RECENT ADVANCES (& CONTINUING CHALLENGES) IN COMBUSTION CHEMISTRY WILLIAM H. GREEN Co-Authors: Shamel S. Merchant1, Aaron G. Vandeputte1, Connie W. Gao1, Nick M. Vandewiele1,2, Nathan W. Yee1, Marko R. Djokic2, Kevin M. Van Geem2, & Guy B. Marin2 1) Department of Chemical Engineering, MIT 2) Laboratory for Chemical Technology, UGent, Ghent, Belgium $$$: DOE, AFOSR, FWO, BAEF, Flanders Methusalem, US Navy 1 Combustion is Critically Important • Provides about 80% of our energy With current technology, Developed countries burn 100 GJ/y per person. 1010 people * 1011 J/y = a lot of combustion! Energy/person (GJ/y) – Transportation, heating, electricity production… – …Will still be main energy source in 2040. – Crucial to Economy! GDP scales with energy use. GDP/person ($) 2 Combustion is Biggest Source of Greenhouse Gases We need to keep [CO2] < 550 ppm to have reasonable chance of avoiding catastrophic climate change. Need to drastically reduce slope of this graph very soon! 3 Epidemiology is clear: Soot Kills Mortality Rate 6-Cities Study, USA Dockery et al. N Engl J Med 1993 5 years less Life expectancy North of river Particulate Level in Air Strong correlation between Deaths and Particulates, seen repeatedly in many different locations & situations. Yuyu Chen et al. PNAS 2013;110:12936-12941 Huai River policy: coal burners north of river, no heat south of river. Life span much shorter on north side of river. Health impacts significantly slow economic growth. 4 What is needed? • Big Increases in fuel-to-work efficiency! – Reduces CO2 emissions and fuel cost – Less fuel burnt: reduces other emissions – Major approach: premixed low-T combustion • Avoid fuel-rich pyrolysis forming soot • Low T: Less heat losses, less NOx formation • But sensitive to ignition delay & flame extinction • Renewable (i.e. non-fossil) fuels – How to make them cheaply, in huge volumes… – … and predictions of their performance • Ways to reduce Soot (Particulate) emissions – Based on understanding of soot formation/oxidation 5 Where are we with soot? • Existing soot formation models include irreversible reactions… …something is fundamentally wrong. • See e.g. Hai Wang, Proc. Combust. Inst. (2011) • Existing soot oxidation models are extremely simplistic. • More work is needed! • Soot burn-out is usually incomplete: why? Good progress on early steps of polycyclic formation And graphene-sheet edge chemistry, e.g. P23 See talks by Klippenstein & Ross, PES calculations by Mebel 6 C5H5+C5H5 naphthalene (C10H8) …how? Many models say C5H5+C5H5 C10H8 + H + H But this is too slow Instead: C5H5 + C5H5 = C10H10 C10H10 + R C10H9 +RH C10H9 C10H8 + H 7 See poster by Marko Djokic for relevant expts Where are we with Flame Extinction? • Crucial in turbulent flames – Flame-holding, stability – Acoustic noise, pressure oscillations – Maybe important in soot break-through? • Some understanding of strain-induced flame extinction [e.g. S.H.Won et al. Combust. Flame (2012) ]… …but so far we haven’t demonstrated we can predict flame extinction for new fuels 8 For the rest of this talk, I’ll focus on Methodology for Predicting (nonsooting, unstrained) Combustion Chemistry of new fuels and on Low-T Ignition 9 Combustion Chemistry Mechanisms are Huge 4500 PRF (Curran et al.) 4000 800 3000 n-heptane (Curran et al.) 2500 600 2000 400 hydrogen 1500 1000 methane (GRIMech3.0) propane (Marinov) 200 n-butane (ENSIC Nancy) 500 0 0 0 1 2 3 4 5 6 7 8 9 Carbon Number Never enough experimental data to determine all the k(T,P): must work in predictive mode, based primarily on quantum chemistry. 10 Number of Species iso-octane (Curran et al.) 3500 Number of Reactions Use computer to build the model! 1000 Commercial software can solve large kinetic simulations… …if one can supply the reaction mechanism. Simulation predictions Diff. Eq. solver Very long list of reactions with rate parameters Interpreter (CHEMKIN, Cantera, KIVA, GTPower) Simulation equations dY/dt = … 11 How we construct chemistry models Sensitivity Analysis Chemistry knowledge Unambiguous documentation of assumptions about how molecules react Simulation predictions High-accuracy quantum calculations on sensitive parameters Diff. Eq. solver Very long list of reactions with rate parameters Interpreter (CHEMKIN, Cantera, KIVA, GTPower) Simulation equations dY/dt = … 12 RMG method: computer builds the kinetic model, based on first-principles. represent species unambiguously determine reactions that species undergo estimate rates from quantum chem determine which species belong in model 13 RMG software has several advanced features, all automatically & consistently applied pressure-dependent kinetics estimation solvation thermochemistry , some kinetics Sulfur chemistry (and Nitrogen too) automatic quantum chemistry for cyclics 14 Chemical Kinetic Modeling Challenges • Identify all important reactions & species – But not unimportant species & reactions: how to distinguish? • Compute all reaction rate coefficients (and properties, e.g. thermochemistry) to sufficient accuracy. – We use Functional Group extrapolations & Quantum Chemistry • Large models pose numerical and computer problems – Very challenging for humans to handle, interpret, debug… …SO WE TRY TO AUTOMATE EVERYTHING We build on prior efforts by large research community, e.g. Thermochemical Kinetics (1974) Comprehensive Chemical Kinetics 35 (1997) Advances in Chemical Engineering 32 (2007) Cleaner Combustion: Developing Detailed Models (2013) 15 RMG algorithm: Faster pathways explored further, growing the model Before: “Current Model” inside. RMG decides whether or not to add species to this model. Final model typically ~500 species, 8000 rxns After: Open-Source RMG software. Download from rmg.sourceforge.net 16 Rate-Based Algorithm is Sensitive to Errors in Rate & Thermo Estimates • Particularly Important to get the Thermo Right: In Combustion typically many species are in partial equilibrium with each other 17 Quantum Enthalpy Predictions Improve a Lot with Empirical Bond-Additivity Corrections (BAC) Hcorrected = DFT (B3LYP) CBSQB3 CCSD(T)-F12 /TZ CCSD(T)-F12 /QZ Error (Expt(ATcT) – Quantum) 2 kcal/tick mark Hquantum + correction for each C-C bond, each C-O bond, etc. 18 BAC (mostly) fix enthalpies, but leave discrepancies in computed barrier heights Compute slightly different Barrier depending on which direction you compute the reaction. Significant at Room T. In this case the inconsistency is ~0.6 kcal/mole: = 35% error at 1000 K, factor of 2.7 at 300 K. 19 Kinetic Model Predictions Rely on Quantum Chemistry for Rates: These are not Perfect! • Functional Group approximation – Compute a few examples of each reaction type with quantum, then use same barrier, A factor for analogous reactions. • Most of our calculations at CBS-QB3 level – Geometries, Vibrational Frequencies from DFT – Single point energies at stationary points at higher level – Extrapolation to Basis Set Limit • Recent calculations use F12 methods – Explicit dependence on distance between every pair of electrons – Much faster basis set convergence • Most calculations rely on several common approximations – – – – Rigid-Rotor Harmonic-Oscillator approximation Conventional Transition State Theory (dividing surface at saddle point) Simple corrections for internal rotors and tunneling Modified Strong Collision approx. for k(T,P) Are computed thermo, rates accurate enough?? 20 Are computed thermo, rates accurate enough?? • Conventional Quantum Chemistry methods have errors: a few kcal/mole in energies a few cal/mole-K in entropies and heat capacities perhaps a factor of 2 due to internal rotor approximations about a factor of 2 due to TST approximations • Several different errors, each about factor of 2 uncertainty • Lucky if a computed rate is within a factor of 2 of the truth • Can we live with a factor of 2 uncertainty in each of 104 rate coefficients? • Fortunately, most sensitivities d(ln(observable))/d(ln(k)) are 0.5 or less – many uncertainties are uncorrelated so they might “average out”. Need to test if this really works out OK! 21 Testing Accuracy of Model Predictions vs. Experiment: Extensive Data available on Pyrolysis, Combustion, Oxidation of Butanols flow reactors RCM With collaborations from other institutes e.g. Univ. Ghent, NIST. Pyrolysis (shock tube) Flame Speeds MBMS Shock tube Flame Speed Rare Situationswhere detailed data available at many different conditions! 22 We used RMG to build a mechanism for butanol pyrolysis and combustion. Four isomers, very different octane numbers. n-butanol iso-butanol Octane number = 86 Octane number = 98 More reactive sec-butanol tert-butanol Octane number = 100 Less reactive RMG considered about 30,000 possible species, selected as important: • 372 chemical species • 8,723 reactions Shamel S. Merchant, E.F. Zanoelo, R.L. Speth, M.R. Harper, K.M. Van Geem and William H. Green, Combustion & Flame (2013) Sensitive k’s computed with highest-level quantum chemistry we could manage; others from group additivity 23 RMG model quantitatively predicts formation of alkenes and 1-ring aromatics from iso-butanol (some via rather complicated reaction sequences) Data from K. Van Geem, Ghent pyrolysis of iso-butanol ~1000 K, 2 atm, 2 seconds. Merchant et al. (2013) 24 Species profiles in butanol flame confirm predictive capabilities for small molecules, high T • Synchrotron measures dozens of species in n-butanol flame, all predicted accurately Dozens of additional species traces, variety of flames: all show comparably good agreement. For isobutanol we worked in predictive mode, with similar level of agreement with expt. 25 Can also predict chemistry + flow quantities, such as zero-strain flame speeds Data from Veloo & Egolfopoulos (343 K) , and W. Liu et al. (353 K), both in Proc Combust Inst (2011). Model Prediction 26 Model quantitatively predicts high-T ignition delays for all butanol isomers & conditions Data measured by Stranic et al., Combust. Flame, 2012, 159 (2), 516-527. 27 And it is not just small molecules like the butanols. For example, the computer (RMG or Genesys) can build models for JP-10 (exo-tricyclodecane, C10H16) pyrolysis and combustion. Radical chain initiation 4EBO Radical chain JP-10 2CC 28 + Predicted major pathways for steam-cracking of exo-tricyclodecane (C10H16, “JP-10”) 29 Steam-cracking of exo-tricyclodecane experiment vs. model predictions T~1000 K, P~2 bar C2H4 CH4 H2 C3H6 cycloC5H6 cycloC5H8 Similar level of agreement for many other species See Vandewiele et al. Energy & Fuels (2015). For model & expts with JP-10 + O2, see Gao et al. Combust. Flame (2015) 30 At this point we are feeling very good: Computer-constructed model based on quantum chemistry can predict many observables quantitatively for several high T C,H,O gas phase systems! 31 We don’t know everything: model less accurate below ~900 K, and completely misses [O2] sensitivity of low-T ignition delay of iso-butanol! Const. [Fuel] In Air Model predicts [fuel] dependence reasonably well Model: No [O2] dependence Expts: τ ~ [O2]-1.5 Data measured by B. Weber and C.J. Sung 32 Possible causes of this Discrepancy • Low T: Energy errors more important −𝐸𝑎 −𝐸𝑡𝑟𝑢𝑒 ±𝛿𝐸 −𝐸𝑡𝑟𝑢𝑒 ∓𝛿𝐸 𝑅𝑇 𝑒 =𝑒 =𝑒 𝑒 • Internal Rotors: Intramolecular H-bonding causes large coupling between rotors 𝑅𝑇 𝑅𝑇 𝑅𝑇 – See e.g. Sharma et al. J.Phys.Chem. A (2010) – As Don Truhlar showed, conformations can be non-intuitive • Omissions or inaccuracies in the reaction mechanism – Computer-built models are not infinite, can omit reactions – Reminder: Computed rates are not perfect! – See talk by Samah Mohamed later this morning • New peroxy reaction types (not known when reaction mechanism was generated) 33 Missing some peroxy chemistry? • Still discovering new peroxy reactions, e.g. – Welz et al., J. Phys. Chem. Lett. (2013) – Jalan et al., J. Am. Chem. Soc. (2013) – Judit Zador’s talk on Monday • How can we discover new (unexpected) reactions? • Can we make a computer do it automatically? 34 Are we missing other important reactions? RMG can list all the low-energy species with same number of atoms as reactant. Many potential products unreachable by any known reaction. Example: Reactant Computer Says: 155 Possible Low-energy Product Channels 35 We Automated Search for New Reactions (Freezing String, followed by Berny TS search) Program automatically found new saddle points connecting reactant to 7 of the 155 possible low-energy products 7 completely new reactions! We don’t know how many the computer missed… …a lot of work still to be done on automatic discovery of new chemistry! 36 Let’s be Optimistic: Suppose our model already includes all the important reactions, just has the wrong rate coefficients for some of them. Which of the 8,723 rate coefficients in the model should we carefully check? Which are really important?? 37 Suppose our model already includes all the important reactions, just have the wrong rate coefficients for some of them. Which of the 8,723 rate coefficients in the model should we carefully check? Which are really important?? A quick primer on what is Important in Low-T ignition chemistry 38 Ignition Delay (log plot) Typical Ignition Delay Curves show 3 parts: High T (>900 K) and Low T (<700 K) are nearArrhenius, plus something in between 10 bar, f=1 in air, adiabatic 39 1000/T Multiple Stages of Ignition, each with different dominant chemistry. “1st-stage Ignition” Hot “Second-Stage” ignition Propane Methanol Exponential Growth in Concentrations 40 The Low-T “QOOH” Amplifier: 1 OH in, 3abg OH out g S.S. Merchant et al., Combust. Flame (accepted) • Reactive Intermediate Concentrations rapidly rise ~5 orders of magnitude: chemical amplifier • If abg=1, l ~ sqrt(2kdecompkROO=QOOH) 41 Methanol’s Amplifier: H2O2 (1 HO2 in, 3 HO2 out) This is why HO2 + fuel is important: it is often a key chain-branching pathway 42 Exponential-Growth Stage (“1A”) Linear kinetics, an eigenvalue > 0 Stage ends when HO2 + HO2 becomes significant Propane: • OH amplified by QOOH cycle Methanol: • HO2 amplified by H2O2 cycle Stage 1A Stage 1A 43 QOOH cycle continues to amplify despite HO2+HO2 during Stage 1B Propane during stage 1B: • HO2 in QSS due to self reaction • OH gain from QOOH cycle Stage 1B ends when QOOH cycle gain drops to 1 (mostly due to T increase) Stage 1B Stage 1A Stage 1A 44 Can write down analytical formulas for Stage 1 ignition delays • Dotted lines are from the analytical formulas. • Depends on 10 rate coefficients (a lot less than the 8,723 in the butanols model!) • Merchant et al., Combust. Flame (accepted) 45 Stage 2: HO2+Fuel H2O2 causes chain branching, tempered by HO2 + HO2 Propane: • QOOH chemistry continues, but OH gain < 1; coupled with H2O2 cycle Methanol: • HO2 amplified by H2O2 cycle • Product (H2CO) more reactive Stage 2 Stage 1B Stage 1A Stage 2 Stage 1A 46 Methanol, Stage 2 HO2+HO2 short-circuits chain-branching 47 Typical 2nd Stage Chemical Amplifier for hydrocarbons Coupled loops. Different fuels give different yields of HO2 and OH from R+O2 Complicated, but not impossible. 48 Summary • We know a lot of combustion chemistry – Can quantitatively predict many experiments – But we need very large models to do it! • Still some important things we don’t know – Soot formation/oxidation – Flame extinction chemistry – Some aspects of low-T ignition • We have important tools in hand, though all need improvement.... – – – – Quantum Chemistry for thermo & rates Automated /Systematic Mechanism Generation Ways to analyze complex models, focus on key issues Automated search for new chemical reactions 49