High-throughput computational-experimental screening protocol for the discovery of bimetallic catalysts

Computational simulations, particularly density functional theory (DFT), are vital in materials science for predicting and designing materials, including catalysts. DFT enables the prediction of reaction pathways on catalyst surfaces. However, accurately estimating crucial properties like reaction barriers is computationally expensive, often taking months even for a single metallic catalyst system. This raises questions about the efficiency of DFT-based catalyst discovery compared to purely experimental approaches. To optimize the combined computational-experimental approach, appropriate descriptors linking computation and experiment are needed. The d-band center theory, correlating the d-band center with gas adsorption energy, has been widely used. However, more comprehensive descriptors, incorporating d-band shapes and sp-band properties, which provide information on both values and shapes of states, can improve prediction accuracy. This study explores the use of the full electronic density of states (DOS) patterns as a descriptor, leveraging the principle that materials with similar electronic structures tend to exhibit similar properties. Previous studies have shown that alloys with similar DOS patterns to Pt or Pd exhibit similar catalytic properties. Therefore, bimetallic alloys with a DOS pattern similar to Pd should exhibit similar catalytic performance, forming the hypothesis for this high-throughput development of bimetallic catalysts. This study uses the full DOS pattern as a key descriptor in high-throughput computational-experimental screening protocols, focusing on Pd as a reference material for hydrogen peroxide (H2O2) synthesis.

The literature review highlights the importance of computational methods, especially DFT, in catalyst design and discovery. It discusses the limitations of relying solely on computationally expensive DFT calculations for comprehensive reaction mechanism elucidation, emphasizing the need for efficient descriptors. Existing descriptors, such as the d-band center theory, are reviewed, along with advancements towards more comprehensive descriptors that include not just the d-band center but also the shape of the d-band and the sp-band properties. Previous studies demonstrating the correlation between similar electronic structures (DOS patterns) and similar catalytic properties in alloys, such as Ir50Au50 and Pt for H2 dissociation, and Rh50Ag50 and Pd for hydrogen storage, are referenced to support the study's rationale. The use of Pd as the reference material is justified by its well-established role as a prototypical catalyst for H2O2 synthesis.

The study employed a three-step high-throughput screening protocol (Figure 1). First, 4350 crystal structures of bimetallic alloys were generated from 30 transition metals in periods IV, V, and VI, considering 435 binary systems with 1:1 composition and 10 ordered phases each. DFT calculations determined the formation energy (ΔE) for thermodynamic stability screening, filtering out structures with ΔE > 0.1 eV. Second, DOS patterns for the remaining 249 thermodynamically stable alloys were calculated and compared to Pd(111) using a defined similarity measure (ΔDOS2−1), which quantifies the difference between the DOS patterns near the Fermi energy. A Gaussian distribution function, g(E;σ) was introduced into the calculation to put more weights on the DOS near the Fermi energy. The standard deviation σ was set to 7 eV because most of the d-band centers for the bimetallic alloys are distributed from −3.5 eV to 0 eV relative to the Fermi energy. Third, synthetic feasibility and economic considerations were evaluated. Seventeen candidates with ΔDOS2−1 < 2.0 were initially selected, with some eliminated based on synthetic feasibility and material cost. Thirteen candidates underwent experimental synthesis using a butyllithium reduction method. Inductively coupled plasma (ICP) analysis confirmed alloy compositions and X-ray powder diffraction (XRD) confirmed crystal structures, closely matching DFT predictions (Figure 4). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was used to verify alloy homogeneity at the nanoscale. Eight bimetallic NPs were successfully synthesized and tested for H2O2 direct synthesis at mild conditions (20°C, 1 atm). Catalytic performance was measured using iodometric titration for H2O2 concentration and gas chromatography for H2 concentration. The catalytic properties (H2O2 productivity, H2 conversion, and H2O2 selectivity) were evaluated, and cost-normalized productivity (CNP) was calculated to assess economic viability. DFT calculations were conducted to explore the Langmuir-Hinshelwood mechanism and an electron-proton transfer mechanism for H2O2 production on Ni50Pt50(111) and Pd(111) surfaces, elucidating the reaction pathways (Figure 5). Projected crystal orbital Hamilton population (pCOHP) analysis and differential charge density calculations were performed to understand the binding characteristics between adsorbates and catalyst surfaces.

The study successfully synthesized and characterized eight bimetallic NPs (Au51Pd49, Ni61Pt39, Pd50Cu50, Pd52Ni48, Pt58Co42, Pt52Pd48, Rh56Ni44, Rh56Pt44) with crystal structures consistent with DFT predictions. Four alloys (Au51Pd49, Ni61Pt39, Pd52Ni48, and Pt52Pd48) exhibited catalytic properties comparable to Pd for H2O2 direct synthesis. The Ni61Pt39 alloy demonstrated superior performance, achieving a 9.5-fold enhancement in CNP compared to Pd. This enhancement is attributed to the high Ni content, which reduces the overall cost. The turnover frequency (TOF) for Ni61Pt39 was also superior to that of Pd. DFT calculations provided mechanistic insights, showing that the rate-determining step for H2O2 synthesis on Ni50Pt50(111) is the desorption of H2O2*, unlike Pd(111) where it is the first hydrogenation step. PCOHP analysis suggested that back-bonding between electron-rich Ni atoms in the alloy and the H2O2* adsorbate contributes to the enhanced stability of H2O2* on Ni-Pt surfaces compared to Pd(111). The selectivity for H2O2 production is comparable for both Pd (63.0%) and Ni50Pt50 (61.2%). The Ni-Pt bimetallic catalyst was identified as a cost-effective alternative for H2O2 direct synthesis due to its superior catalytic activity. Composition optimization through experiments demonstrated that Ni61Pt39 yielded the best catalytic performance regarding H2O2 productivity, H2 conversion, and H2O2 selectivity.

The successful identification of Ni61Pt39 as a superior catalyst validates the efficacy of the high-throughput screening protocol and the suitability of DOS pattern similarity as a descriptor for catalyst discovery. The protocol's efficiency lies in its ability to quickly narrow down a large number of candidates using computational methods before proceeding to experimental synthesis and testing, thus saving time and resources. The superior performance of Ni61Pt39, despite Ni and Pt not typically being considered for H2O2 synthesis due to their tendency to dissociate O2, underscores the power of this approach. The study's findings have implications for replacing or reducing the use of expensive platinum-group metals (PGMs) in catalysis, especially considering the increasing price of Pd. The protocol can be extended by using the Ni-Pt alloy as a new reference material to discover other catalysts, opening opportunities for catalyst development in other chemical reactions.

This research presents a successful high-throughput screening protocol for bimetallic catalyst discovery, leading to the development of a cost-effective Ni61Pt39 catalyst that outperforms Pd for H2O2 direct synthesis. The method’s success highlights the effectiveness of using DOS pattern similarity as a descriptor. Future work could involve exploring other bimetallic systems and further optimizing the Ni-Pt system for broader applications.

The study focused on a 50:50 composition for initial DFT screening, with composition optimization done experimentally. A more comprehensive DFT investigation across a wider range of compositions could further optimize catalyst performance. The study's focus on H2O2 synthesis might limit the generalizability of the DOS pattern similarity descriptor to other catalytic reactions. Additionally, the experimental synthesis employed might not be scalable for large-scale production.