Machine learning-driven optimization of plasma-coupled photocatalytic CO₂ reduction system for syngas production

Abstract

Dielectric barrier discharge (DBD) plasma-coupled photocatalysis offers a promising route for CO₂-to-syngas conversion, but rational optimization remains challenging because operating variables are strongly coupled. Here, we develop a small-sample machine-learning framework to predict and optimize a plasma-coupled photocatalytic CO₂ conversion system using a Cu-Pd/TiO₂ photocatalyst. Using 120 experimental runs, five operating parameters, including discharge power, catalyst dosage, gas flow rate, relative humidity, and light intensity, were evaluated against CO₂ conversion, CO yield, and H₂ yield. Among the evaluated models, a shared-weight Gradient Boosting Regressor (GBR)-Kernel Ridge Regression (KRR) hybrid model achieved the great predictive performance, with test-set R² values of 0.947, 0.950, and 0.954 for CO₂ conversion, CO yield, and H₂ yield, respectively. Permutation importance and SHapley Additive exPlanations (SHAP) analyses identified relative humidity, light intensity, and discharge power as the most influential variables and revealed distinct model-predicted trends across the three outputs. The trained model was further used for constrained optimization under target H₂/CO ratios of 1 and 2, followed by experimental validation of the selected operating conditions. Overall, this work establishes a data-driven strategy for interpreting nonlinear plasma-photocatalytic CO₂ conversion and provides practical guidance for syngas-ratio regulation under experimentally relevant conditions.

Keywords

Plasma-photocatalysis / CO₂ conversion / syngas / machine learning

References

[1] Friedlingstein P,Jones MW.Global carbon budget 2024.Earth Syst Sci Data2025;17:965-1039

[2] Grim RG,Braunecker WA.The chemistry of CO₂ conversion: a review.Chem Rev2026;126:5028-82

[3] Garg S,Chen JG.Tandem reactors and reactions for CO₂ conversion.Nat Chem Eng2024;1:139-48

[4] Yu S,Wang S.CO₂-to-methanol electroconversion on a molecular cobalt catalyst facilitated by acidic cations.Nat Catal2024;7:1000-9

[5] Chen H,Li J.Epitaxially grown silicon-based single-atom catalyst for visible-light-driven syngas production.Nat Commun2023;14:1719 PMCID:PMC10050177

[6] Zhou W,Zhao W.Photocatalytic CO₂ reduction to syngas using metallosalen covalent organic frameworks.Nat Commun2023;14:6971 PMCID:PMC10620383

[7] Xie Z,Li L.Well-defined diatomic catalysis for photosynthesis of C₂H₄ from CO₂.Nat Commun2024;15:2422 PMCID:PMC10948895

[8] Weber JL,de Jong KP.Recent advances in bifunctional synthesis gas conversion to chemicals and fuels with a comparison to monofunctional processes.Catal Sci Technol2024;14:4799-842 PMCID:PMC11347923

[9] Hou X,Wei K.Syngas production from CO₂ and H₂O via solid-oxide electrolyzer cells: fundamentals, materials, degradation, operating conditions, and applications.Chem Rev2024;124:5119-66

[10] Dossow M,Umeki K,Spliethoff H.Electrification of gasification-based biomass-to-X processes - a critical review and in-depth assessment.Energy Environ Sci2024;17:925-73

[11] Zhang Y,Chen J.Recent progress in and future perspectives on high-density single-atom electrocatalysts.Electrochem Energy Rev2025;8:257

[12] Xu C,Ran Y,Lei Y.Two-in-one integrated CO₂/N₂ conversion and related systems: potential, status, and future.Electrochem Energy Rev2025;8:264

[13] Fang S,Bharti J.Photocatalytic CO₂ reduction.Nat Rev Methods Primers2023;3:243

[14] He H,Zhang L.Synergistic modulation of charge dynamics and mass transfer optimization via heterogeneous interface engineering in photothermal catalytic CO₂ reduction within continuous flow systems.Nano Energy2025;142:111290

[15] Choi SH,Dong WJ.Recent progress of photothermal catalysts for carbon dioxide conversion.Energy Mater2025;5:500062

[16] Ray D,Vinodkumar T.g-C₃N₄ promoted DBD plasma assisted dry reforming of methane.Energy2019;183:630-8

[17] Gonzalez-Casamachin DA,Rangarajan S.Mechanistic insights for plasma-catalytic CO₂ reduction over TiO₂ in a dielectric barrier discharge reactor.ACS Eng Au2026;6:334-44

[18] Yang Y.CO₂ conversion using non-thermal plasmas: the path towards industrialisation.Curr Opin Green Sustain Chem2025;51:100994

[19] Huang Q,Shen Y.Boosting CO₂ conversion by synergy of lead-free perovskite Cs₂SnCl₆ and plasma with H₂O.J Phys Chem Lett2023;14:8922-9

[20] Feng X,Fu C.Enhanced CO₂ conversion in dielectric barrier discharge plasma coupled with a heterojunction photocatalyst.Chem Commun2024;60:8900-3

[21] Bi R,Wang D.Plasma-coupled photocatalytic CO₂/H₂O conversion to tunable syngas over Cu_xPd_y/TiO₂ with electronic modulation.Appl Catal B Environ Energy2027;402:127470

[22] Wanten B,De Meyer R.Plasma-based CO₂ conversion: how to correctly analyze the performance?.J Energy Chem2023;86:180-96

[23] Shen Y,Luo W,Wang Z.Machine learning for CO₂ conversion driven by dielectric barrier discharge plasma and Cs₂TeCl₆ photocatalysts.Green Chem2023;25:7605-11

[24] Cai Y,Chen C.Machine learning prediction of small molecule passivators and their impacts on the passivation and photocatalytic performance of organic-inorganic hybrid perovskite interfaces.Energy Mater2025;5:500043

[25] Han N.AI-driven material discovery for energy, catalysis and sustainability.Natl Sci Rev2025;12:nwaf110 PMCID:PMC11983685

[26] Luo W,Fu C,Huang Q.Exploring the CO₂ conversion activated by the dielectric barrier discharge plasma assisted with photocatalyst via machine learning.J Environ Chem Eng2024;12:114428

[27] Taniike T,Nakanowatari S,Takahashi K.Automatic feature engineering for catalyst design using small data without prior knowledge of target catalysis.Commun Chem2024;7:11 PMCID:PMC10786848

[28] Li J,Rebrov E.Machine learning-based prediction and optimization of plasma-catalytic dry reforming of methane in a dielectric barrier discharge reactor.Chem Eng J2025;507:159897

[29] Hu Y,Spatschek R.Practical feature filter strategy to machine learning for small datasets in chemistry.Sci Rep2024;14:20449 PMCID:PMC11379859

[30] Friedman JH.Greedy function approximation: a gradient boosting machine.Ann Statist2001;29:1189-232

[31] Rupp M,Müller KR.Fast and accurate modeling of molecular atomization energies with machine learning.Phys Rev Lett2012;108:058301

[32] van der Laan MJ,Hubbard AE.Super learner.Stat Appl Genet Mol Biol2007;6:25

[33] Chicco D,Jurman G.The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation.PeerJ Comput Sci2021;7:e623 PMCID:PMC8279135

[34] Williamson BD,Simon NR.A general framework for inference on algorithm-agnostic variable importance.J Am Stat Assoc2023;118:1645-58 PMCID:PMC10652709

[35] Fumagalli F,Hüllermeier E.Incremental permutation feature importance (iPFI): towards online explanations on data streams.Mach Learn2023;112:4863-903

[36] Altmann A,Sander O.Permutation importance: a corrected feature importance measure.Bioinformatics2010;26:1340-7

[37] Bilodeau B,Koh PW.Impossibility theorems for feature attribution.Proc Natl Acad Sci U S A2024;121:e2304406120 PMCID:PMC10786278

[38] Štrumbelj E.Explaining prediction models and individual predictions with feature contributions.Knowl Inf Syst2014;41:647-65

[39] Lundberg SM,Chen H.From local explanations to global understanding with explainable AI for trees.Nat Mach Intell2020;2:56-67 PMCID:PMC7326367

[40] Gao B,Li C,Fang Y.Defect-engineered Ru-Co/TiO_2-x catalysts for highly efficient plasma-catalytic ammonia synthesis under ambient conditions.Green Chem2026;28:8235-48

[41] Huang Z,Ma M.Facile synthesis of TiO₂ supported Pd nanoparticles for efficient photocatalytic CO₂ reduction to CH₄ with H₂O.Sustain Mater Technol2025;43:e01247

[42] Liu S,Luo C.Green hydrogen and acetaldehyde production via photo-induced thermal dehydrogenation of bioethanol over a highly dispersed CuS/TiO₂ catalyst.Int J Hydrogen Energy2024;86:1414-23

[43] Liang J,Li Y,Yan X.Single-atom Cu anchored phosphorus-doped graphitic carbon nitride for selective photocatalytic CO₂ reduction.Environ Res2025;285:122324

[44] Huang Q,Liang Z.CO₂ conversion synergistically driven by radiofrequency inductively-coupled plasma and lead-free halide perovskite photocatalyst.J Phys Chem C2023;127:11550-8

[45] Han Y,Guo Y.Plasma-Driven Efficient conversion of CO₂ and H₂O into pure syngas with controllable wide H₂/CO ratios over metal-organic frameworks featuring in situ evolved ligand defects.Angew Chem Int Ed2024;63:e202406007

[46] Haghshenas Y,Gunawan D.Predicting the rates of photocatalytic hydrogen evolution over cocatalyst-deposited TiO₂ using machine learning with active photon flux as a unifying feature.EES Catal2024;2:612-23

[47] Wu X,Zhang X,Liu C.Dielectric surface engineering via TiO₂ coating for enhanced plasma-driven efficient CO₂ decomposition.Chem Eng J2025;505:159743

[48] Trampel CP.Control of electromagnetic edge effects in electrically-small rectangular plasma reactors.J Vac Sci Technol A2012;30:051305

[49] Vertongen R.How important is reactor design for CO₂ conversion in warm plasmas?.J CO2 Util2023;72:102510

[50] Deo JV.Study of electric field distribution on plasma and plasma catalysis reactor for different electrode configurations and pellet sizes.PPT2024;11:54-60

[51] Lisi N,Chierchia R,Stendardo S.A water cooled, high power, dielectric barrier discharge reactor for CO₂ plasma dissociation and valorization studies.Sci Rep2023;13:7394 PMCID:PMC10164120

[52] Liang C,Liu Z,Lam YM.Thermodynamic and microkinetic insights into potential-dependent competition between CO₂ reduction and HER on Mo₂CT_x MXene.Chem Eng J2025;520:165927

[53] Chawdhury P,Conway M.Enhancing the reaction of CO₂ and H₂O using catalysts within a nonthermal plasma.ACS Catal2025;15:7053-65

[54] Klein M.Non-thermal plasma upgrading of humidified CO₂ into syngas in a dielectric barrier discharge reactor: tuning H₂/CO ratios via specific energy input and gas flow rate.Energy Adv2026;5:354-64

[55] Snoeckx R,Reniers F.The quest for value-added products from carbon dioxide and water in a dielectric barrier discharge: a chemical kinetics study.ChemSusChem2017;10:409-24

[56] Behroozi AH.Photocatalytic CO₂ reduction: photocatalysts, membrane reactors, and hybrid processes.Chem Catal2023;3:100550

[57] Alamdari A,Aghaeinejad-Meybodi A.Thermodynamic assessment of tri-reforming of methane with optimization of operating conditions to achieve suitable syngas for methanol production.Sci Rep2026;16:14257 PMCID:PMC13139425

[58] Bonde A,Ahrens A.Integrating hydroformylations with methanol-to-syngas reforming.Chem2025;11:102396

[59] Yoshizawa T,Sato T,Honma T.A data-driven generative strategy to avoid reward hacking in multi-objective molecular design.Nat Commun2025;16:2409 PMCID:PMC11897179

[60] Janet JP,Yang T,Kulik HJ.A quantitative uncertainty metric controls error in neural network-driven chemical discovery.Chem Sci2019;10:7913-22 PMCID:PMC6764470

[61] Haddadnia M,Strieth-Kalthoff F.BoTier: multi-objective Bayesian optimization with tiered objective structures.Digit Discov2025;4:1417-22

[62] Frömbgen T,Dölz J,Kirchner B.Uncertainty quantification for in silico chemistry.Chem Rev2026;126:4189-236 PMCID:PMC13067284