Microkinetic Modeling
Microkinetic modeling describes a catalytic reaction as a set of elementary steps (adsorption, surface reaction, desorption), each with its own rate constant, usually of Arrhenius form with a pre-exponential factor and an energy barrier. Solving the resulting rate equations for a reactor gives predicted conversions or product flows that can be compared with experiment. Used in reverse, fitting the model to kinetic data measured over a range of temperatures and partial pressures yields estimates of the elementary pre-exponents and barriers, which can then be correlated with catalyst properties such as the d-band center. The approach requires a known mechanism that is simple enough for the parameters to be identifiable from the data.
Lateral interactions change the RDS and MARI of NH3 decomposition on Ru (2004)
Mhadeshwar, Kitchin, Barteau and Vlachos built a thermodynamically consistent, mean-field microkinetic model for NH3 decomposition on Ru, with activation energies from the UBI-QEP method and only pre-exponential factors fit to atmospheric fixed bed data, and added DFT-estimated N-N repulsion (about 35 kcal/mol/ML in the model). Without the interactions, N* was the most abundant intermediate and N2 desorption the rate determining step, as in earlier models; with them, H* became the MARI, N* stayed below about 0.7 ML, and NH2* decomposition became rate determining. A refit model without interactions matched the conversion data almost as well but with very different pre-exponentials, a reminder that kinetic parameter sets are not unique and that conversion data alone may not discriminate between mechanisms (Mhadeshwar et al. 2004). Summary: mhadeshwar2004.
Simulating TPD with a DFT-informed rate equation (2014)
Simulated TPD spectra for oxygen on Pt(111) were produced by numerically integrating a single elementary-step rate law (second-order associative desorption, pre-exponential factor 3 x 1012 1/(ML s)) with a coverage-dependent barrier whose shape came from DFT, and fitting only two Brønsted-Evans-Polanyi parameters to experiment (Miller et al. 2013). This illustrates using first-principles energetics to constrain the functional form of a kinetic parameter, so that sparse experimental data suffice to determine the remaining constants. Summary: miller2014.
Barriers for H2 adsorption and desorption across CuPd compositions (2015)
Gumuslu and co-workers fit a two-step mechanism for H2-D2 exchange (dissociative adsorption and recombinative desorption on two sites, neglecting isotope effects and coverage dependence) to HD production measured at 100 CuxPd1-x compositions, 14 temperatures and three feed conditions, minimizing the squared conversion error. Pre-exponents were nearly composition independent and consistent with transition state theory, so they were fixed and only the barriers refit. The uncertainty of a fitted barrier also served as a diagnostic: where the desorption barrier’s uncertainty exceeded its value (x > 0.8), the authors concluded the kinetics had become insensitive to desorption, marking a switch from desorption-limited to adsorption-limited exchange near x ≈ 0.8 (Gumuslu et al. 2015). Summary: gumuslu2015; see also High-throughput experimentation.
Identifying rate-controlling steps with automatic differentiation (2022)
Yang, Achar and Kitchin analyzed three microkinetic models with AD-computed degrees of rate control: a hypothetical two-step reaction with an analytical transient solution, a seven-step redox water-gas shift mechanism, and a 17-step propylene partial oxidation mechanism (Yang et al. 2022). For water-gas shift only OH* dissociation (0.88) and CO2* formation (0.12) controlled hydrogen production. For propylene oxidation at 500 K, adsorption of propylene in its type III configuration and desorption of PO2* promoted propylene oxide formation while O2* desorption inhibited it. Sampling the forward rate constants over an uncertainty of about 2.9 kJ/mol in barrier with DELSA showed that a step’s importance can be bimodal across the uncertain parameter space. Summary: yang2022b.
Periodically forced CSTR microkinetic model (2024)
A unimolecular adsorption, surface reaction and desorption microkinetic model in a continuous stirred tank reactor, with BE-dependent rate constants and a site-balance algebraic equation, served as the testbed for comparing sequential (SciPy solveivp, MATLAB) and simultaneous (Pyomo/IPOPT) solutions of the cyclic steady state under binding-energy oscillation (Tedesco et al. 2024). Both gave the same limit cycles. Summary: tedesco2024.
Microkinetic models under periodic forcing (2025)
Colombo Tedesco and co-workers used rate constants derived from a microkinetic model under an oscillating binding energy (0.1 eV amplitude, 1.4 eV offset, 1000 Hz square wave) for a three-step, two-adsorbate mechanism, and showed that cycle-averaged coverages and limit cycles can be obtained analytically rather than by numerical integration. The analytical result was essentially instantaneous, compared with about 1 s for a boundary value problem solve and about 40 s for forward initial-value integration (Colombo Tedesco et al. 2025). Summary: colombotedesco2025; see also Dynamic catalysis.
An ammonia synthesis volcano from ML-accelerated transition states (2025)
Wander and coworkers reproduced a microkinetic ammonia synthesis volcano on stepped Ru, Fe, Pd, Co, Ni and Rh using an OC20-pretrained EquiformerV2 for intermediate relaxations and N2 dissociation NEBs, with DFT single points for energies. The exhaustive configurational search found systematically lower energy intermediates and transition states than the original study, shifting the volcano and placing Ru more distinctly at the apex, consistent with experiment. The work took about 10 GPU hours, against an estimated 1.7 GPU years for the same NEBs with DFT (Wander et al. 2025). Summary: wander2025a.
Catalyst kinetics feeding a system cost model (2025)
In the Systems-to-Atoms study, CatMAP microkinetic models built on literature DFT energetics gave formic acid decomposition rates on Pd(111) and Pd(211) (Yuan et al. 2025). Arrhenius fits gave apparent barriers of about 89.5 kJ/mol for dehydrogenation and 156.6 kJ/mol for dehydration, with dehydrogenation about six orders of magnitude faster on Pd(111). Rates showed negligible pressure dependence from 1 to 100 bar. Facet rates were combined into a nanoparticle TOF that sized the reactor in a technoeconomic model. Pd and Pt responded similarly to temperature, but pressure raised the modeled Cu rate by one to two orders of magnitude, which made Cu cost competitive with Pd at 100 bar. Summary: yuan2025.
Ensembles of multiscale models compared with experiment (2026)
This perspective argues that linking microkinetic models to experiments is an ill-conditioned inverse problem with a many-to-one mapping, since kinetic, reactor, measurement and atomistic models together can have thousands of parameters (Medford et al. 2026). It highlights an ensemble of 2000 automatically generated light-off profiles for exhaust gas oxidation over Pt(111), built from the BEEF-vdW functional ensemble and RMG, and semiquantitative agreement between transient DRIFTS and microkinetic simulations on Pd/γ-Al2O3, as models for routinely generating and testing large ensembles of mechanisms. Summary: medford2026.