How the correlated quantum chemical calculation changes the uncertainty of theoretically predicted rate coefficients and branching ratios Can Huang 1,2,3 , Zijun Zhou 2 , Bin Yang 2 , Feng Zhang 1 1 Hefei National Laboratory for Physical Sciences at the Microscale, P. R. China, 2 Center for Combustion Energy and Key Laboratory for Thermal Science and Power Engineering of MOE, P. R. China, 3 Chair of Technical Thermodynamics, Germany. Quantum chemical methods and rate theories (e.g. transition state theory and RRKM/master equation method) have been well acknowledged in developing kinetic models for combustion, atmospheric, interstellar, and catalytic synthesis systems. Corresponding uncertainty quantification (UQ) for computed rate coefficients is mandatory to guide the uncertainty propagation during kinetic modeling. The present work quantified correlation in quantum chemical calculations in several prototypical reaction systems based on systematic DFT and post- HF calculations as well as literature data and databases. The results show that a notable correlation exists in energies and imaginary frequencies. The correlations in H abstraction, H-addition and thermal decomposition reactions, regarding radical sites, molecular types and reaction types, were further quantified by the Pearson correlation coefficient. The correlation factors were then incorporated into the global uncertainty analysis for TST and RRKM/master equation calculations to unravel the effect of correlation in the uncertainty of rate coefficients and branching ratios. Including correlation among input parameters largely reduces the predicted uncertainty, with the largest reduction of ~30% for absolute rate and ~45% for branching ratio in TST calculations and ~33% for absolute rate and ~50% for branching ratio in RRKM/master equation calculations. Sensitivity analysis reveals that in TST calculation the reduced uncertainty is solely originated from the reduction of the sampling space due to correlated inputs. In RRKM/ME calculations the uncertainty is more complicated. The reduced uncertainty in rate coefficients arises only from reduced sampling space, but the reduced uncertainty in branching ratios is from both parameters canceling effect and correlation. Understanding the uncertainty propagation behavior from input parameters to the final rate coefficients or product branching is of great benefit to avoid treating those theoretical tools as a “black-box” without a proper sense of errors. References 1. Klippenstein, S. J., From theoretical reaction dynamics to chemical modeling of combustion. Proc. Combust. Inst. 2017, 36 (1), 77-111. 2. Tomlin, A. S., The role of sensitivity and uncertainty analysis in combustion modelling. Proc. Combust. Inst. 2013, 34 (1), 159-176. 3. Yang, B., Towards predictive combustion kinetic models: Progress in model analysis and informative experiments. Proc. Combust. Inst. 2021, 38 (1), 199-222. 4. Proppe, J.; Husch, T.; Simm, G. N.; Reiher, M., Uncertainty quantification for quantum chemical models of complex reaction networks. Faraday Discuss. 2017, 195, 497-520. 5. Miller, J. A.; Sivaramakrishnan, R.; Tao, Y.; Goldsmith, C. F.; Burke, M. P.; Jasper, A. W.; Hansen, N.; Labbe, N. J.; Glarborg, P.; Zádor, J., Combustion chemistry in the twenty-first century: Developing theory-informed chemical kinetics models. Prog. Energy Combust. Sci. 2021, 83, 100886.
P27
© The Author(s), 2022
Made with FlippingBook Learn more on our blog