[Paper Review] ChemXDyn: Dynamics-informed species and reaction detection methodology from atomistic simulations

Raj Maddipati, Dhruthi Boddapati+3 more|arXiv (Cornell University)|Jan 13, 2026

Machine Learning in Materials Science0 citations

TL;DR

ChemXDyn introduces a dynamics-aware framework that uses time-resolved interatomic distance signatures to reliably identify bonds, species, and reactions from reactive MD trajectories, improving kinetic measurements over threshold-based methods.

ABSTRACT

Accurate identification of chemical species and reaction pathways from molecular dynamics (MD) trajectories is a prerequisite for deriving predictive chemical-kinetic models and for mechanistic discovery in reactive systems. However, state-of-the-art trajectory analysis methods infer bonding from instantaneous distance thresholds, which can misclassify transient, nonreactive encounters as bonds and thereby introduce spurious intermediates, distorted reaction networks, and biased rate estimates. Here, we introduce ChemXDyn, a dynamics-aware computational methodology that leverages time-resolved interatomic distance signatures as a core principle to robustly identify chemically consistent bonded interactions and, consequently, extract meaningful reaction pathways. In particular, ChemXDyn propagates molecular connectivity through time while enforcing atomic valence and coordination constraints to distinguish genuine bond-breaking and bond-forming events from transient, nonreactive encounters. We evaluate ChemXDyn on ReaxFF MD simulations of hydrogen and ammonia oxidation and on neural-network potential MD simulations of methane oxidation, and benchmark its performance against widely used trajectory analysis methods. Across these cases, ChemXDyn suppresses unphysical species prevalent in static analyses, recovers experimentally consistent reaction pathways, and improves the fidelity of rate constant estimation. In ammonia oxidation, ChemXDyn removes unphysical intermediates and resolves key NOx- and N2O-forming and -consuming routes. In methane oxidation, it reconstructs the canonical progression from CH4 to CO2. By linking atomistic dynamics to chemically consistent reaction identification, ChemXDyn provides a transferable foundation for MD-derived reaction networks and kinetics, with potential utility spanning combustion, catalysis, plasma chemistry, and electrochemical environments.

Motivation & Objective

• Identify chemically consistent bonded interactions in MD trajectories by leveraging time-resolved interatomic distances.
• Distinguish genuine bond-breaking/formation events from transient encounters using temporal averages and valence constraints.
• Construct chemically valid species and extract reaction networks with rate constants from MD data.
• Benchmark performance against existing trajectory analyzers across multiple reactive systems to demonstrate improved fidelity.

Proposed method

• Step A: analyze time-windowed interatomic distances to compute forward and backward window-averaged distances (or BOs) for each atom-pair within clustered proximity.
• Step B: assign bond multiplicities by comparing time-averaged distances to equilibrium bond lengths and enforce coordination-number constraints to avoid weak/non-bonded interactions.
• Step C: build a connectivity graph from validated bonds and apply a depth-first search to identify molecular species at each timestep, encoding species canonically.
• Step D: track species across timesteps to detect reactions, map to canonical representations, and compute rate constants using k = Nreac / ∫0^T [P][Q] dt for bimolecular cases (extendable to other orders).
• Step A–D are designed to suppress spurious bonds, recover chemically meaningful intermediates, and enable transferable MD-derived kinetic networks.

Experimental results

Research questions

• RQ1 Can time-resolved interatomic distance signatures, combined with valence constraints, reliably distinguish true bond events from transient encounters in reactive MD?
• RQ2 Do dynamics-informed bonds yield more accurate species inventories, reaction pathways, and rate constants than threshold-based methods across diverse chemistries?
• RQ3 How does ChemXDyn perform relative to ChemTrYzer and ReacNetGenerator in H2/O2, NH3/N2/O2, and CH4/O2 systems?
• RQ4 Can the approach reconstruct canonical oxidation cascades (e.g., CH4 to CO2) and remove unphysical intermediates observed in static analyses?

Key findings

• ChemXDyn reduces unphysical species and over-counting of reactions compared with threshold-based tools in H2/O2 and NH3/N2/O2 systems.
• In H2/O2, ChemXDyn yields rate constants that align more closely with experimental data and reference kinetic models than ChemTrYzer-based pipelines.
• For NH3/N2/O2, ChemXDyn identifies fewer species and reactions than ChemTrYzer but captures key NHx–NOx chemistry that CTY misses, offering a more consistent mechanism set.
• In NH3/N2/O2 benchmarks, ChemXDyn detects essential reactions (e.g., NH2+HO2→H2NO+OH) and aligns with established ammonia oxidation pathways.
• In methane oxidation using an MLIP, ChemXDyn demonstrates robustness across force-field paradigms, reconstructing canonical oxidation sequences (CH4 → CH3 → CH2 → CH → CHO/CH2O → CO → CO2).
• Overall, ChemXDyn provides an accurate and transferable framework for MD-derived reaction networks and kinetics, with improved fidelity over existing tools.