This tutorial explores the combustion of ethanol (C2H6O) in oxygen:
C2H6O + 3O2 → 2CO2 + 3H2O
While stoichiometrically simple, actual combustion involves complex reaction networks. Marinov's study identified over 300 elementary reactions in this system.
Modern reactive MD methods now enable efficient analysis of such complex systems, reducing computation time from months to ~30 minutes.
We construct a simulation box containing:
filetype xyz output model.xyz tolerance 4.0 structure ethanol.xyz number 100 inside box 0 0 0 67.18 67.18 67.18 end structure structure oxygen.xyz number 300 inside box 0 0 0 67.18 67.18 67.18 end structure
packmol < packmol.inp
The final periodic system XYZ file includes boundary conditions:
1500 Lattice="67.18 0.0 0.0 0.0 67.18 0.0 0.0 0.0 67.18" Origin="0.026168 0.001867 0.024439" Properties=id:I:1:species:S:1:pos:R:3 1 C 49.2771926179 25.3604183225 27.8483112501 2 H 49.4570955095 25.3162389282 28.928943456 ...
Initial system configuration
When system prepared, we will use GPUMD to perform a 200 ps molecular dynamics simulation. According to the adiabatic combustion temperature of ethanol, which is 2082 ℃, ~2350 K, we will set the simulation temperature as 2350 K. And the ensemble is NVT, using langevin thermostat.
The GPUMD running parameter file run.in as below:
potential nep.txt velocity 2350 time_step 0.2 ensemble nvt_lan 2350 2350 100 dump_thermo 10 dump_exyz 500 run 1000000
We will use the newest Neuroevolution Potential (NEP) parameter set NEP-89 for simulation. Before simulation, the working directory contains files below:
run.in nep.txt model.xyz
Use command gpumd directly to invoke an automatic GPUMD task in current directory.
The simulation cost 10~20 minutes on GTX 4090D machine, after simulation terminated, the system structure looks like below:
System configuration after simulation.
After simulation completion, GPUMD outputs a standard trajectory file (dump.xyz
) in the current
directory. This file is compatible with ReaxTools (File format reference).
Analysis steps:
Configuration tip: For this case, we use -r 1
to apply stricter bond cutoff
criteria for molecular identification. For advanced customization, see Input
options reference.
Loading input file with analysis options
Key results from the simulation are presented below:
Program execution log output
Species count versus time (complete view)
Species count versus time (focused view)
Atom type count versus time
Atom type legend:
OH fragment reaction network (radical/anion focused)
OH-related reaction pathways table
The network visualization displays the main reaction components. Hover over elements to view associated reactions and molecules.
Key molecule reaction fluxes
Key molecule reaction partners table
Key molecules (those with high reaction frequency) are identified through their flux values. The visualizations and tables help analyze: