Home Journals Articles Updates About
CN
image
收藏切换
Neural Network Accelerated Investigation of the Dynamic Structure–Performance Relations of Electrochemical CO2 Reduction over SnOx Surfaces
收藏切换
PDF
Lulu Li1, 2, 3, 4, Zhi-Jian Zhao1, 2, 3, 4, Gong Zhang1, 2, 3, 4, Dongfang Cheng1, 2, 3, 4, Xin Chang1, 2, 3, 4, Xintong Yuan1, 2, 3, 4, Tuo Wang1, 2, 3, 4, 5, Jinlong Gong1, 2, 3, 4, *
Research. Vol 6 Article ID 0067
Less
收藏切换
Research. Vol 6 Article ID 0067
• Research Article •
Neural Network Accelerated Investigation of the Dynamic Structure–Performance Relations of Electrochemical CO2 Reduction over SnOx Surfaces
Full
Lulu Li1, 2, 3, 4, Zhi-Jian Zhao1, 2, 3, 4, Gong Zhang1, 2, 3, 4, Dongfang Cheng1, 2, 3, 4, Xin Chang1, 2, 3, 4, Xintong Yuan1, 2, 3, 4, Tuo Wang1, 2, 3, 4, 5, Jinlong Gong1, 2, 3, 4, *
Affiliations
  • 1School of Chemical Engineering and Technology, Key Laboratory for Green Chemical Technology of Ministry of Education, Tianjin University, Tianjin 300072, China.
  • 2Collaborative Innovation Center for Chemical Science and Engineering (Tianjin), Tianjin 300072, China.
  • 3National Industry-Education Platform of Energy Storage, Tianjin University, Tianjin 300072, China.
  • 4Haihe Laboratory of Sustainable Chemical Transformations, Tianjin 300192, China.
  • 5Joint School of National University of Singapore and Tianjin University, International Campus of Tianjin University, Binhai New City, Fuzhou 350207, China.
About Author:

† These authors contributed equally to this work.

Published: 2023-03-20 doi: 10.34133/research.0067
Outline
收藏切换

Heterogeneous catalysts, especially metal oxides, play a curial role in improving energy conversion efficiency and production of valuable chemicals. However, the surface structure at the atomic level and the nature of active sites are still ambiguous due to the dynamism of surface structure and difficulty in structure characterization under electrochemical conditions. This paper describes a strategy of the multiscale simulation to investigate the SnOx reduction process and to build a structure–performance relation of SnOx for CO2 electroreduction. Employing high-dimensional neural network potential accelerated molecular dynamics and stochastic surface walking global optimization, coupled with density functional theory calculations, we propose that SnO2 reduction is accompanied by surface reconstruction and charge density redistribution of active sites. A regulatory factor, the net charge, is identified to predict the adsorption capability for key intermediates on active sites. Systematic electronic analyses reveal the origin of the interaction between the adsorbates and the active sites. These findings uncover the quantitative correlation between electronic structure properties and the catalytic performance of SnOx so that Sn sites with moderate charge could achieve the optimally catalytic performance of the CO2 electroreduction to formate.

Lulu Li, Zhi-Jian Zhao, Gong Zhang, Dongfang Cheng, Xin Chang, Xintong Yuan, Tuo Wang, Jinlong Gong. Neural Network Accelerated Investigation of the Dynamic Structure–Performance Relations of Electrochemical CO2 Reduction over SnOx Surfaces[J]. Research, 2023 , 6 (3) : 0067 - . DOI: 10.34133/research.0067
Full Text
Less
Introduction
Less
The ever-increasing CO2 concentration in the atmosphere has caused a series of problems including climate change and ocean acidification. Many countries have set the goal to achieve carbon neutrality, i.e., net zero CO2 emission, at about the middle of the 21st century. One of the potential ways to achieve carbon neutrality is CO2 electroreduction (CO2ER) by renewable electricity. It is also seen as a promising strategy to create valuable chemicals (CO, formate, and multicarbon products) due to its ambient operating conditions, environmental friendliness, and high reaction rates [1–5]. Among a variety of researched electrocatalysts, Sn-based electrocatalysts have been paid plenty of attention [6] over the past decades due to their high selectivity to the liquid product formate and the low cost of the catalysts themselves. Great endeavors have been devoted to rational catalyst design and synthesis, such as size optimization [7,8], component regulation [9–14], and electrolyte governing [15,16]. Nevertheless, Sn-based electrocatalysts for CO2ER still suffer from high overpotential, hydrogen evolution side reaction, and limited knowledge of active structures.
It is generally believed that cognition of the active sites and catalytic mechanism at the atomic level can provide theoretical guidance for the rational design of effective electrocatalysts. Numerous theoretical researches have been dedicated to exploring the intrinsic properties of Sn-based electrocatalysts for the improvement of energy efficiency and selectivity toward formate [17,18]. For instance, our previous work found that in-plane oxygen vacancy (Ov) can regulate formate production and the post-reduction stage of SnO2 corresponds to the optimal state [19]. Furthermore, previous operando technique detection by Hu et al. has demonstrated that partially reduced tin oxide (SnOx) is the catalytically active species existing in Sn-based electrocatalysts [20–22] and the oxidation state can be attributed to tuning the electrocatalytic activity of CO2ER to formate [23]. It must be highlighted that the surface structure and corresponding mechanisms are only being investigated from a nearly static perspective by the present theoretical calculation and experiment. However, as a cathode catalyst, SnOx catalyzing CO2ER is also accompanied by obvious self-reduction, leading to the ambiguous perception of SnOx structural dynamics and site diversity [24]. The dynamism of catalyst structure determines the corresponding catalytic activity [25]. Moreover, the dynamic surface structure and site distribution bring great challenges to conventional local optimization methods at the static density functional theory (DFT) level [26]. To accurately provide atom-level structure information and pave the rational way to design effective and efficient SnOx-based catalysts, a comprehensive calculation method should be developed to systematically search for the dynamic reduction process and feature structure of metal oxide in large and long-term simulation scale.
This paper describes a combined molecular dynamics simulation with neural network fitted potential (MD-NN) [27,28] and first-principles calculations to explore the SnO2 reduction process, with SnO2(110) as the starting material because it is the thermodynamically most stable surface. The indentation of structural dynamics of SnOx followed by surface reconstruction and the mechanism exploration of CO2ER and hydrogen evolution reaction (HER) are probed. A simple regulatory factor, the net charge, is identified through analyzing model sites of SnO2(110) and can be utilized for predicting and regulating the product distribution of CO2ER and HER over SnOx catalysts. This discovery is also in line with the experimental measurements of the chronoamperometry test, x-ray diffraction (XRD), and quasi in situ x-ray photoelectron spectroscopy (XPS) semiquantitative analysis. Electronic analyses illustrate that the increased net charge in the process of SnOx reduction is filled in the antibonding states between key intermediates (HCOO*, *COOH, and H*) and active sites, and the simultaneous change of adsorption strength alters the rate-determining step and the theoretical limiting potential, thus leading to the fluctuation of activity and selectivity over dynamic SnOx surfaces. These results disclose a quantitative correlation between electronic structure properties and the catalytic performance of dynamic SnOx surfaces with diverse active sites. A promising strategy is proposed to investigate the adsorbate–surface interactions and the dynamic structure–performance relation in electrocatalysis.
Results
Less
SnOx surface reduction process
MD-NN was utilized to simulate the entire SnO2 surface reduction process, from SnO2(110) to metallic Sn, without any layers fixed (Fig. 1). To rule out the possibility of size dependence, model sizes were examined (Fig. S1 and Supplementary Materials). The reduction of SnO2(110) involves 3 different types of oxygen: bridging oxygen (Obri), in-plane oxygen (Oip), and first-layer subsurface oxygen (Osub) (Fig. S2). Specifically, we defined the reduction degree by the ratio of the lost oxygen atoms to the total of 3 kinds of oxygen atoms involved in the reduction process. The MD-NN simulation of 2 ns with the canonical ensemble (NVT) on each surface was conducted to find the equilibrium surface structure at 300 K. Radial distribution functions (Fig. S3) and potential trajectory (Fig. S4) were employed to determine whether the MD-NN was in equilibrium. The accuracy of NN potential is shown in Fig. S5, Tables S1 and S2, and Supplementary Materials. This strategy has successfully located the dynamic active site in our previous study about oxygen-derived Cu catalysts [29]. The reduction process includes the following several stages: the surface reduction: SnO2(110) (0%), SnOx/SnO2(110) (25%), SnOx/SnO2(110) (50%), SnOx/SnO2(110) (61%), and Sn/SnO2(110) (75%); deep reduction: SnOx (86%) (lost half Osub) and SnOx (100%) (lost all Osub), shown in Fig. 1.
The MD-NN simulation indicates the removal of Obri and half Oip did not cause remarkable surface reconstruction at the early reduction stage (from 0% to 50%). Subsequently, slight reconstruction occurred on the SnOx/SnO2(110) surface (61%, corresponding to the loss of all surface Obri and Oip) (Fig. 1) and the Sn atoms tend to gather. Surface reconstruction becomes more obvious and Sn atoms aggregate to form the reconstructive structure from cluster to rod as the degree of SnO2 reduction increases (from 75% to 100% where the loss of Osub occurs) through MD-NN. Additionally, the reduced structures agree with the in situ spherical aberration corrected scanning transmission electron microscopy (STEM) experimental study [30], which proves the reliability of our simulated structure. The surface Pourbaix diagram was also estimated on different reduced surfaces obtained from MD simulation, which shows that SnO2(110) (0%), SnOx/SnO2(110) (50%), SnOx/SnO2(110) (61%), SnOx (86%), and SnOx (100%) are thermodynamically stable under various potentials (Fig. S6). Furthermore, we also performed global optimization based on the stochastic surface walking (SSW) algorithm (Fig. S7). The optimized structures by both methods confirm the reconstruction feature (cluster or rod) during the reduction process, except for a few O evolutions from deeper metal oxide layers and aggregation with Sn atoms, indicating that the diffusion barrier of subsurface O to surface might cause SnO2 reduction to be sluggish during MD-NN simulation. The reliability of the SSW-NN method was also verified by increasing the number of steps (5,000) of the SSW-NN simulation and by comparison with the MD-NN results, as detailed in Fig. S8. Besides, simulated annealing confirmed the stability of the SnOx surfaces, which demonstrates that annealing may be utilized to surmount the energy barrier of MD-NN simulations and has a propensity to generate structures comparable to SSW-NN. However, the structure of active site on the surface level can still be maintained (Figs. S9 and S10).
Electronic and geometric structure of the active sites
The surface reconstruction during the reduction process produces a variety of possible active sites with reduced Snδ+. Taking the SnOx(100%) surface as the example in Fig. 2A with the most variety of active sites (other surfaces shown in Fig. S11), obvious surface reconstruction occurred on SnOx(100%), and the coordination number (CN) of Sn-O covers 0, 1, 2, 3, 4, and 5 (Fig. 2B). The Sn-O CN on all surfaces is counted in Fig. 2B in order to further differentiate these spots geometrically (the graphical example is shown in Fig. S12). Two key conclusions can be drawn from Fig. 2B: (a) the reduction of SnOx leads to a decreasing trend of Sn-O CN, and (b) the surface reconstruction makes the CN of Sn-O distribution more dispersed, resulting in a variety of Sn active sites.
The loss of oxygen in SnOx leads to the decrease of the oxidation states of surface Sn (Fig. 2A). Therefore, the statistics of the Sn oxidation state distribution on all reduced surfaces were also calculated (Fig. 2B and C). We use the net charges for the stoichiometric ratio SnO2, SnO, and Sn as the reference to estimate the oxidation state of Sn. As shown in Fig. 2C, the surface site of Snδ+ can be approximately categorized by the following types: Snα+ (α = 2 to 4, ~2.32 e− to ~1.20 e−), Sn2+ (~1.20 e−), Snβ+ (β = 0 to 2, ~1.20 e− to ~0 e−), and Sn0 (~0 e−). The proportion of different types of active sites as a function of reduction degree is illustrated in Fig. 2B, demonstrating that the overall values of the net charge show an upward trend with the increased degree of SnOx reduction. The dispersed distribution occurs at the post-reduction stages, mainly from Snβ+ to Sn0, leading to multiple types of active sites. It can be concluded that the trends of the net charge are consistent with the CN of Sn-O, which means the dynamism in geometric structure leads to the redistribution of electron density.
Structure–performance relations
To explore the effect of surface reduction and reconstruction on CO2ER and side reaction (HER) mechanisms, possible pathways and key intermediates were taken into consideration to explore their mechanisms and catalytic activity.
Previous studies have revealed that the selectivity can be determined by the competitive adsorption of the 3 initial hydrogenation intermediates, namely, HCOO* for CO2ER to formate, *COOH for CO2ER to CO, and H* for HER to H2. It should be added that for the attribution of the reaction intermediates of formate and CO, the current reasonable interpretation is that the C-bound mode (*COOH, carbophilic) corresponds to the production of CO and the O-bound mode (HCOO*, oxidophilic) corresponds to the production of formate [11,31] (see details in the Supplementary Materials). The adsorption configuration is shown in Fig. 3A. The adsorption strength of the key intermediates of CO2ER, as well as HER, was calculated at the DFT level and the corresponding net charge of active sites was also related in Fig. 3B. Besides, the effect of dispersion-corrected functionals on adsorption energies was tested in Fig. S13 and Table S3. Each adsorbed species and the net charge of active sites are taken from no less than 40 sets of data among the 210 sites that occurred during the whole reduction process. Interestingly, we discovered that ∆EH*, ∆E*COOH, and ∆EHCOO* can be described as a function of the net charge of Sn active sites. These results can lead to the conclusion that the net charge most likely plays a direct and important role in regulating the adsorption strength of all 3 intermediate species (∆EH* = –0.30 e− + 0.58, ∆E*COOH = –0.31 e− + 0.66, and ∆EHCOO* = –0.62 e− – 0.17)
Further research was done on the effect of adsorption energy (∆Eads) on the trends of the CO2ER and HER (H2, CO, and formate) activities. It is commonly accepted that the theoretical overpotential that ensures each elementary step of a reaction becomes exergonic [32] can be used to estimate the reaction activities theoretically [33]. Besides, CO2ER to formate and CO, as well as HER, include only 2 elementary reaction steps. Therefore, the theoretical overpotential can be obtained from the free energies of intermediates (∆Gads), or the difference between the equilibrium potential and ∆Gads. The correction of ∆Gads from ∆Eads is provided in the Supplementary Materials and Table S4. In this way, the theoretical overpotential of different products, such as CO2ER to formate in Fig. 3C, can be detected based on the relations of net charge and ∆Eads. The results show that the theoretical overpotential depends on the second reaction step when the net charge is large, due to the strong adsorption of HCOO*. The adsorption capacity of HCOO* decreases as the net charge decreases, at which point the potential-determining step shifts to the first one; in this way, a volcano-shaped activity of CO2ER to formate emerges and the theoretical optimal net charge range is approximately from 0.77 to 0.08 e−. The trends of CO and H2 are also calculated, as shown in Fig. S14. The theoretical overpotential of 3 products is summarized in Fig. 3D, demonstrating that when the value of net charge for Sn sites is high, strong binding of HCOO* at Sn sites leads to a highly energy-consuming step of HCOO* to formate in thermodynamics. Besides, excessively weak *COOH adsorption results in a high overpotential of CO. Nevertheless, the adsorption strength of H* is optimal for HER, indicating that the generation of H2 is preferable to that of formate and CO (green area in Fig. 3D). As the net charge decreased (blue area), the weakening of HCOO* adsorption capability could make the generation of HCOO* from CO2 become the potential-determining step and the theoretical overpotential of formate becomes the lowest. It leads to the priority of the target product and inhibition of H2. This means that the related catalytic selectivity is likewise impacted by the dispersion of the net charge distribution brought on surface reduction and reconstruction. Specifically, we can draw the theoretical overpotential as a function of net charge and then directly derive the relevant catalytic activity and selectivity from those profiles owing to the linear relations between the net charge of active sites and the adsorption energy of key intermediates. Thus, the net charge may be considered as a streamlined descriptor to forecast the dynamic structure–performance relations.
Next, we further predicted the theoretical catalytic performance on various SnOx surfaces and directly validated it with electrochemical experiment, based on the redistribution of active sites induced by SnOx reduction and surface reconstruction. As was stated earlier, the only input necessary for DFT calculations is the net charge of Sn sites. Thus, the fitted ∆Eads directly leads to the theoretical overpotential after the correction of free energy. The theoretical Faradaic efficiency (FE) of CO2ER to formate and CO, as well as HER, can be simulated by substituting the prefactors, obtained by fitting the experimental value by –1.2 V vs. reversible hydrogen electrode (RHE) (Fig. 3E and Fig. S15a, details in the Supplementary Materials) [34,35]. The corresponding net charge and the number of surface atoms (Nsurf) exposed to the different reduced degree surfaces are counted 5 times during the MD equilibrium stage. Figure 3E (top) shows that the average FE of formate increases first and then decreases with the reduction of SnO2(110) and shows the volcanic peak in a moderate reduction stage (61% and 75% of SnOx surfaces). To verify the predicted trends of the CO2ER on SnOx, we prepared electrodes using commercial SnO2 (50 to 70 nm) loaded onto commercial gas diffusion layers (GDLs) to represent the SnOx model used in the simulations. The FE of generated formate is quantified by the chronoamperometry test at −1.2 V RHE. The results are shown in Fig. 3E (bottom), which demonstrates that the formate FE trend is consistent with the theoretical simulation. The XRD pattern in Fig. S16 does not show the phase change of SnO2, indicating the stability of the SnO2 bulk phase structure during the reaction time range of 0.0 to 2.0 h. The chemical state of SnOx is examined by quasi in situ XPS studies [36,37]. The binding energy of Snδ+ (chemical state between 0 and 4+) varies linearly with the reduction time as shown in Fig. S18c, indicating that the reduction time can be used to express the surface reduction degree of SnOx. The percentage of Snδ+ is also detected quantitatively by theoretical statistics and quasi in situ XPS studies in Fig. 3F and Fig. S18 to correlate with the formate FE trend. Both theoretical calculations and experimental results show that there is a direct linear positive correlation between the performance of the main product formate and the proportion of the partially reduced Snδ+. More precisely, the chemical state of Sn is between 0 and 2+ as illustrated in Fig. 3F. Besides, the operando Raman and x-ray absorption spectroscopy (XAS) methods [21] monitoring the structural changes of SnOx also provide evidence that the Snδ+ state during CO2ER corresponds to the optimal formate production.
The origin of adsorbate–surface interaction
To gain deep insights into the specific impact of the net charge on the catalytic performance of CO2ER as well as HER, electronic structures at the DFT level were further conducted.
What needs to be illustrated here is that we design a representative model by changing the CN of Sn-O on the SnO2(110) surface from a geometry perspective to construct distinct and definite active sites (Fig. S19). Additionally, other facets such as SnO2(101) and SnO2(100) are also in line with the universal relation of the net charge and ∆Eads (Fig. S20). The linear correlations between ∆Eads and the integrated crystal orbital Hamilton population (iCOHP) can be quantitatively explained initially such that the adsorption strength is directly related to the interaction between the adsorbates and Sn active sites (Fig. S21 and Tables S5 to S7). The projected COHP (pCOHP) of Sn-O (HCOO*) bonding (Fig. 4A and Fig. S22) shows that the antibonding orbitals of s–p interaction move down with the increase in reduction degree, from one Ov (1Ov_d) to 4Ov_d. For details, see Fig. 4B and Fig. S23, where the pCOHP of s–p orbitals demonstrates that the antibonding states close to and below the Fermi level primarily stem from Sn 5s as well as Sn 5pz, which interacted with O 2pz atomic orbitals. This means that Sn 5s and 5pz atomic orbitals essentially contributed to the enhanced charge density of Sn sites. Meanwhile, the pCOHP analyses on *COOH and H* with Sn sites reveal that the weakening of Sn-C and Sn-H interaction are also due to the increase in the filling degree of the corresponding antibonding, namely, Sn 5s with H 1s and Sn 5s with C 2pz (Figs. S24 and S25). Therefore, the key intermediates of CO2ER to C1 products and HER on the active sites can be modulated by the occupation of antibonding orbital near the Fermi level, which is related directly to the geometrical SnO2 reduction together with the net charge of Sn sites.
The chemical bonding configuration of active sites and adsorbates was conducted by periodic natural bond orbital (pNBO) (Tables S8 and S9). The σ* bond occupancy of Sn-O in HCOO* is 0.20 e− with 1Ov_d and rose to 0.88 e− with 4Ov_d in Fig. 4C. This increasing trend of σ* bond occupancy in Fig. 4D is also consistent with the analyses of antibonding from iCOHP, indicating that the adsorption interaction weakens as SnO2(110) is reduced. By evaluating the hybridization of σ* (1Ov_d: 88.92 %, 2Ov_d: 87.55 %, 3Ov_d: 90.73 %, and 4Ov_d: 87.36 %, Table S8), it can be shown that the σ* orbital makes up the major contribution from Sn atoms. The pNBO analyses of Sn-C in *COOH also demonstrate that the σ* bond occupancy displays a positive relation with the concentration of Ov (Table S9), further confirming that an increase in the net charge caused by SnO2 reduction affects the antibonding filling, thus weakening the interaction strength between active sites and adsorbates.
Discussion
Less
Based on the above analyses, Fig. 5 outlines that SnO2 reduction in geometry leads to the increase in the charge density of Sn sites essentially. In particular, when interacting with the adsorbate (HCOO*), the increased charge in 5s and 5pz orbitals injects into the antibonding formed with O 2pz orbital, which weakens the adsorption capacity of SnOx and accounts for the differing adsorption strength of HCOO* on SnO2 with various Ov. It can be summarized that when the degree of SnOx reduction is low, the strong binding of HCOO* leads to the step of HCOO* to formate becoming the theoretical limiting step. As the degree of SnOx reduction deepens, the increased net charge of Sn site is injected into the antibonding orbital between Sn-O (HCOO*), remarkably weakening the binding ability of HCOO*, thus gradually changing the theoretical limiting step to the HCOO* formation step. Therefore, we can infer that the net charge of Sn sites is a key regulatory electronic factor that governs the interaction ability with adsorbates. Tuning the net charge of the active site can be regarded as an efficient strategy to promote the catalytic performance of SnO2 for CO2ER to formate theoretically.
In summary, this paper put forward a method that combined NN potential accelerated MD, global optimization, and first-principles calculations to simulate the dynamic reduction process of SnO2 and investigate the nature of Sn active sites as well as the electronic structure effect on the activity and selectivity of CO2ER and HER. These results demonstrate that the catalytic performance of SnOx changes with the surface reconstruction geometrically together with the charge distribution of Sn active sites. Moreover, a quantitative and dynamic structure–performance relation between the net charge and the ∆Eads was fit. Furthermore, by exploiting this relation, we can statistically estimate the proportion of different sites obtained from MD-NN simulation and thus invert the apparent catalytic performance of CO2ER over different SnOx surfaces, which is also found to be consistent with the experimental test. The model sites reveal that the nature of Sn active sites is responsible for the catalytic performance; that is, the antibonding occupation is enhanced with the increased net charge of Sn active sites, leading to the variation in the interaction between Sn sites with the adsorbates. Finally, a theoretical guideline for the improvement of formate selectivity is proposed: The Sn active sites with a moderate amount of net charge, which can be achieved by tuning the reduction degree of SnOx, performs an excellent catalytic activity of CO2ER. It should be noted that as an electrocatalytic system, the electrical double layer, the electrolyte solution, and the dissolution/redeposition process of electrodes should be considered. However, the force fields currently developed to simulate electrochemical models are still very crude, and the simulation time of MD is limited. We are also currently conducting related research, including potential fitting and application of solid–liquid interface simulation. In this study, the MD-NN simulation method is mainly used to figure out the surface structure and the feature sites of SnOx. These explorations illustrate how a versatile and dynamic simulation approach can explore the key regulatory factor for CO2ER and provide rational guidance to promote the catalytic performance of metal oxide in electrocatalysis.
Materials and Methods
Less
MD-NN and SSW-NN simulations
The large-scale atomistic simulation with neural network potential (LASP) program developed by Huang et al. [38] (http://lasphub.com/%23/lasp/laspHome) was used to resolve the complex oxide structures based on global structure and molecular dynamics simulation (SSW-NN and MD-NN) methods. The SnO.pot and SnOH.pot trained by NN-based SSW PES exploration [39] were adopted over the reduction process of SnO2 to search the stable structures at different reduced stages. The global datasets generated from the high-accuracy first-principles calculations contain 17,343 and 40,143 structures for SnO and SnOH systems, and further details are listed in Tables S1 and S2. The root-mean-square errors of SnO.pot and SnOH.pot potentials are 9.018 and 6.264 meV/atom, respectively. The comparison of NN potential predicted energy and that from DFT calculation of H* adsorption energy (active site: O and Sn) are shown in Fig. S5; the results indicate that the accuracy of NN potential can reach the level of DFT calculation. In order to explore the globally stable and dynamically reduced structures, SSW-NN and MD-NN methods are carried out respectively to optimize the reduction process of SnO2.
As for global optimization, SSW step was set to 2,000 for SSW-NN. As for MD-NN simulations, 2 ns of MD-NN simulation with NVT ensemble at 300 K was carried out for each structure with different reduced degrees [29]. Throughout these processes, no layer of the model system is fixed. After the simulation of the whole reduction process, all the low-energy structure candidates from the SSW-NN and MD-NN exploration and related calculations of CO2ER and HER reaction mechanisms were executed by plane-wave DFT next. The O atoms deeper than the first-layer subsurface oxygen were not considered in the reduction process due to deeper O atoms having little effect on the surface Sn.
DFT calculations
DFT calculations were carried out using the Vienna ab initio Simulation Package (VASP) code [40,41] in conjunction with the computational hydrogen electrode model [42]. We adopted the generalized gradient approximation in the form of the Perdew-Burke-Ernzerhof (PBE) functional and projector augmented wave pseudopotentials [40] to describe the electron exchange and correlation effects [43]. A cutoff energy of 400 eV and an atomic force convergence of 0.02 eV/Å were employed. Four-layer (6×3) periodic slabs of the SnOx/SnO2(110) surfaces were modeled, where the 2 bottom layers were kept fixed, and the Monkhorst-Pack k-mesh of 2×2×1 was adopted for integrations over the Brillouin zone for all systems. In addition, the spin polarization scheme was adopted for all calculations. A vacuum space with at least 15 Å was set between the slabs to the periodic images in the vertical direction. As for the model simulation, a (4×2) slab with 4 layers is constructed. The Monkhorst-Pack k-mesh was set to 3×3×1 and the vacuum space was also at least 15 Å.
The free energy for each intermediate state is calculated as G = EDFT + ZPE + δH0 − TS, where EDFT is the DFT total energy, ZPE is the zero point vibrational energy, δH0 is the integrated heat capacity, T is the reaction temperature, and S is the entropy. For gas and liquid phase species, the δH0 and entropy S at 298.15 K were obtained from Dean [44]. For adsorbates, the entropy was calculated with the harmonic oscillator approximation (see Table S4). The solvation corrections of the water–solid surface interface are included in this work: –0.5 eV for OH*, –0.2 eV for COOH*, and –0.25 eV for HCOO* [42,45]. The theoretical UL is numerically equal to the negative value of the biggest Gibbs free energy change of CO2ER to formate or CO and HER (E = –ΔGmax/zF, where z is the number of electrons transferred, F is the Faradaic constant, and E is the limiting potential). The computational details can be found in the Supplementary Materials.
The COHP analysis was conducted through the Lobster 3.2.0 code, upon a transformation of the (plane) wave functions from VASP into a localized basis set [46,47]. Besides, the pNBO formalism developed by Dunnington and Schmidt [48] was also applied using VASP software to analyze surface chemical bonding via natural bond orbitals.
The SnO2 electrode preparation
The obtained SnO2 was airbrushed onto the commercial carbon-based GDLs at an approximate loading of 0.9 mg/cm2, measured through weighing GDLs before and after airbrushing. The catalyst ink was prepared by dispersing 200 mg of SnO2 and 50 μl of Nafion solution (Sigma-Aldrich) in 750 μl of isopropyl alcohol and 250 μl of ultra-purity water and sonicated for 1 h before airbrushing (H-SET, Paasche).
Electrochemical test
All electrochemical experiments were conducted in a gas-tight H-type electrochemical cell machined from poly(methylmethacrylate) (PMMA). The cell was thoroughly rinsed with ultra-purity water prior to all experimentation. The working and counter electrodes were parallel and separated by a bipolar membrane (fumasep, FBM). Gas dispersion frits were incorporated into both electrode chambers to provide ample gas-electrolyte mixing. The exposed geometric surface area of the working electrode (the SnO2 electrodes) was 1 cm2 and the electrolyte volume of each electrode chamber was 10 ml. The counter electrode was an IrRu alloy deposited Ti mesh (5 cm2). The working electrode potential was referenced against a leak-free Ag/AgCl electrode (Harvard Apparatus) that was calibrated against a homemade standard hydrogen electrode. Metallic impurities in the as-prepared electrolyte (0.1 M KHCO3) were removed before electrolysis by chelating the solution with Chelex 100 (Na form, Sigma Aldrich). Both electrode chambers were sparged with CO2 at a rate of 5 sccm for at least 30 min prior to and throughout the duration of all electrochemical measurements. Upon saturation with CO2, the pH of the electrolyte was 6.8, which was maintained throughout the duration of chronoamperometry (PGSTAT204, Autolab).
During electrolysis, gas products were quantified using an online gas chromatography system (GC7890B, Agilent Technologies, Inc.). The thermal conductivity detector was connected to a MolSieve 5A packed column (Agilent Technologies, Inc.) to detect H2, O2, and N2, and a back flame ionization detector (FID) was connected to a Porapak Q packed column (Agilent Technologies, Inc.) to detect CO. A methanizer was installed to enable the back FID to detect CO with 1,000 times higher sensitivity. A front FID was connected to an HP-PLOT Al2O3 capillary column (Agilent Technologies, Inc.) to detect hydrocarbons (C1 to C3). Ar was used as the carrier gas. After passing through the reactor, the gas was allowed to flow directly into the gas sampling loop of the gas chromatography for online gaseous product analysis. The signal response was calibrated by analyzing a series of standard gas mixtures (Messer Group). The mixture includes H2 (299.1 ppm, 1,047 ppm, 18,376.3 ppm, and 2.01%), CO (103 ppm, 199.5 ppm, 994.9 ppm, and 998.6 ppm), CH4 (20 ppm, 198.1 ppm, 500.4 ppm, and 9,917.4 ppm), C2H4 (20 ppm, 99.4 ppm, 487.5 ppm, and 1,001.6 ppm), and C2H6 (19.9 ppm, 99.0 ppm, 521.2 ppm, and 968.2 ppm).
The liquid products were collected from the cathode and anode chambers after electrolysis and analyzed by headspace gas chromatography, high-performance liquid chromatography, and 1H-NMR. Headspace gas chromatography measurements were carried out using a BCHP HS-2 Headspace Sampler with GC2060 gas chromatography (Shanghai Ruimin Instrument Co., Ltd.). Typically, 10-ml vials were filled with 3 ml of the liquid sample and sealed. They were heated to 70 °C over 20 min in the headspace sampler and 1 ml of the headspace gas composition was automatically injected into the GC. The sample loop (110 °C) and transfer line (110 °C) were both heated to avoid condensation. N2 was used as the carrier gas. An HP-INNOWax capillary column (length: 60 m; ID: 0.32 mm; film: 0.5 μm, Agilent) was used to separate the compounds in the sample. The peak areas of methanol, ethanol, and n-propanol were converted to mole concentration (mol/L, M) using calibration curves that were obtained using methanol, ethanol, and n-propanol standard solutions (1,000 μg/ml, purchased from Beijing Yihuatongbiao Tech. Co., Ltd., and have been certified by the China National Institute of Metrology) diluted with ultra-purity water to different concentrations (0.1, 1, 5, 10, and 15 mM). In the high-performance liquid chromatography analysis (1260 Infinity II with a UV detector at 210 nm, Aglient), a mixed eluent (pH 2.5) of ultra-purity water, H2SO4 (2 mM), and Na2SO4 (25 mM) flowed through a C18-inverse column (4.6 mm × 150 mm, InertSustain AQ-C18, Shimadzu) at 1 ml/min. Before performing the analysis, the electrolyte sample is first acidified with 0.5 M H2SO4 in the same volume as the sample. The peak areas of formate and acetate were converted to mole concentration using calibration curves that were obtained using standard solutions (home-prepared) of each product at a concentration of 0.1, 2.5, 5, and 10 mM. 1H-NMR was performed using AVANCE IIITM HD 400 MHz NanoBAY with solvent (water) suppression. Electrolyte (400 μl) was mixed with 100 μl of a solution of 10 mM DMSO and 50 mM phenol in D2O as internal standards for the 1H-NMR analysis. The internal standards, phenol and DMSO, were chosen because they did not interfere with peaks arising from CO2 reduction products and because of their nonvolatility, which allowed the use and storage of the same internal standards solution for all of the product measurements without appreciable change in concentration. The area of product peaks to the right of the water peak was compared to the area of DMSO (at a chemical shift of 2.6 ppm), and the area of product peaks to the left of the water peak was compared to the area of phenol (at a chemical shift of 7.2 ppm).
Characterization
Quasi in situ XPS measurements
Quasi in situ XPS studies were used to detect the chemical state of SnOx. The SnO2 electrodes were potentiostatically polarized in a glovebox (SUPER, MIKROUNA) using a 3-electrode PMMA cell with an IrRu alloy deposited Ti mesh counter electrode and a calibrated leak-free Ag/AgCl reference electrode. A selected potential (−1.2 V vs. RHE) was applied for a certain duration, followed by emersion under potential control and flushing with deionized water. Afterward, the specimen was transported to the XPS analyzing chamber (ESCALAB Xi+, Thermo Fisher Scientific) in a vacuum transfer module (Thermo Scientific™) for characterization of their surface chemistries, which avoids air exposure as well as surface re-oxidation. The picture of the transfer cell is shown in Fig. S17. Peak positions were further corrected by referencing the C 1s peak position of adventitious carbon for the sample (284.8 eV) and by shifting all other peaks in the spectra accordingly.
XRD measurements
The SnO2 electrodes were potentiostatically under –1.2 V vs. RHE using a 3-electrode PMMA cell with an IrRu alloy deposited Ti mesh counter electrode and a calibrated leak-free Ag/AgCl reference electrode. A selected potential was applied for different durations, followed by flushing with deionized water. The crystal structure was determined by an x-ray diffractometer (XRD, Bruker D8 Focus) with Cu Kα radiation (λ = 1.54056 Å) at 40 kV and 40 mA. XRD spectra were collected over a 2θ range of 30° to 60° at a scanning speed of 8°/min.
References
Less
1.
De Luna P, Hahn C, Higgins D, Jaffer SA, Jaramillo TF, Sargent EH. What would it take for renewably powered electrosynthesis to displace petrochemical processes? Science. 2019;364(6438):Article eaav3506.
2.
Seh ZW, Kibsgaard J, Dickens CF, Chorkendorff I, Nørskov JK, Jaramillo TF. Combining theory and experiment in electrocatalysis: Insights into materials design. Science. 2017;355(6321):Article eaad4998.
3.
Nitopi S, Bertheussen E, Scott SB, Liu X, Engstfeld AK, Horch S, Seger B, Stephens IEL, Chan K, Hahn C, et al.Progress and perspectives of electrochemical CO2 reduction on copper in aqueous electrolyte. Chem Rev. 2019;119(12):7610–7672.
4.
Gao D, Li W, Wang H, Wang G, Cai R. Heterogeneous catalysis for CO2 conversion into chemicals and fuels. Trans Tianjin Univ. 2022;28(4):245–264.
5.
Jiang G, Han D, Han Z, Gao J, Wang X, Weng Z, Yang QH. Rational manipulation of intermediates on copper for CO2 electroreduction toward multicarbon products. Trans Tianjin Univ. 2022;28(4):265–291.
6.
Lei F, Liu W, Sun Y, Xu J, Liu K, Liang L, Yao T, Pan B, Wei S, Xie Y. Metallic tin quantum sheets confined in graphene toward high-efficiency carbon dioxide electroreduction. Nat Commun. 2016;7:Article 12697.
7.
Gu J, Heroguel F, Luterbacher J, Hu X. Densely packed, ultra small SnO nanoparticles for enhanced activity and selectivity in electrochemical CO2 reduction. Angew Chem Int Ed. 2018;57(11):2943–2947.
8.
Zhang S, Kang P, Meyer TJ. Nanostructured tin catalysts for selective electrochemical reduction of carbon dioxide to formate. J Am Chem Soc. 2014;136(5):1734–1737.
9.
Li Q, Fu J, Zhu W, Chen Z, Shen B, Wu L, Xi Z, Wang T, Lu G, Zhu JJ, et al.Tuning Sn-catalysis for electrochemical reduction of CO2 to CO via the core/shell cu/SnO2 structure. J Am Chem Soc. 2017;139(12):4290–4293.
10.
Luc W, Collins C, Wang S, Xin H, He K, Kang Y, Jiao F. Ag–Sn bimetallic catalyst with a core–shell structure for CO2 reduction. J Am Chem Soc. 2017;139(5):1885–1893.
11.
Vasileff A, Zhi X, Xu C, Ge L, Jiao Y, Zheng Y, Qiao SZ. Selectivity control for electrochemical CO2 reduction by charge redistribution on the surface of copper alloys. ACS Catal. 2019;9(10):9411–9417.
12.
Wen G, Lee DU, Ren B, Hassan FM, Jiang G, Cano ZP, Gostick J, Croiset E, Bai Z, Yang L, et al.Orbital interactions in bi-Sn bimetallic electrocatalysts for highly selective electrochemical CO2 reduction toward formate production. Adv Energy Mater. 2018;8(31):Article 1802427.
13.
Sarfraz S, Garcia-Esparza AT, Jedidi A, Cavallo L, Takanabe K. Cu–Sn bimetallic catalyst for selective aqueous electroreduction of CO2 to CO. ACS Catal. 2016;6(5):2842–2851.
14.
Bai X, Chen W, Zhao C, Li S, Song Y, Ge R, Wei W, Sun Y. Exclusive formation of formic acid from CO2 electroreduction by a tunable Pd-Sn alloy. Angew Chem Int Ed. 2017;56(40):12219–12223.
15.
Zhu Q, Ma J, Kang X, Sun X, Liu H, Hu J, Liu Z, Han B. Efficient reduction of CO2 into formic acid on a lead or tin electrode using an ionic liquid catholyte mixture. Angew Chem Int Ed. 2016;55(31):9012–9016.
16.
Lee S, Ocon JD, Son Y-i, Lee J. Alkaline CO2 electrolysis toward selective and continuous HCOO– production over SnO2 nanocatalysts. J Phys Chem C. 2015;119(9):4884–4890.
17.
Wei Y, Liu J, Cheng F, Chen J. Mn-doped atomic SnO2 layers for highly efficient CO2 electrochemical reduction. J Mater Chem A. 2019;7(34):19651–19656.
18.
Cui C, Han J, Zhu X, Liu X, Wang H, Mei D, Ge Q. Promotional effect of surface hydroxyls on electrochemical reduction of CO2 over SnO/Sn electrode. J Catal. 2016;343:257–265.
19.
Li L, Zhao Z-J, Hu C, Yang P, Yuan X, Wang Y, Zhang L, Moskaleva L, Gong J. Tuning oxygen vacancies of oxides to promote electrocatalytic reduction of carbon dioxide. ACS Energy Lett. 2020;5(2):552–558.
20.
Baruch MF, Pander JE, White JL, Bocarsly AB. Mechanistic insights into the reduction of CO2 on tin electrodes using in situ ATR-IR spectroscopy. ACS Catal. 2015;5(5):3148–3156.
21.
Dutta A, Kuzume A, Kaliginedi V, Rahaman M, Sinev I, Ahmadi M, Roldán Cuenya B, Vesztergom S, Broekmann P. Probing the chemical state of tin oxide NP catalysts during CO2 electroreduction: A complementary operando approach. Nano Energy. 2018;53:828–840.
22.
Jiao X, Li X, Jin X, Sun Y, Xu J, Liang L, Ju H, Zhu J, Pan Y, Yan W, et al.Partially oxidized SnS2 atomic layers achieving efficient visible-light-driven CO2 reduction. J Am Chem Soc. 2017;139(49):18044–18051.
23.
Hu H, Gui L, Zhou W, Sun J, Xu J, Wang Q, He B, Zhao L. Partially reduced Sn/SnO2 porous hollow fiber: A highly selective, efficient and robust electrocatalyst towards carbon dioxide reduction. Electrochim Acta. 2018;285:70–77.
24.
Peng M, Dong C, Gao R, Xiao D, Liu H, Ma D. Fully exposed cluster catalyst (FECC): Toward rich surface sites and full atom utilization efficiency. ACS Cent Sci. 2020.
25.
Chee SW, Arce-Ramos JM, Li W, Genest A, Mirsaidov U. Structural changes in noble metal nanoparticles during CO oxidation and their impact on catalyst activity. Nat Commun. 2020;11(1):2133.
26.
Sun JJ, Cheng J. Solid-to-liquid phase transitions of sub-nanometer clusters enhance chemical transformation. Nat Commun. 2019;10(1):5400.
27.
Zhang XJ, Shang C, Liu ZP. From atoms to fullerene: Stochastic surface walking solution for automated structure prediction of complex material. J Chem Theory Comput. 2013;9(7):3252–3260.
28.
Li YF, Zhu SC, Liu ZP. Reaction network of layer-to-tunnel transition of MnO2. J Am Chem Soc. 2016;138(16):5371–5379.
29.
Cheng D, Zhao Z-J, Zhang G, Yang P, Li L, Gao H, Liu S, Chang X, Chen S, Wang T, et al.The nature of active sites for carbon dioxide electroreduction over oxide-derived copper catalysts. Nat Commun. 2021;12(1):Article 395.
30.
Li G, Li S, Han ZK, Zou C, Wu H, Yuan W, Zhu B, Gao Y, Yang H, Zhang Z, et al.In situ resolving the atomic reconstruction of SnO2 (110) surface. Nano Lett. 2021;21(17):7309–7316.
31.
Hori Y, Wakebe H, Tsukamoto T, Koga O. Electrocatalytic process of CO selectivity in electrochemical reduction of CO2 at metal electrodes in aqueous media. Electrochim Acta. 1994;39(11–12):1833–1839.
32.
Peterson AA, Nørskov JK. Activity descriptors for CO2 electroreduction to methane on transition-metal catalysts. J Phys Chem Lett. 2012;3(2):251–258.
33.
Nørskov JK, Rossmeisl J, Logadottir A, Lindqvist L, Kitchin JR, Bligaard T, Jónsson H. Origin of the Overpotential for oxygen reduction at a fuel-cell cathode. J Phys Chem B. 2004;108:17886–17892.
34.
Yang P, Li L, Zhao Z-J, Gong J. Reveal the nature of particle size effect for CO2 reduction over Pd and au. Chin J Catal. 2021;42(5):817–823.
35.
Eom T, Kim WJ, Lim H-K, Han MH, Han KH, Lee EK, Lebègue S, Hwang YJ, Min BK, Kim H. Cluster expansion method for simulating realistic size of nanoparticle catalysts with an application in CO2 electroreduction. J Phys Chem C. 2018;122(16):9245–9254.
36.
Foelske-Schmitz A. X-ray photoelectron spectroscopy in electrochemistry research. In: Encyclopedia of interfacial chemistry. New York (NY): Elsevier; 2018. p. 591–606.
37.
Griesser C, Li H, Wernig E-M, Winkler D, Shakibi Nia N, Mairegger T, Götsch T, Schachinger T, Steiger-Thirsfeld A, Penner S, et al.True nature of the transition-metal carbide/liquid Interface determines its reactivity. ACS Catal. 2021;11(8):4920–4928.
38.
Huang SD, Shang C, Kang PL, Zhang X-J, Liu Z-P. LASP: Fast global potential energy surface exploration. Wiley Interdiscip Rev Comput Mol Sci. 2019;9(6):e1415.
39.
Huang SD, Shang C, Zhang XJ, Liu ZP. Material discovery by combining stochastic surface walking global optimization with a neural network. Chem Sci. 2017;8(9):6327–6337.
40.
Blochl PE. Projector augmented-wave method. Phys Rev B. 1994;50(24):17953–17979.
41.
Kresse JFG, Furthmüller J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput Mater Sci. 1996;6(1):15–50.
42.
Peterson AA, Abild-Pedersen F, Studt F, Rossmeisl J, Nørskov JK. How copper catalyzes the electroreduction of carbon dioxide into hydrocarbon fuels. Energy Environ Sci. 2010;3(9):1311–1315.
43.
Generalized gradient approximation made simple. Phys Rev Lett. 1996;77(18):3865–3868.
44.
J. A. Dean, Lange's handbook of chemistry. 3rd ed. New York (NY): McGraw-Hill Professional; 1998.
45.
Karamad M, Hansen HA, Rossmeisl J, Nørskov JK. Mechanistic pathway in the electrochemical reduction of CO2 on RuO2. ACS Catal. 2015;5(7):4075–4081.
46.
Dronskowski R, Bloechl PE. Crystal orbital Hamilton populations (COHP): Energy-resolved visualization of chemical bonding in solids based on density-functional calculations. J Phys Chem. 1993;97(33):8617–8624.
47.
Deringer VL, Tchougreeff AL, Dronskowski R. Crystal orbital Hamilton population (COHP) analysis as projected from plane-wave basis sets. J Phys Chem A. 2011;115(21):5461–5466.
48.
Dunnington BD, Schmidt JR. Generalization of natural bond orbital analysis to periodic systems: Applications to solids and surfaces via plane-wave density functional theory. J Chem Theory Comput. 2012;8(6):1902–1911.
49.
He M, Xu B, Lu Q. Probing the role of surface speciation of tin oxide and tin catalysts on CO2 electroreduction combining in situ Raman spectroscopy and reactivity investigations. Chin J Catal. 2022;43(6):1473–1477.
50.
Choi YW, Scholten F, Sinev I, Roldan Cuenya B. Enhanced stability and CO/Formate selectivity of plasma-treated SnOx/AgOx catalysts during CO2 electroreduction. J Am Chem Soc. 2019;141(13):5261–5266.
Appendix
Less
Year 2023 volume 6 Issue 3
PDF
51
28
Cite this Article
BibTeX
Article Info
doi: 10.34133/research.0067
  • Receive Date：2022-09-19
  • Online Date：2026-10-27
  • Published：2023-03-20
Article Data
Affiliations
History
  • Received：2022-09-19
  • Accepted：2023-01-15
Affiliations
    1School of Chemical Engineering and Technology, Key Laboratory for Green Chemical Technology of Ministry of Education, Tianjin University, Tianjin 300072, China.
    2Collaborative Innovation Center for Chemical Science and Engineering (Tianjin), Tianjin 300072, China.
    3National Industry-Education Platform of Energy Storage, Tianjin University, Tianjin 300072, China.
    4Haihe Laboratory of Sustainable Chemical Transformations, Tianjin 300192, China.
    5Joint School of National University of Singapore and Tianjin University, International Campus of Tianjin University, Binhai New City, Fuzhou 350207, China.

Corresponding:

* Address correspondence to:
References
Share
https://castjournals.cast.org.cn/joweb/research/EN/10.34133/research.0067
Share to
QR

Scan QR to access full text

Cite this article
BibTeX
Citations
表12种不同金属材料的力学参数
科
Family		 属数
Number of
genus		 种数
Number of
species		 占总种数比例
Percentage of
total species (%)		 属
Genus		 种数
Number of
species		 占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏
  • BibTeX
  • EndNote
  • RefWorks
  • TxT
Links
Articles
Latest Articles
Most Read
Collections
Updates
Events
News
Multimedia
About
About Us
Contact
No. 86 Xueyuan South Road, Haidian District, Beijing
100081
010-62662699(Operation and Technical Support)
qkjq@cast.org.cn

Copyright © 2025 China Association for Science and Technology. All rights reserved. For all open access content, the relevant licensing terms apply.
Sponsored by the Office of the Leading Group for Cybersecurity and Informatization of CAST, and supported by Science and Technology Review Publishing House