Data-Centric Heterogeneous Catalysis: Identifying Rules and Materials Genes of Alkane Selective Oxidation.

Artificial intelligence (AI) can accelerate catalyst design by identifying key physicochemical descriptive parameters correlated with the underlying processes triggering, favoring, or hindering the performance. In analogy to genes in biology, these parameters might be called "materials genes" of heterogeneous catalysis. However, widely used AI methods require big data, and only the smallest part of the available data meets the quality requirement for data-efficient AI. Here, we use rigorous experimental procedures, designed to consistently take into account the kinetics of the catalyst active states formation, to measure 55 physicochemical parameters as well as the reactivity of 12 catalysts toward ethane, propane, and n-butane oxidation reactions. These materials are based on vanadium or manganese redox-active elements and present diverse phase compositions, crystallinities, and catalytic behaviors. By applying the sure-independence-screening-and-sparsifying-operator symbolic-regression approach to the consistent data set, we identify nonlinear property-function relationships depending on several key parameters and reflecting the intricate interplay of processes that govern the formation of olefins and oxygenates: local transport, site isolation, surface redox activity, adsorption, and the material dynamical restructuring under reaction conditions. These processes are captured by parameters derived from N2 adsorption, X-ray photoelectron spectroscopy (XPS), and near-ambient-pressure in situ XPS. The data-centric approach indicates the most relevant characterization techniques to be used for catalyst design and provides "rules" on how the catalyst properties may be tuned in order to achieve the desired performance.

Materials genes of heterogeneous catalysis from clean experiments and artificial intelligence.

ABSTRACT: The performance in heterogeneous catalysis is an example of a complex materials function, governed by an intricate interplay of several processes (e.g., the different surface chemical reactions, and the dynamic restructuring of the catalyst material at reaction conditions). Modeling the full catalytic progression via first-principles statistical mechanics is impractical, if not impossible. Instead, we show here how a tailored artificial-intelligence approach can be applied, even to a small number of materials, to model catalysis and determine the key descriptive parameters ("materials genes") reflecting the processes that trigger, facilitate, or hinder catalyst performance. We start from a consistent experimental set of "clean data," containing nine vanadium-based oxidation catalysts. These materials were synthesized, fully characterized, and tested according to standardized protocols. By applying the symbolic-regression SISSO approach, we identify correlations between the few most relevant materials properties and their reactivity. This approach highlights the underlying physicochemical processes, and accelerates catalyst design. IMPACT STATEMENT: Artificial intelligence (AI) accepts that there are relationships or correlations that cannot be expressed in terms of a closed mathematical form or an easy-to-do numerical simulation. For the function of materials, for example, catalysis, AI may well capture the behavior better than the theory of the past. However, currently the flexibility of AI comes together with a lack of interpretability, and AI can only predict aspects that were included in the training. The approach proposed and demonstrated in this IMPACT article is interpretable. It combines detailed experimental data (called "clean data") and symbolic regression for the identification of the key descriptive parameters (called "materials genes") that are correlated with the materials function. The approach demonstrated here for the catalytic oxidation of propane will accelerate the discovery of improved or novel materials while also enhancing physical understanding. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1557/s43577-021-00165-6.

Bond Energies of Adsorbed Intermediates to Metal Surfaces: Correlation with Hydrogen-Ligand and Hydrogen-Surface Bond Energies and Electronegativities.

Understanding what controls the strength of bonding of adsorbed intermediates to transition-metal surfaces is of central importance in many technologies, especially catalysis and electrocatalysis. Our recently measured bond enthalpies of -OH, -OCH3 , -O(O)CH and -CH3 to Pt(111) and Ni(111) surfaces are fit well (standard deviation of 7.2 kJ mol-1 ) by a predictive equation involving only known parameters (gas-phase ligand-hydrogen bond enthalpies, bond enthalpies of adsorbed H atoms to that surface, electronegativities of the elements, and group electronegativities of the ligands). This equation is based upon Pauling's equation, with improvements introduced by Matcha, derived here following manipulations of Matcha's equation similar to (but going beyond) those introduced by Schock and Marks to explain ligand-metal bond enthalpy trends in organometallic complexes.

Water adsorption on the Fe3O4(111) surface: dissociation and network formation.

We monitored adsorption of water on a well-defined Fe3O4(111) film surface at different temperatures as a function of coverage using infrared reflection-absorption spectroscopy, temperature programmed desorption, and single crystal adsorption calorimetry. Additionally, density functional theory was employed using a Fe3O4(111)-(2 × 2) slab model to generate 15 energy minimum structures for various coverages. Corresponding vibrational properties of the adsorbed water species were also computed. The results show that water molecules readily dissociate on regular surface Fetet1-O ion pairs to form "monomers", i.e., terminal Fe-OH and surface OH groups. Further water molecules adsorb on the hydroxyl covered surface non-dissociatively and form "dimers" and larger oligomers, which ultimately assemble into an ordered (2 × 2) hydrogen-bonded network structure with increasing coverage prior to the formation of a solid water film.

Measurements of deuterium quadrupole coupling in propiolic acid and fluorobenzenes using pulsed-beam Fourier transform microwave spectrometers.

The pure rotational spectra of deuterated propiolic acids (HCCCOOD and DCCCOOH), 1-fluorobenzene (4-d1), and 1,2-difluorobenzene (4-d1) in their ground states have been measured using two Fourier transform microwave (FTMW) spectrometers at the University of Arizona. For 1-fluorobenzene (4-d1), nine hyperfine lines of three different ΔJ = 0 and 1 transitions were measured to check the synthesis method and resolution. For 1,2-difluorobenzene (4-d1), we obtained 44 hyperfine transitions from 1 to 12 GHz, including 14 different ΔJ = 0, 1 transitions. Deuterium quadrupole coupling constants along the three principal inertia axes were well determined. For deuterated propiolic acids, 37 hyperfine lines of Pro-OD and 59 hyperfine lines of Pro-CD, covering 11 and 12 different ΔJ = - 1, 0, 1 transitions, respectively, were obtained from 5 to 16 GHz. Deuterium quadrupole coupling constants along the three inertia axes were well resolved for Pro-OD. For Pro-CD, only eQq(aa) was determined due to the near coincidence of the CD bond and the least principal inertia axis. Some measurements were made using a newer FTMW spectrometer employing multiple free induction decays as well as background subtraction. For 1-fluorobenzene (4-d1) and 1,2-difluorobenzene (4-d1), a very large-cavity (1.2 m mirror dia.) spectrometer yielded very high resolution (2 kHz) spectra.

Gas phase measurements of mono-fluoro-benzoic acids and the dimer of 3-fluoro-benzoic acid.

The microwave spectrum of the mono-fluoro-benzoic acids, 2-fluoro-, 3-fluoro-, and 4-fluoro-benzoic acid have been measured in the frequency range of 4-14 GHz using a pulsed beam Fourier transform microwave spectrometer. Measured rotational transition lines were assigned and fit using a rigid rotor Hamiltonian. Assignments were made for 3 conformers of 2-fluorobenzoic acid, 2 conformers of 3-fluorobenzoic acid, and 1 conformer of 4-fluorobenzoic acid. Additionally, the gas phase homodimer of 3-fluorobenzoic acid was detected, and the spectra showed evidence of proton tunneling. Experimental rotational constants are A(0(+)) = 1151.8(5), B(0(+)) = 100.3(5), C(0(+)) = 87.64(3) MHz and A(0(-)) = 1152.2(5), B(0(-)) = 100.7(5), C(0(-)) = 88.85(3) MHz for the two ground vibrational states split by the proton tunneling motion. The tunneling splitting (ΔE) is approximately 560 MHz. This homodimer appears to be the largest carboxylic acid dimer observed with F-T microwave spectroscopy.

Calculations and measurements of the deuterium tunneling frequency in the propiolic acid-formic acid dimer and description of a newly constructed Fourier transform microwave spectrometer.

The concerted proton tunneling frequency for the propiolic acid-formic acid dimer was calculated using a relaxed ab initio double-well potential in the imaginary-frequency mode of the saddle point, and new measurements were made for the deuterated propiolic acid-formic acid (ProOD-FAOD) isotopologue. It is important to have consistent calculated tunneling frequency values between normal and deuterated isotopologues since parameters can be readily adjusted to get good agreement with one isotopologue. High-resolution rotational spectra of deuterated (ProOD-FAOD) dimer were measured using a newly constructed Fourier Transform microwave spectrometer. The new spectrometer has mirror size: 30 cm in diameter with a radius of curvature of 59 cm and is equipped with multiple-FID data collection (5-10 FID's for each gas pulse). For the deuterated (ProOD-FAOD) isotopologue, 45 rotational lines (a type: 34; b type: 11) were measured in the lowest tunneling states range between 6.5 GHz and 15.5 GHz. With the new high-resolution measurements of the tunneling doublets (b-dipole transitions), the double potential well responsible for the deuterium tunneling was depicted much more precisely. The two tunneling states are separated by 3.48 MHz. The rotational constants obtained in this work are quite helpful for further structure analysis as well.

Microwave structure for the propiolic acid-formic acid complex.

New microwave spectra were measured to obtain rotational constants and centrifugal distortion constants for the DCCCOOH···HOOCH and HCCCOOD···DOOCH isotopologues. Rotational transitions were measured in the frequency range of 4.9-15.4 GHz, providing accurate rotational constants, which, combined with previous rotational constants, allowed an improved structural fit for the propiolic acid-formic acid complex. The new structural fit yields reasonably accurate orientations for both the propiolic and formic acid monomers in the complex and more accurate structural parameters describing the hydrogen bonding. The structure is planar, with a positive inertial defect of Δ = 1.33 amu Å(2). The experimental structure exhibits a greater asymmetry for the two hydrogen bond lengths than was obtained from the ab initio mp2 calculations. The best-fit hydrogen bond lengths have an r(O1-H1···O4) of 1.64 Å and an r(O3-H2···O2) of 1.87 Å. The average of the two hydrogen bond lengths is r(av)(exp) = 1.76 Å, in good agreement with r(av)(theory) = 1.72 Å. The center of mass separation of the monomers is R(CM) = 3.864 Å. Other structural parameters from the least-squares fit using the experimental rotational constants are compared with theoretical values. The spectra were obtained using two different pulsed beam Fourier transform microwave spectrometers.