Installation

The kinetics modeling package is currently in alpha-testing and is available only via SVN. In order to access the package you will need the following dependencies:

svn (for installing and updating)
python (version 2.5 or greater)
python modules:
*numpy
*scipy
*mpmath
*ase
*gmpy (optional - gives 2-3x speedup for multiple precision calculations)

You can then install the module via svn with the following command in the terminal:

svn co svn://suncatls1.slac.stanford.edu/kinetics/trunk <packagename>

where <packagename> is the name of the local directory where the module will be stored. After installing you can get the most current version of the package by going into the <packagename> directory and typing:

svn update

As the package is currently under development, some issues and bugs are to be expected. If you are interested in working on the module as a developer please contact Andrew Medford (ajmedfor@stanford.edu) to obtain read/write access.

Documentation

Descriptor based analysis is a powerful tool for understanding the trends across various catalysts. In general, the rate of a reaction over a given catalyst is a function of many parameters - reaction energies, activation barriers, thermodynamic conditions, etc. The high dimensionality of this problem makes it very difficult and expensive to solve completely and even a full solution would not give much insight into the rational design of new catalysts. The descriptor based approach seeks to determine a few "descriptors" upon which the other parameters are dependent. By doing this it is possible to reduce the dimensionality of the problem - preferably to 1 or 2 descriptors - thus greatly reducing computational efforts and simultaneously increasing the understanding of trends in catalysis. 

The "kinetics" Python module seeks to standardize and automate many of the mathematical routines necessary to move from "descriptor space" to reaction rates. The module is designed to be both flexible and powerful. A "reaction model" can be fully defined by a configuration file, thus no new programming is necessary to change the complexity or assumptions of a model. Furthermore, various steps in the process of moving from descriptors to reaction rates have been abstracted into separate Python classes, making it easy to change the methods used or add new functionality. The downside of this approach is that it makes the code more complex. The purpose of this guide is to explain the general structure of the code, as well as its specific functions and how to begin using it. 

Useful Definitions: (Note symbols may not appear in Safari/IE)

• descriptor - variable used to describe reaction kinetics at a high level. Most commonly these are the binding energies of atomic constituents in the adsorbates (e.g. carbon/nitrogen/oxygen adsorption energies) or other intermediates. However, in the general sense other variables such as electronic structure parameters (e.g. d-band center) or thermodynamic parameters (e.g. temperature/pressure) could also be descriptors.
• descriptor space ( ?? ) - the space spanned by the descriptors. Usually ?¹ or ?².
• parameter space ( ? ) - the space spanned by the full reaction parameters for the reaction model. Usually ?²ⁿ where n is the number of elementary steps (one reaction energy and one reaction barrier per elementary step).
• reaction rates ( r? ) - an n-vector of rates corresponding to each of the n elementary reactions.
• descriptor map - a map of some variable (rates,coverages,etc.) as a function of descriptor space.
• reaction model - a set of elementary steps and conditions which define the physics of the kinetic system, along with the mathematical methods and assumptions used to move from "descriptor space" to reaction rates.
• model definition - a file used to define the reaction model
• input file - a file used to store other data which is likely common to many reaction models (e.g. energetics data)

Code Structure: 

• The interface to the code is handled through the ReactionModel class. This class acts as a messenger class which broadcasts all its attributes to the other classes used in the kinetics module. The class also handles common functions such as printing reactions/adsorbates or comparing/reversing elementary steps. The attributes of ReactionModel are automatically synchronized with the parser, scaler, solver, and mapper so it is useful to think of ReactionModel as a "toolbox" where all the necessary information and common functions are stored. 

The Parser class serves to extend the "model definition" file by reading in various quantities from an "input file". Technically the use of a parser is optional, but in practice it is extremely helpful for reading in common data such as adsorption energies or vibrational frequencies rather than re-typing them for every reaction model. 

The process of creating a "descriptor map" is abstracted into three general processes, which are handled by the following classes within the kinetics module: 

Scaler: Projects descriptor space into parameter space: ?? ? ? 

Solver: Maps parameter space into reaction rates: ? ? r? 

Mapper: Moves through descriptor space. This becomes important for practical reasons since it is often necessary to use a solution at one point in descriptor space as an initial guess for a nearby point. 

The ThermoCorrections class is responsible for applying thermodynamic corrections to the electronic energies which are used as direct inputs. This includes the contributions of entropy/enthalpy and zero-point energy due to temperature, pressure, or other thermodynamic variables. There are also a variety of analysis classes which allow for automated analysis and visualization of the reaction model. These include the RateAnalysis which creates "volcano plot" (descriptor maps of the rate), CoverageAnalysis which creates descriptor maps of the coverages of various intermediates, ScalingAnalysis which can give a visual representation of how well the scaler projects descriptor space to parameter space, and several other useful tools.

Using the code: 

Some examples can be found in the "demos" folder, and these should explain the syntax necessary and serve as a good starting point. The currently implemented features are also briefly described below in order to allow a better understanding of the demos and creating original reaction model definitions. 

Using the kinetics module to conduct a descriptor analysis requires (at least) 2 files: the "model definition" which defines the reaction model, as well as another script to initialize the ReactionModel class and conduct analyses. The "model definition" is generally a static file (i.e. it is not a program) while the submission script will be an actual python program. Model definition files typically end in ".mkm" for micro-kinetic model (although this is not required) while submission scripts end in .py since they are just python scripts. In addition it is very useful to also have an "input file" which contains the raw data about the energetics of the reaction model. An example of such an input file is provided in "published_data.txt" in the demos folder. 

Each class described in the Code Structure section will require some specialized parameters. Some of these parameters are common to all variants of these classes, while others are specific to certain implementations. The possible inputs for the currently implemented variants of each class are listed below. Required attributes are underlined. 

• ReactionModel:

  • +rxn_expressions\*+&nbsp;\- These expressions determine the elementary reaction, and are the most important part of the model. They must be defined unless the elementary_rxns, adsorbate_names, and transition_state_names are all explicitly defined since the rxn_expressions are parsed into these 3 attributes. It is much easier to just define rxn_expressions, although it is important to note the syntax. There must be spaces between all elements of each expression (i.e. C*+O\* is not okay, but C\* + O\* is), and species ending with \_g are gasses. Adsorbed species may end with * or \_x where * designates adsorption at the "s" site, while \_x designates adsorption at the "x" site (note that "x" may be any letter except "g"). Transition-states should include a \-, and reactions with a transition-state are specified by 'IS <-> TS \-> FS' while reactions without a transition-state are defined as 'IS \-> FS' (where IS,TS,FS are expressions for the Initial/Transition/Final State). When the model initializes it checks the expressions for mass/site balances, and if it finds that they are not balanced it prints a warning but does not raise an exception. \[list of strings\]

  • elementary_rxns - list version of rxn_expressions. These will be automatically populated if rxn_expressions are defined. \[list of lists of lists\]

  • adsorbate_names - list of adsorbate names included in the analysis. Automatically populated if rxn_expressions are defined.\[list of strings\]

  • transition_state_names\* - list of transition-state names included in the analysis. Automatically populated if rxn_expressions are defined. \[list of strings\]

  • +surface_names+ \- list of surface names to be included in the analysis. \[list of strings\]

  • +gas_names+ \- list of gas names included in the analysis. \[list of strings\]

  • +gas_pressures+ \- list of gas pressures to be used for the analysis. May not be defined if ThermodynamicScaler is being used with pressure as a descriptor and gas_ratios is defined. \[list of numbers in bar\]

  • gas_ratios - list of gas ratios to be used for the analysis. Only necessary if ThermodynamicScaler is being used to determine the overall pressure. \[list of dimensionless numbers\]

  • +temperature+ \- temperature used for the analysis. May not be defined if ThermodynamicScaler is being used with temperature as a descriptor. \[number in Kelvin\]

  • +site_definitions+ \- definitions for all sites used in rxn_expressions. Note that * defaults to 's' so 's' should always be defined. The values of the dictionary can be strings corresponding to the "site" field in the input file, or a list of strings if a site definition corresponds to multiple values of the "site" field. \[dictionary of string:(string OR list of strings)\]

  • +site_totals+ \- dictionary of total which each site must sum to. \[dictionary of string:(floatable string OR dimensionless number)\]

  • +descriptor_names+ \- names of variables to be used as descriptors. \[list of strings\]

  • parser - name of class to use for solver. Defaults to KineticsParser. \[string\]

  • mapper - name of class to use as a mapper. Defaults to MinResidMapper. \[string\]

  • scaler - name of class to use for scaler. Defaults to GeneralizedLinearScaler. \[string\]

  • solver - name of class to use for solver. Defaults to SteadyStateSolver. \[string\]

  • thermodynamics - name of class to use for thermodynamic corrections. Defaults to ThermoCorrections. \[string\]

  • coverage_map_file - file to save coverage map data. \[filepath string\]

  • rate_map_file - file to save rate map data. \[filepath string\]

• Parser:
  • input_file - file where input data is stored. File must be in the correct format for the parser used.
• KineticsParser:

  • estimate_frequencies - attempt to estimate frequencies when they are not listed in the input_file. First try to find adsorbates on different sites, then by using the frequencies of atomic species. If no frequencies are found for transition-states, use the combined frequencies of the dissociated state. Defaults to False. \[boolean\]

  • adsorbate_definitions - dictionary which allows adsorbates to have a different name in the reaction model than in the input file e.g. \{'CH3O':'H3CO'\} if methoxy was called 'CH3O' in the reaction model but 'H3CO' in the input file. Defaults to \{\}. \[dictionary of string:string\]

• Scaler:

  • gas_thermo_mode - Approximation used for obtaining gas-phase free energy corrections. Defaults to ideal_gas. Other possibilities are: shomate_gas (use Shomate equation), zero_point_gas (zero-point corrections only), fixed_entropy_gas (include zero-point and assume entropy is 0.002 eV/K) , frozen_gas (no corrections), frozen_zero_point_gas (no zero-point and entropy is 0.002 eV/K). \[string\]

  • adsorbate_thermo_mode - Approximation used for obtaining adsorbate free energy corrections. Defaults to harmonic_adsorbate (use statistical mechanics+vibrational frequencies). Other possibilities are: zero_point_adsorbate (zero-point corrections only), frozen_gas (no corrections). \[string\]

  • transition_state_scaling_parameters - Used if transition-state scaling is used. Many published values are hard-coded, and parameters can often be input directly so it is usually not necessary. Read scalers/_init_.py for syntax.
• GeneralizedLinearScaler:

  • descriptor_dict - dictionary of known descriptor values for various surfaces. Usually populated by the parser. \[dictionary of string:list of numbers\] where list of numbers is the length of descriptor_names and is ordered to correspond to the order in "descriptor_names".

  • parameter_dict - dictionary of known parameter values for various adsorbates/reactions. Should be either adsorption energies (if parameter_mode=adsorption) or reaction energies/barriers (if parameter_mode=reaction). This is normally populated to be the adsoprtion energies by the parser. \[dictionary of string:list of numbers\] where list of numbers is the length of surface_names and is ordered to corespond to the order in "surface_names"; if the value for some surface is not known the entry should be '-' in order to "hold its place".

  • parameter_mode - scale to adsorption energies (adsorption) or directly to reaction energies/barriers (reaction). Note that to use reaction a custom parser is needed to make the "parameter_dict" contain reaction energies instead of adsorption energies. Defaults to adsorption. \[string\]

  • default_constraints - list of "constraints" on scaling coefficients to be used when constraints are not explicitly specified. This should be a list of length m+1 where m is the number of descriptors (the extra constraint is on the intercept). Constraints can be single numbers (e.g. 0) in order to constrain the coefficient to that value, colon separated numbers (e.g. 0:1) to constrain the coefficient to that range, '+' to constrain to positive, '-' to constrain to negative, or None to use no constraints. Defaults to \['+','+',None\]. \[list of numbers/strings\]

  • scaling_constraint_dict - dictionary of constraint lists (see default_constraints) for adsorbates which do not use the "default_constraints". Can also specify transition-state scaling by using 'TS(A+B):\[slope,intercept\]' where A and B are species in the initial/final state and slope, intercept are the transition-state scaling parameters. See the "methanation" demo for an example and other syntaxes: \[dictionary of string:(list of strings OR string)\]

  • use_input_energies - if the descriptor points happen to be in "descriptor_dict" then use the exact energies instead of the scaled energies. This usually doesn't matter in practice since it is unlikely that the descriptor map will land exactly on one of the points; it is mostly used for the RateAnalysis where the rate of the unscaled parameters is calculated and compared to the scaled rate. Defaults to False. \[boolean\]

• Solver:
• SteadyStateSolver:

  • decimal_precision - number of decimals to explicitly store. Calculation will be slightly slower with larger numbers, but will become completely unstable below some threshhold. Defaults to 50. \[integer\]

  • tolerance - all rates must be below this number before the system is considered to be at "steady state". Defaults to 1e-50. \[number\]

  • max_rootfinding_iterations - maximum number of times to iterate the rootfinding algorithm (multi-dimensional Newtons method). Defaults to 50. \[integer\]

  • internally_constrain_coverages - ensure that coverages are greater than 0 and sum to less than the site total within the rootfinding algorithm. Slightly slower, but more stable. Defaults to True. \[boolean\]

  • residual_threshold - the residual must decrease by this proportion in order for the calculation to be considered "converging". Must be less than 1. Defaults to 0.5. \[number\]

• Mapper:
• MinResidMapper:

  • search_directions - list of "directions" to search for existing solutions. Defaults to \[\[0,0\],\[0,1\],\[1,0\],\[0,-1\],\[-1,0\],\[-1,1\],\[1,1\],\[1,-1\],\[-1,-1\]\] which are the nearest points on the orthogonals and diagonals plus the current point. More directions increase the chances of findinga good solution, but slow the mapper down considerably. Note that the current point corresponds to an initial guess coverage provided by the solver (i.e. Boltzmann coverages) and should always be included unless some solutions are already known. \[list of lists of integers\]

  • max_bisections - maximum number of time to bisect descriptor space when moving from one point to the next. Note that this is actuall the number of iterations per bisection so that a total of 2{^}max_bisections^ points could be sampled between two points in descriptor space. Defaults to 3. \[integer\]

  • descriptor_decimal_precision - number of decimals to include when comparing two points in descriptor space. Defaults to 2. \[integer\]

• ThermoCorrections:

  • thermodynamic_corrections - corrections to apply. Defaults to \['gas','adsorbate'\]. \[list of strings\]

  • thermodynamic_variables - variables/attributes upon which thermo corrections depend. If these variables do not change the corrections will not be updated. Defaults to \['temperatures','gas_pressures'\]. \[list of strings\]

  • frequency_dict - used for specifying vibrational frequencies of gasses/adsorbates. Usually populated by the parser. Defaults to {}. \[dictionary of string:list of numbers in eV\]

  • gas_param_dict - used for parameters in gas correction. Should be ideal gas parameters for ideal gas, or shomate parameters for shomate gas. Most common gasses are hard-coded, so it is usually not necessary. Read thermodynamics.py for syntax.
  • atoms_dict - used for ideal gas correction. Defaults to ASE structure database for common gasses so it is usually not necessary. Read thermodynamics.py for syntax.

• Analysis:
• MechanismAnalysis:

  • rxn_mechanisms - dictionary of lists of integers. Each integer corresponds to an elementary step. Elementary steps are indexed in the order that they are input with 1 being the first index. Negative integers are used to designate reverse reactions. \[dictionary of string:list of integers\]