TL;DR: Researchers Clement Soulié, Judit Zádor and Leonid Sheps introduced KiMecO, an uncertainty-informed, master-equation-based optimizer that fits a hydrocarbon-oxidation submechanism directly to multi-species, multi-condition concentration–time data.

What was reported on October 7, 2026

On October 7, 2026, the Combustion Research Facility posted a report titled “Accelerating discovery by more efficient coupling of theory and experiment,” according to the Combustion Research Facility.

The post described a paper titled “KiMecO: A Kinetic Mechanism Optimizer for Experimentally Constrained Master Equation Models,” according to the Combustion Research Facility.

The paper was published in The Journal of Physical Chemistry A, according to the Combustion Research Facility.

The authors listed for the advance were CRF researchers Clement Soulié, Judit Zádor, and Leonid Sheps, according to the Combustion Research Facility.

The announcement provided a DOI link for more information: https://doi.org/10.1021/acs.jpca.6c04372, according to the Combustion Research Facility.

The central claim in the announcement was that the team had taken an important step toward making links between experiment and fundamental explanation more direct, according to the Combustion Research Facility.

Why coordinated modeling, experiment and theory matter

Coordinating effort across modeling, experiment, and theory is a powerful paradigm for discovery of fundamental interactions within complex reaction systems, according to the Combustion Research Facility.

Linking the experiment reliably to the most fundamental explanations can be challenging, however, according to the Combustion Research Facility.

That challenge is the starting point for the reported work, according to the Combustion Research Facility.

The reported response to that challenge was to constrain a more fundamental description with controlled experimental records, rather than adjusting a less fundamental description alone, according to the Combustion Research Facility.

The announcement frames that shift as a way to accelerate discovery by more efficient coupling of theory and experiment, according to the Combustion Research Facility.

What KiMecO is

This paper introduces KiMecO, an uncertainty informed, master equation based kinetic mechanism optimizer, according to the Combustion Research Facility.

KiMecO constrains the rate coefficients of a “submechanism,” according to the Combustion Research Facility.

A “submechanism” is a subset of the set of reactions that would govern a full hydrocarbon oxidation system, according to the Combustion Research Facility.

In other words, KiMecO does not attempt to refit every reaction that would appear in a full hydrocarbon oxidation system in the same operation, according to the Combustion Research Facility.

It focuses optimization on a selected subset while retaining the context of complex reaction systems, according to the Combustion Research Facility.

The optimizer is described as uncertainty informed, according to the Combustion Research Facility.

It is also described as master equation based, according to the Combustion Research Facility.

The combination of uncertainty information with a master-equation foundation distinguishes KiMecO from optimization that works only at the level of rate-coefficient parameterization, according to the Combustion Research Facility.

What data constrain the models

KiMecO constrains the rate coefficients against concentration–time profiles, according to the Combustion Research Facility.

The concentration–time profiles cover multiple species, according to the Combustion Research Facility.

The profiles span multiple conditions, according to the Combustion Research Facility.

The profiles come from controlled experiments, according to the Combustion Research Facility.

That multi-species scope means the optimizer is tested against more than one tracked product or intermediate evolving in time, according to the Combustion Research Facility.

That multi-condition scope means the optimizer is tested against more than one experimental setting, according to the Combustion Research Facility.

The controlled-experiment setting means the concentration–time records are obtained under defined laboratory conditions rather than uncontrolled field observations, according to the Combustion Research Facility.

How machine learning and parallel computation are used

KiMecO leverages machine learning and parallel computation to explore the sensitive master equation parameter space, according to the Combustion Research Facility.

The master equation parameter space is described as sensitive, according to the Combustion Research Facility.

Machine learning is therefore presented as a tool for navigating that sensitivity efficiently, according to the Combustion Research Facility.

Parallel computation is presented as a tool for exploring that space across concurrent calculations, according to the Combustion Research Facility.

The use of both approaches supports examination of a space that would be difficult to map by serial or uninformed search, according to the Combustion Research Facility.

The announcement identifies that exploration capability as part of what makes the coupling of theory and experiment more efficient, according to the Combustion Research Facility.

Why the master-equation level matters

The master equation parameter space is a physically more fundamental description than rate coefficient parameterization, according to the Combustion Research Facility.

Constraints at this level enforce physically consistent correlations among rate coefficients, according to the Combustion Research Facility.

That correlation point is central to the reported advance, according to the Combustion Research Facility.

Adjusting individual rate coefficients in isolation can miss relationships that must hold for physical consistency, according to the Combustion Research Facility’s description of constraints at the more fundamental level.

By imposing constraints at the master-equation level, KiMecO maintains those physically consistent correlations among rate coefficients, according to the Combustion Research Facility.

The result is an optimized submechanism whose component rates remain mutually consistent at the underlying physical level, according to the Combustion Research Facility.

The ethyl + O2 demonstration

Use of the KiMecO framework on the reaction sequence initiated by the ethyl + O2 reaction is reported as the demonstration case, according to the Combustion Research Facility.

The ethyl + O2 sequence initiates a reaction sequence that the framework was used to optimize, according to the Combustion Research Facility.

That demonstration connects the method to hydrocarbon oxidation chemistry, because the submechanism concept is defined relative to a full hydrocarbon oxidation system, according to the Combustion Research Facility.

The choice of an ethyl + O2 initiated sequence provides a concrete test in which multiple downstream species and time-dependent behavior can be compared with calculation, according to the Combustion Research Facility’s description of multi-species concentration–time fitting.

The announcement does not present the ethyl + O2 test as a full hydrocarbon oxidation system, according to the Combustion Research Facility.

It presents it as a reaction sequence used to exercise the KiMecO framework, according to the Combustion Research Facility.

Ensembles versus single models

Use of the KiMecO framework on the ethyl + O2 initiated sequence shows that ensembles of master equation models are generally more robust than single models, according to the Combustion Research Facility.

That robustness comparison concerns master-equation models produced within the optimization workflow, according to the Combustion Research Facility.

A single-model result can reflect one satisfactory parameter choice in a sensitive space, according to the Combustion Research Facility’s characterization of the master-equation parameter space as sensitive.

An ensemble retains multiple satisfactory descriptions and is reported as generally more robust, according to the Combustion Research Facility.

The announcement presents that ensemble robustness as a finding from the ethyl + O2 application, according to the Combustion Research Facility.

Uncertainty analysis of optimized rates

The same ethyl + O2 application shows that ensembles enable uncertainty analysis of the optimized rate coefficients, according to the Combustion Research Facility.

The uncertainty analysis applies to the optimized rate coefficients produced for the submechanism, according to the Combustion Research Facility.

Because KiMecO is uncertainty informed, the resulting ensemble supports examination of spread or confidence in those coefficients, according to the Combustion Research Facility.

That capability links the fitting step to downstream use, because users of a kinetic mechanism need to know both nominal rates and their associated uncertainty, according to the Combustion Research Facility’s description of an uncertainty-informed optimizer.

The announcement identifies that uncertainty-analysis capability as an outcome of using ensembles rather than a single model, according to the Combustion Research Facility.

How this relates to other experiment-plus-theory work

Independent work in combustion kinetics has also combined experiments with master-equation simulations for alkyl-radical plus oxygen systems, as documented in the University of Helsinki Research Portal record titled “Kinetics of the reaction between propan-2-yl radical and molecular oxygen: Experiments and master-equation simulations.”

That record lists Arkke J. Eskola, Jukka S. Ruotsalainen, Timo T. Pekkanen, Akshat Shirish Zalte, Matti P. Rissanen and William H. Green as contributors, and identifies Proceedings of the Combustion Institute, volume 42, article 106445, with DOI https://doi.org/10.1016/j.proci.2026.106445, according to the University of Helsinki Research Portal.

The Helsinki-portal example concerns propan-2-yl radical and molecular oxygen, while the KiMecO demonstration concerns the sequence initiated by the ethyl + O2 reaction, according to the Combustion Research Facility and the University of Helsinki Research Portal.

Both examples pair controlled kinetic observations with a master-equation treatment of pressure- and energy-dependent behavior, according to the University of Helsinki Research Portal record’s title and the Combustion Research Facility description of a physically more fundamental master-equation description.

The KiMecO contribution adds an optimizer that fits a submechanism against multi-species concentration–time profiles across multiple conditions, according to the Combustion Research Facility.

How this relates to prior uncertainty and correlation work

An earlier application of Bayesian inference and maximum entropy methods estimated the joint probability density for Arrhenius rate parameters for the H + O2 chain-branching reaction H + O2 -> OH + O, using nominal values and error bars from shock-tube experiments and OH concentration profiles, as described in the record titled “Inference of Reaction Rate Parameters based on Summary Statistics from Experiments.”

That work generated data consistent with published summary statistics using Approximate Bayesian Computation and Markov chain Monte Carlo, built surrogates from Padé and polynomial approximants, used Gauss–Hermite quadrature for moment computation, and pooled posterior densities before propagating the pooled density through stoichiometric hydrogen–air auto-ignition computations, according to “Inference of Reaction Rate Parameters based on Summary Statistics from Experiments.”

It also reported nearly Gaussian conditional parameter densities despite strong nonlinearity, accounted for nuisance parameters in the form of Arrhenius parameters of other rate coefficients, and illustrated the need to account for correlation among Arrhenius parameters of one reaction and across different reactions, according to “Inference of Reaction Rate Parameters based on Summary Statistics from Experiments.”

The KiMecO announcement addresses a related correlation theme by enforcing physically consistent correlations among rate coefficients through constraints at the master-equation level, according to the Combustion Research Facility.

The earlier H + O2 study worked from published nominal values and error bars and OH profiles, while KiMecO works against concentration–time profiles of multiple species across multiple conditions, according to “Inference of Reaction Rate Parameters based on Summary Statistics from Experiments” and the Combustion Research Facility.

What the announcement establishes and what it does not

The announcement establishes the name, authors, journal, optimizer type, data type, computational tools, physical level of constraint, correlation effect, test reaction, and ensemble finding, according to the Combustion Research Facility.

The name is KiMecO, according to the Combustion Research Facility.

The authors are Clement Soulié, Judit Zádor, and Leonid Sheps, according to the Combustion Research Facility.

The journal is The Journal of Physical Chemistry A, according to the Combustion Research Facility.

The optimizer type is uncertainty informed and master equation based, according to the Combustion Research Facility.

The data type is concentration–time profiles of multiple species across multiple conditions in controlled experiments, according to the Combustion Research Facility.

The computational tools are machine learning and parallel computation, according to the Combustion Research Facility.

The physical level of constraint is the sensitive master equation parameter space, described as physically more fundamental than rate coefficient parameterization, according to the Combustion Research Facility.

The correlation effect is enforcement of physically consistent correlations among rate coefficients, according to the Combustion Research Facility.

The test reaction sequence is initiated by ethyl + O2, according to the Combustion Research Facility.

The ensemble finding is greater robustness and enabled uncertainty analysis for optimized rate coefficients, according to the Combustion Research Facility.

The announcement does not provide numerical values for the optimized coefficients, uncertainty ranges, experimental temperatures, pressures, species names beyond the ethyl + O2 initiation, or software availability details beyond the DOI link, according to the Combustion Research Facility.

Why it matters for U.S. combustion science

For readers following U.S. combustion science, the practical significance is a tighter connection between what is measured in time-resolved experiments and what is represented in fundamental kinetic descriptions, according to the Combustion Research Facility.

A submechanism that respects physically consistent correlations is more likely to behave plausibly when conditions change, because its rates share an underlying physical basis, according to the Combustion Research Facility’s account of master-equation-level constraints.

An ensemble that supports uncertainty analysis gives model users a way to assess confidence in optimized rates, rather than receiving only a single best-fit set, according to the Combustion Research Facility.

A workflow that uses machine learning and parallel computation to search a sensitive space offers a more efficient route to those results than manual tuning of individual coefficients, according to the Combustion Research Facility.

Together, those points explain why the Combustion Research Facility presented KiMecO as an important step toward more direct links across modeling, experiment, and theory, according to the Combustion Research Facility.

Sources referenced