High-throughput methods using composition and structure spread libraries
J. R. Kitchin and A. J. Gellman, “High-throughput methods using composition and structure spread libraries”, AIChE Journal 62(11), 3826-3835 (2016). https://doi.org/10.1002/aic.15294
This summary was written by a language model from the paper and has not been checked line by line. Rely on the published paper for details.
Question
Properties of multicomponent materials such as catalytic activity, corrosion resistance and stability depend on composition and on operating environment, but mapping them one discrete sample at a time is slow and costly, and the number of degrees of freedom makes the problem combinatorial. This Perspective asks how libraries with continuous gradients in composition, surface orientation or nanoparticle size, combined with spatially resolved characterization and computation, can speed up the mapping of composition/structure/property relationships for chemical engineering. It is explicitly not a comprehensive review of high-throughput methods; it focuses on the authors’ own work and closely related examples (Kitchin 2016).
Approach
The paper is organized around examples. The central library is the composition spread alloy film (CSAF), a thin film (typically under 1 µm) on a substrate whose composition varies continuously, so that a ternary film can contain the pure metals at the corners, the three binary spreads along the edges and the full ternary space inside. Gellman’s group built a rotating shadow mask deposition tool that co-deposits up to four metals with controlled gradients. The authors then review studies of corrosion, gas-phase catalysis, surface structure spread single crystals (S4C), nanoparticle size gradients and electrocatalysis, and close with an outlook on what high-throughput methods need (Kitchin 2016).
Key findings
For corrosion, an AlxFeyNi1-x-y CSAF oxidized in dry air at 700 K was mapped by EDX, Raman, XPS depth profiling and SEM. Four regions of oxidation behavior were identified: a protective surface Al2O3 scale, two regions passivated by a continuous subsurface Al2O3 scale under an Fe2O3 overlayer (smooth or rough), and an unpassivated region with deep oxidation. The boundary of that last region traces the composition-dependent critical Al concentration, a trajectory through composition space rather than a single number, obtained from a single library (Kitchin 2016).
For catalysis on CuxPd1-x CSAFs, XPS and low energy ion scattering spanning xCu = 0.05 to 0.95 and 400 to 1000 K showed Cu segregation at all compositions, maximal at 700 K, and Langmuir-McLean fits gave composition-dependent segregation enthalpies and entropies. Core level shift measurements revealed a dip in the Cu shift for 0.33 < xPd < 0.55 that DFT attributed to the ordered B2 phase. A 100-channel microreactor array (a 10 × 10 gasket of 700 × 900 µm2 holes over the CSAF, with mass spectrometric analysis) measured H2-D2 exchange from 300 to 600 K, and microkinetic fits gave adsorption and desorption barriers versus composition. UPS valence band energies, whose trends DFT showed follow the d-band center, were then correlated with the H2 adsorption barrier, one of the first such experimental correlations across continuous composition. The barrier fell roughly linearly with Pd content below xPd = 0.40, which DFT explained by hydrogen-induced Pd segregation, and became composition independent above it, possibly because of Pd hydride formation (Kitchin 2016).
A Cu S4C exposing orientations within 10° of (111) showed by XPS at 100 orientations that oxygen uptake after 30 L O2 at room temperature depends only on step density, not on step orientation or kink density. For nanoparticles, the paper reviews Hayden and co-workers’ Pt size gradients on TiO2 (0.2 to 2 nm equivalent thickness) showing that small particles oxidize CO more slowly and reduce oxygen poorly, and Au particles of 1 to 6 nm with peak CO electro-oxidation near 3 nm on titania but no activity on carbon. Electrocatalysis examples include PdAu films with maximum HER/HOR activity near 70 at% Au and PdBi films where an intermetallic phase matched Pd activity without forming a hydride (Kitchin 2016).
Significance
The outlook identifies three experimental legs (the library, spatially resolved characterization at least as careful as for discrete samples, and parallel measurement of the functional property) and argues that library equipment need not be prohibitively expensive. It names open needs: trustworthy tools for modeling, fitting and visualizing the resulting data, design of experiments for efficient sampling of composition space, simulation methods (cluster expansions, accurate force fields, multiscale models) that can handle continuous multicomponent spaces, and data archiving beyond the journal article, citing a 1.8 GB DFT dataset archived on Zenodo (Kitchin 2016).