Chemical Reaction Path Diagram Tool


Introduction: Chemical Reaction Path Diagrams

      The chemical reaction path diagram tool (ChemPathTool) aids in
the generation of chemical path diagrams for reacting flow systems.
More generally, Pathtool is used to manipulate graphically the layout
of a bi-directed graph.  For reacting flow systems, the nodes of this
graph represent chemical species, and the arrows (or, edges)
connecting the nodes represent the flow of a conserved quantity
between species, such as the transfer rate of a particular atom.
      In our systems, atoms are transferred among species through
chemical reactions.  For example, when the reaction
CH4(+M)<=>H+CH3(+M) proceeds, the impact of an arbitrary molecule (M)
knocks one H atom from the CH4 molecule.  Thus, a carbon atom is
"transferred" from the set of all CH4 molecules to the set of CH3.
The progress rate of this reaction (and its reverse, ie sticking an H
atom onto a CH3 molecule) depends on the local state.  In a typical
reacting flow calculation there are hundreds of such reactions, and
the goal is to understand how the state of the fluid affects the
reaction balance.  The path diagram turns out to be a good way to
visualize the state of the chemical system.
    
   The edges in a path diagram are obtained from a numerically
simulated reacting flow using three basic steps:
1. A full state is obtained from the simulation.  The state might
represent an instantaneous snapshot from a time-dependent simulation
or a steady solution.  The state includes temperature, pressure,
density and species distribution.
2. Forward and reverse reaction rates are computed for every relevant
chemical reaction in the system using this state.
3. The reactions are "decomposed" into path diagram edge contributions
and summed, and are thereby converted to values representing the rate
that a particular atom is exchanged among chemical species.  The
summing, or integration, may extend over a section of a
multidimensional reacting flow domain, or may be conditioned on any
parameter the investigator may find useful for the analysis.

   The chemical species and atomic flux rates are loaded into our tool
as graph nodes and directed edges, respectively.  The user then
manipulates the layout of this graph into a geometry that is visually
meaningful.  The resulting path diagram is written out as an
Encapsulated PostScript file for printing or further manipulation, and
the layout of the nodes can be saved and reused with another set of
edge strengths.  By changing the input data (simulation parameters,
region of integration, atom of interest, etc), one can generate a
series of path diagrams that are useful for identifying patterns in
the chemical system.
      There are a number of automated path generation tools available
on the web.  But from our point of view, they all had a serious flaw
that rendered them unusable.  All the tools were too smart, in that
they found all sorts of nifty optimized ways to arrange the nodes
based on the strength and connectivity of the edges.  Each run
resulted in a different nodal layout, and any two cases were
impossible to compare side by side.  (Moreover, there are standard
species layouts in the combustion literature one might like to
emulate).  The ChemPathTool allows (or forces, depending your point of
view) one to specify the location of the nodes regardless of the edge
strengths.  Since the user must now design graph layouts, we found it
most simple to provide a simple customized drawing tool.  In the end,
our simple concept led to a piece of software with two aims: easy
creation of a nice layout, and application of that layout to multiple
sets of edge strengths.
      You may have picked up on one omission from the previous
paragraph: how to generate the data in practice.  Like all analysis,
this can be arbitrarily complicated.  In a later section, we present a
simple example that will guide you in creating your own data.


General ChemPathTool Usage

The node names and edge strengths are loaded into the tool as a file
containing a list of names and numbers.  This may be done while
starting the tool from the python command-line, or after the tool has
been started.  The main window of the tool features a prominent
drawing canvas, and a number of toggle switches and menus.  Beginning
with a blank canvas, the user makes a subset of the nodes visible and
arranges them interactively using the mouse pointer ("align",
"distribute", "group", "lock", "nudge" and "rubber-band select" tools,
such as those found in common illustration software, make it easy to
generate publication-quality results).  The directed edges connected
to the nodes follow them as they are moved around the canvas.  Only
edges with strength above a threshold are drawn; the threshold value
is controlled with a "slider" widget as a percentage of a reference
edge value.  The layout may be saved to a file, and recalled for a new
dataset.  If the new dataset has a different set of species, only
those in common with the saved set are placed on the canvas (this
allows side-by-side comparisons of results from different mechanism
reductions, for example).  The remaining nodes may be omitted, or
placed on the canvas in a separate step.  The resulting path diagram
is then saved in EPS format, allowing annotations and touch-ups by
fancier illustration packages.  Click here for a demonstration of the
ChemPathTool in action.
        
        
ChemPathTool Data Requirements
    
ChemPathTool requires data input in the form of a list of nodes and a list
edges.  Data for each node includes the text string that will be used
as a label.  Edge data includes the labels of the "from" and "to"
nodes, and the strength of the edge (thickness of the arrow to be
drawn).  Optionally, two values may be associated with each edge.  The
data is given to the ChemPathTool in the form of an ascii text file, or
"ChemPathTool Edge Data" (CED) file, and is loaded via the "File->Load
Edge Data" menu while the tool is running:

Line 1->Case ID
Line 2->sp1 sp2 sp3 sp4 sp5 sp6
Line 3->sp2 sp4 1.9 -2.3
Line 4->sp1 sp3 2.3 -100
Line 5->...

The first line of the file contains an ID string that can optionally
be printed onto the canvas.  The second line contains the set of
space-delimited names associated with each node.  Each line after that
contains two species names (each must be on the list given at the
top), and one or two values.  Often the two values will be of opposite
sign and indicate the production and depletion of a species--a single
value would represent their sum.  The ChemPathTool will normalize these
values linearly and draw edge arrows connecting the nodes on the
canvas using a correspondingly scaled arrow thickness.  A toggle on
the panel allows one to coalesce the two on the fly.  Another toggle
allows one to shut off the auto-scaling (when auto-scaling is off,
then a weight of 100 produces an arrow 20pts thick).  Some nodes may
not be connected in a given edge data set.

Input Data Generation - Example

Generation of the data can be arbitrarily complex, depending of course
on the quantity you're trying to visualize.  In the simplest case, one
has a single state, which includes the pressure, temperature and
species distribution.  With this state you will associate a chemical
reaction mechanism and set of thermodynamic relations between state
properties.  With these one may compute chemical reaction rates. Then,
given a choice for which atom to trace through the system, one can
form the complete set of path diagram edges, and proceed to evaluate
them into the CED file described in the section above.

As this part of the analysis is outside the scope the ChemPathTool
intentions, we borrow heavily from Dave Goodwin's Cantera package to
ease the generation of sample data.  We basically create a filter that
turns a set of state data into a PED file using the appropriate
ChemKin(TM) input file and some handy scripts (which are included with
the ChemPathTool, by the way).

Line  1-> from mechinfo import *
Line  2-> from CanteraChemEdgeAnalysis import *
Line  3-> from Cantera import OneAtm
Line  4-> 
Line  5-> readMechanism('gri30.inp')
Line  6-> Nspec = numSpecies()	
Line  7-> 
Line  8-> # Set state
Line  9-> X = [1.0 / Nspec] * Nspec
Line 10-> setState(2000.0, OneAtm, X)
Line 11-> 
Line 12-> # Compute reaction rates
Line 13-> qf = fwdRatesOfProgress()
Line 14-> qr = revRatesOfProgress()
Line 15-> qnet = qf - qr
Line 16-> 
Line 17-> # Decompose reactions into edges
Line 20-> edges = getEdges('C')
Line 21-> 
Line 22-> # Dump out nodes and edges
Line 23-> print join(speciesNames())    
Line 24-> for e in edges:
Line 25->     valf = 0.0
Line 26->     valr = 0.0
Line 27->     for [r,c] in e.rateWeightList:
Line 28->         valf += c*qf[r]
Line 29->         valr += c*qr[r]
Line 30->     print e.sp1,e.sp2,valf,-valr

This script is written in the python language, but should be easy
tofollow.  In lines 1-3, we import a number of useful functions from
Cantera.  We also include CanteraChemEdge-Analysis provided with this
distribution, which includes a special support function to be
discussed in momentarily.  Line 5 is the Cantera call to read up a
ChemKin(TM) style reaction mechanism and thermodynamics database file.
Line 6 defines a handy variable, and Line 9 uses it to create a
reasonable value for all the species mole fractions.  Line 10 sets the
state inside the Cantera package, and then Lines 13-14 use that state
to compute the forward and reverse reaction rates for all the
reactions specified in "gri30.inp".  The two calls result in two
arrays of floating point numbers.
    
Line 20 uses the reaction data from "gri30.inp" to generate a set of
edges that represent the flow of "C" or Carbon through the system.
This is accomplished by looking at each reaction involving
carbon-containing species, and identifying the atoms shuffled between
species as the reaction progress.  For all but a very few cases, such
a decomposition is unambiguous.  There are exceptions in the automatic
processing though.  For example, when tracing nitrogen, the reaction

                 HNO + NO2 <=> HONO + NO

may remove an H atom from HNO resulting in NO.  Alternatively, HNO
could gain an O atom and become HONO.  The code in the
CanteraChemEdgeAnalysis.py file provided with this distribution
arbitrarily chooses to transfer the lighter of the two species.
However, such a decision is certain to be wrong in some
cases.....buyer beware.

Each edge in the edge list provided by the function "getEdges"
contains the "to" and "from" species, and a list of reaction id's and
corresponding multipliers to indicate the number of trace atoms
transferred on each occurrence of the reaction.  The loop in Lines
24-29 sums the forward and reverse atomic fluxes separately for each
edge.  Finally, Lines 23 and 30 print the desired PED file to
stdout. The resulting text file is then input into the ChemPathTool
directly.


ChemPathTool Installation

<NB: A complete install on a Win2K machine takes about 10 minutes or less>

The path tool is fully contained in a simple (albeit somewhat lengthy)
Python script.  Python is an interpreted, interactive, object-oriented
programming language.  It is often compared to Tcl, Perl, Scheme or
Java.  There are Python interfaces to many system calls and libraries,
as well as to various windowing systems (X11, Motif, Tk, Mac,
MFC). The Python implementation is portable: it runs on many brands of
UNIX, on Windows, DOS, OS/2, Mac, Amiga.  Therefore, our ChemPathTool
should be fully portable across all platforms supported by Python.

A Python script is a text file that is interpreted by the Python
interpreter, in a fashion similar to PostScript, MatLab, Mathematica,
etc.  There is a large body of freely available modules for extending
Python's basic functionality.  For our puposes, the most useful of
these are Tkinter and the Python Mega Widgets (Pmw).  Tkinter provides
a very simple and uniform interface to the local window system
(X-windows, Microsoft, MacOS, etc) to enable simple GUI construction,
and is included with the latest Python distribution on most
platforms. Pmw provide an even simpler interface to some handy buttons
and such, and is installed by copying a directory into a special
location in the Python directories built during the default Python
install.

To install Cantera's Python interface, it's best to go directly to
Dave Goodwin's Cantera site after you finish the first three steps
below successfully. See step (4) below for more information.


To install ChemPathTool:

1. Go to http://www.python.org/download and get/install the latest
version of Python (version 2.2.1 as of this writing).
  
For windows in particular, you will get the windows
installer. Double-click on the .exe file you download, verify the
default install location, and click finish.  There are rpm's for Linux
and Mac binaries for the Macintosh that are similarly easy to intall.
For other systems, you've got do the the usual "gunzip; tar;
configure; make install" deal.  After Python is installed, add the
Python directory (eg, C:\Program Files\Python22) to your PATH so you
can run it at the command prompt.

2. Go to http://pmw.sourceforge.net and grab the Pmw tarball (version
2.0.0 as of this writing)

Unzip and untar, and then put the resulting Pmw directory in your
Python site-packages area (alternatively be sure that your PYTHONPATH
environment variable points to a directory that includes the Pmw
directory).  On my Windows 2000 machine, all the Pmw stuff ended up in
a directory called

      C:\Program Files\Python22\Lib\site-packages\Pmw

Alternative, and perhaps better, if you have sudo, do 

      sudo python setup.py install

3. Download the ChemPathTool-1.1 tarball from the CCSE software website. 

Unpack ChemPathTool-1.1.tgz, and copy the files that end up in
the directory "ChemPathTool-1.1" over to the same site-packages directory
used for (2) above.  On my Win2K machine, I ended up, for example,
with the file:

      C:\Program Files\Python22\Lib\site-packages\pathTool.py

and a few others like it.

(For a windows machine, you may alternatively grab the Windows install
file, and double-click it.  This will not only put the ChemPathTool files
in the correct place, but will provide an uninstall capability
available from the "Start->Settings->Control Panel->Add/Remove
Programs" window to clean out the whole thing later if desired.)

4. The sample of python code above is compatible with version 1.3 of
the Python interface to Cantera, which should be available from the
Cantera website. Dave was good enough to allow me to have an
advance copy of the Python components of Cantera 1.3, and to make the
windows-executable forms available here.  So, for windows machines,
download the Cantera-1.3.zip file here, and unzip it (for all other
platforms, you�ll have to get the official 1.3 [or greater] release).  Move the
resulting Cantera folder into the same site-packages directory used
for (2) above, or to a place on your PYTHONPATH.  On my windows
machine, I end up with a directory called

      C:\Program Files\Python22\Lib\site-packages\Cantera


To run ChemPathTool:

ChemPathTool may be started two ways:
1. Double-click on the pathTool.py file in the Python site-packages
area (requires that your PATH variable includes the directory with the
python executable).
2. Import ChemPathTool into a python script, and invoke the tool as a
function call.  A simple example is included in the distribution as a
script file, "pathToolScr.py".  From a command-line prompt in the
directory containing the file �pathToolScr.py�, type �python
pathToolScr.py�


Comments and Acknowledgements

The key to the path diagram tool existence has been in its
simplicity. We are not software packagers, and have only put a minimal
effort into its development.  However, we strongly encourage
suggestions or contributions in terms of additional features and bugs
reports (to: [email protected]).  General feedback and criticisms of the
tool are welcome, nay encouraged.  We hope that you find this drawing
and analysis tool useful and inspiring of new analysis techniques and
tools.

Also, we should mention that the path tool was developed upon ideas
originally programmed by Bill Crutchfield.  Bill pointed us to Tkinter
and created a first-cut implementation that has evolved into its
present shape.  Also, Joe and I fleshed out our thoughts on reaction
decomposition in part while talking with Dave Goodwin, the developer
of the Cantera chemistry package.  Without input from these two, we
may well have never built the tool.