(238c) Transforming Chemistry From Postdictive to Predictive: Extensible Automated Reaction Mechanism Generation for Butanol Pyrolysis

Chemical kinetics' paradigm is primarily postdictive: the ignition time of a potential alternative fuel source as a function of temperature is obtained by running a series of experiments and fitting the data to a global kinetics model. These set of experiments consume substantial financial and personnel resources, including the fuel. Unfortunately, many obstacles arise that may cause the series of experiments, and thus the alternative fuel source, to be deemed unsuccessful, e.g. slow ignition time or increased particulate formation. Sometimes, a successful set of experiments can be cause for even greater concern, because there is usually very little understanding of why the fuel performed in the manner that it did: Will all C₄ alcohols behave similarly; what is the effect of adding an additional carbon atom to the structure's backbone; and how will impurities affect the results, are questions that may arise and, until recently, could only be answered by running additional experiments.

Shifting the chemical kinetics' paradigm from postdictive to predictive involves obtaining validated, detailed chemical mechanisms, or a system of coupled ordinary differential equations (ODEs), that accurately predict the fuel's performance[1, 2]. Answering the previously posed questions then only requires changing the initial conditions or parameters passed to the ODE solver, without having to run a single experiment. Constructing these detailed chemical mechanisms by hand is possible, however this method is difficult to repeat, extremely time costly, and susceptible to human error. Over the past decade, a handful of groups have developed automated tools to facilitate a more robust means of constructing reaction mechanisms[3]. Herein we discuss the algorithm and applications of a free, open-source, extensible software package, Reaction Mechanism Generator[4, 5].

The Reaction Mechanism Generator (RMG) software package is an automated, rate-based kinetic model generator that constructs pressure-dependent reaction mechanisms for isothermal, isobaric batch systems. The user must supply: the system temperature and pressure; the initial species concentrations; and a termination goal, either time of reaction or species conversion; the user may also input a list of preferred species' thermochemistries and a list of preferred reaction rate coefficients. The output files of a RMG simulation include species concentration profiles and a reaction mechanism (in the form of a CHEMKIN file); pressure-dependent reactions are represented by Chebyshev polynomials and high-pressure limit reactions are represented by modified Arrhenius parameters.

The RMG software enlarges the reaction mechanism one species at a time, starting with the initial species. For a given RMG iteration, the software has two species classifications: (1) ?core species,? or the species that RMG predicts will be significant in the overall reaction mechanism, and (2) ?edge species,? or all other species generated in the mechanism. Utilizing a list of extensible reaction family templates, the RMG software generates all possible reactions for the ?core? species present in the mechanism. For each species generated, its rate of formation, as defined by the change in concentration with respect to time, is computed. The largest of these rates is then compared to the ?minimum rate of formation,? R_min, derived using a user-defined tolerance. If the rate of formation is greater than R_min, the species is labeled a ?core? species; if not, it remains an ?edge? species. The preceding represents a single iteration in a typical RMG simulation. The software continues this process until the user-specified time of reaction or species conversion is reached.

To perform the rate analysis and to predict species concentrations as a function of time, the software requires reaction rate coefficients and thermodynamic parameters for all reactions and species generated in the model, respectively. When a new species is generated, RMG initially checks for any user supplied thermochemical parameters; this allows the user to override any thermochemical value RMG would implement in a mechanism with their personal preference. If the user has not supplied thermochemical data for the given species, RMG next checks against the community's recommended data set, as defined by the Process Informatics Model (PrIMe) project (discussed in the next paragraph). If this search is also unsuccessful, RMG will estimate the species' thermochemistry using Benson's group-additivity approach. A similar hierarchy is employed for estimating reaction rate coefficients.

The PrIMe project [6, 7] is a community-collaborative initiative that collects, stores, and validates data with the ultimate goal of processing and assembling the data into kinetic models. Two components of the PrIMe project are the data depository, which houses the community's raw data, and the data library, which stores evaluated data. The data library is the data set alluded to previously; RMG mines this data set for thermochemical data after any user-specified inputs but before estimation routines. RMG and PrIMe communicate using a unique species identifier, the IUPAC International Chemical Identifier (InChI). RMG's description of a species is converted to an InChI, which is used as the query when mining the PrIMe data library. The PrIMe data depository has approximately 11,000 species, over 7,000 of which are labeled with an InChI, including all C/H/N/O species. Mining the PrIMe data library is performed offline and all evaluated data is converted into appropriate RMG format and stored within the RMG database.

Case studies on 1- and 2-butanol have been performed. While there is serious commercial interest in developing 1-butanol into a high-volume, bio-based alternative transportation fuel[8, 9], 2-butanol is manufactured primarily as a precursor to the industrial solvent methyl ethyl ketone. A detailed pressure-dependent reaction mechanism was constructed for both species using the RMG software and PrIMe data library. The mechanisms were validated against experimental data, including new pyrolysis experiments performed in a plugged-flow reactor at the Laboratory of Chemical Technology of Ghent University and butanol-doped methane diffusion flames previously reported by McEnally and Pfefferle[10]. The differences in the two C₄ alcohol decomposition pathways will be discussed.

1. Green, W.H., Building and solving accurate combustion chemistry simulations. Nihon Nensho Gakkaishi, 2008. 50(151): p. 19-28.

2. Van Geem, K.M., et al., Automatic reaction network generation using RMG for steam cracking of n-hexane. AIChE Journal, 2006. 52(2): p. 718-730.

3. Pierucci, S. and E. Ranzi, A review of features in current automatic generation software for hydrocarbon oxidation mechanisms. Computers & Chemical Engineering, 2008. 32(4-5): p. 805-826.

4. Song, J., Building robust chemical reaction mechanisms: next generation of automatic model construction software, Ph.D. thesis in Chemical Engineering. 2004, Massachusetts Institute of Technology: Cambridge, MA.

5. Green, W.H., Reaction Mechanism Generator (RMG), http://rmg.sourceforge.net/. 2009.

6. Frenklach, M., PrIMe: Process Informatics Model, http://www.primekinetics.org/. 2006-2009.

7. Frenklach, M., et al., Process Informatics Model (PrIMe): A customer for ThermoML. Abstracts of Papers, 231st ACS National Meeting, Atlanta, GA, United States, March 26-30, 2006, 2006: p. CINF-036.

8. Technology Review: BP's Bet on Butanol, http://www.technologyreview.com/energy/18443/.

9. ButylFuel, LLC http://www.butanol.com/.

10. McEnally, C.S. and L.D. Pfefferle, Fuel decomposition and hydrocarbon growth processes for oxygenated hydrocarbons: butyl alcohols. Proceedings of the Combustion Institute, 2005. 30(Pt. 1): p. 1363-1370.