Predicting Adsorption Properties of Catalytic Descriptors on Bimetallic

Letter pubs.acs.org/JPCL

Cite This: J. Phys. Chem. Lett. 2019, 10, 1852−1859

Predicting Adsorption Properties of Catalytic Descriptors on Bimetallic Nanoalloys with Site-Specific Precision Tej S. Choksi,†,‡ Luke T. Roling,† Verena Streibel,†,‡ and Frank Abild-Pedersen*,‡ †

SUNCAT Center for Interface Science and Catalysis, Department of Chemical Engineering, Stanford University, Stanford, California 94305, United States ‡ SUNCAT Center for Interface Science and Catalysis, SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park, California 94025, United States

J. Phys. Chem. Lett. Downloaded from pubs.acs.org by UNIV PARIS-SUD on 04/05/19. For personal use only.

S Supporting Information *

ABSTRACT: Bimetallic nanoparticles present a vastly tunable structural and compositional design space rendering them promising materials for catalytic and energy applications. Yet it remains an enduring challenge to efficiently screen candidate alloys with atomic level specificity while explicitly accounting for their inherent stabilities under reaction conditions. Herein, by leveraging correlations between binding energies of metal adsorption sites and metal−adsorbate complexes, we predict adsorption energies of typical catalytic descriptors (OH*, CH3*, CH*, and CO*) on bimetallic alloys with site-specific resolution. We demonstrate that our approach predicts adsorption energies on top and bridge sites of bimetallic nanoparticles having generic morphologies and chemical environments with errors between 0.09 and 0.18 eV. By forging a link between the inherent stability of an alloy and the adsorption properties of catalytic descriptors, we can now identify active site motifs in nanoalloys that possess targeted catalytic descriptor values while being thermodynamically stable under working conditions.

B

monometallic catalysts,26−33 neither approach reveals the thermodynamic stabilities of predicted active site ensembles under reaction conditions. Furthermore, generalized coordination numbers lack clear extensions to bimetallic catalysts as, per definition, they are independent of the local chemical environment. In addition to elucidating the impact of structural variations on adsorption energies, the effect of local chemical composition on adsorption energies has been probed theoretically using approaches ranging from theories of bonding on metals34−36 and cluster expansions37,38 to more recent models inspired by machine learning.39−43 For example, Xin et al. predict composition-dependent OH* binding energies on fcc (111) surfaces of Pt-based alloys for improved oxygen reduction catalysts by parametrizing the d-band model in terms of physical characteristics of elements.34,35 This parametrization, however, is limited to alloys having small perturbations in d-states relative to platinum. Alternatively, it has been shown that perturbations in local electronic structure in bimetallic materials can be captured by alchemical derivatives relying on electrostatic potentials of constituting atoms.36 Extensions to an expanded composition space on fcc (100) and (111) surfaces have been formulated using an artificial neural network that employs fingerprints like the first four moments of the d-band distribution and local electro-

imetallic nanoparticles can be tuned in a vast structure and composition space. This tunability, together with synergistic interactions between individual components,1−3 renders them into promising materials for conventional and emerging catalytic applications.4−14 Candidate materials across this diverse combinatorial space are identified through descriptor-based screening.15,16 Therein, catalytic properties are mapped to one or two descriptors, which are usually adsorption energies of species like OH*, CH3*, CH*, and CO*.17−20 The resulting reactivity or selectivity maps identify a range of descriptor adsorption energies for the best catalysts. These maps, however, lack a direct link to structural and compositional features of active site ensembles and, more crucially, to the thermodynamic stabilities of active site ensembles under reaction conditions. Integrating rigorous considerations of alloy stability into these reactivity maps can explicitly include any dynamic restructuring under reaction conditions within catalyst design paradigms. Concurrent considerations of reactivity and stability will reduce the materials gap between model catalysts and their working state. More importantly, these three-way mappings between structural characteristics of alloys, inherent stabilities of atoms constituting the alloy, and descriptors for catalytic reactivity will engender the longstanding goal of reverse-engineering catalytic motifs. First attempts in this direction have mapped catalytic descriptors to structural features of active sites using generalized21−23 and orbitalwise24,25 coordination numbers. Despite recent successes in identifying optimal active sites for © XXXX American Chemical Society

Received: February 19, 2019 Accepted: April 2, 2019 Published: April 2, 2019 1852

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

Figure 1. (a) OH* adsorbed on a 147 atom cuboctahedral PtAu nanoparticle. Site characteristics such as the elemental composition of an adsorption site (Au-top), its morphology (coordination number (CN) of 9), and chemical composition of the first coordination shell (Pt and Au) are illustrated. Pt, Au, O, and H are depicted in gray, yellow, red, and white. (b) Normal distribution functions highlighting the decreasing importance of variations in site identity (Pt or Au), morphology (CN varies from 3 to 9), and chemical composition of the first coordination shell (Pt, CN:6, and Au, CN:7) on adsorption energies. Full width at half maximum values for each distribution are marked.

negativities.41,42,44 Despite recent advances, existing screening protocols remain limited by two key aspects. First, these paradigms are not sufficiently generalizable across bimetallic alloys having arbitrary structures and compositions. Second, since current screening paradigms do not explicitly include the energy space of bimetallic alloys, they cannot concurrently determine both reactivity trends and thermodynamic stabilities of candidate alloys under reaction conditions. The thermodynamic stability of adsorption sites, defined as the binding energies of individual atoms constituting a given site (top, bridge, or hollow), is a descriptor that explicitly incorporates the energy space of metal atoms. Site stabilities directly indicate the likely existence of a given site under reaction conditions. Site stabilities also unify systematic variations in both local morphology and composition into a single continuous descriptor space. More crucially, site stabilities can be expeditiously generated using a coordination-based framework for mono- and bimetallic systems.45,46 Using this model, Roling et al. revealed a straightforward mapping between binding energies of metal−adsorbate complexes (M-ads*) and site stabilities (M*) in monometallic systems, which can predict site-specific adsorption energies on monometallic nanoparticles.47 These correlations arise from the inherent link between binding energies of surface metal atoms and adsorption strengths of metal−adsorbate complexes, which previously has been explored using experiments.48,49 Herein, by strategically constructing a family of energy correlations spanning coordination and composition space, we propose a highly generalizable predictive framework that determines site-specific adsorption properties on bi- and trimetallic alloys. We examine its versatility by forecasting adsorption energies with atomic site resolution on nanoalloys having wide-ranging morphologies and compositions. Planewave density functional theory (DFT) calculations were performed using the Quantum ESPRESSO package50 within the Atomic Simulation Environment.51 Self-consistent total energies were computed using the RPBE functional52 with core states represented by ultrasoft Vanderbilt pseudopotentials.53 Kohn−Sham equations were solved in reciprocal space on a k-point grid generated using the Monkhorst−Pack method.54 Further computational details are described on pp S2−S5 in the Supporting Information. Thermodynamic stabilities of adsorption sites (BEM1*) are defined as the binding energies of a given ensemble (top, bridge or hollow)

referenced to isolated metal atoms in the gas phase as shown in eq 1. Binding energies of site-adsorbate complexes (BEM1−OH*) are defined in eq 2. In eq 1 and eq 2, EM1* and EM1−OH* represent ground state energies of surfaces containing the adsorption site and the site−adsorbate complex, respectively, E* represents the ground state energy of the surface without the adsorption site, and EM1gas, Egas OH are the ground state energies of metal atoms and adsorbates in the gas phase. All data generated in this study are maintained in the SUNCAT Catalysis Hub database located at https://www.catalysis-hub. org/.55 gas BE M1* = E M1* − E − E M 1 *

(1)

gas gas BE M1− OH* = E M1− OH* − E − E M − EOH 1 *

(2)

Adsorption sites in bimetallic alloys are characterized by their elemental composition, morphology (coordination number, (CN) of atoms constituting the site), and the local chemical environment (elemental composition of the first coordination shell). Using OH* as an example we compare relative impacts of each site characteristic on metal−adsorbate interactions in Figure 1b. We use normal distribution functions to fit variations in adsorption energies of OH* on Pt (solid lines) and Au (dashed lines) as a function of changes to the morphology (blue and red lines in Figure 1b) or chemical environment (green and pink lines in Figure 1b). Additional details about the fitting procedure are provided on pp S5−S10 in the Supporting Information. As indicated by the full width at half-maximum for each distribution in Figure 1b, we find that changing the elemental composition of the site has the greatest effect on adsorption strengths, followed by its coordination and finally the chemical environment, which can be modified by dynamic processes like segregation. The decreased sensitivity to perturbations in characteristics away from the adsorbate as seen in Figure 1b is consistent with the local nature of adsorbate-metal interactions, as longer-range interactions are screened out by itinerant d-electrons.56−58 This “nearsightedness”56−58 suggests that adsorption properties on generic bimetallic systems can be fully understood by confining our model to represent local coordination and composition in the immediate proximity of an adsorption site, yielding a substantially smaller parameter space than usually prevalent in nonlinear regression-based models. This deconstruction of 1853

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

Figure 2. (a) Representative structures of Pt−OH*, Pt2−OH*, and PtAu−OH*. The blue atoms correspond to a given local chemical environment that is varied from Ag, Au, Cu, Ir, Pd, Pt, to Rh. Pt, Au, O, and H atoms are depicted in gray, yellow, red, and white. The step edge of the (211) surface is marked with a dotted black line. (b), (c), (d) Energy correlations between stability of adsorption sites (top and bridge sites for OH*) and metal−OH* complexes. Modulating the local chemical environment for a given site coordination (e.g., CN: 3 for Pt in (b)) results in site-specific linear scaling relations. Slopes and intercepts are reported in Tables S40−S46 in the Supporting Information.

Figure 3. (a) Representative structures of Pt-CH3*, Pt3−CH*, and Pt−CO*. The blue atoms depict the local chemical environment which is varied along each scaling line. Pt, C, O, and H atoms are depicted in gray, brown, red, and white. (b), (c), (d) Energy correlations mapping site stability (top sites for CO*, CH3*, and fcc-hollow sites for CH*) to adsorption strengths of metal−adsorbate complexes.

1854

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

Figure 4. (a) Integrating site-specific scaling with traditional thermochemical scaling to predict site-specific adsorption energies of larger adsorbates. Scaling relations are constructed using adsorption site stabilities derived from either DFT calculations (scheme-1) or generated using αZi 1,(Z1Z2) “parameters (scheme-2). Scheme-2 is explained on pp S71 in the Supporting Information. (b), (c) Parity plots comparing predictions in CH3* adsorption (scheme-1 and scheme-2) with DFT derived numbers on Pt, Au sites in bimetallic alloys.

chemical compositions encompassing reactive metals like Hf, Os, Re, Ru, Y, and Zr also lie on the same scaling line as late transition metals. Site-specific scaling relations are also noted on bridge sites composed of Pt2 (Figure 2c) and PtAu (Figure 2d) as well as on top and bridge sites consisting of Pd and Au atoms (shown in Figures S16−S19 in the Supporting Information). Site stabilities, binding energies of metal− adsorbate complexes, and regression specifics of all correlations are provided in Tables S4−S67 in the Supporting Information. The site-specific scaling relations show a mean average error (MAE) on the order of 0.1 eV. In lieu of explicit DFT calculations as used in Figure 2b−d, site stabilities, which are descriptor values on the x-axis, can also be computed using the coordination-based alloy stability model45,46 resulting in scaling relations requiring one-third fewer DFT calculations with minimal loss of statistical accuracy. These correlations are shown in Figures S25−S37 in the Supporting Information. Figure 3b−d illustrates similar scaling relations linking site and metal−adsorbate stabilities for CH3* (top sites), CH* (hollow sites), and CO* (top sites). For a given site and coordination (e.g., Pt3 with CNs 5−5−5 in Figure 3c), variations in local compositions are all represented on a single scaling line. Correlations for Au-based sites are shown in Figures S17−S37 in the Supporting Information. Using CH* as an example, we show in Figure S38 in the Supporting Information that geometric relaxations of sites and adsorbates have minimal impact on slopes and intercepts of scaling relations. Previous studies have found that for such structureinsensitive scaling relations, slopes are proportional to the ratios of bond orders of the corresponding species.17 Because of bond order conservation, we anticipate that the bond orders of metal−adsorbate complexes (e.g., Pt−OH*) denoted by γM−OH*, will be slightly smaller than the bond orders of metal

an active site motif in a generic bimetallic alloy, allows us to quantify the stability of individual sites or site motifs through a recently introduced alloy stability model.45,46 Using these site stabilities to account for generic variations in structure and composition in bimetallic alloys, we identify a family of linear energy correlations that accurately predicts adsorption trends with site-by-site precision. We model dilute surface alloys by placing adsorption sites of a given elemental composition (e.g., Pt) and coordination number (e.g., 9 in fcc (111)) in different local chemical environments of late transition metals, ranging from Ag, Au, Cu, Ir, Pd, Pt, to Rh. In this study, we focus on top, bridge, and fcc-hollow sites consisting of Pt, Pd, and Au to represent reactive and inert late transition metals. Our analysis is generalizable to other late transition metals. We systematically modify site morphologies through the addition of adatoms on fcc (111), (100), and (211) surfaces such that their coordination numbers range from three to nine. Examples of selected sites are depicted in Figure 2a and Figure 3a with an exhaustive list provided in Figures S7−S15 in the Supporting Information. We focus on adsorption trends for typical catalytic descriptors, namely, OH* (top and bridge sites), CH3* (top site), CH* (fcc-hollow site), and CO* (top site). In Figure 2b, linear energy correlations between adsorption energies of metal−adsorbate complexes (Pt−OH*) and site stabilities (Pt*) are shown. Each correlation corresponds to a specific site coordination (e.g., CN: 5) with points along the line representing variations in local chemical environment. We observe a single correlation for a given site coordination, irrespective of both local chemical composition (Ag, Au, Cu, Ir, Pd, Pt, and Rh) and morphological arrangements in the second coordination shell and beyond (e.g., all sites with CN: 6 on fcc (100) and (211) surfaces are on one correlation). In Figures S20 and S22 in the Supporting Information, we show that local 1855

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

Figure 5. (a) Representative top and bridge sites displaying different local coordination on cuboctahedral (CUBO), octahedral (OCT), and decahedral (DEC) nanoparticles. Pt, Au, O, and H atoms are depicted in gray, yellow, red, and white. (b), (c) Parity plots comparing OH* adsorption energies predicted using site-specific scaling to DFT-derived energies. Site-specific scaling relations are derived using either site Z ,(Z Z ) stabilities calculated from DFT (scheme-1) or predicted with αi 1 1 2 parameters (scheme-2).

atoms (e.g., Pt*), which are denoted by γM. To a first approximation, we assume that a metal atom in an fcc stacking (e.g., Pt*) forms the equivalent of 12 bonds. Metal−adsorbate complexes (e.g., Pt−OH*) involving adsorbates that lack one valence electron (e.g., OH*, CH3*) will thus form the equivalent of 11 bonds. Hence, we expect a slope of around 11/12 (0.92) for the site-specific scaling relations (Figure 2b− d and Figure 3b−d): slope =

γM − OH* γM*

≈

γM* − 1 γM*

≈

11 12

scope of the study, we hypothesize it may arise from rehybridized electronic states at low-coordinated sites.59 By strategically deconstructing active sites in bimetallic alloys in terms of (1) elemental composition, (2) morphology, and (3) the chemical environment of surrounding atoms, we create a family of site-specific scaling lines that systematically encode variations in local coordination and the chemical environment across a vast combinatorial space. We will now illustrate two applications of these scaling relations. First, we will integrate site-specific scaling with traditional thermochemical scaling relations to predict adsorption energies of complex adsorbates with site-by-site precision. Second, we will utilize site-specific scaling relations constructed on idealized dilute alloys to predict adsorption properties of nanoparticles having generic structures and compositions. Structure- and composition-dependent adsorption energies for larger adsorbates can be predicted by combining sitespecific scaling relations (M−CH* vs M*) with traditional thermochemical scaling relations (CH3* vs CH* as an illustrative example). Figure 4a shows a flowchart for such an integrated scheme, which can predict composition- and structure-dependent energies of homologues reaction intermediates and transition states, leading to site-specific reactivity

(3)

Our analysis reveals that in the limit of 9-fold coordinated sites, slopes of site-specific scaling relations indeed approach 0.92. For lower coordinated sites (CN: 3, 4, 5), however, we observe a monotonic increase in the slopes. For instance, in the case of Pt−OH* vs Pt*, slopes increase from 0.93 to 1.18 as the site coordination decreases from 9-fold to 3-fold. Other examples are illustrated in Figure S39 in the Supporting Information. The slightly higher slopes for low-coordinated sites (CN: 3, 4, 5) reflect a larger than anticipated bond order for metal−adsorbate complexes relative to the metal atom. While a detailed explanation for this behavior is beyond the 1856

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

We rigorously compare our approach to existing adsorption models on the basis of generality of materials space screened, size of the parameter space, computational cost for training the model, and predictive performance on test sets in the Supporting Information. This comparison reveals that our model has two distinct advantages. First, we identify a single descriptor that links generic changes in site morphology and the local chemical environment to catalytic activity, thereby expanding the space of materials that can be screened. Second, we develop an operando computational framework in which dynamical changes of the nanoparticles and catalytic turnovers can be estimated on the fly since site stabilities per definition reflect the inherent stability of a given structure. By propagating site-specific scaling relations within microkinetic models, our approach will present a direct mapping between catalytic turnovers, site stabilities, and structural characteristics of active site ensembles like their morphology or chemical composition. In addition to reducing the materials gap between model catalysts and their working state, such mappings will pave the way toward the inverse design of alloy nanoparticles through robust property−structure relationships. In summary, by exploiting the natural correlation between binding energies of metal atoms and metal−adsorbate complexes, we directly calculate adsorption energies of catalytic descriptors as a function of local coordination and chemical environment in generic bimetallic alloy structures. By deconstructing adsorption sites in alloys in terms of their identity, coordination, and local chemical environment, we show that a family of site-specific energy correlations, established on idealized dilute alloy surfaces, can successfully predict adsorption energies on bimetallic nanoparticles spanning a vast structural and compositional design space. We further integrate site-specific correlations with traditional thermochemical scaling relations, to efficiently screen adsorption energies of larger reaction intermediates with siteby-site resolution. Our approach opens up pathways for nanoengineering thermodynamically stable active sites with atomic resolution to discover the next generation of bimetallic catalysts.

trends. Site-specific scaling relations (M−CH* vs M* as an example) are constructed using two approaches: scheme-1, wherein site stabilities (M*) are determined through DFT, or scheme-2, wherein site stabilities (M*) are predicted using Z ,(Z Z ) bond-associated parameters (αi 1 1 2 ) from the recently formulated coordination-based alloy stability model by Roling et al.45,46 (see pp S56−S68 in the Supporting Information for more details). Within this model, coordination numbers are defined as the number of bonds formed to atoms in the first coordination shell of unrelaxed structures that have bulk-like interatomic distances. For all structures in the present study, this definition implied counting the number of neighboring atoms having interatomic distances between 0.9 and 1.1 times the nominal interatomic distance. We note that using this model in scheme-2 reduces the number of required DFT calculations by one-third when compared to scheme-1. We then combine these site-specific relations with known scaling relations (CH3* vs CH*) between adsorption energies of larger adsorbates (CH3*) to catalytic descriptors (CH*).17 Parity plots obtained using scheme-1 and scheme-2 are shown in Figure 4b,c and reveal MAEs of 0.11 and 0.14 eV, respectively. The comparable MAEs between scheme-1 and scheme-2 validate the applicability of the less computationally intensive scheme-2 for creating site-specific scaling relations. Such integrated approaches can expeditiously predict thermodynamic changes of surface reactions within reaction networks with site-by-site resolution. We further evaluate the predictive power of site-specific scaling relations derived from simple slab calculations of dilute alloys by estimating adsorption energies on PtAu random alloy nanoparticles as a function of local structure and composition. The dissimilar lattice constants and differences in inherent reactivities of Pt and Au can cause local distortions, making PtAu nanoparticles an unbiased test system that is distinctive from the idealized dilute alloys within the training set. We calculate OH* adsorption energies on 61 different top and bridge sites of PtAu nanoparticles having cuboctahedral (CUBO, 147 atoms), octahedral (OCT, 146 atoms), and decahedral (DEC, 147 atoms) morphologies. Considered sites encompass wide-ranging coordination numbers and local chemical environments. Selected sites are indicated in Figure 5a with additional structures shown in Figures S41 and S42 in the Supporting Information. Adsorption energies of OH* are predicted from scaling relations generated using both scheme-1 and scheme-2, with parity plots shown in Figure 5b,c, respectively. DFT-derived and model-predicted adsorption energies and residuals for all sites are listed in Table S72 in the Supporting Information. To minimize computational cost, we directly evaluate the site stabilities of the 61 top and bridge sites in PtAu random alloys using the coordination-based alloy stability model.45,46 The scaling relations generate remarkably accurate site-specific adsorption energies with MAEs of 0.09 eV (0.11 eV) for top sites and 0.16 eV (0.18 eV) for bridge sites using scheme-1 (scheme-2). We see larger errors on bridge sites because of substantial geometric distortions in Au−Au bonds that decrease DFT-calculated energies in comparison with predictions from scaling together with finite size effects that are known to be present for nanoparticles of this size.47,60,61 We anticipate that these effects will vanish with increasing nanoparticle size, further improving the accuracy of our model. Parity plots in Figure 5b-c indicate that our framework can successfully estimate adsorption properties with atomic site resolution on generic bimetallic alloys.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpclett.9b00475. Computational details; description of binding sites and their adsorption energies (with both DFT and coordination-based model); illustrations of frequency and energy distributions, tables of binding energies, tables of standard deviations and FWHM for site coordination numbers and chemical ordering of the first coordination shell, adsorption structures; slopes, intercepts, and regression specifics of scaling relations; predictions of CH3* and OH* with site specificity; comparison to other models (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Tej S. Choksi: 0000-0002-9520-019X 1857

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

(15) Nørskov, J. K.; Abild-Pedersen, F.; Studt, F.; Bligaard, T. Density Functional Theory in Surface Chemistry and Catalysis. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 937−943. (16) Nørskov, J. K.; Bligaard, T.; Rossmeisl, J.; Christensen, C. H. Towards the Computational Design of Solid Catalysts. Nat. Chem. 2009, 1, 37−46. (17) Abild-Pedersen, F.; Greeley, J.; Studt, F.; Rossmeisl, J.; Munter, T. R.; Moses, P. G.; Skúlason, E.; Bligaard, T.; Nørskov, J. K. Scaling Properties of Adsorption Energies for Hydrogen-Containing Molecules on Transition-Metal Surfaces. Phys. Rev. Lett. 2007, 99, 16105. (18) Abild-Pedersen, F. Computational Catalyst Screening: Scaling, Bond-Order and Catalysis. Catal. Today 2016, 272, 6−13. (19) Calle-Vallejo, F.; Martínez, J. I.; García-Lastra, J. M.; Rossmeisl, J.; Koper, M. T. M. Physical and Chemical Nature of the Scaling Relations between Adsorption Energies of Atoms on Metal Surfaces. Phys. Rev. Lett. 2012, 108, 116103. (20) Hansen, H. A.; Varley, J. B.; Peterson, A. A.; Nørskov, J. K. Understanding Trends in the Electrocatalytic Activity of Metals and Enzymes for CO2 Reduction to CO. J. Phys. Chem. Lett. 2013, 4, 388−392. (21) Calle-Vallejo, F.; Martínez, J. I.; García-Lastra, J. M.; Sautet, P.; Loffreda, D. Fast Prediction of Adsorption Properties for Platinum Nanocatalysts with Generalized Coordination Numbers. Angew. Chem., Int. Ed. 2014, 53, 8316−8319. (22) Calle-Vallejo, F.; Loffreda, D.; Koper, M. T. M.; Sautet, P. Introducing Structural Sensitivity into Adsorption−Energy Scaling Relations by Means of Coordination Numbers. Nat. Chem. 2015, 7, 403−410. (23) Bazhenov, A. S.; Lefferts, L.; Honkala, K. Adsorption and Activation of Water on Cuboctahedral Rhodium and Platinum Nanoparticles. J. Phys. Chem. C 2017, 121, 4324−4331. (24) Ma, X.; Xin, H. Orbitalwise Coordination Number for Predicting Adsorption Properties of Metal Nanocatalysts. Phys. Rev. Lett. 2017, 118, 036101. (25) Wang, S.; Omidvar, N.; Marx, E.; Xin, H. Coordination Numbers for Unraveling Intrinsic Size Effects in Gold-Catalyzed CO Oxidation. Phys. Chem. Chem. Phys. 2018, 20, 6055−6059. (26) Calle-Vallejo, F.; Pohl, M. D.; Reinisch, D.; Loffreda, D.; Sautet Cd, P.; Bandarenka, A. S. Why Conclusions from Platinum Model Surfaces Do Not Necessarily Lead to Enhanced Nanoparticle Catalysts for the Oxygen Reduction Reaction. Chem. Sci. 2017, 8, 2283−2289. (27) Calle-Vallejo, F.; Tymoczko, J.; Colic, V.; Vu, Q. H.; Pohl, M. D.; Morgenstern, K.; Loffreda, D.; Sautet, P.; Schuhmann, W.; Bandarenka, A. S. Finding Optimal Surface Sites on Heterogeneous Catalysts by Counting Nearest Neighbors. Science 2015, 350, 185− 189. (28) Calle-Vallejo, F.; Pohl, M. D.; Bandarenka, A. S. Quantitative Coordination−Activity Relations for the Design of Enhanced Pt Catalysts for CO Electro-Oxidation. ACS Catal. 2017, 7, 4355−4359. (29) Rück, M.; Bandarenka, A.; Calle-Vallejo, F.; Gagliardi, A. Oxygen Reduction Reaction: Rapid Prediction of Mass Activity of Nanostructured Platinum Electrocatalysts. J. Phys. Chem. Lett. 2018, 9, 4463−4468. (30) Pohl, M. D.; Watzele, S.; Calle-Vallejo, F.; Bandarenka, A. S. Nature of Highly Active Electrocatalytic Sites for the Hydrogen Evolution Reaction at Pt Electrodes in Acidic Media. ACS Omega 2017, 2, 8141−8147. (31) Yan, Z.; Taylor, M. G.; Mascareno, A.; Mpourmpakis, G. Size-, Shape-, and Composition-Dependent Model for Metal Nanoparticle Stability Prediction. Nano Lett. 2018, 18, 2696−2704. (32) Jørgensen, M.; Grönbeck, H. Scaling Relations and Kinetic Monte Carlo Simulations To Bridge the Materials Gap in Heterogeneous Catalysis. ACS Catal. 2017, 7, 5054−5061. (33) Jørgensen, M.; Grönbeck, H. The Site-Assembly Determines Catalytic Activity of Nanoparticles. Angew. Chem., Int. Ed. 2018, 57, 5086−5089. (34) Holewinski, A.; Xin, H.; Nikolla, E.; Linic, S. Identifying Optimal Active Sites for Heterogeneous Catalysis by Metal Alloys

Luke T. Roling: 0000-0001-9742-2573 Frank Abild-Pedersen: 0000-0002-1911-074X Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We acknowledge financial support from the U.S. Department of Energy, Chemical Sciences, Geosciences, and Biosciences (CSGB) Division of the Office of Basic Energy Sciences, via Grant DE-AC02-76SF00515 to the SUNCAT Center for Interface Science and Catalysis. V.S. gratefully acknowledges financial support from the Alexander von Humboldt Foundation.

■

REFERENCES

(1) Boes, J. R.; Gumuslu, G.; Miller, J. B.; Gellman, A. J.; Kitchin, J. R. Estimating Bulk-Composition-Dependent H2 Adsorption Energies on CuxPd1−x Alloy (111) Surfaces. ACS Catal. 2015, 5, 1020−1026. (2) Greeley, J.; Mavrikakis, M. Alloy Catalysts Designed from First Principles. Nat. Mater. 2004, 3, 810−815. (3) Kitchin, J. R.; Nørskov, J. K.; Barteau, M. A.; Chen, J. G. Modification of the Surface Electronic and Chemical Properties of Pt(111) by Subsurface 3d Transition Metals. J. Chem. Phys. 2004, 120, 10240−10246. (4) Andersson, M. P.; Bligaard, T.; Kustov, A.; Larsen, K. E.; Greeley, J.; Johannessen, T.; Christensen, C. H.; Nørskov, J. K. Toward Computational Screening in Heterogeneous Catalysis: Pareto-Optimal Methanation Catalysts. J. Catal. 2006, 239, 501−506. (5) Dathar, G. K. P.; Shelton, W. A.; Xu, Y. Trends in the Catalytic Activity of Transition Metals for the Oxygen Reduction Reaction by Lithium. J. Phys. Chem. Lett. 2012, 3, 891−895. (6) Schumann, J.; Medford, A. J.; Yoo, J. S.; Zhao, Z.-J.; Bothra, P.; Cao, A.; Studt, F.; Abild-Pedersen, F.; Nørskov, J. K. Selectivity of Synthesis Gas Conversion to C2+ Oxygenates on fcc(111) TransitionMetal Surfaces. ACS Catal. 2018, 8, 3447−3453. (7) Snider, J. L.; Streibel, V.; Hubert, M. A.; Choksi, T. S.; Valle, E.; Upham, D. C.; Schumann, J.; Duyar, M. S.; Gallo, A.; Abild-Pedersen, F.; et al. Revealing the Synergy between Oxide and Alloy Phases on the Performance of Bimetallic In−Pd Catalysts for CO2 Hydrogenation to Methanol. ACS Catal. 2019, 9, 3399−3412. (8) Greeley, J.; Stephens, I. E. L.; Bondarenko, A. S.; Johansson, T. P.; Hansen, H. A.; Jaramillo, T. F.; Rossmeisl, J.; Chorkendorff, I.; Nørskov, J. K. Alloys of Platinum and Early Transition Metals as Oxygen Reduction Electrocatalysts. Nat. Chem. 2009, 1, 552−556. (9) Luo, L.; Zhang, L.; Henkelman, G.; Crooks, R. M. Unusual Activity Trend for CO Oxidation on PdxAu140−x@Pt Core@Shell Nanoparticle Electrocatalysts. J. Phys. Chem. Lett. 2015, 6, 2562− 2568. (10) Rankin, R. B.; Greeley, J. Trends in Selective Hydrogen Peroxide Production on Transition Metal Surfaces from First Principles. ACS Catal. 2012, 2, 2664−2672. (11) Studt, F.; Sharafutdinov, I.; Abild-Pedersen, F.; Elkjær, C. F.; Hummelshøj, J. S.; Dahl, S.; Chorkendorff, I.; Nørskov, J. K. Discovery of a Ni-Ga Catalyst for Carbon Dioxide Reduction to Methanol. Nat. Chem. 2014, 6, 320−324. (12) Wang, C.; Li, D.; Chi, M.; Pearson, J.; Rankin, R. B.; Greeley, J.; Duan, Z.; Wang, G.; van der Vliet, D.; More, K. L.; et al. Rational Development of Ternary Alloy Electrocatalysts. J. Phys. Chem. Lett. 2012, 3, 1668−1673. (13) Wilson, N. M.; Priyadarshini, P.; Kunz, S.; Flaherty, D. W. Direct Synthesis of H2O2 on Pd and AuxPd1 Clusters: Understanding the Effects of Alloying Pd with Au. J. Catal. 2018, 357, 163−175. (14) Tang, W.; Zhang, L.; Henkelman, G. Catalytic Activity of Pd/ Cu Random Alloy Nanoparticles for Oxygen Reduction. J. Phys. Chem. Lett. 2011, 2, 1328−1331. 1858

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859

Letter

The Journal of Physical Chemistry Letters

Electronic Structure Database for Surface Reactions. ChemRxiv 2018, DOI: 10.26434/chemrxiv.7336094.v2, (Preprint) . (56) Kohn, W. Density Functional and Density Matrix Method Scaling Linearly with the Number of Atoms. Phys. Rev. Lett. 1996, 76, 3168−3171. (57) Prodan, E.; Kohn, W. Nearsightedness of Electronic Matter. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 11635−11638. (58) Nørskov, J. K.; Bligaard, T.; Hvolbæk, B.; Abild-Pedersen, F.; Chorkendorff, I.; Christensen, C. H. The Nature of the Active Site in Heterogeneous Metal Catalysis. Chem. Soc. Rev. 2008, 37, 2163− 2171. (59) Stenlid, J. H.; Brinck, T. Extending the σ-Hole Concept to Metals: An Electrostatic Interpretation of the Effects of Nanostructure in Gold and Platinum Catalysis. J. Am. Chem. Soc. 2017, 139, 11012− 11015. (60) Kleis, J.; Greeley, J.; Romero, N. A.; Morozov, V. A.; Falsig, H.; Larsen, A. H.; Lu, J.; Mortensen, J. J.; Dułak, M.; Thygesen, K. S.; et al. Finite Size Effects in Chemical Bonding: From Small Clusters to Solids. Catal. Lett. 2011, 141, 1067−1071. (61) Li, L.; Larsen, A. H.; Romero, N. A.; Morozov, V. A.; Glinsvad, C.; Abild-Pedersen, F.; Greeley, J.; Jacobsen, K. W.; Nørskov, J. K. Investigation of Catalytic Finite-Size-Effects of Platinum Metal Clusters. J. Phys. Chem. Lett. 2013, 4, 222−226.

Based on Molecular Descriptors and Electronic Structure Engineering. Curr. Opin. Chem. Eng. 2013, 2, 312−319. (35) Xin, H.; Holewinski, A.; Linic, S. Predictive Structure Reactivity Models for Rapid Screening of Pt-Based Multimetallic Electrocatalysts for the Oxygen Reduction Reaction. ACS Catal. 2012, 2, 12−16. (36) Saravanan, K.; Kitchin, J. R.; Von Lilienfeld, O. A.; Keith, J. A. Alchemical Predictions for Computational Catalysis: Potential and Limitations. J. Phys. Chem. Lett. 2017, 8, 5002−5007. (37) Eom, T.; Kim, W. J.; Lim, H.-K.; Han, M. H.; Han, K. H.; Lee, E.-K.; Lebègue, S.; Hwang, Y. J.; Min, B. K.; Kim, H. Cluster Expansion Method for Simulating Realistic Size of Nanoparticle Catalysts with an Application in CO2 Electroreduction. J. Phys. Chem. C 2018, 122, 9245−9254. (38) Wang, L.-L.; Tan, T. L.; Johnson, D. D. Configurational Thermodynamics of Alloyed Nanoparticles with Adsorbates. Nano Lett. 2014, 14, 7077−7084. (39) Jinnouchi, R.; Asahi, R. Predicting Catalytic Activity of Nanoparticles by a DFT-Aided Machine-Learning Algorithm. J. Phys. Chem. Lett. 2017, 8, 4279−4283. (40) Jinnouchi, R.; Hirata, H.; Asahi, R. Extrapolating Energetics on Clusters and Single-Crystal Surfaces to Nanoparticles by MachineLearning Scheme. J. Phys. Chem. C 2017, 121, 26397−26405. (41) Li, Z.; Wang, S.; Chin, W. S.; Achenie, L. E.; Xin, H. HighThroughput Screening of Bimetallic Catalysts Enabled by Machine Learning. J. Mater. Chem. A 2017, 5, 24131−24138. (42) Ma, X.; Li, Z.; Achenie, L. E. K.; Xin, H. Machine-LearningAugmented Chemisorption Model for CO2 Electroreduction Catalyst Screening. J. Phys. Chem. Lett. 2015, 6, 3528−3533. (43) Wang, S.; Omidvar, N.; Marx, E.; Xin, H. Overcoming Site Heterogeneity In Search of Metal Nanocatalysts. ACS Comb. Sci. 2018, 20, 567−572. (44) Li, Z.; Ma, X.; Xin, H. Feature Engineering of MachineLearning Chemisorption Models for Catalyst Design. Catal. Today 2017, 280, 232−238. (45) Roling, L. T.; Choksi, T. S.; Abild-Pedersen, F. A CoordinationBased Model for Transition Metal Alloy Nanoparticles. Nanoscale 2019, 11, 4438−4452. (46) Roling, L. T.; Li, L.; Abild-Pedersen, F. Configurational Energies of Nanoparticles Based on Metal−Metal Coordination. J. Phys. Chem. C 2017, 121, 23002−20010. (47) Roling, L. T.; Abild-Pedersen, F. Structure-Sensitive Scaling Relations: Adsorption Energies from Surface Site Stability. ChemCatChem 2018, 10, 1643−1650. (48) Campbell, C. T.; Sellers, J. R. V. Anchored Metal Nanoparticles: Effects of Support and Size on Their Energy, Sintering Resistance and Reactivity. Faraday Discuss. 2013, 162, 9−13. (49) Campbell, C. T. The Energetics of Supported Metal Nanoparticles: Relationships to Sintering Rates and Catalytic Activity. Acc. Chem. Res. 2013, 46, 1712−1719. (50) Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.; Cavazzoni, C.; Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.; Dabo, I.; et al. QUANTUM ESPRESSO: A Modular and Open-Source Software Project for Quantum Simulations of Materials. J. Phys.: Condens. Matter 2009, 21, 395502. (51) Larsen, A. H.; Mortensen, J. J.; Blomqvist, J.; Castelli, I. E.; Christensen, R.; Dułak, M.; Friis, J.; Groves, M. N.; Hammer, B.; Hargus, C.; et al. The Atomic Simulation Environmenta Python Library for Working with Atoms. J. Phys.: Condens. Matter 2017, 29, 273002. (52) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Improved Adsorption Energetics within Density-Functional Theory Using Revised Perdew-Burke-Ernzerhof Functionals. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 7413. (53) Vanderbilt, D. Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue Formalism. Phys. Rev. B 1990, 41, No. 7892R. (54) Monkhorst, H. J.; Pack, J. D. Special Points for Brillonin-Zone Integrations. Phys. Rev. B 1976, 13, 5188. (55) Winther, K.; Hoffmann, M. J.; Mamun, O.; Boes, J. R.; Nørskov, J. K.; Bajdich, M.; Bligaard, T. Catalysis-Hub.Org: An Open 1859

DOI: 10.1021/acs.jpclett.9b00475 J. Phys. Chem. Lett. 2019, 10, 1852−1859