleksandr Malyi, University of Oslo, Norway Wenping Sun, Zhejiang University, China Jianqing Zhao, Soochow University, China

Citation: Tang, Y., Malyi, O., Sun, W., Zhao, J., eds. (2020). Next-generation Electrochemical Energy Storage Devices. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-569-6

# Table of Contents


Xiaoxia Liu, Yufei Ren, Lan Zhang and Suojiang Zhang


Xiaosheng Tang, Jie Yang, Shiqi Li, Weiwei Chen, Zhiping Hu and Jing Qiu

*69* In-situ *Functionalization of Metal Electrodes for Advanced Asymmetric Supercapacitors*

Leimeng Sun, Xinghui Wang, Yurong Wang, Dongyang Xiao, Weifan Cai, Yuan Jing, Yanrong Wang, Fangjing Hu and Qing Zhang


Liying Deng, Wangyang Li, Hongnan Li, Weifan Cai, Jingyuan Wang, Hong Zhang, Hongjie Jia, Xinghui Wang and Shuying Cheng

*103 First-Principle Insights Into Molecular Design for High-Voltage Organic Electrode Materials for Mg Based Batteries*

Johann Lüder and Sergei Manzhos

*121 Building High Power Density of Sodium-Ion Batteries: Importance of Multidimensional Diffusion Pathways in Cathode Materials* Mingzhe Chen, Yanyan Zhang, Guichuan Xing and Yuxin Tang

# Enhanced Cycling Stability of Cation Disordered Rock-Salt Li1.2Ti0.4Mn0.4O<sup>2</sup> Material by Surface Modification With Al2O<sup>3</sup>

Baojun Huang, Rui Wang\*, Yansheng Gong, Beibei He\* and Huanwen Wang

Faculty of Materials Science and Chemistry, China University of Geosciences, Wuhan, China

Cation disordered rock-salt lithium-excess oxides are promising candidate cathode materials for next-generation electric vehicles due to their extra high capacities. However, one major issue for these materials is the distinct decline of discharge capacities during charge/discharge cycles. In this study, Al2O<sup>3</sup> layers were coated on cation disordered Li1.2Ti0.4Mn0.4O<sup>2</sup> (LTMO) using atomic layer deposition (ALD) method to optimize its electrochemical performance. The discharge capacity after 15 cycles increased from 228.1 to 266.7 mAh g−<sup>1</sup> for LTMO after coated with Al2O<sup>3</sup> for 24 ALD cycles, and the corresponding capacity retention enhanced from 79.7 to 90.9%. The improved cycling stability of the coated sample was ascribed to the alleviation of oxygen release and the inhibition on the undesirable side reactions. Our work has provided a new possible solution to address some of the capacity fading issues related to the cation disordered rock-salt cathode materials.

Keywords: lithium-ion batteries, cathode, cation disorder, rock-salt, Li-excess

### INTRODUCTION

Cathode materials with high energy densities are crucial for next generation of lithium-ion batteries, especially when used in hybrid and electric vehicles (Zu and Li, 2011; Goodenough and Park, 2013; Rui et al., 2015; Wang et al., 2017; Lv et al., 2018; Tan et al., 2018; Zhang et al., 2019). For this reason, Layer-structured Li-excess materials, which could deliver capacities as high as 300 mAh g−<sup>1</sup> , have been researched for more than 10 years (Lu et al., 2002; Wang et al., 2013; Yu et al., 2014; Hu et al., 2018). But these materials undergo irreversible O loss in the first cycle, which may cause structure densification and voltage degradation in subsequent cycles (Xu et al., 2011). Recently, cation disordered rock-salt Li-excess materials, sharing similar chemical compositions with layered Li-excess materials, have also attracted lots of attentions because of their high capacities (Yabuuchi et al., 2011, 2015, 2016a,b; Lee et al., 2014, 2015, 2017, 2018; Twu et al., 2015; Freire et al., 2016, 2017; Cambaz et al., 2018; Kitchaev et al., 2018; Zhao et al., 2019). Besides, it is reported that the O redox may undergo reversible reactions in the first cycle in this material (Wang et al., 2015), which provides the possibility to achieve reversible changes during the lengthy electrochemical cycles. However, many of these materials have problems in their cycling performances. Wang et al. prepared a cation disordered Li1.23Ni0.155Ru0.615O<sup>2</sup> material, and the capacity drops from 295.3 to 250 mAh g−<sup>1</sup> only after 5 charge-discharge cycles (Wang et al., 2017). Okada's group prepared another cation disordered Li1.2Mn0.4Ti0.4O<sup>2</sup> material, but its capacity drops from 226 to about 200 mAh g−<sup>1</sup> just after 6 cycles (Kitajou et al., 2016).

#### Edited by:

Yuxin Tang, Nanyang Technological University, Singapore

#### Reviewed by:

Xianhong Rui, Guangdong University of Technology, China Jianqing Zhao, Soochow University, China Renheng Wang, Shenzhen University, China

#### \*Correspondence:

Rui Wang wangrui@cug.edu.cn Beibei He babyfly@mail.ustc.edu.cn

#### Specialty section:

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

Received: 25 December 2018 Accepted: 11 February 2019 Published: 04 March 2019

#### Citation:

Huang B, Wang R, Gong Y, He B and Wang H (2019) Enhanced Cycling Stability of Cation Disordered Rock-Salt Li1.2Ti0.4Mn0.4O2 Material by Surface Modification With Al2O3. Front. Chem. 7:107. doi: 10.3389/fchem.2019.00107

Similar problems happened to layered Li-excess materials, and some literatures have reported that atomic layer deposition (ALD) method may be an effective method to alleviate the problem. ALD is a powerful technique to precisely render a uniform and conformal layer at Å level on arbitrary substrate surfaces due to its pulsing and controllable reaction. Belharouak's team coated Li1.2Ni0.13Mn0.54Co0.13O<sup>2</sup> porous powder with ultrathin Al2O<sup>3</sup> film using ALD technique, and the coated material shows higher first cycle coulombic efficiency and improved cycling performance (Zhang et al., 2013). Xiao et al. reported that the AlPO<sup>4</sup> coating layer by ALD can effectively protect the Li1.2Mn0.54Co0.13Ni0.13O<sup>2</sup> against the attack from the electrolyte, and can significantly improve its initial coulombic efficiency and thermal stability (Xiao et al., 2017a). Meanwhile, Oxide-based coatings at the surface of different cathode materials via ALD method have been demonstrated to be very conformal and uniform as reported in the literatures (Zhao and Wang, 2012; Zhao et al., 2012; Zhao and Wang, 2013a,b), thus they can effectively prevent from the electrolyte attack for enhanced cycling stability of the coated cathode.

In this study, we coated a cation disordered rock-salt Li1.2Ti0.4Mn0.4O<sup>2</sup> material with different thicknesses of Al2O<sup>3</sup> layers by ALD technique and studied their effects on the cycling stabilities. Li1.2Ti0.4Mn0.4O<sup>2</sup> is a typical cation disordered rocksalt material but with poor cycling performances (Kitajou et al., 2016; Yabuuchi et al., 2016a). In our previous study, we found that the valence of Mn in disordered materials may be lower than 3+ after the first cycle (Wang et al., 2015), and this low valence may cause Mn dissolution into the electrolyte (Nicolau et al., 2018). In this case, we infer that Al2O<sup>3</sup> coating may be an effective method to increase the reversibility and cycling performances of this cation disordered material.

#### EXPERIMENTAL

Li1.2Ti0.4Mn0.4O<sup>2</sup> was synthesized by the traditional solid-state reaction using the precursors of Li2CO<sup>3</sup> (99%, Alfa Aesar), Mn2O<sup>3</sup> (98%, Alfa Aesar), and TiO<sup>2</sup> (99%, Sigma-Aldrich). Stoichiometric amounts of precursors were ball-milled for 4 h and then pressed into a pellet. The pellet was calcinated at 900◦C for 12 h in Ar, then pulverized and ball-milled at 300 rpm for 4 h. The obtained material was denoted as LTMO.

Coating was achieved by the Atomic Layer deposition (ALD) method using an Ensure Nanotech ALD system (LabNano-9100). During the ALD process, as-prepared LTMO sample was first placed in a home-made sample holder and heated to 200◦C in the reaction chamber under ca. 1.0 mbar. Nitrogen with a flow rate of 20 sccm was used as the carrying and purge gas. The Al(CH3)<sup>3</sup> and water were employed as aluminum and oxygen precursors, respectively. In each cycle, pulse time of Al(CH3)<sup>3</sup> was controlled at 0.02 s, and exposure time water precursor was 0.02 s. 8 s nitrogen purge steps were used after Al(CH3)<sup>3</sup> and water exposures. The LTMO sample was coated with Al2O<sup>3</sup> layers for 16, 24, 40 ALD cycles. The coated samples were denoted as LTMO/nAl2O3, where n stands for the number of ALD cycles.

Morphologies and crystal structures of LTMO with or without Al2O<sup>3</sup> layers were analyzed by scanning electron microscope (SEM, SU 8010) and powder X-ray diffractometer (XRD, Bruker D8 Advance), respectively. High Resolution Transmission Electron Microscope (HRTEM, TF 20) was used to record detailed crystalline structures of the samples. Xray photoelectron spectroscopy (XPS, ESCALab 250Xi) with monochromatic Al K-α radiation was carried out to investigate the valences of the species in the samples.

To fabricate the cathode electrode, different Al2O3-coated LTMO samples were firstly ball-milled with Ketjen black at 150 rpm for 4 h, then manually mixed with polytetrafluoroethylene (PTFE) binder. The mixture was rolled into a thin film, in which the weight ratios of LTMO/nAl2O3, Ketjen black, and PTFE are 70: 20: 10. The surface mass density of each electrode film is about 4.4 mg cm−<sup>2</sup> . Cells were assembled according to our previous study (Wang et al., 2015). Typically, a two electrode swagelok cell was fabricated using the Al2O3-coated LTMO thin film and lithium foil as the working electrode and the counter electrode, respectively. Borosilicate glass fiber membrane (Whatman) was used as the separator, and 1 M solution of LiPF<sup>6</sup> dissolved in ethylene carbonate/dimethyl carbonate (1:1 by volume) was used as the electrolyte. All cells were cycled on an LANHE CT2001A system (Wuhan LAND Electronics Co.) between 4.8 and 1.5 V with a current density of 10 mA g−<sup>1</sup> at room temperature. The electrochemical impedance spectroscopy (EIS) data were recorded using a Gamry Reference 3,000 equipment. Cyclic voltammetry (CV) was measured on a CHI760E electrochemical workstation with the scanning rate of 0.5 mV s−<sup>1</sup> and the potential range of 4.8–1.5 V.

#### RESULTS AND DISCUSSION

X-ray diffraction (XRD) patterns of the samples are shown in **Figure 1a**, and all the results correspond well to the disordered rock-salt structure. According to previous studies, cations are randomly placed on the 4b position (1/2, 1/2, 1/2), and the O atom is placed on the 4a position (0, 0, 0) (Wang et al., 2015). In cation disordered rock-salt materials, Li excess could build the 0-TM channels, through which Li ions could move out and back into the cathode material (Lee et al., 2015). According to the XRD patterns, no signal of Al2O<sup>3</sup> is detected, and the reason may be that the amount of coated Al2O<sup>3</sup> is too small or the coated Al2O<sup>3</sup> is amorphous. The morphologies of LTMO before and after Al2O<sup>3</sup> coating are shown in **Figures 1b,c**. No obvious differences are observed, indicating the coated Al2O<sup>3</sup> particle may be too small to be surveyed in SEM. Energy Dispersive Spectrometers of the two samples are shown in **Figure 2a**. The characteristic peak of Al element could clearly be observed in the LTMO/24Al2O<sup>3</sup> sample, while didn't show up in the LTMO sample. The LTMO/24Al2O<sup>3</sup> sample was further examined by the elemental mapping experiments. and the results are shown in **Figures 2b-f**. Results show that Ti, Mn, O and Al elements distribute uniformly in the sample, which infer the Al2O<sup>3</sup> layer exists uniformly in the sample.

In order to directly observe the coating Al2O<sup>3</sup> layers on the surface of LTMO particles, we investigate bare LTMO and LTMO/24Al2O<sup>3</sup> samples using high resolution transmission

electron microscopy (HR-TEM). **Figure 3a** is the HR-TEM image of bare LTMO particles, showing well-defined lattice fringes in the surface region as well as those in the bulk region. In contrast, an obvious coating film in the surface region is clearly observed for LTMO/24Al2O<sup>3</sup> sample (**Figure 3b**). It is different from the inner areas, and we infer it may be the Al2O<sup>3</sup> coating layer. The thickness of the coating layer is around 3–5 nm.

In order to confirm Al2O<sup>3</sup> layer exists on the sample surfaces, XPS was used to probe the surface compositions of the LTMO and LTMO/24Al2O3. All spectra were calibrated with the C 1s peak at 284.6 eV. **Figure 4A** presents the elemental XPS spectra of LTMO and LTMO/24Al2O<sup>3</sup> samples, showing both samples contain the characteristic peaks of Ti, Mn, O, and C. From the partial enlarged figure (inset of the **Figure 4A**), the elemental peak of Al is only observed in the LTMO/24Al2O<sup>3</sup> spectrum. The core level binding energies of Ti are aligned at 457.9 eV (2p3/2) and 463.6 eV (2p1/2) (**Figure 4B**), which is in good agreement with Ti4<sup>+</sup> ions in other titanium-based compounds (Kim et al., 2017). The Mn 2p XPS spectra of LTMO and LTMO/24Al2O<sup>3</sup> reveal peaks at 641.6 eV and 653.2 eV (**Figure 4C**), which are characteristic peaks of Mn3<sup>+</sup> 2p3/<sup>2</sup> and 2p1/<sup>2</sup> (Das et al., 2011, 2013). This result confirms the valences of Mn in both samples are 3+. The Al 2p3/<sup>2</sup> core level XPS spectrum displays a binding energy of 74.7 eV for the LTMO/24Al2O<sup>3</sup> (**Figure 4D**), affirming that the chemical composition of the surface coating is Al2O<sup>3</sup> for the LTMO/24Al2O<sup>3</sup> sample.

The charge and discharge profiles of uncoated and coated LTMO samples for the first cycle are shown in **Figure 5A**, showing all the samples have the identical profiles except for the voltage platform and capacity. It is obviously seen that all the coated LTMO samples have higher first discharge plateau and coulombic efficiency, and we think this may be the effect of Al2O<sup>3</sup> layer. Though the O reaction in LTMO may happen in the form of O redox, there might still be little amount of oxygen release in the first cycle. We infer the Al2O<sup>3</sup> layer may alleviate this oxygen release and improve the O reaction reversibility, thus improve the discharge plateau and coulombic efficiency. The charge and discharge profiles for the 10th cycle are shown in **Figure 5B**, and the LTMO/24Al2O<sup>3</sup> sample presents the largest discharge capacity. **Figure 5C** shows the cycling stability of the pristine and ALD coated samples. Bare LTMO shows an initial discharge capacity of 286.3 mAh g−<sup>1</sup> (818.6 Wh kg−<sup>1</sup> ) and the capacity drops rapidly to 228.1 mAh g−<sup>1</sup> (587.7 Wh kg−<sup>1</sup> ) after 15 cycles, which is only 79.7% for its initial discharge capacity. After coated with Al2O<sup>3</sup> for 16 ALD cycles, an initial discharge of 289.1 mAh g−<sup>1</sup> (856.6 Wh kg−<sup>1</sup> ) is observed and the capacity retention reaches 85.2%. It is worth noting that the LTMO/24Al2O<sup>3</sup> sample could deliver a capacity of 293.4 mAh g−<sup>1</sup> and energy density of 884.5 Wh kg−<sup>1</sup> in the first cycle. After 15 cycles, the capacity and energy density could still maintain 266.7 mAh g−<sup>1</sup> and 709.5 Wh kg−<sup>1</sup> , respectively. And the capacity retention after 15 cycles is 90.9%, which is 11.2% higher than that of the pristine LTMO. Besides, the coulombic efficiencies also increased from 91.5 to 93.0% after the 24 cycles ALD coating. However, further increasing the ALD cycle numbers of coated Al2O3, the inferior performance for LTMO/40Al2O<sup>3</sup> is obtained, and the capacity retention decreases to 84.6%. The reason should

be the thick insulated Al2O<sup>3</sup> layer may inhibit the electric and ionic transportation (Jung et al., 2010). Based on above results, it can be concluded that proper thickness of Al2O<sup>3</sup> layer can greatly improve the cycling stability of LTMO electrodes, while the over-thick Al2O<sup>3</sup> layer leads to the reverse effect on the electrode.

**Figure 6** show the cyclic voltammetric (CV) curves of LTMO and LTMO/24Al2O<sup>3</sup> samples in the first three cycles at a

scanning rate of 0.5 mV s−<sup>1</sup> . For the LTMO electrode (shown in the **Figure 6A**), the anodic peaks located at about 3.95 V corresponds to the oxidation of Mn3<sup>+</sup> to Mn4<sup>+</sup> (Xiao et al., 2017b) during the initial charge process. Another anodic peaks at around 4.63 V may be attributed to the oxygen loss from the crystal structure and formation of O2<sup>−</sup> 2 or O<sup>2</sup> from O2−, which is usually observed on the Mn-based Li-excess cathode materials (Ma et al., 2016). While this peak disappears in the subsequent cycles, showing the irreversibility of the oxidation of O2<sup>−</sup> in the LTMO electrode. In the following cathodic process, the peak at around 2.89 V corresponds to the reduction of Mn4<sup>+</sup> to Mn3<sup>+</sup> (Kong et al., 2017). It can be clearly seen that this peak moves to lower voltage range in the following cycles, indicating the decreasing of the valence of Mn. As to LTMO/24Al2O<sup>3</sup> (shown in **Figure 6B**), two differences can be observed. First, the anodic peak at around 4.63 V still can be observed in the second and third cycles. Second, the cathodic peak located at around 3.17 V doesn't move during the cycles. These results infer the oxygen loss from the crystal structure in LTMO/24Al2O<sup>3</sup> is milder than in LTMO. As a result, the Mn redox reaction in LTMO/24Al2O<sup>3</sup> happens between Mn4+ and Mn3+. While Mn in pristine LTMO reduces to lower than 3+ because of the O loss and related densification in the first cycle.

In order to investigate the intrinsic factor of the improvement in the electrochemical performance of the LTMO/24Al2O<sup>3</sup>

sample, electrochemical impedance spectra (EIS) were collected on the bare LTMO and the LTMO/24Al2O<sup>3</sup> after charging to 4.8 V and resting for 4 h at the different cycles, shown in **Figures 7A,B**, respectively. All the Nyquist plots comprised a depressed semicircle from high to middle frequencies and an inclined line at low frequency. The simulated equivalent circuit is presented as an inset. The R<sup>1</sup> represents the Ohmic resistance coming from the separator, electrolyte and other components. The semicircle shows the charge transfer reaction composed of a charge transfer resistor (R2) and a constant phase element (CPE1), the inclined line stands for the Warburg diffusion impedance (ZW). The LTMO/24Al2O<sup>3</sup> electrode shows the smaller R<sup>2</sup> of 101.6 before cycling and remains at 228.4 , 433.8 and 851.3 after 1st, 5, and 10th cycling, respectively. Nevertheless, the much larger R<sup>2</sup> values of the LTMO electrode are seen (i.e., 255.8 before cycling and changed to 557.7 after 1st cycling, even to 10,207 for the 5th cycling, respectively), which infer complex side reactions may happen on the electrode surface and these reaction products may have blocked the ionic transfer process and result in worse cycling performance. Thus, it can be concluded that an appropriate thickness of Al2O<sup>3</sup> layer guarantees the stable charge transfer and structural integrity of the cathode electrodes, which led to good cycling performance.

### CONCLUSION

ALD technique was successfully used to deposit ultrathin Al2O<sup>3</sup> coating layer onto the surface of the LTMO particles. Comparing with the uncoated LTMO, ALD process can reduce the polarization, restrain the undesirable side reactions, and suppress the increasing charge transfer resistance during cycling, which results in the significantly improved electrochemical performance of Al2O3-coated LTMO (LTMO/24Al2O3). The controllable ALD technology provides a promising guideline for the surface modification of disordered rock-salt cathode materials with high electrochemical performance.

### AUTHOR CONTRIBUTIONS

RW conceived and designed the research. BaH carried out the experiments. YG and HW contributed to the discussion. RW wrote the manuscript with the help of BeH. All authors reviewed the manuscript.

#### ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (51702294 and 51702295).

#### 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 © 2019 Huang, Wang, Gong, He and Wang. 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 Electrochemical Performances of n-Type Extended Lattice Spaced Si Negative Electrodes for Lithium-Ion Batteries

Moonsang Lee1†, Dockyoung Yoon2,3†, Uk Jae Lee<sup>4</sup> , Nurzhan Umirov <sup>3</sup> , Aliya Mukanova<sup>5</sup> , Zhumabay Bakenov <sup>5</sup> and Sung-Soo Kim<sup>3</sup> \*

*<sup>1</sup> Korea Basic Science Institute, Daejeon, South Korea, <sup>2</sup> SK Innovation, Daejeon, South Korea, <sup>3</sup> Graduate School of Energy Science and Technology, Chungnam National University, Daejeon, South Korea, <sup>4</sup> School of Integrative Engineering, Chung-Ang University, Seoul, South Korea, <sup>5</sup> National Laboratory Astana, School of Engineering, Nazarbayev University, Institute of Batteries, Astana, Kazakhstan*

#### Edited by:

*Jianqing Zhao, Soochow University, China*

#### Reviewed by:

*Yong Min Lee, Daegu Gyeongbuk Institute of Science and Technology (DGIST), South Korea Zhiqiang Xie, University of Tennessee Space Institute (UTSI), United States*

> \*Correspondence: *Sung-Soo Kim kimss@cnu.ac.kr*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *18 February 2019* Accepted: *14 May 2019* Published: *31 May 2019*

#### Citation:

*Lee M, Yoon D, Lee UJ, Umirov N, Mukanova A, Bakenov Z and Kim S-S (2019) The Electrochemical Performances of n-Type Extended Lattice Spaced Si Negative Electrodes for Lithium-Ion Batteries. Front. Chem. 7:389. doi: 10.3389/fchem.2019.00389* The electrochemical performances of lithium-ion batteries with different lattice-spacing Si negative electrodes were investigated. To achieve a homogeneous distribution of impurities in the Si anodes, single crystalline Si wafers with As-dopant were ball-milled to form irregular and agglomerated micro-flakes with an average size of ∼10µm. The structural analysis proved that the As-doped Si negative materials retain the increased lattice constant, thus, keep the existence of the residual tensile stress of around 1.7 GPa compared with undoped Si anode. Electrochemical characterization showed that the As-doped Si anodes have lower discharge capacity, but Coulombic efficiency and capacity retention were improved in contrast with those of the undoped one. This improvement of electrochemical characteristics was attributed to the increased potential barrier on the side of Si anodes, inherited from the electronic and mechanical nature of Si materials doped with As. We believe that this study will guide us the way to optimize the electrochemical performances of LIBs with Si-based anodes.

Keywords: lithium-ion batteries, silicon, dopant, arsenic, discharge capacity, retention

## INTRODUCTION

Lithium-ion batteries (LIBs) have been considered as an important energy storage device for modern portable electronics and electric vehicles (Goodenough and Park, 2013; Nitta et al., 2015; Kim et al., 2016). Until now, the graphite is typically used as anode material in commercial LIBs. However, its theoretical capacity (372 mAh g−<sup>1</sup> ) is not sufficient for the next-generation mobile applications requiring high energy density (Peng et al., 2010; Ma et al., 2014; Ko et al., 2015). To overcome the energy density problem of current LIBs, the use of Si as the anode in LIBs can be an alternative way due to its unique physical properties such as high theoretical capacity of 3,580 mAh g −1 (Li15Si<sup>4</sup> state) and low discharge potential (<0.5 V vs. Li/Li+) (Wang and Dahn, 2006; Liu et al., 2012; Shi et al., 2016). Even if the Si material has such good advantages for LIBs, the commercial success of LIBs with Si anodes has been impeded by fast capacity fading of Si anodes. This fading results from large volume change more than 300% of Si anodes during lithiation and delithiation, leading to their disintegration (Ryu et al., 2004; Kim et al., 2011). Furthermore, semiconductor characteristics of Si make it even more difficult to be utilized in the practical application of LIBs as

**13**

a negative electrode owing to the electrical losses due to large ohmic resistance (Wen and Tian, 2013; Wang et al., 2015).

Recently, various research groups have reported that the electrochemical performances of LIBs with Si anodes can be improved by using Si/Metal alloy electrode (Umirov et al., 2019a,b), the formation of nanostructured-anodes (Mukanova et al., 2018a,b), hollow and yolk-shell structured-composites (Yao et al., 2011; Wang et al., 2013), and doping of Si with impurities such as copper (Cu) (Wen and Tian, 2013), arsenic (As) (McSweeney et al., 2014), aluminum (Al) (Legrain and Manzhos, 2015), silver (Ag) (Talla et al., 2015), boron (B) (Long et al., 2011; Rousselot et al., 2012; Yi et al., 2013), nitrogen (N) (Han et al., 2016), and phosphorus (P) (Domi et al., 2016). Among other approaches, incorporating dopants into Si architectures has gained considerable attention as a suitable route for improving the rate capability, electrical conductivity, and capacity retention of the Si anodes (Mukanova et al., 2018b).

To date, many methods have been proposed for the Si powder preparation, including solution and solid-state based techniques, etc. Taking into account that the electronic/silicon-based semiconductor technologies have widely entered into society and industry, it is easy to imagine how much semiconductor class Si containing waste is produced every day. These silicon materials can be recycled for use in LIBs.

In this paper, we used the Si wafer in order to demonstrate the possibility of the Si electronic wastes to be reused in LIBs. As-modulated Si micro-flake anodes were investigated as an anode. To achieve a homogeneous distribution of As-impurity in the Si anodes, As-doped Si wafer was prepared by ballmilling and employed for the anode in a lithium cell. Moreover, this work covers the comparative analysis in the structural and electrochemical responses of As-doped and undoped Si microflake anodes to investigate the effects of the lattice spacing in Si anodes.

We believe that this paper will shed lights on the trends in the electrochemical performance [cycle retention, Coulombic efficiency (CE), and rate capability] of LIBs with the Si microflake anode with different lattice spacing.

#### EXPERIMENTAL

The commercial As-doped single crystalline (100) Si wafers with a doping concentration of 1 × 10<sup>19</sup> cm−<sup>3</sup> were provided by SK Hynix and used as the anode material for LIBs. The doping concentrations of Si wafer were confirmed by secondary ion mass spectrometry (SIMS) (not shown in this paper). Furthermore, the intrinsic (undoped) Si wafers were used as the control samples to compare the effects of As doping. To remove native oxide on Si surface, Si wafers were immersed in diluted 10% hydrofluoric acid solution for 3 min and followed by rinsing in deionized (DI) water of ultra-purity (resistance 18.2 Mcm) for 5 min. Si wafers were mechanically grounded to achieve Si microflakes using zirconia balls for 2 h in paint shaker (KM-2000T, Nara Sci).

The undoped and As-doped Si based electrodes were fabricated via direct casting of slurry onto Cu foil using the Dr. Blade method. The slurry was prepared by thoroughly

FIGURE 1 | Plan-view SEM image of (A) undoped and (B) As-doped Si powders after ball-milling.

mixing of active material and PAI (Polyamide-Imide, Solvay Torlon 4000T) binder with the wt.% ratio of 9:1, respectively, in NMP (N-Methyl-2-pyrrolidone, Merck) solvent. After casting the electrodes were dried in convection oven at 110◦C to remove NMP. The loading level of electrodes were adjusted to 5 mg/cm<sup>2</sup> .

Electrochemical performances were studied using coin-type half configuration cells with Li metal as the counter electrodes, which were assembled in an Ar-filled glove box with <1 ppm of both oxygen and moisture. The electrolyte solution used was 1.3 M lithium hexafluoro phosphate (LiPF6) dissolved in a solvent mixture of ethylene carbonate (EC) and diethyl carbonate (DEC) with a 3:7 volume ratio. Galvanostatic measurements were carried out to determine charge-discharge capacity and cycling retention in the potential range between 0.005 and 2.000 V vs. Li/Li<sup>+</sup> under a current density of 1.0 A g−<sup>1</sup> (0.28 C) using a computer-controlled battery measurement system (TOSCAT 3000 U).

The surface morphologies of Si microflake anodes were evaluated by field emission scanning electron microscope (FE-SEM). The stress evolution, phase transition, and crystallinity of Si microflake anodes were investigated by micro Raman spectroscopy and X-ray diffraction (XRD).

#### RESULTS AND DISCUSSION

**Figures 1A,B** shows the plan-view SEM images of undoped and As-doped Si negative electrodes, respectively, after ball-milling for 2 h. The irregular and agglomerated particles were made up with the average size of 10µm for both undoped and As-doped Si anodes. The particle size and distributions were measured by particle size analyzer (not shown in this paper). The similar morphology and size of micro particles of undoped and Asdoped Si anodes were observed, indicating that As-dopants do not influence the morphology and particle size distribution of negative anodes.

**Figure 2** exhibits XRD patterns and micro-Raman spectra of undoped and As-doped Si negative electrodes. The phase and crystallinity of Si anodes used in this work were investigated by XRD as shown in **Figure 2A**. It should be noted that all XRD peaks are sharp. This feature could be ascribed to the property of

single crystalline Si. It is found that all the peak positions of XRD patterns can be indexed to those of cubic Si. It is interesting to note that the peak positions of As-doped Si anodes were shifted to the lower angles, which is attributed to the increased lattice

constant by As-doping into Si structure as depicted in the inset of **Figure 2A**. Typical stress profiles of undoped and As-doped Si negative electrodes were studied by micro-Raman spectroscopy measurements at room temperature, as shown in **Figure 2B**. It is well-known that F2g-phonon modes in crystalline Si can be identified with peaks at around 520 cm−<sup>1</sup> (Shimizu et al., 2015). We can easily notice that the peak position for the As-doped Si anode has been red-shifted, indicating the residual tensile stress of approximately ∼1.7 GPa compared to the undoped one. The residual stress can be calculated using the following equation (Wu et al., 2007):

$$\sigma(\text{MPa}) = 434 \times \Delta\omega \tag{1}$$

where σ is the biaxial stress, 1ω is the F2g phonon peak shift. These data are in good agreement with the XRD results in **Figure 2A**.

In order to evaluate the electrochemical characteristics of Si negative electrodes with and without dopants, the various electrochemical examinations were performed. **Figure 3** illustrates the first charge and discharge behavior of undoped and As-doped Si thick film anodes. It is clearly seen that there are plateaus in the curves, which is attributed to lithiation and delithiation by the reactions of Si atoms with Li ions (Chevrier et al., 2010). The initial charge potential plateau is located near 0.1 V, which is in a good agreement with the formation of LixSi via lithiation of crystallized Si (Han et al., 2016). The first charge and discharge capacity for undoped and As-doped Si negative electrodes are 3,733, 2,782, 3,578, and 2,722 mAh g−<sup>1</sup> , respectively. These values correspond to the initial CE of 74.5 and 76.1%, respectively.

**Figure 4A** illustrates the cycle performance of the cells with the undoped and As-doped Si negative electrodes. The discharge capacities of the LIBs with both types of the anodes were evaluated in a cut-off potential range of 0.005–2.000 V as seen in **Figure 3**. The discharge capacity of LIB with undoped Si negative electrode was higher than that of As-doped Si in all cycle ranges. One can easily observe that the discharge capacity fading of undoped Si anode is comparable with that of As-doped Si. Notably, a fading rate of discharge capacity with the undoped Si negative anode changed more sharply near 10 cycles. The

initial discharge capacities of both anodes were decreased by approximately 50% after 30 cycles and their discharge capacities arrived at similar extent. **Figure 4B** shows the CE of both investigated materials. The subsequent cycle CE of cells with undoped and As-doped Si anodes start from 98.5 and 99.6% and further reaches over 95.0 and 95.8% after 30 cycles, respectively. In addition, the capacity retentions of both cells were measured as shown in **Figure 4C**. Despite the discharge capacity of undoped Si anode was higher than that with As-doped Si one, its capacity retention displays similar trends, corresponding to about 50% of retention. This phenomena can be explained by the fact that the As-doped Si anodes consumed more Li-ions for the generation of a solid-electrolyte interface (SEI) compared with the undoped Si anode and high energy barrier in energy band diagram, which will be mentioned later.

**Figure 5** depicts 2theta X-ray rocking curves of the undoped and As-doped Si negative electrodes with various cut-off voltages (OCV) in charging/discharging states after the 1st cycle. After the formation process, there is no sign of the Li15Si<sup>4</sup> phase presence in XRD curves for undoped and As-doped Si anodes. Besides, XRD patterns of both undoped and As-doped Si anodes discharged up to a potential of 0.005 V clearly revealed the presence of the peaks at 39◦ and 41◦ , indicating the formation of Li15Si<sup>4</sup> silicate (Li and Dahn, 2007). It is essential to note that the intensity of XRD patterns of undoped Si negative electrodes was much higher than that of As-doped Si one. This implies that the phase transition of undoped Si anodes into Li15Si<sup>4</sup> phase is easier to occur compared to the doped electrode. In the discharging states, the peak intensity of Li15Si<sup>4</sup> reduced steadily with increasing the OCV, and finally disappeared at OCV of 2 V for undoped and As-doped Si anodes.

In order to estimate the phase transition of undoped and As-doped Si anodes, we performed XRD analysis as a function of cycle number as illustrated in **Figure 6**. No Li silicide peaks were found in the XRD spectra of the samples before cycling (not shown in this paper). The Li15Si<sup>4</sup> phase peaks in XRD

curves after 10 and 30 cycles were easily confirmed as seen in **Figures 6A,B**. Indeed, the XRD intensity of Li15Si<sup>4</sup> phase in undoped Si anode is much higher than that in As-doped Si, which is consistent with the results obtained in **Figure 5**. In addition, the gradual increase of Li15Si<sup>4</sup> phase upon 30 cycles is assigned to the activation process of Li-ion pathway in the electrode-electrolyte system, despite the fact that activation was accompanied by the fast capacity fading of both electrodes as observed in galvanostatic cycling.

Furthermore, we investigated the morphology of the electrodes at their cross-section and surface before and after 30 cycles using SEM imaging. Cross-sectional SEM images of undoped and As-doped Si (**Figures 7A,B**) showed that both electrodes have the thickness of approximately 20µm before cycling, while after 30 cycles, thickness increased drastically (>200%) that can be observed in **Figures 7C,D**. Besides, both samples exhibited similar changes in volume expansion. From the top-view images of undoped Si vs. Asdoped Si (**Figures 7E–H**), one can see that the morphology of electrodes looked the same before and after cycling. The extensive formation of cracks and partial pulverization of active electrode material after 30 cycles was observed in both cases shown in **Figures 7G,H**. Accordingly, SEM results supported the capacity fading phenomena revealed by the galvanostatic tests earlier for the undoped and As-doped Si anodes.

**Figure 8** compares the discharge capacity retentions of undoped and As-doped Si electrodes at various current densities from 0.1 to 4.0 C (1C = 1.0 A/g), respectively. Here, undoped Si exhibits considerable fading in discharge capacity at higher current densities compare with As-doped Si. More specifically, at the highest current density of 4.0 C, the undoped Sibased electrode shows 69% in capacity retention vs. 0.1 C. By contrast, the As-doped Si demonstrates better rate performance with ∼80% capacity retention. Obvious improvement in rate capability can be ascribed to the As-doping effect in Si structure and will be further discussed with the electronic energy band diagram.

To understand the charge/discharge characteristics of undoped and As-doped Si negative electrodes, we took the electronic energy band diagram of both Si anode-electrolyte interfaces into consideration as illustrated in **Figure 9**. It is well-known that when the semiconductor-electrolyte interface is formed, electrolyte and semiconductor reach electrochemical equilibrium by matching their electrochemical potentials. This occurs by charge transfer between both media (Prados et al., 2014). Considering this, each barrier heights of undoped and Asdoped Si anodes are represented in **Figures 9A,B**, respectively. It is noteworthy to mention that the energy barrier height from As-doped Si anode side, qVAs−doped, was found to be higher than qVundoped of undoped Si anode after equilibrium state. Hence, the flux of electrons across the electrolyte in As-doped Si with the increased potential barrier height can be impeded, thus hindering charge flow in the cell. In previously published works, the addition of dopants such as B and P decreased the lattice spacing of Si anodes and led to the increase of the insertion energy of Li ions, thus suppressing the discharging and charging capacity in LIBs (Long et al., 2011; Domi et al., 2016). However, the charging/discharging capacities of As-doped Si anodes have a different aspect, apart from volume change. To understand the electrochemical behaviors of cells with As-doped Si anodes, the energy band diagrams were illustrated, as shownv in **Figure 9**. It was found that the potential barrier of cells with As-dopant is much higher than that of undoped Si anode. We can infer that the higher potential barrier on the anode side by highly n-type doping results in the reduced charging/discharging capacity of As-doped Si anodes, compared to that of undoped Si ones. This is in a fair agreement with the results of **Figures 5**, **6**. Moreover, we consider that slightly higher CE of As-doped Si

#### REFERENCES


negative electrode is related to the potential barrier. A lower charge/discharge capacity leads to the suppression of the phase transition from c-Si to a-Si, which may result in the improved CE, corresponding to the XRD results.

#### CONCLUSION

In summary, the electrochemical performance of the cells with As-doped Si negative electrodes was investigated. The ballmilling of single-crystalline Si wafers was used to fabricate the anodes, which particles were irregular and agglomerated microflakes. The XRD and Raman analysis revealed that As-doped Si anodes have larger lattice spacing compared to undoped Si samples, indicating the residual tensile stress of approximately ∼1.7 GPa. Despite the increased lattice spacing in As-doped Si anode, its charge/discharge capacity was lower compared with that of undoped Si anodes. The Coulombic efficiency and cyclic retention of As-doped Si negative electrodes, however, were comparable to those of the undoped Si anode. This effect was attributed to the increased potential barrier height on the side of As-doped Si, affecting the electrochemical performance of LIBs with the As-doped anode. We believe that this study will provide an effective approach to achieve the enhanced electrochemical performances of LIBs with Si-based anode derived from electronic waste materials.

#### DATA AVAILABILITY

The datasets generated for this study are available on request to the corresponding author.

#### AUTHOR CONTRIBUTIONS

ML and DY conceived design of the study and carried out the experiment; UL, ZB, and S-SK contributed to the interpretation of the results; NU, AM, ZB, and S-SK wrote sections of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

#### ACKNOWLEDGMENTS

The authors acknowledge the financial support of the KETEP (Grant No.20164010201070) and the Nazarbayev University by the research grant No.110119FD4504 Development of 3D all solid-state thin film Li-ion microbatteries.

Goodenough, J. B., and Park, K. -S. (2013). The Li-ion rechargeable battery: a perspective. J. Am. Chem. Soc. 135, 1167–1176. doi: 10.1021/ja3091438


**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 Lee, Yoon, Lee, Umirov, Mukanova, Bakenov and Kim. 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.

Frontiers in Chemistry | www.frontiersin.org

# High-Performance Solid Composite Polymer Electrolyte for all Solid-State Lithium Battery Through Facile Microstructure Regulation

Jingjing Yang<sup>1</sup> \* † , Xun Wang<sup>1</sup> , Gai Zhang<sup>1</sup> , Aijie Ma<sup>1</sup> , Weixing Chen<sup>1</sup> , Le Shao<sup>2</sup> \* † , Chao Shen<sup>3</sup> and Keyu Xie<sup>3</sup> \*

*<sup>1</sup> School of Materials and Chemical Engineering, Xi'an Technological University, Xi'an, China, <sup>2</sup> Shaanxi Coal Chemical Industry Technology Research Institute Co. Ltd., Xi'an, China, <sup>3</sup> State Key Laboratory of Solidification Processing, School of Materials Science and Engineering, Center for Nano Energy Materials, Northwestern Polytechnical University and Shaanxi Joint Laboratory of Graphene (NPU), Xi'an, China*

#### Edited by:

*Jianqing Zhao, Soochow University, China*

#### Reviewed by:

*Wangwang Xu, Louisiana State University, United States Xiaowei Wang, National University of Singapore, Singapore*

#### \*Correspondence:

*Jingjing Yang yangjingjing@xatu.edu.cn Le Shao shaole@sxccti.com Keyu Xie kyxie@nwpu.edu.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *27 March 2019* Accepted: *14 May 2019* Published: *31 May 2019*

#### Citation:

*Yang J, Wang X, Zhang G, Ma A, Chen W, Shao L, Shen C and Xie K (2019) High-Performance Solid Composite Polymer Electrolyte for all Solid-State Lithium Battery Through Facile Microstructure Regulation. Front. Chem. 7:388. doi: 10.3389/fchem.2019.00388* Solid composite polymer electrolytes are the optimal candidate for all solid-state lithium batteries, because of their enhanced ionic conductivities, long-life cycle ability and compatibility to lithium anode. Herein, we reported a kind of solid composite polymer electrolyte comprised of poly(ethylene oxide), graphitic-like carbon nitride and lithium perchlorate, which was prepared by a facile solution blending method. Microstructure of the solid composite polymer electrolyte was regulated by thermal annealing and interaction among components and was characterized by XRD, DSC, FTIR-ATR, and ROM. The obtained solid composite polymer electrolyte achieved an ionic conductivity as high as 1.76 ×10−<sup>5</sup> S cm−<sup>1</sup> at 25◦C. And the electrochemical stable window and the lithium ion transference number, t+, were also obviously enhanced. LiFePO4/Li solid-state batteries with the annealed PEO-LiClO4-g-C3N<sup>4</sup> solid polymer electrolyte presented a high initial discharge capacity of 161.2 mAh g−<sup>1</sup> and superior cycle stability with a capacity retention ratio of 81% after 200 cycles at 1C at 80◦C. The above results indicates that the thermal annealing treatment and g-C3N<sup>4</sup> as a novel structure modifier is crucial for obtaining the high-performance solid composite polymer electrolytes used in the all solid-state lithium battery.

Keywords: all solid-state lithium battery, solid composite polymer electrolyte, microstructure, graphite-like carbonitride, electrochemical property

## INTRODUCTION

During recent years, high-energy-density storage batteries are urgently needed to satisfy the increasing demand in electric vehicles, consumer electronics and grid energy storage (Manthiram et al., 2017; Fan et al., 2018). Lithium metal secondary batteries have been considered as the potential candidate for high-energy-density storage batteries, since lithium metal anode possessed the highest theoretical specific capacity (3,862 mAh g−<sup>1</sup> ) and the lowest reduction potential (−3.04V VS. standard hydrogen electrode) (Shen et al., 2018; Wu et al., 2018). Moreover, lithium metal used asthe anode can even act as the lithium source in the battery with non-lithiated materials such as sulfur or oxygen as the cathode to achieve higher storage capacity compared to the current

**21**

commercial lithium-ion batteries (Pan et al., 2019). However, practical and large-scale development of lithium metal batteries with liquid electrolytes is limited by safety concerns and the rapid capacity attenuation and short cycling capability (Fan et al., 2018). During the charge-discharge process of lithium metal secondary batteries with organic liquid electrolytes, the propensity of faster lithium dendrites formation was resulted from the plating and stripping of Li ions on the anode, which would result in the capacity attenuation and poor cycling capability and short-circuit of cells.

In order to further develop the practical use of lithium metal as anode, scientists made great efforts to solve the lithium dendrites problems and achieved much progress on suppressing dendrite formation and growth (Li D. et al., 2018; Nie et al., 2018; Shen et al., 2018; Wu et al., 2018; Pan et al., 2019). Strategies such as electrolyte modification, multifunctional barriers, composite metallic lithium electrode, and 3D current collectors were proposed to suppress the formation of Lithium dendrites (Wu et al., 2018; Xiao et al., 2018). Replacement of current organic liquid electrolytes with the solid-state electrolytes is considered as the most promising way to realize excellent performance of lithium metal secondary batteries (Goodenough, 2018; Nie et al., 2018). Because the organic liquid electrolyte is flammable, toxic and environmental contamination, solidstate electrolytes are therefore of crucial importance because of their excellent safety and mechanical ability for dendritic growth inhibition (Ban et al., 2018). And all solid-state lithium metal batteries are widely regarded as promising candidates for next generation of energy storage devices with improved energy density and superior safety performances.

Solid-state electrolytes are generally divided into inorganic superionic conductors (or the solid ceramic electrolytes) and solid polymer electrolytes. Compared with the inorganic superionic conductors, solid polymer electrolytes, benefiting from shape versatility, flexibility, light weight and low processing costs, are being investigated as promising candidates to replace currently available organic liquid electrolytes in lithium metal batteries (Bae et al., 2018b), (Zhang X. et al., 2018). However, due to the inherent low ionic conductivity at RT and poor mechanical property of solid polymer electrolytes, polymer-based solid composite electrolytes taking advantage of the merits of both inorganic and polymeric materials have been attracting more and more attention in all solid-state lithium metal battery applications (Tan et al., 2018). And Poly(ethylene oxide) (PEO) based solid composite electrolytes were widely considered as promising candidates solid electrolytes for high energy density lithium metal batteries (Xue et al., 2015; Tan et al., 2018). Numerous fillers were added into the PEO-based composite solid electrolyte to improve the ionic conductivity and the interfacial properties in contact with the electrodes (Tan et al., 2018). And it was the simplest method to achieve the composite polymer electrolyte, whose physical properties can be easily controlled by compositional change (Tan et al., 2018; Xiao et al., 2018; Zhang J. et al., 2018). However, the particle agglomeration especially at high concentration of fillers would inhibit the further improvement of properties of the solid polymer electrolyte. New kind of effective filler or percolated nanofiller structure needs to be developed by researchers.

Graphitic-like carbon nitride (g-C3N4) is a stacked 2D structure, which was mainly composed of carbon, nitrogen and a few residual –NH<sup>2</sup> or –NH groups, and was metal-free and lightweight (Wang et al., 2012; Shi et al., 2014). And its intrinsic polarity and semiconductor property may be helpful to modulate the spatial and interface distribution of ionic charge carriers (Hu et al., 2017). These properties make g-C3N<sup>4</sup> a potential candidate as polymer electrolyte filler for all-in-solid lithium batteries. Shi et al. proved that the introduction of 6.0 wt% g-C3N<sup>4</sup> nanosheets in the sodium alginate (SA) nanocomposite films could greatly enhance the thermal stability and mechanical properties of SA biopolymer electrolyte nanocomposite films (Shi et al., 2014). Hu et al. reported about a kind of composite electrolyte with bis(triffuoromethanesulfonimide) lithium salt (LiTFSI), di(ethylene glycol) dimethyl ether (DGM) and g-C3N4, which was used to suppress lithium dendrite growth during cycling and improve the long-term cycling property of the composite electrolyte (Hu et al., 2017). But quantitative and even qualitative descriptions of the interactions between separate components of the state electrolyte system are still lacking. Understanding of the different interactions in the system and its influence on the microstructure and final electrochemical properties of the solid composite polymer electrolyte is crucial for achieving high performance solid polymer electrolyte.

In this article, we present for the very first time, to the best of our knowledge, the addition of g-C3N<sup>4</sup> as fillers in PEO and its influence on the lithium ionic dissociation in lithium perchlorate (LiClO4) and transport and the crystallization of PEO in solid composite polymer electrolytes. Thermal annealing treatment and the interactions regulation among components by addition of g-C3N<sup>4</sup> were introduced to tailor the microstructure of the solid composite polymer electrolyte. And through controlling the microstructure of solid composite polymer electrolytes, excellent improvement of the desired physical and chemical properties and electrochemical properties was achieved. When the weight percent of g-C3N<sup>4</sup> was 10%, the annealed solid composite polymer electrolyte gave the best performance in the all solidstate lithium battery with lithium metal as the anode and LiFePO<sup>4</sup> as the cathode.

### EXPERIMENTAL SECTION

### Materials

Poly(ethylene oxide) (Mw = 6 × 10<sup>5</sup> , Alfa Aesar), lithium perchlorate (LiClO4, Aladdin, battery grade), poly(vinylidene fluoride) (PVDF, Solef <sup>R</sup> 5130, Solvay), and LiFePO<sup>4</sup> (DY-3, Shenzhen Dynanonic Co., Ltd) were commercially obtained and were dried before usage. Acetonitrile (ACN, Aladdin, AR), Nmethly-2-pyrrolidone (NMP, AR), and urea (Aladdin, AR) were used as obtained. The Li metal used in our experiments was commercially obtained with a diameter of 15.8 and 0.6 mm of thickness.

## Preparation of g-C3N<sup>4</sup>

The g-C3N<sup>4</sup> filler was synthesized by thermal polymerization as we reported in the previous article (Wang et al., 2019). Twenty gram of urea was put into a crucible with a cover in the muffle furnace. And then the crucible was heated to 580◦C to stay for 4 h, with a heating rate of 3◦C/min. And the powder product was in light yellow color. The detailed characterization of pure g-C3N<sup>4</sup> was displayed in **Figures S1**, **S3**.

### Solid Composite Polymer Electrolyte Films Preparation

Firstly, the prepared g-C3N<sup>4</sup> filler was ultrasonic dispersed in ACN. Then PEO and LiClO<sup>4</sup> (molar ratio of ethylene oxide (EO) to LiClO<sup>4</sup> is 15:1) were added into the solution under stirring, in order to obtain uniform mixture solution. The weight percentage of g-C3N<sup>4</sup> compared to PEO was controlled as 0, 5, 10, 15, and 20%, respectively. The mixture solution was casted into a homemade Teflon mold and dried in the vacuum oven at 40◦C for 48 h. And the above mentioned state polymer electrolytes films were firstly annealed at 120◦C for 2 h and then quenched to RT before use. In contrast, the state polymer electrolyte films were also prepared without thermal annealing process. The thickness of obtained solid state polymer electrolyte films was regulated to about 100µm. And the state polymer electrolytes films were cut into rounds with diameter of 19 mm before loaded into the glovebox.

### Characterization

Bruker Vertex 70 FTIR spectrometer with Access ATRTM attachment was used to characterize the composition of samples. The melting temperature of PEO in samples was collected by a differential scanning calorimeter (DSC, Mettler DSC823e) under nitrogen atmosphere. Samples with a mass of 5∼10 mg were sealed into aluminum crucible, and then heated at a rate of 10◦C/min. X-ray powder diffraction (XRD) patterns were collected by a XRD-6000 (Shimazu, Japan) X-Ray diffractometer equipped with Cu K<sup>α</sup> radiation in the range of 5∼50◦ at a scanning rate of 4◦ /min. The reflection optical microscopy (ROM) observations were performed with a Leica DM2500P optical microscope equipped with a C-5050ZOOM camera.

The ionic conductivity of SPEs was measured by electrochemical impedance spectroscopy (EIS) by using a Princeton PARSTAT 4000A (AMETEK) electrochemical workstation over frequency range from 1 to 10<sup>5</sup> Hz with an applied voltage of 10 mV. The measured SPEs were sandwiched between two stainless steel electrodes, which were assembled into CR2025 coin cells. And the temperature was controlled by an oven in the range of 25∼120◦C. The electrochemical stability window was determined by linear sweep voltammetry (LSV) by a Princeton PARSTAT 4000A (AMETEK) electrochemical workstation. The cell was assembled by sandwiching SPEs between a stainless steel and lithium metal (SS/SPE/Li) as the counter and reference electrode, respectively. And the measurement was cycled from 2.0 to 6.0 V (VS. Li/Li+) at a scan rate of 10 mV/S. The lithium transference number (t+) was tested in a symmetric cell (Li/SPE/Li) by chronoamperometry combined with electrochemical impedance spectroscopy according to the method provided by Bruce et al. (Evans et al., 1987). In order to characterize the battery performances, the prepared state polymer electrolytes were fabricated as CR2025 coin-type cells with lithium metal anode and LiFePO<sup>4</sup> cathode. And LiFePO<sup>4</sup> was blended with carbon black (Super-P) and

poly(vinylidene fluoride) at a weight ratio of 8:1:1 with NMP as the solvent. The weight of cathode used in the test was controlled to about 2.40 mg. As a result, the weight of active material was about 1.92 mg. The battery performances of coin cells were examined at 80◦C using a Battery Testing System (NEWARE CT-4008, Shenzhen, China), including cycle property and C-rate capability (0.1, 0.2, 0.5, 1, 2, and 5C). The cut-off voltages were chosen as 4.2 V (charge) and 2.5 V (discharge) during the charge-discharge cycle performance.

### RESULTS AND DISCUSSION

### Microstructure Regulation by Thermal Annealing and Interaction Among Components in Solid Composite Polymer Electrolytes

In order to avoid the effect of thermal history on the crystallization of PEO during the solid composite polymer electrolyte preparation process to an extreme, 120◦C annealing and then quenching method was introduced, as shown in **Figure 1**. And we have reported on this preparation method in detail in a previous work (Wang et al., 2019). During the solid polymer electrolyte preparation process, samples were firstly dried at 40◦C for 48 h, which would provide an isothermal crystallization condition for PEO. As a result, the crystallinity of PEO in the samples was high. However, when the sample was quenching from the amorphous state, the non-isothermal crystallization might happened, which gave a lower crystallinity of PEO in the sample. And this deduction was further proved by the following XRD and DSC results.

In order to clarify the influence of thermal annealing and the interaction among components of the solid composite polymer electrolyte on the microstructure of solid composite polymer electrolytes, XRD was used to characterize the crystallization of samples as shown in **Figure 2**. As shown in **Figure 2A**a, pure g-C3N<sup>4</sup> gave a typical diffraction peak at 27.5◦ , which was the typical peak of (002) crystal plane of g-C3N4. Moreover, the interlamellar spacing of g-C3N<sup>4</sup> was calculated as about 0.32 nm by Bragg equation. On the other hand, XRD patterns of the solid composite polymer electrolyte samples shown in **Figures 2A**b,c displayed three typical peaks at 19.3◦ , 23.4◦ and 27.5◦ , respectively. The two characteristic peaks at 19.3◦ and 23.4◦ were corresponded to (120) and (132)/(032)/(212)/(112) crystal planes of PEO, respectively (Wang et al., 2019). It was obvious that the intensity of typical diffraction peak of PEO at

19.3◦ and 23.4◦ greatly decreased after the sample annealing at 120◦C and then quenching to RT, as shown in **Figures 2A**b,c.

Furthermore, the diffraction peaks of crystalline phase and amorphous phase in XRD patterns were fitted by the Jade 6.0 software. And the crystallization degree of PEO in samples was calculated by divided the summation of peak areas of crystallization phase and amorphous phase by the summation of peak areas of crystalline phase. Therefore, the crystallization degree of PEO was decreased from about 15.49% to about 8%, when the solid composite polymer electrolyte was thermal treated. This result demonstrated that the thermal annealing process could greatly suppress the crystallization of PEO.

As shown in **Figure 2B**, the intensity of diffraction peak observed at 27.5◦ increased with the content of g-C3N<sup>4</sup> increasing. **Figures 2B**a–e showed that intensity of typical peaks of PEO crystals was firstly decreased and then increased due to the addition of g-C3N4, with a minimum value appeared in the solid composite polymer electrolyte with 10% g-C3N4. Moreover, when the content of g-C3N<sup>4</sup> was chosen as 0, 5, 10, 15, and 20%, the crystallization degree of PEO was calculated as about 28, 25, 8, 19, and 20%, respectively. It was obvious that the introduction of g-C3N<sup>4</sup> could significantly decrease the crystallinity of PEO. But the crystallization degree of PEO was firstly decreased and then increased as the content of g-C3N<sup>4</sup> increased, with a minimum value in the solid composite polymer electrolyte with 10% g-C3N4. In the state polymer electrolyte, the porous microstructure of g-C3N<sup>4</sup> sheets (as shown in **Figure S1**) might serve as the hard limitation and inhibit the ordered chain arrangement of PEO, while on the other hand also serve as the nucleation sites for PEO chains. As a result, the heterogeneous nucleation could be enhanced, when the content of g-C3N<sup>4</sup> sheets was over 10%, which would improve the crystallization of PEO. Furthermore, the g-C3N<sup>4</sup> nanosheets might change the priority growth of PEO crystal planes, as shown by the relative intensity of fitting peaks at 19.3◦ and 23.4◦ in **Figures 2B**a,e. It implied that the interaction between g-C3N<sup>4</sup> and PEO could greatly influence the crystallization preference of crystal planes in PEO.

The effect of thermal annealing and the interaction among components of the solid composite polymer electrolyte on the crystallization of PEO was further proved by DSC. As shown in **Figures 3A**a,b the melting temperature of PEO was decreased from 52.4 to 48.7◦C, when the state polymer electrolyte was annealed at 120◦C and then quenched to RT. Also shoulder peak appeared at 46.0◦C in **Figure 3A**a. It meant that PEO crystals in the sample undergone thermal annealing formed lamellae with thinner thickness and the thickness of lamellae gave a wider distribution. It implied that the thermal annealing process could regulate the crystallization of PEO.

Furthermore, the melting behavior of PEO in PEO homopolymer, PEO-LiClO<sup>4</sup> electrolyte and the solid composite polymer electrolytes with different content of g-C3N<sup>4</sup> (5, 10, 15, and 20%), which were treated by the thermal annealing process, was displayed in **Figure 3B**. Due to the interaction between PEO chains and LiClO4, the melting temperature of PEO crystals decreased from 66.7 to 53.3◦C, as shown in **Figures 3B**a,b. Moreover, the addition of g-C3N<sup>4</sup> further restrained the crystallization of PEO, as shown in **Figures 3B**c–f. The melting temperature of PEO crystals in the solid composite polymer electrolyte was 51.5, 48.7, 52.3, and 53.1◦C, respectively, with the content of g-C3N<sup>4</sup> increasing from 5 to 20%. The melting temperature of PEO firstly decreased and then increased, with a minimum value of 48.7◦C in the sample with 10% g-C3N4. Also the crystallinity of PEO was calculated by dividing the melting enthalpy obtained from DSC curves in **Figure 3B** by the equilibrium enthalpy of PEO fusion (197 J/g). And the crystallinity of PEO in different samples shown in **Figure 3B** was about 54, 25, 22, 7, 18, and 20%, respectively. With the addition of LiClO4, the coordination effect of EO unit and Li<sup>+</sup> would decrease the crystallization of PEO. Moreover, the introduction of g-C3N<sup>4</sup> further decreased the crystallization of PEO. The residual –NH<sup>2</sup> or –NH groups presenting in g-C3N<sup>4</sup> can act as active sites for hydrogen bonding with ether oxygen in PEO.

same [EO]: [Li+] ratio of 15:1.

FIGURE 4 | ROM images of solid composite polymer electrolytes prepared with thermal annealing at 120◦C [(a) PEO-LiClO4; (b) PEO-LiClO4-5% g-C3N4; (c) PEO-LiClO4-10% g-C3N4; (d) PEO-LiClO4-15% g-C3N4; (e) PEO-LiClO4-20% g-C3N4] and without thermal annealing [(f) PEO-LiClO4-10% g-C3N4].

And the hydrogen bonds formation restrained the chain folding of PEO, and then reduced the crystallinity of PEO in the solid composite polymer electrolytes. On the other hand, the existence of g-C3N<sup>4</sup> also served as the heterogeneous nucleation sites for PEO crystallization. As a result, the crystallinity degree of PEO firstly decreased and then increased slightly in the solid composite electrolytes, with 10% as the optimum content. This result was in good accordance with the foregoing XRD results.

The influence of thermal annealing and the interaction among components of the solid composite polymer electrolyte on the microstructure morphology of solid composite polymer electrolyte was observed by ROM. Perfect spherulites with diameter of about 100µm were formed in the solid PEO-LiClO<sup>4</sup> polymer electrolyte, as observed in **Figure 4a**. **Figures 4b–e** showed that the morphology perfection and the crystal's size and number of PEO obviously changed with the content of g-C3N4. PEO spherulites were firstly disappeared and then re-formed with the weight percent of g-C3N<sup>4</sup> increasing, with 10% as a critical content. It was ascribed to the hydrogen bonding interaction between g-C3N<sup>4</sup> and PEO that could inhibit the crystallization of PEO, while g-C3N<sup>4</sup> could also serve as the heterogeneous nucleation sites for PEO. Therefore, when the content of g-C3N<sup>4</sup> was over 10%, more and more spherulites were formed in the sample. It implied that the interaction between PEO and g-C3N<sup>4</sup> could greatly affect the morphology of the solid composite polymer electrolyte.

Moreover, compared **Figure 4c** with **Figure 4f**, the thermal annealing treatment might greatly change the morphology of PEO crystals from spherulites to lamellae stack. It demonstrated that the thermal treatment also influenced the morphology of the solid composite polymer electrolyte. The morphology of the solid composite polymer electrolyte was tailored by the thermal treatment and the interactions among components in the sample. And the morphology would play an important role in the compatibility between solid electrolytes and electrodes. Uniformly distributed PEO crystals might control the nucleation of lithium, which would help to inhibit the lithium dendrite formation.

### Effects of Interaction Among Components in Solid Composite Polymer Electrolytes on the Lithium Salt Dissociation

FTIR spectra of PEO, pure g-C3N4, the PEO-LiClO<sup>4</sup> solid polymer electrolyte and the PEO-LiClO4-g-C3N<sup>4</sup> solid composite polymer electrolyte were shown in **Figure S4**. And the typical band of LiClO<sup>4</sup> was identified at 624 cm−<sup>1</sup> . In order to study the effect of g-C3N<sup>4</sup> on the dissociation of LiClO<sup>4</sup> in the solid composite polymer electrolytes, FTIR-ATR spectra in the range of 610–650 cm−<sup>1</sup> were carefully analyzed. It was reported that the peaks at ∼624 and ∼635 cm−<sup>1</sup> were assigned to the dissociated free anion and the bonded ion pair in LiClO4, respectively, Lin et al. (2016). As shown in **Figure 5**, the absorbance peak of dissociated free anion shifted from 624 to 623 cm−<sup>1</sup> , as the weight percent of g-C3N<sup>4</sup> increased to 15%. It demonstrated that the vibration energy level transition of free ClO<sup>−</sup> <sup>4</sup> was easier, due to the interaction of g-C3N<sup>4</sup> and LiClO4. It is because the residual –NH<sup>2</sup> and –NH could coordinate with the lithium ion, which would accelerate the dissociation of LiClO4. Moreover, the relative absorbance of peaks at ∼624 and ∼635 cm−<sup>1</sup> (labeled as A624/A635) was obtained from curves in **Figure 5**. And A624/A<sup>635</sup> was 2.24, 3.01, 4.79, 4.34, and 4.41, respectively, with the weight percent of g-C3N<sup>4</sup> increasing from 0 to 20%. It demonstrated that the introduction of g-C3N<sup>4</sup> could enhance the dissociation of LiClO4, with 10% as the optimum

content. Also the residual –NH<sup>2</sup> and –NH groups in g-C3N<sup>4</sup> could form hydrogen bonds with PEO, which would limit the ordered folding of PEO chains and helped to enhance the chain segment motion. As a result, the dissociation of LiClO<sup>4</sup> was accelerated, due to the coordination between Li<sup>+</sup> and the ether oxygen. The above factors contributed to the separation of Li<sup>+</sup> and ClO<sup>−</sup> 4 in the solid composite polymer electrolyte. And the dissociation of LiClO<sup>4</sup> in the solid composite polymer electrolyte is critical for the good ionic conductivity.

### Ionic Conductivity and the Electrochemical Properties of the Solid Composite Polymer Electrolytes

In order to confirm our hypothesis of the increasing conductivity by thermal annealing treatment and the addition of g-C3N4, the ionic conductivity and electrochemistry properties of the solid composite polymer electrolytes were further studied. The ionic conductivity of the solid composite polymer electrolytes with different content of g-C3N<sup>4</sup> was investigated by EIS. The Arrhenius plot of ionic conductivity with temperature from 25 to 120◦C was displayed in **Figure 6**. Compared with that of PEO-LiClO<sup>4</sup> polymer electrolyte (**Figure 6**e), significant improvement

of ionic conductivity was observed in the annealed solid composite polymer electrolyte. Moreover, the ionic conductivity was firstly increased and then decreased at each temperature with the content of g-C3N<sup>4</sup> increasing from 5 to 20%, as shown in **Figures 6**a–d. The solid composite polymer electrolyte with 10% g-C3N<sup>4</sup> exhibited the maximum ionic conductivity (1.76 × 10−<sup>5</sup> S cm−<sup>1</sup> ) at 25◦C, and the conductivity reached to about 1.08 × 10−<sup>3</sup> S cm−<sup>1</sup> at 100◦C. And this ionic conductivity at room temperature was much higher than that of the recently reported PEO/g-C3N4/LiTFSI solid electrolytes (2.3 × 10−<sup>6</sup> S cm−<sup>1</sup> , at 30◦C) (Sun et al., 2019). Also this ionic conductivity is nearly close to that (2.2 × 10−<sup>5</sup> S cm−<sup>1</sup> at 28◦C) of the PEO based solid composite electrolyte, which needed LiTFSI as the lithium salt and 2D MXene as the filler (Pan et al., 2019). DSC and XRD results demonstrated that the crystallinity of PEO in the solid composite polymer electrolytes was greatly affected by the content of g-C3N4. And when the content of g-C3N<sup>4</sup> was 10%, PEO gave the minimum crystallinity degree, which was preferred for the high ionic conductivity. Besides effects on the crystallinity of PEO, the residual –NH<sup>2</sup> or –NH groups exited on the 2D structure of g-C3N<sup>4</sup> could coordinate with Li+. Also the porous structure might provide a potential transport path for Li+. When the content of g-C3N<sup>4</sup> was 10%, continuous percolating pathways for Li<sup>+</sup> diffusion might be formed. Nonetheless, when the content of g-C3N<sup>4</sup> was over 10%, the redundant inactive fillers would hinder lithium ionic conduction according to effective medium theory (EMT) (Li W. et al., 2018). Moreover, the modification of PEO-LiClO<sup>4</sup> using inorganic nanoparticles fillers were always enslaved to the aggregation of particles, which was also harmful for the ionic conducting (Bae et al., 2018a), Evans et al. (1987). As a result, it demonstrated that the thermal annealed PEO-LiClO4-g-C3N<sup>4</sup> solid composite polymer electrolyte using 10% g-C3N<sup>4</sup> would give an easy way to achieve excellent ionic conductivity in the solid polymer electrolyte.

FIGURE 7 | Chronoamperometry of the Li/ PEO-LiClO4/Li (A) and Li/ PEO-LiClO4-10% g-C3N4/Li (B) cells at a potential of 10 mV at 80◦C. Inset: the AC impedance spectra of the same battery before and after the polarization.

Since the PEO-LiClO4-10% g-C3N<sup>4</sup> gave the best ionic conductivity, it was chosen for the following electrochemical studies. The lithium ion transportation in the solid composite polymer electrolytes was studied as well. And the lithium ion transference number, t+, was analyzed according to the method provided by Bruce and coworkers (Evans et al., 1987). **Figure 7** showed the relation between time and current crossing a symmetric Li/PEO-LiClO4/Li battery (A) and Li/PEO-LiClO4- 10% g-C3N4/Li (B) battery polarized by a small voltage of 10 mV at 80◦C. The inset was the AC impedance spectra of the same battery before and after the polarization. And the t+ can be calculated by the equation: t<sup>+</sup> = I s (1V−I 0R 0 ) I 0 (1V−I <sup>s</sup>R<sup>s</sup> ) , where the I 0 and I s are the initial current and the steady-state current, respectively; 1V is the potential applied across the cell, and R 0 and R s are the initial and steady-state interfacial resistances, respectively. Therefore, the t<sup>+</sup> of the PEO-LiClO<sup>4</sup> electrolyte was 0.241, while the t<sup>+</sup> of the annealed PEO-LiClO4-g-C3N<sup>4</sup> electrolyte was 0.37. It indicated that the thermal annealing treatment coupled with the addition of g-C3N<sup>4</sup> would greatly enhance the mobility of Li+. And this result was in good accordance with the FTIR-ATR results.

Effects of g-C3N<sup>4</sup> on the electrochemical stability of the solid composite polymer electrolytes were characterized by the electrochemical window analysis. The electrochemical stability window of PEO-LiClO<sup>4</sup> electrolyte and annealed PEO-LiClO4-g-C3N<sup>4</sup> electrolyte at 80◦C were measured by LSV, as shown in **Figure 8**. Detected current of PEO-LiClO<sup>4</sup> electrolyte increased sharply as the applying voltage exceeded 3.5 V, as seen from **Figure 8**a. But **Figure 8**b showed that the detected current of PEO-LiClO4-g-C3N<sup>4</sup> electrolyte was more stable and displayed a decomposition voltage at about 4.8 V. It implied that the annealed solid composite polymer electrolyte might guarantee a higher working voltage of electrode. And due to the interaction between g-C3N<sup>4</sup> and PEO, the stability of the solid polymer electrolyte was greatly enhanced. It was reasonable to modify the properties of PEO electrolyte through physically thermal annealing treatment and using g-C3N<sup>4</sup> as the filler.

The battery performance of all-solid-state lithium battery (LiFePO4/Li) based on the solid polymer electrolyte was also performed. **Figure 9** showed the long-term cycling performance and coulombic efficiencies of LiFePO4/Li batteries assembled using the annealed state polymer electrolyte with 10% g-C3N<sup>4</sup> and PEO/LiClO<sup>4</sup> with a current density of 1 C at 80◦C. As shown in **Figure 9A**, the cell based on PEO-LiClO<sup>4</sup> electrolyte gave an initial discharge capacity of only 133.3 mAh g−<sup>1</sup> , and displayed a capacity retention ratio of 80% after 115 cycles. On the other hand, **Figure 9B** demonstrated that the cell based on the solid composite polymer electrolyte with a capacity retention ratio

of 80% after 200 cycles possessed an initial discharge capacity of 161.2 mAh g−<sup>1</sup> , which was very close to the theoretical capacity of LiFePO<sup>4</sup> (170 mAh g−<sup>1</sup> ). Moreover, this cycling performance was better than the reported LiFePO4/Li solid-state batteries based on PEO-LiTFSI-g-C3N<sup>4</sup> polymer electrolytes, which gave an initial discharge specific capacity of 161.3 mAh g <sup>−</sup><sup>1</sup> but cycled for 120 cycles at 60◦C (Sun et al., 2019). The coulombic efficiency of batteries based on the solid composite polymer electrolyte during charge-discharge cycling test was as high as 99.7% in **Figure 9B**, which was higher than that of the battery based on the PEO-LiClO<sup>4</sup> electrolyte shown in **Figure 9A**. Moreover, the prevailing experiments using LiFePO<sup>4</sup> cathode in SPE based solid state lithium batteries were cycled within 3.8 V, in order to guarantee the cycle stability (Nie et al., 2018). In this article, the LiFePO4/Li solid battery could operate in the range of 4.2∼2.5 V and gave excellent cycle stability. Furthermore, the charge-discharge voltage profiles of the corresponding LiFePO4/Li solid battery during the cycling

performance were displayed as the insert in **Figures 9A,B**. The voltage plateau at around 3.48 V was steady in **Figure 9B**, while a sloping voltage plateau was observed in **Figure 9A**. And the attenuation of specific capacity during charge-discharge cycling process was more clearly in the LiFePO4/Li batteries with PEO/LiClO<sup>4</sup> electrolyte shown in **Figure 9A**, compared with that in the LiFePO4/Li batteries with annealed PEO-LiClO4-g-C3N<sup>4</sup> electrolyte shown in **Figure 9B**. It implied that the addition of g-C3N<sup>4</sup> enhanced the electrochemical stability of PEO-based solid polymer electrolytes.

Moreover, in order to ascertain the usefulness of the solid composite polymer electrolyte, the C-rate cycling property of PEO-LiClO<sup>4</sup> electrolyte and the annealed PEO-LiClO4-g-C3N<sup>4</sup> electrolyte at 80◦C were characterized and results were shown in **Figure 10**. As shown in **Figure 10A**, the discharge specific capacity of the all solid-state battery based on the PEO-LiClO<sup>4</sup> electrolyte was 148.0, 150.6, 139.5, 123.1, 80.4, and 22.2 mAh g −1 , respectively, with the C-rate increasing from 0.1, 0.2, 0.5C, 1, 2 to 5C. And it was obvious that the discharge specific capacity was gradually decreased at higher C-rates, which was reported as a typical characteristic of LiFePO<sup>4</sup> material (Brutti et al., 2012). The low discharge specific capacity at 5C-rate was attributed to the electrode polarization and solid electrolyte interface caused by the low electronic conductivity, limited diffusion of Li<sup>+</sup> into the LiFePO<sup>4</sup> structure (Kumar et al., 2014). On the other hand, **Figure 10B** showed the performance of the battery with PEO-LiClO4-g-C3N<sup>4</sup> electrolyte. It delivered an initial discharge specific capacity of 161.0 mAh g−<sup>1</sup> at 0.1C-rate. And the discharge specific capacity was 159.5, 159.7, 158.8, and 155.6 mAh g−<sup>1</sup> , respectively, at 0.2C-rate, 0.5C-rate, 1C-rate, and 2C-rate without any obvious attenuation. The battery was even able to deliver a specific capacity of 59.7 mAh g−<sup>1</sup> even at 5C-rate, which was 168% higher than that of the battery based on the PEO-LiClO<sup>4</sup> solid polymer electrolyte. Moreover, when the rate returned to 0.1C, the average discharge capacity of the battery was still as high as 157.2 mAh g−<sup>1</sup> . Also the coulombic efficiencies were close to 100%. This result implied that the PEO-LiClO4-g-C3N<sup>4</sup> solid polymer electrolyte could facilitate the cycling performance of LiFePO<sup>4</sup> cathode, due to its good compatibility with the cathode and good lithium ionic transport capability. Moreover, the solid state battery based on the solid composite polymer electrolyte was able to light up several LED lamps under a normal condition as shown in **Figure S5**. It demonstrated that the annealed PEO-LiClO4-g-C3N<sup>4</sup> electrolyte could be an outstanding candidate used in all solid-state lithium batteries.

### CONCLUSIONS

To conclude, we reported a novel method to prepare the solid polymer electrolyte with high performance using facile thermal annealing treatment and by addition g-C3N<sup>4</sup> as the modification filler. XRD and DSC results demonstrated that the crystallinity degree of PEO was greatly decreased by the thermal annealing treatment and by interaction between PEO chains and the g-C3N<sup>4</sup> filler. Moreover, the dissociation of LiClO<sup>4</sup> was obviously enhanced as proved by the FTIR-ATR results. The final morphology of the obtained solid composite polymer electrolyte was even changed with the content of g-C3N<sup>4</sup> increasing and changed by the thermal annealing treatment. When the weight percent of g-C3N<sup>4</sup> was 10% in the annealed solid composite polymer electrolyte, spherulites morphology of PEO crystals was totally destroyed, which would facilitate the compatibility between the solid electrolyte and electrodes. And the crystallinity of PEO was only about 8%, which was helpful for the lithium ion transport in the solid composite polymer electrolyte.

As a result, the obtained solid composite polymer electrolyte with 10% g-C3N<sup>4</sup> achieved an ionic conductivity as high as 1.76 × 10−<sup>5</sup> S cm−<sup>1</sup> at 25◦C, due to the above microstructure regulation.

### REFERENCES

Bae, J., Li, Y., Zhao, F., Zhou, X., Ding, Y., and Yu, G. (2018b). Designing 3D nanostructured garnet frameworks for enhancing ionic conductivity and The electrochemical stable window was increased to about 4.8 V. The lithium ion transference number, t+, was changed from 0.24 to 0.37, and increased about 54.8%. Finally, the annealed PEO-LiClO4-10% g-C3N<sup>4</sup> electrolyte was also assembled into an all solid-state battery using LiFePO<sup>4</sup> as the cathode and lithium metal as the anode. And the battery demonstrated excellent performance, with a high initial discharge capacity of 161.2 mAh·g −1 and superior cycle stability with a capacity retention ratio of 81% after 200 cycles at 1C at 80◦C. Furthermore, the C-rate cycling results also demonstrated that the stability of the assembled solid state battery was markedly improved by using PEO-LiClO4-10% g-C3N<sup>4</sup> electrolyte. The above results indicates that the thermal annealing treatment and g-C3N<sup>4</sup> as a novel structure modifier provide an useful and facile way to prepare the state polymer electrolytes used in the all solid-state lithium battery.

### DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

### AUTHOR CONTRIBUTIONS

LS and JY developed the concept and designed the experiment. They contributed equally to this work. XW and JY conducted the experiments. KX, GZ, and AM co-supervised the experiments. JY and LS wrote the manuscript. KX, CS, and WC helped to revise the work critically. All authors listed have made a substantial, crucial and direct contribution to the work, and approved it for publication.

### FUNDING

This work was supported by the project sponsored by the Key R&D Program of Shaanxi (2017ZDCXL-GY-08-03), the National Natural Science Foundation of China (61604120, 51804259 and 21501139), Scientific Research Program Funded by Shaanxi Provincial Education Department (17JK0380 and 18JK0390), Natural Science Foundation of the province of Shaanxi (No.2018JQ2037), Innovation and entrepreneurship training program for college students (1070214082) and the Presidential Research Fund of Xi'an Technological University (No. XAGDXJJ18007).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00388/full#supplementary-material

flexibility in composite polymer electrolytes for lithium batteries. Energy Storage Mater. 15, 46–52. doi: 10.1016/j.ensm.2018.03.016

Bae, J., Li, Y., Zhang, J., Zhou, X., Zhao, F., Shi, Y., et al. (2018a). A 3D nanostructured hydrogel framework-derived high-performance composite polymer lithium-ion electrolyte. Angew. Chem. Int. Ed. 57, 1–6. doi: 10.1002/anie.201710841/


electrolyte nanocomposite films, a novel investigation. ACS Appl. Mater. Interf. 6:429. doi: 10.1021/am4044932


**Conflict of Interest Statement:** LS was employed by the company Shaanxi Coal Chemical Industry Technology Research Institute Co. Ltd.

The remaining 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 Yang, Wang, Zhang, Ma, Chen, Shao, Shen and Xie. 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.

# Advances in Microwave-Assisted Production of Reduced Graphene Oxide

Xinxin Xie1,2, Yanping Zhou1,2 \* and Kama Huang1,2

*<sup>1</sup> College of Electronics and Information Engineering, Sichuan University, Chengdu, China, <sup>2</sup> Key Laboratory of Wireless Power Transmission of Ministry of Education, Sichuan University, Chengdu, China*

Efficient reduction of graphene oxide to obtain high-quality graphene nanosheets is desirable for energy storage, catalysis, electronics and environmental remediation. In this brief review, we mainly focus on the microwave-assisted production of reduced graphene oxide in three categories: (1) microwave-assisted chemical reduction of graphene oxide; (2) microwave-assisted thermal reduction of graphene oxide; (3) microwave-assisted simultaneous thermal exfoliation & thermal reduction of graphite oxide. We also summarize common techniques for characterizing reduction efficiency and quality of as-obtained rGO.

Keywords: microwave-assisted synthesis, reduced graphene oxide, chemical reduction, thermal reduction, exfoliation of graphite oxide

#### Edited by:

*Wenping Sun, University of Wollongong, Australia*

#### Reviewed by:

*Jiantie Xu, South China University of Technology, China Chao Wu, University of Wollongong, Australia*

> \*Correspondence: *Yanping Zhou ypzhou11@scu.edu.cn*

#### Specialty section:

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

Received: *21 February 2019* Accepted: *30 April 2019* Published: *04 June 2019*

#### Citation:

*Xie X, Zhou Y and Huang K (2019) Advances in Microwave-Assisted Production of Reduced Graphene Oxide. Front. Chem. 7:355. doi: 10.3389/fchem.2019.00355* INTRODUCTION

Graphene is a two-dimensional sheet of sp2-hybridized carbon. Owing to its theoretically high mechanical strength (1,060 GPa), thermal conductivity (∼5,000 W m−<sup>1</sup> K −1 ), electron mobility (2 × 105 cm<sup>2</sup> V −1 s −1 ), Young's modulus (∼1 TPa), surface area (2,630 m<sup>2</sup> g −1 ), electrical conductivity (∼2,000 S m−<sup>1</sup> ), and extraordinary optical properties, graphene has been attracting increasing application in various areas including energy storage [like fuel cells (Li et al., 2012; Hur and Park, 2013; Li and Wu, 2015), batteries (Li et al., 2015; Bak et al., 2016; Kaur et al., 2018), supercapacitors (Le et al., 2011; Akhavan, 2015)], sensors (Shao et al., 2010; Wang and Arash, 2014; Chatterjee et al., 2015), catalysis (Han et al., 2017; Hu et al., 2017), electronics (like liquid crystal displays (Novoselov et al., 2005; Lin et al., 2015; Basu et al., 2016; Narayan et al., 2016), touch panels (Das and Prusty, 2013; Liu et al., 2014; Katkov and Osipov, 2017), electromagnetic interference shielding (Eswaraiah et al., 2011; Thomassin et al., 2013; Song et al., 2014; Cao et al., 2015), environmental remediation (Bi et al., 2012; Qin and Brosseau, 2012; Chabot et al., 2014), and so on. Clearly, there will be an increasing demand for high quality graphene.

Currently, reduction of graphene oxide or simultaneous exfoliation and reduction of graphite oxide are regarded to be most promising for large scale production of chemically derived graphene, which can be defined as reduced graphene oxide (rGO). In general, there are mainly two strategies used for graphene oxide reduction: chemical reduction and thermal reduction, both of which are facing certain challenges. In chemical reduction methods, the reduction efficiency is unsatisfactory, requiring long treating time and yielding rGO with oxygen content usually higher than 15% (Stankovich et al., 2007; Shao et al., 2010; Mattevi et al., 2011; Park et al., 2011b; Xiao et al., 2013; Zhang et al., 2013; Wen et al., 2014). In thermal reduction process, long-time high-temperature treatment is not only energy consuming but also results in fragmentation of graphene and lots of structural defects (Jones et al., 2002; Chen and Yan, 2010; Chen et al., 2010; Gao et al., 2010; Lin et al., 2010; Zhu et al., 2010b; Zhang et al., 2011).

Microwave heats materials directly through dielectric loss rather than heat convection as in the conventional heating method, and hence can enable fast heating and selective heating. It is regarded to be promising in shortening the reaction time and yielding hot spots with extraordinary high temperature (Thostenson and Chou, 1999; Kappe, 2004; Tompsett et al., 2006; Schwenke et al., 2015). As such, material scientists have devoted much effort to explore the potential application of microwave in overcoming the current problems and facilitating the fabrication of high-quality rGO in large-scale. To this end, we present an up-to-date critical survey of literature relevant to MW-assisted production of rGO. This brief review is organized as follows. First, techniques for characterizing quality of asobtained rGO are discussed. Next, we provide a comprehensive summary of microwave-assisted chemical reduction of graphene oxide, microwave-assisted thermal reduction of graphene oxide, and microwave-assisted simultaneous thermal exfoliation and reduction of graphite oxide**,** respectively.

### TECHNIQUES FOR CHARACTERIZING QUALITY OF RGO

Generally, X-ray diffraction, X-ray photoelectron spectroscopy, Raman spectroscopy, electrical conductivity, thermal gravity analysis are common techniques used for characterizing reducing efficiency and quality of rGO.

### X-Ray Diffraction

The X-ray diffraction (XRD) pattern of graphene/graphite oxide usually shows a characteristic diffraction peak at 2θ = 10.9◦ with the corresponding d-spacing of 0.81 nm (Stobinski et al., 2014; Strankowski et al., 2016), due to insertion of hydroxyl and epoxy groups between the carbon sheets. After reduction, two types of XRD spectra may be obtained: (1) The XRD pattern with no peaks, suggesting the complete reduction and the formation of single-layered rGO (Hassan et al., 2009). (2) The XRD pattern showing a broad peak around 26.5◦ (Zedan et al., 2010; Liu et al., 2011; Park et al., 2011b; Wang et al., 2011; Pokharel et al., 2014). Since the oxygen-containing functional groups could lead to a reduction in crystallinity, the higher the broad peak's intensity is, the higher the crystallinity degree is, and the better graphene/graphite oxide is reduced.

### X-Ray Photoelectron Spectroscopy

X-ray photoelectron spectroscopy (XPS) could help to obtain qualitative and quantitative analysis results of the content of oxygenated groups. For example, the carbon/oxygen atomic ratio (C/O ratio) could be evaluated from the areas of C1s and O1s peaks and the atomic sensitivity factor. Usually, the C/O ratio of graphene oxide is ∼2.2–2.7 (Stankovich et al., 2007; Chen et al., 2010; Park et al., 2011a; Li et al., 2013; Wen et al., 2014; Han et al., 2015), attributed to the abundant oxygen-containing functional groups introduced during oxidation of graphite. Further, the C1s peak can be decomposed into three peaks: C=C (284.7 eV), C-O (286.9 eV), C=O (287.77 eV) (Zhao et al., 2014), through the areas of which existing statuses of oxygen could be analyzed in a quantitative manner.

#### Raman Spectroscopy

The D band around 1,350 cm−<sup>1</sup> in Raman spectra could reflect the disorder degree of the crystal structure of carbon while the G band around 1,580 cm−<sup>1</sup> represents a first-order scattering E2g vibration mode for characterizing the sp2 bond structure of carbon. ID/I<sup>G</sup> is the ratio of D band intensity to G band intensity, which could be used to evaluate the quality of the graphene structure. The higher ID/I<sup>G</sup> is, the more defects of C atom crystal there are. In the Raman spectrum of graphene, the D-band peak could not be observed. However, the ID/I<sup>G</sup> of graphene oxide could be as high as ∼0.8–1 (Chen et al., 2010; Lin et al., 2010; Liu et al., 2011; Park et al., 2011a; Zhao et al., 2014). Besides, the number of rGO layers can be estimated by observing the shape and position of the 2D peak in Raman spectra which is the second-order two-phonon process. For example, single-layer graphene exhibits a single, sharp 2D band located below 2,700 cm−<sup>1</sup> , while bilayer sheets have a broader 2D peak around 2,700 cm−<sup>1</sup> , and sheets more than five layers have a broad 2D peak above 2,700 cm−<sup>1</sup> (Hassan et al., 2009).

### Electrical Conductivity

Though the conductivity of graphene is as high as 1×10<sup>8</sup> S m−<sup>1</sup> , that of graphene oxide is about 0.02–0.07 S m−<sup>1</sup> (Li et al., 2010; Zhu et al., 2010a; Park et al., 2011a) due to the existence of oxygen-containing functional groups which disrupted sp2 bonding networks. During the reduction process, electrical conductivity can be recovered by restoring the p-network as oxygen-containing groups leave, leading to a conductivity improvement by 3–6 orders of magnitude (Chen et al., 2010; Dreyer et al., 2010; Zhao et al., 2014; Han et al., 2015). Therefore, the restoration degree of conductivity can indirectly reflect the reduction degree and quality of rGO.

### Thermal Gravity Analysis

Graphene/Graphite has a very good thermal stability even when being heated up to 900◦C (Bastiurea et al., 2015). For graphene oxide, it starts to have significant mass loss from ∼200◦C due to the decomposition of labile oxygen functional groups leads to poor thermal stability (Hassan et al., 2009; Chen et al., 2010; Zhu et al., 2010a; Gannavarapu et al., 2018). Therefore, the removal of oxygen-containing functional groups can help to restore its thermal stability. The significant mass loss temperature of rGO usually happens around 500–800◦C, depending on the reduction degree of rGO.

### MICROWAVE-ASSISTED PRODUCTION OF REDUCED GRAPHENE OXIDE

Methods for producing rGO could generally be categorized into chemical reduction of graphene oxide, thermal reduction of graphene oxide, and simultaneous thermal exfoliation and reduction of graphite oxide. Here, we would also discuss microwave-assisted production of rGO in these three categories separately.

### Microwave-Assisted Chemical Reduction of Graphene Oxide

Strong reducing reagents have been widely used to reduce graphene oxide. However, the reduction process is very slow. For example, hydrazine hydrate enabled reduction of graphene oxide in the oil bath for 12 h and 24 h resulted in a C/O ratio about 5 (Park et al., 2011b) and 10.3 (Stankovich et al., 2007), respectively. Microwave irradiation was found to be capable of accelerating the reducing rate significantly. For example, Hassan et al. (2009) reported microwave assisted hydrazine hydrate reduction of graphene oxide. After microwave treatment for 2 × 30 s (on for 10 s, off and stirring for 20 s), the ID/I<sup>G</sup> in Raman spectra was 0.1–0.12, indicating a high reduction degree even after such a short treating time This was confirmed by the improved thermal stability, showing no significant mass loss up to 750◦C. Elazab et al. (2015) obtained Pd/Fe3O<sup>4</sup> nanoparticles supported on graphene nanosheets by a one-pot microwave heating with the existence of hydrazine hydrate. After microwave irradiation for 2 min, the C/O of rGO was determined to be 8.1. Kumar et al. (2015) reported microwave-enhanced chemical reduction of graphene oxide using HI/CH3COOH as reducing reagent. They found the effect of 4 h microwave irradiation was comparable to 48 h conventional reaction process.

Since strong reducing regents are very much expensive and hazardous, organic solvents have also been explored to reduce graphene oxide. Zedan et al. (2010) reported MW-assisted reduction of graphene oxide in dimethyl sulfoxide (DMSO), a solvent with high microwave absorbing capability. The reduction degree of the sample prepared from 2 min MW irradiation at 1,200 W is even much higher than that of the sample derived from 7 h conventional heating treatment, as evidenced by XRD results of both samples shown in **Figures 1a,b**. After 2 min of MW irradiation in DMSO, the color of the product changed to black (**Figure 1c**), and the XRD pattern showed no characteristic peak (10.9◦ ) of GO, indicating that most of the GO has been converted into rGO. In contrast, under conventional heating at 180 ◦C, an apparent peak located at 10.9◦ was found even after 7 h treatment. It was not until 12th h that the GO was completely reduced, as can be seen from the disappearance of the peak at 10.9◦ in the XRD pattern and the color change of the sample to black (**Figure 1d**). Chen et al. (2010)reported the microwave-enhanced chemical reduction of graphene oxide to rGO in a mixed solution of N, N-dimethylacetamide (DMAc) and water without any additional reductant under nitrogen purge. The XRD pattern of the graphene oxide showed a wide peak, indicating the damage of the regular crystalline pattern of graphite during the oxidation. After microwave treatment for 2 min × 5 times (2.45 GHz, 800 W), no peak was observed indicating microwave-assisted formation of single-layered graphene. XPS results showed that C/O ratio changed from 2.09 to 5.46, confirming that the removal of oxygen containing groups happened. However, the ID/I<sup>G</sup> in Raman spectra changed from 0.95 to 0.96, suggesting that the reduction led to an increase of aromatic domains of smaller overall size in graphene. Similarly, Liu et al. (2011) also reported an increase of the product's ID/I<sup>G</sup> in Raman spectra from 0.80 to 0.94 after being treated by microwave at 750 W for 2 min in N, N-dimethylacetamide (DMF) solution. Zhao et al. (2014) applied DMF to reduce graphite oxide under microwave irradiation at 100 W for 30 min. The XPS results showed that C/O ratio rose from 2.13 to 6.13 while Raman results showed that ID/I<sup>G</sup> remained almost unchanged.

The characteristics of the as-obtained rGO products derived from different microwave-assisted chemical reduction methods are summarized and listed in **Table 1**. Since microwave heat rapidly, microwave-assisted rate enhancement in chemical reduction of graphene oxide has been widely observed, with most of the reaction time <10 min. However, as in the conventional heating method, the as-obtained rGO is still unsatisfactory, facing the same problem of containing a high oxygen content and a high ID/I<sup>G</sup> value when organic solvent was used as the reducing reagent. Microwave-enhanced quality of chemically reduced rGO is yet to be explored. Also, since the type and amount of reducing agents were different in different work, it is difficult to draw a conclusion regarding the effect of microwave power and treating time here.

### Microwave-Assisted Thermal Reduction of Graphene Oxide

In conventional thermal reduction process, long-time hightemperature treatment in a protecting atmosphere (e.g., N2/Ar) is usually required, which is not only very energy consuming but also leads to the formation of defects in the graphene basal plane because of the evolution of the oxygen functional groups during reduction, like nanoscopic holes brought by carbon loss as CO or CO<sup>2</sup> and Stone-Wales types of defects lead by rearrangement of carbon atoms in the graphene basal plane. Besides, the reducing efficiency is always suppressed by the highly stable ether and carbonyl groups formed between the oxygen functional groups.

In 2010, Li et al. (2010) for the first time reported a microwaveassisted thermal reduction of graphene oxide. Specifically, they made the graphene oxide into a free-standing film via filtration membrane, and then treated it with a microwave irradiation with a frequency of 6.425 ± 1.150 GHz at a power of 500 W without purging any inert gas. As shown in **Figures 2a,b**, after 1 s irradiation, the temperature rose only a little and no distinct difference were found in the electrical properties and the Raman spectra of the sample. However, after 2 s irradiation, the surface temperature of the graphene oxide film ramped up to 400◦C and the conductivity of the graphene oxide film increased from 0.07 to 1 × 10<sup>4</sup> S m−<sup>1</sup> ; the ID/I<sup>G</sup> in Raman spectra decreased from 1 to 0.3; the mass loss temperature increased from 181 to 668◦C (**Figure 2c**). In addition, they found that in the MW treatment process, fierce degassing and a high gas pressure inside the graphene oxide was generated, which was beneficial for exfoliating the graphene oxide into thinner graphene sheets. Afterwards, they also applied a household MW oven (with a frequency of 2.45 GHz and a power of 1,000 W) to treat the free-standing graphene oxide film. They found that there was an optimal irradiation time of 8 min as evidenced by Raman spectra (**Figure 2d**) and TGA results (**Figure 2e**). Longer treatment time (e.g., 10 min) caused lower mass loss temperature (from 700◦C to around 800◦C), which

could possibly be attributed to the oxidation by air when there was no protecting gas.

Considering the weak microwave absorption of graphene at low frequency, while 2.45 GHz and 915 MHz are most commonly used in industry, Voiry et al. (2016) introduced a conventional mild annealing process before the microwave irradiation process. Firstly, the single-layered graphene oxide solution was slowly injected into an aqueous solution that contained 1 wt% CaCl2, which helps to coagulate GO nanosheets and promote solidification via bridging with the oxygen-containing groups from two GO nanosheets, yielding uniform and continuous gelatinous GO. The coagulated graphene oxide was then washed with DI water and dried, which was subsequently annealed at 300◦C for 1 h under Argon to yield mildly reduced graphene oxide that was capable of absorbing microwaves at 2.45 GHz effectively. After 1–2 s long microwave irradiation in microwave oven at 1,000 W in argon, the efficient absorption of microwave led to rapid heating of the mildly reduced GO and consequently brought large arcing as shown in **Figure 3a**, causing desorption of oxygen functional groups and reordering of the graphene basal plane. As such, the as-obtained rGO exhibited pristine CVD graphene-like features in the Raman spectrum showing sharp G an 2D peaks and a nearly absence of D peak. In contrast, the rGO derived from conventional thermal reduction remains highly disordered, as indicated by the presence of an intense and broad disorder D band and the absence of the 2D band in the Raman spectra (**Figure 3b**). The XPS results were well in agreement with the Raman results. Owning to the large arcing which may possibly resulted in an extremely high temperature, the oxygen content of the MWrGO was about 4%, lower than that theoretically predicted for rGO after annealing at 1,500 K (15–25%). Aberration corrected HR-TEM (**Figure 3c**) also suggested a highly ordered structure, which suggests that there is some reorganization of the carbon bonding during microwave reduction, along with removal of oxygen facilitated by achieving exceptionally high temperatures.

Jiang et al. (2018) further modified Voiry's strategy and developed a triggered microwave-assisted reduction of graphene oxide. In their work, 1 wt% CaCl<sup>2</sup> aqueous solution was used to treat one side of the filter paper while graphene oxide was dropped onto the other side of the filter paper. Subsequently, TABLE 1 | Characteristics of rGO products derived from different microwave-assisted chemical reduction.


graphene oxide was treated with CaCl<sup>2</sup> solution and then washed with DI water, dried in air and pulled out from the filter paper. Finally, a small piece of rGO "paper" obtained by thermal annealing was placed on the as-obtained large graphene oxide paper to act as a trigger, as shown in **Figure 4a**. After 800 W microwave irradiation for 2 s in air, arc discharge process started. The irradiation was maintained for extra 3– 5 s and high-quality microwave reduced graphene oxide paper was obtained. In XPS spectrum, the characteristic peak for C-O vanished, suggesting the significant removal efficiency of oxygencontaining group. Raman spectra of MW-rGO showed sharp G and 2D bands, and low-intensity D band. In contrast, the Raman spectra of rGO derived from conventional thermal reduction at 800◦C showed a strong D band and none 2D band, indicating microwave heating is very beneficial for recovering the highly ordered graphene-like structures which was also confirmed by the conductivity. The sheet resistance of the MW-rGO was measured to be about 40 cm−<sup>2</sup> , far less than that (about 796 cm−<sup>2</sup> ) of rGO prepared by 800◦C annealing. They further examined the Raman spectra of graphene-triggered microwave reduction at different positions, and found the part of graphene oxide located nearest to the trigger was not reduced the best, as shown in **Figure 4b.** Hence, they proposed that the microwave reduction can be not only governed by a simple thermal

reduction process but the result of the thermal effect and arc discharge (Jiang et al., 2018).

Instead of putting a small piece of graphene as the trigger, Wan et al. (2018) put a fully bended copper wire inside graphene oxide to absorb 2.45 GHz microwave irradiation intensively and hence trigger immediate arcs and fire to reduce the graphene oxide, as shown in **Figures 5a,b**. Microwave treatment is applied for 2 s one time. The decreasing D band in Raman spectra under different microwave treating times indicate that high crystalline graphene was obtained after microwave treatment.

Low-temperature reduction of graphene oxide is of great importance when graphene oxide film was deposited on glass or plastic substrates. Han et al. (2015) demonstrated microwaveassisted low-temperature thermal reduction of few-layered graphene oxide film. Specifically, they put the graphene oxide film at the center of the highest magnetic field in a single mode microwave reactor which was operated at 2.45 GHz with a power of 42 W to ensure that the temperature was below 250◦C during the whole process. After 5 min treatment in the air, ID/I<sup>G</sup> in the Raman spectrum of the product decreased from 2.89 to 1.56 and C/O derived from XPS results increased from 7.8 to 17.5. In addition, microwave-enabled exfoliation leading to thinner pieces was also observed in their work.

The characteristics of the as-obtained rGO products derived from different microwave-assisted thermal reduction of graphene oxide methods are summarized and listed in **Table 2**. Arc discharge caused by microwave irradiation is very beneficial for obtaining high-quality rGO, as evidence by the short reaction time, the low ID/I<sup>G</sup> values, the high C/O values and the conductivity values of most as-derived rGO. It is worth to note that rGO containing only 4% oxygen and exhibiting pristine CVD graphene-like features in the Raman spectrum was successfully fabricated via this reducing strategy. Besides, owning to the ultra-rapid treating process, protecting atmosphere is not always necessary. However, too long treatment time would result in re-oxidation and consequently rGO with worse quality when

the reduction is done in the air. Since the amounts of GO samples and the microwave heating chambers in different work were different, which would affect the power intensity imposed on the GO, it is also difficult to draw a conclusion regarding the effect of microwave power here.

permission from Wan et al. (2018), copyright 2018 John Wiley and Sons].

### Microwave-Assisted Simultaneous Exfoliation and Reduction of Graphite Oxide

Beside of direct reduction of graphene oxide to yield rGO, thermal shock like flash-assisted simultaneous exfoliation and reduction of graphite oxide at temperatures up to 1,050◦C has also been used. Theoretically, microwave heating which is well-known for its rapid heating feature is very promising in enabling simultaneous exfoliation and reduction of graphite oxide. In 2010, Zhu et al. (2010a) treated graphite oxide powder in microwave oven at 700 W for 1 min in ambient atmosphere. Large volume expansion as shown in **Figures 6a**,**b** accompanied with violent fuming followed by sparking was observed. They found that a minimum power of 280 W was required to expand the graphite oxide powder successfully, which was probably due to the insulating feature that resulted in inferior microwave absorption ability. After microwave treatment, the as-obtained rGO had a worm-like morphology, in which many regions are wrinkled and folded (**Figures 6c,d**). XPS results showed that C/O increased from 0.79 to 2.75 (**Figure 6e**). XRD spectrum showed a broad peak centered at 2θ = 25◦ . The conductivity was measured to be about 274 S m−<sup>1</sup> .

Under microwave irradiation, graphite oxide itself might not be heated to a high enough temperature to cause its exfoliation and reduction efficiently due to its inferior microwave absorption ability. Scientists have tried several ways to come over this problem.

Park et al. (2011a) mixed graphite oxide powder with 10% graphene nanosheets to realized efficient exfoliation and reduction under microwave irradiation. The temperature rising rate was estimated to be 2,000◦C/min, which was high enough to build up pressure to overcome the van der Waals forces between the graphene sheets in graphite oxide for exfoliation. Huge volume expansion was observed after 10 s, followed by arc after 40 s. SEM images (**Figure 7**) showed that the prepared rGO also had a worm-like structure composed of

ultra-thin sheets, which is different from randomly agglomerated and wrinkled sheets prepared from solution-based reduction syntheses. HR-TEM images suggested the presence of single or a few layers of graphene nanosheets as shown in **Figure 7**. They further compared the effect of reduction atmosphere on the product quality. The rGO prepared in air, Ar, H2+Ar (1:9) showed a C/O ratio of 3.6, 11.45, 18.5, respectively and ID/I<sup>G</sup> values of 0.853, 0.842, and 0.785, respectively in Raman spectra, respectively. The rGO prepared in Ar, H2+Ar showed a conductivity of 7.41 × 10<sup>2</sup> and 1.25 × 10<sup>3</sup> S m−<sup>1</sup> , respectively. Hydrogen containing atmosphere is the most effective because H2O was formed instead of CO2, which prevents the formation of vacancies and defects due to the loss of carbon atoms.

Wen et al. (2014) applied a sequential chemical reduction and microwave exfoliation and reduction of graphite oxide to


TABLE 2 | Characteristics of rGO products obtained by different microwave-assisted thermal reduction.

TABLE 3 | Characteristics of rGO products derived from different microwave-assisted simultaneous thermal exfoliation and reduction.


produce rGO. Specifically, graphite oxide powder was treated by NaBH<sup>4</sup> at 95◦C for 2 h, and subsequently made into a film, which was microwave irradiated at 5.8 GHz with a power of 1,400 W for 10 min in N2. After NaBH<sup>4</sup> treatment, a C/O value of 3.84, an ID/I<sup>G</sup> value of 1.05 in Raman, and a conductivity of 1.41 × 10<sup>2</sup> S m−<sup>1</sup> were obtained. During the microwave treatment, the sample expanded to a loose floc and transmitted bright orange light, leading to ultrathin sheets with a C/O ratio of 10.33, an ID/I<sup>G</sup> ratio of 1.07 in the Raman spectrum and a conductivity of 1.19 × 10<sup>4</sup> S m−<sup>1</sup> . Sole microwave irradiation at 5.8 GHz with a power of 1,400 W for 10 min without a preceding chemical reduction process gave yield to a C/O of 7.24 and an ID/I<sup>G</sup> of 0.89 in Raman spectrum. This demonstrated that solid-state microwave irradiation not only removed oxygen-containing groups, but also repaired defects on the graphene sheets while NaBH<sup>4</sup> though effective for removing oxygen-containing groups may induce more defects. Further, through XPS analysis of content of C-O and C=O groups, they found that microwave irradiation removed C=O more readily than C-O while the chemical reduction method removed C-O more efficiently.

Pokharel et al. (2014) developed a multi-step reduction technique to obtain rGO from graphite oxide. In the first step, graphite oxide with a C/O of 0.83 was exfoliated within 1 min under microwave irradiation at 2.45 GHz with a power of 1,000 W in N<sup>2</sup> without becoming red. The expanded graphite

microwave irradiation synthesis. (f) HR-TEM images of GNS. [Reprinted with permission from Park et al. (2011a), copyright 2011 The Royal Society of Chemistry].

oxide had a C/O of 2.42. However, it was so loose that microwave could not concentrate on a definite area for further reduction of graphite oxide. Hence, in the second step, the expanded graphite oxide was collected and then exposed to additional microwave irradiation for 1 and 2 min, resulting in rGO showing a C/O ratio of 5.78 and 10.39, respectively.

The characteristics of the as-obtained rGO products derived from different microwave-assisted simultaneous thermal exfoliation and reduction of graphite oxide methods are summarized and listed in **Table 3**. As can be seen from the C/O values, ID/I<sup>G</sup> values, mass loss temperatures and conductivities, simultaneous exfoliation and reduction was not able to yield rGO with as high quality as those obtained from thermal reduction of graphene oxides. However, when combined with reducing atmosphere (i.e., H2) or pretreatment with strong reducing reagents (NaBH4), the reduction degree could be greatly enhanced.

### PERSPECTIVES

It is expected that the explored methods for microwave-assisted reduction of graphene/graphite oxide into high-quality rGO could be extended to fabrication of graphene composites. The microwave-induced arc facilitated short reaction time at high temperature is not only beneficial for getting high-quality rGO, but also promising in facilitating the formation of highquality nanomaterials (such as metal nanocrystals, metal oxides nanocrystals, and so on) with advanced performance. In addition, the methods could be extended to fabrication of other conductive materials that could trigger sparks under microwave irradiation, like graphene does.

### AUTHOR CONTRIBUTIONS

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

#### FUNDING

This work was financially supported by the National Natural Science Foundation of China (Grant No.

#### REFERENCES


61801314), National Natural Science Foundation of China (Grant No. 61731013), Fundamental Research Funds for the Central Universities of China (Grant No. YJ201703).

supporting metal nanocrystals in aqueous and organic media. J. Mater. Chem. 19, 3832–3837. doi: 10.1039/B906253J


**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 Xie, Zhou and Huang. 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.

# Functional Ionic Liquid Modified Core-Shell Structured Fibrous Gel Polymer Electrolyte for Safe and Efficient Fast Charging Lithium-Ion Batteries

Xiaoxia Liu, Yufei Ren, Lan Zhang\* and Suojiang Zhang\*

*Beijing Key Laboratory of Ionic Liquids Clean Process, CAS Key Laboratory of Green Process and Engineering, Institute of Process Engineering, Chinese Academy of Sciences, Beijing, China*

#### Edited by:

*Oleksandr Malyi, University of Oslo, Norway*

#### Reviewed by:

*Xianhong Rui, Guangdong University of Technology, China Weimin Kang, Tianjin Polytechnic University, China Huarong Xia, Nanyang Technological University, Singapore*

#### \*Correspondence:

*Lan Zhang zhangl@ipe.ac.cn Suojiang Zhang sjzhang@ipe.ac.cn*

#### Specialty section:

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

Received: *16 March 2019* Accepted: *21 May 2019* Published: *12 June 2019*

#### Citation:

*Liu X, Ren Y, Zhang L and Zhang S (2019) Functional Ionic Liquid Modified Core-Shell Structured Fibrous Gel Polymer Electrolyte for Safe and Efficient Fast Charging Lithium-Ion Batteries. Front. Chem. 7:421. doi: 10.3389/fchem.2019.00421* Fast charging is of enormous concerns in the development of power batteries, while the low conductivity and lithium ion transference number in current electrolytes degraded the charge balance, limited the rate performance, and even cause safety issues for dendrite growth. Combine inorganic fillers and ionic liquid plasticizer, here in this paper we prepared a core-shell structured nanofibrous membrane, by incorporating with carbonate based electrolyte, a gel polymer electrolyte (GPE) with high conductivity, outstanding Li<sup>+</sup> transference number was obtained. Notably, the Li/electrolyte/LiNi0.6Co0.2Mn0.2O<sup>2</sup> (NCM622) half-cell with this composite electrolyte delivers a reversible capacity of 65 mAh/g at 20C, which is 13 times higher than that of with Celgard 2325 membrane. It also shows enhanced long-term cycle stability at both 3C and 5C for the suppression of lithium dendrite. This organic-inorganic co-modified GPE guarantees the fast charging ability and safety of LIBs, thus provides a promising method in high performance electrolyte design.

Keywords: lithium ion battery, fast charging, core-shell structure, nanofibrous separator, safety

### INTRODUCTION

With the increasing demand for energy consumption and the continuous improvement for environmental protection, the development of sustainable and renewable energy is urgently needed currently. Lithium ion batteries (LIBs) have been widely used in the power battery, energy storage applications, portable electronic equipment and other fields, owing to its inherent advantages such as high energy density, high power density, long cycle life, no memory effect, and environmental friendliness (Gao et al., 2009; Goodenough and Kim, 2010; Etacheri et al., 2011; Scrosati et al., 2011; Lu et al., 2014a; Varzi et al., 2016; Tan et al., 2018). Nevertheless, high-capacity and high-power LIBs are facing more and more challenges such as high power output, rapid charge/discharge capability and safety under dynamic conditions (Zhu et al., 2013; Xie et al., 2014; Zhang J. et al., 2014; Li et al., 2018, 2019), and it is of particularly importance to investigate methods to guarantee the safety of LIBs with enhanced fast charge capability (Rui et al., 2016; Tan et al., 2018).

LIBs generally compose of anode, cathode, electrolyte, and packing material. The electrolyte, which including polymer separator and liquid electrolyte in commercial LIBs, mainly plays two key roles: blocking the direct contact between the electrodes to avoid internal electrical short circuit and acting as the Li<sup>+</sup> transport passage (Tarascon and Armand, 2001; Xiao et al., 2009; Huang, 2010; Lee et al., 2014). Therefore, the electrolyte not only directly affects the LIB electrochemical performance, such as cycle stability and rate capability, but also, it has significant influence on the safety of LIBs. Most commercial LIBs adopt polyolefin [such as polyethylene (PE) and polypropylene (PP)] membranes (Huang, 2010; Kim et al., 2015; Shen et al., 2018) and LiPF6/carbonates electrolyte, based on which satisfying cycle performances have achieved, while the rate capability and safety, especially during quick charge, is yet to be enhanced.

The major issue in fast charging LIBs is the poor kinetics caused by the sluggish ion transport in the cell, which will probably lead to large polarization and low reversible capacity, furthermore, when the anode potential is lower than that of Li plating, dendrite will be generated and may short the cell thus cause safety concerns.

Electrolyte holds significant influence on both rate capability and safety, on one hand, ionic conductivity and Li<sup>+</sup> transference number is the major factor to control the battery kinetics (Diederichsen et al., 2017), on the other hand, the separator, which blocks the direct contact between electrodes, should also work as a physical barrier to suppress the dendrite from shorting the battery (Jana et al., 2015; Shin et al., 2015; Zheng et al., 2018). Therefore, scientist made lots of efforts in electrolyte modification, including surface coating of the separator (Ghazi et al., 2017; Liu et al., 2017a,b; Zuo et al., 2017), doping inorganic nanoparticles in polymer to obtain composite electrolyte (Xiao et al., 2012; Chen et al., 2017; Kim et al., 2017; Wang et al., 2017; Shen et al., 2018), and electrostatic spinning to get nanofibrous membrane (Choi et al., 2003; Wu et al., 2015; Ma et al., 2017; Cheng et al., 2018). Thereinto, the nanofibrous membrane has large specific surface area, three-dimensional (3D) porous structure and high electrolyte uptake ability, on which the rate capability can be enhanced (Xiao et al., 2015; Liang et al., 2016; Park et al., 2017). What's more, functional groups can be easily introduced in the electrospining process, thereby endows the electrolyte with other properties, such as dendrite suppression (Lu et al., 2015a,b; Cheng et al., 2018; Deng et al., 2018), flame retardant (Lu et al., 2017; Jia et al., 2018; Sun et al., 2018), et al. Poly(vinylidene fluoride-co-hexafluoropropene) (PVDF-HFP) is considered as a potential polymer matrix for electrolyte not only as it has a low crystallinity which could promotes rapid ion conduction (Ali et al., 2018; Zhao et al., 2019), but also, it has high dielectric constant and low surface energy that may promote the compact deposition of metal Li (Lopez et al., 2018). It is reported the ionic conductivity of PVDF-HFP-based polymer electrolytes is up to 3.9 mS/cm (Xiao et al., 2012).

Ionic liquid modified gel polymer electrolytes (IL-GPE) have attracted much attention due to their good thermal stability and mechanical properties. Singh et al. studied imidazolyl ionic liquids and found that the EMIMFSI-based GPEs have excellent electrochemical stability, good compatibility and thermal stability(Singh et al., 2018). Guo et al. prepared the PVDF-HFP-LiTFSI/SiO2/EMITFSI GPE with high thermal stability and good electrode compatibility (Guo et al., 2018). However, IL-GPE also has some shortcomings, such as lower ionic conductivity and lithium ion mobility (Zhou et al., 2015). In this work, a piperidine ionic liquid, 1-methyl-1-propylpiperidinium chloride (PPCl), and Li2SiO<sup>3</sup> (LSO) nanoparticles were introduced into the PVDF-HFP matrix via coaxial electrospinning technology. The as prepared membrane has high porosity and electrolyte uptake, remarkable ionic conductivity, and outstanding electrochemical performance especially in quick charging. Commercial NCM622 cathode adopting this IL-GPE delivers a high reversible capacity of 65 mAh/g in 20C rate charge/discharging, which is 13 times higher than that of the cell adopting Celgard 2325 membrane.

### EXPERIMENTAL SECTION

### Materials

Poly (vinylidene fluoride-co-hexafluoropropene) (PVDF-HFP, Mw. ∼455,000), 1-methylpiperidine (97%), and (3-chloropropyl) trimethoxysilane (98%) were provided by Sigma-Aldrich. Li2SiO<sup>3</sup> (LSO, 99%, Strem Chemicals), N, N-dimethylformamide (DMF, 99.5%, Beijing Chemical Works), LiNi0.6Co0.2Mn0.2O<sup>2</sup> (NCM622, Beijing Dangsheng Material Technology Co., Ltd.), Super P (Imerys Graphite & Carbon), Polyvinylidene difluoride (PVDF, Solvay 5130), N-Methyl-2-pyrrolidone (NMP, 99.0%, Sinopharm Chemical Reagent Co., Ltd.), lithium tablet (Li, Tianjin Zhongneng Co., Ltd.), and polypropylene (PP, Japan Ube) were commercially available and used without further purification.

### Preparation of the GPEs

The ionic liquid PPCl was synthesized according to the procedure reported before (Lu et al., 2012; Korf et al., 2014; Cheng et al., 2018), its structure and purity was also testified in our previous work (Xu et al., 2018). The core-shell structured 3D porous nanofiber membrane was prepared by the coaxial electrospinning technique on an ET-2535H machine (Ucalery Tech Inc., China), as shown in **Scheme 1**. To make composite nonwoven membrane, 80% (wt., similarly hereinafter) DMF was adopted in all the spinning solutions. The slurry for core spinning was prepared by mixing PPCl and PVDF-HFP in DMF solvent, wherein the weight ratio of PPCl: PVDF-HFP: DMF was fixed at 1:19:80. Correspondingly, the slurry for shell spinning was prepared by mixing LSO and PVDF-HFP in DMF solvent with a ratio of 2:18:80. The coaxial electrospinning equipment mainly contained a variable positive voltage of 15 kV and a negative voltage of −2kV, two syringe pumps, a spinneret consisting of two chambers and a collector. The pumped speed of the shell and core solutions supply were fixed at 0.25 and 0.15 mL/h, respectively, the distance between the needle tip and aluminum foil collector was 13 cm. The as prepared membrane has a thickness of 50 ± 5µm, and was entitled by PHP@PHL. Accordingly, PVDF-HFP, PVDF-HFP-PPCl (PHP), and PVDF-HFP-LSO(PHL) nanofiber membrane was prepared by mixing PVDF-HFP in DMF (20:80), PPCl, and PVDF-HFP in DMF (1:19:80) or LSO and PVDF-HFP in DMF (2:18:80), respectively. The as-prepared nonwoven fiber membranes were cut into disc with a diameter of 16 mm, which were then dried in a vacuum oven at 60◦C for 20 h to remove the residual solvent.

In the end, the membranes were filled and swollen in a liquid electrolyte, 1.2 M LiPF<sup>6</sup> in ethylene carbonate (EC) and ethyl methyl carbonate (EMC) (3:7, weight ratio), for 30 min in an argon filled glove box to obtain the relevant GPEs.

### Characterization of the Membranes and GPEs

The thickness of various films was recorded by measuring membrane apparatus (CH1ST, Shanghai Milite Precise Instrument Co., Ltd., China), and morphology of the membrane was studied by field-emission scanning electron microscopy (FE-SEM, JSM-7001F, JEOL, Japan). The field emission transmission electron microscopy (TEM, JEOL, JEM-2100) was used to test the core-shell structure of PHP@PHL nanoporous fiber membrane. The surface chemical composition of PHP@PHL was analyzed by X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi, Thermo Fisher Scientifec, America). Differential scanning calorimetry (DSC, Mettler-Toledo, Switzerland) was carried out to analyze the thermal behavior of all kinds of membranes. Samples were put into aluminum pans and the test temperature was set from 50 to 250◦C with a heating rate of 5◦C/min, under N<sup>2</sup> atmosphere. The porosity of various films was measured by soaking n-butanol for 2 h, then calculated using Equation (1): P = (mb/ρb)/(mb/ρ<sup>b</sup> + ma/ρa) × 100%, where m<sup>a</sup> and m<sup>b</sup> are the weights of separators and n-butanol, ρ<sup>a</sup> and ρ<sup>b</sup> are the density of separators and n-butanol, respectively (Xiao et al., 2012; Zhou et al., 2013). In an argon filled glove box, the electrolyte uptakes were analyzed by the mass difference of separators before and after soaking in electrolyte for 30 min and then calculated using Equation (2): EU = (W–W0)/W<sup>0</sup> × 100%, in which W<sup>0</sup> and W are the weights of the films before and after immersing in the liquid electrolyte, respectively. Ten samples are tested to measure the electrolyte uptake and porosity of each membrane. Wettability of the separators were researched by contact angle measurements (DSA 100S, Germany KRUSS).

The various GPEs' effective ionic conductivities (σ) were calculated using Equation (3): σ = d/(R<sup>b</sup> × S), where d is the thickness of the film, S is the area and R<sup>b</sup> is the bulk impedance acquired by electrochemical impedance spectroscopy (EIS) in Stainless Steel (SS)/electrolyte/SS symmetric cells. The bulk resistances (Rb) were investigated by a CHI660E electrochemical workstation with a frequency range of 0.01– 105Hz and an amplitude of 5 mV at room temperature. The lithium ion transference numbers (tLi <sup>+</sup>) of different electrolytes were tested using chronoamperometry (CA) and EIS (both by CHI660E), and then calculated using Equation (4): tLi <sup>+</sup> = IS(1v–I0R0)/I0(1v–ISRS) (Li et al., 2018), where I<sup>0</sup> and I<sup>S</sup> are the initial and steady state currents obtained by CA testing of lithium symmetrical cell, R<sup>0</sup> and R<sup>S</sup> are the interfacial impedance before and after polarization, and 1V (10 mV) is the applied voltage difference (Yang et al., 2014; Zhang F. et al., 2014). The electrochemical performances were researched in coin cells with lithium foil anode and LiNi0.6Co0.2Mn0.2O<sup>2</sup> (NCM622) cathode. The interfacial stability between the electrode and electrolytes was investigated using EIS after standing for 1, 10, 20, and 30 days, respectively. The long term cycling stability, as well as rate performances were tested in a voltage range between 2.8 and 4.4 V with different C rates by LAND battery cycle system (Wuhan Blue Electric Co., LTD, China).

### RESULTS AND DISCUSSION

#### Morphologies of the Membranes

The morphologies of the Celgard 2325 (which is a PP/PE/PP trilayer separator), nanofibrous PVDF-HFP, PHL, and PHP@PHL membranes, as well as microscopic changes after the thermal stability analyze in 130–170◦C temperature range, were tested by using SEM, as shown in **Figures 1A,D**. The Celgard 2325 membrane has tensile holes in the lamellar matrix, but the electrostatic spinning films show 3D porous structure with 80–160 nm diameter nanofiber which interlacing with each other to form interconnected networks. The energy dispersive X-ray spectroscopy (EDS) data shown in **Figure S1** gives the elemental distribution in PHL and PHP@PHL complex films. As shown in **Figure S1a**, Si and O are uniformly distributed in the membrane, which proves that Li2SiO<sup>3</sup> is evenly dispersed in the polymer matrix. EDS results of PHP@PHL in **Figure S1b** shows little signal of N or Cl as the content of PPCl is pretty low (<2%). TEM image of PHP@PHL nanofiber further proved its core-shell structure, and the inorganic nanoparticle Li2SiO<sup>3</sup> is dispersed in the out layer of the PVDF-HFP matrix, with the thickness of several to tens of nanometers as shown in **Figure 1B**. The XPS spectra in **Figure S2** detected no Cl element on the surface of PHP@PHL nanofiber, which further proves that the PPCl ionic liquid was fully encapsulated by the out PHL layer.

The thermal stability of various membranes was investigated by storing them in an air circulation oven at a series of temperatures between 130 and 170◦C, each for 20 min. **Figure 1C**

FIGURE 2 | Electrolyte uptake and porosity (A), as well as the contact angle images with liquid electrolyte droplet (B) of the Celgard 2325, PVDF-HFP, PHL, PHP, and PHP@PHL separators.

shows that Celgard 2325 begins to curl for the melting of PE and molecular tanglement of polyolefin at about 135◦C, it turns transparent at 160◦C for the melt of PP, meanwhile PVDF-HFP shrink at 150◦C and melt at 170◦C. However, the PHL and PHP@PHL keeps their porous structure even at 170◦C. It can be concluded that the inorganic nano-particle, Li2SiO3, in the shell structure significantly improved the thermal stability of the separator, which may potentially contribute in enhancing the safety of the battery. In order to further clarify the morphology changes of the films after heat treatment, FE-SEM pictures of the four separators are given in **Figure 1D** for microscopic description. Celgard 2325 was fully shutdown at 160◦C, and PVDF-HFP nanofibrous films melt under 170◦C, but the PHL and PHP@PHL films still keeps certain amount of pores remain in 170◦C (**Figure 1D**). From the microscopic point of view, the core-shell structure nanofibrous membrane expresses excellent thermal stability, hence makes it have distinguished heat resistance and advance the safety of lithium battery under high temperature condition. According to DSC test results of four separators, membranes based on PVDF-HFP shows better thermal stability than that of Celgard 2325, and the core-shell structured PHP@PHL, released the least heat even on melting (**Figure 1E**) (Kang et al., 2016; Wang et al., 2019).

Porosity of the membranes are tested and calculated by Equation (1) as shown in **Figure 2A**, it's about 54.4, 66.1, 68.5, 72.5, and 74.0% for Celgard 2325, PVDF-HFP, PHL, PHP, and PHP@PHL, respectively. Obviously, fibrous separators have higher porosity than Celgard 2325, which may contribute in electrolyte uptake and thus enhance the rate capability. Contact angle of liquid electrolyte to various membranes were also studied and shown in **Figure 2B**. The contact angle of Celgard 2325, PVDF-HFP, PHL, PHP and PHP@PHL is 41.30, 28.48, 21.52 13.51, and 15.61◦ , respectively. All fibrous membranes showed smaller contact angle with electrolyte for the high dielectric constant of PVDF-HFP (Lopez et al., 2018), besides, they also beneficial from the fibrous structure. Smaller contact angle indicated that the PHP@PHL composite nanofibrous membrane has better affinity and is easier to be wetted by the liquid electrolyte, thus better rate capability might be obtained.

The effect ionic conductivity of the five separators was probed by EIS test at ambient temperature and calculated by Equation (3) as shown in **Figure 3A**. R<sup>b</sup> obtained from EIS data is 2.1105, 3.8473, 2.5165, 3.3936, and 0.6801 for Celgard 2325, PVDF-HFP, PHL, PHP, and PHP@PHL, respectively, and the corresponding effect ionic conductivity are 0.63, 0.64, 1.03, 1.02, and 4.05 mS/cm, respectively. Consistent with the results of porosity and wettability, this data further guaranteed the enhanced rate capability in cells.

The lithium ion transference number (tLi <sup>+</sup>) of Celgard 2325 and the as-prepared membranes with liquid electrolyte was tested and calculated via Equation (4) as shown in **Figures 3B–F**, where the corresponding tLi <sup>+</sup> are 0.42, 0.39,

TABLE 1 | The film thickness, electrolyte uptake, porosity, effective ionic conductivity, and lithium ion transference number of Celgard 2325, PVDF-HFP, PHL, PHP, and PHP@PHL separators.


FIGURE 3 | (A) Nyquist plot of AC impedance measurements (SS/separator/SS) of the Celgard 2325, PVDF-HFP, PHL, and PHP@PHL separators; Lithium ion transference numbers of (B) the Celgard 2325, (C) PVDF-HFP, (D) PHL, (E) PHP, and (F) PHP@PHL membranes as demonstrated by CA polarization curves and EIS plots before and after polarization.

0.56, 0.45, and 0.62, respectively. tLi <sup>+</sup> reduced from 0.42 to 0.39 when Celgard 2325 was replaced by PVDF-HFP, which proves that the polymer itself in fact has little contribute in Li<sup>+</sup> transference. The addition of Li2SiO<sup>3</sup> enhanced the number to 0.56, which demonstrated that the inorganic nanoparticle may facilitate the desolvation of Li+. While tLi <sup>+</sup> of PHP is only 0.45, slightly higher than that of Clegard 2325 and PVDF-HFP, the reason lays on that although PPCl was mixed with PVDF-HPF and formed a membrane, but the Cl anion could still partly dissociated into the electrolyte thus influence the transportation of Li<sup>+</sup> (Xu et al., 2018). PHP@PHL shows the highest tLi <sup>+</sup> among the five electrolytes for the synergistic effect of LSO and PPCl, encapsulated by the PHL layer, the Cl anion could not influence the Li<sup>+</sup> transport but adjust the polarization of the separator polymer matrix. Therefore, PHP@PHL nanofiber membrane prominently improves the tLi +, which may contribute in reduces anion aggregation near the electrode surface, thus reduces the concentration polarization and stabilizes the electrochemical deposition of lithium ions, improving the safety of LIB.

The film thickness, electrolyte uptake, porosity, ionic conductivity, and lithium ion transference number of the Celgard 2325, PVDF-HFP, PHL, and PHP@PHL separators summarized in **Table 1**. It is worth noting that the electrolyte uptake of PHP@PHL was about 597%, which is much higher than

TABLE 2 | Ionic conductivity and *tLi*+ in previous relevant works.


that of Celgard 2325 (70%), PVDF-HFP (366%), PHL (466%), and PHP (498%) and further contribute in the better rate capability when adopting in the LIBs. PHP@PHL composite membrane demonstrated the highest porosity and electrolyte uptake ability, the smallest electrolyte contact angle, and there are several factors that contribute to its remarkable performances. Firstly, the molecular structure of PVDF-HFP, as well as the 3D nanofibrous morphology facilitated the electrolyte uptake, for its high polarity and the nanoporous structure. Secondly, Li2SiO<sup>3</sup> and PPCl reduced the order of molecular arrangement, which produced more chances for the electrolyte to be taken. Moreover, PPCl also worked as a plasticizer which has perfect affinity with the electrolyte, on which the contact angle is also reduced. Therefore, PHP@PHL shows the best physical performances. We also compare the ionic conductivity and the tLi <sup>+</sup> with some of the previous publications as listed in **Table 2**, obviously, by the synergistic effect of LSO and PPCl, lithium ion transport efficiency was strongly enhanced.

Galvanostatic cycling measurements in Li/electrolyte/Li symmetrical cells can probed into lithium plating/stripping process and analyze the interfacial stability between the electrolyte and lithium electrode. The charge/discharge cycling test was performed at a fixed current density of 0.5 mA/cm<sup>2</sup> with a total capacity of 1 mAh/cm<sup>2</sup> and the results are shown in **Figure 4**. As can be seen from the Celgard 2325 voltage-time profile, it exhibits a gradual increase in hysteresis (overpotential between Li deposition and dissolution) as the time increases, and the voltage increases to 0.65 V after 300 h. In other words, the SEI film, or dead lithium is thickening with cycling, which means an out-off-balance lithium plating/stripping. The voltagetime profiles of PVDF-HFP, PHL, and PHP@PHL electrolytes were much stable compared with that of Celgard 2325. In particular, although the tLi <sup>+</sup> of PVDF-HFP is lower than that of Celgard 2325, while it turns quite stable in the initial 500 h owing to the fibrous structure of the membrane which could balance the Li<sup>+</sup> flux adjacent to the electrode. PHL presents the least hysteresis after 1,000 h plating/stripping, while as it shows several abnormal convex during the cycling, therefore, we consider that the PHP@PHL cell demonstrated the best stability. As discussed before, the enhanced performance of PHP@PHL can be attributed to several reasons, firstly, the

nanofibrous structure of PVDF-HFP balanced the Li<sup>+</sup> flux and suppressed the mossy deposition of Li; secondly, the synergistic effect of LSO and PPCl reduce the crystallinity of the polymer matrix, thus enhanced the Li<sup>+</sup> transportation efficiency; lastly, Cl anion from the IL may partly be involved in the formation of SEI, which further enhance the cycle stability of the symmetric cell. Hence, the formation of Li dendrites is inhibited and the speed of dendrite growth is reduced as well. These results manifest that using 3D porous electrostatic spinning nanofiber membrane could effectively lower the hysteresis, stabilize cycling behavior and elongate the cell lifetime.

The batteries tested with the above four electrolytes were then disassembled in an argon-protected glove box, the electrodes were cleaned with diethyl carbonate (DEC) and dried strictly, and then put into a vacuum transfer box before FE-SEM analysis. The experimental results are shown in **Figure S3**. For the symmetrical lithium battery assembled with Celgard 2325 after galvanostatic charge/discharge cycle for 300 h, the FE-SEM diagram indicates that a lot of clavate-shaped lithium dendrite is generated on the surface of the lithium electrode. These dendrites will eventually penetrate the separator, causing short circuit and safety problems. By contrast, the symmetrical lithium battery assembled with the electrostatic spinning films, PVDF-HFP, PHL, and PHP@PHL, even cycled for 1,000 h under the same current density, shows much smoother surface with smooth edges, which was not easy to penetrate the separators, greatly improved the safety of LIBs. The electrospinning separators have large specific surface area, which could balance the local current density on the electrode surface, thus inhibited the formation and growth of lithium dendrite. Therefore, the core-shell nanofiber separator prepared by coaxial electrostatic spinning can significantly improve the safety performance of LIB.

Linear sweep voltammetry (LSV) was used to investigate the electrochemical windows of the electrolytes with a scan rate of 0.5 mV/s over the range of 2.8–6 V in Li/electrolyte/SS battery, and the results are shown in **Figure 5**. The oxidation potentials of the Celgard 2325, PVDF-HFP, PHL, and PHP@PHL electrolyte is 4.27, 3.93, 5.5, and 5.45 V (vs. Li+/Li), respectively. As all these membranes are plasticized by the same electrolyte, LiPF6/EC+EMC, therefore, most parasitic reactions should origin from the membrane. Therefore, it can be concluded that both LSO and PPCl could prevent PVDF-HFP from being oxidized and effectively enhance the electrochemical stability of composite membrane, which is pivotal for practical application. No obvious decomposition of Li/PHP@PHL/SS battery was observed below 5.45 V, which manifested PHP@PHL may potentially be able to be coupled with high voltage cathode materials, such as NCM622.

The interfacial stability between the electrode and the electrolyte is fatal for long-term cycle stability and rate performance, therefore, the separators are adopted and Li/electrolyte/NCM622 half-cells were assembled to further test their compatibility with the electrode materials. Static EIS tests were performed in 1, 10, 20, and 30 days after cell assembled at room temperature and the data was deal with an equivalent circuit fitting (**Figure S4**), the results are shown in **Figure 6** and **Table S1**. R<sup>b</sup> is corresponding to the bulk resistance originated from electrolyte and other cell components, R<sup>1</sup> and R<sup>2</sup> can be assigned to the interface resistance between electrolyte and anode or cathode, respectively. R<sup>2</sup> is smaller than R<sup>1</sup> because the battery is never charged after assembly and is left standing, therefore, little reaction would happen in the storage process. The cell with PHP@PHL shows the most stable R<sup>1</sup> value among the four cells, it is 18.32 after 30 days of storage, while it is 61.40, 35.45, and 35.42 for Celgard 2325, PVDF-HFP and PHL, respectively, which indicates that PHP@PHL has the best interfacial compatibility with Li among these electrolytes (Cheng et al., 2018). Its smallest increase in interface resistance is mainly attributed to the following reason: firstly, PVDF-HFP has higher dielectric constant than PP or PE, which endows it with better Li compatibility (Lopez et al., 2018); secondly, the introduction of Li2SiO<sup>3</sup> partially suppresses the decomposition of LiPF<sup>6</sup> and carbonate solvents during long term storage (Fu et al., 2017); and the third, the slow dissociation of PPCl would generate some Cl anion, which would react with Li to form a more stable SEI component, LiCl, thus further stabilized the electrode/electrolyte interface and reduce its resistance (Lu et al., 2014b).

Electrochemical performances of the above mentioned GPEs were tested in Li/electrolyte/NCM622 half cells by different C rates. **Figure S5** shows the 0.5C rate performance of Celgard 2325 and PHP@PHL, we can see that the cells delivers quite similar capacity in the initial 300 cycles, both cells kept more than 100 mAh/g reversible capacity in 500 cycles, which proves that the cathode material works well in this electrolyte system. It also can be notice that the Celgard one shows lower coulombic efficiency after about 350 cycles, which might cause by the mossy Li deposition after long term striping/plating. High rate tests were performed to testify the performance of nanofibrous GPEs. **Figures 7A,B** show the long term cycle performances of the Li/electrolyte/NCM622 half cells at 3C and

5C, respectively. Obviously, the reversible capacity and cycle stability of PHP@PHL system were significantly improved at these C rates. The initial discharge capacities of Celgard 2325, PVDF-HFP, PHL, PHP, and PHP@PHL were 156.6, 160.6, 161.4, 156.5, and 150.6 mAh/g at 3C rate, respectively. It drops sharply to 15.8 mAh/g after 300 cycles for the Celgard 2325 one, while that of PVDF-HFP, PHL, PHP, and PHP@PHL maintained at 85.3, 87.9, 84.1, and 98.5 mAh/g, respectively. The initial discharge capacities of Celgard 2325, PVDF-HFP, PHL, PHP, and PHP@PHL at 5C rate were 156.0, 156.0, 156.2, 158.7, and 160.2 mAh/g, respectively. It declined dramatically to 27.8 mAh/g after 300 cycles for Celgard 2325, and that of PVDF-HFP and PHL were also significantly reduced to 26.3 and 27.2 mAh/g after 400 cycles, respectively. But the one with PHP@PHL remained 78.5 mAh/g even after 500 cycles. The rate performances (from 0.2 to 20C) of Li/electrolyte/NCM622 half-cells are shown

in **Figure 7C**. It can be seen that the difference between the electrospining GPEs and the Celgard separator was not obvious under low current density, but the diversity significantly improved under large current density, especially for PHP@PHL. The Li/electrolyte/NCM622 half-cell with PHP@PHL can still deliver a reversible capacity of 65 mAh/g at 20C, while that of Celgard 2325, PVDF-HFP, PHL, and PHP were 5, 30, 38, and 29 mAh/g, respectively, which means the PHP@PHL delivers 13 times higher capacity than that of Celgard 2325. Chargedischarge profiles of the Li/electrolyte/NCM622 half-cells with Celgard 2325 and PHP@PHL at different C rates were compared in **Figure S6**. It can be easily conclude from these results that the as prepared core-shell structure PHP@PHL nanofibrous electrolyte has the best electrochemical performances, which can realize safe and efficient quick charging, thus has great potential in the application of high-power density LIBs.

#### CONCLUSIONS

In summary, a core-shell structured nanofibrous membrane with PPCl ionic liquid plasticizer in the core and inorganic nano-particle Li2SiO<sup>3</sup> as the filler in shell, PHP@PHL, was prepared by a facile coaxial electrospining method. The electrolyte uptake, porosity, ionic conductivity and lithium ion transference number of PHP@PHL nanofiber membrane were about 597%, 74.0%, 4.05 mS/cm, and 0.62, respectively. By the introduction of PPCl and Li2SiO3, ionic conductivity of the electrolyte increases by an order of magnitude and wettability between the separator and liquid electrolyte also significantly improved compared to commercial Celgard 2325 or even electrospining PVDF-HFP separator. In addition, it has good thermal stability, low interfacial impedance and wide electrochemical window. Symmetrical lithium cells with PHP@PHL demonstrated excellent plating/stripping cycling stability for about 1,000 h without short-circuit, which is 5 times longer than that of Celgard 2325. Moreover, the PHP@PHL electrolyte presents outstanding rate capability, it delivers a reversible capacity of 65 mAh/g at 20C compared to the 5 mAh/g in the case of Celgard 2325. The long term cycle performance was also significantly improved, as demonstrated in 3C and 5C. The core-shell structured nanofibrous membrane, in which different fillers or plasticizers could be used thus different functions can be realized in one single membrane, provides and effective method to enhance the overall performances of LIBs.

#### REFERENCES


### DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

#### AUTHOR CONTRIBUTIONS

SZ helps to improve ideas and experimental platform. LZ helps to solve some problems of experience. XL helps to do mainly works. YR helps to do any experience.

### FUNDING

This work was financially supported by the National Key Research and Development Program of China (No. 2016YFB0100303), National Natural Science Foundation of China (No. 21706261), Beijing Natural Science Foundation (No. L172045), and the Ford-China University Research Program.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00421/full#supplementary-material

activity for neurotransmitter dopamine. J. Phys. Chem. C 113, 15175–15181. doi: 10.1021/jp904788d


as an ionic liquid host: enhanced ion transport for electrochromic devices and lithium-ion batteries. J. Mater. Chem. A 3, 16040–16049. doi: 10.1039/C5TA02154E


separator for lithium–sulfur batteries. J. Mater. Chem. A 5, 10936–10945. doi: 10.1039/C7TA02245J

**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 Liu, Ren, Zhang 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.

# Bimetallic Metal-Organic Framework-Derived Carbon Nanotube-Based Frameworks for Enhanced Capacitive Deionization and Zn-Air Battery

Wenhui Shi 1†, Xilian Xu2†, Chenzeng Ye1†, Dongyong Sha<sup>2</sup> , Ruilian Yin<sup>2</sup> , Xuhai Shen<sup>2</sup> , Xiaoyue Liu<sup>1</sup> , Wenxian Liu<sup>2</sup> , Jiangnan Shen<sup>1</sup> \*, Xiehong Cao<sup>2</sup> \* and Congjie Gao<sup>1</sup>

*<sup>1</sup> Center for Membrane Separation and Water Science and Technology, Ocean College, Zhejiang University of Technology, Hangzhou, China, <sup>2</sup> College of Materials Science and Engineering, Zhejiang University of Technology, Hangzhou, China*

#### Edited by:

*Yuxin Tang, University of Macau, China*

#### Reviewed by:

*Xihong Lu, Sun Yat-Sen University, China Shengjie Peng, Nanjing University of Aeronautics and Astronautics, China Xingtao Xu, Hohai University, China*

#### \*Correspondence:

*Jiangnan Shen shenjn@zjut.edu.cn Xiehong Cao gcscaoxh@zjut.edu.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *03 May 2019* Accepted: *04 June 2019* Published: *19 June 2019*

#### Citation:

*Shi W, Xu X, Ye C, Sha D, Yin R, Shen X, Liu X, Liu W, Shen J, Cao X and Gao C (2019) Bimetallic Metal-Organic Framework-Derived Carbon Nanotube-Based Frameworks for Enhanced Capacitive Deionization and Zn-Air Battery. Front. Chem. 7:449. doi: 10.3389/fchem.2019.00449* Carbon-based materials have attracted intensive attentions for a wide range of energy and environment-related applications. Energy storage/conversion devices with improved performance have been achieved by utilization of metal-organic-framework (MOF)-derived carbon structures as active materials in recent years. However, the effects of MOF precursors on the performance of derived carbon materials are rarely investigated. Here, we report that the incorporation of small amount of Fe or Ni in Co-based MOFs leads to a significant enhancement for the derived carbon nanotube-based frameworks (CNTFs) in Na+/Cl<sup>−</sup> ion electrosorption. Further investigation revealed the enhanced performance can be attributed to the improved specific surface area, electrical conductivity, and electrochemical activity. Notably, the CoFe-CNTF derived from bimetallic CoFe-MOFs achieves a high ion adsorption capacity of 37.0 mg g−<sup>1</sup> , superior to most of recently reported carbon-based materials. Furthermore, the CoFe-CNTF also demonstrates high catalytic activity toward oxygen evolution reaction (OER) with a Tafel slope of 87.7 mV dec−<sup>1</sup> . After combination with three-dimensional graphene foam (3DG), the resultant CoFe-CNTF-coated 3DG is used as air-cathode to fabricate a flexible all-solid-state Zn-air battery, which exhibits a high open circuit potential of 1.455 V. Importantly, the fabricated flexible battery can light a light-emitting diode (LED) even when it is bent. This work provides new insights into designs of high-performance and flexible electrode based on MOF-derived materials.

Keywords: metal-organic frameworks, carbon nanotubes, capacitive deionization, oxygen evolution reaction, Zn-air battery, flexible devices, three-dimensional graphene, hierarchical structures

## INTRODUCTION

Metal-organic frameworks (MOFs), as a group of porous crystalline materials with tailorable composition and structure, are versatile precursors for carbon-based porous materials (Tang et al., 2015; Xu et al., 2017, 2018). MOF-derived carbon-based materials have attracted intensive attentions for a wide range of applications, benefiting from their high surface area and well-defined porosity (Wang et al., 2016, 2017b, 2019; Li et al., 2018b; Liu et al., 2019a). Capacitive deionization

**54**

Shi et al. MOF-Derived CNT Frameworks

(CDI) has emerged as a potential water treatment technology owing to its low cost and energy consumption (Song et al., 2019; Xu et al., 2019). In the CDI process, upon the application of an external voltage, ions in the solution are adsorbed and stored in the electrodes by forming electrical double layers (Suss et al., 2015; Hassanvand et al., 2018). A series of carbon-based materials have been extensively studied as CDI electrodes, such as activated carbon (Wu et al., 2016; Tang et al., 2019), graphene (Cao et al., 2014; Shi et al., 2016; Liu et al., 2017a; Li et al., 2018a), mesoporous carbon (Tsouris et al., 2011; Wang et al., 2014; Gao et al., 2019b), carbon nanotubes (Nie et al., 2012; Liu et al., 2015) and their composites (Xu et al., 2016b; Huang et al., 2017; Hu et al., 2018). Recently, MOF-derived porous carbon has demonstrated great potential for CDI applications. However, MOF-derived carbon materials generally have poor electrical conductivity because of low crystallinity of carbon (Gao et al., 2018, 2019a). Furthermore, abundant micropores within MOF-derived carbon structures limit the diffusion of ions into the electrode, leading to unsatisfactory CDI performance (Ding et al., 2017; Shi et al., 2018, 2019a). Therefore, novel strategies are required to further improve electron and ion transport of MOF-derived carbons (Wang et al., 2017a; Shi et al., 2019b).

In addition, MOF-derived nanostructured carbons are also attractive noble-metal-free electrocatalysts, which are promising in oxygen evolution reaction (OER) and have been used as cathode materials for Zn-air batteries recently (Liu et al., 2018, 2019b; Zhang et al., 2018b; Fang et al., 2019; Guo et al., 2019). Although the bright prospects of MOF-derived carbon materials, their practical applications are still restricted. First, micropores within MOF-derived porous carbon materials contribute to the majority of specific surface area, which have limited diffusion rate for electrolyte ions. Second, the development of nextgeneration flexible energy storage/conversion devices requires electrode materials with robust structure and strong mechanical properties under deformation (Han et al., 2018; Lv et al., 2018; Peng et al., 2018; Zhu et al., 2018; Ji et al., 2019). However, MOFderived carbons are normally in powder form, which is difficult to form a firm interaction within a flexible electrode. Third, previous reports have shown that metallic content within MOFderived carbons play an essential role in their catalytic properties (Liu et al., 2017b; Zhang et al., 2018a; Zheng et al., 2019). The effects of bimetallic MOF precursors on the performance of derived carbon materials are rarely investigated. Therefore, it is still a challenge to develop hierarchically porous, flexible, highion/charge-transport-rate MOF-derived carbon-based materials for high-performance Zn-air battery.

In this work, unique carbon nanotube-based frameworks (CNTFs) were obtained by a facile annealing process of rationally designed bimetallic MOF crystals (i.e., CoFe and CoNi-MOFs) as precursors. By the incorporation of a small amount of Fe or Ni in Co-based frameworks, the obtained CoFe- and CoNi-CNTFs both exhibited improved specific surface area and electrical conductivity, as compared to those of Co-CNTF. Due to the hierarchically porous structure with abundant exposed active sites, high conductivity for fast electron transfer, and unique nanostructures with metal nanoparticle embedded in CNTs, the CoFe- and CoNi-CNTF electrodes exhibited remarkably enhanced Na+/Cl<sup>−</sup> ions adsorption property. Furthermore, high catalytic activity toward oxygen evolution reaction (OER) was also achieved by CoFe-CNTF. As a proof-of-concept application, an air-cathode based on CoFe-CNTF coated 3D graphene foam (CoFe-CNTF@3DG) was fabricated, and then used for flexible all-solid-state Zn-air battery, which displayed a high open circuit potential of 1.455 V.

### EXPERIMENTAL SECTION

#### Material Synthesis

Preparation of Co-MOF: Typically, 1.4553 g of Co(NO3)2·6H2O was dissolved in 50 mL of methanol, which was mixed with 50 mL of methanol containing 3.2860 g of 2-methylimidazole. The obtained mixture was stirred at room temperature for 12 h. The sample was collected by centrifugation and washed with methanol for several times, followed by drying at 60◦C for 12 h in vacuum oven.

Preparation of bimetallic CoFe and CoNi-MOFs: CoFe-MOF was synthesized by using Co(NO3)2·6H2O and Fe(NO3)3·6H2O with a molar ratio of 9:1. Typically, 1.3097 g of Co(NO3)2·6H2O and 0.2019 g of Fe(NO3)3·6H2O were dissolved in 50 mL of methanol. Then, 50 mL of methanol dissolved with 3.2860 g of 2 methylimidazole was added into the above solution followed by continuously stirring for 12 h at room temperature. CoFe-MOF was collected by centrifugation and washed with methanol for several times, followed by drying at 60◦C for 12 h in vacuum oven. CoNi-MOF was synthesized by a similar procedure using Co(NO3)2·6H2O and Ni(NO3)2·6H2O with a molar ratio of 9:1.

Preparation of CNTFs: The as-prepared MOF precursors were pyrolyzed in a tube furnace by a two-step annealing process. The MOF precursors were first heated to 500◦C with a heating rate of 5◦C min−<sup>1</sup> , followed by bubbling ethanol vapor using H<sup>2</sup> for 30 min. Then, the samples were heated at 800◦C for 1 h under N<sup>2</sup> atmosphere, followed by immersing the sample in 1 M HCl for 12 h to obtain CNTFs.

Preparation of CoFe-CNTF@3DG: 3D graphene foam (3DG) was obtained by using our previously reported method (Cao et al., 2011). Then 3DG was treated with a concentrated HNO<sup>3</sup> solution at 80◦C for 4 h, followed by washing with deionized water to obtain hydrophilic 3DG. The 3DG was placed into the precursor solution for CoFe-MOF and reacted for 12 h at room temperature. The CoFe-MOF@3DG composite was obtained by washing with methanol and then drying at 60◦C for 12 h. The CoFe-CNTF@3DG was obtained according to the aforementioned annealing procedures for CNTF samples.

### Material Characterization

The morphologies of the samples were investigated by fieldemission scanning electron microscopy (FE-SEM, Hitachi SU-8010), equipped with energy-dispersive X-ray spectroscopy (EDX) and transmission electron microscopy (TEM, JEM-100CX II). The crystal structures were characterized by powder Xray diffraction (XRD, PNAlytical X'Pert PRO) with Cu Kα radiation. Nitrogen adsorption/desorption measurements were carried out by Micromeritics ASAP 2020. The specific surface areas were calculated by the multipoint Brunauer-Emmett-Teller (BET) method.

#### Electrosorption Measurements

Batch-mode CDI tests were conducted by a circulating system, which includes a CDI module, a peristaltic pump (Longer Pump, YZ-1515x), a conductivity meter (Leici, DDSJ-308F) and a source meter (Keithley, SMU-2,400) (**Figure S1A**). The CDI electrodes were fabricated by coating a mixture of active materials, carbon black and poly(vinylidene fluoride) (PVDF) with a mass ratio of 8:1:1 in n-methylpyrrolidone (NMP) solution onto graphite paper, followed by drying at 60◦C for 12 h in vacuum oven. As shown in **Figure S1B**, the CDI module was assembled with a CDI electrode, cation exchange membrane (Hangzhou Grion Environmental Technology Co., Ltd., LEHeCM-1, Type 1), spacer, anion exchange membrane (Hangzhou Grion Environmental Technology Co., Ltd., LE-HeAM-I, Type 1) and another CDI electrode. The total volume of NaCl aqueous solution is maintained at 50 mL. The electrosorption performance was measured at a flow rate of 30 mL min−<sup>1</sup> , varied applied voltages (1.0–1.6 V) and concentrations of NaCl solution (125–1,000 mg L−<sup>1</sup> ). The electrosorption capacity, Γ (mg g−<sup>1</sup> ), is calculated based on the following equation:

$$
\Gamma = \frac{\left(\text{C}\_0 - \text{C}\_t\right)V}{\text{m}}
$$

where C<sup>0</sup> is the initial concentration of NaCl solution (mg L−<sup>1</sup> ), Ct is the final concentration (mg L−<sup>1</sup> ), V is the total volume of the solution (L), and m is the total mass of the electrodes (g).

#### Electrochemical Measurements

The electrochemical properties were measured by an electrochemical workstation (Chenhua, China, CHI760E) in a three-electrode system at room temperature, using a carbon paper coated with prepared catalyst as the working electrode, a graphite rod as the counter electrode, a Saturated Calomel Electrode (SCE) as the reference electrode and 1.0 M KOH as the electrolyte. The catalyst slurry was obtained by adding 5 mg of CoFe-CNTF powder into a mixed solution of 768 µL DI water, 200 µL ethanol and 32 µL nafion, followed by ultrasonication for 30 min. The electrode was fabricated by coating the obtained slurry onto carbon paper with a loading mass of 1 mg cm−<sup>2</sup> . The Linear Sweep Voltammetry (LSV) was tested at a scan rate of 5 mV s−<sup>1</sup> with a potential range from 1.2 to 1.7 V vs. Reversible Hydrogen Electrode (RHE). The Electrochemical Impedance Spectroscopy (EIS) measurement was conducted by an electrochemical station (Autolab PGSTAT302N) at a potential of 1.33 V vs. RHE in a frequency range from 100 kHz to 0.01 Hz with an amplitude of 5 mV.

### Fabrication of Zn-Air Battery

Zn-air battery was assembled using a carbon paper loaded with CoFe-CNTF catalyst as the air cathode, Zn plate as the anode, and a mixed solution of 6 M KOH and 0.2 M Zn(Ac)<sup>2</sup> as the electrolyte, which was tested on a battery testing system (Land CT2001A). The flexible solid-state Zn-air battery was assembled by using CoFe-CNTF@3DG as the air electrode, nickel foam as the current collector, Zn plate as the anode, and polyacrylic acid (PAA) gel containing 11.25 M KOH and 0.25 M ZnO as the solid electrolyte.

### RESULTS AND DISCUSSION

The preparation of CNTF is schematically shown in **Scheme 1**. Bimetallic MOF crystals (i.e., CoFe- and CoNi-MOFs) were first synthesized using the mixtures of metal salts with a proper molar ratio (route 1, see Experimental Section). For comparison, monometallic Co-MOF was also prepared. Then, the as-prepared MOF precursors were undergone a two-step annealing process for pyrolysis of MOFs and catalytic growth of CNTs, followed by an acid treatment to obtain the CNTbased frameworks (i.e., CoFe-CNTF, CoNi-CNTF, and Co-CNTF). Furthermore, CoFe-MOFs were in-situ grown on 3D graphene foam (3DG) followed by similar annealing process to produce a flexible air cathode (CoFe-CNTF@3DG, route 2, see Experimental Section).

The scanning electron microscopy (SEM) images of prepared bimetallic MOF crystals exhibit typical dodecahedron shape (**Figures S2A–C**), which is similar with that of Co-MOF. In addition, the energy-dispersive X-ray spectroscopy (EDX) spectra in **Figures S2D–F** indicate all of the MOFs show signals of C, N, O, and Co. Besides, CoFe- and CoNi-MOFs contain additional Fe or Ni elements, suggesting the successful incorporation of Fe/Ni in the Co-based frameworks. The X-ray diffraction (XRD) patterns in **Figure S3** show the same diffraction peaks for CoFe-, CoNi-MOFs, and Co-MOF, demonstrating the same crystal structure (Kaur et al., 2016). SEM images in **Figure S4** further reveal that CNTFs derived from CoFe, CoNi, and Co-MOFs maintained a similar dodecahedron shape after the annealing process, which consist of interconnected CNT frameworks. Further investigations on the morphology and microstructure of CNTFs were performed by transmission electron microscopy (TEM). As shown in **Figures 1A–C**, the obtained CNTFs exhibited hierarchically porous structures assembled by CNTs, which have nanoparticles with diameters of 10–18 nm encapsulated on the top of tubes (Li et al., 2016). XRD patterns of all CNTFs exhibit similar characteristic peaks located at ∼26 and 44◦ (**Figure 1D**), which can be assigned to the (002) and (101) planes of graphitic carbon (JCPDS: 41-1487) (Xu et al., 2016a).

The electrosorption performance of the prepared three CNTFs samples for Na+/Cl<sup>−</sup> ions were evaluated by measuring change of solution conductivity at applied voltages using a CDI device (**Figure S1**, see experimental section). As displayed in **Figure 1E**, when an external voltage was applied, Na+/Cl<sup>−</sup> ions were adsorbed and trapped in the electrodes, along with the gradually decreased solution conductivity. The NaCl electrosorption capacity of the CoFe-CNTF electrode is 37.0 mg g−<sup>1</sup> , which is much higher than that of Co-CNTF (19.9 mg g−<sup>1</sup> ). **Figure 1F** depicts CDI Ragone plot of the three electrodes, in which the adsorption

capacity and rate for Na+/Cl<sup>−</sup> ions are used to assess the electrosorption behavior of the electrodes. Notably, in the CDI Ragone plot, the curve of CoFe-CNTF electrode

is located in the upper right corner, suggesting both the highest Na+/Cl<sup>−</sup> ions adsorption capacity and adsorption rate (Lee et al., 2018).

The aforementioned results demonstrate that the remarkably enhanced Na <sup>+</sup>/Cl<sup>−</sup> ions adsorption performance was achieved by the bimetallic MOF-derived CNTFs. To further reveal the origins of the improvement of CDI performance, further investigations on the properties of samples in terms of electrochemical capacitance, electrical conductivity, and specific surface area were carried out. The cyclic voltammetry (CV) profiles of CoFe-, CoNi-, and Co-CNTF electrodes are shown in **Figure 2A**. Obviously, the CoFe-CNTF electrode exhibited the largest integrated area of CV curve, which is consistent with the calculated specific capacitances. The CoFe-CNTF electrode showed the highest specific capacitance at various scan rates, i.e., 153 F g−<sup>1</sup> at 2 mV s−<sup>1</sup> and 72 F g−<sup>1</sup> at 100 mV s−<sup>1</sup> (see **Figure 2B**). To further study the electrical conductivity and internal resistivity of CNTF electrodes, electrochemical impedance spectroscopy (EIS) measurement was conducted. **Figure 2C** exhibits the Nyquist plots of CoFe-, CoNi-, and Co-CNTF electrodes in 1 M NaCl solution, which include a semicircle in the high frequency region and an inclined line in the low frequency region. The inclined line of the CoFe-CNTF electrode shows the largest slope, suggesting faster ion diffusion into the porous structure. In addition, the intercept of the curve in the high-frequency region with the real axis represents the equivalent series resistance (ESR) of the electrodes. The CoFe-CNTF electrode showed an ESR value of 2.38 , smaller than that of Co-CNTF (2.83 ). The smaller ESR and fast ion diffusion of CoFe-CNTF electrode enhance the capacitive performance.

Additionally, to investigate the specific surface area and pore structure of CNTFs, N<sup>2</sup> adsorption-desorption measurements were carried out. As shown in **Figure 2D**, both CoFe- and CoNi-CNTFs displayed type-IV isotherms with sharply increased adsorption at low pressures and distinct hysteresis loops at medium pressure region, indicating existence of micropores, and abundant mesopores (Ding et al., 2018). Compared with CoFeand CoNi-CNTFs, Co-CNTF exhibited a smaller hysteresis loop at high relative pressure, indicating that CoFe- and CoNi-CNTFs possesses more mesopores. Moreover, the pore size distributions of Co-CNTF, CoNi-CNTF, and CoFe-CNTF in **Figure S5** also suggest more mesopores existed in CoNi-CNTF and CoFe-CNTF than that of Co-CNTF. The Brunauer-Emmett-Teller (BET) surface areas of CoFe and CoNi-CNTF are 410 and 369 m<sup>2</sup> g −1 , respectively, which is higher than that of Co-CNTF (334 m<sup>2</sup> g −1 ). Obviously, the CoFe-CNTF with the highest specific surface area and abundant mesopore is in favor of the diffusion of ions, thus, it is promising electrode materials for CDI.

The CDI performance of the CoFe-CNTF electrode was further studied in terms of the effects of various applied voltages and concentrations of NaCl solution. As shown in **Figure 3A**, with the increase of the applied voltage, the electrosorption capacity was increased, indicating that a larger amount of ions were removed from the solution due to stronger electrostatic interaction. The corresponding transient current curves at different applied voltages in **Figure 3B** also show an increased current at higher voltages. Furthermore,

electrosorption experiments in NaCl solutions of various initial concentrations ranging from 125 to 1,000 mg L−<sup>1</sup> were carried out. **Figure 3C** demonstrates the calculated adsorption capacities of CoFe-CNTF electrode in different concentrations of NaCl solution at cell voltages of 1.0, 1.2, 1.4, and 1.6 V, respectively. With increment of the concentration of NaCl solution, the adsorption capacity increases as well. Notably, the CoFe-CNTF electrode exhibited a superior adsorption capacity of 37.0 mg g−<sup>1</sup> in a 500 mg L−<sup>1</sup> NaCl solution at 1.2 V, which is superior to most of the recently reported CDI electrode materials (see **Table S1**). Furthermore, the regeneration stability of the electrode was investigated by repeated electrosorption-desorption experiments at a voltage of 1.2 V. As shown in **Figure 3D**, CoFe-CNTF electrode exhibited excellent stability for 12 cycles without obvious capacity drop.

The above studies clearly demonstrate that CoFe-CNTF has an extraordinary specific surface area, conductivity, and electrochemical activity. This could be attributed to the hierarchically porous structure with abundant exposed active sites, high electrical conductivity for fast electron transfer, and unique nanostructures with Co/Fe nanoparticles embedded in CNTFs. These excellent properties endow them promising performance as air electrode of Zn-air batteries. As shown in **Scheme 1**, in order to apply it in flexible Zn-air batteries, the OER performance of CoFe-CNTF was first investigated. The OER activity of the as-obtained CoFe-CNTF was evaluated by a conventional three-electrode cell in 1.0 M KOH solution. Commercial IrO<sup>2</sup> was also tested for comparison. As shown in **Figure S6A**, the CoFe-CNTF catalyst require a potential of 1.574 V vs. RHE (corresponding to overpotential of 344 mV) to achieve a current density of 10 mA cm−<sup>2</sup> , which is better than that of commercial IrO<sup>2</sup> (1.658 V). Moreover, the Tafel slope of CoFe-CNTF is 87.7 mV dec−<sup>1</sup> , which is much smaller than that of commercial IrO<sup>2</sup> (106.5 mV dec−<sup>1</sup> ), suggesting faster OER catalytic kinetics of CoFe-CNTF (see **Figure S6B**). The Nyquist plots in **Figure S6C** demonstrate that the charge transfer resistance of CoFe-CNTF is much smaller than that of commercial IrO2, indicating that charge transfer process within CoFe-CNTF is more effective. Notably, CoFe-CNTF also exhibits excellent OER stability. As shown in **Figure S6D**, CoFe-CNTF shows nearly constant operating potential with a current density of 10 mA cm−<sup>2</sup> for 12 h.

To demonstrate the practical application of the obtained CoFe-CNTF, a rechargeable Zn-air battery was fabricated. **Figure 4A** shows a schematic diagram of the configuration of rechargeable Zn-air battery, in which CoFe-CNTF catalyst loaded on carbon paper was used as air-cathode, Zn plate as anode, and a mixed solution of 6 M KOH and 0.2 M Zn(Ac)<sup>2</sup> as electrolyte. Two rechargeable Zn-air batteries connected in series can light up a light-emitting diode (LED) bulb successfully (**Figure 4B**). Furthermore, the fabricated Zn-air battery exhibited remarkable cycling stability after operation for 420 h at a current density of 15 mA cm−<sup>2</sup> (**Figure 4C**). In addition, a flexible Znair battery based on CoFe-CNTF is also demonstrated. CoFe-MOFs were grown on 3DG (see **Figure S7**) followed by a similar annealing process to produce the flexible air cathode

FIGURE 5 | (A) SEM image of CoFe-MOF@3DG. Inset is high-magnification SEM image showing a CoFe-MOF crystal grown on the surface of 3DG. (B) SEM image of CoFe-CNTF@3DG. Insets: photograph of a bent CoFe-CNTF@3DG, and high-magnification SEM image (Bottom). (C) Open-circuit voltage of a fabricated all-solid-state Zn-air battery based on CoFe-CNTF@3DG. (D) Photographs of the lightened LED by two Zn-air batteries connected in series under different blending status.

of CoFe-CNTF@3DG. The morphologies and XRD patterns of obtained CoFe-MOF@3DG and CoFe-CNTF@3DG are shown in **Figures 5A,B** and **Figure S8**. The photograph in the top inset of **Figure 5B** clearly demonstrates good flexibility of the obtained CoFe-CNTF@3DG electrode. The open-circuit voltage of the fabricated all-solid-state Zn-air battery is 1.455 V (**Figure 5C**). Importantly, a LED bulb could be successfully lightened by two Zn-air batteries connected in series under different blending status (**Figure 5D**), demonstrating its excellent flexibility and promising application in Zn-air batteries.

### CONCLUSION

In summary, we developed hierarchically porous carbon nanotube-based frameworks (CNTFs), which were derived from rationally designed bimetallic MOFs followed by a two-step annealing process. The CNTFs showed excellent Na+/Cl<sup>−</sup> ion adsorption property. Especially, the CoFe-CNTF exhibited a remarkable adsorption capacity of 37.0 mg g−<sup>1</sup> in a 500 mg L <sup>−</sup><sup>1</sup> NaCl solution at 1.2 V, superior to that of Co-CNTF (19.9 mg g−<sup>1</sup> ). Such enhanced performance can be attributed to the larger surface area and improved electrical conductivity of the CoFe-CNTF. Moreover, due to the unique porous framework structure and abundant metal active sites, the CoFe-CNTF displayed excellent electrocatalytic performance in OER. CoFe-CNTF@3DG as air electrode demonstrated excellent flexibility and potential application for flexible all-solid-state Zn-air battery. We believe our work paves a way for the design of nanostructured carbon materials for the applications in sustainable energy conversion and storage.

### REFERENCES


### DATA AVAILABILITY

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

### AUTHOR CONTRIBUTIONS

WS, XC, JS, CG, and WL designed the project. WS, CY, XS, and XL performed the synthesis, CDI experiment, and data analysis. XX, DS, and RY performed the Zn-air battery experiment and data analysis. All authors contribute to editing and discussion of the manuscript.

### FUNDING

This work was supported by the financial support from Zhejiang Provincial Natural Science Foundation of China (LQ17B030002, LR19E020003) and the National Natural Science Foundation of China (51702286, 51602284), General Scientific Research Project of the Department of Education of Zhejiang Province, China (Y201839638), and the Thousand Talent Program and Qianjiang Scholars program of Zhejiang Province in China.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00449/full#supplementary-material

electrochemical supercapacitor and capacitive deionization. Electrochim. Acta 263, 85–93. doi: 10.1016/j.electacta.2018.01.044


**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 Shi, Xu, Ye, Sha, Yin, Shen, Liu, Liu, Shen, Cao and Gao. 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.

# CsPbBr3/CdS Core/Shell Structure Quantum Dots for Inverted Light-Emitting Diodes Application

Xiaosheng Tang, Jie Yang, Shiqi Li\*, Weiwei Chen, Zhiping Hu and Jing Qiu\*

*Key Laboratory of Optoelectronic Technology and Systems, College of Optoelectronic Engineering, Chongqing University, Ministry of Education, Chongqing, China*

Novel CsPbBr3/CdS core/shell structure quantum dots (QDs) were successfully synthesized using a facile hot-injection method. The corresponding CsPbBr3/CdS QDs based light-emitting diodes (QLEDs) were further prepared, which demonstrated the maximum luminance of 354 cd/m<sup>2</sup> and an external quantum efficiency (EQE) of 0.4% with the current efficiency (CE) of 0.3 cd/A. Moreover, the optoelectronic performance of the CsPbBr3/CdS QDs based QLEDs exhibited a comparable enhancement in contrast to the pure CsPbBr<sup>3</sup> QDs based QLEDs. Hypothetically, the novel CsPbBr3/CdS structure QDs introduced one new route for advanced light emission applications of perovskite materials.

#### Edited by:

*Yuxin Tang, University of Macau, China*

#### Reviewed by:

*Du Yuan, Nanyang Technological University, Singapore Jin Zhao, Nanyang Technological University, Singapore Wang Hui Bo, University of Macau, China*

#### \*Correspondence:

*Shiqi Li shiqili@cqu.edu.cn Jing Qiu jingqiu@cqu.edu.cn*

#### Specialty section:

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

Received: *25 April 2019* Accepted: *28 June 2019* Published: *12 July 2019*

#### Citation:

*Tang X, Yang J, Li S, Chen W, Hu Z and Qiu J (2019) CsPbBr3/CdS Core/Shell Structure Quantum Dots for Inverted Light-Emitting Diodes Application. Front. Chem. 7:499. doi: 10.3389/fchem.2019.00499* Keywords: semiconductors, luminescence, core/shell structure, quantum dots, light-emitting diode

## INTRODUCTION

During the past two decades, huge efforts have been devoted to various QDs as the lightemitting layer of QLEDs (Coe et al., 2002; Sun et al., 2007; Kwak et al., 2012; Seth and Samanta, 2016; Chen et al., 2017). Recently, perovskite QDs have attracted great attention due to their excellent optoelectronic performance, due to their tunable bandgap, high absorbance coefficient, and high photoluminescence quantum yield (PLQY) (Protesescu et al., 2015; Song et al., 2015; Tang et al., 2016; Heejae et al., 2017). However, the electroluminescence (EL) EQE of halide perovskite QLEDs is relatively lower to that of CdSe based QLEDs, which seriously limits further commercial application (Shirasaki et al., 2013; Dai et al., 2014; Wang et al., 2018). It is known that the photoluminescence (PL) performance of the QDs emitting layer plays a crucial role in QLED. Therefore, how to synthesize QDs with high quality is one critical step for preparing high performance QLEDs. It was recognized that the formation of core/shell structure between semiconductors is one effective approach to improve the performance of nanocrystals and QLEDs devices (Hines and Guyot-Sionnest, 1996; Li et al., 2011). Many research groups have reported that the non-radiative Auger recombination of CdSe QDs could be efficiently suppressed by being coated with larger bandgap materials such as CdS and ZnS, where better optical properties have been achieved (Chen et al., 2008; Bae et al., 2013; Efros and Nesbitt, 2016). However, few studies on the core/shell structure for perovskite materials have been reported. Therefore, it is still a challenge to investigate a perovskite-based core/shell structure and the corresponding QLEDs.

Herein, we synthesized novel colloidal perovskite core/shell QDs by covering CsPbBr<sup>3</sup> QDs with a CdS shell. Moreover, the inverted QLEDs (ITO/ZnO:Mg/QDs/CBP(4,4′ -Bis(N-carbazolyl)- 1,1′ -biphenyl)/MoO3/Al) based on CsPbBr3/CdS core/shell QDs and pure CsPbBr<sup>3</sup> QDs were fabricated, respectively. The CsPbBr3/CdS QDs based QLED exhibited the maximum luminance of 354 cd/m<sup>2</sup> with a CE of 0.3 cd/A and the best EQE of 0.4% was 5.4 times of the pure CsPbBr<sup>3</sup> QDs based QLEDs.

### MATERIALS AND METHODS

### Chemicals

Cs2CO<sup>3</sup> (Aldrich, 99.9%), PbBr<sup>3</sup> (ABCR, 98%), Cadmium (II) oxide (Sigma Aldrich, 99.5%), Oleic Acid (OA, Sigma Aldrich, 90%), 1-octadecene (ODE, Sigma Aldrich, tech. 90%), Sulfur (Sigma Aldrich, 99.98%), Oleylamine (OME, Sigma Aldrich, tech. 70%), Toluene (Sigma Aldrich, 99.8%), CBP (Xi'an Polymer Light Technology Co., Ltd.), Molybdenum Oxide (MoO3, from Aladdin-reagent), ZnO:Mg (Xingshuo Nano Technology Co., Ltd.).

### Cd-oleate Solution Synthesis

For the synthesis of Cd-oleate solution a protocol by Li et al. (2011) was adopted. A 0.38 M Cd-oleate solution was made by dissolving 383 mg CdO in 3.9 ml oleic acid and 3.9 ml ODE at 280◦C under N<sup>2</sup> flow. After 1 h the CdO was dissolved and the clear solution was degassed for 30 min at 110◦C.

## CsPbBr3/CdS Core/Shell QDs Synthesis

For the synthesis of CsPbBr<sup>3</sup> core QDs a protocol by Protesescu et al. (2015) was adopted. One hundred milligram Cs2CO<sup>3</sup> was loaded into a 100 ml 3-neck flask along with 4 ml 1-octadecene, and 0.5 ml OA (oleic acid), then heated under N<sup>2</sup> to 120◦C until the powder was completely dissolved. Five milliliter ODE and 69 mg PbBr<sup>2</sup> were loaded into a 100 ml 3-neck flask and heated under N<sup>2</sup> to 120◦C 1 h. 0.5 ml OME and 0.5 ml OA was injected at 120◦C under N2. After the PbBr<sup>2</sup> salt was completely dissolved, the temperature was raised to 150◦C and the Cs-oleate solution (0.4 ml, 0.125 M in ODE, prepared as described above) was quickly injected and, 5 s later, the reaction finished. Four milliliter ODE, 1 ml Cd-oleate solution, and 0.4 ml 1 M sulfur in OME solution, was mixed when the reaction for CsPbBr<sup>3</sup> QDs finished and added dropwise over 20 min to the CsPbBr<sup>3</sup> solution at 150◦C under N<sup>2</sup> flow. After the addition was complete, the mixture was allowed to react for 20 min at 150◦C and was subsequently cooled by an ice-water bath, washed with toluene several times, and dispersed in toluene.

### Preparation of QLED

The cleaned ITO/glass was treated under UV–ozone for 30 min. The ZnO:Mg nanoparticles were spin-coated onto ITO/glass at 3,000 rpm for 40 s, and annealed at 100◦C for 10 min. The perovskite QDs were deposited by spin-coating at 2,000 rpm for 60 s. CBP (40 nm), MoO<sup>3</sup> (10 nm), and Al (100 nm) electrodes were deposited using a thermal evaporation system through a shadow mask under a high vacuum of ≈1 × 10−<sup>4</sup> Pa. All device operations were performed in a nitrogen-filled glove box.

### Optical Characterization

Photoluminescence spectra were measured by Agilent Cary Eclipse spectrograph FLS920P. XRD characterization was done by Shimadzu/6100 X-ray diffractometer, using a Cu Kα radiation source (wavelength at 1.5405 Å). TEM images were recorded on a Zeiss/Libra 200 FE. The PLQY measurements were carried out in solutions with an Edinburgh Instruments fluorescence spectrometer (FLS920), which included a xenon lamp with monochromator for steady-state PL excitation. A calibrated integrating sphere was used for PLQY measurements. The EL spectra and luminance (L)—current density (J)—voltages (V)

Tang et al. CsPbBr3/CdS QDs Based QLED

characteristics were collected using a Keithley 2400 source and a PR-670 Spectra Scan spectrophotometer (Photo Research) at room temperature.

### RESULTS AND DISCUSSION

Based on the continuous injection method, a novel approach to overcome a non-radiative recombination of CsPbBr<sup>3</sup> QDs was developed by growing a thin CdS shell on the surface of the CsPbBr<sup>3</sup> core. **Figure 1A** shows the simulation of the CsPbBr3/CdS core/shell structure. The X-ray diffraction (XRD) patterns of the CsPbBr3/CdS core/shell QDs clearly show that two types of crystalline structures were formed (**Figure 1B**). Compared to the pure CsPbBr<sup>3</sup> QDs (**Figure S2**), the extra peaks of the CsPbBr3/CdS XRD patterns are fairly consistent with the standard XRD patterns of a CdS-liked zinc blende structure. From the transmission electron microscope (TEM) images of CsPbBr<sup>3</sup> (**Figure S1b**) and CsPbBr3/CdS QDs (**Figure 1C),** it can be seen that the point angle of the cubic shape become a circular arc. The high resolution TEM image of CsPbBr<sup>3</sup> (the inset of **Figure S1b**) and CsPbBr3/CdS QDs (the inset of **Figure 1C)** clearly exhibit the crystal lattice (110) of the CsPbBr<sup>3</sup> nanoparticles and further verify the CsPbBr3/CdS core/shell structure. In addition, the CsPbBr3/CdS QDs exhibited a typical cubic shape with a larger average size (22.1 nm) than that of the CsPbBr<sup>3</sup> QDs (12.7 nm; **Figure S1**). To further demonstrate the successful formation of the CsPbBr<sup>3</sup> core and CdS shell, the mapping analysis and energy dispersive X-ray (EDX) spectroscopy of CsPbBr3/CdS QDs was employed (**Figures 1D,E**) with the S and Cd elements uniformly distributed on the surface of the CsPbBr<sup>3</sup> core. Moreover, the PL spectra of CsPbBr3/CdS QDs has a slightly blueshift (from 519 to 514 nm) compared with pure CsPbBr<sup>3</sup> QD (**Figure 1F**), which was probably induced by the CdS shell. A similar phenomenon was found and ascribed for the cationic inter-diffusion at the interface between the CsPbBr<sup>3</sup> core and CdS shell (Oladeji and Chow, 2005; Li et al., 2011; Kwak et al., 2015). On the other side, the CsPbBr3/CdS showed a high photoluminescence quantum yield (PLQY) of 88% (**Figure S3**) and the full width at half-maximum (FWHM) was as narrow as 19 nm, which is similar to pure CsPbBr<sup>3</sup> QDs.

As shown in **Figure 2A**, the inverted QLEDs structure was fabricated with a sandwich structure of ITO (In2O3- SnO2)/ZnO:Mg/QDs/CBP/MoO3/Al, which was constituted by the transparent electrode ITO as a cathode, Mg doped ZnO as an electron transport layer, perovskite QDs as the lightemitting layer, CBP as a hole transport layer, Al as an anode, and MoO<sup>3</sup> as a hole injection layer. The flat-band energy levels of devices are shown in **Figures 2B,C**. After perovskite QDs were coated on the ZnO:Mg electron transport layer, the perovskite film was still soluble in organic solvents due to the presence of aliphatic ligands on the QDs. The structure above the light emitting layer was prepared by evaporation (**Materials and Methods section**).

As shown in **Figures 2D,E**, the peaks of the QLEDs EL spectra based on CsPbBr<sup>3</sup> QDs and CsPbBr3/CdS QDs were consistent with the PL spectra, and the luminescence intensity

of CsPbBr3 QDs (D) and CsPbBr3/CdS QDs (E) based QLED, the insets shows an operating device driven at 4 V.

increased with the voltage increase. The EL of pure CsPbBr<sup>3</sup> QDs based LEDs (**Figure 2D**), peaking at 519 nm with a FWHM of ∼20 nm, was observed. And there is no emission peak from the hole/electron transport layer. The CsPbBr3/CdS QDs based QLEDs also performed with a peak at 516 nm with a FWHM of ∼20 nm. As depicted in **Figure 2E**, the emission from the hole/electron transport layer was hardly shown. The insets showed an operating device driven at 4 V, and the devices both exhibited a saturated and pure green color.

For the potential application in a light emission device, the characteristic of emitter layers is extremely important. For the inverted QLEDs with the CsPbBr**<sup>3</sup>** QDs emitting layer, the pure CsPbBr<sup>3</sup> QDs worked with poorer luminous characteristics, exhibiting a lower maximum luminance intensity of 65 cd/m<sup>2</sup> and the maximum CE of 0.14 cd/A (**Figures 3A,B**). It is worth mentioning that a maximum luminance of 354 cd/m<sup>2</sup> was obtained when the CsPbBr3/CdS QDs were used as the emitting layer, and the maximum CE was boosted to 0.3 cd/A (**Figures 3C,D**). The overall performance of two kinds of devices were also described with the average EQE. The EQE of QLEDs with the CsPbBr<sup>3</sup> QDs emitting layer (0.4%) was 5.4 times larger than the pure CsPbBr<sup>3</sup> QDs (0.07%). The results are tabulated in

TABLE 1 | Comparison of performance parameters between CsPbBr3 QDs and CsPbBr3 QDs based QLED.


**Table 1** for comparison. Based on the same device structure with a different emitting layer, the turn-on voltage of CsPbBr3/CdS based QLEDs was slightly lower than CsPbBr<sup>3</sup> based QLEDs. Moreover, the maximum CE of the CsPbBr3/CdS based QLEDs was 2.1 times larger than the pure CsPbBr<sup>3</sup> based QLEDs, which could be ascribed to the valence bands (conduction bands) of the CdS shell which are lower (or higher) than that of CsPbBr<sup>3</sup> core, and a non-radiative Auger recombination of CsPbBr<sup>3</sup> was effectively suppressed (Park et al., 2014; Nasilowski et al., 2015; Niu et al., 2017). Additionally, the probable reason could be attributed to the fact that the extra carriers cannot escape from the surface of the QDs as the protecting layer of CdS, which leads to a higher radiative recombination rate, higher luminance and EQE.

## CONCLUSION

In conclusion, we have developed novel all-inorganic core/shell perovskite QDs and applied it as an emitting layer in inverted QLEDs (ITO/ZnO:Mg/QDs/CBP/MoO3/Al). The results show that the fabricated CsPbBr3/CdS QDs based QLEDs exhibited enhanced performance compared with pure CsPbBr<sup>3</sup> QDs based QLEDs. The EQE of CsPbBr3/CdS QDs was 5.4 times higher than that of pure CsPbBr<sup>3</sup> QDs, which demonstrates that the introduction of a CdS shell can increase the optoelectronic performance. The core/shell structure perovskite QDs present a new route of perovskite materials for light emission applications.

### DATA AVAILABILITY

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

## AUTHOR CONTRIBUTIONS

XT and JY contributed equally to this work. XT and JQ designed the experiment. JY and SL conducted the experiments and characterization. XT and JQ wrote and revised the paper. JY, WC, and ZH participated in the discussion. XT funded some of the subject experiments.

### FUNDING

This work is supported by the Fundamental Research Funds for the Central Universities (Grant No. 2018CDYJSY0055, 106112017CDJQJ128837), the National Natural Science Foundation of China (Grant Nos. 61520106012, 61674023, 51775070, and 51602033), the Chongqing Research Program of Basic Research and Frontier Technology (Grant Nos. cstc2017jcyjB0127, cstc2017jcyjAX0197), and the Open Fund of the State Key Laboratory of High Field Laser Physics (Shanghai Institute of Optics and Fine Mechanics).

#### REFERENCES


### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00499/full#supplementary-material

with thick shells. Nano Res. 10, 1149–1162. doi: 10.1007/s12274-016- 1287-3


**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 Tang, Yang, Li, Chen, Hu and Qiu. 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.

# In-situ Functionalization of Metal Electrodes for Advanced Asymmetric Supercapacitors

Leimeng Sun<sup>1</sup> \*, Xinghui Wang<sup>2</sup> , Yurong Wang<sup>1</sup> , Dongyang Xiao<sup>1</sup> , Weifan Cai <sup>3</sup> , Yuan Jing<sup>3</sup> , Yanrong Wang<sup>3</sup> , Fangjing Hu<sup>1</sup> \* and Qing Zhang<sup>3</sup> \*

<sup>1</sup> MOE Key Laboratory of Fundamental Physical Quantities Measurement & Hubei Key Laboratory of Gravitation and Quantum Physics, PGMF and School of Physics, Huazhong University of Science and Technology, Wuhan, China, <sup>2</sup> College of Physics and Information Engineering, Institute of Micro-Nano Devices and Solar Cells, Fuzhou University, Fuzhou, China, <sup>3</sup> NOVITAS, Nanoelectronics Centre of Excellence, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, Singapore

#### Edited by:

Yuxin Tang, University of Macau, China

#### Reviewed by:

Xiehong Cao, Zhejiang University of Technology, China Cao Guan, National University of Singapore, Singapore Xiaolei Sun, Leibniz-Institut für Festkörper- und Werkstoffforschung (IFW Dresden), Germany Baihua Qu, Xiamen University, China

#### \*Correspondence:

Leimeng Sun sunleimeng@hust.edu.cn Fangjing Hu fangjing\_hu@hust.edu.cn Qing Zhang eqzhang@ntu.edu.sg

#### Specialty section:

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

> Received: 26 May 2019 Accepted: 03 July 2019 Published: 16 July 2019

#### Citation:

Sun L, Wang X, Wang Y, Xiao D, Cai W, Jing Y, Wang Y, Hu F and Zhang Q (2019) In-situ Functionalization of Metal Electrodes for Advanced Asymmetric Supercapacitors. Front. Chem. 7:512. doi: 10.3389/fchem.2019.00512 Nanostructured metal-based compound electrodes with excellent electrochemical activity and electrical conductivity are promising for high-performance energy storage applications. In this paper, we report an asymmetric supercapacitor based on Ti and Cu coated vertical-aligned carbon nanotube electrodes on carbon cloth. The active material is achieved by in-situ functionalization using a high-temperature annealing process. Scanning and transmission electron microscopy and Raman spectroscopy confirm the detailed nanostructures and composition of the electrodes. The TiC@VCC and CuxS@VCC electrodes show a high specific capacity of 200.89 F g−<sup>1</sup> and 228.37 F g −1 , respectively, and good capacitive characteristics at different scan speeds. The excellent performance can be attributed to a large surface area to volume ratio and high electrical conductivity of the electrodes. Furthermore, an asymmetric supercapacitor is assembled with TiC@VCC as anode and CuxS@VCC as cathode. The full device can operate within the 0–1.4 V range, and shows a maximum energy density of 9.12 Wh kg−<sup>1</sup> at a power density of 46.88 W kg−<sup>1</sup> . These findings suggest that the metal-based asymmetric electrodes have a great potential for supercapacitor applications.

Keywords: asymmetric supercapacitors, carbon nanotubes, carbon cloth, in-situ functionalization, metal electrode

### INTRODUCTION

Supercapacitors (SCs) with outstanding power densities and cycling performances have become one of the most promising power sources for next generation microelectronics and portable electronic products (Wang, 2010). However, their low energy density is one of major barriers for commercialization and practical applications of SCs. For example, the SCs based on carbon materials could hardly satisfy the energy demands for most practical applications in comparison with metal ion batteries. According to the energy (E) and capacitance (C) relationship, i.e., E = 1 2 CV<sup>2</sup> , where V stands forthe working voltage of SC, asymmetric supercapacitors (ASCs) based on novel materials are considered to be capable of improving the energy densities from two aspects. One is that the operation potential window is effectively broadened through the asymmetric design. The other one is that, by applying functional nanomaterials to electrodes to introduce the faradic capacitance, the specific capacitance of the electrodes can be significantly enhanced. As a result, ASCs based on nanostructured metal oxides have shown a significant improvement in energy density, and an operating potential of 2.0 V for ASCs have been achieved (Xiao et al., 2012). However, the metal oxides based ASCs suffer from poor conductivities and low power densities (Zhu et al., 2015). In order to solve these problems, new electroactive materials apart from transition metal oxides need to be investigated, to optimize the performance of electrodes in ASCs.

Transition metal oxides were considered as promising candidates for new energy storage material due to the introduction of pseudo capacitance that was able to significantly increase specific capacitance when compared with traditional carbon based electrodes. However, transition metal oxide are usually of poor conductivity and could not deliver high current and power densities. Transition metal carbides exhibit both fascinating energy storage performances and outstanding conductivities, when compared with transition metal oxides. Two-dimensional metal carbides have demonstrated ultrahigh specific volumetric capacitances (Lukatskaya et al., 2013). In addition, compared with metal oxides, transition metal carbides usually have better cycling performances. For example, supercapacitor based on a tubular TiC fiber nanostructured electrode was fabricated and tested for more than 150,000 cycles at a high temperature of 65◦C (Xia et al., 2015). Moreover, layered titanium carbide Ti3C<sup>2</sup> has also been proved to be a promising negative electrode material with a high mass loading of 7.6 mg cm−<sup>2</sup> and a high specific capacitance of 112 F g−<sup>1</sup> in a stable potential window (−0.9 to −0.3 V refer to Ag/AgCl electrode) (Lin and Zhang, 2015). Transition metal sulfides (e.g., MoS2, NiCo2S4, Ni3S2) based electrodes have also been extensively studied with ever improving intrinsic conductivity (Acerce et al., 2015; Fu et al., 2015; Li et al., 2015). Copper sulfides (CuxS) have been used in electrochemical devices for gas sensing application (Sagade and Sharma, 2008) and lithiumion batteries (Chung and Sohn, 2002), owing to their superior conductivity and great specific capacitance. Furthermore, acting as positive electrodes of supercapacitors, CuxS presents a metallike conductivity of ∼1 × 10<sup>3</sup> S cm−<sup>1</sup> (Mazor et al., 2009) and remarkable specific capacitance of 110 F g−<sup>1</sup> (Zhu et al., 2012). To better evaluate the capacitance of CuS based electrodes, an asymmetric supercapacitor cell, constructed with nanostructured CuS networks as the cathode and activated carbon as anode, was demonstrated with a high specific capacity of 49.8 mAh g −1 at a current density of 1 A g−<sup>1</sup> , and the maximum energy density is 17.7 Wh kg−<sup>1</sup> at a power density of 504 W kg−<sup>1</sup> (Fu et al., 2016).

The structures of electrodes are also important for optimizing the performance of supercapacitors. Recently, three-dimensional nanocarbon electrodes made from carbon nanotubes (CNTs) on carbon cloth (CC) were applied to electrochemical cells. A nickel-zinc battery based on a 3D hierarchical carbon nanofiber-CC electrode was reported to have a power density of 6.09 mWh cm−<sup>3</sup> and an energy density of 355.7 Wh kg−<sup>1</sup> (Liu et al., 2016). A lithium-ion battery with 3D carbon nanostructures as its electrodes can be consistently operated for more than 8,000 cycles (Wang et al., 2015). Furthermore, low-dimensional metal-organic frameworks (LD MOFs) have attracted increasing attention in recent years, which successfully combine the unique properties of MOFs, with the distinctive physical and chemical properties of LD nanomaterials (Xu et al., 2017, 2018; Liu et al., 2019).

Herein, we report on an asymmetric supercapacitor constructed with TiC and CuxS as its anode and cathode, respectively. TiC@VCC and CuxS@VCC are chosen as negative and positive electrode, respectively, due to the potential window of the materials. To optimize the electrochemical performance of the device, we have incorporated a verticalaligned carbon nanotube (VACNT) array on carbon cloth as the electrodes. Different from other reported CNT arrays, the VACNT array employed here is of low density and superior conductivity, and has demonstrated as promising electrochemical electrodes (Sun et al., 2015, 2016, 2017). The array with well-distributed VACNTs on carbon cloth (VCC) forms a 3D nanostructure with a large surface to volume ratio and ultra-straight morphology (Wang et al., 2016). A high-temperature annealing process is conducted to in-situ functionalize the metal-coated VCC electrodes for the anode and cathode. A large specific capacitance of 200.89 F g−<sup>1</sup> in a potential window of −0.7–0.1 V and 228.37 F g−<sup>1</sup> in −0.1– 0.7 V are obtained. Moreover, a full device based on these electrodes shows a high energy density of 9.12 Wh kg−<sup>1</sup> and power density of 46.88 W kg−<sup>1</sup> . Our findings suggest a feasible approach to achieve SCs with both high energy densities and high power densities.

### EXPERIMENTAL

### Preparation of VACNT Array on Carbon Cloth

Well-distributed VACNTs were grown on a piece of flexible carbon cloth as current collectors. Firstly, an ultrathin Ni/Al2O<sup>3</sup> bi-layer catalyst was deposited through a plasma-enhanced CVD (PECVD) system (Wang et al., 2014a) and the CNT grew in a mixture of ammonia/acetylene (240/60 sccm) gas atmosphere under 120 W plasma at 800◦C.

## Preparation of TiC@VCC and CuxS@VCC Electrodes

After the fabrication of VCC, 500 nm titanium and copper were coated on the VACNT array using a Denton RF/DC magnetron sputtering system in which a Ti target (99.999%) and a Cu target (99.99%) were sputtered at a current of 0.7 A and 0.4 A under a pure Argon atmosphere, respectively. In-situ functionalization processes for both electrodes were conducted in a thermal chemical vapor deposition (TCVD) furnace. The sulfur powder was placed in an Al2O<sup>3</sup> ceramic boat and heated in a low temperature zone and the two metal-coated VCC electrodes were

placed in a high temperature zone of 700◦C for about 30 min in a mixed gas (Ar/H2) atmosphere.

### Materials Characterization and Electrochemical Measurement

The structure and morphology of the electrodes were characterized with a SEM system (LEO 1550 Gemini) and TEM (JEM 2100 FJEOL), respectively. A Raman system (WITec) with a 532 nm wavelength excitation was applied for Raman spectrum measurements. The mass loading of these as-grown TiC or CuxS nanocomposite were 0.12 and 0.65 mg/cm<sup>2</sup> , respectively, which were determined by the difference before and after material deposition and functionalization with an analytical balance (Mettler Toledo XP 26, 0.002 mg).

Electrochemical measurements of the electrode were carried out through an electrochemical workstation (Autolab/M101) in a 1.0 M LiCl aqueous electrolyte under a three-electrode measurement setup with a standard Ag/AgCl reference electrode. For full device test, a standard CR-2032 coin cell testing system was built and tested as a whole, in which TiC@VCC and CuxS@VCC electrodes were used as the anode and cathode, respectively. A membrane was used as the separator and 1.0 M LiCl as the aqueous electrolyte.

### RESULTS AND DISCUSSION

The fabrication process of the TiC@VCC and CuxS@VCC electrodes is shown in **Figure 1**. A 20 nm nickel thin film functioning as catalyst for VACNT growth was deposited on the carbon cloth substrate through an electron beam evaporation system. The VACNT array synthesis was conducted in a plasmaenhanced chemical vapor deposition system (Wang et al., 2014b) to form 3D nanostructured VCC electrodes.

SEM images of the VCC substrate in **Figures 2a–c** show that VACNTs were well distributed on the surface of the carbon cloth. The density and uniformity of the as-prepared VACNTs are much better than randomly grown carbon nanotube networks (De Volder et al., 2013) and common carbon nanotube arrays grown via thermal CVD system (Jiang et al., 2013). In addition, large interspacings in the VACNT array in **Figure 2c** can reserve space for later accommodation of active materials.

After the fabrication of VCC electrodes, titanium and copper were deposited on these VCC substrates via sputtering. SEM

images of Ti@VCC (**Figure 2d**) and Cu@VCC (**Figure 2g**) electrodes suggest that the diameters of 1D nanostructures in both electrodes were uniformly increased, indicating that conformable coatings of titanium and copper on VCC electrodes have been achieved. Subsequently, high temperature annealing processes were conducted for Ti@VCC and Cu@VCC electrodes, as described in Section Preparation of TiC@VCC and CuxS@VCC electrodes. It is confirmed that the proposed in-situ functionalization process has successfully converted the two metal-coated VCC electrodes into TiC@VCC (Yildirim and Ciraci, 2005) and CuxS@VCC (Vas-Umnuay et al., 2015) electrodes, respectively. In **Figures 2e,h**, it is seen that the diameters of metal coated VACNTs have not been significantly changed. From zoom-in images as seen in **Figures 2f,j**, the surface morphologies of 1D nanostructures in both electrodes have been altered, especially for the CuxS@VCC electrode. It is also noticed that a notable aggregation occurred in the functionalized VACNT array during the annealing process, and part of the nanowires have stick together to form thicker bundles (Sun et al., 2015). Nevertheless, enough interspacing and porosity were reserved inside of these 3D nanostructured electrodes although the 1D nanostructured array would aggregate during the annealing step.

To obtain detailed nanostructures and materials compositions of the electrodes, transmission electron microscopy (TEM) characterizations were performed. **Figures 3a,c** reveil the coreshell nanostructure within a single TiC-coated VACNT. The inset of a high resolution TEM (HRTEM) image indicates a lattice interspacing of 2.196 Å, corresponding to the (002) planes of TiC (Xia et al., 2015). TEM image of a single CuxS-coated VACNT (**Figure 3d**) is consistent with the SEM characterization results, and the aggregation phenomenon in CuxS@VCC further escalated during the high-temperature functionalization. In this process, as-deposited Cu thin film was converted into CuxS nanoparticles attached to the VACNTs with an approximate average diameter of 100 nm. The inset of **Figure 3b** shows a dspacing of 3.127 Å, which coincides with other findings of hightemperature fabricated CuxS (Quintana-Ramirez et al., 2014; Bulakhe et al., 2016). Low-magnification images of TiC@VCC and CuxS@VCC are further provided as shown in **Figures 4C,D** to confirm the material of electrodes.

The Raman spectra of the TiC@VCC electrode and bare VCC electrodes are shown in **Figures 4A,B**, respectively. Strong characteristic peaks of 258, 430, and 620 cm−<sup>1</sup> are attributed to TiC after functionalization (Lohse et al., 2005; Xia et al., 2015). Concurrently, the intensities of the Raman disordered band (D band) and graphitic band (G band) are significantly reduced, confirming that the TiC thin film shell was conformably coated on VACNTs. The Raman spectra of CuxS@VCC electrodes illustrate a strong copper sulfide peak at 470 cm−<sup>1</sup> (**Figure 4C**), and weakened D and G band (**Figure 4D**), which are wellconsistent with other reported results (Munce et al., 2007; Quintana-Ramirez et al., 2014; Bulakhe et al., 2016).

The electrochemical characterizations of the proposed electrodes were performed in a standard three-electrode testing system with Pt plate as a counter electrode, and Ag/AgCl as

a reference electrode. TiC@VCC anode/CuxS@VCC cathode were used as the working electrodes, and LiCl as an electrolyte. For comparison purpose, the CV curves of VCC electrodes at different scan speeds from 20 to 5,000 mV s−<sup>1</sup> are first shown in **Figures 5A,D**. Correspondingly, the CV curves of the TiC@VCC anode (from −0.7–0.1 V, vs. Ag/AgCl) and CuxS@VCC cathode (from −0.1–0.7 V, vs. Ag/AgCl) are presented in **Figures 5B,E,C,F**, respectively. It is confirmed that both TiC@VCC and CuxS@VCC electrodes exhibit good capacitive behaviors at these scan speeds. **Figure 5G** further shows the real and imaginary parts of the electrochemical impedance for both electrodes, validating highly conductive properties with small impedance. The specific capacitance of TiC@VCC and CuxS@VCC electrodes can be calculated from these CV curves, yielding much higher values in comparison with VCC electrodes for a scan rate up to 200 mV/s (**Figure 5H**). It is noticed that the specific capacitance of the CuxS electrode decreases sharply after 200 mV/s, which may limit the applications of the proposed electrode at lower scan rate. This is attributed to the limited migration of electrolyte ions at higher scan rates, and some similar results have been observed in other metal sulfide based electrodes (Choudhary et al., 2015). Nevertheless, the specific capacitance is still higher than carbon only electrodes at higher scan rates. Measured CV curves of TiC@VCC and CuxS@VCC electrodes in the range of −0.7–0.7 V and scan speeds at 10 and 500 mV/s (**Figure 5I**) show a better rectangular shape than those of the VCC electrodes, suggesting dominating capacitive characteristics.

To evaluate the applicability of the proposed electrodes, a full device was assembled using the TiC@VCC as anode and the CuxS@VCC as cathode, having a working range from 0 to 1.4 V. **Figures 6A,B** illustrate the measured CV curves of the full device at scan rates from 2 to 100 mV/s. At these scan speeds, good rectangular shapes and large curve areas are obtained. The CV curves as a function of the bias voltage (from 0.7 to 1.4 V) at a scan rate of 10 mV/s (**Figure 6C**) confirm that the full device can operate at different bias voltages, and therefore, can operate effectively as a micro-supercapacitor. After the charging and discharging test of 3,000 cycles, it is found that the specific capacitance retention is from ∼80% to 110% of its original value (**Figure 6D**), which is relatively unstable when compared with other demonstrated micro-supercapacitors. This phenomenon is attributed to the side electrochemical reactions occurred during the first several hundred cycles within the

copper sulfide in positive electrode (Zhu et al., 2012; Hsu et al., 2014; Bulakhe et al., 2016). Experimental results of copper sulfide based supercapacitors suggest that the unstable cycling performance in full device was a common issue and usually only 1,000∼2,000 stable cycles of electrochemical test can be achieved. Nevertheless, all these results suggest that the full device shows an excellent electrochemical performance when compared to other recently reported full micro-supercapacitors.

### CONCLUSIONS

In this paper, we designed, fabricated and experimentally demonstrated a metal-based asymmetric supercapacitor using TiC@VCC as anode and CuxS@VCC as cathode. TiC and CuxS were prepared through in-situ functionalizations on the VCC substrate, providing a large interspacing and porosity to the 3D nanostructured electrodes. The electrodes showed a specific capacitance of 200.89 and 228.37 F g−<sup>1</sup> in the potential window of −0.7 to 0.1 V and −0.1 to 0.7V, respectively. A full device assembled from the electrodes was able to work within a potential window of 0–1.4 V at a scan speed up to 100 mV/s, and demonstrated a maximum energy density of 9.12 Wh kg−<sup>1</sup> at a power density of 46.88 W kg−<sup>1</sup> . Cycling measurements showed that the capacitance retention was between 80 and 110% of its original value. The proposed all-solid-state asymmetric supercapacitor demonstrated a high applicability and can be used as efficient energy-storage devices.

## DATA AVAILABILITY

The datasets generated for this study are available on request to the corresponding author.

## AUTHOR CONTRIBUTIONS

LS and QZ conceived the idea. LS, XW, and YuW designed and fabricated the sample, and conducted the experiment. All the authors contributed to the analysis of data and the draft of the manuscript.

## FUNDING

The project was financially supported by the National Key R&D Program of China (No. 2018YFC0603301), the National Natural Science Foundation of China (Nos. 11704071, 61801185), MOE AcRF Tier1 (MOE 2016-T1-001-197, RG102/16), MOE AcRF Tier1 (2018-T1-005-001) and MOE Tier2 (MOE2018-T2-2- 005), Singapore.

## ACKNOWLEDGMENTS

LS, XW, and YuW contributed equally to this work. The project was financially supported by the National Key R&D Program of China (No. 2018YFC0603301), the National Natural Science Foundation of China (Nos. 11704071, 61801185), MOE AcRF Tier1 (MOE 2016-T1-001-197, RG102/16), MOE AcRF

REFERENCES


Tier1 (2018-T1-005-001) and MOE Tier2 (MOE2018-T2-2- 005), Singapore.


**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 Sun, Wang, Wang, Xiao, Cai, Jing, Wang, Hu 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.

# Hollow Mesoporous Fe2O<sup>3</sup> Nanospindles/CNTs Composite: An Efficient Catalyst for High-Performance Li-O<sup>2</sup> Batteries

Hairong Xue<sup>1</sup> , Yiou Ma<sup>1</sup> , Tao Wang<sup>1</sup> \*, Hao Gong<sup>1</sup> , Bin Gao<sup>1</sup> , Xiaoli Fan<sup>1</sup> , Juanjuan Yan<sup>1</sup> , Xianguang Meng<sup>2</sup> , Songtao Zhang<sup>3</sup> and Jianping He<sup>1</sup> \*

*<sup>1</sup> Jiangsu Key Laboratory of Materials and Technology for Energy Conversion, College of Materials Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, China, <sup>2</sup> Photofunctional Materials Research Platform, College of Materials Science and Engineering, North China University of Science and Technology, Tangshan, China, <sup>3</sup> Testing Center, Yangzhou University, Yangzhou, China*

#### Edited by:

*Jianqing Zhao, Soochow University, China*

#### Reviewed by:

*Yuxin Tang, University of Macau, China Bo Jiang, National Institute for Materials Science, Japan Jianhua Zhou, Guilin University of Electronic Technology, China*

#### \*Correspondence:

*Tao Wang wangtao0729@nuaa.edu.cn Jianping He jianph@nuaa.edu.cn*

#### Specialty section:

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

> Received: *16 May 2019* Accepted: *03 July 2019* Published: *25 July 2019*

#### Citation:

*Xue H, Ma Y, Wang T, Gong H, Gao B, Fan X, Yan J, Meng X, Zhang S and He J (2019) Hollow Mesoporous Fe2O3 Nanospindles/CNTs Composite: An Efficient Catalyst for High-Performance Li-O2 Batteries. Front. Chem. 7:511. doi: 10.3389/fchem.2019.00511* The design of mesoporous or hollow transition metal oxide/carbon hybrid catalysts is very important for rechargeable Li-O<sup>2</sup> batteries. Here, spindle-like Fe2O<sup>3</sup> with hollow mesoporous structure on CNTs backbones (Fe2O3-HMNS@CNT) are prepared by a facile hydrolysis process combined with low temperature calcination. Within this hybrid structure, the hollow interior and mesoporous shell of the Fe2O<sup>3</sup> nanospindles provide high specific surface area and abundant catalytical active sites, which is also beneficial to facilitating the electrolyte infiltration and oxygen diffusion. Furthermore, the crisscrossed CNTs form a three-dimensional (3D) conductive network to accelerate and stabilize the electron transport, which leads to the decreasing internal resistance of electrode. As a cathodic catalyst for Li-O<sup>2</sup> batteries, the Fe2O3-HMNS@CNT composite exhibits high specific capacity and excellent cycling stability (more than 100 cycles).

Keywords: hollow mesoporous structure, carbon support, transition metal oxides, cathodic catalyst, Li-O2 batteries

#### INTRODUCTION

To meet the global energy demand, the development of the clean and sustainable energy storage or conversion devices is very important (Tarascon and Armand, 2011; Lu et al., 2014; Wang et al., 2017; Zhang et al., 2018; Gao et al., 2019). Rechargeable Li-O<sup>2</sup> battery has attracted wide attention as a new energy storage device, due to its high theoretical energy density (∼3,500 Wh kg−<sup>1</sup> ) (Tarascon and Armand, 2011; Lu et al., 2014). However, the practical application of Li-O<sup>2</sup> batteries still suffer a series of problems, including high overpotentials, low rate capacity and poor cycle stability, which primarily originates from its sluggish kinetics for oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) (Bruce et al., 2012; Wang et al., 2014). At the cathode (air electrode), the gradual formation of the insoluble discharge products Li2O<sup>2</sup> may block the inward oxygen diffusion of and electrolyte infiltration, which results in the rapid decline of battery performance (Zhao et al., 2015; Wang et al., 2016). To overcome these challenges, the design and development of the high-performance catalysts for oxygen-involved reactions are highly desired for the Li-O<sup>2</sup> batteries.

In recent years, a lot of the efforts have been devoted to investigate the highly active and stable catalysts for the Li-O<sup>2</sup> batteries, such as carbons, precious metals, and transition metal oxides. As the common catalysts, the inexpensive carbon materials have high surface area, nevertheless, their

**78**

limited catalytic activity for both OER and ORR restricts the battery performance of the Li-O<sup>2</sup> batteries (Girishkumar et al., 2010; Shui et al., 2013). Since the nanoporous gold (NPG) was used as a cathode catalyst, various precious metals (e.g., Ru, RuO, and Pd) have been adopted in the Li-O<sup>2</sup> batteries (Peng et al., 2012; Lu et al., 2013; Ottakam Thotiyl et al., 2013; Li et al., 2014, 2015). Although the cyclic stability can be distinctly enhanced by precious metal catalysts, the battery capacity is severely restricted because of the highly chemical formula weight. Moreover, high price of precious metals also hinders the largescale commercialization in the Li-O<sup>2</sup> batteries. Benefiting from the low cost, high stability and good catalytic performance, transition metal oxides have been proposed as the promising catalysts for the Li-O<sup>2</sup> batteries (Wang H. et al., 2012; Chen et al., 2016; Gong et al., 2016, 2018a,b; Xue et al., 2016a,c; Dai et al., 2017; Tan et al., 2017; Feng et al., 2019). Many researches have indicated that Fe-based materials possess high catalytic activities for ORR in fuel cells and OER in water electrolysis (Bates et al., 2016; Song et al., 2019). Recently, some reports began to focus on iron oxides (Fe2O3), which can serve as the cathode catalyst in Li-O<sup>2</sup> batteries (Zhang et al., 2014). These works show the enhanced electrochemical performance (e.g., higher capacity and lower overpotentials) of the Li-O<sup>2</sup> batteries, however the cycling performance still needs to further improve. Therefore, it is necessary to explore an effective approach to enhance the catalytic performance of Fe2O3-based materials.

Tailoring of the morphology is an important method for obtaining the high-performance catalysts in various electrochemical application. Mesoporous hollow architectures show high surface area and large pore volume, which offers fast electron transfer paths and facilitates the electrolyte infiltration (Kresge et al., 1992; Inagaki et al., 2002; Malgras et al., 2016). Normally, mesoporous or hollow metallic oxide are prepared through the template-based methods, using either hard templates (e.g., carbon sphere and mesoporous silica) or soft templates (e.g., surfactants and polymer) (Attard et al., 1997; Crossland et al., 2013; Liu et al., 2015; Xue et al., 2016d). For these template methods, the multi-step processes are unavoidable, and the mesoporous of hollow structures may be damaged after removing the templates. Moreover, the resultant mesoporous or hollow frameworks often show poor crystalline degree and even amorphous, which limits their electrochemical performance (Lin et al., 2015). It should be noted that the low electronic conductivity is an intrinsical characteristic of the most metallic oxides. Although most of pure carbon materials possess low catalytic activity for OER and ORR, they are identified as the good catalyst support due to their high electric conductivity and large specific surface area (Hsin et al., 2007; Stein et al., 2009; Wu et al., 2009; Xia et al., 2018). Among various carbon materials, the carbon nanotubes (CNTs) show low density, very high strength, high chemical stability, and excellent conductivity (Sathiya et al., 2011; Wang Z. et al., 2012; Ma et al., 2018). These advantages are beneficial to fabricate hybrid or composite materials in many applications by using CNTs as useful substrates. However, the fabrication of mesoporous hollow Fe2O<sup>3</sup> with high crystalline degree supported on CNTs by a simple and effective method is still an important challenge.

Inspired by the above idea, we proposed a Fe2O3/CNTs composite (denoted as Fe2O3-HMNS@CNT) prepared by using a simple hydrolysis reaction combined with heat treatment, in which spindle-like Fe2O<sup>3</sup> with hollow mesoporous structure grown on CNTs backbones. Their hollow interior and mesoporous shell with high specific surface area offer abundant catalytical active sites for OER and ORR, which also promotes the diffusion and infiltration of electrolyte. Furthermore, a great deal of the crisscrossed CNTs form the three-dimensional (3D) conductive network, which benefits the fast and stable electron transport. As a cathodic catalyst for Li-O<sup>2</sup> batteries, the Fe2O3- HMNS@CNT exhibits good battery performance, especially excellent outstanding cycling stability (100 cycles).

#### EXPERIMENTAL METHODS

#### Hollow Mesoporous Fe2O<sup>3</sup> Nanospindles on CNT Backbones

Carbon nanotubes (CNTs, 30–60 nm in diameter and 5–15µm in length) were purchased from Shenzhen Nanotech Port Co. Ltd (Shenzhen, China). CNTs were refluxed in HNO<sup>3</sup> and then rinsed with distilled water. After drying, the obtained CNTs (10 mg) was dispersed in FeCl<sup>3</sup> solution (20 mL, 0.12 M) under sonication for 1 h. Then, the suspension was heated at 75◦C for 6 h in an oil bath with stirring. After several rinsing combined with sonication, the products were dried at 60◦C. Finally, the above products were annealed at 400◦C for 4 h in air by using a slow heating rate (0.5◦C min−<sup>1</sup> ) to form hollow mesoporous Fe2O<sup>3</sup> nanospindles on CNT backbones.

#### Materials Characterization

The X-ray diffraction (XRD, Bruker D8 advance) with Cu Ka radiation (λ = 1.5406 Å) is used to analyze the crystal structure of the sample. The N<sup>2</sup> adsorption–desorption measurements conducted on a ASAP-2010 analyzer to investigate the pore structure. The Brunauer-Emmet-Teller (BET) method is used to calculate the specific surface area. The morphology and microstructure are observed by using field-emission scanning electron microscope (FE-SEM, Hitachi S-4800) and highresolution transmission electron microscope (HR-TEM, JEOL JEM-2100), respectively.

#### Electrochemical Measurement

The electrochemical measurements are tested under twoelectrode system at room temperature in the pure O<sup>2</sup> atmosphere, using Li plate as the counter and reference electrode. The work electrode is prepared by mixing Fe2O3-HMNS@CNT catalyst (50 wt %), Super P (45 wt %) and polytetrafluoroethylene (PTFE) binder (5 wt %), followed by vacuum drying at 100◦C for 12 h. The capacity and current density of the sample is calculated by the whole electrode' weight. For the assembling of Li-O<sup>2</sup> battery, the working electrode and Li plate are separated by a glassfiber separator in coin cell. The tetraethylene glycol dimethyl ether (TEGDME) containing lithium bis(tri-fluoromethanesulfonyl)imide (LiTFSI) (1 M) is used as electrolyte. The Land 2100 Charge/Discharge instruments are carried to test the electrochemical measurements.

### RESULTS AND DISCUSSION

**Figure 1** illustrates the fabrication of the mesoporous hollow Fe2O<sup>3</sup> nanocrystals covered on CNT by a simple hydrolysis reaction combined with heat treatment. In order to facilitate the nucleation and anchoring of the nanocrystals on the CNT surface, these CNTs are functionally modified with carboxylic or hydroxyl groups under acidic reflux. During the subsequent stirring process, the electropositive Fe3<sup>+</sup> ions can preferential adsorb the electronegative oxygen-containing groups on CNTs by the electrostatic attraction. Due to the hydrolysis of Fe3<sup>+</sup> ions followed by the olation/oxolation of the FeO<sup>6</sup> units, the formed spindle-like β-FeOOH nanocrystals can be spontaneously grown on CNT backbones, which avoids the addition of any structuredirecting agent. The TG/DSC is used to explore the reactions in the final annealing step (**Figure S1**). The TG curve show two mainly weight loss processes at 300◦C and 300∼400◦C during the annealing step, which are attributed to the conversion from FeOOH into Fe2O3. The slowly intramolecular dehydration results in the slow weight loss before 300◦C. After 300◦C, a distinct weight loss can be found, which is originated from the fast removal of the H2O molecules. After annealing treatment in air, the spindle-like β-FeOOH nanocrystals are converted to hollow mesoporous α-Fe2O<sup>3</sup> nanospindles on CNT backbones (denoted as Fe2O3-HMNS@CNT) by the thermal dehydroxylation together with the lattice shrinkage.

FIGURE 2 | (A) XRD patterns, (B) N2 adsorption–desorption isotherms of the Fe2O3-HMNS@CNT composite.

The crystalline structure of the Fe2O3-HMNS@CNT composite is confirmed by XRD analysis. As shown in **Figure 2A**, the XRD pattern exhibits several intensively diffraction peaks at the 2θ values of about 24.2, 33.2, 35.6, 39.3, and 40.9◦ , which are indexed to the (012), (104), (110), (006), and (113) facets of the hexagonal α-Fe2O3. This result of XRD is in accord with the standard card of α-Fe2O<sup>3</sup> (JCPDS no. 33-0664). The pore structure of the Fe2O3-HMNS@CNT composite is analyzed by N<sup>2</sup> adsorption–desorption isotherms (**Figure 2B**). It can be seen that there is a type-IV curve of the Fe2O3-HMNS@CNT with a distinctly hysteresis loop on N<sup>2</sup> adsorption–desorption isotherms, which indicates the typically mesoporous structure (Deng et al., 2007; Xue et al., 2016b). Moreover, the nitrogen uptake is found from 0.40 to 0.70 (P/Po), which can be attributed to the mesoporous materials' capillary condensation of nitrogen. Based on N<sup>2</sup> adsorption–desorption isotherms, the specific surface area is calculated to be 97 m<sup>2</sup> g −1 . The above results indicate that the Fe2O3-HMNS@CNT has a typically mesoporous structure with high specific surface area.

The morphologies and structures of the Fe2O3-HMNS@CNT composite are observed by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). As shown in **Figure S2**, the CNT shows the typical one-dimensional tubular structure with a diameter of ∼40 nm. **Figure 3a** reveals a panoramic view of the sample, in which the spindle-like nanocrystals grown on the entire surface of the sinuous CNTs

backbones. It can be noted that there are many obvious porous within the Fe2O<sup>3</sup> nanospindles, as shown in **Figures 3b,c**. TEM observation is used to further investigate the internal structure of the sample. The spindle-like Fe2O<sup>3</sup> nanocrystals have a diameter of around 80 nm and a length of about 250 nm (**Figure 3d**), which show a well-developed hollow interior (marked by orange dotted line) and a typically mesoporous shell (marked by purple dotted line) (**Figure 3e**). During the calcining process, the decomposition-oxidation of the FeOOH can release abundant H2O and gases, which leads to the formation of mesoporous structure. On the other hand, the density of FeOOH (3 g cm−<sup>3</sup> ) is lower than that of the hematite (Fe2O3, 5.3 g cm−<sup>3</sup> ), so some internal mesoporous slowly forms the larger porous to maintain the spindle-like structure during the lattice shrinkage process, thus leading to the formation of the hollow interior together with the mesoporous shell. As shown in **Figure 3f**, some clear lattice

fringes with an interplanar spacing of 0.251 nm in a single Fe2O<sup>3</sup> nanospindle, corresponding to the (110) planes of the hexagonal α-Fe2O3, which confirms the formation of the α-Fe2O<sup>3</sup> with high crystallinity.

Inspired by the superiorities of material and structure, the Fe2O3-HMNS@CNT composite is investigated as the cathodic catalyst for rechargeable Li-O<sup>2</sup> batteries. **Figure 4A** shows the discharge and charge curves of the sample, which is tested under O<sup>2</sup> atmosphere with a current density of 200 mA g−<sup>1</sup> . The Fe2O3- HMNS@CNT-based cathode exhibits a large discharge capacity of 7,730 mAh g−<sup>1</sup> and a high discharge plateau (∼2.70 V) with the low overpotential for discharging (0.26 V). Owing to the low overpotential (1.10 V) for the charge process, the coulombic efficiency of the sample is closed to 100%. The reversible formation/decomposition of the discharging product Li2O<sup>2</sup> can be confirmed by SEM (inset of **Figure 4A**). After fully discharging (defined as II), it is seen that a mass of the Li2O<sup>2</sup> particles with the typically toroid-like shape uniformly cover on the electrode surface, which is observably different from the fresh electrode (defined as I). When the electrode is fully charged, the formed Li2O<sup>2</sup> particles is rarely observed (defined as III), indicating high reversibility of the Fe2O3-HMNS@CNTbased cathode. We further evaluate the cycling performance of the Fe2O3-HMNS@CNT-based cathode, tesed under the limited capacity (1,000 mAh g−<sup>1</sup> ). It can be found that the sample has the small discharge and charge overpotentials of 0.24 and 0.93 V in the first cycle, respectively, which implies the high catalytic activity for both OER and ORR (**Figure 4B**). After 100 cycles, no distinct change of the specific capacities of the sample is observed in discharging/charging curves. Moreover, the discharging and charging terminal voltages can still maintain in 2.6 and 4.3 V. These results exhibit the excellent cycling stability of the Fe2O3- HMNS@CNT-based cathode (**Figure 4C**). As a reference, the cycling stability of the Fe2O3-HMNS@CNT-based cathode is much better than those of previously reported non-precious metal/metal oxide-based materials (**Table S1**).

As the cathodic catalyst for rechargeable Li-O<sup>2</sup> batteries, the as-prepared Fe2O3-HMNS@CNT composite exhibits the excellent cycling stability, which not only benefits by the intrinsic material characteristics but also dependents on the morphology and structure. On the one hand, these largely separated Fe2O<sup>3</sup> nanospindles directly grown on the CNT backbones, which makes them accessible to the electrolyte. Different from the conventional nanoparticle catalysts, the adverse agglomeration of the Fe2O3-HMNS@CNT can be effectively decreased, avoiding the elimination of the active interfaces. The Fe2O<sup>3</sup> with a hollow interior and a mesoporous shell offers high specific surface area and a mass of catalytical active sites, which also facilitates the diffusion and infiltration of electrolyte. On the other hand, the 3D conductive network composed of the crisscrossed CNTs ensures

#### REFERENCES

Attard, G., Bartlett, P., Coleman, N., Elliott, J., Owen, J., and Wang, J. (1997). Mesoporous platinum films from lyotropic liquid crystalline phases. Science. 278, 838–840. doi: 10.1126/science.278. 5339.838

the fast and stable electron transport, leading to the lower internal resistance of electrode.

### CONCLUSION

In summary, we successfully fabricated hollow mesoporous Fe2O<sup>3</sup> nanospindles on CNTs (Fe2O3-HMNS@CNT) though a simple hydrolysis reaction followed by a heat treatment, which are served as cathodic catalyst for Li-O<sup>2</sup> batteries. In this catalyst design concept, the spindle-like Fe2O<sup>3</sup> nanocrystals possess the hollow interior and mesoporous shell, which not only provides high specific surface area and abundant catalytical active sites but also facilitates the diffusion and infiltration of electrolyte. Moreover, the 3D conductive network formed by the crisscrossed CNTs ensures the fast and stable electron transport, reducing internal resistance of electrode. Benefiting from the intrinsic material characteristics and structural superiorities, the Fe2O3- HMNS@CNT catalyst shows high specific capacity and excellent cyclic stability.

#### DATA AVAILABILITY

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

### AUTHOR CONTRIBUTIONS

HX, TW, and JH conceived this research work. HX performed the experiments and wrote the manuscript. YM, HX, HG, BG, XF, and JY tested the electrochemical performance. XM and SZ contributed to analyze the experimental results. All authors read and approved the final manuscript.

#### FUNDING

This work was financially supported by the Zhejiang Provincial Natural Science Foundation of China under Grant no. LQ18B010005, National Natural Science Foundation of China (51602153, 11575084, and 21703065), the Natural Science Foundation of Jiangsu Province (BK20160795), the Fundamental Research Funds for the Central Universities (no. NE2018104), and a project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2019.00511/full#supplementary-material


carbon for superior lithium-ion batteries. Chem. Eur. J. 22, 4915–4923. doi: 10.1002/chem.201504420


**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 Xue, Ma, Wang, Gong, Gao, Fan, Yan, Meng, Zhang and He. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Free-Standing SnO2@rGO Anode via the Anti-solvent-assisted Precipitation for Superior Lithium Storage Performance

Shuli Jiang1,2, Ruiming Huang<sup>3</sup> \*, Wenchang Zhu1,2, Xiangyi Li 1,2, Yue Zhao1,2 , Zhixiang Gao1,2, Lijun Gao1,2 \* and Jianqing Zhao1,2 \*

*<sup>1</sup> College of Energy, Soochow Institute for Energy and Materials Innovations, Soochow University, Suzhou, China, <sup>2</sup> Key Laboratory of Advanced Carbon Materials and Wearable Energy Technologies of Jiangsu Province, Soochow University, Suzhou, China, <sup>3</sup> Department of Chemistry, Rutgers-Newark, The State University of New Jersey, Newark, NJ, United States*

#### Edited by:

*Vito Di Noto, University of Padova, Italy*

#### Reviewed by:

*Abhishek Lahiri, Clausthal University of Technology, Germany Xiaosheng Tang, Chongqing University, China*

#### \*Correspondence:

*Ruiming Huang ruiming407@gmail.com Lijun Gao gaolijun@suda.edu.cn Jianqing Zhao jqzhao@suda.edu.cn*

#### Specialty section:

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

Received: *06 September 2019* Accepted: *05 December 2019* Published: *19 December 2019*

#### Citation:

*Jiang S, Huang R, Zhu W, Li X, Zhao Y, Gao Z, Gao L and Zhao J (2019) Free-Standing SnO2@rGO Anode via the Anti-solvent-assisted Precipitation for Superior Lithium Storage Performance. Front. Chem. 7:878. doi: 10.3389/fchem.2019.00878* Metal oxides have been attractive as high-capacity anode materials for lithium-ion batteries. However, oxide anodes encounter drastic volumetric changes during lithium ion storage through the conversion reaction and alloying/dealloying processes, leading to rapid capacity decay and poor cycling stability. Here, we report a free-standing SnO2@reduced graphene oxide (SnO2@rGO) composite anode, in which SnO<sup>2</sup> nanoparticles are tightly wrapped within wrinkled rGO sheets. The SnO2@rGO sheet is assembled in high porosity via an anti-solvent-assisted precipitation of dispersed SnO<sup>2</sup> nanoparticles and graphene oxide sheets in the distilled water, followed by the filtration and post-annealing processes. Significantly enhanced lithium storage performance has been obtained of the SnO2@rGO anode compared with the bare SnO<sup>2</sup> anode material. A high charge capacity above 700 mAh g−<sup>1</sup> can be achieved with a satisfying 95.6% retention after 50 cycles at a current density of 500 mA g−<sup>1</sup> , superior to reserved 126 mAh g <sup>−</sup><sup>1</sup> and a much lower 16.8% retention of the bare SnO<sup>2</sup> anode. XRD pattern and HRTEM images of the cycled SnO2@rGO anode material verify the expected oxidation of Sn to SnO<sup>2</sup> at the fully-charged state in the 50th cycle. In addition, FESEM and TEM images reveal the well-preserved free-standing structure after cycling, which accounts for high reversible capacity and excellent cycling stability of such a SnO2@rGO anode. This work provides a promising SnO2-based anode for high-capacity lithium-ion batteries, together with an effective fabrication adoptable to prepare different free-standing composite materials for device applications.

Keywords: SnO2, rGO, anti-solvent-assisted precipitation, free-standing anode, lithium-ion battery

### INTRODUCTION

With the rapid development of portable electronic devices, pure electric vehicles and emerging large-scale energy storage systems, lithium-ion batteries are required to at least have high energy and power densities, in order to meet high-grade demands for various practical applications. The exploration of alternative anode materials has become an urgent task to pursue high lithium storage capacity, together with excellent rate capability and cycling stability, because the specific capacity of

**85**

Jiang et al. Free-Standing SnO2@rGO Anode

the commercial graphite anode has been reached to the theoretical limit of 372 mAh g−<sup>1</sup> . The tin-based oxides have been widely reported as promising anode candidates, due to the high capacity, non-toxicity, and natural abundance (Hu et al., 2017; Sahoo and Ramaprabhu, 2018; Cao et al., 2019; Hong et al., 2019). As reported in the literature (Zhao et al., 2016; Ahmed et al., 2017; Cui et al., 2017), lithium storage capacity of the SnO<sup>2</sup> anode material is on the basis of reversible alloying/dealloying processes of SnxLi (0 < x ≤ 4.4, corresponding to the maximum theoretical capacity of 782 mAh g <sup>−</sup><sup>1</sup> when the x = 4.4) after an initial irreversible conversion reaction from original SnO<sup>2</sup> to the metallic Sn (Wang et al., 2012). An impressive capacity up to 1,493 mAh g−<sup>1</sup> based on x = 8.4 has been achieved by Wang and co-authors (Wang et al., 2015) through realizing the fully reversible oxidation from the reduced Sn back to SnO intermediate then to SnO2. In addition, increased capacities can also be obtained via a partial Sn to SnO<sup>2</sup> conversion, coupled with synergistic effects from different carbonaceous materials or functional nanomaterials (Kim et al., 2014, 2017; Sun et al., 2015). However, the practical application of SnO2-based anode materials is impeded by severe volumetric expansion/contraction up to 259% during alloying/dealloying processes of SnxLi variants, leading to the structural degradation and poor electronic conductivity of the anode. Additionally, the undesirable aggregation of reduced Sn nanoparticles into clusters together with the Sn pulverization occurs during prolonged electrochemical cycling, which further brings about the deactivation of active Sn particles, and thus the rapid capacity loss and poor cycling stability (Li et al., 2015; Liu et al., 2015; Min et al., 2019).

Compared with extensively-reported strategies, such as reducing particles size (Park et al., 2007; Wang et al., 2015; Xia et al., 2016; Yao et al., 2019) and dispersing active anode materials into a solid matrix (Xu et al., 2012; Zheng et al., 2016), few attention has been devoted to the binders in the anode (Zhang et al., 2014). The traditional binder, i.e., polyvinylidene fluoride (PVDF) is insulating and electrochemically inactive, which is used to strengthen mechanical connections between active anode materials, conductive additives, and the current collector. However, the presence of the binder decreases the overall electronic conductivity, but increases the electrochemical polarization in the anode. As reported in the literature (Kumar et al., 2019; Pan et al., 2019), polymetric conductive binders with strong mechanical binding force and even self-healing capability have been demonstrated to address detrimental volume effects of oxide/metal-based anode materials. Moreover, such polymetric binders play an additional role in offering the desired pathway for the charge transfer, resulting in free conductive carbon additives in the anode. However, the synthesis of those conductive polymers is expensive and timeconsuming, which requires complex coupling reactions, noble metal catalysts and stringent reaction conditions. In addition to the binders, the metallic current collector, i.e., the copper foil is also needed for the anode fabrication. Within the typical anode, inactive components involving the binder, conductive additive and current collector exceed 50 wt.% of the total electrode mass. Therefore, an advanced electrode structure should be rationally designed to increase the energy density of the anode.

The graphene and reduced graphene oxide (rGO) have been extensively used for energy storage and conversion applications, especially in lithium-ion batteries, owing to their unique two-dimensional structures with excellent flexibility, mechanical strength, chemical stability, and thermal and electronic conductivities (Rong et al., 2014; Deng et al., 2016; Ahn et al., 2019; Riyanto et al., 2019). Both graphene and rGO materials have been demonstrated to act as reliable supporting and buffering matrixes to improve electrochemical performance of high-capacity anode materials, such as SnO<sup>2</sup> and Si, by accommodating their drastic volume changes during the lithium storage (Jiang et al., 2017; Ma et al., 2017; Chen et al., 2018; Deng et al., 2019). Tri-dimensional hybrid materials consisting of graphene (or rGO) sheets and active anode particles can be served as promising free-standing anodes with free conductive additives and binders. The wrinkled structure of graphene or rGO sheets may be also fabricated, in order to ensure the "buffering" capability. On the other hand, active particles are required to distribute in the graphene-based matrix uniformly (Li et al., 2015; Wang et al., 2018). As reported in the literature (Li et al., 2011, 2019), either aerosol spray drying process or solution ionic strength engineering has been demonstrated as an effective route to obtain desired composite materials, but the conductive additive and polymer binder are still added for the electrode preparation. It would be very interesting to explore assembly methods for the preparation of free-standing graphene-based anodes full of pores and wrinkles, in which active particles are uniformly distributed free of the conductive carbon and binder components, resulting in the maximum capacity contribution of such the anode (Xia et al., 2019; Xing et al., 2020).

In this work, we report an effective approach to fabricate a free-standing SnO2@rGO composite anode through an antisolvent-assisted precipitation followed by the suction filtration. The resulting SnO2@rGO anode has sufficient wrinkles and internal channels, which are expected to favor not only the electrolyte permeation but also the accommodation of large volume expansion during cyclic lithium storage of the SnO<sup>2</sup> anode material, and thus contribute to enhanced electrochemical performance compared with the bare SnO<sup>2</sup> anode material. The assembly method developed in this study may be adopted to prepare different free-standing composite materials consisting of a flexible matrix and functional nanoparticles for device applications beyond lithium-ion batteries.

#### EXPERIMENTAL

### Preparation of the Graphene Oxide (GO) and Free-Standing SnO2@rGO Electrode

Graphene oxide was prepared via a modified Hammond method (Marcano et al., 2010). Typically, 0.75 g graphite flakes and 4.5 g KMnO<sup>4</sup> were added to a 100 mL concentrated H2SO<sup>4</sup> and H3PO<sup>4</sup> solution at a volume ratio of 90:10. After heating at 50◦C for 12 h under continuous stirring, the mixed solution was cooled to room temperature and poured onto a 100 mL ice with 1 mL

(2)

30 wt.% H2O2. The obtained slurry was then centrifuged and washed repeatedly with 10% HCl to remove Mn2<sup>+</sup> ion, followed by removing the majority of Cl<sup>−</sup> ion via the successive washing using the acetone. For the complete removal of all the ions, deionized water was used to wash GO until no precipitate was observed when the GO solution was mixed the 10 mM AgNO<sup>3</sup> solution.

In order to prepare the free-standing SnO2@rGO electrode, 4 mg SnO<sup>2</sup> (Sigma Aldrich) and 8.8 mg GO were co-dispersed in 15 mL de-ionized water for 15 min. Twenty milligrams LiCl powder was directly added. The suspension was then sonicated for an additional 1 min. Sixty milliliters acetone was poured into the above suspension at one time shot. The resulted mixture was then collected via vacuum filtration using the PTFE filter paper with the pore size of 0.2µm and the diameter of 15 mm (Sterlitech). The obtained free-standing SnO2@GO sheet on the PTFE paper was then together dried in vacuum at 60◦C for 2 h. After peeling off from the PTFE paper, the dried SnO2@GO sheet was subjected to post-annealing processes to reduce the GO and obtain SnO2@rGO. The heating temperature was increased

$$\begin{aligned} \text{2Li}^+ + \text{SnO}\_2 + 4\text{e}^- &\leftrightarrow 2Li\_2\text{O} + \text{Sn} \quad \text{(capacity delivered: 711 mA/g)}\\ \text{xLi}^+ + \text{Sn} + \text{x}\text{e}^- &\leftrightarrow Li\_2\text{Sn} \ (0 \le \text{x} \le 4.4) \quad \text{(capacity delivered: 782 mA/g)} \end{aligned}$$

100◦C per step with a duration time of 1 h up to 400◦C, followed by heating at 500◦C for 3 h in the Ar flow. The typical mass of asprepared SnO2@rGO electrode is 5.6 mg, corresponding to the loading density of 3.1 mg/cm<sup>2</sup>

### Material Characterizations

Crystallographic structures of as-prepared materials were identified by X-ray diffraction (XRD) on a Bruker D8 Advance automatic diffractometer with Cu Kα radiation. Morphology and structure of different samples were observed by using scanning electron microscopy (SEM, Hitachi S-4800) and transmission electron microscopy (TEM, FEI Tecnai G2T20) at an acceleration voltage of 200 kV, respectively. The chemical environment and valent states of anions and cations within different materials were characterized by X-ray photoelectron spectroscopic (XPS) measurements on an ESCALAB 250Xi XPS equipment. All XPS spectra were calibrated according to the binding energy of the C 1s peak at 284.8 eV. The degree of graphitization of the GO and rGO materials were characterized by Raman spectra on a Horiba JY LabRAM Aramis equipment. The rGO content within the SnO2@rGO electrode was determined by the thermogravimetric analysis on a TG/DTA-7300 thermal analyzer (Seko) in air flow at a temperature range between room temperature and 900◦C.

#### Electrochemical Measurements

Electrochemical measurements were carried out in a twoelectrode system for lithium-ion battery testing. The freestanding SnO2@rGO sheet was directly used as the working anode. All CR2025-type coin cells were assembled in an Ar-filled glove box, using the lithium metal as the counter and reference electrode, and Celgard3501 as the separator. The electrolyte was 1 M LiPF<sup>6</sup> dissolved in a mixture of ethylene carbonate (EC), dimethyl carbonate (DMC) and diethyl carbonate (DEC) at a volumetric ratio of 4:3:3. Galvanostatic charge/discharge of the cells were performed on the MTI BST8-MA-battery analyzer in a voltage range of 0.01–3.0 V vs. Li+/Li. Cyclic voltammetric (CV) curves were recorded at a scanning rate of 0.1 mV s−<sup>1</sup> between 0.01 and 3 V vs. Li+/Li, and electrochemical impedance spectroscopy (EIS) was conducted from the open circuit voltage of testing cells in a frequency range of 10 mHZ-100 kHZ with an AC amplitude of 5 mV using an electrochemical analyzer (CHI 760C).

### RESULTS AND DISCUSSION

The SnO<sup>2</sup> oxide has been extensively reported as a highcapacity anode material for lithium-ion batteries, but suffers from fast capacity decay during cycling, due to drastic volume changes for the lithium storage on the basis of the reversible initial conversion reaction (Equation 1) and successive alloying/dealloying processes (Equation 2) as follows (Huang et al., 2010; Wang et al., 2011; Zhang et al., 2012; Jiang et al., 2017):

$$\text{H} + 4e^- \leftrightarrow 2\text{Li}\_2\text{O} + \text{Sn} \quad \text{(capacity delivered: 711 mA/g)}\tag{1}$$

$$\mathbf{x} \succeq \mathbf{x} \mathbf{:} \mathbf{r} \mathbf{i} \quad \text{(capacuty duenverteu. } \lor \mathbf{o} \mathbf{z} \text{ mizu/g)}$$
 
$$\text{mon} \qquad \text{However it is difficult to maintain the } \mathbf{z} \text{-} \text{ part of } \mathbf{x} \text{ (b.)}$$

However, it is difficult to maintain the full reversibility of the conversion reaction as shown in Equation (1), resulting in a capacity range of 782–1,493 mAh/g for the SnO<sup>2</sup> anode material (Wang et al., 2013; Deng et al., 2016; Cao et al., 2019). In order to obtain the maximum capacity with excellent cycling stability, a carbon-based framework should be employed to support active SnO<sup>2</sup> nanoparticles by serving as a conductive network and a structural cushion to release mechanical strains during lithiation/delithiation of such the oxide anode. In addition, the uniform dispersion of SnO<sup>2</sup> nanoparticles in the carbonaceous framework is also highly required to restrict from the unfavorable aggregation and pulverization of the reduced Sn nanoparticles. The as-prepared GO nanosheets are here used to accommodate commercial SnO<sup>2</sup> nanoparticles, and a free-standing SnO2@rGO anode is obtained through an effective anti-solvent-assisted precipitation followed by a post-annealing process. **Figure 1** shows schematic illustrations for the synthesis of the freestanding SnO2@rGO sheet, together with photographs taken in different stages. Notably, an anti-solvent-assisted precipitation, which is inspired by a frequently used method to purify and/or concentrate RNA and DNA in biochemistry (Zeugin and Hartley, 1985), is adopted to assemble the SnO2@GO composite material. The synergy of ionic strength and dielectric constant change induced by a low concentrated salt and an organic solvent (such as ethanol and acetone) allows for the controllable assembly of well-dispersed graphene oxide (GO) nanosheets and SnO<sup>2</sup> nanoparticles in an aqueous solution. SEM images as shown at **Stage I** reveal distinct aggregations of both original GO nanosheets and SnO<sup>2</sup> nanoparticles. The brown solution-like suspension can be obtained through simply dispersing GO and SnO<sup>2</sup> powders in the deionized (DI) water by the ultrasonic treatment at **Stage II**, on account of numerous hydrophilic groups involving -COOH and -OH ends at the surface of GO

sheets and nano-sized SnO<sup>2</sup> particles. The assembly of GO sheets and SnO<sup>2</sup> nanoparticles can be implemented at **Stage III** by adding a small amount of LiCl salt and subsequently the acetone as an anti-solvent, resulting in the rapid precipitation as shown in the vial. The instant formation of SnO2@GO precipitant is attributed to the considerably strengthened electrostatic force between negatively-charged GO sheets and positively-charged Li<sup>+</sup> ion induced by the anti-solvent acetone. Thus, the stable GO dispersion is disturbed, and GO sheets crumple and fold to minimize the surface energy. With the co-existence of well-distributed SnO<sup>2</sup> nanoparticles in the GO solution, SnO<sup>2</sup> nanoparticles are all wrapped in wrinkled GO nanosheets, resulting in the SnO2@GO precipitation. As shown at **Stage IV**, the free-standing SnO2@rGO sheet is obtained via the facile vacuum filtration of all SnO2@GO precipitants, followed by a post-annealing process.

Morphologic and structural characteristics of resulting SnO2@rGO composite material are examined in **Figure 2**. **Figures 2a,b** show SEM images taken at the intersection area (the side view) and the surface (the top view) of the freestanding SnO2@rGO sheet, respectively, indicating a distinct porous structure. The zoom-in FESEM image as shown in **Figure 2c** clarifies that all SnO<sup>2</sup> nanoparticles are tightly wrapped within rGO sheets full of expected wrinkles, which is very similar to the assembled structure of original SnO2@GO material (**Figure 2d**). The post-annealing process only results in the reduction of GO component to rGO material, while encased SnO<sup>2</sup> nanoparticles are stabilized without the particle growth and phase reduction to either SnO or Sn (will be discussed in **Figure 3**). As captured in TEM observation (**Figure 2e**), SnO<sup>2</sup> nanoparticles all show spherical shapes in a particle size range of 10–50 nm. The HRTEM image of the SnO2@rGO material is also captured as displayed in **Figure 2f**, coupled with corresponding selected area electron diffraction (SAED) pattern as shown in **Figure 2f**I , indicating the single crystal property of SnO<sup>2</sup> particles and a high graphitization degree of the rGO sheet. Lattice fringes of two selected SnO<sup>2</sup> particles (squared in red and pink dashed lines in **Figure 2f**) with the corresponding d-space distance of 0.26 and 0.33 nm can be indexed to (101) and (110) planes of the SnO<sup>2</sup> in a tetragonal rutile structure, as shown in **Figures 2f**II**, f**III, respectively. Additionally, **Figure 2f**IV shows the enlarged selected lattice fringe of the rGO sheet (squared in the navy dashed line in **Figure 2f**), corresponding to (004) planes of the hexagonal graphite phase with an interplanar distance of 0.17 nm. Overall, SEM and TEM observations demonstrate the desired structural integrity of the free-standing SnO2@rGO sheet with the high porosity.

**Figure 3** further analyzes the composition, crystallographic structure and chemical environment of the SnO2@rGO composite material. The rGO content within the SnO2@rGO is determined to be 33.1 wt.%, according to the TG analysis as shown in **Figure 3a**. However, XRD pattern of the SnO2@rGO is identical to that of the pure SnO<sup>2</sup> powder (**Figure 3b**), which can be well indexed to the rutile SnO<sup>2</sup> phase in the tetragonal structure (PDF: 41–1445). The absence of characteristic diffraction peaks from the rGO component implies the homogeneous dispersion of SnO<sup>2</sup> nanoparticles between each rGO layer, in accordance with SEM and TEM observations (**Figures 2c,e**). By contrast, the typical (002) peak can be detected at 2θ = 12<sup>0</sup> in the XRD pattern of GO powder, indicating the ordered stack of original GO sheets. The other broad peak located at 2θ = 23<sup>0</sup> is resulted from the partial reduction of the GO material, possibly owing to the drying process during the material collection. Accordingly, the Raman spectrum of rGO sheets shows an intensity ratio of the D band over G band, i.e., the ID/I<sup>G</sup> equal to 1.06, also indicating a certain degree of graphitization, because the G band peak located near 1,589 cm−<sup>1</sup> is related to the vibration of sp<sup>2</sup> -bonded carbon atoms in an ordered two-dimensional hexagonal lattice of carbon-based materials, while the D band peak around 1,339 cm−<sup>1</sup> is associated with defects and disorder formed in the hexagonal graphitic layers (Wang et al., 2012). However, it is interesting to see that the rGO component in the SnO2@rGO composite material gives a higher ratio of ID/I<sup>G</sup> = 1.14, although it was subjected to the post-annealing process for reducing defects in original GO sheets. The increased disordered domains in the rGO component are probably caused by wrapped SnO<sup>2</sup> nanoparticles through the possible chemical bonding; hence, XPS spectra were carried out on the SnO2@rGO powder to study chemical environments of Sn, O, and C elements. Characteristic C 1s, Sn 3d, and O 1s XPS peaks are plotted in **Figures 3d–f**, respectively. According to fitting patterns, the C 1s peak can be deconvoluted to one dominant C=C contribution at 284.7 eV, together with two weak effects from the C-O at 285.2 eV and C=O at 286.5 eV (**Figure 3d**), indicating a typical chemical environment of the rGO component as reported in the literature (Min et al., 2019).

It is worth noting that two Sn 3d XPS peaks in **Figure 3e** shift to the higher binding energies for 0.8 eV compared with the reported pure SnO<sup>2</sup> powder (Yao et al., 2019). Furthermore, different contributions of O-Sn bond at 531.3 eV, Sn-O-C bond at 532.4 eV and O-O bond at 533.3 eV can be identified in the O 1s peak on the basis of fitting patterns as shown in **Figure 3f**, respectively. Thus, Raman and XPS spectra together indicate the favorable chemical bonding formed between SnO<sup>2</sup> and rGO components within the SnO2@rGO composite material.

The resulting SnO2@rGO sheet (**Figure 1IV**) is directly used as the free-standing anode for lithium storage performance evaluations without adding the conductive carbon, polymer binder and even the copper current collector. For the comparative propose, the bare rGO film is also fabricated as the other free-standing anode, and the bare SnO2-based anode is composed of 70 wt.% SnO<sup>2</sup> nanoparticles as the active material, 20 wt.% acetylene black as the conductive carbon and 10 wt.% polyvinylidene fluoride (PVDF) as the binder. **Figures 4a–c** show cyclic voltammetric (CV) curves of the SnO2@rGO, SnO<sup>2</sup> and bare rGO anodes in the first five cycles recorded at a scanning rate of 0.1 mV s−<sup>1</sup> in a voltage range of 0.01–3.0 V vs. Li+/Li. As referred to CV characteristics of bare rGO and SnO<sup>2</sup> anode materials, the initial cathodic peak at 0.74 V in the first CV discharge of the SnO2@rGO anode is attributed to the conversion reaction of SnO<sup>2</sup> to the metallic Sn and Li2O, as described in Equation (1). The subsequent broad cathodic peak below 0.5 V reveals combined effects from the successive formation of LixSn (0 < x < 4.4) alloys (Equation 2), lithiation of the rGO material (**Figure 4b**) and the formation of solid electrolyte interphase (SEI) film at the surface of the working electrode. Correspondingly, a weak anodic shoulder peak at 0.2 V in the following charge process can be probably assigned to the

reversible delithiation from the rGO component. The dominant anodic peak positioned at 0.55 V corresponds to the dealloying process of as-formed LixSn compounds. It is interesting to find a very wide anodic park cross a voltage range between 1.0 and 2.5 V. As reported in the literature (Chen and Yano, 2013; Kim et al., 2014; Hong et al., 2019), such electrochemical CV behaviors result from the reversible reaction from the reduced Sn to the SnO intermediate near 1.3 V and then to the SnO<sup>2</sup> near 1.9 V. Accordingly, a wide anodic peak appears in a voltage range of 0.75–1.5 V in the second CV cycle, which supports our speculation of the reversible conversion reaction between SnO<sup>2</sup> and Sn as shown in Equation (1). Two following cathodic peaks located at 0.2 and 0.01 V are ascribed to the alloying process of LixSn and lithium ion storage of the rGO, respectively, together with the continuous growth of SEI film. Notably, CV profiles of this SnO2@rGO composite anode in next cycles are almost identical to each other after initial electrochemical activations in the first two cycles, indicating the excellent electrochemical reversibility similar to that of the bare rGO anode (**Figure 4c**). By contrast, the reversible SnO<sup>2</sup> ↔ SnO ↔ Sn conversion reaction cannot be well-maintained in the bare SnO<sup>2</sup> anode, leading to the gradual disappearance of anodic/cathodic CV peaks around 1.5/1.0 V in its CV curves after initial two cycles (**Figure 4b**), which may account for the drastic capacity decay of the bare SnO<sup>2</sup> anode material in initial 10 of charge/discharge cycles. In a short summary, CV results indicate the satisfied electrochemical reversibility for stable lithium ion storage of the free-standing SnO2@rGO anode on the basis of synergistic effects from SnO<sup>2</sup> and rGO components (Cong et al., 2015; Huang et al., 2016).

**Figures 4d,f** plot galvanostatic charge/discharge curves at different cycles of these three anodes at a current density of 500 mA g−<sup>1</sup> between 0.01 and 3.0 V vs. Li+/Li, and corresponding cycling performance are compared as shown in **Figure 4g**. The SnO2@rGO anode can deliver initial discharge and charge capacities of 1,169 and 744 mAh g−<sup>1</sup> in the first cycle, respectively. The moderate initial columbic efficiency of 63.6% is mainly attributed to the undesirable SEI formation and the lithium consumption during the conversion reaction from SnO<sup>2</sup> to metallic Sn and lithiated Li2O in the first discharge process. By contrast, the lower columbic efficiency of 54.6% in the bare SnO<sup>2</sup> anode may result from the limited reversibility from the reduced Sn and Li2O back to SnO<sup>2</sup> and Li<sup>+</sup> ion in the charge process, and the inferior columbic efficiency of 23.5% in the bare rGO anode should be caused by considerably aggravated side reactions. As a result, the SnO2@rGO anode can retain a desired charge capacity of 711 mAh g−<sup>1</sup> in the 50th cycle, corresponding to the capacity retention of 95.6%, much higher than that of the bare SnO<sup>2</sup> anode (reserved 126 mAh g−<sup>1</sup> and 16.8% retention in the 50th cycle as shown in **Figure 4g**). Significantly enhanced cycling stability of the SnO2@rGO anode can be attributed to the free-standing structure, in which active SnO<sup>2</sup> nanoparticles are well-accommodated in the flexible rGO buffer with excellent conductivity and sufficient porosity. **Figure 4h** compares EIS spectra of both the bare SnO<sup>2</sup> and SnO2@rGO anodes before and after cycling. The SnO2@rGO composite anode shows a distinctly lower charge-transfer resistance after 50 cycles compared with that of the bare SnO<sup>2</sup> anode, which can be attributed to high electronic conductivity of the rGO framework even without any carbon additives. In addition, side

500 mA g−<sup>1</sup> of (d) SnO2@rGO, (e) bare SnO2, and (f) bare rGO anodes, (g) cycling performance at a current density of 500 mA g−<sup>1</sup> SnO<sup>2</sup> and SnO2@rGO anodes before and after cycling and (i) comparative high-rate performance of three anodes.

reactions of active SnO<sup>2</sup> and/or LixSn materials may also be impeded in the composite anode through the rGO protection. It is noting that the SnO2@rGO anode has the reduced resistance after cycling in comparison with its original state, possibly owing to the electrolyte infiltration into its internal structure during cycling. Furthermore, improved rate capability is also achieved of the SnO2@rGO anode, which delivers high charge capacities of 1,085, 958, 758, and 480 mAh g−<sup>1</sup> at gradually-increased current densities of 100, 200, 500, and 1,000 mA g−<sup>1</sup> , respectively. The specific capacity around 1,000 mAh g−<sup>1</sup> can be reversed, when such the free-standing anode is cycled back to a low current density of 100 mA g−<sup>1</sup> after high-rate trials (**Figure 4i**). Electrochemical performance verifies that the free-standing SnO2@rGO anode supported by the rGO framework results in enhanced cycling stability and rate capability of SnO2-based anode materials for superior lithium storage.

In order to deeply understand the relationship between the designed structure and superior electrochemical performance of free-standing SnO2@rGO composite material, the cycled anode is reexamined after 50 cycles in the full-charged state. **Figure 5a** shows the SEM image of cycled SnO2@rGO anode, which preserves the well free-standing structure. Tin-based nanoparticles are expectedly localized between rGO sheets as shown in the cross-section view. The cycled electrode shows compressed pores and channels by contrast with the initial structure (**Figures 2a,b**), which can be attributed to the SEI formation during cycling, together with the high-pressure effect from the coin cell assembly on the anode. As shown in **Figure 5b**, XRD pattern of the cycled anode reveals the co-existence of three rGO, SnO<sup>2</sup> and Sn components in the fully-charged state. The reformation of SnO<sup>2</sup> component verifies the partial oxidation of Sn to SnO<sup>2</sup> during the charge reaction, in consistence with CV results (**Figure 4a**), which significantly accounts for high lithium storage capacity of the SnO2@rGO anode as illustrated in Equation (1). TEM and HRTEM images are recaptured on the cycled SnO2@rGO material. The TEM observation in **Figure 5c** clarifies the structural stability of such an assembled composite material, in which tin-based nanoparticles are uniformly and

, (h) EIS spectra of both the bare

tightly wrapped within rGO sheets, in accordance with the SEM image (**Figure 5a**). **Figures 5d,e** presents HRTEM images of SnO<sup>2</sup> and Sn particles through indexing their different lattice fringes, respectively, which is consistent with the XRD results (**Figure 5b**). All characterizations of the cycled anode demonstrate the desired structural stability and the Sn to SnO<sup>2</sup> oxidation reaction during the charge reaction of the free-standing SnO2@rGO anode; hence, improved cycling stability with high specific capacity can be obtained for prolonged cycles.

#### CONCLUSIONS

An anti-solvent precipitation method has been developed to effectively assemble SnO<sup>2</sup> nanoparticles and graphene oxide sheets for the fabrication of a free-standing electrode, which is free of the conductive additive, polymer binder and current collector. The resulting SnO2@rGO composite anode shows significantly improved lithium storage performance compared with the bare SnO<sup>2</sup> anode material. It can deliver an impressive charge capacity of ∼500 mAh g−<sup>1</sup> at a high current density of 1 A g−<sup>1</sup> , and an attractive capacity above 700 mAh g−<sup>1</sup> can be retained after 50 cycles at a moderate current density of 500 mA/g. Enhanced cycling stability and rate capability of the composite anode can be attributed to the unique free-standing structure, in which all SnO<sup>2</sup> nanoparticles are tightly wrapped within rGO sheets full of wrinkles. Material characterizations of the cycled anode indicate the desired structural stability of such a free-standing SnO2@rGO anode, accounting for superior lithium storage performance. This work offers a facile assembly method for the preparation of free-standing composite materials with enhanced performance for device applications.

### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

### AUTHOR CONTRIBUTIONS

JZ, LG, and RH conceived the idea and designed the work. SJ, RH, WZ, XL, YZ, and ZG carried out materials synthesis, characterizations, and performance measurements. SJ, JZ, and RH wrote and revised the paper. All authors have made substantial, direct and intellectual contributions to the work.

#### FUNDING

This work was supported by the National Natural Science Foundation of China [grant number 21703147, U1401248]; the Jiangsu Provincial Natural Science Foundations for

#### REFERENCES


the Young Scientist [grant number BK20170338]. The Open Fund of Jiangsu Key Laboratory of Materials and Technology for Energy Conversion [grant number MTEC-2017M01]. The authors also acknowledge Suzhou Key Laboratory for Advanced Carbon Materials and Wearable Energy Technologies, Suzhou 215006, China, and Testing and Analysis Center, Soochow University, Suzhou 215123, China.


nanosheets towards high-rate and long cycle life lithium-ion batteries. Chem. Eng. J. 361, 329–341. doi: 10.1016/j.cej.2018.08.217


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

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

# A Hierarchical Copper Oxide–Germanium Hybrid Film for High Areal Capacity Lithium Ion Batteries

Liying Deng<sup>1</sup> , Wangyang Li <sup>1</sup> , Hongnan Li <sup>1</sup> , Weifan Cai <sup>2</sup> , Jingyuan Wang<sup>2</sup> , Hong Zhang<sup>1</sup> , Hongjie Jia<sup>1</sup> , Xinghui Wang1,3 \* and Shuying Cheng1,3 \*

*<sup>1</sup> College of Physics and Information Engineering, Institute of Micro-Nano Devices and Solar Cells, Fuzhou University, Fuzhou, China, <sup>2</sup> NOVITAS, Nanoelectronics Centre of Excellence, School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore, Singapore, <sup>3</sup> Jiangsu Collaborative Innovation Center of Photovolatic Science and Engineering, Changzhou, China*

#### Edited by:

*Jianqing Zhao, Soochow University, China*

#### Reviewed by:

*Wangwang Xu, Louisiana State University, United States Qingshui Xie, Xiamen University, China*

#### \*Correspondence:

*Xinghui Wang seaphy23@fzu.edu.cn Shuying Cheng sycheng@fzu.edu.cn*

#### Specialty section:

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

Received: *16 October 2019* Accepted: *03 December 2019* Published: *08 January 2020*

#### Citation:

*Deng L, Li W, Li H, Cai W, Wang J, Zhang H, Jia H, Wang X and Cheng S (2020) A Hierarchical Copper Oxide–Germanium Hybrid Film for High Areal Capacity Lithium Ion Batteries. Front. Chem. 7:869. doi: 10.3389/fchem.2019.00869* Self-supported electrodes represent a novel architecture for better performing lithium ion batteries. However, lower areal capacity restricts their commercial application. Here, we explore a facial strategy to increase the areal capacity without sacrificing the lithium storage performance. A hierarchical CuO–Ge hybrid film electrode will not only provide high areal capacity but also outstanding lithium storage performance for lithium ion battery anode. Benefiting from the favorable structural advance as well as the synergic effect of the Ge film and CuO NWs array, the hybrid electrode exhibits a high areal capacity up to 3.81 mA h cm−<sup>2</sup> , good cycling stability (a capacity retention of 90.5% after 150 cycles), and superior rate performance (77.4% capacity remains even when the current density increased to 10 times higher).

Keywords: self-supported electrode, lithium ion battery, CuO, Ge, areal capacity

### INTRODUCTION

Rechargeable lithium ion batteries (LIBs) are identified as the ideal sources of power for wide applications ranging from portable electronic devices to large-scale products on account of their long-life span and high energy density (Liu et al., 2018b; Xu et al., 2018; Zhang et al., 2018; Yan et al., 2019). However, the specific capacity of the electrodes severely restricts their energy density. As a solution to this problem, different anode materials with higher specific capacities have been investigated to take the place of the present commercial graphite (Kim et al., 2017). To date, Ge and CuO have aroused increasing interest as the novel anodes for new generation LIBs due to their high theoretical capacities. Compared with commercial graphite (theoretical gravimetric capacity is 372 mA h g−<sup>1</sup> ), Ge has a high theoretical capacity of 1,624 mA h g−<sup>1</sup> , while CuO has a capacity of 674 mA h g−<sup>1</sup> . Ge has been widely studied because of its high ionic conductivity and low lithiation potential and CuO, as one of the transition metal oxides, has demonstrated to be a promising material for the substitute anodes in LIBs for its earth abundance, commercial benefit, and environmental friendly (Chan et al., 2008; Xiaojun et al., 2012; Huang et al., 2015; Susantyoko et al., 2015; Mironovich et al., 2017).

However, the considerable capacities of Ge and CuO are generally accompanied by drastic volume change upon Li intercalation and deintercalation, thus causing the poor cycling performance (Liu et al., 2015; So et al., 2018). Great efforts have been made to solve the pulverization problem during the Li insertion and extraction process using nanomaterials such as nanoparticles (Hyojin et al., 2005; Mi-Hee et al., 2010; Yang et al., 2010), nanowires (Chan et al., 2008; Chockla et al., 2012; Yuan et al., 2012; Mullane et al., 2013), nanotubes (Chen et al., 2009; Cao et al., 2015; Liu et al., 2018a; Sun et al., 2018), and so on. For example, Li et al. synthesized mesoporous and hollow Ge@C nanostructures via carbon coating and reduced the hollow ellipsoidal GeO<sup>2</sup> precursor into Ge. A stable cycling performance (capacity retention remained 100% at 0.2 C rate for 200 cycles) and high rate capability (805 mA h g−<sup>1</sup> at 20◦C) is finally obtained (Li et al., 2013). Wang et al. fabricated self-supported CuO nanowires (NWs) on Cu foam (CF). The obtained electrodes delivered a specific capacity of 461.5 mA h g−**<sup>1</sup>** after 100 cycles at a current density of 100 mA g−<sup>1</sup> , and a capacity of 150.6 mA h g−<sup>1</sup> even at a high rate of 1,000 mA g−<sup>1</sup> (Wang et al., 2018).

Although these nanoengineering strategies have effectively improved the Li<sup>+</sup> ions storage performance of these highcapacity electrodes (Sun et al., 2018), most of these nanoscale metal oxides/group-IV elements and corresponding composites are mixed with organic binders and conductive carbon and then fabricate into electrode, which complicate the fabrication process (Wang et al., 2012). The bonding force between traditionally used binders and high-capacity active materials is too weak to maintain a stable performance after long-term cycling (Chang et al., 2019).

Recently, self-supported active nanomaterials which is in situ grown on current collectors without any inactive materials represent an unique architecture, which can offer many advantages for LIBs such as large contact area with electrolyte, great electrical conductivity, fast Li-ion transportation, and better performance for the electrodes (Wang et al., 2016). Susantyoko et al. fabricated amorphous Ge on the multiwall carbon nanotube arrays (MWCNT/a-Ge) by the combination of facial chemical vapor deposition and a physical sputtering method, which could give a specific areal capacity of 0.405 mA h cm−<sup>2</sup> at the rate of 0.1 C after 100 cycles (Susantyoko et al., 2014). Kim et al. synthesized nano-Ge/C composite via electrochemical deposition method; the obtained electrode exhibits a capacity of 1 mA h cm−<sup>2</sup> at 0.1 C over 50 cycles (Kim et al., 2017). Ji et al. fabricated binder-free electrodes, which is composed of 3D graphene network and octahedral CuO, the obtained 3D GN/CuO composites, yielding an areal capacity of 0.39 mA h cm−<sup>2</sup> at 0.095 mA cm−<sup>2</sup> (Dong et al., 2016). Xu et al. synthesized CuO mesocrystal entangled with MWCNT composites through a combination of precipitation and an oriented aggregation process. The CuO-MWCNT composites could deliver an areal capacity of 1.11 mA h cm−<sup>2</sup> after 400 cycles at the current density of 0.39 mA cm−<sup>2</sup> (Xu et al., 2016). Great progress for CuO- and Ge-based self-supporting electrodes has been achieved by the above-mentioned effects. Areal capacity is one of the important parameters for practical LIB application. Especially for self-supporting electrodes, it is very important and hard to obtain both high areal capacity and good electrochemical performance. However, the areal capacities of the most obtained electrodes are <2 mA h cm−<sup>2</sup> , which is lower than the commercial specification of 3–4 mA h cm−<sup>2</sup> (Cong et al., 2017). Normally, larger mass loading of active materials will make contribution to higher areal capacities but meanwhile sacrificing electrochemical performance (Chang et al., 2019). There is an increasing concern about fabricating self-supporting electrodes with high areal capacity as well as good electrochemical performance.

Usually, the self-supporting electrodes cannot maintain good electrochemical performance at very high areal capacity. Herein, we report a hierarchical CuO–Ge hybrid film on CF as a self-supporting electrode with ultrahigh areal capacity for LIB application. As shown in **Figure 1**, the integrated film was formed by physical vapor deposition of Ge film on CuO NWs array, which were grown directly on the CF via a facial and scalable solution approach. CuO NWs array with well-defined nanostructure serves as both the active materials and conductive connection for Ge film. The porous feature will not only alleviate the drastic volume change during the Li insertion and extraction process but also facilitate the diffusion of electrolyte into the electrode. Benefiting from the favorable nanostructures as well as the synergic effect of the Ge film and CuO NWs

Frontiers in Chemistry | www.frontiersin.org

array, the integrated electrode delivers ultrahigh areal capacity, extraordinary rate capability, and stable cycling performance. Moreover, this is the first time that CuO NWs combined with Ge film hybrid anode achieved a high areal capacity. It could deliver an ultrahigh charge areal capacity up to 3.45 mA h cm−<sup>2</sup> after 150 cycles at a current density of 0.8 mA cm−<sup>2</sup> and a capacity ∼2.98 mA h cm−<sup>2</sup> at a current density as high as 4 mA cm−<sup>2</sup> .

#### EXPERIMENTAL

#### Fabrication Procedure Fabrication of CuO NWs Array

Typically, the growth of CuO NWs array on CF is fabricated from a simple and scalable method. The CF with dimension of 2 × 2 cm<sup>2</sup> (1 mm, 100 PPI) was first degreased in 1.0 mol L−<sup>1</sup> HCl for 15 min and then washed in ethanol, acetone, and deionized water by ultrasonication for 15 min, respectively. The rinsed CF was then submerged in the mixed solution, which was prepared by adding 8 ml of freshly obtained 10 M NaOH into (NH4)2S2O<sup>8</sup> solution [0.913 g (NH4)2S2O<sup>8</sup> added into 22 ml of deionized water] with magnetically stirring. After 15 min reaction time, the light blue CF can be obtained, which means the formation of Cu(OH)<sup>2</sup> NWs. After rinsing with deionized water for several times, the light blue CF was dried under N<sup>2</sup> gas flow and annealed at 180◦C in air at a ramp rate of 2◦C min−<sup>1</sup> for 2 h. Then, the final product CuO NWs array was obtained with dark brown color. The principal of CuO NWs array in situ grown on CF could be described as following reactions (Cheng et al., 2016):

$$\begin{aligned} \text{Cu} + 4\text{OH}^- + (\text{NH}\_4)\_2\text{S}\_2\text{O}\_8 &\rightarrow \begin{aligned} \text{Cu} \,(\text{OH})\_2 + 2\text{SO}\_4^{2-} \\ &+ 2\text{NH}\_3\uparrow + 2\text{H}\_2\text{O} \\ \text{Cu(OH)}\_2 &\rightarrow \text{CuO} + \text{H}\_2\text{O} \end{aligned} \end{aligned} \quad \text{(1)}$$

#### Fabrication of Hierarchical CuO–Ge Hybrid Film

The CuO NWs array on CF was placed into an radio frequency (RF) sputtering system (Verios G4 UC, Shenyang Lining Co.) using 99.999% pure Ge target. The base pressure was 7.8 × 10−<sup>4</sup> Pa. Then, argon flowed at 50 sccm, and pressure remained at 2.2 Pa. The RF power was 100 W, and the deposition time was 400 min. The mass loading of CuO NWs supported Ge (typically ∼0.67 mg cm−<sup>2</sup> ), CuO NWs array (typically ∼3.64 mg cm−<sup>2</sup> ) and Ge on pristine CF (typically ∼0.45 mg cm−<sup>2</sup> ) were weighed before and after sputtering using a microbalance (OHAUS, AX224ZH) with an accuracy of 0.1 mg. Besides, the Ge film was also deposited on pristine CF under the same deposition parameters.

#### Structural Characterization

The samples were characterized using X-ray diffraction (Rigaku Ultima IV) and Raman spectroscopy (WITEC alpha300 R Confocal Raman system), and the structure and morphology characterization of them were carried out by field-emission scanning electron microscopy (SEM, FEI Inspect F50) with accelerating voltage of 5.00 kV and transmission electron microscopy (TEM, FEI Tecnai G2).

#### Electrochemical Characterization

CR 2032-type coin cells was used to test electrochemical characterizations, test cells were assembled in a high-purity argon filled glove box (Mikrouna Technology) with oxygen and moisture level <0.5 ppm. The fabricated self-supporting electrodes were used as the working electrode and a Li foil as the counter and reference electrode. Lithium hexafluorophosphate (LiPF6) (1 M) in a mixture of ethylene carbonate and diethyl carbonate (1:1 in volume) was used as the electrolyte. All the cells were aged for 12 h so that the electrolyte can fully infiltrate the whole cell before measurement. Lithium storage performance were evaluated by a multichannel battery tester (Neware, BTS-610) in the voltage range of 3.0–0.01 V (Li/Li+). An electrochemical workstation (CHI660c, Shanghai Chenhua Co.) was used to evaluate the cyclic voltammetry (CV) at scan rate of 0.1 mV s−<sup>1</sup> . All the tests were carried out in the thermotank at fixed temperature of 25◦C.

#### RESULTS AND DISCUSSION

The Raman spectra analyses of the Ge film, CuO NWs array, and CuO–Ge hybrid film are shown in **Figure 2A**. For Ge film, a broad peak at 290 cm−<sup>1</sup> was observed, which can be indexed to amorphous form of Ge (Susantyoko et al., 2014). There are three Raman peaks at 280, 324, and 618 cm−<sup>1</sup> for CuO NWs array samples, which can be indexed to the Ag, B(1) <sup>g</sup> , and B(2) <sup>g</sup> modes of CuO (Xu et al., 2015). All the peaks can be found on the CuO–Ge hybrid film samples, demonstrating that the hybrid structure was successfully fabricated. The X-ray diffraction patterns are shown in **Figure 2B**. One can note that the strong diffraction peaks of 43.3, 50.4, and 74.1◦ , which could be indexed to the CF with JCPDS card no. 70-3039. There are two weak but identifiable peaks located at 35.5 and 38.8◦ , corresponding to the (−111) and (111) planes of the monoclinic CuO, with JCPDS card no. 89-5899.

The typical SEM images of the obtained CuO NWs array on CF are shown in **Figures 3A–C**. The low-magnification SEM image in **Figure 3A** shows that CF has a well-organized 3D porous architecture. The magnified image shown in **Figure 3B** indicates the aligned CuO NWs array are 192 nm in diameter, and there are sufficient space available in CuO NWs array as indicated by the white-dashed squares, which can provide room for Ge thin film loading. Besides, the side view SEM image shown in **Figure 3C** demonstrates that all the CuO NWs array with length of ∼6.1µm are firmly rooted from Cu microfibers.

Then, a thick amorphous Ge was sputtered on the CuO NWs array using RF sputtering technique. **Figures 3D–F** show the SEM images of the CuO–Ge hybrid film. A typical image of a part of CF, as shown in **Figure 3D**, indicates that the 3D ordered nanowire architecture are well preserved after Ge coating. From **Figures 3E,F**, it is clearly observed that the average diameters and length of the nanowires are increased to 587.5 nm and 6.6µm, respectively, due to the deposition of Ge film. There are still large space in between these nanowires after sputtering as indicated by the white-dashed squares in **Figure 3E**, which will not only benefit for accommodating the volume change but also facilitate

the diffusion of electrolyte into the electrode. From the side view SEM images shown in **Figure 3F**, it can be seen that the diameter of the synthetic nanowires is gradually decreased from the top to the bottom, which is attributed to the shadowing effect of the RF sputtering technique. It has been demonstrated that this structure is beneficial for Li storage performance (Wang et al., 2016).

We also present the TEM images of the CuO–Ge hybrid film in **Figure 4A**. The obtained distributions of Cu and Ge are shown in **Figure 4B**, the energy dispersive spectroscopy mapping profile obviously pictures that the outer sheath consists of Ge, whereas Cu is perfectly populated in the inner part of the CuO–Ge hybrid film, and the Ge films were grown uniformly and was deposited onto the whole CuO NWs.

**Figure 5A** shows the typical initial discharge and charge profiles of the CuO–Ge hybrid film, CuO NWs array, and Ge film with the voltage window of 0.01–3 V (Li+/Li) at a current density of 0.8 mA cm−<sup>2</sup> . The CuO–Ge hybrid film delivers an initial discharge and charge capacity of ∼5.09 and 3.81 mA h cm−<sup>2</sup> , giving the initial Coulombic efficiency of 74.8%. The irreversible discharge capacity is associated with the formation of solid electrolyte interface layer and the irreversible insertion of Li<sup>+</sup> into CuO and Ge films, which are common for CuO and Ge based anodes (Chan et al., 2008; Chockla et al., 2012; Yuan et al., 2012; Mullane et al., 2013; Liu et al., 2018a). The low Coulombic efficiency may restrict the capacity of anode materials; however, it has been demonstrated that LIBs must undergo a few charge–discharge cycles, which is generally called the "formation process" (Chen et al., 2013). Besides the first cycle, the Coulombic efficiency of the battery was all above 99.2%, indicating excellent recyclability. In contrast, the first discharge and charge capacities are 3.34 and 2.50 mA h cm−<sup>2</sup> for CuO NWs array and 0.71 and 0.59 mA h cm−<sup>2</sup> for Ge film, respectively. The CuO–Ge hybrid film electrode exhibits much higher initial charging areal capacity when compared with the sum of the Ge film and CuO NWs array

electrode. This is because the mass loading of the Ge film on CuO NWs array is higher than that on CF attribute to the larger surface area of the CuO NWs array, demonstrating the structural advantages of the CuO NWs array.

**Figure 5B** compares the cycle performance of the CuO–Ge hybrid film, CuO NWs array and Ge film for subsequent 150 cycles at a high current density of 0.8 mA cm−<sup>2</sup> . For the CuO–Ge synthesized film electrode, a reversible discharge capacity of

Ge film.

3.81 mA h cm−<sup>2</sup> was achieved at the second cycle, corresponding to a specific capacity of 883 mA h g−<sup>1</sup> . The electrode could still deliver a high areal capacity of ∼3.45 mA h cm−<sup>2</sup> with a capacity retention of 90.5% after 150 cycles (the corresponding specific capacity contribution from Ge was 1,462 mA h g−<sup>1</sup> , and CuO was 678 mA h g−<sup>1</sup> ). In contrast, the CuO NWs array electrode could deliver a capacity of 2.47 mA h cm−<sup>2</sup> after 150 cycles, which corresponds to the 90.0% of the original one. While Ge film can only obtain a reversible capacity of 0.2 mA h cm−<sup>2</sup> after 150 cycles with a much lower capacity retention of 59.3%. The CuO–Ge hybrid film electrode exhibits a superior improvement in Li storage performance compared to the other electrodes, which may be attribute to the novel structure design using a hierarchical 3D nanostructure to combine two high theoretical capacity materials. The well-separated CuO NW arrays will not only provide large area for larger mass loading of Ge but also the large void space to overcome the large volume change during charge and discharge.

The initial few CV curves of the CuO–Ge hybrid film electrode were conducted in range of 0.01–3.0 V at a scan rate of 0.1 mV s −1 as displayed in **Figure 5C**. It can be indicated from these peaks that there is a multistep electrochemical reaction between Li and the hybrid electrode. A broad but moderate peak at ∼2.1 V corresponding to the initial formation of LixCuO in the first cathodic sweep. Then, two succession reduction peaks were observed at ∼1.0 and ∼0.66 V, corresponding to the transformation of LixCuO into Cu2O and Cu; then, a moderate peak at ∼0.07 V was observed, which was associated with the formation of the LixGe alloy (Rudawski et al., 2013; Guo et al., 2015). These reduction peaks shifted toward slightly higher voltages in the following scans, which might associate with the drastic Li driven structural modifications during the initial discharge and charge process (Yunhua et al., 2014). A peak was found at 0.44 V during the first anodic scan, attributing to the phase transition of LixGe to Ge; then, two distinct peaks at ∼1.5 and ∼2.5 V and a shoulder peak at ∼2.7 V appeared, which was associated with oxidation of Cu<sup>0</sup> to Cu<sup>+</sup> and Cu2<sup>+</sup> (Dong et al., 2016; Xu et al., 2016; Wang et al., 2018). These results are in agreement with the other reports of electrochemical reactions of Ge and CuO with Li (Seo et al., 2011; Ren et al., 2013; Xinghui et al., 2014; Wei et al., 2017; Lin et al., 2018; Wang et al., 2018). The CV curves were well-overlapped with each other from the second cycle afterwards, suggesting that the electrode has a good reversibility.

The rate capability is further tested for the CuO–Ge hybrid film electrode, which is of significant importance for high power energy storage. The rate performance was evaluated by charging–discharging at varied current densities varying from 0.4 to 4 mA cm−<sup>2</sup> . As shown in **Figure 5D**, after the first five cycles at the current density of 0.4 mA cm−<sup>2</sup> , the obtained electrode showed a high discharge areal capacity of 3.85 m A h cm−<sup>2</sup> ; then, it slightly reduced to 3.68 and 3.45 m A h cm−<sup>2</sup> at current rates of 0.8 and 1.6 mA cm−<sup>2</sup> . Even at a rate as high as 4 mA cm−<sup>2</sup> , the CuO–Ge hybrid film could still deliver a reversible capacity of ∼2.98 m A h cm−<sup>2</sup> , corresponding to the 77.4% capacity of the capacity at 0.4 mA cm−<sup>2</sup> . After the rate returned back to the initial value of 0.4 mA cm−<sup>2</sup> for five cycles, 94.5% of the initial charge capacity was regained. In the comparison, the CuO NWs exhibited a capacity of 2.83, 2.65, 2.35, and 1.86 m A h cm−<sup>2</sup> , respectively, and eventually obtained a capacity of 2.59 m A h cm−<sup>2</sup> . As for the Ge film, it only showed a capacity of 0.98, 0.44, 0.33, and 0.10 m A h cm−<sup>2</sup> , respectively. Indicating the benefit from the favorable nanostructures as well as the synergic effect of the Ge film and CuO NWs array, the hybrid electrode hybrid film electrode has wonderful rate capability far beyond the CuO NWs and Ge film electrode.

To examine the structure stability of the CuO–Ge hybrid film electrode upon repeated discharge/charge process, the electrode was disassembled after 50 cycles. As shown in **Figure 6A**, the CuO–Ge hybrid film were uniformly remained on the CF with no detaching signs. From high magnification SEM shown in **Figure 6B**, one can see that the hierarchical CuO–Ge hybrid films maintain their original structure even after 50 cycles, indicating high structural stability of the hybrid structure, which proves that this novel structure can withstand the dramatic volume change caused during repeated discharge and charge cycles. Therefore, the excellent lithium storage performance of the CuO–Ge hybrid film electrode is mainly due to the following aspects: (1) The wellseparated CuO NW arrays can not only provide large area for higher mass loading of Ge, resulting in higher areal capacity, but also improve the cycle performance of the Ge film by providing sufficient void space to alleviate the large volume change of the Ge film. (2) The hierarchical porous feature in the hybrid film not only provides sufficient space to accommodate the drastic volume change but also facilitates the lithium diffusion into the inner electrodes. (3) The Cu in the lithiated CuO NWs will promote the electronic conductivities, enhancing the rate performance of the electrodes (Yang et al., 2014).

#### CONCLUSIONS

In conclusion, an efficient strategy to prepare self-supporting electrode with ultrahigh areal capacity for LIB application has been introduced. The obtained CuO–Ge hybrid film electrode exhibits excellent lithium storage performance. It can deliver a high areal capacity of 3.81 mA h cm−<sup>2</sup> after 150 cycles, corresponding to 90.5% of the original one. Furthermore, the electrode could deliver high areal capacities of 2.98 mA h cm−<sup>2</sup> even at ultrahigh current density of 4 mA cm−<sup>2</sup> . The hierarchical CuO–Ge hybrid film grown directly on CF could be a novel substitute of graphite for LIBs, and the facial and efficiency synthesis strategy sheds light on improving the areal capacity

#### REFERENCES


of the self-supporting electrodes, which can be applicable for preparation of other high capacity hybrid electrode for energy storage application.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/supplementary material.

#### AUTHOR CONTRIBUTIONS

All authors have contributed in various degrees to the analytical methods used, to the research concept, to the experiment design, to the acquisition of data, or analysis and interpretation of data, to draft the manuscript, or to revise it critically for important intellectual content.

#### FUNDING

This work was supported by the National Natural Science Foundation of China (grant number 11704071), the Excellent Youth Foundation of Fujian Scientific Committee (grant number 2019J06008), the Natural Science Foundation of Fujian Province, China (grant number 2017J01503), and the Award Program for Fujian Minjiang Scholar Professorship.

structure-matched stepwise lithium storage. ACS Nano 9, 11462–11471. doi: 10.1021/acsnano.5b05610


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

Copyright © 2020 Deng, Li, Li, Cai, Wang, Zhang, Jia, Wang and Cheng. 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.

# First-Principle Insights Into Molecular Design for High-Voltage Organic Electrode Materials for Mg Based Batteries

#### Johann Lüder <sup>1</sup> \* and Sergei Manzhos <sup>2</sup> \*

<sup>1</sup> Department of Materials and Optoelectronic Science, National Sun Yat-sen University, Kaohsiung City, Taiwan, <sup>2</sup> Centre Énergie Matériaux Télécommunications, Institut National de la Recherche Scientifique, Varennes, QC, Canada

Low cost, scalability, potentially high energy density, and sustainability make organic magnesium (ion) battery (OMB) technologies a promising alternative to other rechargeable metal-ion battery solutions such as secondary lithium ion batteries (LIB). However, most reported OMB cathode materials have limited performance due to, in particular, low voltages often smaller than 2 V vs. Mg2+/Mg and/or low specific capacities compared to other competing battery technologies, e.g., LIB or sodium ion batteries. While the structural diversity of organic compounds and the large amount of possible chemical modifications potentially allow designing high voltage/capacity OMB electrode materials, the large search space requires efficient exploration of potential molecular-based electrode materials by rational design strategies on an atomistic scale. By means of density functional theory (DFT) calculations, we provide insights into possible strategies to increase the voltage by changes in electronic states via functionalization, by strain, and by coordination environment of Mg cations. A systematic analysis of these effects is performed on explanatory systems derived from selected prototypical building blocks: five- and six-membered rings with redox-active groups. We demonstrate that voltage increase by direct bandstructure modulation is limited, that strain on the molecular scale can in principle be used to modulate the voltage curve and that the coordination/chemical environment can play an important role to increase the voltage in OMB. We propose molecular structures that could provide voltages for Mg insertion in excess of 3 V.

Keywords: magnesium battery, ab initio modeling, organic cathode material, rational design, DFT, molecular design

### INTRODUCTION

The development of smart grids, renewable energy sources (Dunn et al., 2011) and electro-mobility (Lu et al., 2013) are just a few fields that demand improved electric energy storage technologies. New solutions, which meet the growing demands of scalability, better performance, and lower costs, should also be sustainable and nontoxic. Secondary metal ion battery technologies are at the forefront of current electrical energy storage research as they can provide high energy density, be used in mobile as well as stationary/large scale applications and could last thousands of charge-discharge

#### Edited by:

Oleksandr Malyi, University of Oslo, Norway

#### Reviewed by:

Dan Huang, Guangxi University, China Young-Kyu Han, Dongguk University Seoul, South Korea

#### \*Correspondence:

Johann Lüder johann.lueder@mail.nsysu.edu.tw Sergei Manzhos sergei.manzhos@emt.inrs.ca

#### Specialty section:

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

Received: 02 August 2019 Accepted: 27 January 2020 Published: 18 February 2020

#### Citation:

Lüder J and Manzhos S (2020) First-Principle Insights Into Molecular Design for High-Voltage Organic Electrode Materials for Mg Based Batteries. Front. Chem. 8:83. doi: 10.3389/fchem.2020.00083

**103**

cycles (Hameer and van Niekerk, 2015). Metal ion batteries are increasingly used not only in mobile devices but also to palliate load-level balancing of solar- and wind power plants. To fully access the benefits of this type of technology, many aspects need to be improved such as slow ion diffusion kinetics limiting the rate performance or capacity and cyclability limitations as well as user and environmental safety (Goodenough, 2012).

Many open challenges are of course materials related; specifically, the active electrode material's interaction with metalions, e.g., lithium (Li) (Diouf and Pode, 2015), counter ions (Chen et al., 2018) or common electrolytes critically dictates the materials performance. Today, various types of active electrode materials (from here on referred to as electrode for brevity but to be distinguished from contact electrodes) are known that can be manufactured from various sources. Examples are layered structures (Kulish et al., 2014; Sun et al., 2016; Ortiz-Vitoriano et al., 2017), metal oxides (Liu et al., 2017), polymers (Nishide et al., 2004; Koshika et al., 2010; Muench et al., 2016), or metalorganic frameworks (Xue et al., 2016). Electrode materials can differ in their redox-active centers—like compounds with oxygen or sulfur atoms, or functional groups. Thus, the search for highperformance electrode materials extends over a huge search space of materials, and much effort was made and will continue to be necessary to explore it further to find new materials leading to improved battery performance (Mohtadi and Mizuno, 2014; Yabuuchi et al., 2014; Kim et al., 2015a; Wang et al., 2015; Muench et al., 2016; Mauger et al., 2017; Leisegang et al., 2019).

Organic electrode materials have gained much interest (Liang et al., 2012; Lee et al., 2018; Manzhos, 2019) as their production can be sustainable (Xu et al., 2008; Zeng et al., 2014), and non-expensive; they can be recyclable, non-toxic and have a low carbon footprint at the same time (Liang et al., 2012; Deng et al., 2013). They can rely on well-developed chemistries and industrial base. Indigo carmine is an example of a well-known organic material that can be used as an electrode (Yao et al., 2010). Moreover, organic materials are most suitable for sustainable energy storage technologies and do not defeat the purpose of renewable energy technologies (i.e., sustainability and scalability). For instance, load leveling of electric grids in which a fluctuating amount of excess energy from renewable energy sources must be stored and released. In addition, it was shown that by some measures (e.g., extremely high cycling rate) organic materials can outperform commercially available inorganic electrode materials (Koshika et al., 2009). Development of organic electrode materials for metal ion batteries has picked up momentum in recent years, both through experimental and theoretical studies (Stephan and Nahm, 2006; Bhatt and O'Dwyer, 2015; Chen and Manzhos, 2016a; Padhy et al., 2018; Tripathi et al., 2019) Moreover, many of their properties including redox activity can be tuned (Oltean et al., 2016; Chen et al., 2017) by molecular design allowing an excellent adaptability for particular technological fields. Remaining challenges to achieve better performance in e.g., capacity, voltage, power, rate capability, and cyclic stability might be solvable.

Among all secondary metal ion battery types, only Li ion batteries are widely used, for instance in commercially available lithium cobalt oxide based batteries due to their high energy density. However, lithium as a resource is very unevenly distributed and is relatively scarce, whereas many other metallic elements, which can be used for so-called post-lithium batteries, like sodium (Na), magnesium (Mg), or aluminum (Al) are abundant (Leisegang et al., 2019). Much research effort has therefore been directed to make use of alternative metal ions that could replace Li. Na ion is a suitable candidate for these batteries because Na is abundant, cheap and it has a similar valance shell structure as Li, i.e., one unpaired valence s electron (Ratnakumar et al., 1990; Wenzel et al., 2011; Lu et al., 2012; Slater et al., 2013). Often, sluggish kinetics, low voltage and low capacity limit the success of Na ion batteries. Moreover, the same inorganic host materials used in Li ion batteries may be unsuitable for reversible Na insertion. Well-known examples are graphite and silicon wherein lithiation but not sodiation can be achieved (Stevens and Dahn, 2001; Ge et al., 2012; Malyi et al., 2013). Nonetheless, organic and inorganic materials for Na ion electrodes were discovered that can outperform commercially available Li ion batteries in some key performance parameters (Zhao et al., 2016; Lee et al., 2018; Padhy et al., 2018).

Multivalent charge carriers like Mg2<sup>+</sup> or Al3<sup>+</sup> (Mohtadi and Mizuno, 2014; Leisegang et al., 2019) could theoretically double and triple the capacity per carrier of the active material compared to Li<sup>+</sup> and Na<sup>+</sup> as two (for Mg) or three (for Al) electrons would be transferred per active cation instead of the one electron from Li or Na. In particular, metallic Mg as anode material is interesting in this regard due to further benefits that are (i) being environmentally friendly, (ii) non-toxic, (iii) biodegradable and (iv) allowing for dendrite-free deposition (Gregory, 1990). For inorganic materials, examples for electrode materials for the storage of multivalent active cations are vanadium oxides or manganese oxide and Prussian blue analogs that work with several metal ions including Mg2<sup>+</sup> (and others; Hayashi et al., 2004; Xu et al., 2015; Tojo et al., 2016; Kulish et al., 2017a,b; Koch and Manzhos, 2019). However, the larger charge on the ions leads to significant polarization and has negative effects on the kinetics/diffusion rate in inorganic solids (Kulish and Manzhos, 2017). Organic electrode materials could in principle achieve fast kinetics with multivalent cations compared to inorganic materials due to larger atomic spacing (Canepa et al., 2017) but it is still a challenging field of research. Few multivalent organic electrode materials with sufficiently high voltage for a cathode were reported, that are, e.g., quinones such as anthraquinones providing voltages of ca. 2 V as Mg ion battery cathodes (Sano et al., 2012; Zhao-Karger et al., 2013; Bitenc et al., 2015, 2016; Pan et al., 2016a,b). The low voltage is a key problem in multivalent batteries and is a direct consequence of multivalence: the average voltage V to achieve a given state of charge is

$$V = -\Delta G/ne,\tag{1}$$

where 1G is the respective free energy change, e the electron charge and n the number of charges transferred per elementary reaction (i.e., n = 2 with Mg; Urban et al., 2016). That is to say, for a cathode, a very high 1G (i.e., high Mg-host interaction energy) should be achieved while preserving stability of the electrode material and reversibility of Mg insertion. Many organic materials that are redox-active with Na or Li atoms such as the well-studied carboxylates (Park et al., 2012; Renault et al., 2013; Sk and Manzhos, 2016) do not result in strong enough interaction that would allow using them in Mg ion batteries.

A systematic approach is required to understand and improve organic electrodes for multivalent Mg (ion) batteries. Rational design based on atomic insights from ab initio computations is critically important for this purpose and in particular to understand and systematically explore the connection between material properties, such as electronic structure, interlayer spacing and crystallinity, and the redox processes in materials (Miroshnikov et al., 2016; Seo et al., 2016; Araujo et al., 2017; Molaei et al., 2017). Different design approaches should be tested computationally for molecules and for solids to modulate the interaction strength with inserted metal ions and the voltage (Kulish et al., 2017a). For instance, p-doping, i.e., the creation of a "hole" state in the electronic structure in the vicinity of the valence band maximum (and below the lowest electronic conduction state) increases the binding strength of Li in aluminum doped silicon and of Li on boron-doped disodium terephthalate by ca. 1 and 2 eV, respectively (Legrain and Manzhos, 2015; Lüder et al., 2017a,b). Adding/replacing electron-withdrawing groups, e.g., F atoms or cyano (CN) groups, in a molecular structure can also increase the voltage by increasing the binding strength to the active cation (Chen and Manzhos, 2016c; Chen et al., 2018). Doping is only one way to affect the performance of batteries and it is particular in that it directly addresses the electronic structure of a material. Other modifications can aim to alter strain effects, the chemical surrounding, or a combination of them. In this work, we focus on materials that work by reduction, i.e., in materials in which the lowest unoccupied molecular orbital (LUMO) of the electrode determines the redox potential. For materials that work by oxidation, the HOMO (highest occupied molecular orbital) determines the voltage and effective strategies there can be different (Geniès et al., 1989; Chen and Manzhos, 2016c; Rodriguez-Pérez et al., 2016; Chen et al., 2018).

In this study, we explore conceptual solutions for high voltage organic molecule based Mg (ion) electrode materials operating by reduction using Density Functional Theory (DFT) calculations. We organize the paper as follows: firstly, we explore conceptually the limitations of the approach of increasing the voltage in organic Mg-ion batteries working by reduction only by direct electronic effects. We show that this mechanism has substantial limitations on achievable voltages and possibly leads to unwanted side reactions. With the help of small organic molecules consisting of five- and six-membered rings with redoxactive carbonyl groups that we employ as model systems, we show the extent to which expected voltage can be modulated by design of electronic structure such as the use of substitutional doping (Lüder et al., 2017a,b). We study how much the electronic effects can be separated from others for these systems. Secondly, the influence of strain (linear, rotational (bending) and dihedral strain) on the expected voltage with a molecule is elucidated (Chen and Manzhos, 2016b). Because of the limitations of design by the modulation of the electronic states, strain design and the effective use of it to increase the interaction strength to metal ions can potentially further contribute to increase in voltage. Third, we discuss a new design aspect that addresses the arrangement of the coordination centers such as ring linking and fusion to design cation environments with high redox potentials.

#### METHODS

The Gaussian (Frisch et al., 2016) package was used to perform DFT (Hohenberg and Kohn, 1964; Kohn and Sham, 1965) calculations. The B3LYP functional (Becke, 1993) and the 6– 31+G(d,p) basis set were used on all atoms. This basis set provides sufficiently accurate energies for the attachment of alkali atoms on organic molecules due to the inclusion of diffuse and polarized functions.

The formation energy per metal atom of m (m = 1, 2, 3, . . . ) metal atoms (e.g., Mg) and one molecular complex was computed from the total energy of m metal atoms attached to the complex at their fully relaxed geometry (E mol+M r ) and the total energies of m isolated metal atoms (E <sup>M</sup>) and the isolated and relaxed molecule (E mol r ).

$$E\_{ad} = \left(E\_r^{mol+M} - E\_r^{mol} - m\,\mathrm{E}^M\right) / m \tag{2}$$

From the formation energy, the voltage V vs. Mg2+/Mg can be estimated according to Ceder et al. (1997) and Urban et al. (2016)

$$V = -\frac{\left(E\_{ad} - E\_{coh}^{M}\right)}{en} \tag{3}$$

which relates the formation energy of the Mg-molecule complex to the cohesive energy of the metal E M coh in its bulk phase. For Mg E M coh is taken from the literature as 1.51 eV (Kaxiras, 2003; Kittel, 2005), and the charge (en) transferred in the redox process with n being the number of electron charges per cation is taken as two e <sup>−</sup> for Mg. This method of estimating the voltage approximates the Gibb's free energy (Equation 1) by neglecting entropy contributions as well as effects from vibrations. Nonetheless, the approach based on electronic energies computed by DFT to compute voltages has been successfully used before (Urban et al., 2016). It can also be used for molecular systems with accurate reproduction of experimental voltages with models of molecular solids, while in isolated molecular models the neglect of environmental effects causes an underestimation of the absolute values of the voltage on the order of 1 V but allows reliable comparative calculations between different molecular structures (Sk and Manzhos, 2016; Chen et al., 2017; Manzhos, 2019). This is also the approach used here—we use molecular calculations to compare the effect of molecular building blocks on the strength of interaction with Mg atoms and on voltage vs. Mg2+/Mg.

The binding strength of a Mg atom to the electrode material should exceed the cohesive energy of metallic Mg, or else the socalled plating will occur. When the binding energy is equal to the cohesive energy, the equivalent voltage based on Equation (3) would be 0 V. When there is no binding between the molecule and the Mg atom the theoretical voltage based on Equation (3) is formally −0.755 V and is negative as long as the binding strength is weaker than E M coh of metal M. Negative voltage would of course not be observed experimentally but will be used in what follows for formal comparison between molecules. Also, as singlemolecule models underestimate the voltage, molecular systems with formally negative voltages might be of practical interest.

The linear (achieved through elongation of bond lengths, e.g., between Mg and the redox-active parts or intramolecular bonds), rotational (caused by bending) and dihedral (i.e., through changes in dihedral angles) strain energy E<sup>s</sup> of a molecule, that is caused by the attachment of the Mg atoms or structural deformations induced otherwise, is computed as the energy difference between the molecular structure it assumes when the metal atom is attached (but single point calculations are done without the metal atom) or the deformation is applied (E mol f ) and the isolated fully relaxed molecule or the molecule before the deformation (E mol r )

$$E\_s = E\_f^{mol} - E\_r^{mol} \tag{4}$$

Linear strain is created by pulling a molecular complex apart along a chosen direction. The relaxation of the structure is performed on all but a minimal set of frozen atoms to control the strain direction. Bending strain generated by pushing a set of atoms within a molecular complex closer to each other and keeping their coordinates fixed during structural relaxation while all other atoms fully relax. Dihedral strain is set by freezing the coordinates of atoms fixed during structural relaxation that define a dihedral angle between connected rings. In this study, all deformations that lead to strain other than by Mg attachment are achieved by keeping atomic coordinates fixed during structural relaxations.

### RESULTS AND DISCUSSION

### Limits of Electronic Effects for Voltage Increase in Organic Mg (Ion) Electrode Materials

It was shown in several studies that the energies of the LUMO significantly correlate to the voltage achieved with materials operating by reduction as the electron from the attached alkali atom, e.g., Li, will occupy the LUMO of the molecule (Burkhardt et al., 2013; Liang et al., 2013; Kim et al., 2015b, 2016; Lüder et al., 2017a). Specifically, a stabilization of the LUMO energy of comparable systems, for instance by substitutional doping, show strong correlations to voltage increase. When one uses in this model the LUMO energy before molecular reduction, we address this model as static. However, the precise value of the voltage increase will depend on several factors due to structural and electronic relaxations during the redox reaction. For Li-ion electrode materials, the energy difference between the energy (eigenvalue) εLi2<sup>s</sup> of the 2s level in Li (or the Fermi level in the metallic anode if it is used as a reference) and the LUMO of the molecule will lead to a gain in total energy (and free energy) that will drive the reaction and largely determine the voltage, see Equation (1). According to this model, any change in the LUMO energy will affect the voltage, i.e., a lower LUMO will lead to a larger voltage. In our previous studies (Lüder et al., 2017a,b; Chen et al., 2019), we showed that substitutional p-doping can be used to lower the LUMO of a molecule. The LUMO shifts to a lower energy and therefore the energy gained in the redox reaction is increased, which was observed for different prototypical organic Li- and Na-ion electrode materials.

The same mechanism can be explored for Mg atom in which the role of the HOMO energy level of the Mg atom (εMg3s) and the electronic structure of the tested molecules are evaluated. In **Figure 1A**, the theoretical voltage computed for the attachment of one Mg atom is plotted vs. LUMO energy for a large number of tested molecular structures. The structures are shown in the Supplementary Material in **Figures S1**–**S5**. One observes that there is indeed a correlation, albeit somewhat loose, between the voltage and the LUMO. We highlighted different areas in **Figure 1A**. The areas are obtained by a k-means cluster finding algorithm on the data. The separation in three groups is based on the assumption that the tested molecules can be separated into non-working, low voltage and high voltage electrode materials. The algorithm employed the HOMO and LUMO energies, as shown in **Figure 1B** together with the HOMO-LUMO gap (black points) as features. The three areas can be roughly distinguished as molecules with higher voltages that do correlate well to very low LUMO energies (blue area, "deep LUMO"), molecules with high LUMO energies and formally negative voltage (i.e., unsuitable as active electrode materials, green area), and molecules with moderately high LUMO energies (magenta). In **Figure 1A**, the scatter of voltage-LUMO data is not surprising as there are other factors than bandstructure energy (we will liberally use "bandstructure" as set language even when talking about molecules).

The red curve in **Figure 1A** shows the linear regression of the LUMO-voltage data. The curve shows the well-known correlation between decreasing the LUMO energy and increasing the voltage, which is frequently observed for Li- and Naion electrode materials (organic and inorganic). Here we see that such correlation also largely holds for Mg attachment as well. It should be noted that the LUMO energies refer to the values of the single molecule without metal atom attached. The curve shows that a LUMO energy of <-3 eV is needed for a working anode as implied in Equation (3). Furthermore, the curve also reveals that for high voltage electrodes (more than 2 V) a LUMO energy of <-8 eV would be required from this mechanism.

The simplified picture above is based on the frozen state approximation. Relaxation effects could be significant. To account for the energy level relaxation, the energy level of the molecular LUMO, which becomes occupied by the valence electrons of the metal atom, may shift to lower energies. This process was observed, for instance, for the boron-doped coronene molecule under Li and Na attachment (Lüder et al., 2017b). Here, we include the LUMO relaxation effect after reduction and assume that the shifted LUMO will not decrease to energies below the HOMO energy. Thus, the change of free energy based on the electron transfer from metal to an

FIGURE 1 | (A) Computed LUMO eigenvalues and theoretical voltages vs. Mg2+/Mg for a set of small organic molecules. The molecules are categorized into three groups (blue, purple and green areas and data points) by a k-means algorithm. A linear regression (red curve) is compared to an electronic approximation (black curve) to Equation (2). The top axis refers to a comparison based on the electronic effects assuming a LUMO relaxation reaching HOMO values. (B) Comparison of the relation between HOMO and LUMO eigenvalues, and HOMO-LUMO gaps for the three categories in the dataset. (C) Comparison of the linear regression (red curve) from (a) different LUMO relaxation values in the electronic approximation to Equation (2). (D) Schematic of the LUMO relaxation process during the redox reaction. For a Mg atom, two electrons are transferred (i.e., n = 2). Arrows indicate electrons with spin-up and spin-down. Color code is H (white), C (gray), N, (blue), O (red), F (light blue), Mg (yellow), and S (dark yellow) here and below.

organic electrode (which is the net effect, even though the electrons transfer through the external circuit) can increase by the electronic relaxation of the LUMO energy during the redox reaction. Of course, the most significant change is achieved when the LUMO energy relaxes to values close to the HOMO energy. This effect becomes less significant for the redox process with decreasing LUMO energy level as displayed in **Figure 1B** due to the decrease of the HOMO-LUMO gap (black dots, right axis) or reduced relaxation. The figure also shows the three different types of materials determined by the kmeans classification, which are distinguishable by their HOMO and LUMO.

**Figure 1C** compares the regression of the computed data with the extended electronic model (i.e., including LUMO energy relaxations) illustrated in **Figure 1D** at different amounts of state relaxations. The free energy that enters the computation of the voltage [or the Ead of Equation (3)] is replaced by a weighted energy difference between HOMO and LUMO energy of the molecule and the HOMO energy of the Mg atom (−5.3 eV). It reads

$$E\_{ad} = a\left\varepsilon\_{HOMO}^{mol} + b\left\varepsilon\_{LUMO}^{mol} - \varepsilon\_{HOMO}^{Mg} \tag{5}$$

in which the weight factors satisfy a + b = 1. The sets of data points (indicated by different coloring) correspond to different weighting factors of the computed data shown in **Figures 1A,B** and the solid lines to their resulting linear regressions. The values of HOMO and LUMO energies are regarded in this model as a key determining factor of Ead and therefore give the limits of the electronic model. This model results in two extreme cases for the computed voltage that can be compared to the computed data. However, it must be noted that the calculated data (Equation 3) use the total energy changes obtained by DFT, while the estimates of the model are based on eigenvalue changes. Their sole purpose is to elucidate possibilities and limitations of voltage modulation by electronic effects. It is a measure of the gain in energy solely due to bandstructure (equivalent to a charge transfer in frozen orbitals), while the computed voltage

data include further relaxation channels (e.g., charge distribution, transfer, bond length changes) that affect the change in free energy, too. The energy of the LUMO defines one electronic limit (light green curve). The corresponding curve in **Figure 1C** is far below the linear regression of the computed data (red curve). For this case, working voltages (V > 0) require HOMO energies of <- 7 eV because the redox process will not benefit from any LUMO relaxation in the static case. With increasing electronic relaxation, i.e., the LUMO after reduction approaching values closer the HOMO energy, the curve approaches voltages close to 2 V. The other limit (black curve) is yielded by the HOMO energy, i.e., electronic relaxation where the occupation of LUMO makes its energy approach the HOMO. The resulting curve intersects with the regression curve of the computed data at ca. −5.5 eV while it overestimates voltages below and underestimates voltages above this energy. The same curve is overlaid in **Figure 1A** as black curve and it can be regarded as the electronic limit. From the figure, it becomes clear that the LUMO must be very deep to result in a positive voltage. In this model, the LUMO must be lower than ca. −8 eV to achieve voltages larger than 2 V.

Besides the apparent 2 V limit by electronic structure design, two more observations should be mentioned: First, neither the extended electronic model nor the static model offers a solution to the intrinsic limitation of electronic structure design for Mg electrode materials. Instead, they can be understood as different mechanisms (or point of views) in the redox process governed by electronic effects. While the static model requires a low LUMO for high voltages, this can also be achieved with a high LUMO material but with sufficient relaxation effects (requiring a low energy of HOMO). Second, while the electronic limit based on the HOMO energy appears similar to the regression curve based on the computed data, the actual LUMO relaxation does not reach energies close to the HOMO of the molecule before reduction. Thus, the electronic model is compensating for other effects that appear simultaneously, as mentioned above.

As a reference of realistic and technologically relevant materials operating by reduction, we consider the LUMO energies of tetracyanoethylene (TCNE) which is known for its electron-accepting property, i.e., a very low LUMO energy in a stable molecule associated with a large absolute electron affinity (Pearson, 1986). The LUMO energy of TCNE is shown as vertical dashed orange line at −5.25 eV (Chen and Manzhos, 2016b) in **Figure 1A**. A lower LUMO would in principle results in stronger reduction activity. Besides the difficulties of designed materials with extremely low LUMO energies, such materials would be extremely unstable. Even as packing effects in a solid state are expected to result in a somewhat higher voltage than computed from a single-molecule model (Manzhos, 2019), this clearly shows the severe limitation of the strategy of LUMO design in the case of Mg ion batteries even as it served well organic material design for Li and Na ion batteries.

Electrodes in organic Mg and Mg-ion batteries often use carbonyl groups, which create the danger of parasitic MgO formation. The formation energy of MgO of 3.07 eV implies an equivalent of about 0.75 V vs. a metallic Mg anode. This estimate gives a target for the minimal voltage that a molecule should achieve to avoid loss of capacity because of redox reactions resulting in voltages above these values, the formation of MgO could be avoided.

**Figure 1** contains data for several classes of small organic molecules with a ring structure, including benzene, 1,4 benzoquinone, biphenyl and biphenylene. They include the unmodified molecules, and, e.g., carbonyl, CN- and Ffunctionalized derivatives; details can be seen in SI. These functional substituents are known as electron-withdrawing groups. Thus, any substitution in benzene, benzene-based molecules or similar ring structures is expected to result in an electro-positive part (e.g., C atoms) and a more electro-negative part (i.e., the functional groups). Besides the redistribution of electronic charge, stabilizing effect on HOMO and / or LUMO energy and therefore changes in the interaction strength between a metal cation and the redox-active molecule can be anticipated, potentially with an increase in voltage. This was observed before, as mentioned earlier, in organic Li- and Na-ion electrode materials. However, the CN/F-functionalized benzene results cannot confirm this for organic Mg (and in principle Mg-ion) electrode as it did not significantly increase the interaction with Mg. In addition, we also investigated other halogen functionalization of benzene (e.g., Cl and Br). Our results indicate that electronic effects achieved by functionalization of materials can lead to small modulations of the voltage if a sufficient number of redox-active groups such as carbonyls are present.

On the other hand, a careful inspection of the data shown in **Figure 1A** reveals a few exceptional prototypical molecules (indicated by arrows and given molecular structures of biphenyl derived molecules). These molecules have (i) a LUMO energy higher than the one of TCNE and are therefore expected to be stable and (ii) their computed voltages are far above the regression curve and even beyond the electronic limit indicating that other processes in these molecules occur that strongly benefit the redox mechanism. Identifying the mechanism behind a higher computed voltage in these molecules and deriving molecule design strategies based on them will be discussed in the following.

### Effect of Structural Modifications Leading to Strain

Reversible structural modifications can influence the total energy of a molecule as well as electronic states. HOMO, LUMO energies and the HOMO-LUMO gap can be modulated by strain. Both, changes in the total energy and that of individual electronic states can in principle lead to changes in the interaction strength between molecules and Mg atom(s). Thus, it will affect the voltage according to Equations (1–3). However, the direct application of this idea for molecular-based electrode design and practical applications in batteries is not frequently explored.

To efficiently modify voltage curves, the changes of the total energy must mainly affect the mentioned interaction between the metal atom and the molecule while the intra-molecular changes and their associated energies are small, or vice versa. When the structural modifications induce total energy changes of similar size and same direction (that means endo- or exothermic), the resulting energy difference in the redox mechanism will cancel out according to Equation (1).

We distinguish between three principal structural changes: linear stretch, bending of the molecular plane, and dihedral rotation. For these three strain designs, we discuss prototypical systems.

#### Linear Strain

Here we study a system of a Mg atom coordinated to two molecules by two carbonyl groups, to account for the transfer of two electrons (one to each group). In a system of linear geometry in which two molecules with a carbonyl group surround a Mg atom, one way to introduce linear strain is by pulling either one of the molecules or both molecules simultaneously away from the Mg atom. Linear strain can also be introduced by elongating the bonds within a molecule. However, our investigation showed that for the latter case, effects on the LUMO or HOMO energy position are minor, and the binding strength to Mg atoms is barely affected. Moreover, the strong covalent bonds within organic molecule prevent a significant elongation by forces with technologically relevant magnitude. Thus, we focus on the approach of changing intermolecular geometry, i.e., the Mg-O bonds.

The chosen molecular complex that is used to investigate the effects of linear strain on voltage modulation is shown in **Figure 2**. We select a ring-fused linear molecule consisting of five six-membered rings. The molecule was laterally functionalized with two carbonyl groups with a distance between them of ca 9.8 Å, i.e., forming a pentacenedione (PAD) isomer. To systematically investigate the effect of linear strain on voltage modulation at the molecular scale and to provide technologically relevant insights, constraint geometry optimizations are required imposing constraints on and offering control of degrees of freedom. We restrict our analysis to a planar configuration of two PADs. This allows us to directly probe the correlation between atomic geometry, electronic effects, and the applied strain. **Figure 2A** shows the equilibrium structure of this (planar) configuration of the complex formed by a Mg atom and two 4,8 pentacendione molecules. We formed a series of geometries (of which three selected cases are shown in **Figure 2B**) in which the intramolecular degrees of freedom are relaxed, and the molecules are made slide in parallel to and past each other in the same plane. The initial structure in **Figure 2B** has the same Mg-O bond lengths as in the equilibrium structure. The middle figure of this panel shows the strained structure at an intermediate strain stage. The right panel of **Figure 2B** shows the most strained structure with a displacement of ca 12 Å and an O-Mg-O bond which is again similar to the one of the initial structures.

**Figure 2C** shows (top panel) changes in the binding strength of the configuration with the strongest binding, and changes in the total energy of the structure with and without the Mg atom (middle panel); binding distances to and Mulliken charges of the Mg atom (bottom panel); HOMO and LUMO eigenvalues of the Mg-complex and the complex without Mg atom. At the geometry with the largest strain energy shown in the middle of **Figure 2B**, there is a significant decrease in binding strength, up to 3.5 eV. The decrease in binding strength translates into a

FIGURE 2 | Constrainted calculation on 6-membered ring-fused PAD structure with two carbonyl groups per molecule. (A) Optimized geometry of the planar system. (B) Selection of strained configurations in which one of the PAD molecules is parallelly shifted with respect to the other molecule: (left) Initial structure with parallel alignment of the PAD molecules shown in (A), (middle) configuration with the highest strain energy equivalent, (right) configuration with final shift with a similar O-Mg-O bond as in initial structure. (C) Overview of electronic changes and properties during the shift: (top) Change of reaction energy (left axis) of Mg insertion (dark blue, square), change of energy of the complexes with (green, circle) and without Mg atom (light green, triangle), as well as computed voltage during parallel shift (right axis, dashed curve). (Middle) Maximal (blue, squares) and average distance (green, circles) between nearest two O atoms and the Mg atom; Mulliken charge on the Mg atom (light green curve) is given on the right axes. (Bottom) Eigenvalues of HOMO before and after Mg attachment (light green and blue, respectively) and eigenvalues of LUMO before Mg attachment (green).

voltage reduction of 1.75 V. At this stage, one Mg-O distance increases to 6 Å, which is too long for any covalent bond, see middle panel of **Figure 2C**. Hence, the complex has lost one of the two bonds while the other Mg-O distance remains at a typical value for strong bonds of ca. 1.9 Å. In addition, the electron charge transfer from Mg to the organic part of the molecule is reduced, which is seen in a drop of Mulliken charge on Mg to almost 0. Then, the decrease of binding strength can be explained by the loss of a Mg-O bond.

In contrast, the average and maximal distance of Mg-O are <2 Å (that is for average and max.) for maximal binding strength, and the Mulliken charge on Mg is ca. +1.3 |e| which corresponds to transfer of two electrons. Besides, the electronic structure (HOMO and LUMO) of the complex without the Mg atom (green curves in the lower panel of **Figure 2C**) is only slightly affected by strain. The LUMO relaxation effect is given by the energy difference between the LUMO energy of the molecular complex before Mg insertion (dark turquoise curve) and HOMO energy of the Mg-molecule complex (blue curve). It is very interesting that this relaxation effect shows a significant increase that correlates with the strain dependent weakening of the binding strength. This suggests that both effects, one being a charge transfer and bond release driven—the other being an electronic effect, act in the presented case in opposite directions.

At the final stage of the applied strain, the average and maximal Mg-O distance, Mulliken charge and binding strength reach values similar to the ones given at the initial stage indicating possible reversibility. Moreover, they imply that a key factor to efficiently design molecular systems, that can employ strain-based voltage modulation, is the ability to control the distance between Mg and the redox-active groups, and the number of formed Mg-O bonds. This suggests that the design of coordination of Mg to several redox-active groups is important; we will address this issue in the section Design of Coordination Environment of Mg Cations. However, the seen strain effects mainly reduce the voltage. Thus, the charging voltage of an organic electrode material can be reduced by applying linear deformation assuming the microstructure of the material supports this reversibly.

The results of the linear strain are in agreement with basic chemistry: at large distances no bonds are formed while at shorter distances the hybridization of states causes the formation of bonds (and consequent charge transfer leading to more ionic interactions). It should be noted that there is no strong correlation between the changes of the LUMO of the molecule only complex, and that changes in V are almost binary to situations when Mg binds to one and to two redox centers preferring two redox centers. For situations in which several redox centers surround the Mg ion—such as in a solid, a large molecule or several molecules—controlling the distances and amount of redox active centers binding to Mg may provide a way to increase voltages.

#### Bending Strain

We analyze the effect of bending in molecular complexes in which Mg binds to two neighboring molecules. Similarly to the linear strain section, constrained geometry optimizations were performed to investigate the effects of a bent molecular structure on the binding strength between organic electrode materials and the inserted Mg atom.

We chose the 1,2-benzoquinone (BQE), 2,3-naphthalenedione (NPH) and 2,3-anthracenedione (ACD), which are molecules similar to PAD but with a reduced number of fused sixmembered rings. BQE, NPH and ACD have one, two and three fused 6-membered rings, respectively. In each initial structure of the molecular complexes, two identical molecules from a linear complex with a Mg atom at its center. A planar configuration was imposed by symmetry during relaxation to simplify the evaluation of the bending effect (the NPH and ACD complexes have dihedral angles different from 0◦ at their lowest energy configurations). The complexes in their fully relaxed planar form are shown in **Figure 3A**. The bending was initialized by decreasing the distance d between the outer most carbon atoms (called the edge of the molecule) in steps of 1 Å until a distance of ca. 4 Å was reached. Four different bending stages of Mg-di-NPH are shown **Figure 3B** in which curved arrows indicate the bending direction. An example of a final bending configuration is shown in **Figure 3C** for the ACD Mg complex with almost parallel molecular parts.

**Figure 3D** shows the increase in strain energy (Es) when the Mg complexes are forced into a bent structure. All other atoms but those at the edges are allowed to fully relax including the bonds between the Mg atom and the carbonyl groups. Thus, any effect leading to a weakening of this bond (i.e., an increase of E<sup>s</sup> or a decrease of Ead) is expected to be accurately represented. After a steep increase at the beginning, the bending curves for the three studied complexes yield similar values of 0.8 to 1.2 eV (corresponding to a voltage decrease of ca. 0.5 V vs. Mg2+/Mg) when the two molecules reach almost parallel alignments (at a distance of ca. 4 Å), shown in **Figure 3D**, left panel. This indicates that the bending strain is not (strongly) depending on the size of the molecule but depends on the bond angle formed between O-Mg-O, shown in **Figure 3D**, right panel.

#### Dihedral Strain

The effect of changing a dihedral angle in a molecular structure on the interaction strength between Mg and an organic molecule can be tested similarly to the effect of bending. However, single fused-ring molecules like biphenylene or naphthalene exhibit strong restoring forces when the molecular structures are twisted. Neither electronic nor structural changes were found to be large enough under this deformation to result in significant voltage changes for these molecules. Thus, we focus on molecular prototypes like biphenyl-derived systems (as seen in the section Limits of Electronic Effects for Voltage Increase in Organic Mg (Ion) Electrode Materials) with a bond between ring structures that can act as a rotational axis. Although biphenyl is known for its sterically hindered full dihedral rotation of the two phenyl rings, we expect to see a rather low energy increase for a wide range of dihedral angles. The mentioned bond on which the dihedral rotation is performed is referred to as the rotation axis and the change of dihedral angle as 1γ.

Two carbonyl groups were introduced into the biphenyl molecule (C12H8O2, referred to in the following as BipO2) shown in **Figure 4A**. The carbonyl groups are on opposite phenyl rings. For this molecule, two principal configurations of the molecule are possible—one with the two carbonyl groups close to each

of the two ACD molecules in the complex. (D) Change of strain energy during bending as function of distance d (left) and as function of the bending angle α (right).

other (as shown in the figure) and the other with them far from each other. This is obtainable by a dihedral rotation. The larger separation of the carbonyl groups is energetically preferred by 0.3 eV. Bond lengths between atoms defining the dihedral rotation are taken at the relaxed molecular geometry and are kept constant when calculating the voltages for different dihedral angles. Similarly to the linear strain case, this degree of freedom allows us to manipulate the (binding) distances between carbonyl groups and a Mg atom.

The constrained calculations were performed with two reference structures to estimate the influence of initial conditions on the computed voltage changes. One structure represents the charged and the other the discharged state of the material. The discharged state is given by BipO2 with an attached Mg atom, i.e., Mg(C12H8O2)—referred to as MgBipO2. The charged state is given by BipO2, using the close-carbonyl group configuration because this is the configuration assumed by the molecule directly after the detachment of the Mg atom. The effect of Mg attachment, i.e., discharging, results in differences in the structure of the organic part of the molecule, which is most noticeable in the bond length and the dihedral angle between the phenyl-like rings. For the charged state (i.e., BipO2 without Mg) shown in **Figure 4A**, the bond length between the phenyl rings is 1.39 Å and has a dihedral angle of 30◦ while for the discharged system (MgBipO2) shown in **Figure 4B**, the bond length and dihedral angle increase to 1.51 Å and 71◦ , respectively. The impact on the computed properties with the different initial configurations in the constrained optimization is analyzed in **Figures 4C–F**. In these figures, we also compare the effect of keeping all atomic positions of the molecular structure fixed (i.e., only the position

FIGURE 4 | Effects of dihedral angle change in a biphenyl-like molecule with two carbonyl groups (BipO2) at connected phenyl-like rings. Differences in bond lengths between phenyl-like rings and dihedral angle γ in (A) BipO2 after Mg detachment and (B) MgBipO2, i.e., BipO2 with one attached Mg atom. (C) Change of interaction strength between Mg and the molecule for frozen and partly frozen, i.e., γ-frozen, relaxations based on the optimized BipO2 structure, as shown in (A), red curves, and based on MgBipO2, as shown in (B), blue curves. The computed voltages are shown in (D). The strain energies given in (E) and (F) are computed without and with one attached Mg atom, respectively.

of the Mg atom is optimized—labeled as "frozen") to the frozen dihedral angle calculation (i.e., all atomic positions but the ones defining the dihedral angle are allowed to relax during the structure optimization—labeled as "γ-frozen").

**Figure 4C** compares the change of interaction energy (Ead) and **Figure 4D** shows the computed voltage vs. Mg2+/Mg for both of the reference structures, i.e., BipO2 and MgBipO2. The dihedral angle change of the configuration with strongest binding differs by around 60◦ between the BipO2 and MgBipO2 references. This is a consequence of the dihedral angle increase under Mg attachment.

For the MgBipO2 reference system, the binding energy between the Mg atom and the BipO2 molecule is −5.5 eV when fully relaxed (i.e., MgC12H8O2), corresponding to a voltage of 2.0 V vs. Mg+<sup>2</sup> /Mg. Under dihedral rotation, the binding energy reduces to −1.1 and −1.7 eV (an energy decrease of 4.4 eV and 3.8 eV) for the frozen and γ-frozen optimization leading to a voltage of −0.2 and 0.1 V vs. Mg2+/Mg (i.e., changes of 2.2 and 1.9 V), respectively. Both maxima are at a dihedral angle change of ca. 120◦ , which yields an almost flat geometry. The steric hindrance in the latter configuration is present in the molecular structure with and without the attached Mg atom. The binding energy weakening and voltage decrease is then a consequence of the removal of one of the Mg-O bonds. The loss of one bond (or reduction of bond order) also results in a reduced charge transfer. This is supported by the binding energy and voltage computed at 1γ of 180◦ . In this configuration, the carbonyl groups have a large distance from each other, and a Mg atom can only bind to one of them. The binding weakens by ca. 3 eV and the voltage takes a value of 0.8 V—values similar to the ones seen with the bond release mechanism in linear strain applied to PAD (see above) indicating that these values are somewhat characteristic of a single Mg-O bond in these molecules and the voltage reduction is caused by the loss of a Mg-O bond.

Similar changes but of smaller magnitude are seen in the BipO2 reference calculations. The curves are shifted by around 60◦ , which follows the change of the dihedral angle under Mg attachment during full optimization. The strongest interaction between the Mg atom and the BipO2 reference molecule yields a binding energy of −3.3 (−2.1) eV corresponding to a voltage of 0.9 (0.2) V vs. Mg2+/Mg for the γ-frozen (frozen) optimization. Dihedral rotation changes these values by 2.6 (1.8) eV setting the lower estimate of the voltage change to 1.3 (0.9) V. This is given at 60◦ for the frozen and γ-frozen BipO2 reference. The voltages reach their maximum because this configuration is most similar to the fully optimized structure of MgBipO2. Interestingly, the voltage curve is changing sign for both cases, frozen and γ-frozen constrained geometry calculations. In principle, this could indicate a polarity change of direct current. Then, the charging cycle could be replaced by a mechanical formation. However, the calculation misses to account for any aggregate state contribution to the voltage which is in the order of 1 V. This may shift the voltage curve above the point of polarity change. Nevertheless, the principle should be transferable to other materials. Depending on the onset of the voltage-capacity curve, the strain induced polarity change may be applicable to low-voltage electrode materials in general.

For the tested γ-frozen configurations, the LUMO shift of the molecule has a range of 0.63 eV and the HOMO shift of the MgBipO2 has a range of 0.4 eV. The frozen cases also have ranges of 0.6 eV or less. The Mulliken charges are between 0.9 and 0.5 q at 0 and 150◦ for MgBipO2 (frozen and γ-frozen) and 0.6 and 0.1 q for the BipO2 configuration at 30 and 150◦ , respectively. This means that the bond between Mg and the molecule becomes less ionic and weakens due to a dihedral rotation. In general, this can be seen in a reduction of bond order or the loss of a bond. In the context of the calculations, it is the loss of one of the Mg-O bonds in MgBipO2. Thus, structural effects are a major factor to be considered besides the observed small changes in electronic structure for dihedral angle controlled voltage changes in this molecule and potentially similar systems.

**Figures 4E,F** compare the effect of strain energy E<sup>S</sup> when the Mg atom is removed and when it is present, respectively. Without an attached Mg atom (**Figure 4E**), the strain is minimized at round 120◦ for BipO2 and MgBipO2, as expected by the energetically preferred structure of BipO2 in which the carbonyl groups have a large separation distance. The transition potential of the rotation of BipO2 is ca. 0.9 and 1.4 eV for the γ-frozen and frozen calculation, respectively. For the MgBipO2 reference system, the molecule stores strain energy at the optimized structure due to the dihedral rotation under Mg attachment. Without Mg atom, this is partly released when the dihedral rotation reaches values similar to the one of the relaxed BipO2. In contrast, the strain energy increases by dihedral rotation when a Mg atom is attached. The max. values increase to more than 2 eV for all cases, see **Figure 4F**, and the increase in strain energy correlates roughly with a decrease in interaction strength with Mg in the case of the MgBipO2 reference. Since the distances between the redox-active groups and the Mg atom increase with increasing change of the dihedral angle, this correlation can be understood from the elongation and eventual loss of bonds between redox-active groups and the Mg atom.

For BipO2, the voltage change by dihedral strain is estimated to be between 1.3 and 1.9 V. To further explore torsion on the molecular scale for voltage modulation, we proceed with three other molecules of a similar structure. **Figure 5A** shows a 1,1′ -bi(1,4-cyclohexadiene-1-yl)-2,2′ ,6,6′ -tetrone (BCT) molecule with a Mg atom attached in two configurations in which the dihedral angle between the 6-membered rings differ by 90◦ . We chose the molecules with a Mg atom attached as initial geometries. The blue arrow indicates the change of dihedral angle and the superimposed structures distinguish the atomic geometry before the rotation (faded out and with red oxygen atoms) and with 90◦ rotation (indicated by benzene-like ring with blue oxygen atoms). The loss of one Mg-bond and the bending of the molecular structure is visible. **Figure 5B** gives an overview of the structural changes after constrained relaxation for which the dihedral angle was changed in steps of 30◦ in a range from 0 to 180◦ . It is kept fixed during relaxation. The loss of one Mg-O bond for geometries with a dihedral rotation close to 90◦ is visible. **Figures 5C–F** show the investigated molecules. The systems were derived from the BipO2 molecule with a few modifications; this includes: **C**) combination of five- and six-membered ring–C10NH7O, **D**) and **E**) a different number of carbonyl groups on the phenyl-like rings and **F**) replacing carbonyl groups with R2C = S functional groups (BipS4). **Figure 5G** shows the changes in interaction strength (1Ead) as a function of change of dihedral angle γ for the structures shown in **Figures 5C–F**. As seen above, the change of interaction strength defines the change in voltage. However, a slight bend in the

with carbonyl as redox-active group. (A) 1,1′ -Bi(1,4-cyclohexadiene-1-yl) -2,2′ ,6,6′ -tetrone (BCT) with chosen rotational axis (black vertical line) and angle (arrow). The initial structure is given with red O atoms and the final position with a changed dihedral angle between 6-membered rings indicated by blue O atoms on the ring. (B) Superposition of investigated rotated structures of BCT 0◦ to 180◦ in steps of 30◦ . Overview of molecules investigated for dihedral strain design: (C) C10NH7O, (D) C12H8O<sup>2</sup> (BipO2), (E) BCT, and (F) C12H6S4. (G) Change of interaction strength with a Mg atom for different dihedral angles γ.

bond between the rings can cause differences in the interaction strengths at 0 and 180◦ due to a reduced structural symmetry.

In general, the tested molecules with carbonyl groups yield larger interaction strengths with Mg. Between BCT and BipO2, the differences are small. They have similar maximum voltages (e.g., 2.1 V for BCT). In contrast to BipO2, BCT and Mg interact via two carbonyl groups in each of the phenyl-like rings. At dihedral rotations of more than 90◦ , the distance between Mg and O atoms will increase again in BCT leading to significantly smaller 1Ead compared to BipO2. The other molecules have smaller maximum binding strengths. The interaction energies (voltages) of BipS4 and C10NH7O with Mg are −1.7 eV (0.1 V) and −3.1 eV (0.8 V), respectively.

For all tested molecules, the curves reach the point of maximal change of the interaction energy at a twisted structure with a dihedral angle changed by around 90 or 180◦ . This can be caused by the presence of one or two redox-active groups at the rotated ring or the bend structure due to Mg attachment on one side of a molecule. All molecules but C10NH7O significantly undergo a bending away from the attached Mg atom, in addition to the change of dihedral angle of the molecular structure. This can result in additional strain during the rotation. C10NH7O has a dihedral angle of 0◦ at full relaxation when Mg is attached. At a rotation of 180◦ , the distance between the redox-active groups and Mg is maximized and the interaction energy decreases by 0.8 eV. A similar decrease was observed for the C12H6S<sup>4</sup> complex. For both complexes, the absolute decrease is smaller than for the MgBipO2 case since the binding strength is weaker. Although the C12H6S<sup>4</sup> complex has two redox-active groups on each 6 membered ring, the maximal change of the interaction energy is not found at 90◦ but at 180◦ caused by a slightly bent structure and the changed interaction between S atoms. In all these cases, the strain can drastically decrease the interaction strength between molecule and Mg atom, and the interaction strength can be reduced by 80 % by torsion on the molecular scale.

Overall, the influence of the relative orientation of molecular planes and their redox-active groups, either within the same molecule or same complex but different molecules, can be a significant factor for changing the voltage-capacity curve if mechanical control on the molecular level can be achieved. In all tested cases, the interaction strength decreases between the molecules and the Mg atom leading to a lower theoretical voltage. While the molecular torsion does not increase the voltage, it could potentially be used to lower the charging voltage due to the weaker interaction strength. At certain dihedral angles, the binding strength can change by up to 2– 3 eV translating to a voltage reduction of 1 to 1.5 V. Large steric hindrance can be expected that prevents a 180◦ rotation for practical cases.

### Design of Coordination Environment of Mg Cations

Mg atoms can bond in different ways to one or more redox-active groups of the active electrode material. They may also experience interaction with other molecules present in the electrode. Already in the example of the linear strain in the organic PAD-Mg complex shown above, the interaction strength with Mg differs when one or when two molecules simultaneously attach via a carbonyl group to Mg. Previously it was reported that the chemical environment can increase the voltage for organic Li and Na ion batteries, for instance, seen in the increase by ca. 1 V due to bulk phase (Manzhos, 2019 and references therein). Thus, it is not surprising if this also applies to organic electrode materials for MIB.

We study the effect of the local environment around the Mg atom created by molecules with carbonyl groups, that is the coordination of a different number and arrangement of carbonyl groups to a central Mg atom. The effect of a different number of redox centers in the same type of geometry was investigated by comparison between the benzoquinone (tBQE) and 1,2-benzoquinone (BQE). BQE and tBQE are structural isomers (C6H4O2) that allow us to probe the effect of the intramolecular carbonyl distance on the voltage. The effect of a different chemical character was tested by comparison between the tBQE and phenyloxidanyl molecule (C6H5O). Three different types of geometries were selected to test the influence of the arrangements, that are: linear, trigonal and tetragonal, displayed in **Figure 6**. It should be noted that in most cases, molecular



systems without a central Mg atom show weak or no binding in our calculations, partly because we do not include van-der-Walls interactions (on purpose, as we are interested in electronic effects). Furthermore, we adapt Equation (2) by substitution E mol <sup>r</sup> with w Emol <sup>r</sup> where w is the number of molecules in the final geometry.

The computed interaction energies and resulting theoretical voltages for the fully relaxed systems (structures shown in **Figure 6**) are presented in **Table 1**. tBQE has voltages of 0.2 to 1.8 V when more than one molecule coordinates to the Mg cation while a single BQE molecule results in a formally negative voltage of −0.5 V. Phenyloxidanyl has a voltage of ca. 1.7–2.8 V that is higher than those of tBQE in comparable configurations possibly due to its radical character. The BQE based configurations result in even higher voltages (0.3 V for one molecule and 2.2–2.8 V for linear to tetragonal configurations) than for tBQE and are of similar magnitude as for phenyloxidanyl. Although BQE does not have a radical character, the increase in voltage compared to BQE could arise through the interaction of the positive Mg cation and the two negatively charged carbonyl groups of each BQE. A possible consequence for the design of active Mg cathode material is to create a chemical environment for the Mg atom with a large number of redox-active groups. The results show that the type of coordination geometry can lead to a, e.g., max. voltage of 2.8 V for BQE that is an increase of 2.7 V compared to a single molecule. The number of redox centers that coordinate to Mg is a key factor to achieve this high voltage and it can change the voltage by more than 2 V. Interestingly, chemical locality design can change a non-working electrode, e.g., single molecule tBQE, into a working electrode, e.g., a trigonal complex of tBQE.

#### A Promising Molecular Design

The idea of combining structural and electronic design is applied to a complex ring system, a cycloparaphenylene molecule consisting of three fused biphenyl-like units (C36H24), shown in **Figure 7**. This molecular system combines the previously described effects of dihedral rotation, chemical environment, and it is possible to introduce electronic modifications, for example, by functionalization. We discuss possible modifications to an inner (with six H atoms, not shown) and an outer ring (with 18 H atoms). The bonds between the phenyl rings allow rather easy partial rotational changes. Up to six carbonyl groups can be in the inner ring. Then, the O atoms can take, e.g., a linear, trigonal, square-planar, tetragonal and bipyramidal-like symmetry for a central Mg ion inserted into the center of the molecule. Regarding the relative positions of the O atoms, these geometries are similar to the one shown in the section Design of Coordination Environment of Mg Cations. Of course, different combinations of sites of the carbonyl group are possible for two to four O atoms in the inner ring. The optimized geometries of the ring molecule with one to six carbonyl groups and an inserted Mg atom are shown in **Figures 7A–F**, respectively.

**Figure 8** shows the voltage vs. Mg2+/Mg computed for the insertion of one Mg atom in the inner ring of the C36H<sup>24</sup> molecule and its derivatives with different numbers of O atoms. There is no strong interaction between Mg atoms and the C36H<sup>24</sup> molecule (no O atoms), resulting in a theoretical voltage of −0.75 V, i.e., the molecule is not suitable for Mg storage. The first carbonyl group increases the interaction strength and results in a slightly positive voltage. The most significant increase in the voltage is observed when increasing the number of inner carbonyl groups from one to two. This is similar to the cases of individual molecules with their carbonyl groups around the Mg atom. Introduction of more inner ring carbonyl groups increases the voltage to a value of slightly more than 2.53 V vs. Mg2+/Mg. Depending on the number of carbonyl groups, the structure of the molecule exhibits different dihedral angle rotations between six-membered rings when the Mg atom is inserted (the strain energy is estimated at 2.7 eV). In contrast, the presence of carbonyl groups without an inserted Mg atom has a minor effect on these angles.

Introducing carbonyl groups in the inner ring can serve two purposes. First, it introduces redox-active groups. Choosing O atoms is obvious since carbonyl groups are known to result in working organic electrode materials, e.g., BQE. Second, the O atoms have a larger electronegativity than H atoms, thus they have an electron-withdrawing effect and they can stabilize the LUMO energies in organic compounds. From zero to six O atoms, the LUMO decreases from −1.43 to −3.81 eV, a change of −2.38 eV. As a result, the electronic effect on the voltage increase is at most 1.19 V while the computed voltage change is 3.28 V. This indicates that also for this system, electronic effects alone cannot fully explain the voltage increase and the number and character of bonds formed between the Mg atom and the carbonyl groups must be considered.

Outer ring functionalization gives another example of the interplay between electronic and structural effects on the voltage. We explore functionalization with F and with O atoms, and the resulting effect on the voltage for the insertion of one Mg atom. In this way, the contribution from the chemical locality (inner ring structure) and electronic changes through functionalization can be combined. The former option (substitution with F atoms leading to C36F18O6, structure is shown in **Figure 9**) can increase the theoretical voltage to 3.51 V vs. Mg2+/Mg (an increase by 1 V) at full outer ring fluorination and the latter option (O atoms– C36O24) results in a maximal voltage of 3.17 V (an increase by 0.64 V). An overview of voltages and voltage changes between the F-functionalized and the non-functionalized ring structure

FIGURE 7 | Optimized structures of MgC36OnH24−<sup>n</sup> n = {1,2,3…6} where n = 1 (A) increases to n = 6 (F).

function of the number of O atoms in the inner ring of the molecule.

is shown in **Figure 10**, top and bottom panel, respectively. The right panel of **Figure 10** shows the voltage changes depending on the number of carbonyl groups in the inner-ring. The LUMO of the F functionalized molecule is stabilized by 1.46 eV and the O functionalized one is stabilized by 2.9 eV. The counterintuitive relation between an increase in voltage and LUMO stabilization requires to extend a model that is only based on LUMO energy changes to understand the atomistic reasons behind the observed effects.

While the change of LUMO energy can often explain any voltage increase by electronic effects, we introduce a contribution accounting for the chemical environment to extend the LUMO based empirical model. Since F atoms have a larger electronegativity than O and H atoms, a charge redistribution can take place when the molecule is functionalized with F atoms. To consider this in a simple approach, we measure the charge redistribution on the redox-active atoms (here the O atoms in the inner ring). Since we are only interested in changes due to functionalization of the outer ring, the charge redistribution can be estimated by the Mulliken charges.

We provide a simple model based on the LUMO energy changes and the changes in the Mulliken charges of the redoxactive atoms (nearest neighbor to inserted Mg atom) to estimate a voltage increase for Mg batteries. The model is based on the assumption that the energy level of the LUMO and the Coulombic charges estimated as point charges surrounding a central Mg ion will contribute to the interaction strength. Furthermore, we assume that the change of the voltage is

proportional to the change of LUMO energy, and that it is indirectly proportional to the averaged change of charge on the redox-active centers due to functionalization.

The linear regression (shown as solid blue line in the bottom panel of **Figure 10**) for organic Mg electrode materials operating by reduction confirms a relatively linear trend over the ratio of change in LUMO eigenvalue and change in Mulliken charges on the redox-active groups that are involved in the redox process. The suggested model can also explain our previous observation that electronic modification in the coronene molecule does not result in increased voltages for Mg batteries while it affects the interaction with Li and Na ions (Lüder et al., 2017b). Furthermore, the model was confirmed with the result of the O-functionated ring structure that is indicated as red dot in **Figure 10** and it may find application in other systems, too.

Among all investigated structures, the C36O<sup>24</sup> molecule might be the most promising one for further investigation aiming to optimize its potential vs. Mg2+/Mg. It has a conceptual similarity to oxocarbon salts (Zhao et al., 2016) which showed promise for monovalent batteries. For this system, we computed multiple Mg atom attachments since the O functionalization introduced more carbonyl groups that can be used to store more Mg atoms. The capacity corresponding to the attachment of up to 5 Mg atoms results still in voltages of ca. 2 V or more vs. Mg2+/Mg. As the attachment takes place at the outer ring carbonyl groups, these voltages are a lower estimate and chemical locality of bulk materials constituted by neighboring molecules could further increase the voltage at higher capacities.

#### CONCLUSIONS

Organic Mg ion batteries are still in their early stages of development. Specifically, it is challenging to design insertion type active high-voltage electrode materials; many types of molecules that show good performance in monovalent batteries do not work for Mg ion batteries and do not bind Mg. Even for molecular materials with appreciable interaction energy with Mg, bivalency means very low voltages. In this work, we have explored computationally design strategies that could be used to increase voltages, applying them to several prototypical molecular systems. We considered electronic structure design, structural changes/functionalizations, strain design as well as the design of Mg coordination environment.

We find that it is possible to increase the voltage of organic electrodes for Mg batteries by several volts by a combination of electronic and structural design. Electronic structure design based on LUMO stabilization, however, shows severe limitations to voltages of not more than 2 V for most materials. In particular, we demonstrate that an increase of voltage by 4.3 V to a maximal voltage of 3.5 V vs. Mg2+/Mg can be achieved when electronic effects are combined with engineering of the local chemical environment. Designing coordination environment for Mg atoms can change the voltage by 2.0 to 2.5 V already for small molecules and appears to be a promising strategy.

Three types of strain that are linear strain, bending and dihedral strain can influence the interaction strength to Mg; here, we found they lead to lower interaction energies/voltages. Strain design may find potential application in lowering the charging potential if mechanical control at the molecular level is achievable. A reduction of 0.5 V for bending and 1–2 V for linear and dihedral strain may be achievable.

Based on the results obtained with the ring structures that combine electronic effects with effects of the chemical locality, we provide an empirical model based on two computable molecular properties, LUMO energy change and changes of Mulliken charges, to estimate the voltage change due to structural modifications such as fluorination.

In this paper, we explored what would be theoretical limits of voltage increase by modifications of molecular and electronic structure. Some of these modifications could be achieved in a way which is intrinsic to a single molecule (e.g., doping, use of functional groups causing steric hindrance to induce strain, or coordination to several redox-active centers of the same molecule) and is expected to be stable under battery cycling. Other modifications that we explored (e.g., linear or dihedral strain or coordination to redox-active centers of several molecules) would require development of practical design strategies to stabilize these environments (e.g., anchoring (Chen and Manzhos, 2015) to approach the theoretical limits delimited here.

#### REFERENCES


#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

### AUTHOR CONTRIBUTIONS

JL: calculations, data analysis, and text. SM: conceptualization, data analysis, and text.

#### FUNDING

JL acknowledges funding through the Ministry of Science and Technology (MOST) Taiwan under grant MOST108-2112-M-110-001-MY2 and was grateful to the National Center for Highperformance Computing for computer time and facilities.

#### ACKNOWLEDGMENTS

We thank Dr. Yingqian Chen for performing some test calculations and for discussions.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2020.00083/full#supplementary-material


sodium ion batteries. Adv. Mater. 24, 3562–3567. doi: 10.1002/adma.201 201205


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

Copyright © 2020 Lüder and Manzhos. 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.

# Building High Power Density of Sodium-Ion Batteries: Importance of Multidimensional Diffusion Pathways in Cathode Materials

#### Mingzhe Chen, Yanyan Zhang, Guichuan Xing and Yuxin Tang\*

*Institute of Applied Physics and Materials Engineering, University of Macau, Macau, China*

Emerging sodium-ion batteries (SIBs) devices hold the promise to leapfrog over existing lithium-ion batteries technologies with respect to desirable power/energy densities and the abundant sodium sources on the earth. To this end, the discoveries on novel cathode materials with outstanding rate capabilities are being given high priority in the quest to achieve high power density SIBs devices, and the multi-dimensional Na<sup>+</sup> migration pathways with low diffusion energy barriers are crucial. In light of this, the recent development of Prussian blue analogs and sodium superionic conductor (NASICON)-type materials with 3D Na<sup>+</sup> diffusion pathways for building high power density NIBs are provided in this perspective. Ultimately, the future research directions to realize them for real applications are also discussed.

#### Edited by:

*Cheng Zhong, Tianjin University, China*

#### Reviewed by:

*Xifei Li, Xi'an University of Technology, China Shiyou Zheng, University of Shanghai for Science and Technology, China*

> \*Correspondence: *Yuxin Tang yxtang@um.edu.mo*

#### Specialty section:

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

Received: *12 December 2019* Accepted: *18 February 2020* Published: *28 February 2020*

#### Citation:

*Chen M, Zhang Y, Xing G and Tang Y (2020) Building High Power Density of Sodium-Ion Batteries: Importance of Multidimensional Diffusion Pathways in Cathode Materials. Front. Chem. 8:152. doi: 10.3389/fchem.2020.00152* Keywords: high power density, multidimensional diffusion pathways, cathode materials, sodium-ion batteries devices, materials design

### INTRODUCTION

Although Li-ion batteries have received tremendous success in electronic and portable devices, the high price of lithium due to its limited and unequally distributed resources (**Figure 1A**) may hinder its further application in novel large-scale electrochemical energy storages systems (EESs) (Choi and Aurbach, 2016; Hwang et al., 2017; Vaalma et al., 2018). The rapid development of renewable energy sources such as wind energy, solar energy, and tide energy requires more cost-efficient and highly reliable large-scale EESs. SIBs have been widely considered as promising candidates for next-generation low-cost energy storage devices due to the almost unlimited resources of sodium (Li et al., 2017; Chen et al., 2019b,c). Due to the large atomic radius of Na (0.97 Å for Na<sup>+</sup> and only 0.68 Å for Li+), however, hurdles that hinder the further applications of SIBs are mainly found in their sluggish sodium diffusion kinetics and large volume expansion, which are both critical parameters for revolutionary power density performance that can meet the real needs of power grids and large-scale EESs (Dai et al., 2017; Chen et al., 2019a). Therefore, it is important to explore novel cathode materials for SIBs with improved Na kinetics. From the solid-state diffusion perspective, the dimensions of diffusion pathways are the extremely important factors determining the diffusion coefficient of Na<sup>+</sup> as well as the diffusion energy barriers. For facile and rapid ionic transportation, the cathode materials that possess three-dimensional (3D) Na<sup>+</sup> migration pathways accompanied by low diffusion energy barriers are promising for building the high-power density SIBs (You et al., 2016; Chen S. et al., 2017). To this end, two emerging groups of Prussian blue analogs (PBAs) and sodium superionic conductor (NASICON)-type materials are recently

developed. They both possess open 3D stable frameworks, exceptionally high ionic conductivities, and 3D sodium diffusion pathways. More importantly, these two types of materials usually undergo minimal volume changes during cycling, and their structures are highly tunable with different electrochemical properties (Lee et al., 2014; Wang B. et al., 2018). We compared the five important aspects for the real applications of these two types of materials in **Figures 1B–E**. We have witnessed an outstanding increase in publications regarding cathode materials with 3D diffusion pathways (Kim et al., 2012; You et al., 2016; Fedotov et al., 2018; Hurlbutt et al., 2018). Specifically, strategies such us nanosizing particles, introducing a uniform carbon coating/carbon matrix, and heteroatom doping are heavily used for better electrochemical properties (**Figure 1F**). These strategies have been developed to regulate the microstructures and surface atomic configurations to accelerate the charge transfer and kinetic properties as the same time.

In this perspective, we highlight (i) the specific key issues that are hindering the wider applications of these 3D Na<sup>+</sup> diffusion pathway materials, (ii) the recent novel and potential strategies to achieve high-performance SIBs, and (iii) the approaches to achieve higher energy/power densities and long-term cycling stability, and (iv) the new emerging SIB devices. Our perspective sharps the future development of cathode materials for the highpower density of SIBs and provides effective methodologies for other related next-generation energy storage devices.

### PBAs WITH 3D Na<sup>+</sup> DIFFUSION PATHWAYS

Low-cost Na-hosting Prussian blue analogs usually possess the general structural formula of NaxM2[M1(CN)6]y·nH2O (0 ≤ x ≤ 2, 0 ≤ y ≤ 1), where M<sup>1</sup> and M<sup>2</sup> are transition metal such us Mn, Fe, Co, etc. (Li et al., 2015). Within this structure, PBAs allow various modifications of their chemical compositions without changing the overall crystalline structure. Normally, most of the studied PBAs occurs as cubic phases with 6-fold coordination of FeII,III-O<sup>6</sup> octahedra participating in the redox reactions (Xiao et al., 2015). Theoretically, PBAs allow a maximum number of two Na<sup>+</sup> ions per formula unit if both the M<sup>1</sup> and the M<sup>2</sup> transition metal elements are electrochemically active, which can lead to a potential capacity over 170 mAh g−<sup>1</sup> , completely comparable with all other kinds of cathode material for SIBs (Wu et al., 2017). Such high capacity is not easy to achieve, however, since the defects, vacancies, and water molecules will greatly influence the crystal structures of PBAs. Their zeolitic water molecules tend to occupy the interstitial sites that are essential for the sodium ion diffusion. The ideal situation is that all the zeolitic water can be completely removed, while the coordination water can be retained to render structural stability. In addition, the poor intrinsic electronic conductivity of PBAs also remains a great challenge for better C-rate performance, although they possess 3D diffusion pathways. From the previous density function theory (DFT) calculations, the intercalation of Na<sup>+</sup> with a larger ionic radius will incur the cost of increased insertion

potential, since the steric interactions have a close correlation with the hydration energy. Both sodium and potassium possess sufficiently high redox potentials that an appropriate energy density can be secured. The role of water in ionic conductivity is still an open question for all researchers. A paddlewheel mechanism was previously proposed, in which the alkali metal ions move through channels and vacancies (Wessells et al., 2011). Coordinated water is beneficial to facilitate ion migration, while zeolitic water has a negative influence, obstructing improvement of ionic conductivity. Both computational and experimental insights into the effects of the presence of water in PBAs are urgently needed.

One promising approach is to synthesize and optimize the Na-rich phase, Prussian white. When M<sup>2</sup> is Mn and M<sup>1</sup> is Fe in NaxM2[M1(CN)6]y·nH2O (0 ≤ x ≤ 2, 0 ≤ y ≤ 1), a higher content of Na can be achieved. The valence of Fe is +2, however, and it is vulnerable to being oxidized to +3, so an inert atmosphere is required during both fabrication and centrifugal separation processes. By facile sodium citrate added precipitation method, Na-rich monoclinic phase NaxMnFe(CN)<sup>6</sup> was successfully synthesized by Shen et al. (2018). Uniform microsize cuboid particles were obtained, and all the involved elements were uniformly distributed (**Figures 2A–C**). The high reversible capacity of 133.1 mAh g <sup>−</sup><sup>1</sup> was obtained, and satisfactory rate performance was achieved from 0.1 C to 10 C (1 C = 150 mA g−<sup>1</sup> ), as shown in **Figures 2D,E**. The uniform cuboid particle morphology provides extra convenience for fast ion diffusion apart from the original 3D pathways. In recent years, rhombohedral PBAs have been receiving more and more attention, since the concentration of coordinated water can be reduced due to the rhombohedral lattice symmetry, which is particularly favorable for the Na-rich phases. A zero-strain nickel hexacyanoferrate (NiHCF) was introduced by Ji et al. (2016) using a simple coprecipitation method. It delivered a high operation voltage, good cycling performance, and superior C-rate capability. They found that the unexpected high operation voltage was mainly attributable to the asymmetric residence of Na<sup>+</sup> ions along with the low charge density around Fe2+. Different charge density analyses indicated that the Na<sup>+</sup> ions in cubic-structured NiHCF (c-NiHCF) tend to stay exactly in the center of interstitial channels (**Figure 2F**), while for rhombohedral-structured NiHCF (r-NiHCF), the Na<sup>+</sup> ions prefer to asymmetrically stay at N-coordinated corners (**Figure 2G**). Also the Na-N distance in r-NiHCF is shorter than that in c-NiHCF. Dramatic electron polarization round N atoms occurs, leading to a charge redistribution between adjacent Ni2<sup>+</sup> and Fe2<sup>+</sup> ions.

Therefore, the operation voltage of r-NiHCF is slightly higher than that of c-NiHCF. This phenomenon also facilitates fast sodium diffusion with excellent C-rate performances. In addition, the open-channel 3D diffusion pathways can guarantee satisfactory low-temperature performance. You et al. synthesized the Na2FeFe(CN)<sup>6</sup> materials with carbon nanotubes (CNT) to achieve potential low-temperature properties (You et al., 2016). The well-decorated CNT network provides excellent electrical contact with the current collector. In addition, they performed a DFT study and found that the evenly distributed enriched 24d sites can provide uniform 3D Na<sup>+</sup> diffusion and feasible diffusion pathways, as shown in **Figures 2H,I**. There, PBAs are ideal candidates for low-temperature SIBs (Song et al., 2015; Peng et al., 2018). Further exploration should be focused on the inherent relationship between the crystal structure and the vacancies and water, and how these factors influence the sodium diffusion under various circumstances also needs to be urgently understood in detail.

### POLYANIONIC COMPOUNDS WITH 3D Na<sup>+</sup> DIFFUSION PATHWAYS

The polyanionic-type cathode materials are one of the important branches among all the kinds of available electrodes for SIBs. Their merits are obvious and distinctive, such as outstanding cycling stability, high safety, and suitable operating voltages (Lu et al., 2017; Chen et al., 2020). Most of the polyanionic compounds consist of the corner- or edge-sharing MO<sup>6</sup> (M = transition metal) octahedra with connected XO<sup>4</sup> (X = P, S, Si, etc.) tetrahedra. This unique structure can provide sufficient stability and ionicity around the MO<sup>6</sup> octahedra, and, in turn, the ionic conductivity can be sustained within its 3D framework (Guo et al., 2017). Their sturdy framework can also render only small or even tiny volume shrinkage during cycling, since the topotactic reaction mechanism is dominant. Most of the polyanionic compounds are thermally stable due to their non-flammable nature (Ni et al., 2017). Also, polyanionic compounds normally possess high initial cycle Coulombic efficiency (ICE), which is an important parameter when making full cells. Nevertheless, most of the polyanionic compounds suffer from inherent low electronic conductivity, and therefore a well-decorated carbon coating and network are essential for their C-rate performance. According to the previous report, the majority of polyanionic composites possess one- or two-dimensional (1D or 2D) sodium diffusion pathways (Kim et al., 2013, 2015; Chen M. et al., 2017; Panigrahi et al., 2017; Chen et al., 2018; Fang et al., 2018; Zhu et al., 2019). The low-dimensional sodium diffusion pathways cannot guarantee satisfactory C-rate capabilities, especially in low temperature environments. Therefore, it is necessary to highlight those polyanionic compounds that possess unique 3D sodium diffusion pathways that are conducive toward real applications of SIBs and relevant energy storage devices. One class of these polyanionic composites comprises the sodium superionic conductor (NASICON)-type materials. Due to their high ionic conductivities, the NASICON-type materials also have been investigated as solid-state electrolytes (Li et al., 2018; Lu et al., 2018). Numerous papers have reported various NASICON-type materials, including both cathodes and anodes for SIB devices (Bui et al., 2015; Wang et al., 2017; Gao et al., 2018; Hu et al., 2018; Wang E. et al., 2018). Recently, we have introduced a new Fe-based polyanionic compound, Na4Fe3(PO4)2(P2O7), which can be classified as a new NASICON-type material (Chen et al., 2019a). With the help of ethylenediaminetetraacetic acid (EDTA) as the complexing agent in a sol-gel approach, the

*(Continued)*

FIGURE 2 | (K) High-angle annular dark-field (HAADF) image with line profile in the inset, and (L) STEM-EDS mapping of as-obtained Na4Fe3(PO4)2(P2O7) (NFPP) nanoplate. (M) C-rate performance and cycling stability (inset) of NFPP-E nanoplate. (N) Galvanostatic intermittent titration technique (GITT) curve of NFPP-E electrode and corresponding calculated diffusion coefficient of Na<sup>+</sup> ions. Reproduced from paper (Chen et al., 2019a) with permission from the Nature Publishing group. (O,P) Two typical 3D sodium diffusion pathways in Na3V(PO3)3N material and the (Q) corresponding activation barriers. Reproduced from paper (Kim et al., 2017) with permission from the American Chemical Society. (R) NASICON-type Na3V2(PO4)<sup>3</sup> material with energy isosurface, showing interweaving of 3D Na<sup>+</sup> pathways and (S) migration pathways of Na<sup>+</sup> in the open 3D framework involving M1-M2-M1 hopping. Reproduced from paper (Wong et al., 2017) with permission from the Royal Society of Chemistry.

Na4Fe3(PO4)2(P2O7) nanoplates (denoted as NPFF-E) can be well-obtained with a high degree of crystallinity, as shown in **Figures 2J,K**. All the detected elements are uniformly distributed in one NPFF-E particle (**Figure 2L**). Outstanding C-rate performance can be obtained, and satisfactory cycling stability can be achieved as well (**Figure 2M**). In addition, we also determined the sodium diffusion coefficient through a galvanostatic intermittent titration technique (GITT) study. From **Figure 2N**, the sodium diffusion coefficients range from 10−<sup>13</sup> to 10−<sup>10</sup> cm<sup>2</sup> s −1 , which is highly competitive with other recognized NASICON-type materials. These high diffusion coefficients are critical for low-temperature performance.

Another newly recognized family of NASICON-type cathode materials with 3D sodium diffusion pathways comprises the Vbased N-substituted polyanionic materials. Kang et al. recently reported a new member of the 4 V-class of zero-strain cathode materials, Na3V(PO3)3N, with unique 3D diffusion pathways for Na<sup>+</sup> ions (Kim et al., 2017). One PO3N tetrahedron is connected to three VO<sup>6</sup> octahedra, leading to a strong inductive effect which, in turn, influences the operation voltage during charge/discharge. The relevant sodium ion diffusion pathways and diffusion energy barriers were calculated via a DFT study, and illustrations are presented in **Figures 2O–Q**. This exciting discovery has opened up new interest in the highvoltage N-substituted cathode materials with unique unexpected 3D sodium diffusion pathways. As for the well-recognized NASICON-type Na3V2(PO4)<sup>3</sup> material, great attention has been extensively paid to it due to its outstanding merits (high operation voltage, acceptable specific capacity, and excellent cyclability). Its 3D sodium diffusion properties have been well-investigated, with the 3D network involving M1-M2- M1 hops and low energy barriers. Each Na in an M1 site is actively connected to six M2 sites. The corresponding illustrations are shown in **Figures 2R,S**. In addition, we would like to see more polyanionic compounds discovered with 3D Na<sup>+</sup> diffusion pathways in the search for advanced SIB storage devices with high power densities. It is suggested that further exploration should be concentrated on improving the reaction kinetics based on a more comprehensive understanding. Nevertheless, due to the intrinsic low electron conductivity of polyanionic material, how to maintain the well-designed carbon matrix or network and large-scale synthesis is still very challenging.

### DISCUSSION AND PERSPECTIVES

The ultimate goal for advanced sodium ion battery devices has never changed, that they should be thoroughly competitive with the lithium ion battery in energy density, power density, and lower overall manufacturing cost (Tang et al., 2017, 2018; Zhang et al., 2017, 2019; Guo et al., 2020). For fullcell sodium ion battery devices, the emphasis always falls on the cathode parts, since they possess key parameters such as high enough operation voltages, large theoretical capacity, and most of all, acceptable high-rate capability. Materials with 3D sodium diffusion pathways are a promising branch among all the reported ones, and their distinctive physical and chemical properties deserve more integrated investigations and a deeper and more comprehensive understanding. Both the PBAs and the polyanionic compounds with 3D sodium diffusion pathways have shown outstanding fast kinetics, leading to superior C-rate performance at various current densities. Further investigations are suggested to be focused on the improvements of their intrinsic kinetic properties as well as mass-production issues. Specifically, for PBAs, as mentioned above, the structurekinetics-performance relationship is subject to the influence of each crystallographic component: binary carbon- and nitrogencoordinated transition metal ions, the density distribution of electronic states at each sodium site, and the vacancies created by either zeolitic water or coordinated water. When dealing with the specific concerns about the actual Na<sup>+</sup> diffusion channels and ionic conductivity, the bond length between carbon and carbon-coordinated transition metal elements is one of the key parameters for channel size, which, in turn, affects the ionic conductivity. The content of vacancies also has a strong influence on the Na<sup>+</sup> diffusion channels and ionic conductivity. The appropriate amount of vacancies will provide extra channels for Na<sup>+</sup> diffusion without cracking the local crystal structure. Nevertheless, how these parameters quantitatively influence the ionic conductivity as well as the crystal stability is still poorly understood. Another important issue that should not be neglected is the electronic conductivity. Up until now, only a few papers have focused on improving electronic conductivity directly by using the generalized gradient approximation with the Hubbard correction (GGA + U) method. Experimentally speaking, the electronic conductivity is strongly correlated with all the factors mentioned above and needs to be determined in an equational relationship. In the case of polyanionic compounds, the electronic conductivity is the most troublesome problem, and a well-designed carbon coating and carbon network are necessary even when the materials have high ionic conductivities. Therefore, how to maintain the delicate carbon network in a large-scale manufacturing procedure is a critical issue for the real application of polyanionic composites. Unlike the PBAs, only some of the reported materials possess 3D sodium diffusion pathways, although almost all the polyanionic compounds have open 3D frameworks. Compared to the PBAs, the activation energy barriers of these polyanionic compounds are slightly higher. Heteroatom doping can be considered as an effective approach to alter the pristine electron state density distributions around both MO<sup>6</sup> octahedra and XO<sup>4</sup> tetrahedra, which, in turn, may lower the charge barriers around each sodium site. The transition metal migration is also poorly understood to date [such as for Na4M3(PO4)2(P2O7) (M = Fe, Mn, etc.)], and how the phenomenon affects the related ion diffusion pathways and ionic conductivity should be determined in the future. It is also important to continuously discover more novel polyanionic cathode materials with 3D sodium diffusion pathways that have higher capacity and operation voltages. Besides, another important research direction is that how to combine both the PBAs and polyanions with the solid-state electrolyte properly so that their unique 3D diffusion properties can be more significantly magnified.

In **Table S1**, we summarize some of the reported cathodes for SIB devices with unique 3D diffusion pathways and their corresponding diffusion coefficients and activation barriers. These types of cathodes are especially important for highpower SIB devices for real applications. We can reasonably anticipate more technical breakthroughs, such as precise control of the water content in PBAs, ways to maintain a welldecorated carbon network at the industrial level, efficient organic/solid state electrolytes to fully achieve their theoretical capacities, and, in turn, the advanced SIB devices can be further

#### REFERENCES


designed and developed from a more technology based point of view.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

#### AUTHOR CONTRIBUTIONS

MC, YZ, and GX prepared the manuscript. YT supervised the project.

#### FUNDING

This work was funded by The Science and Technology Development Fund, Macau SAR (Grant Nos. 0057/2019/A1 and 0092/2018/A2).

#### ACKNOWLEDGMENTS

The authors thank Dr. Tania Silver for critical reading of the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem. 2020.00152/full#supplementary-material


for fast charging high voltage battery electrode. Adv. Mater 29:1701828. doi: 10.1002/adma.201701828


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

Copyright © 2020 Chen, Zhang, Xing and Tang. 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.