Kinetic Modelling of CH4/CO2 Dry Reforming in a Dielectric Barrier Discharge for Hydrogen Production: An Open, Sensitivity- Guided Validation Study

Technical Summary: Kinetic Modelling of CH4/CO2 Dry Reforming in a Dielectric Barrier Discharge

Problem Statement
The conversion of methane (CH4) and carbon dioxide (CO2) via plasma-assisted dry reforming (DRM) in dielectric barrier discharges (DBDs) offers a non-thermal pathway to produce hydrogen and syngas. However, kinetic modelling tools for this process often rely on specialized, non-modular platforms and frequently utilize illustrative, order-of-magnitude rate constants rather than rigorously validated data. Furthermore, there is a lack of open frameworks that explicitly couple 0D homogeneous chemistry with 1D axial-dispersion transport, making it difficult to rapidly prototype operating parameters or validate models against experimental benchmarks with quantitative precision.

Methodology
The authors developed an open, modular Python framework (0D/1D) to simulate CH4/CO2 conversion in a DBD.

• Kinetic Mechanism: A reduced mechanism comprising 23 neutral species and 38 reactions was employed. This includes electron-impact dissociation, bimolecular neutral-neutral reactions, three-body associations, and unimolecular decompositions.
• Electron Kinetics: Unlike previous iterations using analytical fits, electron-impact rate coefficients (E1–E5) were derived from a genuine two-term electron Boltzmann equation solve (using the BOLOS solver) fed with real electron-neutral scattering cross-sections from the LXCat/Morgan database.
• Plasma Forcing: The DBD was modeled as a periodic pulse train representing filamentary microdischarges. A time-averaging approximation was applied to replace instantaneous pulse-resolved rates with cycle-averaged values, validated to reproduce pulse-resolved results with <1% error while significantly reducing computational cost.
• Sensitivity Analysis: A global Morris sensitivity analysis was performed on the 33 neutral-neutral rate constants. This identified six specific reactions as the primary drivers of CO2 conversion, H2/CO ratio, and energy efficiency.
• Validation Strategy: The six identified critical reactions (R2, R11–R14, and R33) were replaced with values from evaluated combustion and atmospheric kinetics literature. The remaining 27 rate constants were retained as illustrative order-of-magnitude parameters.
• Model Structure: The framework couples a 0D Lagrangian model (tracking a gas parcel) with a 1D Eulerian axial-dispersion plug-flow reactor model to simulate spatial profiles.

Key Results

• Mechanism Validation: Replacing illustrative guesses with literature values corrected errors ranging from 6-fold to over 3000-fold. Specifically, the reaction CO + O (+M) → CO2 (+M) (R33), which is spin-forbidden and difficult to measure, was anchored to a 200 K literature value (the best available, though the model operates at 400 K).
• Conversion Trends: With the validated mechanism, CO2 conversion respects the ~25% ceiling observed across 19 atmospheric-pressure plasma sources in the literature for Specific Energy Input (SEI) up to ~30,000 J/L. CH4 conversion rises more steeply, approaching near-complete conversion at high SEI.
• Selectivity: The H2/CO ratio stabilizes between 1.1 and 2.1, reflecting a CH4-limited regime where CH4 dissociates faster than CO2, consistent with experimental observations.
• Benchmark Comparison:
  • At a matched SEI of 5.7 eV/molecule, the model initially over-predicted CH4 conversion by ~5.8-fold and CO2 by ~2.7-fold compared to the Ozkan et al. [55] experimental benchmark.
  • Applying a "power-corrected" SEI (accounting for total power transfer rather than just electron-impact energy cost) reduced these gaps to ~3.3-fold (CH4) and ~1.6-fold (CO2).
  • Crucially, at lower, experimentally typical SEI values (938–2,500 J/L), the model showed close agreement with experiment, under-predicting CH4 conversion slightly and matching CO2 closely. The divergence increases only at high SEI, suggesting the model is robust in the low-to-moderate SEI regime typical of laboratory DBDs.
• Uncertainty Sources: The sensitivity analysis confirmed that even after validation, R33 remains the dominant source of uncertainty for CO2 conversion and H2/CO ratio. The remaining over-prediction of CH4 conversion is attributed to unvalidated CH4-radical-pool chemistry (R1, R3–R10) and the assumption of channel-resolved branching ratios for CH4 dissociation.

Significance and Claims
The paper presents a significant step toward a quantitatively validated, open-source tool for plasma DRM. Its primary contributions are:

1. Open Framework: Providing a reproducible Python codebase that couples 0D chemistry with 1D transport, accessible for rapid prototyping and teaching.
2. Methodological Upgrade: Moving from arbitrary analytical fits to genuine Boltzmann-solved rate coefficients based on real cross-sections.
3. Targeted Validation: Demonstrating that a sensitivity-guided replacement of just six key rate constants with literature values significantly improves model fidelity, removing a major source of error and bringing predictions closer to experimental reality.
4. Realistic Limitations: The authors modestly claim that while the model is not yet fully quantitative across all SEI ranges, it provides a "qualitative-to-semi-quantitative" basis for the low-to-moderate SEI regime. They explicitly identify the remaining roadmap for full validation: obtaining lower-temperature data for R33, validating CH4-radical chemistry, and calibrating plasma forcing parameters against specific reactor electrical measurements.

The work does not claim to have solved the DRM problem quantitatively but rather provides a transparent, open computational basis with a specific, narrowed roadmap for achieving a fully validated model.