PREDICTING ORGANOSULFUR CHEMISTRY IN FUEL SOURCES ARCHE$ by CALEB ANDREW CLASS MASSACHUSETTS INSTITUTE OF rECHNOLOLGY M.S. Chemical Engineering Practice Massachusetts Institute of Technology, 2011 B.S. Chemical Engineering Purdue University, 2009 JUN 162015 LIBRARIES Submitted to the Department of Chemical Engineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemical Engineering at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY February 2015 0 2015 Massachusetts Institute of Technology. All rights reserved. Signature of Auth or Signature redacted Department of Chemical Engineering January 29, 2015 Certified by__ ignature redacted K Accepted by William H. Green Hoyt C. Hottel Professor of Chemical Engineering T hesis Supervisor Signature redacted j Richard D. Braatz Edwin R. Gilliland Professor of Chemical Engineering Chairman, Committee for Graduate Students Abstract 3 3 Abstract Predicting Organosulfur Chemistry in Fuel Sources by Caleb A. Class Submitted to the Department of Chemical Engineering on January 29, 2015 in partial fulfillment of the requirements for the degree of Doctor of Philosophy Abstract Desulfurization of fossil fuels with supercritical water (SCW) has been the topic of many studies over the past few decades. This process does not require the use of any catalyst, eliminates the need for a hydrogen feed, and minimizes coke formation. Previous research has shown that it has the potential to be a viable commercial process, and recent experimental studies have proven that water acts as one hydrogen source for sulfur removal in this process. However, the exact desulfurization mechanism is largely unknown, as are many other reaction mechanisms involving sulfur compounds. Recent work has greatly expanded our ability to build comprehensive reaction mechanisms automatically for the decomposition of organic sulfur compounds using the automated Reaction Mechanism Generator (RMG). This thesis presents the implementation of this and other tools to investigate chemical processes relevant to our use of fuel sources containing sulfur compounds, and it shows some steps that have been taken to improve our predictions for these mechanisms and those that will be generated in the future. Previous investigations had focused on the pyrolysis of small sulfur compounds containing less than six heavy atoms, so RMG is first used to study the pyrolysis of t-butyl sulfide. A detailed reaction mechanism is then presented for the SCW desulfurization of hexyl sulfide. Comprehensive kinetic mechanisms for these larger molecules are likely to include thousands of reactions, so RMG builds this model in a systematic and unbiased way using a database of ab initio data. This database is expanded with potentially relevant thermochemical and kinetic parameters using transition state theory and quantum chemical calculations at the CBS-QB3 and CCSD(T)-F12 levels of theory. With these data, as well as previously calculated rates for Abstract 4 hydrocarbon and sulfur kinetics, RMG is used to build a reaction mechanism for the conversion of hexyl sulfide to hydrogen sulfide, pentane, and carbon monoxide in the presence of SCW. This mechanism is validated with results from batch and flow reactor experiments, and predictions are accurate within a factor of two for reactant and major product concentrations. Analysis of the proposed mechanism shows that the molecular addition of water to the carbonsulfur double-bond in hexanethial is a key step in the SCW process, as this not only leads to the desulfurization of the compound, but also prevents the thioaldehyde from undergoing addition reactions with other hydrocarbons in a process that could eventually form coke. Thus, this work not only has implications in the SCW desulfurization process, but in the overall crude oil upgrading process as well. The calculated kinetic and thermochemical parameters are used to generate predictive reaction mechanisms for other processes relevant in fuel chemistry, such as the geological formation of oil and gas from kerogen. This not only allows us to model experimental work investigating the effect sulfur compounds have on the oil-to-gas process, but we also explore how these effects differ at geological conditions and timescales. And as the possible applications of RMG grow, the need for accurate parameters in mechanism generation become even more critical. A thermochemical database is generated for a wide variety of sulfur compounds using the highaccuracy CCSD(T)-F12/cc-pVTZ-F12 method, and this provides a basis for the investigation of organosulfur chemistry with tighter uncertainty. Thesis Supervisor: William H. Green Title: Hoyt C. Hottel Professor of Chemical Engineering Acknowledgements 5 Acknowledgements Thank you to my thesis advisor, Prof. William H. Green. Your encouragement through the ups and downs of a difficult project has been critical to my success, and your input has pushed us to exciting new discoveries. I'm extremely grateful for your mentorship, not only in helping me achieve success in kinetic modeling, but in my future professional life as well. Thank you to my thesis committee members, Profs. Ahmed Ghoniem, Yuriy Romdn, and Michael Timko. Your suggestions at committee meetings, videoconference preparation meetings, and other interactions have not only been valuable in guiding the direction of my research, but have also improved my ability preparing and giving presentations to a variety of audiences. Prof. Timko was also the director for the upgrading project for my first few years at MIT, and our interactions during this period were critical to my success in the remaining years. Thank you to Dr. Yuko Kida, whose curiosity, diligence, and attention to detail in the lab led to the discoveries that made my thesis scientifically satisfying, and whose presence in the lab made it a truly enjoyable place to work. Thank you to Dr. Aaron Vandeputte for writing such a good thesis and providing a strong basis for the predictive modeling of organosulfur chemistry. It was a pleasure working with you, even when the models weren't such a pleasure. Thank you the RMG developers, server administrators, and quantum chemists, including Shamel, Connie, Jorge, Josh, Ray, Nick, Mike, Richard, Enoch, Nate, and Amrit. This thesis is only possible thanks to your commitment to removing bugs, repairing hardware, and making sure my rate calculations weren't ridiculous. Thank you to the past and present members of the SCW Upgrading team, including Adam, AJ, Lawrence, Raj, Andrew, and Ashwin. Your contributions have allowed us to overcome numerous experimental and modeling challenges, and my thesis, as well as our overall knowledge of the upgrading process, is better for it. Acknowledgements 6 Thank you to Prof. Shuhei Ono and Dr. Eoghan Reeves. It has been both fascinating and enjoyable parsing and interpreting your experimental results with you, and gaining new insight in the oil-to-gas process. Thank you to Prof. Ilse Ipsen for giving this chemical engineer the chance to be a mathematician for a week. Thank you to Gwen, Alison, and Barb, for helping the group's operations run smoothly (probably in many more ways than I'm even aware of). Your problem-solving, willingness to lend an ear, and even your pizza choices are greatly appreciated. Thank you to all of the past and present members of the Green Group. You're not only an intelligent, curious, and hard-working bunch, but a fun group of people to spend an evening with. The value of going to a conference, meeting, or regular workday with helpful, kind, and pleasant coworkers cannot be understated, and I'm incredibly grateful for this. Thank you to all of the friends I've met while at MIT. Even when a day at the office went poorly, I always looked forward to spending time with you guys, whether doing something simple like a dinner, pub quiz, or Lame Night, or something incredibly stupid like a Pieathlon. Thank you to my parents, siblings, and the rest of my family for loving me, supporting me, and listening to me (even when I wasn't in the mood to talk). This is the point where I would put a joke acknowledgement for a random celebrity, but I've learned not to do that anymore. Table of Contents 7 Table of Contents Chapter 1: Introduction................................................................................................................. 11 1.1 Form ation and composition of fuels........................................................................... 12 1.2 Pyrolysis of sulfur compounds.................................................................................... 13 1.3 Supercritical w ater (SCW ) treatment of crude oil...................................................... 15 1.4 Autom atic m echanism generation.................................................................................. 19 1.5 Param eter estim ation.................................................................................................. 20 1.6 Ab initio calculations.................................................................................................. 21 1.7 Thesis Overview ............................................................................................................. 22 1.8 References ...................................................................................................................... 24 Chapter 2: M odeling the pyrolysis of t-butyl sulfide............................................................... 27 2.1 Abstract .......................................................................................................................... 27 2.2 Introduction .................................................................................................................... 28 2.3 M ethods.......................................................................................................................... 29 2.4 Results and Discussion................................................................................................ 30 2.4.1 Calculation of Unim olecular D ecom position Rates ............................................ 30 2.4.2 Pyrolysis of Neat t-Butyl Sulfide........................................................................ 32 2.4.3 Pyrolysis of t-Butyl Sulfide with Cyclohexene ................................................... 35 2.4.4 M echanism Comparison ......................................................................................... 37 2.5 Conclusions .................................................................................................................... 38 2.6 References ...................................................................................................................... )9 Chapter 3: A kinetic database for organic sulfur and oxygen compounds ............................... 41 3.1 Abstract .......................................................................................................................... 41 3.2 Introduction .................................................................................................................... 42 3.3 Methods.......................................................................................................................... 44 3.3.1 Calculation of Rate Constants for Reactions with Submerged Transition States... 46 3.3.2 Basis Set benchm arking for CCSD(T)-F 12 Calculations .................................... 3.4 Results and D iscussion................................................................................................ 47 48 3.4.1 Molecular Addition of Water (Hydration of Double Bonds)............................... 48 3.4.2 M olecular Addition of Hydrogen Sulfide .......................................................... 52 Table of Contents 8 3.4.3 Hydrogen Abstraction Reactions ........................................................................ 56 3.4.4 Radical Addition to Double Bonds (Reverse Beta-Scission) ............................. 60 3.4.5 Tautom erization of Thiocarboxylic Acids ........................................................... 63 3.4.6 Therm ochem ical Library ...................................................................................... 65 3.5 Conclusions .................................................................................................................... 66 3.6 References ...................................................................................................................... 67 3.7 Appendix: Calculations for reactions 21 and 37 ........................................................ 70 Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water...................... 71 4.1 Abstract .......................................................................................................................... 71 4.2 Introduction .................................................................................................................... 72 4.3 M ethods.......................................................................................................................... 73 4.3.1 Batch Reactor Experim ent .................................................................................... 73 4.3.2 Batch Reactor M odel .......................................................................................... 74 4.3.3 Continuous Flow Stirred Reactor (CSTR) Experiment ...................................... 74 4.3.4 CSTR Reactor M odel........................................................................................... 75 4.3.5 Principles of Autom ated M echanism Generation................................................... 75 4.3.6 Quantum Calculations............................................................................................. 76 Results: Quantum Calculations ...................................................................................... 77 4.4 4.4.1 Water-Catalyzed Elim ination of H 2 S ................................................................. 77 4.4.2 Hydrogen M igration............................................................................................. 79 4.4.3 Radical Addition to M ultiple Bond...................................................................... 80 4.4.4 Cyclic Sulfide Formation 4.4.5 Therm ochem istry Calculations ............................................................................. 84 Results: RM G Model Perform ance ............................................................................. 84 4.5 ....................................... 4.5.1 Reaction Path Analysis ........................................................................................ 84 4.5.2 M odel validation for hexyl sulfide conversion in a CSTR .................................. 89 4.5.3 Model validation for product distributions in a batch reactor.............................. 90 4.5.4 Effect of W ater Concentration ............................................................................. 93 4.6 Conclusions .................................................................................................................... 94 4.7 References ...................................................................................................................... 95 Chapter 5: Modeling the decomposition of alkylaromatic compounds .................................... 98 Table of Contents 9 5.1 A bstract .......................................................................................................................... 98 5.2 Introduction .................................................................................................................... 99 5.3 M ethods.......................................................................................................................... 99 5.3.1 Reaction Sim ulation............................................................................................. 5.3.2 Quantum Calculations 5.4 ......................................... Results and D iscussion................................................................................................. 99 100 101 5.4.1 Quantum Calculations........................................................................................... 101 5.4.2 RM G M odel Perform ance..................................................................................... 102 5.5 Conclusions..................................................................................................................105 5.6 References .................................................................................................................... Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition......................... 106 108 6.1 Abstract........................................................................................................................ 108 6.2 Introduction 109 6.3 M ethods........................................................................................................................ ................................................... 110 6.3.1 Reaction Sim ulation.............................................................................................. 110 6.3.2 Quantum Calculations........................................................................................... 1-10 Results and D iscussion................................................................................................. 111 6.4 6.4.1 Diethyl Disulfide Decomposition M echanism ...................................................... 111 6.4.2 Phenyldodecane D ecom position M echanism ....................................................... 117 6.4.3 Phenyldodecane Decomposition in the Presence of Diethyl Disulfide ................ 120 6.5 Conclusions .................................................................................................................. 123 6.6 References.................................................................................................................... 123 Chapter 7: A therm ochem ical database for organosulfur com pounds........................................ 7.1 A bstract ........................................................................................................................ 7.2 Introduction..................................................................................................................126 7.3 M ethods........................................................................................................................ 7.3.1 7.4 Regression of Bond Additivity Corrections (BA C's)........................................... 125 127 128 Results and D iscussion.................................................................................................129 7.4.1 BAC Regression and V alidation........................................................................ 7.4.2 Therm ochem ical Database..................................................................................131 7.5 125 Conclusions .................................................................................................................. 129 137 ................. Table of Contents 10 7.6 References .................................................... 138 7.7 Appendix: Calculated Thermochemical Parameters.................................................... 140 C hapter 8: C onclusions ............................................................................................................... 150 8.1 S u m mary ...................................................................................................................... 150 8.2 Recommendations for future work............................................................................ 151 8.3 R eferen ces .................................................................................................................... 152 ------------ - .. Chapter 1: Introduction Chapter 1: Introduction 11 11 Chapter 1: Introduction As the need for affordable sources of energy increases throughout the developing and developed world, high-value fuel sources like light crude oil have become rarer and more difficult to obtain. Existing petroleum reservoirs contain heavier oils that are more expensive to process,' while even more expensive processing is required to meet stricter environmental regulations on the impurities that are allowed in commercial fuels. 2 This increased cost, in addition to the increased cost of nontraditional drilling, makes petroleum exploration a much riskier affair, as the margin of error has shrunk dramatically. Understanding the chemistry of these fuel sources can greatly enhance our ability to find and harness these hydrocarbons. Reservoir models currently employ very general models to make billion-dollar drilling decisions. Refinery models use much more detailed kinetic mechanisms, but these are also frequently fit to match their exact conditions, and may therefore be prone to errors when applied to different conditions. Sulfur compounds in fossil fuels and their precursors are of particular interest. Sulfur can be present at up to 10 wt% in crude oil, and these compounds must be removed to comply with environmental regulations and prevent the poisoning of catalytic converters. A detailed kinetic Chapter I.: Introduction 12 understanding of how various sulfur compounds react with other hydrocarbons, as well as how they can be removed most efficiently, would potentially improve our ability to find and process fuel sources. 1.1 Formation and composition of fuels Fossil fuels are produced by the decomposition of plant and animal remains at high pressures, deep beneath the surface of the earth. While coal was mainly formed from the remains of ferns and other plants in swampy areas, the organisms in the ocean mainly decompose to form kerogen, a heavy, waxy hydrocarbon. 3 This kerogen reacts further to form natural gas and crude oil, which may also continue to produce more natural gas. The kerogen-to-oil-to-gas process is likely affected by the presence of metals, 4 water,' 6 and reactive sulfur compounds,7 and geochemists use this information to make drilling decisions. This research is focused on the compounds found in crude oil, which can be separated into four main categories: saturates, aromatics, resins, and asphaltenes. 8 The first category includes simple alkanes and organosulfur molecules like thiols, sulfides, and disulfides, while the aromatics fraction includes thiophenes and benzothiophenes. The resins and asphaltenes are larger compounds that generally contain multiple aromatic rings, a saturated portion, and some heteroatoms. Asphaltenes are larger than resins, and they precipitate out of the mixture when nalkanes are added. While some constituent units in these larger species have been identified,9 their full molecular structures are not well known. One example of a possible asphaltene molecule can be seen in Figure 1. N Figure 1. Hypothetical asphaltene structure. The sulfur types present in crude oil vary widely from reservoir to reservoir. Table 1 shows the relative abundance of five sulfur types in the various fractions obtained from two different crude Chapter 1: Introduction 13 oils.' 0 Sulfides are generally the easiest to desulfurize, but recent work has shown that these compounds can react to form aromatic compounds in simple pyrolysis." Thiophenes are much more stable in noncatalytic treatment, but the removal of alkyl chains connected to these aromatic rings may be a key to overall upgrading chemistry. A more detailed understanding of the chemistry of each of these compounds will be necessary to improve our utilization of crude oil. Table 1. Relative abundance of different sulfur types in two representative crude oils.'0 Sulfide 1.2 Thiophene Sulfoxide Sulfone Sulfate CAL Asph 15 29 50 5 1 CAL Resin CAL Oil KUW2 Asph KUW2 Resin KUW2 Oil 11 24 40 40 45 27 27 55 52 47 59 46 2 5 5 1 1 1 1 1 1 1 1 1 1 Pyrolysis of sulfur compounds An understanding of the decomposition mechanisms of individual organosulfur compounds, including both aliphatic and aromatic sulfur, is critical to gaining insight on how these species affect overall processes in fuel chemistry. Much of the research in the pyrolysis of aliphatic sulfur compounds is focused on sulfides. In particular, a variety of alkyl t-butyl sulfides and alkyl allyl sulfides have been studied by Martin et al., and the overall Arrhenius parameters for these reaction networks were tabulated in Martin's review on organosulfur compound pyrolysis.12 The results for alkyl t-butyl sulfide pyrolysis supported the transition state suggested by Benson and Haugen," which is shown in Figure 1. In this type of scheme, the reaction would occur more quickly if the transition state is stabilized by resonance effects; this is true in Martin's work, as the overall pyrolysis of t-butyl phenyl sulfide is found to have a lower activation energy than the fully non-aromatic compounds. 4 14 Chapter 1: Introduction (CH3) 2C----S--Ri H-S-R S-R C- ----H H2 -J 3 Figure 2. Proposed reaction pathway for the decomposition of an alkyl t-butyl sulfide.' Researchers have also proposed free radical mechanisms for the pyrolysis of some sulfide compounds. Many past studies were not able to quantitatively validate their proposed mechanisms with experimental data, but there has been some recent progress in this area. Zheng et al. proposed a kinetic mechanism for the pyrolysis of diethyl sulfide, and it compared reasonably well with the experimental results.15 Work by Vandeputte et al. has provided additional insight into decomposition of small organosulfur compounds, using computational chemistry techniques to model the pyrolysis of dimethyl disulfide and diethyl sulfide.' 6' 17 The study provided a validation of an expansive thermochemistry of organosulfur compounds.' 8-21 second database for the kinetics and Thiols are important intermediates in many sulfur pyrolysis mechanisms.18-21 A few of the compounds previously studied include t-butanethiol, I-butanethiol, and bicyclo[2,2, 1 ]heptane-2thiol. The pyrolysis of a thiol is believed to be initiated by fission of the C--SH bond, and H 2S is quickly formed as a major product, along with an olefin.2 Some research has also been conducted on disulfide pyrolysis. Available studies include the pyrolysis of dimethyl disulfide in a static system, as well as di-t-butyl disulfide and a few aryl t-butyl disulfides, and these results can be used to validate computational pyrolysis models.22 -24 As a major component of coal and heavy oil sands, a significant amount of research has been done on the high temperature chemistry of aromatic sulfur compounds. In one of the coal-related studies, Bruinsma et al. investigated the gas phase pyrolysis of thiophene, furan, pyrrole, 27 pyridine, and benzene, as well as each of their benzo- and dibenzo- derivatives.25- The main finding of this work was that the sulfur-containing rings were significantly more stable than the other compounds. As shown in Figure 2, temperatures of over 900'C were required for 10% conversion of thiophene and benzothiophene in five seconds (dibenzothiophene required a slightly lower temperature). The authors concluded that the initiation of thiophene pyrolysis occurs by ring-H scission, which accounts for the similar stability of the benzene molecules; the 15 Chapter 1: Introduction pyrolysis of other heterocycles is likely initiated by bond fission between a carbon and the heteroatom. A computational study of thiophene decomposition predicted very high barriers in the likely decomposition mechanism, suggesting that this decomposition would not appreciably occur in refinery processes without catalysts. 28 0--a '73-b r-r 97 1073 1173 - Q273 T/K - 873 Figure 3. The temperature required for 10-50% conversion (in five seconds) of each heterocycle (a), benzo29 derivative (b), and dibenzo- derivative (c). Figure from Bruinsma et al. More research has been done specifically on the topic of benzothiophene and dibenzothiophene pyrolysis. Dartiguelongue et al.28 looked at dibenzothiophene pyrolysis between 375 and 500'C, and this work showed that desulfurization occurred more at lower reactant conversions, while radicals combined to form heavy products and high conversions. A mechanism was proposed to explain this process, but it was not confirmed experimentally. 1.3 Supercritical water (SCW) treatment of crude oil The critical point of water is at 374'C and 22.06 MPa.30 As this point is approached and exceeded, the properties of water vary in ways that make it a candidate medium for crude oil treatment. One of the most important properties for this application is the dielectric constant, which continuously decreases until shortly after the critical temperature, where it settles at a very low value (as seen in Figure 3). This is evidence that hydrogen bonding becomes less prevalent as the temperature increases. The decrease is significant enough that supercritical water can be 16 Chapter 1: Introduction used as a nonpolar solvent for hydrocarbons. The mixing of supercritical water and various ' hydrocarbons has been further studied using transport models.3 a x 10; p /kg m 3 -logKw 10 1400 . 3 ~14 1000 218 22 600 - 26 200 0 473 673 T)K ionic product (3) of water Figure 4. Temperature dependences of the dielectric constant (1), density (2), and 32,33 & Lunin. at a pressure of 24 MPa. Figure from Galkin As the temperature increases within the subcritical region, the ionic product of water increases to a maximum, three orders of magnitude greater than at standard conditions. However, this value quickly decreases as you pass the critical point, and it monotonically decreases in the supercritical region. This means that subcritical water (near the critical point) is a much greater source of H+ and OH- ions, and ionic chemistry should be less important in the critical region. Using this knowledge, multiple researchers have attempted to remove sulfur from organic compounds by SCW treatment. Katritzky et al. studied the kinetics of thiophene and benzothiophene, as well as a variety of sulfides and thiols containing aromatic rings, in SCW at 4600C." The thiols and sulfides were desulfurized to varying degrees, with diphenyl sulfide being notably less reactive than the others, likely due to the lack of abstractable hydrogen to form a reactive intermediate in decomposition. Thiophene was found to be completely unreacted after one hour in SCW (and in 15% solutions of formic acid or sodium formate), while only 1.9% of benzothiophene was desulfurized after one hour in formic acid solution, which may be within the margin of error of the experiment. In addition, Vogelaar et al. 4 studied pretreated gasoil that had been spiked with benzothiophene, dibenzothiophene, diphenyldisulfide, and octadecanethiol. The results found were similar to those of Katritzky et al., as only the thiol and WS Chapter 1: Introduction 17 disulfide were desulfurized at 673 K. Finally, Townsend et al.35 treated dibenzothiophene in 550'C SCW, and found that this also resulted in the same small conversion, about 2% after four hours, as dry pyrolysis. These studies provide evidence that the addition of SCW alone is not sufficient to desulfurize aromatic hydrocarbons like benzothiophene. The desulfurization of aromatic hydrocarbons has also been explored using alkaline SCW, and these experiments have provided significantly better results. Yoshida et al.36 treated thiophene in NaOH concentrations ranging from 1.0 to 5.0 mol/dm3 , and a maximum conversion of 58% was reported in 20 minutes at 400'C, with a NaOH concentration of 4.0 mol/dm 3 . Kishita et al.3 7 also studied this type of desulfurization using NaOH and KOH. The model compound study in an NaOH solution showed that about 70% of benzothiophene was converted in one hour at only 300'C, while 60% conversion of dibenzothiophene was reported at 430'C. Kishita also explored the desulfurization of bitumen-hydrocarbon mixtures containing a much greater heavy fraction-in alkaline SCW at 430'C. This resulted in about 60% desulfurization after one hour using a 5 mol/dm 3 KOH concentration, and the reaction led to an overall decrease in the pH of the aqueous phase. Because of this, the authors concluded that the base was consumed while causing the bitumen to break down into benzothiophenes and dibenzothiophenes. From here, the benzothiophenes were decomposed to result in most of the desulfurization, while the dibenzothiophenes were desulfurized to a much lesser extent due to their greater stability. Kishita proposed that the sulfur was removed as S 2 , and this conclusion was supported by Yoshida et al., who found that this ion was the only sulfur product in the alkali SCW desulfurization of thiophene. However, it is important to note that the metallic reactor walls may have been acting as catalysts in these studies. Recently, Patwardhan et al. used a continuous stirred tank reactor to study the kinetics of the SCW treatment of hexyl sulfide and benzyl sulfide, and the resulting 3/2 order conversion suggested that the decomposition rate was controlled by a free-radical mechanism. 38 ' 39 Kida et al. proved that SCW is a reactant in the desulfurization mechanism of hexyl sulfide, and intermediate studies were used to propose a combined free-radical and molecular mechanism, presented in Figure 4, to explain the observed products.4 0 This not only has the effect of removing some sulfur from the reaction mixture, but it also prevents the reactive thioaldehyde 18 Chapter 1: Introduction compounds (containing a carbon-sulfur double-bond) from undergoing addition and cyclization reactions that could form undesired heavy products. CSH COCO 1 1 S C5 H11 C5H12 . C 5H 11 CSH1I AO CSH1 CA11-tS C6H14 C 5H 1 _11 C6H 13 C 5H 11 S OH H2S CSH 11SHH2 (H20/ Figure 5. Proposed mechanism for conversion of hexyl sulfide to pentane and hydrogen sulfide in SCW treatment. Figure from Kida et al' Work in the Green Group has agreed with the previous findings that thiophene rings are recalcitrant to SCW treatment," but the alkyl chains connected to these rings are likely to break at these conditions. Analysis of crude oil fractions (before and after SCW treatment) by twodimensional gas chromatography, as shown in Figure 5, has demonstrated a shift in the peaks of sulfur compounds from the heavier to the lighter range, while a much heavier product distribution was obtained from anhydrous pyrolysis of the same feedstock. This decrease in density, or upgrading, would provide a significant increase in the value of the oil, suggesting that 40 41 SCW treatment has the potential to be a viable commercial process. ' Chapter 1: Introduction Chapter 1: Introduction 19 19 Figure 6. GCxGC-SCD chromatograms for crude oil bottoms before (a) and after (b) SCW treatment at 450 *C. x-axis shows volatility (heavier compounds on right), y-axis shows polarity. Figure from Kida et al.42 1.4 Automatic mechanism generation Over the past few decades, automatic mechanism generation has become a viable replacement to the manual creation of free radical mechanisms. Systematization of the model-building process allows for the use of the vast chemical knowledge available in literature, provides mechanisms much more quickly than possible by hand, and prevents many of the common pitfalls in manual work, such as biases and transcription errors. Several other groups have developed software that also automates this process. 42 Our effort has focused on the development of the open-source Reaction Mechanism Generator (RMG). 43 44 RMG-Java was used for most of the work in this thesis, although a version is also under development in Python. The algorithm and use of this software have been extensively described in the thesis of Joshua Allen, 45 so only a general introduction will be provided here. The key feature of RMG is its iterative, flux-based model expansion algorithm, which allows for the detailed pursuit of reaction pathways that are predicted to have a greater flux, while placing less emphasis on those that are predicted to be minor. Only a small amount of input is required from the user to generate an initial mechanism, including the reactant geometries and concentrations, temperature, pressure, and termination criteria. More advanced features have been implemented recently, including 20 Chapter 1: Introduction pressure-dependent rate calculations, on-the-fly quantum calculations, solvation chemistry, and chemistry for molecules containing nitrogen and sulfur, the latter if which is the topic of this thesis. 1.5 Parameter estimation Accurate kinetic and thermochemical data have been collected experimentally for thousands of reactions over the past hundred years, while reasonably accurate calculations have been conducted to calculate parameters for perhaps tens of thousands more. Unfortunately, the entire database of the kinetics community does not come close to capturing the billions of reactions possible in nature, or even the millions of reactions that may be considered in one RMG run! Thus, algorithms are implemented to use the available data to estimate parameters that are as accurate as possible. Enthalpies, entropies, and heat capacities can be captured by group additivity schemes, of which the most widely-used was defined by Benson and Buss46 and expanded by Vandeputte et al. for organosulfur compounds. 47 An example application of this Group Additivity Value (GAV) scheme is presented in Figure 7. These GAV's were regressed from experimental measurements and ab initio calculations for a variety of compounds, so they can be updated when additional data are available. Additional corrections are available to estimate thermochemical parameters for rings, radicals, and many other special cases. 0 lI \C= 0 /CH CH 2 S 0 0 AHfi 98 = + H1 98 + Hf,298+ 2H 2 98 +4Hf 1 9 8 Figure 7. Application of group additivity scheme to estimate enthalpy of formation for 2,3-dihydrobenzothiophen-3-one. 21 Chapter I: Introduction The estimation of reaction rate parameters can be more complicated, especially for bimolecular reactions. When RMG identifies a species or set of species as matching a specific reaction type, a tree structure is used to determine the most similar reaction for which reaction rate data are available. This works well in some cases, but in others it leads to estimation by averaging a large number of other rate constants, providing parameters with uncertainties of multiple orders of magnitude. If a poorly-estimated reaction rate is determined to be important to the overall predictions of a mechanism, either by sensitivity analysis or chemical intuition, improved parameters must be obtained. This can be done via experiments that isolate the reaction of interest, or via ab initio calculations. As the former method is generally more accurate but also more expensive, it is most valuable when there is a small number of reactions that need to be known with very high accuracy. The second method was utilized for this thesis, as rate parameters needed to be calculated for a large number of new reactions. 1.6 Ab initio calculations Ab initio is Latin for "from the beginning," and in the field of Physical Chemistry, it refers to the determination of molecular properties by solving the Schr6dinger equation, HW = EW, where H is the Hamiltonian operator, E is the energy of state W, and T is the wave function, which describes the state of the electrons in the system. The solution of this equation gives us exact physical and thermochemical parameters for a molecule of interest, but unfortunately, this exact solution is only possible for one-electron systems. A wide variety of models have been proposed to approximate the solution, including Hartree Fock (HF) and Density Functional Theory (DFT).2 Composite methods, such as CBS-QB3, have also been successful in the past,48 but CCSD(T)-F12 has recently been shown to provide more accurate energy calculations with a similar computational cost. 49'50 Quantum calculations in this thesis were conducted using the Gaussian 0351, 52 and Molpr5 3 software packages. The exact methods used vary, so the exact details can be found in each chapter. For species optimizations, N-dimensional scans, where N is the number of rotatable bonds in the compound, at the B3LYP/6-3 IG(d) level of theory or higher, were first conducted Chapter 1: Introduction 22 to roughly identify the conformer with the lowest energy. Local optimizations, frequencies, and single-point energies were then calculated using higher levels of theory. The CanTherm software package5 4 was used to calculate reaction rate constants using transition state theory, assuming reactions were occurring in the gas phase. 1.7 Thesis Overview This thesis focuses on the use of automatic mechanism generation to gain new insights into processes relevant to our use of the planet's fossil fuels, which previously had only been understood at a high level from bulk experimental data. In particular, we focus on generating mechanisms for the decomposition of sulfur compounds, exploring how these compounds interact with other species present in fuel sources, and improving the data available for generating accurate predictions in the future. Chapter 2 presents an application of the RMG algorithm on the pyrolysis of t-butyl sulfide. Previous modeling work had focused on modeling the decomposition of sulfides containing an alpha-hydrogen (hydrogen bonded to the carbon adjacent to the sulfur atom), the donation of which provides a key propagation step in their decomposition mechanisms. This chapter validates the database for sulfur compounds without an alpha-hydrogen, and additionally demonstrates the ability of RMG to explain and quantitatively test the key reaction steps for mechanisms that could only be hypothesized in the past. In Chapter 3, the RMG databases for thermochemistry and reaction kinetics are expanded to allow for the simulation of organic mixtures containing both oxygen and sulfur. Experimental and calculated data for this type of chemistry are very sparse in literature, so accurate coupledcluster calculations were conducted to provide rate parameters for hydrogen abstraction and radical addition reactions that involve both heteroatoms. A database was also built for the addition reactions of water and hydrogen sulfide to double bonds, as these reactions were proposed as key steps in the supercritical water desulfurization of sulfides. Chapter 4 shows the application of the expanded RMG database to study the pyrolysis and supercritical water treatment of hexyl sulfide. The resulting mechanisms provided elementary Chapter 1: Introduction 23 reaction pathways to explain the formation of all of the major products from both treatments, as well as most of the minor products, and product predictions agreed well with quantified experimental product distributions. The main difference between the two mechanisms was found to be the fate of the reactive thioaldehyde intermediate, which can continue to form aromatic thiophenes in the absence of water, and potentially undesired heavy products as well. This intermediate is desulfurized in SCW to form lighter products. Thus, this work not only shows one of SCW's effects in desulfurization chemistry, but in the overall crude oil upgrading process as well. In Chapter 5, an RMG mechanism is presented for the dealkylation of hexyl benzene in pyrolysis and SCW treatment. As thiophenes and benzenes are known to be particularly stable in the absence of catalysts, the breaking of alkyl chains connecting multiple aromatic rings may be a key to the overall crude oil upgrading process. The work in this chapter shows that SCW alone does not directly react with hexyl benzene or its intermediates, as the model predicts the same product distribution in the presence and absence of SCW, a result validated by experimental data. Chapter 6 presents an application of RMG on understanding the effects of sulfur compounds on the geological dealkylation of aromatic compounds. Decomposition mechanisms were built for phenyldodecane and diethyl disulfide, which have previously been used as model heavy oil and sulfur compounds in experimental studies. Mechanisms were generated to be valid both at conditions previously studied experimentally and at conditions more relevant geologically, where experiments cannot be completed on a reasonable timescale. As expected, the disulfide and alkyl aromatic compound reacted on vastly different timescales, and this could have important implications in organic geochemistry. Chapter 7 provides a database of high-accuracy thermochemical parameters for organosulfur compounds. Enthalpies, entropies, and heat capacities were first calculated for a training set of species for which reasonably accurate experimental thermochemical parameters are available, and the comparisons between calculated and experimental values were used to generate additivity corrections for enthalpy calculations. The database was then expanded to include parameters for additional sulfur compounds, which are used as a starting point for the generation Chapter I.: Introduction 24 of additivity parameters for the estimation of sulfur compounds in pyrolysis and oxidation chemistry. Chapter 8 presents overall conclusions of this work, as well as some thoughts on future areas of research in the predictive modeling of organosulfur chemistry. This not only includes challenges that may be faced computationally, but also experimental work that will be critical to decreasing the uncertainty, and increasing the value, of models built using these techniques. References 1. 2. 3. V. A. Gembicki, T. M. Cowan and G. R. Brierley, Hydrocarbon Processing, 2007, 86. C. S. Song, in National Meeting of the American Chemical Society, Boston, MA, 2002. H. H. Schobert, in Chemistry of Fossil Fuels and Biofuels, Cambridge University Press, New York, 2013. S. J. M. Butala, J. C. Medina, T. Q. Taylor, C. H. Bartholomew and M. L. Lee, Energy Fuels, 2000, 14, 235-259. M. D. Lewan, Geochimica et Cosmochimica Acta, 1997, 61, 369 1-3723. M. D. Lewan and S. Roy, Organic Geochemistry, 2011, 42, 31-41. M. D. Lewan, Nature, 1998, 391, 164-166. C. A. Islas-Flores, E. Buenrostro-Gonzalez and C. Lira-Galeana, Energy & Fuels, 2005, 19, 2080-2088. O. P. Strausz, T. W. Mojelsky and E. M. Lown, Fuel, 1992, 71, 1355-1363. S. Mitra-Kirtley, 0. C. Mullins, C. Y. Ralston, D. Sellis and C. Pareis, Applied Spectroscopy, 1998, 52, 1522-1525. Y. Kida, C. A. Class, A. J. Concepcion, M. T. Timko and W. H. Green, Phys Chem Chem Phys, 2014, 16, 9220-9228. G. Martin, in The chemistry of sulphur-containingfinctionalgroups, John Wiley & Sons, New York, 1993, pp. 395-437. S. W. Benson and G. R. Haugen, J. Am. Chem. Soc., 1965, 87. G. Martin, H. Martinez and J. Ascanio, InternationalJournalof Chemical Kinetics, 1989, 21, 193-206. X. Zheng, J. W. Bozzelli, E. M. Fisher, F. C. Gouldin and L. Zhu, Proceedingsof the Combustion Institute, 2011, 33, 467-475. A. G. Vandeputte, C. A. Class, M.-F. Reyniers, W. H. Green and G. B. Marin, Submitted, 2014. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, J. Phvs. Chem. A, 2010, 114, 1053 110549. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, ChemPhysChem, 2013, 14, 17031722. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, ChemPhysChem, 2013, 14, 37513771. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. & 1.8 Chapter 1: Introduction 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 25 A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Phys Chem Chem Phys, 2012, 14, 12773-12793. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Chemistry-A European Journal, 2011, 17, 7656-7673. W. Tsang, J. Chem. Phys., 1964, 40, 1498-1505. T. 0. Bamkole, J. Chem. Soc. Perkin Trans., 1977, 2, 439-443. D. E. Johnson and A. F. Dimian, J. Chem. Soc. Chem. Comm., 1987, 416-417. J. A. R. Coope and W. A. Bryce, Can. J. Chem., 1954, 32, 768-779. G. Martin and N. Barroeta, J. Chem. Soc. Perkin Trans. 2, 1976, 1421-1424. G. Martin and J. Ascanio, J. Phys. Org. Chem., 1991, 4, 579-585. 0. S. L. Bruinsma, P. J. J. Tromp, H. J. J. D. Nolting and J. A. Moulijn, Fuel, 1988, 67, 334-340. X. Song and C. A. Parish, J. Phys. Chem. A, 2011, 115, 2882-2891. C. Dartiguelongue, F. Behar, H. Budzinski, G. Scacchi and P.-M. Marquaire, Org. Geochem., 2006, 37, 98-1.16. A. A. Galkin and V. V. Lunin, Uspekhi Khimii, 2005, 74, 24-40. S. Dabiri, G. Wu, M. T. Timko and A. F. Ghoniem, JournalofSupercriticalFluids, 2012, 67, 29-40. G. Wu, S. Dabiri, M. T. Timko and A. F. Ghoniem, Journal ofSupercriticalFluids, 2012, 72, 150-160. A. R. Katritzky, R. A. Barcock, M. Balasubramanian, J. V. Greenhill, M. Siskin and W. N. Olmstead, Energy & Fuels, 1994, 8. B. M. Vogelaar, M. Makkee and J. A. Moulijn, Fuel ProcessingTechnology, 1999, 61, 265-277. S. H. Townsend, M. A. Abraham, G. L. Huppert, M. T. Klein and S. C. Paspek, Ind. Eng. Chem. Res., 1988, 27, 143-149. S. Yoshida, K. Takewaki, K. Miwa, C. Wakai and M. Nakahara, Chem. Lett., 2004, 33, 330-331. A. Kishita, S. Takahashi, F. Jin, Y. Yamasaki, T. Moriya and H. Enomoto, J. Jpn. Pet. Inst., 2005, 48, 272-280. A. Kishita, S. Takahashi, Y. Yamasaki, F. Jin, T. Moriya and H. Enomoto, J. Jpn. Pet. Inst., 2006, 49, 177-185. P. R. Patwardhan, M. T. Timko, C. A. Class, R. E. Bonomi, Y. Kida, H. H. Hernandez, J. W. Tester and W. H. Green, Energy & Fuels, 2013, 27, 6108-6117. Y. Kida, Massachusetts Institute of Technology, 2014. Y. Kida, A. G. Carr and W. H. Green, Energy & Fuels, 2014, 28, 6589-6595. W. H. Green, in Advances in Chemical Engineering, ed. G. B. Marin, Academic Press, vol. 32. A. S. Tomlin, T. Turinyi and M. J. Pilling, in Comprehensive Chemical Kinetics, ed. M. J. Pilling, Elsevier, 1997, vol. 35, pp. 293-437. W. H. Green, J. W. Allen, R. W. Ashcraft, G. J. Beran, C. A. Class, C. Gao, C. F. Goldsmith, M. R. Harper, A. Jalan, G. R. Magoon, D. M. Matheu, S. S. Merchant, J. D. Mo, S. Petway, S. Raman, S. Sharma, K. M. Van Geem, J. Song, J. Wen, R. H. West, A. Wong, H.-W. Wong, P. E. Yelvington and J. Yu, Reaction Mechanism Generator (R.MG), (2013). Chapter I.: Introduction 46. 47. 48. 49. 50. 51. 52. 53. 54. 26 J. W. Allen, Massachusetts Institute of Technology, 2013. S. W. Benson and J. H. Buss, Journalof Chemical Physics, 1958, 29, 546-572. Y. Zhao and D. G. Truhlar, Accounts of Chemical Research, 2008, 41, 157-167. J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, Journalof ChemicalPhysics, 2000, 112, 6532-6542. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, Theor Chem Account, 2009, 123, 391-412. J. Aguilera-Iparraguirre, H. J. Curran, W. Klopper and J. M. Simmie, Journalof Physical Chemistrv A, 2008, 112, 7047-7054. W. Klopper, R. A. Bachorz, D. P. Tew, J. Aguilera-Iparraguirre, Y. Carissan and C. Hattig, Journalof Physical Chemistry A, 2009, 113, 11679-11684. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokurna, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and M. SchUtz, Wiley InterdisciplinaryReviews: ComputationalMolecularScience, 2011, 2, 242-253. Chapter 2: Modeling the pyrolysis of t-butyl sulfide 27 Chapter 2: Modeling the pyrolysis of t-butyl sulfide 2.1 Abstract The automated Reaction Mechanism Generator (RMG) is used to build reaction networks for the thermal decomposition of t-butyl sulfide. Simulation results were compared with data from pyrolysis experiments with and without the addition of a cyclohexene inhibitor. Purely freeradical chemistry did not properly explain the reactivity of t-butyl sulfide, as the previous experimental work showed that the sulfide decomposed via first-order kinetics in the presence and absence of the inhibitor. With additional reaction rate data from a library of unimolecular decomposition reactions of sulfides and thiols, as well as other published kinetic parameters for small-molecule sulfur chemistry, the agreement between model and data was significantly improved for the inhibited case. The unimolecular decomposition of t-butyl sulfide to form isopropene and t-butanethiol was found to be a key reaction in both cases, as it explained the first-order sulfide decomposition. Cyclohexene had a significant effect on the available radical pool, and this led to dramatic changes in the resulting product distribution. Chapter 2: Modeling the pyrolysis of t-butyl sulfide 2.2 28 Introduction Sulfur compounds are important in many aspects of life, including food, fuels, and the environment.1-3 Sulfur in fuel sources can lead to problems in processing and usage. Sulfur in crude oil leads to some challenges, as a sufficient amount of these compounds must be removed during refinement to satisfy governmental regulations and prevent the release of toxic sulfur compounds into the environment and the poisoning of catalytic converters.4- 6 In pyrolysis and steam cracking, sulfur compounds have a significant impact on the initiation, termination, and product distribution7 , leading to undesired process variability. Because of the importance of these compounds, significant experimental efforts have been undertaken to understand their chemistry, from pyrolysis -14 to oxidation"' ' to decomposition in aqueous and supercritical environments. 3-16 Due to the complexity in building accurate mechanisms to model these phenomena, most of the available literature on sulfur chemistry includes only speculative mechanisms without quantitative product predictions, with the exception of the study of diethyl sulfide pyrolysis by Zheng et al. 7 The development of automated reaction mechanism generation software' 8' 19 has greatly aided in this effort. For decades, this software was limited to the study of hydrocarbon species without sulfur.2 0 21 With extensions to the software and the availability of improved estimation methods for the thermochemistry and reaction rates of elementary organosulfur reactions, -24 it is now possible to use computational tools to shed mechanistic insight on past experimental studies of sulfur chemistry. Martin et al. previously studied the pyrolysis of a variety of alkyl t-butyl sulfides,2 5 and most of the observed product distributions supported the four-centered transition state suggested by Benson and Haugen, which is presented in Figure 1. t-butyl sulfide8 , which was pyrolyzed between 360 and 413 'C with and without the presence of a supposed radical inhibitor, cyclohexene, was the main exception to the trend. The expected products of the four-center reaction were observed in the presence of cylcohexene, but significantly different product distributions were obtained in the case of neat pyrolysis. In this chapter, kinetic mechanisms built by RMG are analyzed to clarify the key reaction steps for the two cases. . ... Chapter 2: Modeling the pyrolysis of t-butyl sulfide (CH S-R 29 -S-R I ] + H-S-R H2 Figure 1. Proposed reaction pathway for the decomposition of an alkyl t-butyl sulfide. 2.3 Methods The Java version of RMG was used to generate the mechanisms for this work.' 8 RMG uses an iterative, flux-based algorithm to build reaction mechanisms. With a user-supplied set of "core" species, RMG searches a database containing a set of specific reactions and a more general library of reaction recipes, to determine all possible reactions for the given species. The products of these reactions are then added to the model's "edge" if they are not already present in the mechanism. The reaction mixture is then simulated with the user-specified conditions and initial concentrations, until an edge species reaches the required flux to be added into the core. The database is then searched for reactions with the new core, and the whole process repeats until the specified termination condition is reached without any of the edge species reaching the required flux for addition to the core. The final mechanism only contains the core species and reactions, and this is output as a CHEMKIN input file. 27 Thermochemical parameters were calculated using the group additivity values developed by Vandeputte et al.2 4 The CBS-QB3 database of rate parameters for organosulfur compounds, including hydrogen abstraction, 2, 2 29 beta-scission,3 and homolytic substitution reactions, 3 was used to provide rate constant estimations for reacting sulfur compounds. Other parameters for free-radical and molecular reactions, available in the RMG database, were used estimate rate constants for reactions where sulfur was not present. Reactions found to be of particular importance in the model predictions were refined further by calculating single-point energies at the CCSD(T)-Fl2a/cc-pVDZ-F12 level of theory after geometry optimizations, frequency calculations, and hindered rotor scans using B3LYP/6311 G(2d,d,p), with a scaling factor of 0.99 used for the frequency analysis. This coupled-cluster method with F12 has been found to provide basis-set errors that are generally below 2 kJ/mol using triple-zeta and greater basis sets for small molecule calculations.32-34 As will be shown in Chapter 3, double-zeta calculations are within about 2 kJ/mol of the same using triple-zeta, and 30 Chapter 2: Modeling the pyrolysis of t-butyl sulfide should therefore be accurate within roughly 4 kJ/mol or 1 kcal/mol, although the true uncertainty is likely somewhat greater. A double-zeta basis set was chosen for this work because of the relatively large systems under consideration, which contain up to 10 heavy atoms. After quantum calculations were completed in Gaussian 0335 and Molpro, 36 the open-source CanTherm software package 3 7 was used to calculate rate constants between 300 and 2000 K, including a tunneling correction using the Eckart method.38 The rate constants were then fit to the modified Arrhenius form, k(T) = A x T' x exp ( RTa, where T is the temperature in Kelvin. 2.4 Results and Discussion 2.4.1 Calculation of Unimolecular Decomposition Rates Martin & Barroeta proposed a set of unimolecular reactions for the decomposition of t-butyl 39 disulfide to explain the formation of isobutene and hydrogen disulfide from the initial reactant. Transition states were found using quantum chemistry methods for these two reactions, and their geometries can be seen in Figure 2. The analogous mechanism is also possible in the pyrolysis of t-butyl sulfide, and optimized transition states for these reactions are presented in Figure 3. [ ~~H S ks -CH2 H -H -H 2C o _ k SH SH --- H2S2 H 1. 171.8 LOW 9 2.950 S 2.896 H H Figure 2. Proposed reaction pathway 39 (top) and optimized transition state geometries (bottom) for the molecular decomposition of t-butyl disulfide. Distances (A). 31 31 Chapter 2: Modeling the pyrolysis of t-butyl sulfide Chapter 2: Modeling the pyrolysis of t-butyl sulfide H ---CH2 S H2C- -H 10- SH jw- H2S L 1.9 .19 &I =ME@ H H H1 6 H Figure 3. Proposed reaction pathway (top) and optimized transition state geometries (bottom) for the molecular decomposition of t-butyl sulfide. Distances (A). Table 1. Calculated rate constants (using CCSD(T)-F12/cc-pVDZ-F12) for molecular elimination reactions. A [s-], n [unitless], E. [kJ/molI, k s-1]. I S "S< l H2 S2 S logOA n Ea k(380 'C) 11.65 1.04 214.5 2.6E-03 12.71 0.39 233.8 1.3E-05 12.51 0.89 239.4 7.7E-05 12.88 0.36 256.0 2.6E-07 + SH SH + S SH + S'SH - - - H2 S The calculated rate parameters are presented in Table 1. While the calculated unimolecular rate constant for sulfide consumption falls within the experimental uncertainty of the overall rate constant,8 the calculated rate for disulfide consumption is more than an order of magnitude slower than what was observed experimentally. 39 Calculations using a larger basis set might bring this prediction closer to the observation. The overall trend provides some insight into the differences between the reaction mechanisms of t-butyl sulfide and t-butyl disulfide. While tbutyl disulfide undergoes the full molecular mechanism to form isobutene and hydrogen disulfide-some of which can react further to form hydrogen sulfide-the elimination of H2 S from t-butanethiol is slower than the other three reactions in Table I by two orders of magnitude, suggesting that the consumption of t-butanethiol may occur more quickly by a free-radical Chapter 2: Modeling the pyrolysis of t-butyl sulfide 32 mechanism. This would explain the equal consumption rate of t-butyl sulfide with and without the radical inhibitor, with the cyclohexene inhibiting the radical pathway for H 2 S formation from t-butanethiol. The exact mechanism can be explored in more detail using RMG. 2.4.2 Pyrolysis of Neat t-Butyl Sulfide RMG simulated the pyrolysis of neat t-butyl sulfide at 380 'C and 217 Torr, using no diluent, a goal reactant conversion of 60%, and a core tolerance of 0.10. Complete convergence of the mechanism was not possible with the available memory, so the simulation was terminated with 289 species and 2749 reactions. The resulting CHEMKIN input file was used to simulate the conditions of the experiments by Martin & Barroeta.8 The main reaction pathways are presented in Figure 4, with major products in boxes and intermediate products, which continue to form a variety of other minor products, in dashed boxes. Pathway (a), homolytic scission of a C-S bond, accounts for an appreciable amount (20%) of the overall sulfide decomposition, and it provides most of the predicted isobutene production. This reaction occurs much quicker than analogous bond-scissions of other hydrocarbons due to the weakness of the C-S single-bond; this one is particularly fast due to the production of a tertiary t-butyl radical in addition to the thiyl compound. Pathways (b) and (c) provide the majority of the main product, isobutene. Pathway (b), which starts with the molecular elimination reaction that directly forms t-butanethiol and isobutene, is predicted to account for 66% of sulfide conversion. This is a sensible result based on the experimental data, as this reaction would explain the overall first-order consumption of t-butyl sulfide observed in the presence and absence of cyclohexene. Much of the thiol undergoes abstraction of the hydrogen bonded to the sulfur to form a thiyl radical. This radical is also produced in smaller amounts through pathway (c), which requires hydrogen abstraction from one of the six methyl groups on t-butyl sulfide, prior to a beta-scission reaction that also forms isobutene. The t-butanethiyl radical abstracts a hydrogen atom from one of the adjacent methyl groups, and the resulting radical species undergoes beta-scission to form the mercapto radical, which then abstracts hydrogen to form hydrogen sulfide. Comparisons between the experimental data and RMG predictions are presented in Figure 5 and Figure 6. These plots show excellent agreement between model and experimental data, as reactant conversion is predicted within 20% of the experimental observation. Isobutene 4SH 33 Chapter 2: Modeling the pyrolysis of t-butyl sulfide production is predicted with slightly greater error, suggesting that the rate of production and recombination of 2-methylallyl radicals from isobutene is slightly overpredicted. ZS R (a) (b) 11% -W + -- (c) 11% 66% 20% 20% RH I'l -) S RH 33% - 3% RR S S- SH H 4% 17% R RH 64% SH 4% 17%R RH 64% RY Rt- R Figure 4. Major reaction pathways for neat pyrolysis of t-butyl sulfide. Percentages represent proportion of reacted sulfide proceeding through a pathway over 40 minutes. 34 Chapter 2: Modeling the pyrolysis of t-butyl sulfide ) - 80 40 - 0 U 20 - - ,-,60 S 0 0 40 30 20 10 Time (min) *1 100 - 100 - Figure 5. Experimental 8 and simulated results for conversion of neat t-butyl sulfide at 380 *C. 0 ) 0( 0 10 - 10 H 2S 0 1 0 - S 0 1 10 30 20 40 -t I 0 10 Time (min) 40 30 - 100 - 100 20 Time (min) S 0 0 .......... I............ - 10 - 10 0 0 +SH 1 1 0 10 20 Time (min) 30 40 0 10 20 30 40 Time (min) Figure 6. Experimental 8 and simulated results for major products of neat t-butyl sulfide pyrolysis, presented in logarithmic scale as a percentage of initial reactant concentration. Chapter 2: Modeling the pyrolysis of t-butyl sulfide 35 2.4.3 Pyrolysis of t-Butyl Sulfide with Cyclohexene A mechanism was generated for the pyrolysis of a 40:60 mixture of t-butyl sulfide and cyclohexene (by mole), using the same termination criteria. The resulting CHEMKIN file from the converged run includes 69 species and 392 reactions, and the predicted overall fluxes over 40 minutes are presented in Figure 7. The same three major pathways are predicted as in the pyrolysis mechanism in the absence of cyclohexene. Again, pathway (b) dominates the sulfide consumption mechanism, so we would expect this mechanism to follow the same first-order kinetics (with roughly the same rate) as in the absence of the inhibitor, and this is what was observed experimentally. LI S R RH (b) (a) R C) 33% 2% 75% 22% RH 2% S. + 8% 25% 32% S RH -R*U R R 5% RH -- SH 3% H* + SH RH-N 11- 5% c RH R 0 RH-> 3% R 32% RH Figure 7. Major reaction pathways for pyrolysis of t-butyl sulfide in the presence of cyclohexene. Percentages represent proportion of reacted sulfide proceeding through a pathway over 40 minutes. 36 Chapter 2: Modeling the pyrolysis of t-butyl sulfide The main difference between the predicted reaction mechanisms for the two cases is seen in the relative rate of pathway (c) of Figure 7, in comparison with the pyrolysis case. While this radical is produced at a similar rate as the pyrolysis case, cyclohexene readily donates hydrogen atoms to reverse this reaction, causing the overall flux to be much lower. Cyclohexene also provides hydrogen in the production of H2S and isobutane, resulting in a resonance-stabilized radical, which can further donate hydrogen atoms to eventually form a small amount of benzene. These compounds will also donate hydrogen atoms to the thiyl radical produced through pathway (a), substantially decreasing the net flux from the thiol to the thiyl radical. Due to this decreased consumption rate of the thiol, this compound should have a concentration roughly equal to isobutene, as was observed experimentally. A comparison of the experimental and predicted sulfide conversion is presented in Figure 8, and the product predictions are compared with experiments in Figure 9. Conversion is predicted with good accuracy, with the same small error as in the neat pyrolysis case. Major products are also predicted reasonably accurately. The production of hydrogen sulfide is overpredicted with approximately the same error as the underprediction of t-butanethiol, suggesting a slight error in the t-butanethiol desulfurization pathway. 80 _ - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 80 0 - 40 20 - 0 S 0 0 10 20 30 40 Time (min) Figure 8. Experimental8 and simulated results for conversion of t-butyl sulfide in the presence of cyclohexene at 380 *C. Chapter 2: Modeling the pyrolysis of t-butyl sulfide Chapter 2: Modeling the pyrolysis of t-butyl sulfide 37 37 100 S 0 -SH 10 1 0 40 30 20 10 Time (min) 100 100 S 0 10 0 0 - :1< 0 10 - S 0~ I H 2S I 0 10 20 30 Time (min) 40 0 10 20 30 40 Time (min) Figure 9. Experimental and simulated results for products of t-butyl sulfide pyrolysis in the presence of cyclohexene, presented in logarithmic scale as a percentage of initial sulfide concentration. 2.4.4 Mechanism Comparison As seen in the previous section and in the experimental work, the addition of a cyclohexene inhibitor has little effect on the overall sulfide decomposition rate, as the dominating reaction for this process is the unimolecular decomposition reaction to form isobutene and t-butanethiol. This is further emphasized using sensitivity analysis, which shows this reaction as the most sensitive for sulfide decomposition by a wide margin. However, the presence of cyclohexene and the radicals subsequently produced has been shown to have a significant effect on the resulting product distributions. This can be seen in Figure 10, where major radical species concentrations are plotted for the two cases. While the total concentration of radical compounds is similar in the two cases, the presence of cyclohexene suppresses the concentration of radicals other than 38 Chapter 2: Modeling the pyrolysis of t-butyl sulfide cyclohexenyl. These stable radicals will abstract hydrogen more slowly than others, leading to a slower radical-decomposition of intermediate products, primarily the thiol. 1.E-05 - 1.E-07 0 4a-J L 1.E-09 i.E-13 0.00001 0.001 0.1 10 1000 Time (s) Figure 10. Radicals produced in pyrolysis of t-butyl disulfide, in the presence (gray) and absence (black) of cyclohexene. Isobutenyl (solid), t-butanethiyl (dashed), mercapto (dotted), and cyclohexenyl (double). 2.5 Conclusions In this work, we have demonstrated the ability of automated mechanism generation software to propose and validate mechanisms for organosulfur pyrolysis mechanisms, where only postulations based on bulk experimental data were available in the past. In particular, RMG was able to identify the most important reaction controlling the rate of t-butyl sulfide in the presence and absence of a compound expected to inhibit the reaction rate. Coupled-cluster calculations suggested that the unimolecular decomposition pathway could fully account for the decomposition mechanism of the disulfide and the first step of the sulfide mechanism. RMG was used to elucidate the reaction mechanism for t-butyl sulfide, demonstrating the free radical mechanism for the decomposition of t-butane thiol and the lack of this decomposition in the presence of the cyclohexene inhibitor. Chapter 2: Modeling the pyrolysis of t-butyl sulfide 39 References 1. S. Patai and Z. Rappoport, The Chemistry of sulphur-containing functional groups, Wiley, New York, 1993. C. Song, Catalysis Today, 2003, 86, 211-263. D. T. Johnston, Earth-Science Reviews, 2011, 106, 161-1,83. V. Mohnen, Sci. Am., 1988, 259. N. A. Fishel, R. K. Lee and F. C. Wilhelm, Environmental Science & Technology, 1974, 8, 260. G. Corro, Reaction Kinetics and Catalysis Letters, 2002, 75, 89-106. M. Bajus, Sulfur Rep., 1989, 9, 25. G. Martin and N. Barroeta, International Journal of Chemical Kinetics, 1980, 12, 699716. L. Xu, J. Yang, Y. Li and Z. Liu, Fuel Process. Technol., 2004, 85, 1013-1024. J. K. Winkler, W. Karow and P. Rademacher, J. Anal. Appl. Pyrolysis, 2002, 62, 123141. M. T. Timko, E. Schmois, P. R. Patwardhan, Y. Kida, C. A. Class, W. H. Green, R. K. Nelson and C. M. Reddy, Energy & Fuels, 2014, 28, 2977-2983. H. Mei, B. W. Mei and T. F. Yen, Fuel, 2003, 82, 405-414. A. R. Katritzky, M. Balasubramanian and M. Siskin, Energy & Fuels, 1992, 6, 431-438. A. R. Katritzky, R. A. Barcock, M. Balasubramanian and J. V. Greenhill, Energy Fuels, 1993, 8, 498-506. Y. Kida, C. A. Class, A. J. Concepcion, M. T. Timko and W. H. Green, Phys Chem Chem Phys, 2014, 16, 9220-9228. P. R. Patwardhan, M. T. Timko, C. A. Class, R. E. Bonomi, Y. Kida, H. H. Hernandez, J. W. Tester and W. H. Green, Energy & Fuels, 2013, 27, 6108-6117. X. Zheng, J. W. Bozzelli, E. M. Fisher, F. C. Gouldin and L. Zhu, Proceedings of the Combustion Institute, 2011, 33, 467-475. W. H. Green, J. W. Allen, R. W. Ashcraft, G. J. Beran, C. A. Class, C. Gao, C. F. Goldsmith, M. R. Harper, A. Jalan, G. R. Magoon, D. M. Matheu, S. S. Merchant, J. D. Mo, S. Petway, S. Raman, S. Sharma, K. M. Van Geem, J. Song, J. Wen, R. H. West, A. Wong, H.-W. Wong, P. E. Yelvington and J. Yu, Reaction Mechanism Generator (RMG), (2013). A. S. Tomlin, T. Turitnyi and M. J. Pilling, in Comprehensive Chemical Kinetics, ed. M. J. Pilling, Elsevier, 1997, vol. 35, pp. 293-437. M. R. Harper, K. M. Van Geem, S. P. Pyl, G. B. Marin and W. H. Green, Combust. Flame, 2011, 158, 16-41. J. W. Allen, A. M. Scheer, C. W. Gao, S. S. Merchant, S. S. Vasu, 0. Welz, J. D. Savee, D. L. Osborn, C. Lee, S. Vranckx, Z. Wang, F. Qi, R. X. Fernandes, W. H. Green, M. Z. Hadi and C. A. Taatjes, Combust. Flame, 2014, 161, 711-724. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, Theor Chem Account, 2009, 123, 391-412. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Phys Chem Chem Phys, 2012, 14, 12773-12793. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Chemistry-A European Journal, 2011, 17, 7656-7673. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. & 2.6 Chapter 2: Modeling the pyrolysis of t-butyl sulfide 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40 G. Martin, in The chemistry of sulphur-containing functional groups, John Wiley & Sons, New York, 1993, pp. 395-437. S. W. Benson and G. R. Haugen, J. Am. Chem. Soc., 1965, 87. CHEMKIN-PRO 10131, (2013) Reaction Design, San Diego. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, ChemPhysChem, 2013, 14, 37513771. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers, V. Van Speybroeck, M. Waroquier and G. B. Marin, J. Phys. Chem. A, 2007, 111, 11771-11786. A. G. Vandeputte, University of Ghent, 2012. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, ChemPhysChem, 2013, 14, 17031722. J. Aguilera-Iparraguirre, A. D. Boese, W. Klopper and B. Ruscic, Chemical Physics, 2008, 346, 56-68. J. Aguilera-Iparraguirre, H. J. Curran, W. Klopper and J. M. Simmie, Journal of Physical Chemistry A, 2008, 112, 7047-7054. W. Klopper, R. A. Bachorz, D. P. Tew, J. Aguilera-Iparraguirre, Y. Carissan and C. Hattig, Journal of Physical Chemistry A, 2009, 113, 11679-11684. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, 1. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and M. Schiitz, Wiley Interdisciplinary Reviews: Computational Molecular Science, 2011, 2, 242-253. S. Sharma, M. R. Harper and W. H. Green, CanTherm open-source software package, (2010). C. Eckart, Physical Review, 1930, 35, 1303-1309. G. Martin and N. Barroeta, J. Chem. Soc. Perkin Trans. 2, 1976, 1421-1424. Chapter 3: A kinetic database for organic sulfur and oxygen compounds 41 Chapter 3: A kinetic database for organic sulfur and oxygen compounds 3.1 Abstract Potential energy surfaces and reaction kinetics were calculated for 40 gas-phase reactions involving sulfur and oxygen. This includes 11 H 2 0 addition, 8 H2 S addition, 11 hydrogen abstraction, 7 beta scission, and 3 elementary tautomerization reactions, which are potentially relevant in the combustion and desulfurization of sulfur compounds found in various fuel sources. Geometry optimizations and frequencies were calculated for reactants and transition states using B3LYP/CBSB7, and potential energies were calculated using CBS-QB3 and CCSD(T)-Fl2a/VTZ-F12. Rate coefficients were calculated using conventional transition state theory, with corrections for internal rotations and tunneling. Additionally, thermochemical parameters were calculated for each of the compounds involved in these reactions. With few exceptions, rate parameters calculated using the two potential energy methods agreed reasonably, with calculated activation energies differing by less than 5 kJ/mol. The computed rate coefficients and thermochemical parameters are expected to be useful for kinetic modeling. Chapter 3: A kinetic database for organic sulfur and oxygen compounds 3.2 42 Introduction Sulfur compounds can be found in almost every aspect of life, and their interactions with oxygenated species play an important role in fuels, geochemistry, and environmental chemistry." 2 The formation of petroleum in geochemical reservoirs may be accelerated by the presence of weak carbon-sulfur bonds, and the reaction mechanisms of these species can be affected by the presence of water.3 5 One of the most important sources of sulfur compounds is crude oil, and these compounds will react to form toxic sulfur dioxide if not removed prior to combustion. The desulfurization of crude oil has become a very important topic of study, as sulfur emission standards have tightened and the availability of sulfur-lean feedstock has lessened.6 The current industry standard, hydrodesulfurization, requires the use of hydrogen and expensive catalyst to achieve the proper sulfur level, so multiple alternatives are being studied to potentially achieve similar results at a lower cost. Oxidative desulfurization converts thiophenic compounds into more easily removable polar compounds using hydrogen peroxide and a catalyst.7 Microbial desulfurization removes sulfur from organic compounds at ambient temperature and pressure. 8 Treating oil with supercritical water accomplishes desulfurization without the requirement of any catalyst.9 Work in supercritical water upgrading has demonstrated that water generates products with reduced sulfur content and molecular weight.' 11 Water's involvement in this process has been explored via model compound experiments, and some investigators have proposed pathways to explain the reactivity of various sulfur compounds in aqueous and supercritical systems.12 Additional experiments and the advancement of computational chemistry techniques have assisted in the elucidation of this mechanism, showing water to be both a reactant and a hydrogen-transfer catalyst in the mechanism of alkyl sulfide desulfurization.9 Based on intermediate studies and quantum chemistry calculations, a plausible pathway for water-aided desulfurization was proposed, and this is shown schematically in Figure 1. In the proposed mechanism, the water prevents the conversion of the reactive thioaldehyde (reactant 3) to an oligomer, which is known to occur in the absence of water.1 3 Water participates by adding to the carbon-sulfur double-bond in reaction c to form reactant 4, which readily reacts at high temperature to form hydrogen sulfide, carbon monoxide, and a smaller alkane. __WE4_z_ - __ - . ", Chapter 3: A kinetic database for organic sulfur and oxygen compounds - 43 0 Co9 CC. a 12 CH S8 * C 5H 6 H14 11 1 'CSH 11 b 0 * C S C 5H 1 QO 1 2 C5 H 11 O d H2 S C 5 H 1< S CH 13 C111 4 OH C C5HII> SH H 2 0) 20 Figure 1. Proposed mechanism 9 for conversion of hexyl sulfide to pentane and CO 2 Many other pathways are possible, and a full kinetic mechanism of the system based on accurate thermochemical and kinetic data is necessary to evaluate and validate them. Extensive libraries of thermochemical data and reaction rate parameters for hydrogen abstraction, beta scission, and substitution reactions involving organosulfur compounds have been generated by Vandeputte et al. 14-16 Rate constants have also been calculated for small-molecule reactions involved in combustion to form SOx compounds.1 7' 18 However, these data are not sufficient for accurately modeling the reactions of thiols, sulfides, and thiophenes with oxygenated species. This work focuses on the reactions of sulfur compounds and other species that are likely to be produced in the presence of water at high temperatures. Many of the reactions considered here could also be relevant to organosulfur combustion systems. Rate parameters in modified Arrhenius form were calculated for 40 reactions that involve organic sulfur and oxygen. These provide rate constants for use in simulations of hydrocarbon mixtures including both sulfur and oxygen, as well as in training sets to develop more general rate estimation rules. Thermochemical parameters, which are required for the calculation of equilibrium constants and temperature changes in reacting systems, have also been computed for each of the species involved in the reactions and compared to the limited data already available. _-AAd Chapter 3: A kinetic database for organic sulfur and oxygen compounds 3.3 44 Methods Thermochemical data were computed using the Gaussian 03 and Molpro quantum chemistry packages.' 9 ',20 All species with an even number of electrons were calculated in their singlet state, and radical compounds were calculated in their doublet states. Geometry optimizations and frequency calculations were conducted using B3LYP/CBSB7 , 2 and it was tested that all the reactants and products were indeed minima on the potential energy surface and that all the transition states showed one and only one imaginary frequency that corresponded to the expected reaction coordinate. These geometries were then used for single point energy calculations at higher levels of theory. Electronic energies were calculated using both the composite CBS-QB3 method22 ' 23 and the explicitly-correlated CCSD(T)-Fl2a/VTZ-F12 method 24 -2 7 (this will be referred from now on as CCSD(T)-F12). The slow convergence of CCSD(T) with the basis set size has been known for a long time.28, That restricted its application to very small systems. 3 0, 31 29 In the last few years explicitly- correlated methodologies have been introduced to circumvent this problem. 2 ,' 3 They directly address the fact that conventional coupled-cluster methods approximate wavefunctions based on one-electron basis functions and can hardly describe the electron-electron correlation. This drawback was overcome with the introduction of functions depending explicitly on the interelectronic distance, as used in the CCSD(T)-F12 family. That makes the basis set convergence much faster and allows us to describe medium-sized systems with basis-set error of less than I kcal/mol. These properties have allowed it to be successfully applied in all sorts of fields, 34 3 6 including thermochemistry and kinetics. CBS-QB3 has previously been used in a variety of kinetic studies, including some relevant to sulfur chemistry, and the reaction barriers calculated have been shown to have an uncertainty of a few kcal/mol. 2 2 , 23, 37, 38 CBS-QB3 thermochemistry is usually more accurate due to the availability of empirical Bond Additivity Corrections (BAC). 3 9 It appears that CBS-QB3 is becoming obsolete, as new density functionals like M06 and BMK provide comparable accuracy at a much lower cost, 4 0' 41 and CCSD(T)-F12 methods provide improved accuracy at a still- reasonable cost. We include it here nevertheless since a big part of the data available in the literature from the last two decades has been calculated in this way, so a good assessment of its accuracy is still useful. Chapter 3: A kinetic database for organic sulfur and oxygen compounds 45 Partition functions were calculated using the CanTherm software package,42 using a scaling factor of 0.99 for the frequency analysis. Enthalpies, entropies, and heat capacities were calculated using CanTherm, including CBS-QB3 BAC's that are available in literature, 3 9 No correction was available for the C=S bond due to the scarcity of experimental data for thiocarbonyl compounds. Calculated parameters were used to generate NASA polynomials for each of the reactants and products. These calculations were used to extend the group additivity scheme for thermochemical properties, which was originally proposed by Benson and Buss, and extended by Vandeputte et al. using CBS-QB3 for compounds containing sulfur.' 6 ' 43, 44 Using the thermochemical parameters calculated in this work, group additivity values (GAV's) of enthalpy and entropy of formation, and heat capacities between 300 and 1500 K for 15 groups containing both sulfur and oxygen were derived using the regression method discussed by Vandeputte et al.' 6 Hydrogen Bond Increments (HBI's), as defined by Lay et al. , were derived for two radical groups including sulfur and oxygen. The values for groups with previously calculated GAV's (i.e. those that do not contain all of sulfur, carbon, and oxygen) were held constant at the literature values. Transition states were optimized for each elementary reaction, and transition state theory in CanTherm was used to calculate rate coefficients under the ideal gas assumption, correcting for the internal rotations of each single bond within the reactants, products, and transition states. One-dimensional hindered rotations were used in the analysis, optimizing the geometries at the B3LYP/6-31 1G(2d,p) level at 20-degree increments for each rotatable bond. Asymmetric Eckart tunneling corrections were also calculated, and these corrections were applied to generate the reaction rate constants between 300 and 2000 K.46' 47 Rate constants were fitted to the modified Arrhenius form, -Ea k(T) = A * T' * eR*T, where T is the temperature in Kelvin, R is the gas constant, A and n are fitted constants, and Ea is the fitted activation energy. It is important to note that the fitted Ea is not the same as the reaction energy barrier AEa, the calculated energy difference between the reactants and transition state including zero-point energies (ZPE's). E, and AE, can differ by multiple kJ/mol. The modified WAWMMM"" .I--_ ................. - --. 1-111111111 Chapter 3: A kinetic database for organic sulfur and oxygen compounds 46 Arrhenius form has been demonstrated to fit rate constants for a variety of organic systems better than the standard Arrhenius form without the r term.48 '49 These rate parameters were calculated assuming reactant activity coefficients a= 1. The activity for water can vary significantly at supercritical conditions: for example, at the conditions of Kida et al., 9 the activity coefficient of water is calculated to be approximately 0.5, reducing the rates by this factor when water is a reactant. Thus, the rate parameters in this work should be adjusted to account for the conditions being modeled to avoid introducing additional errors. 3.3.1 Calculation of Rate Constants for Reactions with Submerged Transition States The reaction barrier was calculated to be significantly negative (i.e. greater than the uncertainty of the calculations) for two of the reactions studied in this work, implying the existence of reactive complexes at lower energy levels than the reactants of the respective reactions. This type of "submerged-barrier" reaction is illustrated in Figure 2. The same methods as discussed previously for reactants and products were used to calculate energies and frequencies for the reactive complex of each reaction. k1 Reactantsk2O Lk. Complex Products Reaction Coordinate--Figure 2. Potential energy surface for a generic reaction with a submerged transition state. The parameters for each submerged reaction were calculated for the high-pressure limit using CanTherm. The rate k, for the formation of complex was assumed to be the collision rate, 1013 cm3/(mol* s), and k-I was calculated using thermochemical consistency. The rate of formation of Chapter 3: A kinetic database for organic sulfur and oxygen compounds 47 products from the pre-reactive complex, k 2, was calculated using transition-state theory. The complex is short-lived, so it can be modeled using the quasi-steady-state approximation. The overall rate of product formation for a reaction with two reactants is therefore dC p k1k 2 dt k 1 + k2 1 R2 and the effective rate constant is keff(T) _k =k 1 1 k +2 k2 The effective rate constant keg(T) was calculated at temperatures between 300 and 2000 K, and modified Arrhenius parameters were fit to these calculations to obtain the values reported in the Tables for Reactions 21 and 37. As our primary interest is in supercritical water reactions (with pressures greater than 200 bar), rate constants are reported in the high pressure limit. In some gas-phase situations, the low-pressure limit might be more appropriate than the high-pressure limit values reported here. 3.3.2 Basis Set benchmarking for CCSD(T)-F12 Calculations A procedure to establish the accuracy of the basis set in the Coupled-Cluster calculations for our particular set of reactions was defined as follows: in each of the different classes of reactions, the one with the smallest number of electrons was used. These are reactions 1 for the molecular additions of water, 12 for the molecular additions of hydrogen sulfide, 20 for the hydrogen abstractions and 31 for the beta-scissions. The restriction on the size of the reactions allows us to perform calculations on a bigger basis that would it be practical otherwise, and use its results as a benchmark. We obtained both reactants and transitions states for our set of reactions and performed a consistent set of CCSD(T)-F12a calculations with the basis set series VDZ-F 12, VTZ-F12, and VQZ-F 12.50 CCSD(T)-F12b energies were also calculated with the VQZ-F12 basis set, and these agreed with the F 12a energies for the same basis set with an average error of 0.17 kJ/mol. The convergence with respect to basis set is shown on Table 1. Triple-zeta F12 barrier heights are converged to better than 1 kJ/mol. Double-zeta basis set on the other hand lead to errors above 1 kJ/mol. This is in good agreement with previous studies.5l In a compromise between 48 Chapter 3: A kinetic database for organic sulfur and oxygen compounds accuracy and computational cost, we chose VTZ-F 12 basis set as the standard for this study. It is important nevertheless to be aware of the error introduced by such a choice. Table 1. Mean absolute difference in barrierheight (kJ/mol) calculated using double-, triple-, and quadruplezeta basis sets with CCSD(T)-F12a. # Reaction |DZ-QZ| ITZ-QZ| 1 1.21 0.01 12 20 31 Average 0.31 0.57 1.57 0.34 1.67 0.16 1.19 0.27 While the calculations reported here are converged with respect to basis set, this does not mean they are exact. CCSD(T) is not full-CI, and there are several small neglected terms (BornOppenheimer breakdown, relativistic, anharmonicity) which can contribute errors on the order of kJ/mol. Still, we expect that the numbers computed here are rather close to the true energies. 3.4 Results and Discussion 3.4.1 Molecular Addition of Water (Hydration of Double Bonds) Reaction coefficients calculated for the ten reactions involving the molecular addition of water to double bonds are presented in Table 2. These reactions progress via a four-membered ring transition state, as depicted in Figure 3 for Reactions 1-9. Transition states for reactions 3, 8 and 9 were previously calculated.9, 52 All the other geometries were determined in this work and are reported in the Supporting Information. Those geometries were used in this study. Reaction 8 corresponds to the addition of water to the thiocarbonyl group of carbonyl sulfide, while reaction 9 is addition to the carbonyl group. Reactions 10 and 11 are for the addition of water to the carbon-carbon double-bond of thiophene, which also occurs through a four-membered ring transition state. S ---- H o R,S + 0I 2 ~L H | HO |2 H Figure 3. H 2 0 Addition to thiocarbonyl compound. SH Chapter 3: A kinetic database for organic sulfur and oxygen compounds 49 The transition state of reaction I is presented in Figure 4. Calculated reaction parameters for the molecular elimination of water from methanediol and ethanol are available in literature, and using available thennochemistry data we can estimate the activation energy of these reactions in the addition direction. 3 ,5 4 These are compared with the activation energies of reaction 1 in Table 3. The instability of thiocarbonyl compounds, which are known to polymerize at room temperature, provides for a low-energy pathway for the conversion of this type of compound.' 3 Table 3 shows that the activation energy in both directions is lowest for the thiocarbonyl case, as the 4-center reaction is much more facile for sulfur-containing systems than for C/H/O systems. Lower A-factors and higher n-factors are calculated when a methyl or ethyl group is substituted, as in reactions 2 and 3. This leads to rate constants that are within an order of magnitude at most temperatures. Reactions 2 and 3 have very similar Arrhenius constant and n-factor, and a difference of 4 kJ/mol for the activation energy in both directions, which suggests that increasing the length of the thiocarbonyl compound has a minor effect, and will likely have a lesser effect as this chain length increases. Reaction 4 has similar activation energies but a lower A-factor than reactions 1-3 due to the presence of a methyl group on both sides of the thiocarbonyl group. We compute slightly lower A-factors for the addition of water to benzenethial when compared with reactions 1-3, but a significantly higher barrier height is predicted in the forward direction. The transition state geometries for reactions 5 and 6, the addition of water to 2-propenethial and benzenethial, are presented in Figure 5. The lengths of the C-S bonds in the two transition states differ by less than 0.01 A, and this similarity is reflected in the rate constant calculations. CBSQB3 calculations on reaction 5 resulted in a reaction barrier of 145 kJ/mol, which is within 1 kJ/mol of the calculated barrier for hydration of benzenethial. As expected, very similar Arrhenius parameters are calculated for the addition of water to a thioaldehyde bonded to an sp2 carbon. I J WIiww --------------------- .1.1 ------------------- 50 Chapter 3: A kinetic database for organic sulfur and oxygen compounds Table 2. Modified Arrhenius coefficients for the molecular addition of water to sulfur-containing compounds. A [cm 3 /(mol*s)i, n (unitless), E., AE, and AH*,,, (kJ/mol). Parameters for reaction 6 computed using CCSD(T)-F12//B3LYP. Forward Rate Parameters Reaction + - 1. SH i0HO 11 log 10A n -0.62 AE 0 3.55 E1 101.8 122.7 AH*rxn -54.1 -2.42 3.96 102.7 123.5 -48.1 -2.58 3.95 101.3 121.6 -46.7 -4.30 4.54 101.7 125.0 -46.9 -1.22 3.75 122.8 140.9 -28.4 -1.78 3.89 122.2 143.5 -29.6 -4.98 4.64 135.4 160.2 -8.3 -6.36 5.40 188.7 187.8 13.8 3.77 2.47 257.7 230.4 35.6 -26.0 11.0 191.9 269.3 33.9 -11.1 6.87 222.9 264.4 37.0 SH 8 + 10 2. H SH 1120 S ------ S 4. 12 - + 1120 OH + - 11- -- i"............ , S S 5. H SH Ph Ph S+ 1120 H SH - H2 0 7. OH H H 0 f l() 8. ~ S 0 9. OH HS cO 1-120 C// OH HO S + 10. S 11. OH H2 0 S + 11()0 H 51 Chapter 3: A kinetic database for organic sulfur and oxygen compounds 1.784 1.4 1.572- Figure 4. Optimized transition state for the hydration of thioformaldehyde. Distances (Angstroms). Table 3. Reaction barriers (kJ/mol) for hydration of thioformaldehyde, formaledehyde, and ethene. Reaction H 2C I + H2)0 S 0 + H2C CH2 SH HO OH H-,O HO H0 O + H 2C HO AE0, AE or Ref. 123 173 this work 166 189 Kent [50] 209 254 Li [49] 1. 8 1.728 1 .747 A1.605 1.165 Figure 5. Transition states for the hydration of 2-propenethial (left) and benzenethial (right). Distances (Angstroms). Reactions 8 and 9 correspond to the addition of water to carbonyl sulfide, as investigated by Deng et al.,5 and transition state geometries for these reactions (as calculated in literature) are presented in Figure 6. The barrier height is calculated to be 43 kJ/mol greater for the addition to the C=O bond than when water attacks the C=S bond. Comparing reaction 8 with the other hydration reactions with thiocarbonyl groups, we see that addition to carbonyl sulfide requires an activation energy more than 80 kJ/mol greater than reactions 1 through 4. 52 Chapter 3: A kinetic database for organic sulfur and oxygen compounds _L256 1,45 1 1. 15 1.128 I 1.152 1;641 09 1. 2 'V Figure 6. Transition states for the hydration of carbonyl (left) and thiocarbonyl (right) group of carbonyl sulfide. Distances (Angstroms). Because of the aromaticity of thiophene, reactions 10 and 11 are endothermic, in contrast to the exothermic addition of water to C-C double-bonds in alkenes. As such, these reactions proceed via much higher-energy pathways, and the parameters calculated in this study show that water will not appreciably react directly with thiophene at temperatures below 1500 K. 3.4.2 Molecular Addition of Hydrogen Sulfide Reaction coefficients for eight reactions involving the addition of H 2 S to a carbonyl bond are presented in Table 4. The optimized transition states for reactions 14 and 19 were available in literature.: The other geometries were determined in this work and are reported in the Supporting Information. This type of reaction progresses in a similar fashion as the molecular addition of water to a thiocarbonyl compound, via a four-member ring transition state as shown in Figure 7. 0 ---- H SH HO 0| R/ + H2S R H Figure 7. H 2 S addition to a carbonyl compound. 53 Chapter 3: A kinetic database for organic sulfur and oxygen compounds Table 4. Modified Arrhenius coefficients for the molecular addition of hydrogen sulfide to carbonyl compounds. A Icm 3 /(mol*s)I, n (unitless), E, AE,, and A H*,,,, (kJ/mol). Parameters for reaction 6 computed using CCSD(T)-F12//B3LYP. Forward Rate Parameters Reaction Reaction 12. H2S + SH HO - 156.7 AE, 170.0 AH*rxn -50.1 2.93 153.5 161.8 -34.1 1.49 2.96 152.0 159.9 -32.9 0.22 3.45 158.0 168.7 -26.1 2.58 2.72 151.8 159.1 -22.0 2.09 2.83 145.1 152.1 -20.5 -0.68 3.60 159.7 170.4 31.9 -9.20 6.38 199.1 204.0 40.7 logjOA n Ea 1.09 3.27 1.78 SH HS + " - 13. H SH + 14. Ii's SH 15. 0 - --- H2 OH + 15. .SH 1-12S + 16. H SH Ph 0 + Ph H2 S - * 17. H SH 112S OH - + 18. - H 0 1 19. C// + 1's HS I OH This reaction occurs via a four-membered transition state, as in the addition of water to a doublebond, but the bond lengths and angles are greatly different. This is shown in Figure 8, which shows the optimized transition state for reaction 12. An IRC scan confirmed that this transition state corresponded to the expected reaction, and the potential energy surface was scanned using stepping the C-S and 0-H bond distances while optimizing the remaining variables. This is presented in Figure 9, and it shows that the reaction happens in a somewhat sequential fashion, with the translation of the hydrogen atom to form an OH group largely complete while the carbon and sulfur atoms are still separated by a distance of 2.7 A (in b3lyp/6-311G(2d,p), comparison with the final C-S bond length of 1.8 A). Thus, we expect that a separate disproportionation pathway exists with a similar transition state, although the addition reaction's transition state is over 100 kJ/mol more stable than the sum of the CH 2OH and SH radicals that would be the intermediates of a disproportion-recombination pathway. This reaction type is also Chapter 3: A kinetic database for organic sulfur and oxygen compounds 54 a likely candidate for a roaming radical pathway, which has previously been investigated for the decomposition of formaldehyde. 5 ' 56 In addition, investigating the possibility of reaction pathway bifurcation5 7 may be an area of future research for this type of reaction system. The carbon-sulfur distance in Figure 8 is calculated to be 46% greater in the transition state than the bond length in the product compound (compared to only a 12% difference for the carbonoxygen distance in reaction 1). This is reflected in the general trend of activation energies, where the addition of water to a thioaldehyde is calculated to be a significantly more favorable reaction than the addition of H2 S to an aldehyde. t.84 1. i 9Q9 2 2.728 2 Figure 8. Transition state for the molecular addition of H2 S to formaldehyde. Distances (Angstroms). Similarly to the case with the addition of water to a thiocarbonyl compound, the reaction barrier in both directions is slightly lower when an alkyl group is substituted on the carbonyl compound, as shown by reactions 13 and 14. The transition states for these two reactions are presented in Figure 10. Again, this effect decreases as the chain length increases, so the calculated rate parameters for reaction 14 should be acceptable approximations for the addition of H 2 S to a longer aldehyde. Substituting an alkyl group on both sides of the carbonyl group leads to the prediction of a lower Arrhenius constant and greater n-factor is predicted for reaction 15. Chapter 3: A kinetic database for organic sulfur and oxygen compounds H SH + CHOH S+CHO 220 - 3.6 55 200 3.4180 160 3 140 -- 28 0 (I) 120 H--S 2-6H 100 6 2-4 80 22 60 - 1 1.2 1.4 40 - 2 2 18 16 0-H Distance (Angstrom) 2.2 2.4 Figure 9. Potential energy surface for Reaction 12. Energies (kJ/mol) relative to the mercaptoalcohol. 1.053 .. 7 -- 12 - 277-7- 2.762 Figure 10. Transition states for the molecular addition of H 2S to acetaldehyde (left) and 2-propanone (right). Distances (Angstroms). Substitution of a phenyl group stabilizes the transition state of this reaction. In contrast to hydration reactions 5 and 6 which had very similar Arrhenius parameters, the energy barrier for reaction 17 is calculated to be 7 kJ/mol lower than that calculated using CBS-QB3 for reaction 16. However, the rate constants estimated using these parameters agree within a factor of two at temperatures above 600 K, and the disagreement will decrease at higher temperatures. The optimized transition states of reactions 18 and 19 are presented in Figure 11. These are the only ones in Table 4 calculated to be endothermic in the addition direction, as these require addition to a stable carboxylic acid or carbon dioxide. The activation energies of these reactions are calculated to be the greatest of the reactions calculated in the addition direction, but the lowest in the H2S elimination direction. These transition states have the shortest carbon-sulfur Chapter 3: A kinetic database for organic sulfur and oxygen compounds 56 distance of any calculated for this type of reaction, and this length is 14% less for the addition of H2 S to CO 2 than for the addition to acetic acid. 10681 1~~~ 919.67 1.69 26982311' Figure 11. Transition states for the endothermic addition of H 2 S to acetic acid (top) and carbon dioxide. Distances (Angstroms). 3.4.3 Hydrogen Abstraction Reactions Hydrogen abstraction reactions proceed when a hydrogen atom is abstracted by a radical species, as shown in Figure 12. Modified Arrhenius parameters for the 10 hydrogen abstraction reactions calculated in this work are presented in Table 5. $t H / + R2 R,----H----R 2 H R1 + R/ Figure 12. A hydrogen abstraction reaction. Reactions 20-25 show the abstraction of hydrogen from a sulfur compound by an oxygen radical center. The first four reactions are favored in the forward direction, due to the much greater hydrogen-affinity of an oxygen atom relative to the sulfur atom. Linear transition states were found for most of these reactions, which is typical for hydrogen abstractions. However, linear and nonlinear transition states were found for reactions 20, and these are presented in Figure 13. IRC scans were conducted for the converged geometries, and they showed that both versions of each transition state corresponded to the correct reaction. Lower potential energies were calculated using the bent transition state, so this geometry was used to calculate rate parameters for this reaction. 57 Chapter 3: A kinetic database for organic sulfur and oxygen compounds Table 5. Modified Arrhenius coefficients for hydrogen abstraction reactions. Reaction 21 calculated for overall pathway including pre-reactive complex at high-pressure limit. A [cm 3 /(mol*s)], n (unitless), Ea, AE,, and AH*, ,, (kJ/mol). Forward Rate Parameters Reaction Ea AE 0 7.80 -2.8 4.4 -114.1 S+ H20 13.0 0.03 1.9 -12.3 -136.7 OH 4.32 2.44 5.0 14.9 -56.3 6.12 2.09 -2.0 -1.0 -78.9 -3.14 4.73 52.4 77.0 22.5 -4.59 5.08 54.1 79.4 -0.10 4.08 2.90 0.74 6.8 -9.8 -3.71 4.68 26.4 44.9 -19.4 SH + H 20. HbS + *0H1 21. SH 22. 1 2S + + sH 23. SH + I 24. H2S + 25. -SH + OH S + ( H + + OH OH + 26. + SH 27. + K SH + f2S SH 0 OH OH 28. + *CH3 + *SH 29. N CH4 -0.29 3.74 15.6 35.2 -55.2 + H2 S 5.04 2.47 3.1 18.1 -0.17 0.13 3.51 -3.6 4.9 -59.3 H SH +S o' H+ OH OH 30. AH*0 n 1.71 1og, A SH + 'CH 3 0___ + CH4 Reaction 20 has previously been studied in experimental 62 and theoretical 63 investigations. The rate constants estimated in this work are compared with experimental data in Figure 13. Although our TST calculation do not capture the negative temperature dependence at very low temperatures (below 300 K), all of the methods employed in this work come within about 20% of the experimental data at the temperatures relevant to combustion and pyrolysis. 58 Chapter 3: A kinetic database for organic sulfur and oxygen compounds 0 1458 IL~ 3 14340. L434 3 1385 1. 1.621 1. 5 Figure 13. Linear (left) and angled (right) transition states optimized for reaction 20. Distances (Angstroms) and angle (degrees). - -10.6 -10.8 -11 Q -114 0.5 1 1.5 2 2.5 3 3.5 4 1000/T Figure 14. Comparison of rate constant calculations (cm 3/molecules/s) with experimental data for Reaction 20. Lafage (o), Michael (x), Perry (0), Westenberg (+), Ellingson (hashed): M06-2X (black), MPWB1K (blue), MPW1K (green), BBIK (red), This Work (solid): CBS-QB3 (red), CCSD(T)-Fl2a/VTZ-F12 (black), CCSD(T)-Fl2a/VQZ-F12 (blue) For reaction 21, a valid transition state was only found for the angled geometry. The energy of the transition state for this reaction was calculated to be 12.3 kJ/mol lower than the initial reactants, and a prereactive complex was optimized at an energy 19.4 kJ/mol lower than that of Chapter 3: A kinetic database for organic sulfur and oxygen compounds 59 the reactants, which is illustrated in Figure 14. The rate of reaction 21 approximately equals the collision rate at temperatures above 400 K, and this is reflected in the optimized effective rate parameters (the actual keis calculated are provided in the Appendix). atL TS 1379 -12.3 6 Reactants7 0 Complex -19.4 Products -135.1 0A~ it Figure 15. Potential energy surface for reaction 21. Energies (kJ/mol), distances (Angstroms), angles (degrees). Nearly linear transition state geometries were found for reactions 24 and 25, as both saddle point geometries had an O-H-S angle greater than 170'. Reaction 24 was the only one found to be exothermic in the direction of hydrogen abstraction from the hydroxyl group, while reaction 25 is isothermal (within the margin of error for the calculations). This is in agreement with - published thermochemistry data, from which standard enthalpies of reaction are estimated to be 6 18.2 and -4.3 kJ/mol for reactions 24 and 25, respectively.1 '64-67 Reactions 26 and 27 represent the abstraction of an aldehydic hydrogen by a sulfur-containing radical. Low activation energies are calculated for reaction 26 in both directions, while abstraction of the hydrogen of the carbon adjacent to a thiol group is found to be significantly less favorable. However, this activation energy is 27 kJ/mol lower than that calculated for the abstraction of hydrogen from propane by acetyl radical to form isopropyl radical and 60 Chapter 3: A kinetic database for organic sulfur and oxygen compounds acetaldehyde 68 , as the alpha radical in a thiol or sulfide is stabilized by the presence of sulfur. These two values are compared in Table 6. Table 6. Forward reaction barriers (kJ/mol) for hydrogen abstraction reactions by the acetyl radical. Reaction SH + + O " SH + AEO Ref. 44.9 this work 67.9 Tsang [55] Reactions 28-30 were calculated as possible intermediate steps in the desulfurization of alkyl sulfides and thiols in supercritical water. Reactions 28 and 29 show significantly lower activation energies than generally observed for the abstraction of a hydrogen from tetravalent carbon, as the resulting radical is stabilized by the neighboring sulfur and oxygen atoms. Reaction 30 is highly exothermic, and the radical formed in this reaction is stabilized by the carbonyl group. A negative activation energy was fit to this reaction, but the ZE,) is positive and the positive relationship between temperature and rate constant is expressed by the n-factor of 3.5. 3.4.4 Radical Addition to Double Bonds (Reverse Beta-Scission) Modified Arrhenius parameters for the seven radical addition reactions calculated in this work are presented in Table 7. These proceed via the pathway presented in Figure 15. f F- R3 R1 R 3 + 2 -R 1 R2- Figure 16. Radical addition to a double bond. R3 R4 61 Chapter 3: A kinetic database for organic sulfur and oxygen compounds Table 7. Modified Arrhenius coefficients for radical addition to double bonds. Reaction 37 calculated for 3 overall pathway including pre-reactive complex at high-pressure limit. A (cm /(mol*s), n (unitless), E,, AE, and AH*,.x (kJ/mol). Reaction 31. HO s + S + F - Forward Rate Parameters i Reaction - S. HO log 10A n Ea AEO AH*rxn 8.45 1.63 11.4 16.0 -132.1 4.36 2.35 23.0 28.5 -99.3 3.22 2.54 16.3 20.7 -96.0 9.30 1.21 -5.3 -0.41 -101.2 9.92 1.23 32.2 38.4 -38.8 6.91 1.68 54.2 59.1 -28.3 13.08 0.00 1.7 -9.3 -41.6 OH 32. HO *CH3 OH 33- HO S + 'C2H Sr OH 34. + OH SH 35. + 36. + - S *CH3 Sr SH 37. + 'SH _ OH Optimized transition states for reactions 31-33 are presented in Figure 16. The reverse of reaction 31, which forms thioformic acid and a hydrogen atom, is calculated to be significantly less favorable than the beta scission reactions (reverse of 32 and 33) that form the same thioformic acid and alkyl radicals. The transition state of reaction 34 is calculated to have a slightly negative activation energy and barrier height. Chapter 3: A kinetic database for organic sulfur and oxygen compounds 62 LH Wj .' 1. 8 2.056 2.334 SJM 'WW 1. 1 Figure 17. Transition states for beta scission reactions 31 (top left), 32 (top right), and 33 (bottom). Distances (Angstroms). Reverse reactions 35 and 36 form the stable carbonyl sulfide. These are calculated to be significantly less endothermic than reverse reactions 31-34; so as expected, much lower activation energies are calculated in the beta scission direction, while greater activation energies are predicted in the addition direction. A significantly submerged reaction barrier was calculated for reaction 37, and a pre-reactive complex was optimized near the transition state geometry. The potential energy surface of this reaction is presented in Figure 17. The conversion of the pre-reactive complex to form the product is calculated to occur significantly faster than the reverse reaction to reform the reactants at temperatures greater than 400 K: thus, the overall kg is calculated to exhibit very little temperature dependence and remain approximately equal to the collision rate (additional details available in the Appendix). Chapter 3: A kinetic database for organic sulfur and oxygen compounds 63 *r . 5 2.327 TS -9.3 Reactants Complex -14.1 2Product 1. 2.9051 Figure 18. Potential energy surface for reaction 37. Energies (kJ/mol), distances (Angstroms). 3.4.5 Tautomerization of Thiocarboxylic Acids Three elementary tautomerization reactions were calculated in this work, and they are shown in Table 8. These occur via the translation of a hydrogen atom from an alcohol group of a thiocarboxylic acid to the sulfur atom, as shown in Figure 18. R SH S, S OH R R Figure 19. Tautomerization of a thiocarboxylic acid. 0 64 Chapter 3: A kinetic database for organic sulfur and oxygen compounds Table 8. Modified Arrhenius coefficients for elementary tautomerization reactions that include sulfur and oxygen. A (s-1), n (unitless), E,, AE0 , and AH*,,,, (kJ/mol). Forward Rate Parameters Reaction SH S 38. OH H) 0 H" S OH Ea AEO AH*rxn 1.72 3.26 83.6 112.2 -8.6 2.02 3.21 78.0 106.2 -9.9 1.94 3.23 78.9 107.1 -8.6 0 CH 3 S SH 40. C2 H5 n SH 39. CH 3 log 10A OH C 2 H5 0 The three reactions calculated in Table 8 proceed via very similar transition states, as shown in Figure 19. Interatomic distances vary by less than 0.03 A between the saddle point geometries for reactions 38 and 39, and the rate parameters calculated vary only slightly. The transition state is stabilized to some extent by the substitution of an alkyl group, but this only leads to a difference of 6 kJ/mol in the forward barrier height of reactions 39 and 40 in comparison with reaction 38. Reactions 39 and 40 are calculated to have nearly identical Arrhenius parameters, and Figure 19 shows that the relevant interatomic distances for these two reactions are nearly identical. We expect that further increasing of the alkyl chain length should have a negligible effect. Thus, the coefficients calculated for reaction 40 should be acceptable for elementary tautomerization reactions of thiocarboxylic acids containing alkyl chains. Based on the rate coefficients calculated for reaction 40, a thiocarboxylic acid with a C=S bond would have a half-life of less than 0.1 s at temperatures above 500 K. It is recommended to include this pathway in any model where this type of compound is likely to be produced. Chapter 3: A kinetic database for organic sulfur and oxygen compounds 65 1.650 1 2 1. 1.667 66 6 1.354 1 1.369 1.647 1.356 Figure 20. Transition states for reactions 38 (top left), 39 (top right), and 40 (bottom). Distances (Angstroms). 3.4.6 Thermochemical Library Thermochemistry Group Additivity Values (GAV) 43 for the 15 groups calculated in this work using CBS-QB3 are presented in Table 9, and Hydrogen Bond Increments (HBI) 45 for the two radical groups are presented in Table 10. Previous comparisons with a small set of sulfur compounds with experimental thermochemistry showed that these calculations are generally accurate within 4 kJ/mol.16, 38 These groups are primarily relevant to the SCW pyrolysis of sulfides and thiols; they represent a small subset of all possible groups containing carbon, sulfur, and oxygen. Future expansion of this group library will be necessary for modeling more oxidized 69-73 systems, for which more extensive experimental data are available for benchmarking. - In addition, regression of BAC and GAV using CCSD(T)-F12 for organic compounds should provide more accurate estimates for thermochemical parameters, and these calculations are discussed in Chapter 6. 66 Chapter 3: A kinetic database for organic sulfur and oxygen compounds Table 9. GAV for groups containing carbon, sulfur, and oxygen. Groups presented in Benson notation AfH* (kJ/mol), S~in (J/mol/K) C,* (J/mol/K) Benson Group Additivity Values Group C-(O)(S)(H) '. 2 C-(C)(O)(S)(H) CO, AfH- 0 298 K 298K 300K 400K 500K 600K 800K 1000K 1500K -11.58 4.58 7.48 9.54 11.00 11.91 12.85 13.54 14.93 -16.14 8.37 10.32 11.10 11.30 11.30 11.21 11.60 8.34 -11.10 sint C-(C) 2 (O)(S) -11.26 -39.73 8.16 10.15 10.69 10.52 9.74 9.01 C-(0) 2(S)(H) -19.72 -13.26 6.36 8.72 10.13 10.88 11.56 11.91 12.53 C-(C)(O) 2 (S) -21.41 -36.70 6.65 8.43 9.23 9.47 9.43 9.20 8.89 CO-(S)(H) CO-(C)(S) CO-(O)(S) CS-(O)(H) -9.84 -14.02 -11.53 2.85 29.36 8.55 5.51 4.37 4.94 4.48 6.16 5.04 6.70 5.51 6.35 6.09 7.17 5.83 6.99 6.82 8.06 6.29 7.59 8.05 8.79 6.48 7.76 8.99 9.83 6.38 8.18 10.37 -1.32 8.62 3.90 4.17 4.60 5.10 6.08 6.76 7.44 -22.72 2.67 3.08 3.59 3.90 4.03 3.99 3.75 3.23 6.98 5.56 8.05 5.78 8.35 6.31 9.10 6.73 9.48 7.00 9.95 7.60 10.38 7.61 10.65 8.34 11.59 8.52 11.62 9.31 12.26 8.99 12.26 9.77 12.99 9.29 13.25 10.14 CS-(C)(O) CS-(0) 2 0-(CS)(H) O-(CS)(C) S-(CO)(H) S-(CO)(C) -31.38 -14.54 -21.06 -15.33 9.61 30.14 32.08 10.02 35.41 11.11 5.63 5.30 Table 10. HBI for radical groups containing carbon, sulfur, and oxygen. AfH* (kJ/mol), S'i, (J/mol/K) C* (J/mol/K) Hydrogen atom bond increment Group AfH 0 298 K Ce-(C)(O)(S) SO-(CO) 92.10 89.86 Sint 298 K 8.16 -1.26 C 300 K -5.77 -9.75 400 K -5.52 -11.80 500 K 600 K 800 K 1000 K 1500 K -4.98 -13.39 -4.77 -14.85 -5.82 -17.41 -8.12 -19.29 -14.23 -21.42 3.5 Conclusions Rate coefficients and thermochemical parameters were calculated for 40 reactions involving sulfur and oxygen compounds. These have applicability in studies of sulfur chemistry in an environment rich in water or other oxygenated species, such as the reactions of organosulfur compounds in supercritical water reactors or in geological formations where water is present. Although the calculation methods employed in this work are among the most accurate available, rate coefficients calculated using these methods can still have greater than factor-of-2 uncertainty. In situations where more accurate rate parameters are required, experiments (if possible) or calculations using higher-level quantum chemistry methods and improved treatments of anharmonicity 74' 75 should be conducted. However, the parameters calculated in this work Chapter 3: A kinetic database for organic sulfur and oxygen compounds 67 provide a good starting point for the kinetic modeling of organosulfur chemistry in supercritical water. 3.6 References 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. & 7. C. Song, Catalysis Today, 2003, 86, 211-263. D. T. Johnston, Earth-Science Reviews, 2011, 106, 161-183. M. D. Lewan, Geochimica et CosmochimicaActa, 1997, 61, 3691-3723. M. D. Lewan, Nature, 1998, 391, 164-166. M. D. Lewan and S. Roy, Organic Geochemistry, 2011, 42, 31-41. ETA. Retrieved 03/19/2014 from http://www.eia.gov/dnav/pet/pet_pnp_crq_dcu -nusm.htm.. L. A. Gonzalez, P. Kracke, W. H. Green, J. W. Tester, L. M. Shafer and M. T. Timko, Energy & Fuels, 2012, 26, 5164-5176. J. D. V. Hamme, A. Singh and 0. P. Ward, Microbiology and Molecular Biology Reviews, 2003, 67. Y. Kida, C. A. Class, A. J. Concepcion, M. T. Timko and W. H. Green, Phys Chem Chem Phys, 2014, 16, 9220-9228. M. Morimoto, Y. Sugimoto, Y. Saotome, S. Sato and T. Takanohashi, The Journal of SupercriticalFluids, 2010, 55, 223. L.-Q. Zhao, Z.-M. Cheng, Y. Ding, P.-Q. Yuan, S.-X. Lu and W.-K. Yuan, Energy Fuels, 2006, 20, 2067. A. R. Katritzky, R. A. Barcock, M. Balasubramanian and J. V. Greenhill, Energy Fuels, 1993, 8, 498-506. N. J. Cooper, Compr. Org. Funct. Group Transform. 11, 2005, 3, 355-396. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Phys Chem Chem Phys, 2012, 14, 12773-12793. A. G. Vandeputte, University of Ghent, 2012. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Chemistry-A European Journal, 2011, 17, 7656-7673. K. J. Hughes, A. S. Tomlin, V. A. Dupont and M. Pourkashanian, Faraday Discuss., 2001, 2001, 337-352. P. Glarborg, D. Kubel, K. DamJohansen, H. M. Chiang and J. W. Bozzelli, International Journalof Chemical Kinetics, 1996, 28, 773-790. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. lyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. & 1. 2. 3. 4. 5. 6. Chapter 3: A kinetic database for organic sulfur and oxygen compounds 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 68 Liashenko, P. Piskorz, 1. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and M. SchUtz, Wiley InterdisciplinaryReviews: ComputationalMolecular Science, 2011, 2, 242-253. A. D. Becke, Journalof Chemical Physics, 1992, 98, 5648-5652. J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, Journal of Chemical Physics, 1998, 110, 2822-2827. J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, Journal of Chemical Physics, 2000, 112, 6532-6542. T. B. Adler, G. Knizia and H.-J. Werner, Journal of Chemical Physics, 2007, 127, 221106. T. B. Adler, H.-J. Werner and F. R. Manby, Journal of Chemical Physics, 2009, 130, 054106. G. Knizia, T. B. Adler and H.-J. Werner, Journal of Chemical Physics, 2009, 130, 054104. K. A. Peterson, T. B. Adler and H.-J. Werner, Journalof Chemical Physics, 2008, 128, 084102. R. J. Bartlett and M. Musial, Reviews of Modern Physics, 2007, 79, 291-352. K. Raghavachari, G. W. Trucks, J. A. Pople and M. Head-Gordon, Chemical Physics Letters, 1989, 157, 479-483. J. D. Watts, J. Gauss and R. J. Bartlett, Journal of Chemical Physics, 1993, 98, 87188733. J. M. L. Martin, Chemical Physics Letters, 1996, 259, 669-678. W. Klopper and J. Noga, in Chemistry and Physics, ed. J. Rychlewski, Dordrecht, 2003. J. Noga and W. Kutzelnigg, Journalof Chemical Physics, 1994, 101, 7738-7762. J. Aguilera-Iparraguirre, A. D. Boese, W. Klopper and B. Ruscic, Chemical Physics, 2008, 346, 56-68. J. Aguilera-Iparraguirre, H. J. Curran, W. Klopper and J. M. Simmie, Journalof Physical Chemistry A, 2008, 112, 7047-7054. W. Klopper, R. A. Bachorz, D. P. Tew, J. Aguilera-Iparraguirre, Y. Carissan and C. Hdttig, Journalof Physical ChemistryA, 2009, 113, 11679-11684. J. Zheng, Y. Zhao and D. G. Truhlar, J. Chem. Theory Comput., 2009, 5, 808-82 1. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, Theor Chem Account, 2009, 123, 391-412. G. A. Petersson, D. K. Malick, W. G. Wilson, J. W. Ochterski, J. A. Montgomery and M. J. Frisch, Journalof Chemical Physics, 1998, 109, 10570-10579. Y. Zhao and D. G. Truhlar, Accounts of Chemical Research, 2008, 41, 157-167. A. D. Boese and J. M. L. Martin, Journalof Chemical Physics, 2004, 121, 3405-3416. S. Sharma, M. R. Harper and W. H. Green, CanTherm open-source software package, (2010). S. W. Benson and J. H. Buss, Journalof Chemical Physics, 1958, 29, 546-572. S. W. Benson, F. R. Cruickshank, D. M. Golden, G. R. Haugen, H. E. O'Neal, A. S. Rodgers, R. Shaw and R. Walsh, Chemical Reviews, 1969, 69, 279-324. I ... Chapter 3: A kinetic database for organic sulfur and oxygen compounds 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 69 T. H. Lay, J. W. Bozzelli, A. M. Dean and E. R. Ritter, J. Phys. Chem., 1995, 99, 1451414527. H. S. Johnston and J. Heicklen, JournalofPhysical Chemistry, 1962, 66, 532-533. C. Eckart, PhysicalReview, 1930, 35, 1303-1309. J. A. Miller and S. J. Klippenstein, Journalof Physical Chemistry A, 2003, 107, 77837799. J. S. Pilgrim, A. McIlroy and C. A. Taatjes, Journal of Physical Chemistry A, 1997, 101, 1873-1880. F. A. Bischoff, S. Wolfsegger and D. P. Tew, MolecularPhysics, 2009, 107, 963-975. A. Jalan, 1. M. Alecu, R. Meana-Pafieda, J. Aguilera-Iparraguirre, K. R. Yang, S. S. Merchant, D. G. Truhlar and W. H. Green, J. Am. Chem. Soc., 2013, 135, 11100-11114. C. Deng, Q.-G. Li, Y. Ren, N.-B. Wong, S.-Y. Chu and H.-J. Zhu, Journal of ComputationalChemistry, 2007, 29, 466-480. J. Li, A. Kazakov and F. L. Dryer, J. Phys. Chem. A, 2004, 108, 7671-7680. D. R. Kent, S. L. Widicus, G. A. Blake and W. A. Goddard, Journalof Chemical Physics, 2003, 119, 5117-5120. L. B. Harding, Y. Georgievskii and S. J. Klippenstein, J. Phys. Chem. A, 2010, 114, 765777. L. B. Harding and S. J. Klippenstein, J. Phys. Chem. Lett., 2010, 1, 3016-3020. D. H. Ess, S. E. Wheeler, R. G. Iafe, L. Xu, N. Celebi-OlgUm and K. N. Houk, Angew. Chem. Int. Ed., 2008, 47, 7592-7601. C. Lafage, J.-F. Pauwels, M. Carlier and P. Devolder, J. Chem. Soc. Faraday Trans. 2, 1987, 83, 731-739. J. V. Michael, D. F. Nava, W. D. Brobst, R. P. Borkowski and L. J. Stlef, J. Phys. Chem., 1982, 86, 81-84. R. A. Perry, R. Atkinson and J. N. Pitts Jr., J. Chem. Phys., 1976, 64, 3237-3239. G. S. Tyndall and A. R. Ravishankara, InternationalJournalof Chemical Kinetics, 1991, 23, 483-527. A. A. Westenberg and N. deHaas, J. Chem. Phys., 1973, 59, 6685-6686. B. A. Ellingson and D. G. Truhlar, J. Am. Chem. Soc., 2007, 129, 12765-12771. M. W. Chase, Jr., J. Phys. Chem. Ref Data, 1998, Monograph 9, 1-195 1. G. da Silva, C.-H. Kim and J. W. Bozzelli, J. Phys. Chem. A, 2006, 110, 7925-7934. D. P. Tabor, M. E. Harding, T. Ichino and J. F. Stanton, J. Phys. Chem. A, 2012, 116, 7668-7676. W. D. Good, J. L. Lacina and J. P. McCullough, J. Phys. Chem., 1961, 65, 2229-2231. W. Tsang, Journalof Physical and Chemical Reference Data, 1988, 17, 887-952. F. Turecek, L. Brabec, T. Vondrak, V. Hanus, J. Hajicek and Z. Havlas, Collect. Czech. Chem. Commun., 1988, 53, 2140-2158. W. K. Busfield, H. Mackle and P. A. G. O'Hare, Trans. Faraday Soc., 1961, 57, 10541057. H. Mackle and D. V. McNally, Trans. FaradaySoc., 1969, 65, 1738-1741. H. Mackle and P. A. G. O'Hare, Trans. FaradaySoc., 1961, 57, 1070-1074. H. Mackle and W. V. Steele, Trans. FaradaySoc., 1969, 65, 2053-2059. B. C. Garrett and D. G. Truhlar, Journalof Physical Chemistry, 1979, 83, 1915-1924. J. Zheng, T. Yu, E. Papajak, 1. M. Alecu, S. L. Mielke and D. G. Truhlar, Phys Chem Chem Phys, 2011, 13, 10885-10907. 70 Chapter 3: A kinetic database for organic sulfur and oxygen compounds 3.7 Appendix: Calculations for reactions 21 and 37 Rate parameters were calculated for the two reactions with submerged transition states using the method described in the chapter, and the component rate coefficients are presented in the following tables. keg provides the effective rate constant calculated at each temperature, while k/,q shows the rate constant obtained using the best-fit modified Arrhenius parameters. The ratio of the fitted rate constants to the rate constants calculated at each temperature show that low fitting error was obtained for temperatures between 400 and 2000 K, but significantly greater error was obtained at 300 K. Thus, the specific ke(300 K) should be used at this temperature. Rate parameters for Reaction 21 T 300 400 500 600 800 1000 1500 2000 k, 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 k1 1.27E+11 1.69E+10 5.36E+09 2.58E+09 1.11E+09 7.12E+08 4.41E+08 3.80E+08 KI 1.27E-02 1.69E-03 5.36E-04 2.58E-04 1. 11E-04 7.12E-05 4.41E-05 3.80E-05 1.89E+10 9.28E+10 2.42E+11 4.62E+11 1.05E+12 1. 76E+12 3.60E+12 5.26E+12 keff kfit 1. 30E+12 8.46E+12 9.78E+12 9.94E+12 9.99E+12 1.OOE+13 1.OOE+13 1.OOE+13 5.58E+12 6.81E+12 7.68E+12 8.34E+12 9.25E+12 9.86E+12 1.08E+13 1. 13E+13 kfit/keff 4.30 0.81 0.79 0.84 0.93 0.99 1.08 1.13 Rate parameters for Reaction 37 _ k-1 K1 k2 keff kfit kfit/keff 5.1OE+11 7.20E+10 2.33E+10 1.14E+10 4.91E+09 3.14E+09 1.96E+09 1.71E+09 5.10E-02 7.20E-03 2.33E-03 1.14E-03 4.91E-04 3.14E-04 1.96E-04 1.71E-04 2.82E+11 3.96E+11 4.93E+11 5.77E+11 7.19E+11 8.34E+11 1.04E+12 1.18E+12 3.56E+12 8.46E+12 9.55E+12 9.81E+12 9.93E+12 9.96E+12 9.98E+12 9.99E+12 6.23E+12 7.36E+12 8.13E+12 1.75 0.87 0.85 0.89 0.95 1.00 1.06 1.10 - T 300 400 500 600 800 1000 1500 2000 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 1.OOE+13 8.69E+12 9.44E+12 9.93E+12 1.06E+13 1.1OE+13 Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 71 Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 4.1 Abstract A detailed reaction network is proposed for the desulfurization of hexyl sulfide in the presence of supercritical water (SCW) using the automated Reaction Mechanism Generator (RMG). Experimental data have shown that pentane, carbon monoxide and carbon dioxide are products of hexyl sulfide desulfurization in SCW, while none of these are detected in the simple pyrolysis of hexyl sulfide. The observation of CO and CO 2 in the reaction products is a key result as it provides evidence that water is acting as a hydrogen source for sulfur reduction. Several pathways to generate these products from hexyl sulfide are proposed, and kinetic parameters for the included reactions are calculated using transition state theory and quantum chemical calculations at the CBS-QB3 level of theory. Using these rate parameters, as well as previously calculated data from hydrocarbon and sulfur kinetic studies, reaction mechanisms were built using RMG for the conversion of hexyl sulfide to H 2 S in the presence and absence of SCW. Predictions from the RMG model agreed with reasonable accuracy with experimental data in the presence and absence of SCW. Because the model and experiments were in good agreement, flux Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 72 analysis was used to identify the most important reaction steps, and sensitivity analysis was used to propose reactions that should be studied further to decrease the model's uncertainty. 4.2 Introduction The ability to generate predictive kinetic models is valuable for many systems, ranging from process and product design' to engines 2' 3to large-scale atmospheric models. 4 The primary goal of chemical kinetic models is to make predictions for what ensues in a particular mixture under specified reaction conditions. The model prediction can then be compared with experimental data for validation. In many systems of technological importance, the chemistry is complex, involving hundreds of reactive intermediates. For this reason, systematic construction of chemical kinetic models by computer algorithms and databases is becoming increasingly attractive. Several research groups have developed software that automates model generation.' 6 Our own effort in this direction has resulted in the development of the Reaction Mechanism Generator (RMG), an open source software package for automatic mechanisms generation.7 RMG offers several advanced features, including estimates of pressure dependent rates, thermochemical estimates of cyclic species, solvation effects on rates and the main subject of this thesis, sulfur chemistry and reaction rates.> Sulfur, a natural component of crude oil, can range in concentration from 0.1 wt% in "sweet" samples up to 10 wt% in "sour" samples.'( The majority of the sulfur found in crude oil is present as over 1000 different molecular structures, ranging from aliphatic sulfides and disulfides to aromatic thiophenes and larger polycyclic benzothiophenes. 1 Sulfur content in crude oil has a severe impact on oil production and refinery processes, making "sour" crude undesirable; moreover, very low sulfur levels are required in most fuels to reduce engine emissions and prevent poisoning of catalytic converters. However, the sulfur content of remaining crude oil reserves is trending upward towards heavy "sour" crude, pushing oil companies to seek alternatives to standard desulfurization techniques, such as hydrodesulfurization (HDS), which . requires a heterogeneous catalyst and high pressure H 2 Supercritical water (SCW) treatment of heavy "sour" crude presents advantages over current desulfurization techniques for several reasons, especially because no catalyst or external ........... Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 73 hydrogen source is necessary. Thermal coker units are used for some heavy, sulfur-rich streams, but an alternative, SCW treatment, is attractive because it minimizes the formation of difficultto-handle solid coke.12 , 13 Previous research suggests that the desulfurization of sulfur compounds in SCW has the potential to be a sustainable, commercial-scale process. 4-2 However, large-scale commercialization of SCW treatment of "sour" crude has been impeded by a lack of mechanistic details and kinetic information. To build a valid model for desulfurization in SCW, RMG was extended to include accurate thermo-chemical and detailed rate information for sulfur compounds. A detailed database for the thermochemistry and kinetics of sulfur compounds has been developed by Vandeputte et al. 2 This has been extended to include reactions and thermochemistry estimations for compounds including both sulfur and oxygen, particularly for use in high-temperature models involving sulfur compounds and water, as discussed in the previous chapter. The sulfur database in RMG was validated by modeling the decomposition of diethyl sulfide at temperatures between 800 and 1000 K, and it succeeded to predict all of the major compounds of these experiments with quantitative accuracy.24 The purpose of this paper is to first present the development and implementation of new sulfur thermochemistry and kinetic calculations, with potential relevance to SCW desulfurization, into RMG software and databases. Secondly, mechanisms are generated for the SCW desulfurization of hexyl sulfide, as well as its pyrolysis without water. Predictions from these models are compared and validated with new and recently published experimental results. 2 5, 26 Sensitivity analysis is conducted to identify important reactions, and potential strategies for further model improvement are proposed. 4.3 Methods 4.3.1 Batch Reactor Experiment The experimental setup has been described in detail elsewhere, 2 6 and only a brief description is provided here. A stainless steel batch reactor with an internal volume of 24 mL was used for all experiments. Heating was provided to the reactor by a fluidized sand bath. For the base experiment, hexyl sulfide and water were added to the reactor in a 1:3 ratio (wt/wt). To Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 74 investigate the concentration effect of water on the production of pentane, the loading of water was adjusted while other components of the reaction mixture were kept constant. A total of 8 water/sulfide ratios were investigated experimentally, holding the sulfide loading constant, from a ratio of 0 to 70 by mole. The base experiments were conducted at 400 'C and approximately 28 5 MPa to ensure that the water was in the supercritical state (374 C and 22.1 MPa). Air was flushed from the reactor with helium, and 20 bar was left in the headspace upon sealing. The same temperature was used for all experiments with different initial water loadings, and as expected, the pressure increased with increasing water load. Once the reaction was quenched, the gas-phase product and liquid phase products were collected and analyzed. The gas-phase product was analyzed using conventional gas chromatography with either a flame ionization detector or a thermal conductivity detector (GC-FID or GC-TCD, respectively). The liquid oil phase was analyzed using two-dimensional GC with both an FID detector and a sulfur chemiluminescence detector (GCxGC-FID and GCxGC-SCD, respectively). Bulk sulfur content was measured with X-ray fluorescence (XRF). Representative water samples were analyzed for sulfur and trace metals content. For more details on the analytical instruments and reactor, see Kida et al.2 6 4.3.2 Batch Reactor Model For simulating the batch reactor experiments, the reaction network was implemented in CHEMKIN-PRO. The batch reactor was simulated using the closed homogenous reactor model.2 7 The reactor was simulated as closely as possible to the experimental conditions, with a five-minute linear heat-up from ambient conditions to the experimental temperature of 400 'C. Initial reactant concentrations were calculated assuming a homogeneous mixture of hexyl sulfide, water, and inert helium. 4.3.3 Continuous Flow Stirred Reactor (CSTR) Experiment The experimental CSTR setup has been described in detail elsewhere,2 and as with the batch reactor, only a brief description is provided here. The main reactor consisted of an Autoclave Engineers (AE) bolted enclosure housing a 600 mL Inconel-625 vessel, rated for 345 bars at 650 C.28 The stirring mechanism was a magnetically driven AE Magnedrive that could achieve stirring speeds up to 2500 rpm. Mixing was achieved using a combination of two mixers: a six- Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 75 blade radial turbine mixer attached at the tip of the impeller and a four-blade axial mixer located - at approximately 10 cm from the tip of the shaft. Water and hydrocarbon were fed separately without pre-mixing - to the top of the reactor through two separate tangential feed side ports. Product was withdrawn from the main bottom port. The gas-phase product was analyzed in real time with an online conventional gas chromatography with a flame ionization detector. The liquid oil phase was analyzed using twodimensional GC with both an FID detector and a sulfur chemiluminescence detector (GCxGCFID and GCxGC-SCD, respectively). Bulk sulfur content was measured with X-ray fluorescence (XRF). 4.3.4 CSTR Reactor Model For simulating the CSTR reactor experiments, the reaction network was implemented in Chemkin-Pro.2 The CSTR reactor was simulated using the perfectly stirred reactor (PSR) model. 4.3.5 Principles of Automated Mechanism Generation Reaction Mechanism Generator (RMG-Java, version 4.0.1) employs advanced methods in thermochemistry and kinetic parameter estimation that allow the construction of complex reaction networks. 7, 29, 30 RMG has demonstrated the ability to generate accurate kinetic models ranging from pyrolysis of hydrocarbons to the low and intermediate temperature oxidation of radicals during autoignition.3 1 ,32 At the heart of RMG is a fast rate-based algorithm that builds chemical kinetic models from an initial set of reactants and initial conditions; temperature, pressure, and species concentration.3 2 RMG reacts the initial species in all possible ways referencing kinetic and thermodynamic information stored in RMG's database and integrates the model in time. Each species generated in the chemical model is classified as either an "edge" or "core" species. RMG tracks the rate or flux of each new species produced and a species is transferred from the model edge to the model core once the species flux exceeds a critical value set by the user. The species transferred from the edge to the core are then reacted with already present core species in the model, to generate a new set of edge species and reactions. In this iterative process the core and edge are expanded Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 76 and the process is repeated. At each step of model generation the resulting differential equations are solved for all the species in the core. In this sequential way, a reaction network for the given initial conditions is formed. A detailed database for the thermochemistry and kinetics of sulfur compounds has been developed by Vandeputte et al.21 23 This has been extended to include reactions and thermochemistry estimations for compounds including both sulfur and oxygen, particularly for use in high-temperature models involving sulfur compounds and water. ' High accuracy calculations are already available for many hydrocarbons and small molecules. 34 Cumulatively, these calculations provide reasonably accurate rate estimations for a very small percentage of all of possible reactions that can occur. The other rate parameters are estimated by analogy, similar to what is done in Benson's textbook,3 5 but often using quantum calculations on a small molecule rather than experimental data as the basis of the analogy. Our experience is that this approach is not completely reliable. Thus, additional quantum calculations are conducted to improve the accuracy of rate parameters for reactions that are found to be important via sensitivity analysis. 4.3.6 Quantum Calculations Thermochemical and kinetic data were calculated at the CBS-QB3 level of theory using the Gaussian 03 quantum chemistry package. 3 6 All stable compounds were calculated in their singlet state, and radical compounds were calculated in their doublet states. Partition functions were calculated using the CanTherm software package,3 7 using the recommended scaling factor of 0.99 for the frequency analysis. 38 One-dimensional hindered rotations were also included in the analysis, using scans at the B3LYP/6-3 I IG(2d,p) level for each rotatable bond. Hindered rotor scans were stepped in 10-degree increments, and all other coordinates were allowed to reoptimize at each step. The effective moment of inertia 1(2,3) for each hindered rotor was calculated at the equilibrium geometry.3 9 Thermochemistry data for most species were estimated using Benson group additivity.4 ( An extensive database has been generated for group additivity values for the estimation of sulfur compounds, and these data are used along with other available estimation data in the RMG database. 23, 33 However, quantum calculations for specific compounds generally provide more ----------- Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 77 accurate thermochemistry data than group additivity schemes, so these calculations have been conducted for 30 of the reactants, proposed intermediates, and products of the hexyl sulfide decomposition mechanism, in addition to data for compounds that were calculated in previous work. 23 , 33 Thermochemical parameters were calculated in CanTherm. Enthalpy and entropy of formation were calculated at 298 K, and heat capacities were calculated at 300, 400, 500, 600, 800, 1000, 1500, 2000, and 2400 K. Bond additivity corrections were applied to obtain the enthalpy values, although no correction is available for the C=S double-bond, due to lack of experimental data.41 This topic will be explored further in Chapter 7. Tunneling corrections were applied to the rate constant calculations using the asymmetric Eckart method, which has been shown to provide accurate corrections for this type of chemistry.42 43 Rate constants were calculated in CanTherm using conventional transition state theory at 59 temperatures between 300 and 2000 K. These were fitted to the modified Arrhenius expression, = A-T"-exp J , k(T) R-T where T is the temperature and R is the gas constant. The three coefficients to be fitted are the Arrhenius constant A, the temperature factor n, and the activation energy E. 4.4 Results: Quantum Calculations 4.4.1 Water-Catalyzed Elimination of H 2 S Kida et al. have recently discussed how water catalyzes the dehydration of germinal mercaptoalcohols, reporting rate parameters for the reactions shown in Table 1.26 The presence of the additional water molecule is calculated to increase the first-order decomposition rate of the mercaptoalcohol by an order of magnitude, which can significantly affect product branching in the resulting reaction mechanisms. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 78 Table 1. Modified Arrhenius coefficients computed for the elimination of H 2 S to form acetaldehyde. A [s-' for 26 reaction 1, cm3 /(mol*s) for reaction 2], n (unitless), and Ea (kJ/mol). From Kida et al. logA OH H -s O HSH [o' OH SH+ H 2 0 Ea 0-H 1. 2. Rate Parameters n + H 2S 0.13 12.62 k(673K) = 0.041 1/s 185.09 3.05 0.54 k(673K)*[H20] = 0.74 1/s 93.20 'o'H - + H 2 S + H 20 Due to the comparative speed of the water-catalyzed reaction in Table 1, a brief study was conducted to determine kinetic parameters for three additional reactions, which are presented in Table 2. The first reaction in this table is the elimination of water from the mercaptoalcohol compound to form a thioaldehyde. A slightly lower activation energy was calculated for this reaction as compared with the H 2S elimination reaction in Table 1, but under the conditions of this study they will occur at similar rates. Thus, we expect a significant amount of pentane production from the water-catalyzed elimination of H 2 S from the mercaptohexanol. We expect this channel to have some competition from the water elimination reaction to reform the thioaldehyde. In Table 2, reactions 4 and 5 are analogous to reactions 2 and 3, respectively, but catalyzed by hydrogen sulfide instead of water. In both cases, the transition state is stabilized by the presence of the additional H 2 S molecule to form a six-member ring. This stabilization effect is about 30 kJ/mol less than in the water-catalyzed version of both reactions. With the lower rate constants and significantly lower concentration of H2 S in the reaction mixture in this study, the H 2 Scatalyzed reactions should be negligible compared with the water-catalyzed reactions. However, this type of reaction could be important in situations where large amounts of H 2 S are being produced, including pyrolysis conditions where water is not present. The analogous hexyl sulfide reactions for those presented in the two tables have been added to the RIG database in the direction presented, but the termolecular reverse reactions are also calculated by RMG using thermodynamic consistency. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 79 Table 2. Modified Arrhenius coefficients for three reactions catalyzed by H20 or H2 S. A [cm 3 /(mol*s)], n (unitless), and E,, (kJ/mol). Rate Parameters n logA OHH 3. SH / +H 2S+H2S -0.79 3.44 115.06 4. ---H+ H 20 1.08 2.59 86.02 20 + H H ' OH S ' H a + H 20 Ea SH+ H 2 S H SHH OHH'' 5. H+ H 2S H-S-', -1 )'SH - H S+H20+H2S0.32 2.94 119.75 S 4.4.2 Hydrogen Migration Four intramolecular hydrogen abstraction (or hydrogen migration) reactions were chosen for this work, based on possible relevance to the hexyl sulfide decomposition mechanism. Due to the stabilization from the neighboring sulfur atom, the most stable product should be the hexyl sulfide alpha-radical produced from hydrogen abstraction of the initial reactant. However, the proposed decomposition mechanisms suggest that other hexyl sulfide radicals-particularly the beta radical-could also be important intermediates in some of the pathways. Thus, rate parameters were calculated for reactions with five- or six-membered cyclic transition states to convert between different hexyl sulfide radicals. Rate parameters for four hydrogen migration reactions are presented in Table 3, and the transition state geometries can be seen in Figure 1. The ring size of the transition state had the greatest effect on the rate constant, as reactions 7 and 9 had greater A-factors but lower n-factors, as well as activation energies approximately 22 kJ/mol less than the similar reactions (6 and 8, respectively) with one fewer carbon in the cyclic transition state. Comparing reactions with the same number of atoms in the transition state ring, we see that the reaction rates at 400 'C will be similar (within a factor of 10) for reactions 6 and 8, and likewise for reactions 7 and 9. . ... . .. ... .... Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 80 7. 6. 1. 6 3 1. 1.403 9 1. 181. 1.357 1.361 1,355' 8. 9. 1. 5 1. 1.365 1.65 1.344- 71376 Figure 1. Transition state geometries for the four hydrogen migration reactions. Distances (Angstroms). Table 3. Calculated rate constants for hydrogen migration reactions. A (s-'), n (unitless), and E, (kJ/mol). Rate Parameters Reactions 6. - SH 7. lo SH 8. SS 9. S , -- - 1W SH SH logA n Ea -4.75 4.50 49.91 -1.29 3.24 29.04 -2.94 3.95 46.73 -1.55 3.28 24.73 4.4.3 Radical Addition to Multiple Bond Six radical addition reactions-beta scission reactions in the reverse direction-were considered for this work. These proceed via the pathway shown in Figure 2, and the calculated modified Arrhenius parameters are presented in Table 4. Reactions 10 through 14 were chosen as possible consumption reactions for the thioaldehyde formed from hexyl sulfide decomposition. Radical addition to the sulfur atom in the C-S double bond were not calculated in this work, as rate Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 81 parameters for many of these reactions have previously been calculated and added to the RMG database.2 Reactions 15 through 17 are relevant to hexyl sulfide decomposition in the beta scission direction, as they are possible final steps in the production of thiophenic compounds from hexyl sulfide. R3 R2 1 R4 Figure 2. A radical addition reaction. Activation energies for the four reactions involving addition of a radical to the thiocarbonyl group were all below 5 kJ/mol, and negative activation energies were fit to reactions 11 and 13 (although the overall rate constants exhibit the usual positive temperature dependence). This is due to the instability of thiocarbonyl compounds, which are known to polymerize at room temperature. 44 The energetics of reaction 14 reveal a submerged transition state, and a complex was optimized near this geometry. A rate constant was estimated for this reaction using the method discussed in the previous chapter. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 82 3 Table 4. Calculated rate constants for radical addition and beta scission reactions. A [cm /(mol*s)I, n (unitless), and E, (kJ/mol). Reactions Rate Parameters logA 10. As+ 14 11. Xs 9.07 n 1.46 Ea 4.92 0.19 3.16 -6.73 3.24 2.50 2.26 1.58 2.82 -5.14 12.2 0.03 2.10 )W 7.40 1.76 9.89 W 1.99 3.25 24.33 4.06 2.56 11.09 Ag 13. - 14. N-& + 'C 2 H5 No + V- SH S S 15. + q 16. + 'C2 H 5 S 17./ 'HS + + + C2 H 3 S - OW / M , 12. & SH + -MSH Greater activation energies were calculated for the addition of a radical to a stable thiophenic compound (reactions 15-17). Of more interest to this work is the reverse direction for reactions 15 and 16, which can be estimated using thermodynamic consistency. Beta scission of the ethyl radical from the initial cyclic radical occurs via a significantly lower energy pathway than beta scission of hydrogen. Thus, if beta scission of this radical were the primary method of generated thiophenic compounds from hexyl sulfide, the production of thiophene over ethyl thiophene would be expected, in disagreement with experimental data. However, disproportionation reactions also promote the generation of ethyl thiophene, and this could help explain the experimental results. The beta scission of the vinyl radical to form thiophene, the reverse of reaction 17, has a high activation energy. This species would be more likely to undergo radicalmediated tautomerization to eventually form ethyl-thiophene, an experimentally observed pyrolysis product of hexyl sulfide. This mechanism is discussed further below. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 83 4.4.4 Cyclic Sulfide Formation A cyclic sulfide formation reaction is a unimolecular radical addition, where a radical attacks the sulfur atom of a thiocarbonyl group to form a cyclic sulfur compound, as shown in Figure 3. Three of these reactions, depicted in table 5, were selected for this work, as possible intermediate steps leading to the generation of the experimentally observed ethyl-tetrahydrothiophene and ethyl-thiophene. R2R2 -<> -Ri Figure 3. A cyclic sulfide formation reaction. Modified Arrhenius parameters are presented in Table 5, and the transition state geometries are presented in Figure 4. Reaction 18 is nearly barrierless in the forward direction, while reactions 19 and 20 have similar calculated activation energies. All three of these reactions are exothermic in the cyclization direction. Reaction 20 has the greatest barrier in the reverse direction, as it requires the breaking of a stable thiophenic ring to form an unstable thioketene. Table 5. Calculated rate constants for cyclic sulfide formation reactions. A (s-1), n (unitless), and E, (kJ/mol). Reactions logA Rate Parameters Ea n 3 6.65 1.17 1.15 * 11.90 0.10 44.78 - S S 18. 11.85 -0.14 45.71 0. S - 19. 20C-' S I f Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water ivia, ty 19. 18. 84 2.374 _374 1.2 LA 20. 2.739 %1. Figure 4. Transition state geometries for the three cyclic sulfide formation reactions. Distances (Angstroms). 4.4.5 Thermochemistry Calculations In addition to thermochemical parameters calculated in previous work33 , parameters were calculated here for 30 additional species. These include the reactants and products of the reactions studied in the quantum chemistry work, as well as calculations that were used to estimate thermochemical parameters for reactants, products, and likely intermediates for the hexyl sulfide mechanisms. These parameters are further refined and discussed in Chapter 7. 4.5 Results: RMG Model Performance 4.5.1 Reaction Path Analysis To investigate the effect of SCW on the high-temperature reactivity of hexyl sulfide, mechanisms were generated by RMG in the presence and absence of water to simulate batch reactor experiments that were previously conducted at 400 OC. 2 6 The SCW mechanism includes 273 species and 5971 reactions. Reaction path analysis of the RMG mechanisms suggests that the formation of the major products in the SCW treatment of hexyl sulfide occurs via the reaction Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 85 pathways proposed by Kida et al., as presented in Figure 5.26 The reactant and observed products are encased in boxes, and predicted products that are not observed experimentally are encased by dashed boxes. This figure neglects reactions with fluxes below one percent for clarity. The highest flux channel for the cracking of hexyl sulfide (species 1 in Figure 5), which is responsible for 61% of the total consumption, starts with a hydrogen abstraction from the parent molecule. The primary H-abstraction occurs from the alpha-carbon (carbon adjacent to the sulfur atom), as the resulting alpha-radical is stabilized by its proximity to the sulfur atom. Small amounts of gamma-, delta-, and epsilon-radicals are predicted, and these all are expected to convert to the alpha-radical via intramolecular hydrogen migration reactions, but mainly react back to form the initial reactant. The alpha-radical then undergoes a beta scission reaction (65% flux) to form hexane (2) and hexanethial, the latter of which will be discussed in more detail. The second major reaction pathway (27% flux) involves a hydrogen abstraction from hexyl sulfide to form a beta radical, which quickly undergoes beta scission to form 1-hexene (4) and hexanethiyl radical. Hexanethiyl radical further reacts to form hexanethiol (3). The 1 -hexene is predicted to convert to 2- and 3-hexene (5 and 6, respectively) via hydrogen abstraction reactions, with the majority predicted to exist as 2-hexene after 30 minutes. Hexanethial produced by the beta scission of the alpha radical is predicted to be consumed by two pathways. The main one (47%) is the pentane (5) and CO production pathway proposed by Kida el al. (Path 1 in Figure 5). Hexanethiyl radical is then predicted to react with some CO to form carbonyl sulfide (OCS), which would likely react further to form CO2.45 RMG also predicts a variety of radical addition reactions, mainly attacking the carbon side of the C=S double-bond, leading to the production of larger minor products, some of which continue to react. Additional hexanethial consumption pathways are also predicted, but with much smaller fluxes. While none of these products is observed experimentally, these pathways could be early steps in the formation of large aromatic compounds, which may explain the dark color of the product solution observed in some of the experiments. 2 6 The pathways will be discussed further in the analysis of the anhydrous decomposition mechanism. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water R RH C5H 13SH 3 R C 6H 13s . 24% 4 27 5 RH CH1 S2% RHCH1 R R R R S'CSH13 'RH RH 24% 29% C7H13 RH 86 CH H13 2% '61% RH C 6H 13 86% 2 C 5H R 6 CGH 13 86% R ( 1 % H R S 7% H2 0 47% Y S OHH C6 H13 CSH 11 SH KR 6% 47%1 RH H2S SH C 5H11 1CS 47% C6H 13 6* RH 2H2 RH 46 RH 6 Figure 5. Predicted reaction fluxes for the SCW desulfurization of hexyl sulfide at 400 *C The mechanism for hexyl sulfide pyrolysis without water includes 231 species and 2773 reactions. The major reaction pathways for pyrolysis without water are largely similar to the SCW case, and the total flux analysis for this mechanism is presented in Figure 6. The hydrolysis of the thioaldehyde shown in Figure 5 is not possible in the absence of water, so there is greater flux toward the diene pathway than in the SCW mechanism. In addition, a lengthy reaction sequence to produce thiophenic compounds is predicted, and this mechanism is presented in Figure 7. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 87 In both models, dienes have been predicted as significant products although they are not observed experimentally. This may correspond to the amount of coke-non-volatile species with a high aromatic content-that was observed in both experiments. Although this content was not quantified, it was clear that more coke was produced in the absence of water, due to the darker color of the pyrolysis product and substantial loss of carbon that could be quantified by GC analysis of gas- and liquid-phase products. 2 6 Thus, we propose that the production of dienes and cyclic species from the thioaldehyde is an early step in the production of these large compounds. Unfortunately, it is extremely difficult to track these species in detailed mechanisms, as there is no single "coke molecule," but many different large species that will individually have very small concentrations. Using the current version of RMG with available computational resources, it is difficult to build a comprehensive mechanism including species with more than 30 heavy atoms. Improving the ability of mechanism generation software to capture large-molecule chemistry should be a topic of future research. 88 Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water SH RH D SH 1%10% RH 6 1<1% -10% RH R R R 3S 6H27%6C 1 3 RH RH RH 27% 27% R 27% 4 R % 5 S'C-6H 13 RH RH N, S'C6H13 27R. R~ RHR ~ C13 RH * RA RH 1% C6H13 %C6H13 1 73%R S C6)K13 C 6H 13 RH 1% R' C 6H 14 2 RH RRH 71% C6H13 71% 31% SH R R* 20% 20% R 11% 20% SH RH 31% R' i See fig. 15 SH 31 %H RH 7 31% 28% RH 28% R . RH Figure 6. Predicted reaction fluxes for the pyrolysis of hexyl sulfide at 400 *C in the absence of water Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 89 C6 H11 RH R SH S SH SH RH s C 6H,1 RH RH S S S RH RH -s S C 2H5 S & R R k' _) R yRH RH F F RH A RH R irS Figure 7. Speculative reaction pathway for the formation of thiophenic compounds from the pyrolysis of hexyl sulfide at 400 *C in the absence of water 4.5.2 Model validation for hexyl sulfide conversion in a CSTR An additional mechanism was generated using RMG for the reaction mixture of hexyl sulfide, hexadecane, and SCW, simulating the reaction conditions used by Patwardhan et al.2 ' The CSTR mechanism was generated to be valid for temperatures between 400 and 450 'C, and it contains 140 species and 1957 reactions. The predicted reaction pathways for hexyl sulfide decomposition are largely similar to the hexyl sulfide + SCW case, although the radical pool is different in this case. C-S bond scission remains the dominant initiation reaction, but as hexadecane was present in a much higher concentration than hexyl sulfide in the CSTR experiments, radicals produced by the hexadecane solvent via bond-breaking or hydrogen abstraction exist in a higher concentration. The model predictions are compared with experimental data for hexyl sulfide conversion in Figure 8. Overall, RMG predicts sulfide conversion accurately, with a slight underprediction at Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 90 the lowest temperature and a more significant overprediction at the highest. This may reflect the uncertainty in the overall activation energy for hexyl sulfide decomposition predicted by RMG. Our model is predicted for reactions that occur at a lower conversion (up to 25%), so it is likely to be inaccurate at very high conversions. It is also important to remember the experimental uncertainty in measuring the conversion of a relatively small amount of sulfide in a bulk of hexadecane. 100 ....................... .................... 80 - - 400C, 500ppm 400C, 1000ppm 0- 60 -425C, o5 ---- 450C, 40 ~Expt 400C, lOO0ppm A t X 20 K 1000ppm 1000ppm p4 Expt 400C, 500ppm A Expt 425C, 1000ppm 0 Expt 450C, 1000ppm 0 0 10 20 40 50 30 Residence Time (min) 60 70 Figure 8. Comparison of RMG model with CSTR data for hexyl sulfide conversion in SCW. 4.5.3 Model validation for product distributions in a batch reactor SCW Mechanism All of the major products are predicted reasonably accurately for SCW treatment, in comparison with previously conducted batch reactor experiments. 6 Comparisons are presented in Figure 9 for hexyl sulfide, hexane, hexenes, hexanethiol, and pentane (note that the first 5 minutes, the reactor heat-up time, are not presented). The conversion of hexyl sulfide and concentrations of hexane and hexanethiol are predicted with excellent accuracy in comparison to experimental data. The RMG model also captures the reactivity of hexanethiol in this case, although the consumption rate of the thiol is overpredicted in comparison with experimental data. The concentration of this sulfur compound reaches a maximum at some point in the reaction timescale, before converting to other products-this is discussed further in the discussion of reaction pathways. In addition, pentane, a key product of the SCW experiments, is predicted Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 91 accurately by the RMG mechanism, suggesting that the SCW-hexanethial pathway is indeed leading to the formation of this product. 1 00 0.9 0 0 S 75 0 0.6 50 0 0 0.3 25 C5H11 0 0 5 12 10 is S 20 C 5H11 25 C6H14 5 30 10 Time (minutes) 15 20 25 30 Time (minutes) 0.6 0.4 C 6 H 13 SH E 0.3 0.4 0.2 0 0.2 0 * 0.1 5 ----10 ,--15 0.0 - 0.0 20 25 5 30 10 Time (minutes) 15 20 Time (minutes) 25 30 0.6 0 C5H12 1 0.0 5 10 15 20 25 30 Time (minutes) Figure 9. Comparison of RMG model with batch reactor data for hexyl sulfide in SCW at 400 *C. Full mechanism (solid), experimental data (o) Normalized sensitivity coefficients for hexyl sulfide concentration were calculated at a time of 6 minutes, and these are presented in Figure 10. At this time, the most sensitive reaction for hexyl sulfide decomposition is the beta-scission of the alpha-radical to form hexanethial and hexyl radical, which are intermediates in the main predicted product pathway to form pentane and hexane, respectively. Hydrogen transfer reactions involving H2S and hexanethial are also Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 92 important to our overall prediction of the sulfide decomposition rate. Although they are involved in minor pathways, our prediction is also sensitive to disproportionation reactions involving hexanethial and subsequent intermediates. The rate estimates for these reactions have large uncertainty in comparison with the other most sensitive reactions. However, it is important to note that the other rate constants also have uncertainties of a factor of 2 or more, and these uncertainties will propagate to a significantly greater uncertainty in the overall model predictions. Decreasing this uncertainty through improved thermochemistry and rate estimates will be a topic of further study. -0.05 0 0.0.5 Ce.,1 1 ' C Hl< S 'CrH 11 S + CbH,1 C 4H9 + C5HII- C5H 11 ' 'S s C4H 4+ S C5H 1 - + .0.1 SS S C,Hi C4HR S SH + C4H9 H2S + 3 CrH-1 SH + CH SH + CSH SH * S H 2S C 4H9 + S bHib HS CH SCH C4HqC C 4 H9 IC4H, C4H9 5 4 H+ SHH + CsH11 S'CH - -- SH C4H] 1 : + SH + C5H,1 S S CA Figure 10. Normalized sensitivity coefficients for hexyl sulfide concentration, calculated at t=360 s Non-SCW Mechanism As discussed in the reaction path analysis section, the computer-generated non-SCW mechanism predicted thiophene production as a major pathway, although this was not observed experimentally. A speculative mechanism was also generated using the same input parameters but disallowing the cyclicization reaction ultimately leading to the production of thiophene, which is presented as the red arrow in Figure 7 (this was done by setting the activation energy of this reaction arbitrarily high). The results of both simulations are presented in this section. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 93 The predicted conversion of hexyl sulfide, and production of hexane, hexenes, and hexanethiol are compared with batch reactor experiments in Figure 11. The pyrolysis model with the fast channel to thiophene blocked shows excellent agreement with experimental data for hexyl sulfide conversion, although the full mechanism disagrees with experimental conversion by about a factor of two. This difference is primarily due to the disproportionation reaction (presented in Figure 7) that forms ethyl-thiophene. This reaction is predicted to decrease the overall radical pool significantly. Major products are predicted with greater accuracy by the mechanism neglecting thiophene formation -- 100 X4-- - ------- In 0.9 za C .2 0 E 4, = (A 4, 25 C51-H1 5 10 is 20 0 C5HI1 S 0.3 C6H114 30 25 5 10 15 20 25 30 Time (minutes) Time (minutes) 0.6 04 In x x C6H 13 SH 0 X 0.3 . 0.4 0.2 / I- 02 0 / 0.0 5 10 C 6H 12 15 20 25 Time (minutes) 30 / 0.1 Is 0 0.0 5 10 15 20 25 30 Time (minutes) Figure 11. Comparison of RMG model with batch reactor data for hexyl sulfide pyrolysis at 400 *C. Full mechanism (solid), mechanism blocking fast cyclization (dashed), experimental data (x) 4.5.4 Effect of Water Concentration Additional batch reactor experiments were conducted for this work to investigate the dependence of pentane production on the initial concentration of water, and mechanisms were generated using RMG to model these experiments. The comparison of the model prediction and experimental data is presented in Figure 12. We see that the RMG model significantly underpredicts the production of pentane at low concentrations of water, and slightly overpredicts Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 94 it at high concentrations. This disagreement underscores the complexity of the chemistry occurring in this system. Experiments below a water/sulfide ratio of 24 (the first 5 experimental points in Figure 12) did not reach the critical pressure. These experiments were therefore conducted in superheated steam. At these low-water conditions, reaction 1 is more important than reaction 2. As the water loading increases in the supercritical range, the production of pentane is predicted to increase but then level off at a pentane yield greater than what is observed experimentally. This overprediction could be due to the absence of important competing pathways for thioaldehyde consumption. This is expected to be a major source of the experimentally observed "coke" (colored, nonvolatile material), which was produced in both the water and no-water cases (but decreased when supercritical water was present). This molecularweight growth chemistry is extremely complex and currently very difficult to predict, but it will be possible to explore these pathways more extensively as computational methods improve. 0.80 0.60 S0.400 0 0.00 0 20 40 60 80 100 Initial Moles Water/Hexyl Sulfide Figure 12. Comparison of experimental data and RMG simulations for production of pentane in 30 minutes in a 400 *C batch reactor, from different initial water/sulfide ratios. Model (solid,+), experimental data (o). 4.6 Conclusions Reaction mechanisms have been generated using the automated Reaction Mechanism Generator to model the reactivity of hexyl sulfide, with and without the presence of supercritical water. Quantum chemistry calculations have been completed to improve the rate constant and thermochemistry estimations used in mechanism generation. Calculations on the formation of carbon monoxide from a thioaldehyde confirm the previous hypothesis that a water-catalyzed pathway is likely to contribute to CO formation at supercritical conditions. Rate constants and Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 95 thermochemistry parameters have also been calculated for intramolecular hydrogen migration reactions, as well as other reactions potentially relevant to the formation of thiophenic compounds from sulfide pyrolysis. CSTR models using the RMG mechanism provide excellent agreement with experimental data for the decomposition of hexyl sulfide in the presence of hexadecane and SCW. Good agreement is also observed between experimental measurements and model predictions for almost all of the products generated in the batch reactor study. Sensitivity analysis for hexyl sulfide conversion shows that reasonable rate constant estimations are employed for the most important reactions in the simulation, but as always, more accurate and computationally expensive calculations could possibly improve agreement with the data. Sensitivity analysis is a useful tool for identifying areas of improvement in a kinetic mechanism, but it alone is not sufficient in confirming the accuracy of a mechanism. A rate constant that is misestimated by multiple orders of magnitude can be identified as sensitive, only to disappear from this analysis when a more reasonable estimate is applied; more worryingly, a poorly estimated (or missing) reaction may not appear in the sensitivity analysis at all even though it is actually quite important. Improvements in the rate estimation method, via improvements in the RMG algorithm or automated transition state searches and rate calculation methods, will provide more useful mechanisms on the first attempt and allow one to focus on refining the reactions that are truly important to the model predictions. 4.7 References 1. S. Brarendregt, P. J. M. Valkenburg, E. S. Wagner, M. Dente and E. Ranzi, in 14th Ethylene Producers'Confirence, AIChE, New Orleans, LA, 2002, pp. 2497-2537. F. Battin-Leclerc, J. M. Simmie and E. Blurock, eds., Cleaner Combustion: Developing DetailedChemical Kinetic Models, Springer, London, 2013. Y. Shi, H.-W. Ge and R. D. Reitz, Computational Optimization of Internal Combustion Engines, Springer, London, 2011. Rethinking the Ozone Problem in Urban and Regional Air Pollution, National Academy of Press, Washington, DC, 1991. W. H. Green, in Advances in Chemical Engineering, ed. G. B. Marin, Academic Press, vol. 32, pp. 1-50. 2. 3. 4. 5. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 96 A. S. Tomlin, T. Turinyi and M. J. Pilling, in Comprehensive Chemical Kinetics, ed. M. J. Pilling, Elsevier, 1997, vol. 35, pp. 293-437. W. H. Green, J. W. Allen, R. W. Ashcraft, G. J. Beran, C. A. Class, C. Gao, C. F. Goldsmith, M. R. Harper, A. Jalan, G. R. Magoon, D. M. Matheu, S. S. Merchant, J. D. Mo, S. Petway, S. Raman, S. Sharma, K. M. Van Geem, J. Song, J. Wen, R. H. West, A. Wong, H.-W. Wong, P. E. Yelvington and J. Yu, Reaction Mechanism Generator (RMG), v. 4.0.1 (2013). http://rmg.sourceforge.net. J. W. Allen, C. F. Goldsmith and W. H. Green, Phys Chem Chem Phys, 2012, 14, 11311155. A. Jalan, R. H. West and W. H. Green, J. Phys. Chem. B, 2013, 117, 2955-2970. R. Hua, Y. Li, W. Liu, J. Zheng, H. Wei, J. Wang, X. Lu, H. Kong and G. Xu, J. Chromatogr. A, 2003, 1019, 101-109. J. Beens and R. Thijssen, J. High Resolut. Chromatogr., 1997, 20, 131-137. A. Kruse and E. Dinjus, Journalof SupercriticalFluids, 2007, 39, 362-380. N. Akiya and P. E. Savage, Chemical Reviews, 2002, 102, 2725-2750. A. R. Katritzky, R. A. Barcock, M. Balasubramanian, J. V. Greenhill, M. Siskin and W. N. Olmstead, Energy & Fuels, 1994, 8. 0. M. Ogunsola and N. Berkowitz, Fuel, 1995, 74, 1485. T. Adschiri, R. Shibata, T. Sato, M. Watanabe and K. Arai, Ind Eng. Chem. Res., 1998, 37, 2634-2638. B. M. Vogelaar, M. Makkee and J. A. Moulijn, Fuel Processing Technology, 1999, 61, 265-277. A. R. Katritzky, M. Balasubramanian and M. Siskin, Energy & Fuels, 1992, 6, 431-438. T. Sato, T. Adschiri, K. Arai, G. L. Rempel and F. T. T. Ng, Fuel, 2003, 82, 123 1-1239. M. Watanabe, S. Kato, S. Ishizeki, H. HInomata and R. L. Smith, Jr., J. Supercrit. Fluids, 2010, 52, 48-52. A. G. Vandeputte. PhD dissertation, University of Ghent, 2012. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Phys Chem Chem Phys, 2012, 14, 12773-12793. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Chemistry-A European Journal, 2011, 17, 7656-7673. A. G. Vandeputte, C. A. Class, M.-F. Reyniers, W. H. Green and G. B. Marin, Submitted, 2014. P. R. Patwardhan, M. T. Timko, C. A. Class, R. E. Bonomi, Y. Kida, H. H. Hernandez, J. W. Tester and W. H. Green, Energy & Fuels, 2013, 27, 6108-6117. Y. Kida, C. A. Class, A. J. Concepcion, M. T. Timko and W. H. Green, Phys Chem Chem Phys, 2014, 16, 9220-9228. CHEMKIN-PRO 10131, (2013) Reaction Design, San Diego. P. A. Marrone. PhD Thesis, Massachusetts Institute of Technology, 1998. J. Song, S. Raman, J. Yu, C. D. Wijaya, G. Stephanopoulos and W. H. Green, Abstr. Pap. Am. Chem. Soc., 2003, 226, U530-U531. J. Song, R. Sumathi, J. Yu and W. H. Green, Abstr. Pap. Am. Chem. Soc., 2004, 228, U233. K. M. Van Geem, M.-F. Reyniers, G. B. Marin, J. Song, W. H. Green and D. M. Matheu, AIChE J., 2006, 52, 718-730. Chapter 4: Modeling the desulfurization of hexyl sulfide by supercritical water 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 97 M. R. Harper, K. M. Van Geem, S. P. Pyl, G. B. Marin and W. H. Green, Combust. Flame, 2011, 158, 16-41. C. A. Class, J. Aguilera-Iparraguirre and W. H. Green, Submitted, 2014. C. F. Goldsmith, G. R. Magoon and W. H. Green, Journal of Physical Chemistry A, 2012, 116, 9033-9057. S. W. Benson, The Foundationsof Chemical Kinetics, McGraw-Hill, New York, 1960. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, 1. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. S. Sharma, M. R. Harper and W. H. Green, CanTherm open-source software package, (2010). http://github.com/GreenGroup/CanTherm A. P. Scott and L. Radom, Journalof.Physical Chemistry, 1996, 100, 16502-16513. A. L. L. East and L. Radom, Journalof ChemicalPhysics, 1997, 106, 6655-6674. S. W. Benson, F. R. Cruickshank, D. M. Golden, G. R. Haugen, H. E. O'Neal, A. S. Rodgers, R. Shaw and R. Walsh, Chemical Reviews, 1969, 69, 279-324. G. A. Petersson, D. K. Malick, W. G. Wilson, J. W. Ochterski, J. A. Montgomery and M. J. Frisch, Journalof Chemical Physics, 1998, 109, 10570-10579. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers, V. Van Speybroeck, M. Waroquier and G. B. Marin, J. Phys. Chem. A, 2007, 111, 11771-11786. C. Eckart, Physical Review, 1930, 35, 1303-1309. N. J. Cooper, Compr. Org. Funct. Group Transform. H, 2005, 3, 355-396. C. Deng, Q.-G. Li, Y. Ren, N.-B. Wong, S.-Y. Chu and H.-J. Zhu, Journal of Computational Chemistry, 2007, 29, 466-480. Chapter 5: Modeling the decomposition of alkylaromatic compounds 98 Chapter 5: Modeling the decomposition of alkylaromatic compounds 5.1 Abstract The Reaction Mechanism Generator (RMG) is used to model the dealkylation of aromatic compounds in pyrolysis and supercritical water (SCW) treatment. Experiments on whole crude oil have shown that heavy hydrocarbons are cracked to form lighter compounds in SCW treatment, increasing the value of the product. Model compound studies were used to investigate the mechanism for this upgrading process, and this chapter presents our efforts to explain the results using ab initio methods and automated mechanism generation. Thermochemical calculations were able to explain the differing reaction rates of 2-hexylthiophene, 3- hexylthiophene, and hexylbenzene. RMG was then used to build detailed reaction mechanisms for the decomposition of hexylbenzene. The key result was that SCW does not have any direct impact on the decomposition of this model compound, as effectively identical product distributions were observed experimentally and predicted by RMG with and without water. This result suggests that SCW is not directly reacting with these compounds in crude oil, but facilitating upgrading by preventing other reactions-such as addition to a thioaldehyde, discussed in Chapter 4-that lead to the formation of large compounds. Chapter 5: Modeling the decomposition of alkylaromatic compounds 5.2 99 Introduction A key finding in the experimental work on the SCW desulfurization of hexyl sulfide was the formation of heavy compounds in the anhydrous pyrolysis experiments, evidenced by the darker product liquid and lack of mole balance closure. The SCW experiment provided a clearer product with better mole balance, suggesting not only that intermediate sulfur products are a possible cause of coking in the high-temperature chemistry of crude oils, but also that SCW could prevent this coking.' We were able to explore some of the mechanistic reasons for that in the previous chapter. Experimentation on heavy crude oil provided a similar result, that SCW treatment resulted in lighter and therefore more valuable product distributions than simple pyrolysis.2 In particular, this was exhibited by the cleavage of alkyl chains connected to aromatic rings, producing a much smaller alkylaromatic compound. Model compound experiments and detailed mechanistic studies would allow us to investigate this process on a fundamental level. In this chapter, a model for hexylbenzene decomposition is generated using RMG. Ab initio calculations allow us to explain the relative reaction rates of three alkylarenes thermochemically, suggesting that the decomposition occurs via free-radical mechanism. The detailed decomposition mechanisms for hexylbenzene with and without SCW are generated and then validated with experimental data, agreeing within the uncertainty of the calculations. The pyrolysis case can then be compared with the SCW decomposition to help explain the role of SCW in the overall crude oil upgrading process. 5.3 Methods 5.3.1 Reaction Simulation The Reaction Mechanism Generator (RMG) was used to build kinetic mechanisms for this work.' The RMG algorithm has been discussed extensively in past work,4 5 and only a brief introduction will be provided here. The key feature is a flux-based algorithm for model generation, pursuing reaction pathways in directions with greater flux while omitting those with lower predictions. RMG produces a file containing NASA polynomials to estimate thermochemical parameters for each of the species in the mechanism, as well as the modified Chapter 5: Modeling the decomposition of alkylaromatic compounds 100 Arrhenius parameters for each reaction. This file can be imported into CHEMKIN-PRO to simulate the kinetic experiments, and conduct flux and sensitivity analyses. 6 For this work, mechanism generation and simulation conditions were chosen to match the experimental conditions as closely as possible. Mechanisms were generated in RMG using the experimental temperature, pressure, and starting concentrations (one with and one without SCW). All reactions were assumed to be in the high-pressure limit. Reactor simulations were conducted using the "Closed Homogeneous Batch Reactor" model in CHEMKIN-PRO, including a 1 0-minute linear heat up time to reach the experimental temperature of 450 'C. 5.3.2 Quantum Calculations Previously calculated Arrhenius and thermochemical are available in literature for the decomposition mechanisms of alkylaromatic 7 compounds, and some of these were added to the RMG database to assist in parameter estimation for this work. In addition, the RMG database is able to estimate reasonably accurate rate parameters for many other relevant reactions using previously calculated data. However, when generating mechanisms containing thousands of reactions, it is not possible to find extremely accurate rate parameters for each of the predicted reactions. Thus, some of them must be roughly estimated, and these estimations can be uncertain by multiple orders of magnitude. These uncertainties are mitigated by calculating more accurate rate constants using ab initio techniques. These calculations can still be uncertain by up to a factor of ten, so it is important to consider these uncertainties when analyzing these reaction mechanisms. In this work, some reactions were identified as having a significant effect on the rate of phenyldodecane decomposition, and the initially estimated parameters were improved using quantum mechanics and transition state theory. Gaussian 038 was used to determine the geometries and vibrational frequencies of stable molecules and reaction transition states at the B3LYP/6-31 1G(2d,p) level of theory, and single point energies were calculated using CBS-QB3. Parameters for particularly important reactions, including the retroene reaction depicted in Figure 1, were refined using CBS-QB3 9' 10 and CCSD(T)-F12/cc-pVDZ-F12"" single point energies. Calculations using the second, coupled-cluster method were conducted using Molpro.1 4 Chapter 5: Modeling the decomposition of alkylaromatic compounds The open-source CanTherm 5 101 software package was used to calculate rate constants and thermochemical parameters using transition state theory. A scaling factor of 0.99 was used for the frequency analysis. One-dimensional hindered rotations were also included in this analysis, using scans at the B3LYP/6-3 1 G(2d,p) level of theory in 10 degree increments. The effective moment of inertia 1(2,3) was calculated for each hindered rotor. Modified Arrhenius constants were then derived for each of these reactions, and these parameters were added to the RMG database to improve model generation. ph [Ph 1401 H 1 + Ph 1.927 2.323 Figure 1. Proposed pathway (top) and optimized transition state (bottom) for the retroene reaction to produce toluene and 1-pentene. Interatomic distances (X). 5.4 Results and Discussion 5.4.1 Quantum Calculations While similar product distributions were obtained in experiments on the pyrolysis of 2hexylthiophene, 3-hexylthiophene, and hexylbenzene, the rate of decomposition was measured to be significantly different. 2 We theorized that this was due to the stability of the radicals arising from each compound, which can be illustrated by resonance in Figure 2. Thermochemistry calculations were conducted at the CBS-QB3 level of theory, for the analogous propylarenes instead of hexylarenes, to save computational time. The calculated enthalpies, presented in Table , show that the stability order of the three compounds is 3-propylthiophene > propylbenzene > Chapter 5: Modeling the decomposition of alkylaromatic compounds 102 2-propylthiophene. This agrees qualitatively with the decomposition rates for the analogous hexylarenes. Rj R C Nr-r U- R* S S R Figure 2. Dealkylation of three alkylarenes, and resonance structures of the resulting radicals. Table 1. Calculated enthalpies of reaction for C-C bond breaking in alkylarenes, and experimentally observed conversions2 in pyrolysis at 450 *C for 30 min. Reaction AH rxn (kcal/mol) Observed Conversion (%) 77.2 93 74.9 9 R R R S/1 R R 70 78.1 ZR 5.4.2 RMG Model Performance The reaction mechanism generated to model the pyrolysis experiment contains 279 species and 583 reactions, while the SCW model contains 275 species and 578 reactions. The major reaction pathways predicted for the two cases are effectively identical, in agreement with the experimental results. The main reaction steps, including fluxes, are presented in Figure 11. Hexylbenzene pyrolysis is predicted to proceed mainly via normal cracking, in accordance with previous modeling research on alkyl aromatics.7 The "retroene reaction," the molecular reaction forming pentene and toluene from the reactants, is predicted to provide appreciable toluene production, although only about 10% of the overall decomposition was via this pathway. This is in agreement with previous studies, which predicted 80% of toluene production via this molecular pathway at 330 'C, but only 20% at 400 oC. 6 ,'17 This reaction, as well as the free- Chapter 5: Modeling the decomposition of alkylaromatic compounds 103 radical toluene production pathway, also account for the foriation of 1-pentene, which then can go on to react further via decomposition or addition reactions. Styrene is predicted as a major product via beta-scission pathway. Experimental results from this and previous work7 ' 18 suggest that this compound is an intermediate in the formation of ethylbenzene, although our experiments at 450 'C show some styrene remaining as a product after 40 minutes. RMG predicts that the reverse disproportionation pathway leads to some production of ethylbenzene, but this reaction is less dominant than at lower temperatures. A comparison of experimental vs. model hexylbenzene conversion is plotted in Figure 12, and the comparison for the main aromatic products is presented in Figure 13. Excellent agreement is observed for the reactant conversion, suggesting that the rate constants and thermochemistry for the main reaction steps are reasonably accurate. Toluene is overpredicted, but well within the model's uncertainty. The sum of ethylbenzene and styrene production predicted by RMG also agrees with what was observed experimentally, which also suggests that the first few reaction steps are predicted well. However, ethylbenzene is underpredicted by the model while styrene is significantly overpredicted, likely due to two factors. First, the expected reverse disproportionation pathway, which seems to be supported by the experimental results, has been predicted to be slower than necessary to account for the observed amount of ethylbenzene. This disagreement is likely due to the uncertainty in the calculation of rate constants for disproportionation reactions, which is about a factor of 10. Second, RIG predicts a number of large molecules (two or more benzene rings) to be produced from these experiments, and each of these compounds is predicted in very small amounts. The low concentration of each of these species is reasonable, as there are many different possible addition and recombination products; no single large compound is likely to be produced in great quantity. On the other hand, we propose that many different large molecules are produced in these experiments, each in rather small quantities. This hypothesis is supported by the small peaks observed on the heavier end of the GC chromatographs, as well as the lack of benzene balance closure, which is also observed for hexyl thiophene. 2 104 Chapter 5: Modeling the decomposition of alkylaromatic compounds 20/At 4% --,C2H 7 7% 28% 10% 3% ~ C 3H6 - C5HIe 42% 6% 649 23% 2% C4H C 4H 40% CH 3 +o CH 4 8 A 2% 5% 2% Figure 3. Reaction fluxes for the pyrolysis of hexylbenzene at 450 *C. 0 75 - 100 40 C 50 25 - a - 45% I~~~ 0 0 10 20 30 I 40 Time (minutes) Figure 4. Hexylbenzene conversion vs. time at 450 *C. (can add more plots if needed) (x) pyrolysis experiment, (o) SCW experiment, (-) pyrolysis model, (--) SCW model. 7% 0.6 - -- 105 - Chapter 5: Modeling the decomposition of alkylaromatic compounds 0. - 'X E X ra 0.2 0.0 0 10 20 30 40 lime (minutes) Figure 5. Model (solid lines) comparison with experiments (pyrolysis x, SCW o) for production of toluene (blue), styrene (black), and ethylbenzene (red). Aromatic molar yield is defined here as fraction of total aromatic compounds present as the given product. Only the pyrolysis model is presented here 5.5 Conclusions In this chapter, a reaction mechanism was presented for the decomposition of hexylbenzene at 450 'C. This compound was predicted to react primarily via normal cracking, with some amount of the conversion occurring via the molecular retroene reaction to form 1 -pentene and toluene directly from the reactant. The predictions for major products agreed reasonably well with the experimental results, but an extended reaction mechanism would likely generate better predictions of minor products. In the RMG mechanism and experimental product distribution, no change was predicted with the addition of SCW. Thus, it is probable that water does not react directly with an alkylbenzene or its intermediates during treatment, but with other reactive species that would lead to the formation of larger compounds. One example of this would be the reactive thioaldehyde discussed in Chapter 4, which leads to the formation of thiophenes in pyrolysis but lighter compounds in SCW treatment. However, there are thousands of other species in these fuels, and additional investigations may uncover additional benefits of this relatively simple process. Chapter 5: Modeling the decomposition of alkylaromatic compounds 106 5.6 References 1. Y. Kida, C. A. Class, A. J. Concepcion, M. T. Timko and W. H. Green, Phys Chem Chem Phys, 2014, 16, 9220-9228. Y. Kida, A. G. Carr and W. H. Green, Energy & Fuels, 2014, 28, 6589-6595. W. H. Green, J. W. Allen, R. W. Ashcraft, G. J. Beran, C. A. Class, C. Gao, C. F. Goldsmith, M. R. Harper, A. Jalan, G. R. Magoon, D. M. Matheu, S. S. Merchant, J. D. Mo, S. Petway, S. Raman, S. Sharma, K. M. Van Geem, J. Song, J. Wen, R. H. West, A. Wong, H.-W. Wong, P. E. Yelvington and J. Yu, Reaction Mechanism Generator (RMG), (2013). J. W. Allen, C. F. Goldsmith and W. H. Green, Phys Chem Chem Phys, 2012, 14, 11311155. M. R. Harper, K. M. Van Geem, S. P. Pyl, G. B. Marin and W. H. Green, Combust. Flame, 2011, 158, 16-41. CHEMKIN-Pro, (2011) Reaction Design Inc., San Diego. V. Burkle'-Vitzthum, R. Michels, G. Scacchi and P.-M. Marquaire, Ind. Eng. Chem. Res., 2003, 42, 5791-5808. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, Journal of ChemicalPhysics, 1998, 110, 2822-2827. J. A. Montgomery Jr., M. J. Frisch, J. W. Ochterski and G. A. Petersson, Journal of ChemicalPhysics, 2000, 112, 6532-6542. J. Aguilera-Iparraguirre, A. D. Boese, W. Klopper and B. Ruscic, Chemical Physics, 2008, 346, 56-68. J. Aguilera-Iparraguirre, H. J. Curran, W. Klopper and J. M. Simmie, JournalofPhysical Chemistry A, 2008, 112, 7047-7054. W. Klopper, R. A. Bachorz, D. P. Tew, J. Aguilera-Iparraguirre, Y. Carissan and C. Hattig, JournalofPhysical ChemistryA, 2009, 113, 11679-11684. H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and M. Schitz, Wiley InterdisciplinaryReviews: ComputationalMolecular Science, 2011, 2, 242-253. S. Sharma, M. R. Harper and W. H. Green, CanTherm open-source sofiware package, (2010). 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. Chapter 5: Modeling the decomposition of alkylaromatic compounds 17. 18. V. Burkkd-Vitzthum, R. Michels, G. Scacchi and P.-M. Marquaire, Industrial Engineering Chemistry Research, 2003, 42, 5791-5808. P. E. Savage and M. T. Klein, Industrial & Engineering Chemistry Research, 1987, 26, 374-376. P. E. Savage and M. T. Klein, Ind. Eng. Chem. Res., 1987, 26, 374-376. & 16. 107 Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 108 Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 6.1 Abstract The automated Reaction Mechanism Generator (RMG) is used to simulate the decomposition of phenyldodecane in the presence of diethyl disulfide. Recent experiments have shown that the presence of diethyl disulfide has a statistically insignificant effect on conversion of the alkylaromatic over 72 hours at 350 'C, suggesting that increased concentration of this radicalproducing compound might not have the effect that was proposed by previous researchers. Detailed kinetic modeling allows us to investigate these results mechanistically, by modeling each elementary step of the process. Mechanisms were first built to model the individual pyrolysis mechanisms for phenyldodecane and diethyl disulfide. Both mechanisms were validated with the available experimental data. Diethyl disulfide is predicted to react within minutes, forming mostly ethane and hydrogen sulfide, as well as methane and carbon dioxide in the presence of water. An RMG mechanism including both compounds was then generated, and reactor simulation showed that accelerated phenyldodecane conversion only occurred during the intial minutes of the experiment, while disulfide conversion was in progress. Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 6.2 109 Introduction Fossil fuels are produced by the decomposition of plant and animal remains deep beneath the surface of the earth. In particular, kerogen is formed by the decomposition of algae, plankton, and woody plants. 1 This heavy, waxy material can decompose further to form petroleum and natural gas, the former of which is the basis for one of the world's most important industries. Unfortunately, despite advances in geochemistry and engineering, petroleum exploration remains a highly uncertain field, as the success of multibillion dollar drilling decisions is largely governed by chance. Improved knowledge of the chemical processes linking kerogen to oil to gas would be one significant step to better utilization of the planet's energy resources. Currently, decisions in oil exploration are assisted by the predictions of lumped kinetic models, which consist of simple mechanisms linking proposed reactants to products of interest. Rate parameters in these mechanisms are fit to match data from experiments conducted at conditions drastically different from geological conditions. This is necessary because useful experimental results at 200 'C could take millions of years to achieve, while the decomposition of similar compounds at 350 'C can be observed in hours. However, the extrapolation of parameters collected at these elevated temperatures to those of interest geologically introduces massive uncertainty into the model, as the relevant chemistry can differ greatly at 200 and 350 'C. Detailed reaction mechanisms, which model a process of interest as a series of elementary reaction steps, have the potential to provide more relevant models for use in conditions relevant in the kerogen maturation process. The open-source Reaction Mechanism Generator2 (RMG) allows a user to automatically generate a reaction mechanism for hydrocarbons containing oxygen, nitrogen, or sulfur. This software is able to propose new reactions that fit templates for known reaction types, and then estimate the necessary reaction rate constants using available experimental or ab initio data. The resulting mechanisms are completely free from bulk fitting to experiments, but new experimental data are extremely useful in the validation of these models. For this work, phenyldodecane was chosen as a representative model compound in crude oil, and diethyl disulfide was used as the sulfur contributor. The resulting high-temperature mechanisms were compared with experimental work to better understand the effects of sulfur at Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition Ito experimentally relevant conditions, and an additional reaction mechanism was generated at geologically relevant conditions to investigate whether sulfur had the same effect. 6.3 Methods 6.3.1 Reaction Simulation RMG 2 has previously been used to model a variety of processes relevant in fuel chemistry, including the combustion and pyrolysis of various candidate biofuels. 3 , 4 RMG builds these mechanisms using a rate-based algorithm, automatically proposing any possible reaction (fitting within the templates in the RMG database) for a given set of reactants. Arrhenius and thermochemical parameters are estimated for every reaction, and these data are used to estimate the rate of each reaction in the model. These rate estimations allow us to continue to follow reaction pathways that are predicted to proceed more quickly. For our initial investigation, we have neglected solvent effects. A reaction mechanism was first built for phenyldodecane conversion at 350 'C and 80 bar for a reaction time of 72 hours, and a separate mechanism was built for diethyl disulfide pyrolysis to 'ensure that the major pathways for each compound were followed. These phenyldodecane and diethyl disulfide mechanisms were combined, and then RMG was employed to predict the cross-reactions between the species contained in the two separate mechanisms, as well as the additional products of these reactions. Mechanisms were also generated for the reaction in the presence of water. For these runs, a "seed mechanism" was included to predict the formation of CO2 from the hydration of thioaldehyde compounds, which has previously been demonstrated to be important in the desulfurization of sulfides by supercritical water. 5 After mechanism generation was complete, reactor simulations were conducted using the "Closed Homogeneous Batch Reactor" model in CHEMKIN-PRO. 6 6.3.2 Quantum Calculations Previously calculated Arrhenius and thermochemical are available in literature for the decomposition reactions of organosulfur 740 and alkylaromatic 1 compounds, and these were available for model generation in this work. However, the RMG database still does not have accurate data available for 100% of the thousands or millions of reactions that could possibly be involved in oil-to-gas mechanism generation. Thus, an estimation algorithm is implemented to Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition III provide as accurate parameters as possible for a given reaction using the data available to RMG, even when specific rate parameters for a reaction are not available. Depending on the data available, these parameters can still be uncertain by multiple orders of magnitude. Reactions that are found to be important for overall mechanism accuracy can be refined using ab initio methods. Quantum calculations were conducted using the Gaussian 03 software package, using optimizations and frequencies computed at the B3LYP/6-31 1G(2d,p) level of theory. Single point energies were calculated using CBS-QB3 for all reactions, and CCSD(T)-F12/cc-pVTZF12 for reactions found to be particularly important to the mechanism predictions. This second method has been found to provide more accurate energies, 13-15 and these coupled-cluster calculations were conducted in Molpro.1 6 The CanTherm software package' 7 was then used to calculate reaction rate constants using transition state theory, including treatment of onedimensional hindered rotations and Eckart tunneling.' 8 6.4 Results and Discussion 6.4.1 Diethyl Disulfide Decomposition Mechanism To better understand the role of sulfur linkages in hydrocarbon cracking, it is first necessary to identify the important reactions in the decomposition mechanisms of these sulfur compounds. To this end, RMG was first used to build a small reaction mechanism for diethyl disulfide decomposition, at conditions applicable to the experiments done by Lewan 19 and in this work, but in the absence of phenyldodecane. Work has previously been conducted on the pyrolysis of dimethyl disulfide, 0 and many of the reaction rate constants and thermochemical parameters from that work are applicable to the current study. These data, as well as other calculated data for the reaction rates and thermochemical parameters of sulfur compounds,7- 9 were used to generate a pyrolysis mechanism for diethyl disulfide that is valid between 200 and 350 'C, but with phenyldodecane replaced by inert argon. The mechanism generation was set to terminate when 60% conversion of the disulfide was achieved, and the core tolerance was set at 10% of the characteristic flux. Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 112 The RMG mechanism contains 101 species and 3261 reactions, and the major reaction pathways at a reaction time of 60 seconds are presented in Figure 1. Scission of the S-S bond is an important radical generator within the first seconds of the simulation, and C-S bond scission becomes more important after this initiation phase (due to recombination reactions decreasing the net flux of the S-S scission pathway). At 60 seconds, 90% of the reacting disulfide is predicted to be consumed by a hydrogen abstraction, followed by a beta scission reaction to produce the ethanethiyl radical and thioacetaldehyde. Much of the ethanethiyl quickly decomposes to ethylene and a mercapto radical, which abstracts a hydrogen atom to form hydrogen sulfide. Thioacetaldehyde, which contains a C-S double-bond, is not as easily desulfurized, and it mostly participates in addition reactions to form stable intermediates, including a cyclic compound containing two sulfur atoms. These intermediates are predicted to react over a longer timescale to form more H 2 S, as well as stable thiophene. The conversion of diethyl disulfide over time is plotted for 200 to 350 'C in Figure 2. At the temperature studied experimentally, the reactant is 50% consumed within the first 15 seconds. This half-life increases to seven minutes, 11 hours, and 10 months at 300, 250, and 200 'C, respectively. H2 S generation occurs somewhat more slowly than reactant conversion, as presented in Figure 3. At temperatures above 300 'C, very fast H 2 S generation is predicted at the initial stages of the reaction time, with significantly slower generation over the remainder of the 72 hours. H 2 S is likely to behave as a chain-transfer agent and have some effect on hydrocarbon cracking products, but the presence of this compound itself is not likely to have a great effect on the total concentration of radical species in the reaction mixture, as the S-H bond is approximately as stable as a C-C bond. Finally, the concentration of various radicals formed in pyrolysis is presented in Figure 4. Reactive ethylthiyl (C 21 5S) and thioformylmethyl (C 2 H3 S) radicals reach a maximum within the first second of the experimental timescale and then decrease for the remaining 72 hours, while relatively stable S 2 (a diradical) and HS 2, which have low hydrogen affinity, are formed later on. HS Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 5% 0- 2H 5 + S + <1% 113 - RH 5%R 1*R' H RH 90% 4% RH 4% R' SH: + C HS S R' 90% )*RH 4% RH R' 77%~ SH77 77% :C2 H4 | RH 20%R R' S 42% 42% RH - + zS 42% S SH S |- SH SRH 77% . 4% 4% RH 42% RF S IS R' -SH 4 RH 4% R Figure 1. Decomposition pathway of neat DEDS at a temperature of 350 *C and a time of 60 seconds. Percentages are presented relative to total decomposition rate of DEDS at this time. Products in dashed boxes are predicted to be consumed through subsequent reactions, while ethane and hydrogen sulfide are predicted to be stable products. 114 Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 100 100 IA 0 wj 0 C 75 75 LI .- - -- 0 50 50 25 25 Of 0 ... . .... 0 0 48 24 100 0 0.001 72 0.1 0.01 1 10 100 Time (Hours) Time (hours) Figure 2. Predicted conversion of diethyl disulfide vs. time. 350 *C (solid), 300 *C (dashed), 250 *C, (double), and 200 *C (dotted) 1.00 Ifl 0 LU 0 0.75-ii 4.. 0.50I 0.250.00 0 24 72 48 Time (hours) Figure 3. Predicted production of H 2 S from diethyl disulfide pyrolysis. 350 *C (solid), 300 *C (dashed), 250 *C, (double), and 200 *C (dotted) 1.E-0 2 - - 1.E+u U 1.E-0 4 0 1.E-0 1.E-0 8 i.E-1 i.E-1 4 1.E-07 1.E-)4 1.E-Ol - 1.E+02 - .E- 2 i.E+i-5 Time (s) Figure 4. Concentration vs. time for representative radicals formed by DEDS pyrolysis at 350 *C. S2 diradical (black solid), HS 2 radical (black dashed), thioformylmethyl radical (gray solid), ethylthiyl radical (gray dashed) Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 115 Experimental data for disulfide pyrolysis are scarce, but the data collected by Lewan and in a previous study of dimethyl disulfide pyrolysis provide evidence that our predicted mechanism is reasonable. This model predicts approximately 90% conversion of diethyl disulfide in two hours at 590 K and 10 kPa, which is in good agreement with the 85% conversion observed experimentally by Coope & Bryce for dimethyl disulfide pyrolysis.2 1 This is evidence that the main reaction rate constants have been estimated with reasonable accuracy; therefore, we can continue explore the effect of diethyl disulfide on hydrocarbon cracking by modeling the pyrolysis of a mixture of this compound and phenyldodecane. To explore the role of water in disulfide decomposition, an additional reaction mechanism was generated to simulate the gold bag experiments conducted with phenyldodecane, diethyl disulfide, and water, but with phenyldodecane replaced by inert argon to simplify the initial simulations. Using the same termination criteria as the pyrolysis case, the resulting mechanism contains 248 species and 26,641 reactions. This is more than double the size of the pyrolysis mechanism in the absence of water, mainly due to the inclusion of the seed mechanism. The predicted decomposition pathway for diethyl disulfide at 350 'C and a time of 60 seconds is presented in Figure 4. The primary reaction steps are largely the same as in the predictions without water, with the most significant difference in the decomposition of the thioaldehyde intermediate. At 60 seconds, the thioaldehyde consumption pathway is similar to the case without water, but the hydration pathway will take priority over the full timescale. The hydration of this compound leads to the production of a mercapto-alohol, which then loses hydrogen sulfide to form acetaldehyde. This pathway can be catalyzed by the presence of additional water molecules.5 The aldehyde can then decompose to form carbon monoxide, which can then react to the experimentally observed carbon dioxide either directly via water gas shift, or by a free radical pathway that proceeds with a carbonyl sulfide (OCS) intermediate. The production of carbon dioxide from carbonyl sulfide has previously been studied,2 2 and can also be catalyzed in the presence of water. Overall, the generated mechanism predicts that approximately one mole each of ethane, methane, and other gas-phase carbon (in the form of carbon dioxide or carbonyl sulfide) to be formed from each mole of diethyl disulfide reacted. This agrees with experimental results, with only slight disagreement in the amount of carbon dioxide vs. carbonyl sulfide predicted, as no carbonyl Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 116 sulfide was observed experimentally. This is likely due to modeling uncertainties, as the water gas shift reaction kinetics were roughly estimated, and rate constants in the carbonyl sulfide to -~s* + <1% -~s* 9% C2H5 -S-8 + carbon dioxide pathway have uncertainties of approximately a factor of two. RH RH 9% 6% RH 87% R' R' 3% C 2H6 -'S' SH - 86% H 20 + 2% 5 2 4% 53% R- C 2H 4 ' RH r SH ~7 SRH 35% R' 20% 38% RH 35%5 ."SH + -3SH H 2S OH S --- 4\s 2% --- ----S <1% H2 0 <1 O <1% RH 2% R' H2*- if 0 2H 5 -1% d+ 2% OH 3 --- F +__ RH OH H20 <1% <1{ R*r I OH C 2 H6 O SH <1% 002 + HR 2S Figure 5. DEDS decomposition pathway at a temperature of 350 *C and a time of 60 seconds, in an equimolar mixture with water. Percentages are presented relative to total decomposition rate of DEDS at this time. Products in dashed boxes are predicted to be consumed later in the reaction timescale, while ethane and hydrogen sulfide are predicted to be stable products. Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 117 6.4.2 Phenyldodecane Decomposition Mechanism The kinetics of phenyldodecane (PDD) pyrolysis in the absence of any other additional reactants has been previously studied,2 3' 24 and thermodynamic data for liquid PDD at elevated temperatures and pressures are also available, allowing for a thorough interpretation of the behavior of this oil surrogate at experimental and reservoir conditions with additional reactants present. The short and long chain products of PDD cracking are also more easily recoverable than shorter chain alkyl substituted aromatics. A kinetic study of the similar phenyldecane has been previously conducted, including experimental and modeling work." The mechanism generated was largely similar to our initial RMG-predicted mechanism for phenyldodecane, and it included a calculated rate constant for the key reverse disproportionation reaction connecting styrene to ethylbenzene formation. In addition, rate parameters for a molecular elimination reaction to form toluene and an alkene (the retroene reaction presented in Figure 5) have been calculated prior to this work. Thus, the parameters for these reactions were added to RMG to generate an improved mechanism, and this improved mechanism was used as a starting point for investigating the effect of diethyl disulfide on conversion. H H Ph R I Ph R Ph- + R Figure 6. Retroene reaction of an alkyl benzene to form toluene and an alkene An RMG mechanism containing 365 species and 15,755 reactions was generated to model the experiment of Savage & Klein at 400 *C, 23 and the comparisons between model and experimental data for the reactant and major products are presented in Figure 6. Reasonable agreement with the experimental data can be seen in this figure, as the rate of PDD pyrolysis is predicted within a factor of two. Major product predictions from the RMG model have been compared with experimental data, and these comparisons along with RMG-predicted reaction mechanisms for these products are presented in Figure 6. The model qualitatively captures the concentration profiles of undecene and toluene. In both the experiment and model, undecene and toluene are produced at the same 118 Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition rate for early time points, as they arise from the same reaction mechanism (shown in Figure 6). As the experiment progresses, the mole fraction of toluene continues to increase while the mole fraction of undecene stabilizes (and decreases slightly in the experiment). RMG predicts undecene to be consumed primarily by radical addition reactions, forming larger hydrocarbon species. 1 X 0.75 Xx 0.5 IL X 0 - PDD Data PDD Model 0.25 0 240 180 120 60 0 Time (min) 0.3 x .2 0.2 -Undecene U- Model Toluene Model .9 A Undecene Data x Toluene Data &A A 0 X 0 80 60 40 20 % Conversion PDD q Ph.-MC 9 H1 9 RH C1H22 RH Ph CH19 4 Ph- f K 2 -1400Ph- Figure 7. Comparison plot (top) and predicted pathway (bottom) for production of undecene and toluene from neat PDD pyrolysis at 400 *C. Experimental data from Savage & Klein. Figure 7 shows the RMG predictions of three major products-decane, styrene, and ethylbenzene-produced by another pathway from PDD. This proceeds by abstraction of 119 Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition hydrogen neighboring the phenyl group, which then undergoes a beta-scission reaction to form the expected cracking products, decane and styrene. Experimental data suggests that the styrene proceeds to form ethylbenzene. This was automatically predicted by RMG: styrene can abstract a hydrogen atom from PDD in a reverse disproportionation reaction to form two stable radical species, leading to the formation of stable ethylbenzene. This pathway is underpredicted by the RMG mechanism, but ethylbenzene is still predicted within an order of magnitude of the experimental data. 0.09 C6 . 0.06 Decane Data X 0 X Styrene Data Ethylbenzene Data X 0.03 Decane Model - 1X -Styrene Model Ethylbenzene Model 0 0 80 60 40 20 % Conversion PDD RH q ClOHH 5 PhM- 9C H 19 ClOH22 RH RH Ph"" > C9H19 Ph- q RH q Ph-Ph styrene, and Figure 8. Comparison plot (top) and predicted pathway (bottom) for production of decane, ethylbenzene from PDD pyrolysis at 400 *C. Experimental data from Savage & Klein. at Comparison with the product distributions collected experimentally by Reeves after treatment 350 'C for 72 hours can be seen in Figure 8. Good agreement is observed for all major products expected except for undecane, which is significantly underpredicted by RMG. There are two to form pathways for formation of this product. The first is bond scission of the initial radical undecane. benzyl and undecyl radicals, the latter of which would abstract hydrogen to form Undecene produced in the major toluene production pathway in Figure 6 could also abstract for these hydrogen in a reverse disproportionation reaction to form undecyl radical. Parameters reactions are reasonably estimated by RMG, but these could be investigated further using more accurate ab initio methods to improve model performance. 120 Chapter 6: Modeling the effect pf sulfur on phenyldodecane decomposition 1000 100 10- % 0 -6 E 0.1 0.01 Figure 9. Comparison of experimentally observed (dashed, x) and RMG predicted (solid, A) products from pyrolysis of neat phenyldodecane at 350 *C for 72 hours. 6.4.3 Phenyldodecane Decomposition in the Presence of Diethyl Disulfide Pyrolysis mechanisms containing 95 and 70 species for phenyldodecane and diethyl disulfide, respectively, were found to provide similar product predictions as the full mechanisms analyzed previously, so these were used as seed mechanisms to build the full mechanism for the decomposition of phenyldodecane in the presence of diethyl disulfide, as studied experimentally by Lewan 19 and by Reeves. RMG was used to determine the additional hydrogen abstraction, addition, and substitution reactions that are possible between species in each mechanism, and model expansion continued using a core tolerance of 0.5. This mechanism was built to be applicable for the conditions and concentrations investigated by Lewan,19 from pure phenyldodecane to a 10:1 mixture (by mole) of phenyldodecane and water. The final mechanism includes 155 species and 2840 reactions. It was also built to be valid at temperatures between 200 and 350 *C, so that the resulting mechanism could be used to make predictions at geological conditions. The decomposition of phenyldodecane is predicted to occur by largely the same mechanism in the presence and absence of the disulfide, via the free radical and molecular pathways that mainly produce toluene, styrene, ethylbenzene, decane, and undedecane. However, as seen in Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 121 Figure 8, very little acceleration is predicted in the phenyldodecane decomposition rate when the concentration of disulfide is increased from two to eight percent in the initial reaction mixture, although some increase is predicted when increasing from no sulfur to two percent. The small predicted effect in decomposition rate could be due to the fast decomposition of diethyl disulfide, which was discussed previously, producing a large amount of relatively stable hydrogen sulfide (and possibly some very stable thiophene compounds). These products are not likely to impact the phenyldodecane decomposition rate after they are produced, although they might affect product selectivities. 100 0 0 00 5 o0 U 0 0 0.02 0.04 0.06 0.08 0.1 Initial Mole Fraction DEDS & Figure 10. Comparison of RMG predictions with experimental data for the conversion of phenyldodecane after 72 hours at 350 *C, in the presence of varying amounts of diethyl disulfide (RMG solid line rectangles, Lewan experiment white circles, Reeves experiment yellow circles). Error bars on Reeves data indicate 95% confidence intervals. The same reaction mechanism was used to simulate conditions relevant to the geological oil-togas process. The predicted effect of diethyl disulfide on phenyldodecane conversion at 200 'C is presented in Figure 10. Nearly the same sulfur effect is predicted at this temperature, as the disulfide conversion and H 2S production are largely complete within the first one percent of the overall reaction time. An increase in phenyldodecane conversion is predicted within this short period, which is illustrated in Figure 12, as the concentration of reactive radicals is greater. After the majority of disulfide has reacted to form stable products, phenyldodecane is predicted to react at same rate in all four simulations. Thus, the presence of other sulfides that produce Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 122 radicals more slowly (such as sulfides and thiols) may provide a more sustained increase in the radical concentration of the mixture. - 100 0 0 75 - (A - 50 0 0 a. 25 ' 0- 0 0.02 0.04 0.06 0.08 0.1 initial Mole Fraction DEDS Figure 11. RMG predictions for the conversion of phenyldodecane after 1000 years at 200 *C, in the presence of varying amounts of diethyl disulfide. 1001 1 75 50 - .2 C 0 - 25 0 I 0 5000 10000 15000 20000 Time (years) Figure 12. Predicted conversion of phenyldodecane (single lines) and diethyl disulfide (double line) vs. time at 9 200 *C with four different initial disulfide concentrations. Initial disulfide loading as described in Lewan' : 0 (black solid), 0.982 (black dotted), 1.767 (gray solid), 3.370 (gray dotted). Predicted disulfide conversion is nearly identical for the different runs, so only 0.982 case is presented. Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 6.5 123 Conclusions The automated Reaction Mechanism Generator has been used to model the cracking of a model oil component, phenyldodecane, in the presence and absence of diethyl disulfide. Individual decomposition mechanisms were generated for each of the two components, and these mechanisms were validated with experimental data before modeling the pyrolysis of a mixture of the two compounds. RMG predicts that diethyl disulfide reacts to form ethane, hydrogen sulfide, and some thiophene in the absence of water; in the presence of water, single-carbon species are predicted in addition to ethane and hydrogen sulfide. Much of this process occurs within the first minutes of the 72 hour experiment, accelerating the conversion of phenyldodecane for this short time due to the increased radical presence. However, as products of disulfide pyrolysis are formed, the radicals recombine with other small radicals and large phenyldodecane-derived radicals. After this time, much of the sulfur is present as H 2S, which does not have a significant effect on phenyldodecane conversion. 6.6 References 1. H. H. Schobert, in Chemistry of Fossil Fuels and Biofuels, Cambridge University Press, New York, 2013. W. H. Green, J. W. Allen, R. W. Ashcraft, G. J. Beran, C. A. Class, C. Gao, C. F. Goldsmith, M. R. Harper, A. Jalan, G. R. Magoon, D. M. Matheu, S. S. Merchant, J. D. Mo, S. Petway, S. Raman, S. Sharma, K. M. Van Geem, J. Song, J. Wen, R. H. West, A. Wong, H.-W. Wong, P. E. Yelvington and J. Yu, Reaction Mechanism Generator (RMG), (2013). J. W. Allen, A. M. Scheer, C. W. Gao, S. S. Merchant, S. S. Vasu, 0. Welz, J. D. Savee, D. L. Osborn, C. Lee, S. Vranckx, Z. Wang, F. Qi, R. X. Fernandes, W. H. Green, M. Z. Hadi and C. A. Taatjes, Combust. Flame, 2014, 161, 711-724. M. R. Harper, K. M. Van Geem, S. P. Pyl, G. B. Marin and W. H. Green, Combust. Flame, 2011, 158, 16-41. Y. Kida, C. A. Class, A. J. Concepcion, M. T. Timko and W. H. Green, Phys Chem Chem Phys, 2014, 16, 9220-9228. CHEMKIN-PRO 10131, (2013) Reaction Design, San Diego. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, ChemPhysChem, 2013, 14, 17031722. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, ChemPhysChem, 2013, 14, 37513771. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Phys Chem Chem Phys, 2012, 14, 12773-12793. 2. 3. 4. 5. 6. 7. 8. 9. Chapter 6: Modeling the effect of sulfur on phenyldodecane decomposition 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 124 A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Chemistry-A European Journal, 2011, 17, 7656-7673. V. Burkle'-Vitzthum, R. Michels, G. Scacchi and P.-M. Marquaire, Ind. Eng. Chem. Res., 2003, 42, 5791-5808. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. J. Aguilera-Iparraguirre, A. D. Boese, W. Klopper and B. Ruscic, Chemical Physics, 2008, 346, 56-68. J. Aguilera-Iparraguirre, H. J. Curran, W. Klopper and J. M. Simmie, Journal of Physical Chemistry A, 2008, 112, 7047-7054. W. Klopper, R. A. Bachorz, D. P. Tew, J. Aguilera-Iparraguirre, Y. Carissan and C. Hattig, Journal of Physical Chemistry A, 2009, 113, 11679-11684. H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and M. Schutz, Wiley Interdisciplinary Reviews: Computational Molecular Science, 2011, 2, 242-253. S. Sharma, M. R. Harper and W. H. Green, CanTherm open-source software package, (2010). C. Eckart, Physical Review, 1930, 35, 1303-1309. M. D. Lewan, Nature, 1998, 391, 164-166. A. G. Vandeputte, M.-F. Reyniers and G. B. Marin, J. Phys. Chem. A, 2010, 114, 1053 110549. J. A. R. Coope and W. A. Bryce, Can. J. Chem., 1954, 32, 768-779. C. Deng, Q.-G. Li, Y. Ren, N.-B. Wong, S.-Y. Chu and H.-J. Zhu, Journal of Computational Chemistry, 2007, 29, 466-480. P. E. Savage and M. T. Klein, Ind. Eng. Chem. Res., 1987, 26, 374-376. F. Behar, F. Lorant, H. Budzinski and E. Desavis, Energy & Fuels, 2002, 16, 831-841. Chapter 7: A thermochemical database for organosulfur compounds 125 Chapter 7: A thermochemical database for organosulfur compounds 7.1 Abstract A database of calculated gas-phase thermochemical parameters has been built for organic compounds containing sulfur. This includes sulfides, thiols, and thiophenes, as well as some oxidized sulfur compounds. Energy calculations were conducted using the high-accuracy CCSD(T)-F12a/cc-pVXZ-F12 method, where double- (X=D) and triple-zeta (X=T) basis sets were used depending on molecule size. A training set was chosen containing compounds for which reasonably accurate experimental enthalpy measurements were available. Calculated enthalpies were compared with these experimental values to regress a set of bond additivity corrections, which are then applied to the calculated enthalpies of the remaining compounds. Entropies and heat capacities were also calculated and validated against the available experimental data. The resulting parameter database can be used to generate new group additivity estimation schemes for studying the pyrolysis and oxidation of organosulfur compounds. ................... Chapter 7: A thermochemical database for organosulfur compounds 7.2 .126 Introduction Thermochemical parameters are critical to our ability to model reacting systems accurately. Gibbs free energies are required to calculate the equilibrium constant of a reaction, which is used to estimate the rate of a reverse reaction when forward rate parameters are available. Heat capacities are necessary to capture temperature changes in nonisothermal systems. Enthalpies, entropies, and heat capacities are generally determined most accurately via experiments, but this is not always feasible (particularly for unstable compounds). Because of this, quantum calculations have been a reasonable way to generate thermochemical data for a large number of species in a relatively small amount of time. Recent advancements, including the use of coupled-cluster calculations with F 12 corrections, have greatly improved the accuracy of the parameters estimated, particularly the enthalpies. When compared with accurate experimental data, such as the Active Thermochemical Tables (ATcT), 2, a systematic error is observed. This error generally scales linearly with molecule size, and it suggests that some This has frequently been accomplished correction can be added to decrease the uncertainty. using either atomic or bond additivity corrections (AAC's and BAC's, respectively), both of which have been shown to substantially decrease deviations between experiments and quantum calculations. 3-6 Substantial data are available for the estimation of thermochemical parameters for species containing carbon, hydrogen, and oxygen, 4' 6-8 and these have been used extensively in model generation for combustion and pyrolysis chemistry. 9' 10 This database was subsequently expanded to include organosulfur compounds 3 and some organosulfur compounds with oxygen." However, many of these early calculations were conducted using DFT methods that generally provided uncertainties greater than 1 kcal/mol for enthalpies of formation, even after additivity corrections were applied. The uncertainties for the parameter estimations of sulfur compounds are even greater, as accurate experimental data for these species are rare. Advances in coupled-cluster calculation methods12 ' 13 have made it possible to calculate single- point energies for molecules to much greater accuracy with reasonable computational cost. These have been validated for a variety of kinetic and thermochemical systems,14-1 6 and the implementation of additivity corrections improves the accuracy further for enthalpy Chapter 7: A thermochemical database for organosulfur compounds 127 calculations.1 7 Application of this technique to sulfur compounds, with additional experimental enthalpies for use as a training set, will allow one to calculate accurate thermochemical parameters for sulfur compounds in a reasonable amount of time. However, when building mechanisms considering millions of possible species, it is still not' reasonable to calculate thermochemical data for each compound using these methods. Methods for automatic quantum calculations without user intervention have made it feasible to perform these calculations for a greater number of species, but group additivity methods 7' 8 remain the best bet for estimating these parameters in the shortest possible time. Expansive databases of group additivity values (GAV's) are available for C/H/O and C/H/S compounds, but there is relatively little available for C/H/O/S species. These are particularly important when studying the oxidation of sulfur compounds, which occurs in combustion and oxidative desulfurization mechanisms 18-20 In this work, a database of enthalpies, entropies, and heat capacities for organosulfur compounds is presented. A set of 60 compounds for which reasonably accurate enthalpies are available are selected from literature, and calculations are conducted for these molecules using CCSD(T)-F 12 with the cc-pVDZ-F12 and cc-pVTZ-F12 basis sets to determine bond additivity corrections (BAC's). Thermochemical parameters are calculated for an additional set of molecules to generate a database of accurate calculations that can be used to regress GAV's and cyclic corrections, as well as hydrogen bond increments (HBI's) for radical species. 7.3 Methods Optimizations and frequency calculations were conducted using density functional theory in Gaussian 032. Molecules were optimized in their singlet state, with the exception of S2, which was found to be at its ground state as a triplet. Radical species were calculated as doublets. The CBSB7 basis set has been found to provide accurate geometries and force constants, so B3LYP/CBSB7 was used for most molecules in this work. Optimizations using CCSD(T)/63 l+G(d') were also performed when the converged CBSB7 geometry was in doubt. 22 Single-point energies were calculated using the Molpro software package . Based on the work by Buesser et al,17 where CCSD(T)-F12a/cc-pVDZ-F12 and CCSD(T)-F12a/cc-pVTZ-F12 were Chapter 7: A thermochemical database for organosulfur compounds 128 shown to provide highly accurate enthalpies of formation when BAC's were applied, these two basis sets were chosen for our work to achieve a balance between accuracy and computation time. Quadruple-zeta basis sets provide even greater accuracy, and these should be considered when the resources are available. Partition functions and thermochemical parameters were calculated using the open-source CanTherm software package23. A scaling factor of 0.99 was used for the frequency analysis. One-dimensional hindered rotations were also included, using scans in 10 degree increments at the B3LYP/6-31 lG(2d,p) level of theory. The effective moment of inertia ](2,3) was calculated for each rotor at the equilibrium geometry. In the final analysis, BAC's were applied to estimate the enthalpy of formation, and the entropy of formation and heat capacities between 300 and 2000 K were also calculated. 7.3.1 Regression of Bond Additivity Corrections (BAC's) BAC's are applied to minimize the error, c, between the calculated and experimental enthalpy of formation: AH2 98 ,expt = A H2 9 8,calc + nIc + 4 where ci is the correction for bond type i (such as a C=C double-bond) and ni is the number of that bond type present in the molecule. Experimental thermochemistry data for sulfur compounds were obtained from the NIST WebBook2 4 . Because these data were obtained by a large variety of research groups using different experimental methods, reported uncertainties vary widely among the dataset. As such, the enthalpies reported in the ATcT tables are known to much greater uncertainty, so the BAC's calculated for C/H/O compounds using these data are likely to have less uncertainty than those regressed using the available data for sulfur compounds. For this work, we will hold the BAC's constant at the previously regressed values for C/H/O compounds, and new BAC's will only be regressed for the bonds including a sulfur atom. 60 acyclic species with reported uncertainties below 0.7 kcal/mol were chosen as the training set for BAC regression. Three additional species containing a C=S double-bond were also included in this analysis, even though they have uncertainties greater than 0.7 kcal/mol. This Chapter 7: A thermochemical database for organosulfur compounds 129 was done due to the lack of experimental data for this type of compound, but it is important to remember the uncertainty introduced due to this decision. 7.4 Results and Discussion 7.4.1 BAC Regression and Validation 60 acyclic species for which gas-phase enthalpy measurements have been reported with uncertainties below 0.70 kcal/mol were selected for the training set from the NIST Chemistry WebBook.2 4 The cumulative density function (CDF) for the deviations of each calculation from the experiment is presented in Figure 1. The mean absolute deviation (MAD) for the original calculations is 1.62 kcal/mol, with an average deviation of 1.41 kcal/mol. This suggests a systematic overprediction of the enthalpy of formation, which should be partially resolved with the implementation of BAC's. Corrections for the S-H, C-S, C=S, S-S, 0-S, and O=S bonds were derived using leastsquares regression, decreasing the mean absolute deviation of the training set enthalpies from 1.62 to 0.72 kcal/mol. 80% of the species in the training set agreed within I kcal/mol of the experimental data, but compounds containing O=S double-bonds were found to have the greatest deviation from experiments, as shown in Figure 2. These deviations were on both ends of the error range, suggesting that a simple O=S correction would not correctly treat the calculation errors for sulfates, sulfites, sulfones, and sulfoxides. Thus, a modified bond additivity correction scheme (MBAC) was proposed, including an additional O=S=O correction to replace the O=S correction for species where the sulfur atom is double-bonded to two oxygen atoms. This additional fitting parameter decreases the mean absolute deviation further to 0.55 kcal/mol, providing the CDF in Figure 3. Note that while the highlighted compounds are much closer to zero error than in Figure 2, they still represent many of the compounds with the greatest disagreement with experiment. Thus, further work is necessary in accurately modeling the thermochemistry of these compounds. Chapter 7: A thermochemical database for organosulfur compounds 00 00 - 1 130 0.75 Q 0.5 0.25 0 -4 4 2 0 -2 Error (Calc - Expt) [kcal/mol] 6 Figure 1. CDF for AHf*(298 K) deviations between experiment and calculation, not including BAC's. 1 a P0 0.75 0.5 0.25 0 0 0 -4 00 2 0 -2 Error (Calc - Expt) [kcal/mol] 4 Figure 2. CDF for AHf*(298 K) deviations between experiment and calculation, including standard BAC treatment. Gray: sulfoxides and sulfites, Black: sulfones and sulfates, White: others. 131 Chapter 7: A thermochemical database for organosulfur compounds 10 V 0.75 Q 0.5 0.25 0 -4 2 0 -2 Error (Calc - Expt) [kcal/mol] 4 Figure 3. CDF for AHr*(298 K) deviations between experiment and calculation, including MBAC treatment. Gray: sulfoxides and sulfites, Black: sulfones and sulfates, White: others. The derived BAC's are presented in Table I for the double- and triple-zeta cc-pVXZ-F12 basis sets. The fitting statistics for the training set are presented in Table 2, and enthalpies and entropies of compounds thermochemistry modeled in the of additional training set are presented in Table species for which experimental thermochemical 3. The data were available, but were not included in the training set for MBAC regression, are presented in Table 4. Sulfides, disulfides, thiols, and cyclic sulfides (including thiophenes) are modeled with excellent accuracy with the MBAC scheme. In addition, entropies are reproduced well by these calculations, deviating by an average of 0.15 cal/(mol*K). Much larger errors are observed for enthalpy calculations of cyclic sulfones, as well as vinyl sulfide and the most Sn cyclic species. These calculations roughly agreed with those that had been reported previously in CBS-QB3 calculations,; so it is possible that errors in the geometry optimization are the cause of the deviations in enthalpy calculations. Thus, care should be taken when building models using thermochemistry estimations for these types of molecules. 7.4.2 Thermochemical Database The same calculations were calculated for an additional set of organosulfur compounds, for which no accurate experimental data were available. These compounds were selected for their potential relevance in the pyrolysis, SCW desulfurization, and oxidation of organosulfur Chapter 7: A thermochemical database for organosulfur compounds 132 compounds. The MBAC's regressed in the previous section were applied to the initially calculated enthalpies of formation, and these refined values are available at rmg.mit.edu. Standard entropies and heat capacities between 300 and 2000 K are also provided. The full database of 282 species is provided in the appendix of this chapter. Table 1. Regressed MBAC's (kcal/mol) regressed to model organosulfur thermochemistry. C-H, C-C, and C-O bonds were held fixed to previously regressed BAC from ATcT data. Regressed MBAC's C-H VDZ -0.49 C-C S-H C-S C=S S-S C-O 0-S O=S O=S=0 -0.71 0.87 0.42 0.51 0.86 -0.29 0.23 -0.53 1.95 VTZ -0.10 -0.35 0.52 0.13 -0.12 0.30 0.06 0.15 -2.61 0.27 Table 2. Fitting statistics for CCSD(T)-F12/cc-pVXZ-F12 enthalpy calculations (kcal/mol), where X=D or T. Method Mean Absolute Deviation Average Deviation Root Mean Square Deviation Number of Molecules Number of Fitting Parameters VDZ 4.73 4.47 5.63 59 0 VDZ-MBAC 0.66 0.02 0.89 59 7 VTZ 1.62 1.41 2.13 59 0 VTZ-MBAC 0.55 0.02 0.77 59 7 Chapter 7: A thermochemical database for organosulfur compounds 133 Table 3. Comparison of calculated (triple-zeta) and experimental enthalpies of formation and standard entropies for 59 gas-phase species selected for training set. Af1 0 -98 [kcal/mol] 34.0 MBAC 34.5 Experiment 46.76 This Work 45.9 -5.6 -4.6 49.18 67.3 68.6 28.4 28.7 -5.7 67.2 68.4 49.1 50.3 18.2 -15.5 17.3 -9.1 -11.3 -16.3 50.32 56.38 56.45 57.6 60.55 65.91 0.15 -17.8 -14.41 i 0.26 -25.99 Experiment SH H 2S -4.9 67 68 28.3 29.88 -5.46 :C=s HC'=S H 2C=S H 3 C-S' H 3 C-SH 0.1 2 2 HS 0.48 0.14 16.41 2 -8.96 i 0.48 -11.03 -16.39 t 0.15 HS -18.39 S S HS S HS- HS HS S' 298 [cal/(mol K)] No BAC 28.1 28.5 -5.3 56.4 55.1 57.1 61.1 65.4 69.68 68.8 69.63 70.9 80.1 -18.6 77.9 78.0 -13.5 -14.4 79.71 79.8 0.21 -23.6 -25.0 83.68 82.8 -22.97 0.19 -21.9 -23.3 87.3 -23.06 0.21 -21.9 -23.2 87.1 -18.8 -20.3 88.0 -20 0.55 -8.8 -11.1 S 86.6 -21.43 0.18 -20.2 -21.7 HS -26.49 0.42 -23.9 -25.8 98.3 HS-k -30.33 0.22 -30.9 93.1 HS -28.91 0.23 -27.5 -29.3 83.9 -27.4 0.28 -25.7 -27.6 95.7 0.23 -25.4 -27.3 96.1 HS -30.76 i 0.21 -29.8 -31.6 90.2 HS -24.42 -22.2 -22.9 -24.2 98.6 -25.03 0.26 0.19 -24.9 98.7 -28.93 0.18 -27.4 -29.4 0.6 -25.0 -27.0 95.4 0.23 -28.1 -30.5 107.7 -35.34 + 0.25 -33.9 -36.3 99.3 0.24 -33.5 -35.9 100.8 -27.42 HS \ -28 -30.9 HS -35.37 85.25 90.14 90.2 -1-1- .......... Chapter 7: A thermochemical database for organosulfur compounds 134 Table 3. Comparison of calculated (triple-zeta) and experimental enthalpies of formation and standard entropies for 59 gas-phase species selected for training set (Continued). ArH Experiment S 298 So 2 9 8 [caV(mol K)] [kcaVmo] No BAC MBAC Experiment This Work -29.1 -30.3 0.6 0.6 -26.5 -26.8 -29.1 -29.4 107.6 107.7 -35.3 0.6 -31.9 -34.4 98.7 -33.93 0.3 -31.4 -34.0 100.3 1.2 1.3 S 1.2 ^' S 6 S 0 53.1 -35.97 0.36 -36.0 -38.9 74.6 -49.1 0.4 -46.2 -50.3 92.7 -60.9 0.4 -54.5 -59.7 111.4 -65.5 0.4 -59.6 -64.8 102.2 -70.939 0.012 -71.2 -70.9 -102.61 0.62 -100.3 -101.5 91.4 -103.6 0.7 -101.2 -102.4 94.2 -115.5 0.5 -110.8 -113.6 90.1 -125.2 0.5 -121.0 -124.4 99.2 -131.9 0.46 -128.4 -132.3 106.9 -164.2 0.5 -165.2 -165.1 93.4 0.0024 -181.0 -182.1 108.6 30.4 30.7 27.6 27.3 59.327 59.4 0 o5 o d '-b 6 * s* 53.045 o o -180.7 SC= SoS HS- 30.74 27.95 0.26 SH HS -SH HS-' SH HS SH 54.54 56.88 54.6 56.9 -2.23 0.55 -3 .0 -2.4 -5.76 -5.8 -5.8 -7.13 0.28 0.44 -7.5 -7.5 80.7 91.3 -12.08 0.25 -11.4 -11.9 102.9 -17.85 -16.8 -17.9 98.0 -15.6 -14.8 -16.7 -16.2 111.3 -19.83 0.36 0.36 0.35 0.35 -17.5 -19.5 105.8 118.9 -28.04 0.26 -25.6 -27.9 116.3 -16.97 -15.58 83.8 80.48 Chapter 7: A thermochemical database for organosulfur compounds 135 Table 4. Comparison of calculated (triple-zeta) and experimental enthalpies of formation and standard entropies for 58 additional species. S AfH029 8 [kcaVmol] Experiment S rI SS No BAC MBAC 2 98 [caV(mol K)] Experiment This Work 19.7 0.24 19.1 18.6 61.0 14.6 0.3 15.8 14.7 68.7 2.7 0.3 11.6 10.5 70.0 0.24 29.5 26.6 66.6 25.33 t 0.96 42.0 39.2 21.67 0.30 21.3 19.0 71.3 20.86 0.28 21.8 19.5 70.9 -8.02 0.28 -6.4 -8.0 72.0 20.16 0.22 22.9 19.4 76.8 19.74 0.22 22.8 19.4 76.7 -15.18+ 0.25 -13.1 -15.2 77.5 -15.27 0.18 -12.3 -14.4 80.2 -14.47 0.20 -12.3 -14.4 80.6 -11.42 0.18 -9.0 -11.0 85.6 0.7 7.0 4.4 93.1 S 27.49 S S S S S 1/ S S S 81.43 81.6 OH S^ 4.3 S SH 26.86 0.21 30.8 26.3 80.5 SH -15.38 0.23 -11.6 -14.3 93.4 -22.88 0.19 -19.9 -22.5 87.0 -42.5 1.2 -41.0 -42.1 81.7 -32.33 0.47 -30.5 -32.3 78.0 -46.89 0.47 -38.9 -40.8 76.3 19.3 13.9 84.8 -52.9 -54.6 90.2 oSH S 0; o S 5.98 0.72 0 -54.45 0.22 Chapter 7: A thermochemical database for organosulfur compounds 136 Table 4. Comparison of calculated (triple-zeta) and experimental enthalpies of formation and standard entropies for 58 additional species (Continued). AfH Experiment 0 S 29 8 S 0 2 9 8 [caV(mol K)] [kcaVmol] No BAC MBAC Experiment -31.79 0.61 -36.4 -38.8 82.2 -37.10 0.65 -37.1 -39.5 82.4 -89.1 0.8 -89.6 -89.7 76.6 -29.7 0.7 -29.7 -31.1 74.6 -97.6 0.8 -94.8 -95.4 87.5 -61.1 0.4 -54.5 -56.5 81.7 -62.61 0.73 -53.3 -55.2 81.0 -31.2 -33.7 86.4 S 0 S 9=0 Sx. This Work 0 '0 sO , S -36 + 0.9 0 -71.5 0.48 -60.8 -63.3 88.8 S OH -69.7 0.52 -60.8 -63.3 88.8 \ -77.03 0.65 -73.0 -75.4 100.6 -109.87 0.63 -103.1 -104.8 106.4 -111.92 0.63 -108.1 -110.4 113.7 -113.1 0.9 -108.8 -110.5 97.5 -94.591 -95.3 -97.6 61.37 61.5 -175.7 -175.0 -174.2 71.41 74.6 25.23 26.0 17.9 60.88 17.3 25.2 17.7 60.9 70.9 -11.8 -11.2 -11.8 91.0 -18.1 -19.2 97.8 -26.8 -26.3 -28.0 101.4 -32._8 -31.6 -3.9 109.0 dl \ S-S S ..os - S -18.9 S < 1.0 70.82 Chapter 7: A thermochemical database for organosulfur compounds 137 Table 4. Comparison of calculated (triple-zeta) and experimental enthalpies of formation and standard entropies for 58 additional species (Continued). AfH%98 [kcaVmol] 0-S S S OI '>=OS S0 29 8 [cal/(mol K)] This Work Experiment Experiment No BAC MBAC _ -13.5 -13.0 -15.6 -5.3 -6.6 74.7 -29.3 76.9 39.8 40.7 63.2 1.6 50.7 49.7 77.1 0.48 28.6 28.2 79.5 15.8 15.6 92.9 28.2 27.3 87.7 34.84 52.6 53.8 71.8 26.14 29.4 30.9 77.9 24.36 36.0 37.8 85.5 27.17 26.3 28.4 94.0 24 32.7 35.1 109.8 -3.6 1.2 -30.1 1.2 63.788 63.9 S S 33.81 S S S S QS S 60.5 22.41 IfS 13.8 S S 18.53 S-S S-s 5S S S 0.64 S s S-S S-S' 7.5 Conclusions Thermochemical parameters have been calculated for an extensive set of organic compounds containing sulfur. A new set of bond additivity corrections were regressed to improve the agreement of calculated enthalpies with experimental data, and entropies and heat capacities were validated with available measurements. Calculations of thermochemical parameters were conducted for a set of additional sulfur compounds, to allow for the regression of a more accurate group additivity scheme, which can be used in thermochemical parameter estimation for pyrolysis and oxidation models involving sulfur. 138 Chapter 7: A thermochemical database for organosulfur compounds The best thermochemical estimation schemes combine high-level ab initio methods with accurate experimental data. In some cases, such as organosulfur chemistry, experimental data are deficient. More accurate experiments, and the inclusion of sulfur compounds in the Active Thermochemical Tables, 2 would allow for the regression of additivity corrections with much lower uncertainty. And while improved quantum chemistry methods, including multireference coupled cluster methods2 5 2 7 and improved treatments of anharmonicity,28, 29 may make the regression of bond additivity corrections redundant in the future, experimental data will remain necessary to confirm that organosulfur thermochemistry is correctly modeled in the future. 7.6 References 1. B. Ruscic, R. E. Pinzon, M. L. Morton, N. K. Srinivasan, M.-C. Su, J. W. Sutherland and J. V. Michael, J. Phys. Chem. A, 2006, 110, 6592-6601. B. Ruscic, R. E. Pinzon, M. L. Morton, G. von Laszevski, S. J. Bittner, S. G. Nijsure, K. A. Amin, M. Minkoff and A. F. Wagner, J. Phys. Chem. A, 2004, 108, 9979-9997. A. G. Vandeputte, M. K. Sabbe, M.-F. Reyniers and G. B. Marin, Chemistry-A European Journal, 2011, 17, 7656-7673. M. K. Sabbe, A. G. Vandeputte, M.-F. Reyniers, V. V. Speybroeck, M. Waroquier and G. B. Marin, Journal of Physical Chemistry A, 2007, 111, 8416-8428. G. A. Petersson, D. K. Malick, W. G. Wilson, J. W. Ochterski, J. A. Montgomery and M. J. Frisch, Journal of Chemical Physics, 1998, 109, 10570-10579. C. F. Goldsmith, G. R. Magoon and W. H. Green, Journal of Physical Chemistry A, 2012, 116, 9033-9057. S. W. Benson and J. H. Buss, Journal of Chemical Physics, 1958, 29, 546-572. S. W. Benson, F. R. Cruickshank, D. M. Golden, G. R. Haugen, H. E. O'Neal, A. S. Rodgers, R. Shaw and R. Walsh, Chemical Reviews, 1969, 69, 279-324. M. R. Harper, K. M. Van Geem, S. P. Pyl, G. B. Marin and W. H. Green, Combust. Flame, 2011, 158, 16-41. K. M. Van Geem, M.-F. Reyniers, G. B. Marin, J. Song, W. H. Green and D. M. Matheu, AIChE J., 2006, 52, 718-730. C. A. Class, J. Aguilera-Iparraguirre and W. H. Green, Submitted, 2014. W. Klopper and J. Noga, in Chemistry and Physics, ed. J. Rychlewski, Dordrecht, 2003. J. Noga and W. Kutzelnigg, Journal of Chemical Physics, 1994, 101, 7738-7762. J. Aguilera-Iparraguirre, A. D. Boese, W. Klopper and B. Ruscic, Chemical Physics, 2008, 346, 56-68. J. Aguilera-Iparraguirre, H. J. Curran, W. Klopper and J. M. Simmie, Journal of Physical Chemistry A, 2008, 112, 7047-7054. W. Klopper, R. A. Bachorz, D. P. Tew, J. Aguilera-Iparraguirre, Y. Carissan and C. Hdttig, Journal of Physical Chemistry A, 2009, 113, 11679-11684. B. Buesser, J. Aguilera-Iparraguirre and W. H. Green, In preparation. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. .................. Chapter 7: A thermochemical database for organosulfur compounds 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 139 M. T. Timko, E. Schmois, P. R. Patwardhan, Y. Kida, C. A. Class, W. H. Green, R. K. Nelson and C. M. Reddy, Energy & Fuels, 2014, 28, 2977-2983. X. Zheng, J. W. Bozzelli, E. M. Fisher, F. C. Gouldin and L. Zhu, Proceedings of the Combustion Institute, 2011, 33, 467-475. X. Zheng, E. M. Fisher, F. C. Gouldin and J. W. Bozzelli, Combust. Flame, 2011, 158, 1049-1058. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. Montgomery, J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, 0. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, 0. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, 0. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez and J. A. Pople, Gaussian 03, (2004) Gaussian, Inc., Wallingford CT. H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and M. Schiltz, Wiley Interdisciplinary Reviews: Computational Molecular Science, 2011, 2, 242-253. S. Sharma, M. R. Harper and W. H. Green, CanTherm open-source software package, (2010). http://github.com/GreenGroup/CanTherm P. J. Linstrom and W. G. Mallard, eds., NIST Chemistry WebBook, NIST Standard Reference Database Number 69, National Institute of Standards and Technology, Gaithersburg, MD. F. A. Evangelista, W. D. Allen and H. F. Schaefer, J. Chem. Phys., 2006, 125, 1-16. F. A. Evangelista, W. D. Allen and H. F. Schaefer, J. Chem. Phys., 2007, 127, 1-17. J. Pittner, J. Chem. Phys., 2003, 118, 10876-10889. J. Zheng, T. Yu, E. Papajak, I. M. Alecu, S. L. Mielke and D. G. Truhlar, Phys Chem Chem Phys, 2011, 13, 10885-10907. B. C. Garrett and D. G. Truhlar, Journal of Physical Chemistry, 1979, 83, 1915-1924. 140 Chapter 7: A thermochemical database for organosulfur compounds Appendix: Calculated Thermochemical Parameters 7.7 P AffFI so [kcaVmnol] 298 K -6.5 65.6 28.0 -5.7 [ca/(nol K)] 298 K 49.1 50.3 55.1 61.1 300 K 8.2 7.2 9.2 12.1 400 K 8.5 7.4 10.3 13.9 500 K 8.9 7.7 11.4 15.7 HS 18.8 -7.9 -10.1 -13.6 65.4 68.8 70.9 80.1 14.1 17.9 17.6 23.3 16.6 20.9 20.9 27.8 19.0 23.7 23.9 32.0 600 K 9.3 7.9 12.4 17.4 21.1 26.3 26.7 35.9 HS -15.9 78.0 23.5 28.3 32.6 36.4 42.5 -11.6 43.8 79.8 81.6 23.0 25.4 27.4 29.9 31.6 33.8 35.4 37.0 HS -20.8 82.8 31.0 37.4 42.8 HS HS -19.0 -18.9 87.3 87.1 29.3 35.2 29.2 -16.0 88.0 -17.3 [caY(nol K)] 800 K 1000 K 10.1 10.8 8.2 8.5 14.1 15.3 20.2 22.3 24.6 27.2 34.2 30.8 31.2 34.6 42.2 47.0 2000 K 12.1 8.7 17.1 25.8 31.1 39.6 40.0 54.2 12.8 8.8 18.1 27.6 33.2 42.5 42.8 58.0 13.1 8.8 18.6 28.5 34.2 43.9 44.1 59.8 47.1 54.3 58.0 59.8 41.7 42.1 46.4 45.8 53.8 51.3 57.7 54.2 59.6 55.6 47.3 54.4 59.7 68.2 72.8 75.1 40.8 45.7 53.7 59.6 68.6 73.3 75.5 35.3 40.8 45.7 53.7 59.6 68.6 73.3 75.5 28.7 34.3 39.7 44.7 52.8 58.9 68.1 72.9 75.2 86.6 29.4 35.3 40.6 45.4 53.1 59.0 68.1 72.9 75.3 -SH 5.3 85.6 27.1 34.9 42.1 48.5 58.5 65.6 76.1 81.3 83.8 10.6 -19.9 93.1 98.3 34.6 35.9 41.3 42.4 47.0 48.8 51.8 54.6 59.5 64.3 65.2 71.6 74.1 82.7 78.9 88.5 81.2 91.1 HS -25.2 93.1 35.3 43.0 49.9 55.8 65.4 72.4 83.0 88.6 91.3 HS -23.6 93.9 35.2 42.5 49.3 55.2 64.9 72.0 82.9 88.5 91.2 HS HS -21.7 95.7 35.3 42.7 49.4 55.3 64.9 72.0 82.9 88.5 91.2 -21.5 96.1 34.9 42.1 48.8 54.8 64.6 71.9 82.8 88.5 91.2 HS -25.9 90.2 35.5 43.3 50.2 56.2 65.7 72.7 83.2 88.7 91.3 90.9 90.9 :C=s H 2C=S H 3C-SH S HS S S S 2400 K 1500 K H2 S S -18.3 -19.0 98.6 98.7 35.1 34.3 41.9 41.3 48.4 47.9 54.2 54.0 63.9 63.9 71.2 71.2 82.3 82.4 88.2 88.2 S 7.6 90.2 35.6 43.3 50.0 55.7 65.0 71.9 82.5 88.2 91.0 -21.1 95.4 35.5 42.5 49.0 54.8 64.4 71.5 82.5 88.2 90.9 80.5 25.4 32.2 38.2 43.2 50.6 55.7 62.8 66.4 68.2 -7.1 93.4 33.1 42.1 50.4 57.7 69.3 77.7 90.1 96.3 99.3 -SH 15.3 87.0 31.6 41.2 49.9 57.6 69.6 78.3 91.0 97.4 100.4 HS -23.0 107.7 42.1 49.7 57.1 64.0 75.4 84.0 97.0 103.7 106.8 HS -29.0 99.3 41.4 50.3 58.4 65.4 76.7 85.0 97.5 103.9 107.0 HS -28.6 100.8 41.6 50.9 59.0 66.0 77.2 85.3 97.6 104.1 107.1 -21.6 -21.9 107.6 107.7 41.4 40.8 49.3 48.8 56.8 56.5 63.7 63.5 75.0 74.9 83.5 83.5 96.6 96.6 103.4 103.4 106.6 106.6 S ' SH I H34.4 S o S 141 Chapter 7: A thermochemical database for organosulfur compounds S ArH [kcalmoll 298 K So [cal(nol K)] 298 K 300 K 400 K 500 K -27.1 98.7 41.4 50.3 58.2 65.1 76.2 -26.6 100.3 42.1 50.7 58.4 65.1 76.0 C0 [caV(mol K)] 800 K 1000 K 600 K 1500 K 2000 K 2400 K 84.4 97.0 103.6 106.8 84.2 96.8 103.4 106.6 S -s -- -0.5 53.1 7.3 7.6 7.8 8.1 8.4 8.6 8.8 8.8 8.9 0* S -35.8 74.6 22.0 26.0 29.2 32.0 36.6 40.0 45.5 48.5 49.9 , SS -39.8 81.7 24.3 29.1 33.2 36.8 42.2 46.1 51.7 54.5 55.7 20.4 84.8 29.3 34.9 39.5 43.1 48.6 52.4 57.8 60.5 61.7 -50.7 90.2 28.0 34.2 39.9 44.9 52.8 58.3 66.1 69.9 71.5 -44.1 92.7 33.6 40.0 45.7 50.7 58.8 64.8 74.1 78.9 81.2 -50.4 111.4 45.4 54.0 62.1 69.3 80.9 89.6 102.6 109.3 112.5 -55.6 102.2 46.3 55.9 64.0 70.9 82.1 90.3 103.0 109.6 112.8 -74.5 59.4 9.6 10.4 11.1 11.7 12.5 12.9 13.4 13.6 13.7 -90.0 76.6 24.5 29.0 32.8 36.1 41.2 45.0 51.0 54.2 55.6 -94.1 87.5 29.8 35.7 40.8 45.2 52.2 57.4 65.2 69.3 71.3 -30.6 86.4 32.5 38.6 43.4 47.2 52.8 56.8 62.7 65.9 67.4 -63.4 91.2 33.2 39.9 45.5 50.3 57.8 63.1 71.1 75.1 77.1 -98.7 91.4 34.1 42.0 49.0 55.2 65.0 72.1 82.6 88.0 90.5 -99.7 94.2 36.0 43.3 49.5 54.9 63.5 69.8 79.4 84.5 86.9 -70.8 100.6 41.4 49.1 55.6 61.0 69.5 75.7 85.3 90.4 92.9 -100.4 106.4 41.9 49.9 57.3 63.8 74.3 82.1 93.8 99.8 102.6 -104.4 113.7 47.1 56.4 65.1 72.8 85.3 94.4 108.0 114.9 118.2 -106.3 97.5 42.4 51.4 58.9 65.3 75.3 82.7 94.0 99.9 102.7 -98.8 61.5 12.2 13.8 15.2 16.2 17.6 18.3 19.1 19.4 19.6 -111.0 90.1 27.3 32.2 36.7 40.5 46.3 50.2 55.8 58.4 59.6 -120.3 99.2 33.4 39.6 45.3 50.2 57.7 62.9 70.3 73.8 75.4 -126.6 106.9 39.4 47.1 54.1 60.2 69.5 76.0 85.1 89.4 91.4 -177.2 74.6 22.1 24.7 26.4 27.5 29.2 30.4 32.4 33.6 34.1 -165.6 93.4 31.2 35.8 40.0 43.7 49.7 54.1 60.6 63.9 65.4 -179.4 108.6 42.9 50.5 57.4 63.5 73.2 80.1 90.3 95.3 97.6 28.9 25.1 -3.6 54.6 56.9 83.8 7.8 10.8 23.8 8.1 11.8 27.2 8.3 12.5 30.2 8.5 13.0 32.9 8.7 13.7 37.2 8.8 14.0 40.5 8.9 14.5 45.6 8.9 14.7 48.2 8.9 14.7 49.4 -5.8 80.7 22.7 26.1 29.2 32.1 36.7 40.1 45.4 48.1 49.4 8 0 0^ S 0 d/ S d/ \ O.,. Ns LO \b 1\0 Od/ \OS'b S < ' \b 0 ~S HO SOH HO O o a & 'b S'-S* S=C=S HS-- SH 142 Chapter 7: A thermochemical database for organosulfur compounds ArH-i1 SO [kcaVmnol] 298 K [ca/(nol K)] 298 K 300 K 400 K 500 K -23.5 101.4 40.6 48.6 55.6 61.6 71.1 -14.3 118.9 45.6 54.0 62.0 69.2 -21.7 116.3 46.2 54.9 63.0 -27.9 109.0 47.2 56.5 -16.7 63.9 10.6 14.0 92.9 17.3 Co* [caV(mol K)J 1000 K 600 K 800 K 1500 K 2000 K 2400 K 78.1 88.6 94.1 96.6 80.9 89.4 102.1 108.6 111.7 70.1 81.7 90.1 102.7 109.0 112.0 64.7 71.7 82.9 91.1 103.2 109.5 112.4 11.4 12.1 12.5 13.1 13.3 13.6 13.8 13.8 31.4 36.1 40.1 43.4 48.5 52.0 57.1 59.5 60.7 77.0 22.9 26.2 29.1 31.7 36.2 39.8 45.5 48.5 49.9 11.4 82.5 29.3 34.7 39.1 42.7 48.5 52.9 60.0 63.9 65.7 59.9 74.3 19.6 22.6 25.1 27.2 30.5 33.0 37.0 39.1 40.1 54.4 81.4 25.4 29.9 33.7 36.8 41.8 45.5 51.2 54.3 55.7 20.0 67.6 15.3 18.1 20.5 22.5 25.6 27.9 31.5 33.4 34.3 12.5 75.3 21.0 25.2 28.8 31.8 36.5 40.0 45.6 48.6 50.0 63.6 14.8 63.0 84.2 14.2 29.3 16.0 34.4 17.4 38.6 18.4 42.1 20.0 47.9 21.1 52.3 22.9 59.6 24.0 63.6 24.5 65.4 9.0 89.8 35.0 42.1 48.0 52.8 60.2 65.7 74.2 78.8 81.2 57.0 82.6 25.6 29.8 33.3 36.4 41.3 45.0 50.9 54.0 55.5 51.8 89.9 31.8 37.2 41.8 45.8 52.4 57.3 65.0 69.2 71.2 16.7 75.4 22.5 26.2 29.4 32.1 36.6 40.0 45.5 48.5 49.8 9.6 82.5 28.4 33.6 37.9 41.5 47.4 52.0 59.5 63.5 65.4 60.6 72.7 19.6 22.6 25.0 27.0 30.2 32.7 36.7 38.9 39.9 32.0 88.5 30.9 39.0 46.1 52.2 61.4 67.7 76.9 81.4 83.5 86.8 79.3 23.2 26.7 29.6 32.0 35.8 38.5 42.6 44.7 45.7 57.5 93.6 34.4 43.4 51.1 57.6 67.4 74.1 83.5 88.1 90.3 133.7 73.8 21.2 24.3 26.4 27.9 29.9 31.3 33.8 35.3 36.0 101.4 91.6 33.0 40.8 47.3 52.7 60.6 66.1 73.9 77.7 79.5 HS- S HS-SH 1.0 -2.6 61.8 72.7 11.7 17.4 13.0 19.8 14.1 21.9 15.0 23.6 16.2 26.5 16.9 28.5 17.8 31.6 18.3 33.2 18.5 33.9 HS-S 20.3 87.7 26.7 30.7 34.3 37.4 42.3 45.9 51.3 53.9 55.2 64.6 84.3 24.9 28.3 31.0 33.2 36.5 38.8 42.4 44.3 45.2 34.0 101.0 36.2 44.5 51.7 57.9 67.2 73.6 82.5 86.8 88.8 3.1 73.5 17.1 18.9 20.3 21.2 22.4 23.1 23.8 24.0 24.0 25.6 27.9 29.8 34.8 37.6 38.9 39.4 36.5 17.7 39.2 18.9 32.8 43.6 20.6 46.7 21.8 51.4 23.5 53.8 24.3 54.8 24.6 ' S S HS HS> HS> HS HS - HS HS S K - - S S S / K> S-S S-S SS HS HS SH S/ -5-/HS S -0.9 -4.9 24.0 83.9 90.5 66.8 22.9 29.7 14.3 33.4 16.2 143 Chapter 7: A thermochemical database for organosulfur compounds ArfI 298 K so [caV(nol K)] 298 K 300 K 400 K 500 K 1500 K 2000 K 2400 K 18.2 46.4 75.5 80.7 19.1 23.3 22.5 27.4 25.4 30.7 27.8 33.4 31.5 37.2 34.1 39.8 37.7 43.4 39.4 45.1 40.1 45.9 93.0 78.2 21.3 24.2 26.3 28.0 30.4 32.1 34.4 35.5 36.0 61.5 96.2 32.8 40.8 47.6 53.3 61.5 67.1 74.6 78.0 79.5 54.3 81.7 20.8 24.0 26.7 28.9 31.9 33.6 35.3 35.7 35.8 HS---SH 28.3 1.9 78.1 73.1 19.7 19.8 21.9 21.4 23.6 22.7 24.9 23.9 26.7 26.0 27.8 28.0 29.2 31.2 29.7 33.0 29.8 33.7 HS '1,SH -4.1 81.8 25.4 28.5 31.1 33.4 37.3 40.4 45.5 48.1 49.3 25.8 75.7 21.9 25.3 27.7 29.5 32.0 33.8 36.9 38.7 39.6 19.7 82.6 27.9 32.9 36.6 39.4 43.5 46.4 51.2 54.0 55.2 31.1 89.9 31.4 39.4 46.5 52.4 61.5 67.9 77.0 81.5 83.6 33.8 100.5 38.5 47.9 56.1 62.9 73.4 80.7 91.2 96.5 99.0 15.1 76.0 19.4 23.2 26.9 30.2 35.5 39.4 45.3 48.3 49.7 10.3 81.8 25.7 30.9 35.7 39.9 46.7 51.7 59.4 63.5 65.4 S 3.7 85.6 31.7 38.8 44.8 50.1 58.4 64.5 73.9 78.8 81.1 S 42.1 68.6 17.2 21.1 24.4 27.2 31.4 34.1 37.8 39.6 40.6 S 88.5 68.1 15.9 18.4 20.4 22.0 24.3 25.8 28.2 29.6 30.2 56.2 82.0 27.5 35.2 41.9 47.5 55.9 61.4 68.9 72.6 74.4 54.1 68.4 15.6 18.6 21.1 23.1 25.9 27.5 29.1 29.7 30.0 9.9 73.0 19.0 22.8 26.4 29.6 35.0 39.0 45.1 48.2 49.7 34.5 77.8 23.6 28.4 32.5 36.0 41.5 45.4 51.3 54.3 55.8 79.5 76.7 21.4 25.0 28.1 30.7 34.8 37.7 42.2 44.6 45.7 49.3 90.9 33.2 42.2 50.0 56.4 66.2 73.0 82.6 87.2 89.4 45.3 79.3 21.4 25.4 28.8 31.7 35.9 38.7 42.6 44.4 45.2 15.6 75.9 19.3 22.5 25.1 27.4 31.0 33.6 37.4 39.2 40.0 13.3 85.2 27.7 32.5 37.0 40.9 47.5 52.4 59.9 63.8 65.6 8.4 92.1 33.4 39.9 45.6 50.6 58.6 64.6 73.9 78.8 81.1 0.9 98.6 38.9 47.0 54.1 60.4 70.3 77.6 88.6 94.4 97.1 78.5 91.3 28.3 32.9 37.1 40.9 46.8 51.1 57.5 60.7 62.3 63.2 84.9 24.7 27.8 30.4 32.6 36.0 38.5 42.3 44.3 45.2 29.9 96.6 37.2 44.4 50.5 55.8 64.0 70.1 79.5 84.5 86.9 46.2 102.1 35.7 41.2 45.6 49.1 54.4 58.0 63.1 65.6 66.7 [kcaYmol] HS, S.-S HS SH SH - HS C0 [ca/(mol K)] 600 K 1000 K 800 K S HS S S SH S S AN 5 NS 5 SH S ASH S NS---- SH '-S r S ' zzS S 144 Chapter 7: A thermochemical database for organosulfur compounds S ArHP So [kcalrnol] 298 K [caV(nol K)] 298 K 300 K 400 K 500 K 1500 K 2000 K 2400 K 14.7 85.1 24.5 29.2 33.4 37.0 42.6 46.4 52.0 54.7 55.9 6.0 92.9 28.8 35.1 40.8 45.7 53.3 58.6 66.2 69.9 71.6 11.5 85.5 24.8 29.4 33.3 36.8 42.1 45.9 51.5 54.3 55.6 26.9 93.8 34.2 42.0 48.9 55.0 64.3 70.9 80.4 85.1 87.3 28.1 85.2 29.7 35.9 41.1 45.5 52.4 57.5 65.2 69.2 71.2 34.6 76.7 22.7 28.1 32.7 36.5 42.4 46.4 52.1 54.8 56.1 75.5 86.4 26.9 31.9 36.2 39.9 45.7 50.0 56.4 59.7 61.3 61.2 109.3 39.4 48.0 55.7 62.2 72.1 78.8 88.0 92.4 94.4 96.6 89.5 27.5 30.8 33.0 34.6 36.8 38.2 39.9 40.8 41.1 38.2 89.1 26.1 31.0 35.5 39.3 45.5 49.8 56.1 59.2 60.7 -20.3 92.1 35.4 42.5 48.3 53.2 60.8 66.3 74.8 79.1 81.1 9.2 86.0 25.6 27.5 29.0 30.2 32.3 34.0 36.8 38.2 38.9 1.9 91.2 32.8 36.5 39.0 41.0 44.3 46.9 51.4 53.7 54.8 -10.9 86.3 31.6 36.5 40.4 43.8 49.2 53.3 60.0 63.5 65.1 -14.2 97.1 36.6 42.9 48.2 52.8 60.1 65.6 74.2 78.7 80.8 -1.2 99.3 38.1 42.9 46.8 50.1 55.3 59.3 65.4 68.7 70.2 71.9 108.2 43.5 56.7 68.0 77.3 91.1 100.5 113.5 120.1 123.2 128.8 119.6 47.3 56.9 65.1 71.9 82.2 89.2 99.0 103.7 105.8 43.9 126.7 47.3 55.7 63.0 69.1 78.3 84.6 93.1 97.2 99.0 -43.1 68.6 18.4 21.1 22.9 24.4 26.4 28.0 30.9 32.7 33.5 -50.7 75.8 24.6 28.7 31.7 34.1 37.6 40.3 45.1 47.9 49.2 -53.9 84.9 31.0 36.0 40.0 43.2 48.5 52.5 59.3 63.0 64.8 -58.8 82.5 31.0 36.5 40.7 44.0 49.0 52.8 59.4 63.1 64.8 -30.9 -27.8 64.5 62.3 13.5 11.7 15.2 13.9 16.7 15.8 18.0 17.5 19.9 20.0 21.3 21.6 23.2 23.7 24.0 24.5 24.3 24.8 -41.8 73.3 18.7 22.0 24.8 27.1 30.7 33.2 37.1 38.9 39.7 C0 [caV(mol K)] 800 K 600 K 1000 K SS HS S S S S 1 S" S S S HS SH HS rS H SH SH SH SH SH HSSH H S / SH S-SH S < SH HO__SH SH SH H H ,TSH OH O'SH HO-'S 0 ASH 145 Chapter 7: A thermochemical database for organosulfur compounds C ArHI S [kcaVmoll 298K [caV(mol K)] 298K 300K 400K 500K -38.0 71.0 16.7 20.1 23.3 26.1 30.4 45.3 82.7 25.1 29.3 33.0 36.2 -41.8 81.0 22.1 26.9 31.3 -89.6 69.5 16.8 19.7 -75.1 66.8 17.0 -101.2 14.2 94.5 76.4 -88.3 [cal(mol K)] 1000K 800K 600K 1500K 2000K 2400K 33.5 37.8 39.6 40.4 41.5 45.3 51.1 54.0 55.2 35.2 41.3 45.6 51.7 54.7 56.0 22.1 23.9 26.4 27.8 29.2 29.6 29.7 20.3 22.9 24.8 27.2 28.3 29.2 29.8 30.0 32.4 20.8 37.8 24.5 42.4 46.4 52.6 57.2 64.1 67.7 69.4 27.9 31.0 36.0 39.7 45.5 48.5 49.9 77.2 21.1 23.9 26.2 28.0 30.8 32.8 35.9 37.5 38.3 -45.7 -36.9 76.0 76.9 24.1 23.6 28.6 27.2 32.0 30.2 34.5 32.7 38.0 36.8 -39.7 -32.7 83.5 71.2 27.6 17.4 32.6 20.6 37.1 23.4 41.1 25.9 47.5 29.8 44.9 45.2 59.7 37.1 47.7 47.9 63.6 39.2 49.0 49.2 65.5 0 40.5 40.0 52.3 32.8 S -16.7 70.5 16.0 19.3 22.3 25.0 29.3 32.4 36.9 39.1 40.0 -46.6 80.2 22.3 27.3 31.8 35.7 41.7 45.8 51.6 54.3 55.5 79.6 21.9 26.2 30.4 34.3 40.7 45.4 52.0 54.9 56.1 50.7 90.2 28.0 34.2 319.9 44.9 52.8 58.3 66.1 69.9 71.5 -22.7 81.1 28.4 33.3 36.8 39.4 43.2 46.1 50.9 53.7 55.0 -8.6 93.9 38.1 47.0 54.1 59.7 67.9 73.6 82.2 86.7 88.8 OH S 0 SH OH 0 OH HS S OH SH HO HOH SH SH OH HO S OH 0 SH . .. T f-31.4 TSH H H / H '- H 40.0 146 Chapter 7: A thermochemical database for organosulfur compounds Cyclic Species Af11P [kcal/moll 298 K so [caV(mol K)] 298 K 300 K 400 K 500 K C0 P [caV(mol K)] 1000 K 800 K 600.K 1500 K 2000 K 2400 K 28.0 31.9 34.0 35.0 19.9 61.0 12.9 16.2 19.1 21.6 25.4 [ 17.3 68.7 17.2 22.1 26.6 30.4 36.4 40.7 46.9 50.0 51.6 S 13.4 70.0 18.9 23.5 27.5 31.0 36.4 40.3 46.1 49.2 50.8 31.0 66.6 17.5 22.7 27.0 30.4 35.3 38.5 43.1 45.5 46.6 23.3 71.3 20.0 25.8 30.8 35.1 41.5 46.0 52.5 55.8 57.4 24.0 70.9 20.0 25.7 30.8 35.0 41.5 46.0 52.5 55.8 57.4 72.0 21.6 28.2 34.2 39.4 47.5 53.3 61.8 66.0 68.1 25.3 76.8 23.0 29.1 34.4 38.8 45.6 50.3 57.0 60.6 62.3 25.3 76.7 23.2 29.2 34.5 39.0 45.7 50.4 57.1 60.6 62.3 -9.5 77.5 26.3 34.6 42.1 48.5 58.7 66.0 76.7 82.1 84.7 -8.7 80.2 27.6 35.7 42.9 49.0 58.7 65.6 75.9 81.3 83.8 -8.7 80.6 27.8 35.7 42.8 49.0 58.6 65.5 75.9 81.3 83.8 -28.4 78.0 23.6 29.7 35.1 39.7 46.5 51.3 58.1 61.7 63.3 -37.1 76.3 23.1 29.3 34.7 39.2 46.2 51.1 58.0 61.6 63.3 -33.2 82.2 28.1 36.0 42.8 48.6 57.5 63.8 73.0 77.6 79.8 -33.9 82.4 28.2 36.0 42.9 48.6 57.6 63.8 73.0 77.6 79.8 -30.1 74.6 22.2 27.5 32.0 35.6 40.9 44.3 48.9 51.5 52.7 -53.6 81.7 26.6 33.7 39.6 44.6 51.9 56.8 63.8 67.5 69.2 -52.5 81.0 26.6 33.6 39.6 44.5 51.8 56.7 63.8 67.4 69.2 88.8 32.6 40.8 47.8 53.7 62.6 68.9 78.0 82.5 84.7 88.8 32.6 40.8 47.8 53.7 62.6 68.9 78.0 82.5 84.7 -7.1 74.7 21.5 25.8 29.1 31.6 34.9 36.9 39.8 41.3 42.0 -29.7 76.9 23.2 28.3 32.5 35.9 40.9 44.2 49.1 51.6 52.7 S S 37.3 63.2 11.6 12.4 13.0 13.3 13.6 13.7 13.8 13.9 13.9 47.7 77.1 22.4 26.7 30.0 32.4 35.5 37.4 40.1 41.5 42.1 cs> :s S / S S S -3.8 01/ S S S S 0 0 0 9=0 S' -59.0 S IS -- O S CS-O 147 Chapter 7: A thermochemical database for organosulfur compounds So Cp* [kcaVnol] 298 K [caV(inol K)] 298 K 300 K 400 K 500 K [caV(nol K)] 1000 K 800 K 600 K 26.3 79.5 24.3 29.4 33.5 36.8 41.5 27.1 87.7 29.3 35.8 41.3 45.9 48.8 71.8 16.9 18.2 18.9 24.0 77.9 20.1 21.8 29.7 85.5 25.6 18.6 94.0 23.8 ArH- S CS S 1500 K 2000 K 2400 K 44.7 49.4 51.7 52.8 52.8 57.5 64.4 67.7 69.3 19.2 19.4 19.5 19.7 19.8 19.8 22.6 23.0 23.3 23.4 23.7 23.8 23.8 27.4 28.3 28.7 29.1 29.3 29.6 29.7 29.7 30.8 32.9 33.9 34.4 34.9 35.2 35.6 35.7 35.7 109.8 37.9 40.2 41.4 42.1 42.8 43.1 43.4 43.6 43.6 -17.2 78.1 25.1 31.3 36.5 40.7 47.0 51.2 57.5 60.8 62.4 -16.6 77.4 25.4 31.7 37.0 41.3 47.4 51.4 57.3 60.7 62.4 S S-S S-SS S S / SSI s-Ss S/ S-S \S \S-S' S OH S OH 148 Chapter 7: A thermochemical database for organosulfur compounds Radical Species H 3C-S' SH HS-S' S HCOS S S S HS S' H 2C-SH CHSH AfH0 So [kcaVmnol] 298 K 29.1 33.5 23.6 [caV(nmol K)] 298 K 57.1 45.9 60.9 300 K 9.6 6.9 9.5 400 K 10.9 7.0 10.2 500 K 12.3 7.0 10.7 C0 [caV(mol K)] 1000 K 600 K 800 K 13.6 15.7 17.3 7.6 7.1 7.4 11.9 12.3 11.2 1500 K 19.9 8.1 13.0 2000 K 21.3 8.4 13.3 2400 K 22.0 8.6 13.5 17.0 70.9 14.4 16.3 18.0 19.5 22.0 23.9 26.7 28.2 28.9 67.9 51.8 89.7 56.4 64.5 58.9 8.8 13.2 12.8 9.4 15.7 14.3 10.0 17.8 15.4 10.5 19.6 16.2 11.2 22.4 17.2 11.8 24.4 17.9 12.6 27.4 19.0 13.1 29.0 19.7 13.3 29.7 20.0 63.0 76.6 23.7 30.2 35.9 40.7 47.8 52.5 58.7 62.0 63.7 53.8 64.6 11.5 12.9 14.1 15.1 16.4 17.2 18.3 18.9 19.2 24.8 37.4 73.8 62.5 14.4 13.0 15.1 14.2 15.7 15.5 16.1 16.6 16.8 18.5 17.3 19.8 17.9 21.8 18.3 22.9 18.4 23.3 31.6 73.2 16.9 19.6 22.2 24.6 28.4 31.3 35.6 37.8 38.8 44.3 45.3 CHSH 47.3 76.5 20.4 24.1 27.4 30.2 34.5 37.6 42.1 CHSH 91.7 74.4 20.0 22.6 24.6 26.2 28.5 30.2 32.9 34.3 35.0 63.2 88.3 31.1 39.0 45.8 51.5 59.9 65.6 73.5 77.3 79.0 45.1 75.0 19.2 22.7 25.3 27.3 30.0 31.7 34.0 35.3 35.8 39.5 77.4 16.5 18.0 19.4 20.8 23.0 24.5 26.7 27.8 28.3 25.0 80.6 21.9 25.6 29.4 32.9 38.7 43.0 49.5 52.8 54.4 40.0 85.6 26.0 30.8 35.2 39.0 45.0 49.5 56.1 59.4 61.0 85.1 84.1 24.7 28.4 31.6 34.4 38.7 41.9 46.7 49.2 50.4 37.0 82.6 23.9 28.6 32.6 35.8 40.6 43.9 48.4 50.4 51.3 33.6 86.5 22.0 24.4 26.9 29.3 33.2 36.2 40.6 42.8 43.9 72.5 67.9 15.4 17.7 19.4 20.8 23.0 24.5 27.1 28.5 29.1 16.7 18.4 21.3 23.4 26.5 28.1 28.8 CH S CHH HS'CHSH SH SH C. SH O H SH SH 'CSH CIs S S S S S 58.3 66.5 12.8 14.7 25.8 68.0 16.0 19.1 21.9 24.4 28.3 31.3 36.0 38.5 39.6 20.9 75.0 21.7 26.4 30.4 33.9 39.5 43.7 50.2 53.6 5 5. 3 14.5 79.4 28.0 34.2 39.4 43.9 51.0 56.3 64.6 68.9 71.0 33.9 79.9 20.8 24.3 27.4 30.1 34.3 37.2 41.5 43.6 44.6 53.3 74.4 18.9 22.8 26.1 28.9 33.4 36.6 41.5 44.2 45.4 62.7 76.0 21.8 26.9 31.3 34.8 40.1 43.6 48.5 50.9 52.0 45.2 72.8 18.7 22.5 25.7 28.5 33.0 36.3 41.4 44.1 45.3 41.9 73.3 18.5 22.1 25.3 28.2 32.9 36.3 41.4 44.1 45.3 42.1 73.0 18.0 21.7 25.1 28.1 32.8 36.2 41.4 44.1 45.3 20.3 84.8 19.1 21.1 22.9 24.5 27.0 28.9 31.7 33.1 33.8 S S S S 149 Chapter 7: A thermochemical database for organosulfur compounds ArH [kcaVnol] 298 K so [caI/(nol K)] 298 K 300 K 400 K 500 K 56.9 86.7 29.1 37.0 43.8 49.5 58.1 S 56.4 85.4 29.0 36.9 43.7 49.5 l-SCH2 34.1 71.2 18.2 21.0 23.4 37.8 84.2 23.6 26.4 41.8 84.5 24.8 28.9 80.7 22.2 25.0 91.0 18.6 S' d O CH 2 HSO 'CH OH 2 Cp0 [caV(mol K)] 800 K 600 K 1000 K 1500 K 2000 K 2400 K. 64.1 72.8 77.1 79.1 58.1 64.1 72.8 77.1 79.1 25.5 29.0 31.7 36.1 38.5 39.6 28.9 31.1 34.5 37.1 40.9 42.9 43.8 27.5 29.7 31.7 34.9 37.5 41.5 43.7 44.8 26.1 29.8 33.2 38.8 42.9 49.3 52.7 54.3 28.1 33.2 38.2 42.7 50.0 55.4 63.7 67.9 69.9 99.6 33.5 39.6 45.6 51.1 60.3 67.1 77.6 82.9 85.5 73.7 103.2 34.9 40.4 45.7 50.6 58.7 64.8 74.1 79.0 81.3 54.6 77.5 18.4 21.5 24.6 27.5 32.1 35.6 40.6 43.2 44.4 -7.2 67.9 13.8 16.2 18.2 19.9 22.4 24.1 26.7 28.1 28.9 -14.2 75.1 20.0 23.8 26.9 29.5 33.6 36.5 41.0 43.3 44.5 -17.8 83.9 25.8 30.8 35.2 39.0 44.8 49.1 55.4 58.6 60.2 -14.4 86.7 28.3 33.5 37.5 40.8 45.7 49.2 54.8 57.8 59.4 -6.3 88.1 29.4 34.1 37.8 40.9 45.7 49.2 54.8 57.9 59.5 7.8 64.2 11.2 12.4 13.5 14.4 15.8 16.7 18.1 18.8 19.1 -4.2 72.8 15.9 18.6 21.0 23.1 26.3 28.6 32.0 33.8 34.6 -7.6 82.6 22.0 25.6 28.9 32.0 36.9 40.6 46.1 48.9 50.2 HO OH - $ S.> - S OH OH "SH OH SH O S' * S 18.4 110.2 40.0 47.6 54.9 61.4 72.2 80.1 92.1 98.2 101.2 41.9 73.3 18.5 22.1 25.3 28.2 32.9 36.3 41.4 44.1 45.3 -10.8 78.4 22.8 26.1 28.9 31.1 34.7 37.3 41.2 43.1 44.1 2.0 74.5 20.7 24.6 27.7 30.4 34.4 37.3 41.6 43.9 44.9 OH SH 6 ',SH Chapter 8: Conclusions Chapter 8: Conclusions 150 150 Chapter 8: Conclusions 8.1 Summary This thesis has provided an important step toward the detailed modeling of complicated fuel mixtures that include sulfur compounds. Where it was previously only possible to build speculative mechanisms manually, reasonably accurate kinetic mechanisms can automatically be generated for a variety of systems involving organic sulfur compounds. The already-extensive database for organosulfur chemistry has been expanded to include the kinetics of reactions involving both sulfur and water, and the thermochemical database has been expanded to make it possible to model oxidative sulfur chemistry. As more accurate rate estimation schemes continue to take hold in the chemistry community, the scope and accuracy of these methods is sure to increase. The sulfur database in RMG has been validated by a variety of experimental studies. Accurate pyrolysis models have been generated for t-butyl sulfide, making this the largest sulfur compound for which quantitatively accurate pyrolysis models are available. RMG was able to identify the pyrolysis and supercritical water desulfurization mechanisms of hexyl sulfide, confirming water's role as a desulfurization agent as well as an inhibitor in the formation of Chapter 8: Conclusions 151 aromatics in the high-temperature treatment of crude oils. Automatic mechanism generation has also proven useful in geochemical modeling, allowing us to study the effects of sulfur compounds on phenyldodecane decomposition, at timescales far beyond those that can be studied experimentally. 8.2 Recommendations for future work As automatic mechanism generation continues to expand to modeling new, industrially relevant systems, accurate kinetic models will be desired for larger and larger compounds. These molecules provide many more opportunities for bonds to be broken, hydrogen atoms to be abstracted, and cycles to be formed; all of this can be quite taxing on the available memory. Mixtures of multiple large species compound the problem further, causing RMG's memory requirement to greatly exceed what is available. Even models for relatively light compounds in crude oil, such as hexyl sulfide, should predict the formation of larger aromatic species in certain conditions: as there is no single "coke" molecule, a pyrolysis system may produce millions of different large molecules in concentrations too small to be considered individually. However, the sum of these species can affect the overall chemistry greatly, and improvements in RMG are necessary to model these systems accurately. "Pruning" (removing some edge products from a model when a flux does not reach a certain cutoff point) should allow RMG to include more important products, but changes in how the structure and reactivity of large species are represented in RMG will let us build models that capture the most information possible. One representation, which has previously been used in coal modeling, consists of modeling a large organic molecule as a group of unreactive nodes (consisting of fused aromatic rings) linked by reactive alkyl chains that may also contain oxygen or sulfur. In this treatment, the exact geometry of the nodes is unimportant, and reactions only need to be considered for the linking chains. Additional lumping can be done if necessary, and the memory saved should allow for the quantitative modeling of more complicated chemistries than what has been investigated previously. As the modeling of larger and larger species becomes feasible, these compounds are likely to remain in the liquid phase, even at the high temperatures frequently studied in pyrolysis and . .... Chapter 8: Conclusions 152 combustion chemistry. Significant efforts have already been undertaken to facilitate mechanism generation for these systems,' and the addition of sulfur groups to these solvation models will allow us to generate reasonable reaction mechanisms for processes involving large sulfur molecules in the liquid phase. As future improvements in automated reaction mechanism generation schemes and ab initio methods make it possible to model the kinetics of larger and larger molecules, the role of uncertainty in these models will continue to increase. Errors in thermochemical property calculations tend to be approximately proportional to molecule size, 2 and as these uncertainties can be mitigated with additivity schemes, the importance of accurate experimental data for benchmarking purposes becomes paramount. This problem has largely been solved for compounds containing carbon, hydrogen, oxygen, and nitrogen (among others) by the Active Thermochemical Tables, 3 but the problem of high experimental uncertainties remains for sulfur compounds. An accurate experimental database for sulfur chemistry will greatly broaden the horizons for the accurate modeling of organosulfur systems, whether they are petrochemical, geological, environmental, or biological. 8.3 References 1. 2. A. Jalan, R. H. West and W. H. Green, J. Phys. Chem. B, 2013, 117, 2955-2970. G. A. Petersson, D. K. Malick, W. G. Wilson, J. W. Ochterski, J. A. Montgomery and M. J. Frisch, Journalof Chemical Physics, 1998, 109, 10570-10579. B. Ruscic, R. E. Pinzon, M. L. Morton, G. von Laszevski, S. J. Bittner, S. G. Nijsure, K. A. Amin, M. Minkoff and A. F. Wagner, J. Phys. Chem. A, 2004, 108, 9979-9997. 3.