Substantial Impact of Charge on Electrochemical ... - ACS Publications

Article Cite This: J. Am. Chem. Soc. 2018, 140, 9127−9131

pubs.acs.org/JACS

Substantial Impact of Charge on Electrochemical Reactions of TwoDimensional Materials Donghoon Kim, Jianjian Shi, and Yuanyue Liu* Texas Materials Institute and Department of Mechanical Engineering, The University of Texas at Austin, Austin, Texas 78712, United States

Downloaded via DURHAM UNIV on July 25, 2018 at 11:13:38 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: Two-dimensional (2D) materials have attracted great interest in catalyzing electrochemical reactions such as water splitting, oxygen reduction, and carbon dioxide reduction. Quantum mechanical simulations have been extensively employed to study the catalytic mechanisms. These calculations typically assume that the catalyst is charge neutral for computational simplicity; however, in reality, the catalyst is usually charged to match its Fermi level with the applied electrode potential. These contradictions urge an evaluation of the charge effects. Here, using the example of hydrogen adsorption on the common 2D electrocatalysts (N-doped graphene and MoS2) and 3D metal catalysts, and employing the grand canonical density functional theory, we show that the charge on 2D materials can have a much stronger impact on the electrochemical reaction than the charge on 3D metals (the reaction energy can differ by >1 eV after including the charge effects). This arises from the charge-induced change in the occupation of electronic states, which is more significant for 2D materials due to their limited density of states. Our work provides a fundamental understanding of the charge effects in 2D materials, calls for re-evaluation of the previously suggested mechanisms by including the overlooked charge effects, and offers practical guidelines for designing 2D catalysts.

■

INTRODUCTION Two-dimensional (2D) materials have emerged as promising electrocatalysts.1−3 These materials feature a thickness of only one or a few atomic layer(s), providing a high ratio of surface atoms that are potentially active or can be tuned for catalysis. For example, doped graphene can efficiently catalyze the oxygen reduction reaction (ORR)1,3,4 for fuel cells and metal− air batteries, the oxygen evolution reaction (OER)1,3,4 and the hydrogen evolution reaction (HER)1−3,5,6 for water splitting, and the carbon dioxide reduction (CO2R)2,7−9 to convert greenhouse gas into more useful products. 2D transition metal dichalcogenides can also efficiently catalyze the HER2,3,10−12 and CO2R.2,13,14 The further developments of 2D electrocatalysts require an improved understanding of the catalytic mechanisms, which urge an accurate atomistic simulation method to provide insights. Density functional theory (DFT) calculations have been widely used to study 2D electrocatalysts.10,15,16 These calculations typically assume that the catalyst is charge neutral for computational simplicity (thus called charge-neutral method, cnm). Consequently, the Fermi level (EF) of the neutral catalyst changes as the reaction proceeds, due to the variation in the adsorbed chemical species. However, in reality, the catalyst is usually charged by accepting/donating electrons from/to the electrode to match the catalyst Fermi level (EF) with the applied electrode potential (μe), and the charge © 2018 American Chemical Society

typically varies as the reaction proceeds. These contradictions cast doubt on the accuracy of the widely used cnm and urge an evaluation of the neglected charge effects. Indeed, it has been shown that the electrostatic interactions (especially the electrical double layer formed on the electrode surface) play important roles in various systems, such as ionic liquid transistors,17,18 silicon electrodes,19 alkoxyamine cleavage,20 and Au(I)-catalyzed hydroarylation.21 Here to study the charge effects for 2D materials, we chose the most common 2D electrocatalystsN-doped graphene and MoS2as representative examples. As mentioned before, the N-doped graphene has been extensively reported to have a high activity especially for ORR and OER,1,3 and MoS2 is a well-known catalyst for HER.2,3,10−12 For comparison, we also included common 3D metal catalysts: Pt, Ni, Ag, and Cu. Since many reactions occur in protic solution or involve proton as the reactant, we considered the Volmer reaction22 (i.e., adsorption of proton from the solution coupled with one electron transfer from the electrode) as a representative example: H+(sol) + e− → H*, where H* indicates hydrogen adsorbed on the catalyst site (e.g., N in graphene). The free energy change (ΔG) of this reaction suggests whether the H adsorption is thermodynamically favorable or not, which is Received: March 18, 2018 Published: June 29, 2018 9127

DOI: 10.1021/jacs.8b03002 J. Am. Chem. Soc. 2018, 140, 9127−9131

Article

Journal of the American Chemical Society important to know for reactions in protic environment, as well as indicates the HER activity.22

■

RESULTS AND DISCUSSION In the charge-neutral method, the ΔG is calculated as ΔGcnm = G(H*) − G(*) − G(H+(sol)) − μe

(1)

where the subscript cnm denotes the charge-neutral method; G(H*) and G(*) are the free energies of the catalyst with and without hydrogen adsorption; G(H+(sol)) is the free energy of the proton in solution at given pH; and μe is the electron energy defined by the absolute electrode potential. Referring the μe to the electron energy in standard hydrogen electrode (SHE) (μSHE) and using the Nernst equation, eq 1 can be rewritten as a more computationally feasible form

Figure 1. Representative structures of pyridinic N-doped graphene. Pyridinic N atoms (blue) are bonded with two C atoms (brown) in the hexagon, and the edge C atoms are saturated with H (white).

ΔGcnm = G(H*) − G(*) − G(H 2(g))/2 + 0.059pH + | e| U

(2)

where U is the applied voltage versus the SHE (i.e., U = μe − μSHE). In this work, we consider the reactions at pH = 0; therefore ΔGcnm = G(H*) − G(*) − G(H 2(g))/2 + |e|U

(3)

As mentioned above, the cnm neglects the charge effects. Moreover, most calculations in the literature were performed with catalysts in vacuum for computational simplicity, thus the solvation effects were also neglected in those calculations. When considering these two effects, the ΔG becomes

Figure 2. (a) Free energy of H adsorption (ΔG) on common 2D electrocatalysts and 3D metals at pH = 0, calculated using the charge neutral method (cnm) and constant potential method (cpm). 0 and 1.23 indicate the electrode potential, U vs SHE. (b) The difference in free energy calculated by using the two methods, ΔGcpm − ΔGcnm, as a function of the electrode potential for various materials.

ΔGcpm = Gsol(H*Q2 ) − Gsol(*Q1) − G(H+(sol)) − (Q1 − Q2 + 1)μe = Gsol(H*Q2 ) − Gsol(*Q1) − G(H 2(g))/2 + |e|U − (Q1 − Q2)μe

(4)

where Q1 and Q2 are the net charges on the catalyst before and after H adsorption, which are determined by the constraint E F(*Q1) = E F(H *Q2 ) = μe

U = 0 V, the equilibrium potential for HER, and U = 1.23 V, the equilibrium potential for OER/ORR in water. Both methods show that the ΔG increases with U, which can be explained by the electron energy decrease as the potential gets higher. However, they give notably different values for 2D materials, especially at higher U. For example, at U = 1.23 V, the cnm gives a ΔG = 0.47 eV for Z and 1.79 eV for A, which suggests that H adsorption on N is thermodynamically unfavorable; in contrast, the cpm gives a ΔG = −0.25 eV for Z and −0.22 eV for A, indicating that the H adsorption is thermodynamically favorable. This can be further seen in Figure 2b, which plots the ΔGcpm − ΔGcnm as a function of the U. The |ΔGcpm − ΔGcnm| tends to increase as |U| increases and can even reach 2.01 eV for A at U = 1.23 V. This increase is due to the increase in the charge induced by matching the Fermi level with the potential (see Table S1). These results indicate that the charge effects in 2D materials are substantial and call for a reconsideration of the previous catalytic mechanisms of 2D materials suggested using the cnm. Interestingly, for all the tested 3D metals, |ΔGcpm − ΔGcnm| is less than 0.05 eV at U = 0 and 0.1 eV at U = 1.23 V, indicating that the charge effects in 3D metals are negligible. To understand why the charge effects become significantly stronger when the material’s dimension is reduced, we decomposed the ΔGcpm − ΔGcnm into the three contributions as follows:

(5)

and thus are U dependent. The subscript “cpm” emphasizes that the μe (and consequently the EF) is fixed during the reaction; thus, this method is called constant-potential method (cpm). The subscript “sol” indicates that the system is embedded in the solution, and the charge on the catalyst is balanced by the counterions in the solution. There are several approaches to perform the constant-potential calculations,23−26 and here we used the grand canonical DFT27 as implemented in the JDFTx code,28 which allows for automatic adjusting of the catalyst charge to satisfy the constraint eq 5. Further details can be found in the Supporting Information. First, we calculated the ΔG for the selected 2D and 3D electrocatalysts as mentioned above. Figure 1 shows the representative structures of N-doped graphene where N can be located in the basal plane (1N or 3N) or at the edges (Z, A, K1, or K2). Since MoS2 is known to be edge active for HER, we considered the H adsorption at its edge. Although the 2D materials are usually supported on electrically conductive 3D substrate, we follow the common practice to neglect the substrate in simulations (see the scientific justifications in the SI). For 3D metals, we considered the most stable surface, i.e., the (111) surface. The supercells used for modeling these structures are shown in Figure S1. Figure 2a lists the ΔG computed using cnm and cpm at two representative voltages: 9128

DOI: 10.1021/jacs.8b03002 J. Am. Chem. Soc. 2018, 140, 9127−9131

Article

Journal of the American Chemical Society ΔGcpm − ΔGcnm = {[Esol(H*) − Esol(*)] − [E(H *) − E(*)]} + {[Esol(H *Q 1) − Esol(*Q1)] − [Esol(H *) − Esol(*)]} + {Esol(H *Q2 ) − Esol(H *Q1) − (Q1 − Q2)μe}

(6)

The derivation is shown in the SI. The first angle-bracket term of the final expression is the solvation effect on H adsorption (ΔG1 − ΔGcnm, see Figure 3a for illustration); the second term

Figure 4. Electronic density of states (DOS) of Z (a−c) and Pt (d−f) for the neutral state (a,d) at U = 0 V (b,e) and at U = 1.23 V (c,f) vs SHE. The occupied states are shown in shadows.

filling remains nearly the same even at a high U = 1.23 V. Since the chemical reactivity is strongly affected by the occupation of the electronic states,11,29−31 the variation in the chemical reactivity for Z ought to be larger than that for Pt. Similarly, other 2D materials also show a notable change in the occupation of electronic states when they are charged to the target U (except for 1N and 3N at U = 0 V because the EF’s of their neutral states are already close to the μe at U = 0 V; see Figures S3−5); thus, they have a large change in the chemical reactivity. We attribute the large variation in DOS filling of 2D materials to their small “quantum capacitance” compared with 3D metals.31,32 Therefore, in order to have a similar charge as 3D metals (see Table S1 for the comparison of charges), the electronic state occupation of 2D materials needs to shift significantly. Note that, when referring to the DOS profile, the EF of 3D metals does not exhibit an obvious shift; however, when referring to the vacuum level, the change is significant due to the electric double layer created by the charge on metal and the counter-charge in solution that has been included in the solvation model. Although only the H adsorption is considered here, we expect that the charge effects in 2D materials can also be strong for other reactions. This is exemplified by the calculations of ORR detailed in the SI, where the cpm gives a better agreement with experiments than the cnm. Beyond the fundamental understandings of the charge effects in low dimensions, here we highlight several practical impacts: (1) our work elucidates whether various N dopants in graphene are bonded with H or not in aqueous electrochemical conditions, which are crucial to understanding the catalytic mechanisms as well as designing catalysts. As mentioned earlier, N-doped graphene has shown great promise for catalyzing various important aqueous electrochemical reactions. However, there is limited understanding of the catalytic mechanisms, which in turn impedes the further developments of the catalyst. An important question is whether various N dopants are bonded with H (or adsorb proton) before the catalytic reaction, which determines the subsequent catalytic pathways and reflects the catalytic activity of a specific dopant. Our calculations show that the “basal plane-3N” type of N dopant/site forms a strong bond with H (Figure 2) and thus is unlikely to be highly active due to the H passivation. In contrast, the K2 type of dopant/site forms a weak bond with H (Figure 2) and thus is likely to have a high activity. These results suggest that, in order to improve the overall activity of N-doped graphene, one should increase the amount of K2 type

Figure 3. (a) Schematic illustration of decomposing adsorption free energy (see eq 6 and the related text). Gray rectangles, pink rectangles, and blue dots represent catalysts, solutions, and adsorbates, respectively. (b,c) ΔGcpm − ΔGcnm and its components as a function of the electrode potential, for (b) Z and (c) Pt. (d) ΔGcpm − ΔGcnm as a function of the chemical reactivity difference for various materials.

can be understood as the chemical reactivity difference between neutral and charged catalyst (ΔG1 − ΔG2, the reactivity for adsorbing one H atom, keeping the catalyst charge to be Q1, Figure 3a); and the last term is the electron reorganization energy to adjust the charge from Q1 to Q2 (ΔG3, Figure 3a). According to this decomposition, the solvation effect is independent of U, while the other two terms depend on U. Figure 3b and 3c shows how they contribute to the ΔGcpm − ΔGcnm for Z and Pt at different U, respectively. Similar plots for other materials can be found in Figure S2. We found that the ΔGcpm − ΔGcnm is strongly correlated with the chemical reactivity difference, while the solvation effect and the electron reorganization energy have relatively small contributions (Figure 3 and Figure S2). This can be further seen from Figure 3d, which shows the ΔGcpm − ΔGcnm as a function of the chemical reactivity difference for all the materials at various U. Clearly, the 2D materials have a large change in the chemical reactivity when they are charged to the target U, which gives rise to a large |ΔGcpm − ΔGcnm|. In contrast, the change in the chemical reactivity of 3D metals is minute, and thus their |ΔGcpm − ΔGcnm| is small. To further understand why the chemical reactivity of 2D materials is much more sensitive to the charge, we compare the electronic density of states (DOS) between 2D materials and 3D metals. Figure 4 shows Z and Pt as examples. For Z (Figure 4a), the occupation of DOS changes significantly when the material is charged to the target U; in contrast, for Pt, the DOS 9129

DOI: 10.1021/jacs.8b03002 J. Am. Chem. Soc. 2018, 140, 9127−9131

Article

Journal of the American Chemical Society

(3) Deng, D.; Novoselov, K. S.; Fu, Q.; Zheng, N.; Tian, Z.; Bao, X. Nat. Nanotechnol. 2016, 11 (3), 218−230. (4) Zhang, J.; Zhao, Z.; Xia, Z.; Dai, L. Nat. Nanotechnol. 2015, 10 (5), 444−452. (5) Jiao, Y.; Zheng, Y.; Davey, K.; Qiao, S.-Z. Nature Energy 2016, 1, 16130. (6) Fei, H.; Dong, J.; Arellano-Jiménez, M. J.; Ye, G.; Dong Kim, N.; Samuel, E. L. G.; Peng, Z.; Zhu, Z.; Qin, F.; Bao, J.; Yacaman, M. J.; Ajayan, P. M.; Chen, D.; Tour, J. M. Nat. Commun. 2015, 6, 8668. (7) Wu, J.; Ma, S.; Sun, J.; Gold, J. I.; Tiwary, C.; Kim, B.; Zhu, L.; Chopra, N.; Odeh, I. N.; Vajtai, R.; Yu, A. Z.; Luo, R.; Lou, J.; Ding, G.; Kenis, P. J. A.; Ajayan, P. M. Nat. Commun. 2016, 7, 13869. (8) Wu, J.; Liu, M.; Sharma, P. P.; Yadav, R. M.; Ma, L.; Yang, Y.; Zou, X.; Zhou, X.-D.; Vajtai, R.; Yakobson, B. I.; Lou, J.; Ajayan, P. M. Nano Lett. 2016, 16 (1), 466−470. (9) Wu, J.; Yadav, R. M.; Liu, M.; Sharma, P. P.; Tiwary, C. S.; Ma, L.; Zou, X.; Zhou, X.-D.; Yakobson, B. I.; Lou, J.; Ajayan, P. M. ACS Nano 2015, 9 (5), 5364−5371. (10) Hinnemann, B.; Moses, P. G.; Bonde, J.; Jørgensen, K. P.; Nielsen, J. H.; Horch, S.; Chorkendorff, I.; Nørskov, J. K. J. Am. Chem. Soc. 2005, 127 (15), 5308−5309. (11) Liu, Y.; Wu, J.; Hackenberg, K. P.; Zhang, J.; Wang, Y. M.; Yang, Y.; Keyshar, K.; Gu, J.; Ogitsu, T.; Vajtai, R.; Lou, J.; Ajayan, P. M.; Wood, B. C.; Yakobson, B. I. Nature Energy 2017, 2, 17127. (12) Li, G.; Zhang, D.; Qiao, Q.; Yu, Y.; Peterson, D.; Zafar, A.; Kumar, R.; Curtarolo, S.; Hunte, F.; Shannon, S.; Zhu, Y.; Yang, W.; Cao, L. J. Am. Chem. Soc. 2016, 138 (51), 16632−16638. (13) Asadi, M.; Kim, K.; Liu, C.; Addepalli, A. V.; Abbasi, P.; Yasaei, P.; Phillips, P.; Behranginia, A.; Cerrato, J. M.; Haasch, R.; Zapol, P.; Kumar, B.; Klie, R. F.; Abiade, J.; Curtiss, L. A.; Salehi-Khojin, A. Science 2016, 353 (6298), 467−470. (14) Asadi, M.; Kumar, B.; Behranginia, A.; Rosen, B. A.; Baskin, A.; Repnin, N.; Pisasale, D.; Phillips, P.; Zhu, W.; Haasch, R.; Klie, R. F.; Král, P.; Abiade, J.; Salehi-Khojin, A. Nat. Commun. 2014, 5, 4470. (15) Jiao, Y.; Zheng, Y.; Jaroniec, M.; Qiao, S. Z. J. Am. Chem. Soc. 2014, 136 (11), 4394−4403. (16) Xu, H.; Cheng, D.; Cao, D.; Zeng, X. C. Nature Catalysis 2018, 1 (5), 339−348. (17) Ren, X.; Schmidt, E.; Walter, J.; Ganguly, K.; Leighton, C.; Frisbie, C. D. J. Phys. Chem. C 2017, 121 (12), 6540−6545. (18) Xie, W.; Zhang, X.; Leighton, C.; Frisbie, C. D. Advanced Electronic Materials 2017, 3 (3), 1600369. (19) Vogel, Y. B.; Zhang, L.; Darwish, N.; Gonçales, V. R.; Le Brun, A.; Gooding, J. J.; Molina, A.; Wallace, G. G.; Coote, M. L.; Gonzalez, J.; Ciampi, S. Nat. Commun. 2017, 8 (1), 2066. (20) Zhang, L.; Laborda, E.; Darwish, N.; Noble, B. B.; Tyrell, J. H.; Pluczyk, S.; Le Brun, A. P.; Wallace, G. G.; Gonzalez, J.; Coote, M. L.; Ciampi, S. J. Am. Chem. Soc. 2018, 140 (2), 766−774. (21) Lau, V. M.; Pfalzgraff, W. C.; Markland, T. E.; Kanan, M. W. J. Am. Chem. Soc. 2017, 139 (11), 4035−4041. (22) Nørskov, J. K.; Bligaard, T.; Logadottir, A.; Kitchin, J. R.; Chen, J. G.; Pandelov, S.; Stimming, U. J. Electrochem. Soc. 2005, 152 (3), J23−J26. (23) Goodpaster, J. D.; Bell, A. T.; Head-Gordon, M. J. Phys. Chem. Lett. 2016, 7 (8), 1471−1477. (24) Xiao, H.; Cheng, T.; Goddard, W. A.; Sundararaman, R. J. Am. Chem. Soc. 2016, 138 (2), 483−486. (25) Taylor, C. D.; Wasileski, S. A.; Filhol, J.-S.; Neurock, M. Phys. Rev. B: Condens. Matter Mater. Phys. 2006, 73 (16), 165402. (26) Bonnet, N.; Marzari, N. Phys. Rev. Lett. 2013, 110 (8), 086104. (27) Sundararaman, R.; Goddard, W. A., III; Arias, T. A. J. Chem. Phys. 2017, 146 (11), 114104. (28) Sundararaman, R.; Letchworth-Weaver, K.; Schwarz, K. A.; Gunceler, D.; Ozhabes, Y.; Arias, T. A. SoftwareX 2017, 6, 278−284. (29) Hammer, B.; Norskov, J. K. Nature 1995, 376 (6537), 238− 240. (30) Hwang, J.; Rao, R. R.; Giordano, L.; Katayama, Y.; Yu, Y.; Shao-Horn, Y. Science 2017, 358 (6364), 751−756.

dopant/site. Note that the conventional cnm gives misleading results due to the neglect of the important charge effects (Figure 2). (2) Our work rules out the possibility that graphitic N is active for oxygen reduction reaction (ORR). As shown in the SI, our calculations indicate that the first step of ORR on graphitic N is highly endothermic (while the cnm incorrectly predicts a moderate energy). This finding suggests that one should focus on other types of N (such as K2 type) rather than on the graphitic ones during synthesis. (3) Our work calls for reconsideration of the catalytic mechanisms by incorporating the charge effects, which have a substantial impact on the reaction energetics but are generally neglected in previous studies.

■

CONCLUSION In summary, we show that the charge on 2D materials can have a much stronger impact on the electrochemical reactions than the charge on 3D metals, which originates from the chargeinduced change in the occupation of electronic states. Our work provides fundamental understanding of the charge effects in 2D materials, calls for re-evaluation of the previously suggested mechanisms by including the overlooked charge effects, and offers practical guidelines for designing 2D catalysts.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.8b03002. Calculation details, derivation of the adsorption free energy decompositions, discussion on the charge effects for other reactions, structure models, potential dependence of H adsorption free energy and the decomposed energetics for various materials, electronic density of states for neutral and charged materials, Fermi levels of neutral materials with respect to vacuum level, charges on the materials at different potentials, structure of OOH on N-doped graphene, data used for Figures 2 and 3, and schemes of 2D materials on 3D support (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Yuanyue Liu: 0000-0002-5880-8649 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Y. L. acknowledges the support from Welch Foundation (Grant No. F-1959-20180324). Y. L. and D. K. thank the startup support from UT Austin. This work used computational resources sponsored by the DOE’s Office of Energy Efficiency and Renewable Energy and located at the National Renewable Energy Laboratory and the Texas Advanced Computing Center (TACC) at UT Austin.

■

REFERENCES

(1) Liu, X.; Dai, L. Nat. Rev. Mater. 2016, 1, 16064. (2) Voiry, D.; Shin, H. S.; Loh, K. P.; Chhowalla, M. Nature Reviews Chemistry 2018, 2, 0105. 9130

DOI: 10.1021/jacs.8b03002 J. Am. Chem. Soc. 2018, 140, 9127−9131

Article

Journal of the American Chemical Society (31) Liu, Y.; Wang, Y. M.; Yakobson, B. I.; Wood, B. C. Phys. Rev. Lett. 2014, 113, 028304. (32) Wood, B. C.; Ogitsu, T.; Otani, M.; Biener, J. J. Phys. Chem. C 2014, 118 (1), 4−15.

9131

DOI: 10.1021/jacs.8b03002 J. Am. Chem. Soc. 2018, 140, 9127−9131