A reaction mechanism is not useful until it has been validated. And a validated mechanism is not useful until it has been assembled into a model that connects the chemistry to something physically observable — in this case, the burn rate and species structure of an HMX monopropellant flame.
This piece picks up where the quantum mechanics investigation left off. Having mapped the initial decomposition pathways of RDX and HMX at the molecular level, the next challenge was to build a complete mechanism, validate it against thermolysis experiments, solve the transport property problem, and assemble everything into a multiphase combustion model. Each of these steps required its own set of decisions.
Building the comprehensive mechanism
A kinetic mechanism for HMX decomposition needs to account for two distinct regimes:
The liquid phase: at the temperatures and pressures relevant to combustion, HMX first melts and then decomposes in the liquid phase before evaporating or gasifying. The initial bond-breaking events and early intermediate reactions happen here. The QM calculations described earlier provided the rate constants for the initial pathways. Expanding the mechanism required identifying and characterizing all the subsequent reactions of the intermediate species — the products of the first reactions become the reactants of the next generation, and so on, until the observed final decomposition products are formed.
The gas phase: the volatile products of liquid-phase decomposition (CH₂O, N₂O, NO₂, NO, HCN, CO) enter the gas phase and continue reacting. The gas-phase mechanism was assembled and updated based on a comparative ab initio study, with particular attention to the early ring-opening and hydrogen abstraction reactions and to the species evolving from the liquid phase.
The final comprehensive mechanism covers the full decomposition network — from the initial N–NO₂ homolysis and ring-opening events through the intermediate species to the final products — in both phases.
Experiments as the validation anchor
Mechanism validation required carefully designed experiments that probe different aspects of the decomposition behavior.
Thermogravimetric analysis (TGA) measures the mass loss of a small sample as it is heated at a controlled rate. As HMX decomposes, the sample loses mass; the rate of mass loss as a function of temperature provides a clear experimental signature that any valid mechanism must reproduce. TGA experiments were conducted at multiple heating rates — varying the heating rate changes the temperature at which the decomposition occurs and provides information about the activation energy of the rate-controlling steps.
Confined rapid thermolysis (CRT) is a more direct probe of the decomposition products. A small sample is rapidly heated to a fixed temperature and the evolved gaseous products are collected and analyzed using FTIR spectroscopy and gas chromatography. This gives the actual species distribution at different temperatures and decomposition extents.
The key experimental findings: CH₂O and N₂O were the dominant decomposition products at all conditions. H₂O, NO₂, NO, HCN, CO, and CO₂ were also detected and quantified. The relative concentrations and their evolution with temperature and heating rate are the signatures against which the computational model was validated.
A homogeneous liquid-phase reactor model was developed to simulate the TGA experiments (time-varying temperature, tracking mass loss) and the CRT experiments (fixed temperature, tracking species evolution). The computational mass loss profiles and species evolution profiles were in reasonable agreement with experiment across all conditions tested — which is a more demanding test than matching a single condition, since the mechanism must be right for the right physical reasons to generalize.
The transport property problem
Combustion modeling requires not just reaction rates but also transport properties: diffusion coefficients, viscosities, and thermal conductivities for all the species in the mechanism. These are needed to describe how molecules move through the gas phase and mix.
The standard approach is to use Lennard-Jones (LJ) collision parameters — two numbers per species ( and ) that characterize the size and interaction strength of the molecule — and then apply kinetic theory correlations. The problem is that LJ parameters are not available for most energetic material intermediates, and the standard "combining rules" (which estimate parameters for unlike-molecule collisions from the pure-component values) introduce systematic errors that vary with the choice of bath gas.
A new approach was developed for this work. Instead of selecting a particular bath gas and combining rule, intermolecular potentials (Buckingham potentials) were computed from quantum mechanics for all relevant organic energetic species. These were then used to obtain LJ collision parameters through a least-squares optimization that used all available bath gases and all common combining rules simultaneously — making the final parameters consistent across the full set of data rather than optimized for any one condition.
A sensitivity analysis on the RDX combustion model revealed that the temperature gradient controlling burn rate is actually more sensitive to transport parameters than to reaction rate parameters — a finding that underscores why getting the transport properties right is not a secondary concern.
The multiphase combustion model
With the liquid-phase mechanism, gas-phase mechanism, thermodynamic data, and transport properties in hand, the final assembly was the multiphase combustion model for HMX monopropellant.
The model describes the burning HMX strand as a one-dimensional structure: the condensed-phase (liquid) region near the surface, the surface itself where evaporation and early gas-phase reactions occur, and the gas-phase flame above. Mass and energy conservation are solved across all three regions simultaneously. The model outputs:
- Burn rate as a function of pressure — the most directly measurable combustion property of a propellant
- Temperature profiles in both the liquid and gas phases
- Species concentration profiles — which species are present where, and in what amounts
- Melt layer thickness — how deep into the solid the liquid layer extends at a given burn rate
The predicted burn rates and melt layer thicknesses were in excellent agreement with experimental measurements across a range of pressures. The model also correctly captured the transition in dominant gas-phase chemistry as pressure changes — a behavior that is invisible to global empirical models but emerges naturally from the detailed mechanism.
The importance of including the liquid-phase decomposition in the combustion model — rather than treating the surface as a simple evaporation boundary — was explicitly demonstrated. Models that omit the liquid-phase chemistry underpredict the burn rate because they miss the exothermic decomposition that occurs in the melt layer and modifies the heat feedback to the solid.
What this kind of model enables
A validated multiphase combustion model with first-principles chemistry is a tool for prediction beyond the conditions it was validated against. New propellant formulations, off-nominal conditions, pressure excursions, and ignition transients can all be explored computationally before committing to expensive experimental programs.
More broadly, this project established that the quantum-mechanics-to-combustion-model pipeline is viable for energetic materials: begin at the molecular level, use QM to avoid relying on poorly constrained empirical fits, validate each component against experiment, and build upward. The approach is more expensive upfront than fitting global models to data, but it produces mechanistic insight that transfers to new conditions and materials.
The full work is described in:
- A Comprehensive Mechanism for Liquid-phase Decomposition of HMX: Thermolysis Experiments and Detailed Kinetic Modeling, Combustion and Flame, 2020
- Intermolecular potential parameters for transport property modeling of energetic organic molecules, Combustion and Flame, 2019
- Modeling of HMX Monopropellant Combustion with Detailed Condensed-Phase Kinetics, AIAA Propulsion and Energy Forum, 2019