Nanoparticle Catalysts

Overview

The science and technology of catalysis is particularly important at this time due to the energy and environmental challenges facing society. Research in the 1990s, showing surprising activity of Au nanoparticles, has largely motivated a search for new catalytic materials on the nanoscale with properties that are different from their bulk counterparts. Another significant factor in the development of new catalysts has been the growth of computing power and improvements in theoretical methods to help understand experiments at the atomic scale and to provide guidelines for catalyst design. Experiments, demonstrating the high activity of nanoparticle catalysts, have inspired the development of theoretical methods for calculating reaction mechanisms and screening for new catalysts. Iterating between theory and experiment is a promising strategy for understanding nanoparticle catalysis and reducing the cost of the development cycle for new catalysts.


Our group tightly collaborate with experiment group lead by Prof. Richard Crooks at UT Austin, where dendrimer-encapsulated nanoparticles (DENs), as a model system are synthesized and characterized at atomic level for direct comparison with theory. Summarized by the figure above, our collaboration leads to refinement of the theory and the prediction of better nanoparticle electrocatalyst candidates. This in turn leads to the next generation of more active electrocatalysts that can be used for a wide variety of applications. Detailed review of our work has been presented in the following two papers:

1. R. M. Anderson, D. F. Yancey, L. Zhang, S. T. Chill, G. Henkelman, and R. M. Crooks, A Theoretical and Experimental Approach for Correlating Nanoparticle Structure and Electrocatalytic Activity, Acc. Chem. Res. 48 1351-1357 (2015). DOI

2. L. Zhang, R. M. Anderson, R. M. Crooks, and G. Henkelman, Correlating Structure and Function of Metal Nanoparticles for Catalysis, Surf. Sci. 640 65-72 (2015). DOI

Our group also collaborates with experimental groups in UT Austin lead by Prof. Simon Humphrey and Prof. Charles Werth. Humphrey's group specializes in microwave assisted synthesis of alloy nanoparticles with conventionally immiscible metals which have catalysis applications like nitrate/nitrite reduction, methane selective activation. We use computational tools to suggest optimal alloy candidates for these catalysis applications.

Nitrate/Nitrite Reduction

Nitrate (NO₃^-) /Nitrite (NO₂^-) contamination of drinking water is a major problem in rural areas (agricultural) around the world mainly due to the extensive use of nitrogen based fertilizers. Ingestion of NO₃^- and NO₂^- leads to the formation of N-nitroso compounds which are carcinogenic, harmful for brain development of infants, causes thyroid diseases and many other health concerns. There are many ways to remove these ions like anion exchange, reverse osmosis etc but these methods have issues like generation of secondary waste, large industrial scale start-up times, high energy consumption and low ion removal efficiencies. Catalytic reduction alleviates these issues compared to the other methods, so we use direct catalytic reduction of these nitrates/nitrites to nitrogen gas (N₂) which makes it easier to treat drinking water on site. One of the key challenges is to control the selectivity of the reaction products, apart from N₂ gas we also get ammonia (NH₃) gas as a byproduct which is not desirable because NH₃ dissolves in the water making this non-potable. To control the selectivity of this reaction we used Pd-Ru (Palladium-Ruthenium) NP catalyst and found the selectivity of the reaction can be controlled by adjusting the composition of the NP. Our results show that high Pd composition favors N₂ selectivity. Experimental collaborators: Dr. Werth’s and Dr. Humphrey’s groups. This project will have a potential impact on the development of cost effective ways to treat drinking water contaminated with nitrates/nitrites.

Methane Selective Activation

Determining Nanoparticles Structure

In order to model the properties of DENs it is necessary to characterize their structure. Determining the overall shape of 1-2 nm particles using only experimental methods is challenging. Validity of common characterization techniques, such as XRD and EXASF, on such small scale is still under estimation. On the other hand, in principle, it is possible to take a purely theoretical approach to determine the particle's geometry by locating the structure that corresponds to the global minimum potential energy as calculated by DFT. However the system is much larger than just the particle(interactions with the solvent and the dendrimer should be considered), greatly increasing the computational cost. Thus there is also a need for experimental structural information input on the theory side. We are developing new structural analysis tools, which intergrates the DFT simulation with experiment measured electro-chemical voltammogram, TEM image, XRD and EXAFS spectrum. This will allow an unprecedented level of feedback of experimental data to support the leading computational models.

• DFT meets XRD-PDF






1.7 nm Pt nanoparticles were synthesized as DENs by the Crooks group and as monolayer-protected clusters (MPCs) by our collaborators. The MPCs produced particles with ligands bound to their surface, while the DENs yielded ligand-free particles. X-ray diffraction (XRD) data obtained at the APL showed that the ligand-bound particles were more disordered than the ligand-free particles. Previous attempts to fit structures to the X-ray data gave disordered particles, which theory predicts to be unstable. Structures selected for their theoretical stability, however, were found to match the X-ray data poorly. Therefore, structures were selected for a combination of their match to theory (lower energy is better) and their match to experiment (lower PDF error is better) in figure 1. The fit of three commonly suggested nanoparticle structures shows that the truncated octahedron structure best fits the data. In fact, a set of global searches showed that the truncated octahedron motif best fit the data for all groups of particles, indicating that surface ligands only slightly perturbed the surface, as shown in figure 2. This technique allows us to better approximate the structure of nanoparticles, which cannot be independently determined by more direct means.
References
M. Welborn, W. Tang. J. Ryu, V. Petkov, and G. Henkelman, A combined density functional and x-ray diffraction study of Pt nanoparticle structure, J. Chem. Phys. 135, 014503 (2011). DOI



• DFT meets Underpotential Deposition Voltammogram



Cu Underpotential Deposition on Pt DENs




We have been using Cu underpotential deposition (UPD) to probe the structure of Pt DENs. UPD provides information about the surface of DENs, and hence complements other characterization tools in our arsenal (principally EXAFS, XRD-PDF, Monte Carlo simulations, and TEM). The largest DENs we examined consisted of ~225 or ~147 Pt atoms, and these materials yielded two distinct Cu UPD waves. Smaller cores, containing just ~55 Pt atoms, revealed a single, broad UPD wave. Because the different crystallographic planes of Pt give rise to Cu UPD waves at different electrochemical potentials, we interpreted these results to mean that distinct facets emerge on the larger particles. Through a combined analysis of the Cu shell coverage, Θ_Cu, determined using electrochemical methods and in-situ EXAFS, we found that Cu UPD onto Pt DENs having 225 and 147 Pt atoms results in core@shell configurations with a complete Cu shell. This was an important finding, because it provides us with a definitive electrosynthetic method for preparing well-defined core@shell nanoparticles having a single-atom-thick shell.

To gain a more complete picture of the Pt@Cu DENs prepared by UPD, we used DFT calculations to interpret the electrochemical and in-situ EXAFS results. Figure 1 shows the calculated potentials for Cu UPD and stripping with and without SO₄ adsorbed to the Pt DEN surface. The calculated results are overlayed onto an experimentally derived voltammogram. Clearly, better agreement between theory and experiment is obtained when SO₄ is present on the Pt surface. These findings have provided important insights on the interplay between shell layers and adsorbed ligands during the UPD process; the extension to catalysis is obvious.
Au₁₄₇@M_x (M=Pb, Pt; x=54, 102) DENs




We recently showed that Au@Pt DENs could be synthesized by the UPD of Cu onto a Au core followed by galvanic displacement of the Cu by Pt. We found that the ORR activity of these core@shell particles was similar to that of Pt DENs of the same size. Since then, we have started to develop more sophisticated methods for synthesizing very well defined core@shell DENs by UPD followed by galvanic displacement for other noble metals. Our first task was to understand the UPD of Pb onto cuboctahedral Au₁₄₇ particle surfaces. To achieve this, we modeled the UPD of Pb on Au₁₄₇ with DFT by taking an atom-by-atom approach. The calculated deposition potentials (relative to the bulk reduction of Pb) were in excellent agreement with the experimental cyclic voltammograms, as shown in Figure 3. The calculations showed that Pb deposits on the (100) surface orientation first, followed by deposition on the (111) surface facets (Figures 4a and 4b). DFT calculations also showed that Pb forms a stable overlayer on the Au₁₄₇ surface facets and that the core@shell structure is stable. A partial shell structure having 54 Pb shell atoms (9 atoms on each of 6 (100) facets) and a full shell structure having an additional 48 Pb atoms (6 atoms on each of 8 (111) facets) were modeled. Pb UPD was not predicted to take place on the edge and corner sites of the Au₁₄₇ particles.

Structures corresponding to those evaluated by UPD were also synthesized and characterized. Subsequently, the Pb was galvanically exchanged for Pt to create either Au₁₄₇@Pt₅₄ (partial shell structure) or Au₁₄₇Pt₁₀₂ (full shell structure) (Figure 3). The surface composition was determined by cyclic voltammetry. Specifically, the Au surface area was determined by integrating the charge required to reduce Au oxides, and the Pt surface area was measured by integrating the charge necessary to adsorb a monolayer of H atoms. From these values, the ratio of Pt surface area to the total surface area (Au + Pt) was determined to be 0.39 and 0.57 for the partial Pt shell and full Pt shells, respectively, which is in good agreement with the DFT-predicted surface structure. The Au₁₄₇@Pt₅₄ and Au₁₄₇@Pt₁₀₂ structures were used to catalyze the ORR, and both exhibited similar activity (0.98 and 0.92 mA/cm2 @ 0.8 V vs RHE, respectively). Initially, this result was perplexing, but DFT molecular dynamics (MD) simulations revealed that Pt deposited on the (100) facets of the particles undergoes a rearrangement to form (111) ordered domains (right side of Figures 4a and 4b). DFT calculations also showed that the (111) diamond shaped facets on Au₁₄₇@Pt₅₄ and Au₁₄₇@Pt₁₀₂ are the most active sites for the ORR, and this explains the similarity in activity of the two different structures. This is a key finding, because it suggests that very subtle changes in catalyst structure control reactivity.


References
1. E. Carino, H.-Y. Kim, G. Henkelman, and R. M. Crooks, Site-selective Cu deposition on Pt dendrimer-encapsulated nanoparticles: Correlation of theory and experiment, J. Am. Chem. Soc. 134, 4153-4162 (2012). DOI
2. D. F. Yancey, L. Zhang, R. M. Crooks, and G. Henkelman, Au@Pt dendrimer encapsulated nanoparticles as model electrocatalysts for comparison of experiment and theory, Chem. Sci. 3, 1033-1040 (2012). DOI

Correlating nanoparticle Strcuture and Function

Alloy nanoparticles have a broad range of applications for many catalytic processes. In this research, we focus on the oxygen reduction reaction (ORR). The ORR is the cathode reaction of proton-exchange membrane fuel cells. It involves at least two elementary processes, O₂ dissociation and the subsequent removal of the dissociated species by further reduction to H₂O. ORR activity can be measured experimentally using electrochemical techniques, including cyclic voltammetry (CV) and rotating disk voltammetry (RDV). Theoretically, the ORR activity of a surface is understood to correlate with the O binding energy. Pioneering work by Nørskov and co-workers have established a volcano-shaped correlation between the O binding energy and ORR activity. On the strong-binding side of the volcano, oxygen species over-bind to the catalyst surface and limit the kinetics of product formation. On the weak-binding side of the volcano the reaction is limited by oxygen dissociation and adsorption. Pt offers the optimal tradeoff of the pure metals and the peak of the volcano corresponds to an oxygen binding energy which is slightly weaker than on a Pt(111) surface. The underlying reason that oxygen binding is a good reactivity descriptor is that there is a Brønsted-Evans-Polanyi (BEP) correlation between the binding and transition state energies, so that a single descriptor can describe both limits.In the following, we consider several nanoparticle alloys strucutures and seek to find structures with oxygen binding energies, and hence ORR activities, that approach or exceed that of Pt.

• Core@Shell Nanoparticles



Charge Redistribution in Core-shell Nanoparticles to Promote Oxygen Reduction
Bimetallic core-shell nanoparticles are a class of near surface alloy catalyst for which there is a high degree of control over size and composition. A challenge for theory is to understand the relationship between their structure and catalytic function, and provide guidelines to design new catalysts that take advantage of materials properties arising at the nanoscale. In this work, we use density functional theory to calculate the energetics of oxygen dissociative adsorption on 1 nm Pd-shell nanoparticles with a series of core metals. The barrier for this reaction and the binding energy of atomic oxygen is found to correlate well with the d-band level of the surface electrons. Noble metal cores lower the barrier and increase the binding, reducing the activity of the Pd-shell as compared to Pt. Reactive core metals such as Co and Mo, on the other hand, lower the d-band of the shell with respect to the Fermi level, giving the Pd-shelled particles oxygen reduction kinetics similar to that of Pt. While both ligand and strain effects determine the d-band center of the Pd shell, a greater surface relaxation reduces the strain in nanoparticles as compared to single crystal near-surface alloys. Charge redistribution between core and shell then becomes an important factor for lowering the d-band center of Pd-shelled particles and increasing their activity for the oxygen reduction reaction.



Optimizing nanoparticle catalysts with a genetic algorithm
Platinum-based fuel cells offer an attractive alternative to internal combustion engines as a future means of utilizing chemical energy. There are, however, shortcomings of such technologies that must be resolved if they are to become practical and widespread. Some of these difficulties include the CO poisoning, the short lifetime of electrodes in acidic environments, the ~30% energy loss due to slow oxygen reduction kinetics at the cathode, and the high material cost and limited supply of platinum itself. Cheaper, more effective electrocatalysts need to be developed, yet the task of discovering novel platinum alternatives has proven to be extremely challenging.



We have implemented a genetic algorithm (GA) within density functional theory to investigate the catalytic properties of a large number of 38- and 79-atom bimetallic core-shell nanoparticles for the oxygen reduction reaction (ORR) in an effort to identify promising platinum replacement catalysts for use in fuel cells. In the GA, each nanoparticle is represented by a two-gene chromosome that identifies its core and shell metals. The fitness of each particle is specified by how closely the d-band level of the shell electrons matches that of the Pt(111) surface, a catalyst known to be effective for the ORR. The GA starts by creating an initial population of random core-shell particles; the fittest particles are then bred and mutated to replace the least-fit particles in the population and form successive generations. Within a few generations, the average energy of the surface d-band electrons in the nanoparticles is tuned to that of Pt(111). These nanoparticles may possess similar catalytic properties as Pt(111) for the ORR. Promising core-shell metal combinations revealed by the GA are shown above.
References
1. Wenjie Tang and Graeme Henkelman, Charge redistribution in core/shell nanoparticles to promote oxygen reduction, J. Chem. Phys. 130 194504 (2009).
2. N. S. Froemming and G. Henkelman, Optimizing core-shell nanoparticle catalysts with a genetic algorithm J. Chem. Phys. 131, 234103 (2009).



• Random Alloy Nanopartcles



Role of Geometric Relaxation



Calculations were carried out to study the role of geometric relaxation in the ORR for nanoparticle of various transition metals. Previous experimental results by Crooks group have shown enhancement of ORR activity by the addition of small amounts of Pd to a Pt DEN containing ~180 atoms. To study the geometric effect, we decomposed the total binding energy (E_tot, Figure 1) into two parts: the energy of geometric relaxation (E_r) and the energy of oxygen affinity (E_a). The energy decomposition in Figure 1 exhibits strong size dependence in E_r. The value of E_r becomes comparable to E_a when the particle size goes down, and the effect is significant at high Pt content. Also, the E_r curve of 79-atom particle is actually not linear but convex. All these factors together can explain why the E_tot trend in 79-atom particle reverses at high Pt ratio and looks so different from the curves of the slab and the 140-atom particle. Our calculations show that the structural deformation induced by atomic oxygen binding can energetically stabilize the oxidized states and thus reduces the catalytic activity. One key finding is that the catalytic performance can be improved by making alloys with less deformable metals.
Role of Electronic Effects



While geometric relaxation is important to understand the ORR activity trend of Pd/Pt random alloy nanoparticles, electronic effects are dominant in random alloy nanoparticles formed by two hard elements, such as Pd and Cu. Geometric and electronic effects were studied to determine their contributions for tuning the oxygen binding. The geometric effect is independent of the Cu ratio, and therefore only the electronic effect is responsible for the oxygen binding trend. The oxygen binding energy trend of Pd/Cu random alloy NP79 was studied for a range of compositions (Cu% = 0, 0.25, 0.5, 0.75 and 1), as shown in Fig. 3. Although Cu NP79 itself binds oxygen more strongly than Pd NP79, the addition of Cu to Pd initially weakens the average oxygen binding until it reaches the extremum at Pd:Cu ratio of 1:1. While all the Pd/Cu random NP79 alloys bind oxygen more strongely than Pt(111) (dashed line in Fig. 3), the highest ORR activity should be achieved at the weakest oxygen binding composition, that is a Pd:Cu ratio near 1:1. This optimal ratio from our calculation matches well with previous experiments, where Cu has been found to promote the ORR activity of Pd at a ratio of 50%.

When Cu is alloyed into Pd, charge shifts from Cu to Pd since Cu has a higher Fermi level. This charge redistribution lowers the Pd d-band and raises the Cu d-band. Thus, the Cu atoms neighboring Pd decrease the Pd-O bond strength, while Pd atoms neighboring Cu increase Cu-O binding. These effects are not symmetric, however, as shown in Fig. 4, where the oxygen binding was decomposed into Pd-O and Cu-O interactions as a function of the Cu ratio according to the d-band model. The slope of the Pd-O interaction, is larger in magnitude than the Cu-O slope. In other words, the alloying of Pd and Cu weakens the Pd-O interaction more than it strengthens the Cu-O interaction. The average binding to an alloy particle can be understood as a linear combination of Pd-O and Cu-O interactions, weighted by the fraction of each metal in the particle. The Pd-O and Cu-O interactions can be estimated as a linear function of composition using their slopes. In the final expression, the first two terms are linear interpolations between the binding at the pure metal particles, and the last term is a quadratic function that describes the relative change in binding due the metals’ influence on each other. The difference in magnitude of slopes of Pd−O and Cu−O gives rises to the nonlinear binding trend with a curvature towards weak binding at intermediate compositions.

We expect the above non-linear binding trend to be a general prop- erty of adsorbate binding to random alloys, providing a prescription for tuning catalytic activity through alloying. The curvature is determined by the sign and relative magnitude of slopes, which can be estimated from the oxygen binding energy on core@shell nanoparticles. As shown in Fig. 5, Pd/X (X = Au, Cu, Ir) random alloy NPs all exhibit non-linear oxygen binding trends. The oxygen binding trend for Pd/Ir alloys has the opposite curvature of Pd/Cu and Pd/Au. This can be understood by considering, for example, the Pd/Au alloy. Au has a lower Fermi level than Pd so charge transfers from Pd to Au when they are alloyed.

References
1. C.-Y. Lu and G. Henkelman, Role of geometric relaxation in oxygen binding to metal nanoparticles, J. Phys. Chem. Lett. 2, 1237-1240 (2011). DOI
2. W. Tang, L. Zhang, and G. Henkelman, Catalytic activity of Pd/Cu random alloy nanoparticles for oxygen reduction, J. Phys. Chem. Lett. 2, 1328-1331 (2011). DOI [Erratum]



• Alloy-Core@Shell Nanoparticles



Motivation for alloy-core@shell Nanoparticles design

As shown in the sections random-alloy and core@shell are two structures that have been well-studied to search for bimetallic catalysts. They are often studied because of the opportunity to tune electrocatalytic properties by controlling the structure of the particles. However, there are some shortcomings of these two structures from the viewpoint of particle synthesis and design. Variations in the core@shell structures are discrete in chemical compound space (elements can only be changed by integer atomic numbers) so that the properties of such particles can not be tuned precisely. As shown in core@shell nanoparticles research page, none of the single core@Pd NP79 achieve the same oxygen binding as Pt(111).

The properties of the random alloy structure, on the other hand, can be finely tuned by varying the composition of the components. However the presence of reactive metals on the surface of the alloy nanoparticle could be unstable under reaction conditions and dissolve into solution. The combined alloy-core@shell structure aims to take advantage of the strengths of the alloy and core@shell particles, and avoid their flaws. Noble metal shells (e.g. Pt, Pd) are stable during synthesis and characterization, and the alloy core offers the opportunity to tune the electronic and geometric properties of the shell by changing its composition. The predictive power of first principles theory in leading the design cycle for novel materials is also demonstrated here for the ORR.
Linear Binding Correlation of Alloy-core@shell Nanoparticles

The average oxygen binding energy at Pd/X (X = Ir, Rh, Cu, Ru and Mo) alloy core with Pd shell nanoparticles are studied as a function of the ratio of metal X in the alloy core, as shown in Fig. 2(a). The red dashed line marks the oxygen binding on Pt(111), which is chosen as the target value. X@Pd (X = Ir, Rh, Cu, Ru and Mo) bind oxygen too strongly as compared to Pt(111), while Cu@Pd binds oxygen too weakly -they are on the opposite branches of the volcano plot. All of these particles are expected to have lower activity as compared with Pt(111). Our calculations show that the oxygen binding energy can be weakened linearly by mixing Cu into these cores. As demonstrated in Fig. 2(b-d), this linear binding correlation is independent of types of core or shell metals, adsorbate molecules, size of particles, thereby providing a general description for tuning principle of alloy-core@shell nanoparticles.This linear tuning mechanism offers a new scheme for catalyst design. Taking the ORR for example, the target oxygen binding energy of X_xY_1-x@Z can be achieved by alloying a metal X and Y, where X@Z and Y@Z are on the opposite sides of the target value. The optimal ratio x^* for X_xY_1-x@Z to achieve target binding is given by its linear interpolation.
Linear Binding Correlation of Alloy-core@shell Nanoparticles

Following the above design principle, several alloy-core compositions with a Pt shell emerge as good ORR catalysts. The optimal compositions can be obtained from Fig. 2(b). One such promising structure is Pd_0.72Au_0.28@Pt. Based on our theoretical prediction, we synthesized a series of PdAu@Pt DENs with different Pd:Au ratios in the alloy core. The Pd/Au random alloy core was prepared by co-reduction. Then Cu UPD was conducted on the Pd/Au random alloy DENs modified electrode to form a homogenous Cu shell. Finally, Pd/Au@Pt DENs were then synthesized by a subsequent galvanic exchange, replacing the Cu shell with Pt.

The ORR activity of these PdAu@Pt DENs was determined using RDV, by measuring the onset potential where the current density reached 0.1 mA/cm². Fig. 3(b) plots the RDV measured ORR onset potential against the corresponding oxygen binding energy calculated from DFT, exhibiting a volcano-like activity trend. Notably, data for Pd@Pt DENs is not shown on Fig. 3 because there was core/shell inversion, which was confirmed by EXAFS and DFT. Alloying Au to the Pd-core not only optimized the catalytic activity but also enhanced its stability. The peak activity was achieved with 75% Pd in the core, in perfect agreement with our prediction from the linear binding model. As shown in the above section, this linearly tuning mechanism is a general property of alloy-core@shell nanoparticles. Therefore, similar catalyst design scheme based on alloy-core@shell nanoparticles is expected to be transferable to other important catalytic processes.
References
1. L. Zhang and G. Henkelman, Tuning the oxygen reduction activity of Pd shell nanoparticles with random alloy cores, J. Phys. Chem. C 116, 20860-20865 (2012). DOI
2. L. Zhang, R. Iyyamperumal, D. F. Yancey, R. M. Crooks, and G. Henkelman, Design of Pt-shell nanoparticles with alloy cores for the oxygen reduction reaction, ACS Nano 7, 9168-9172 (2013). DOI
3. L. Zhang and G. Henkelman, Computational design of alloy-core@shell metal nanoparticle catalysts, ACS Catal. 5, 655-660 (2015). DOI
4. L. Luo, L. Zhang, G. Henkelman, and R. M. Crooks, Unusual activity trend for CO oxidation on Pd_xAu_140-x@Pt core@shell nanoparticle electrocatalysts, J. Phys. Chem. Lett. 6 2562-2568 (2015). DOI