Adding in coverage dependent effects to surface kinetics #2090
Commits on Jul 23, 2021
-
Adding in coverage parameters to SurfaceArrhenius
Adding coverage dependent parameters to surface arrhenius kinetics added in coverage dependence in kinetics model Adding _coverage_dependence to SurfaceArrhenius and SurfaceArrhenius Cython declarations adding variable names to coverage documentation and making it more readable Fixing the updated docstrings for coverage_dependence Adding in coverage_dependence to SurfaceArrhenius and SurfaceArrheniusBEP
-
Use coverage parameters to modify the reaction rates
using coverage dependent parameters multiply kr by cov parameters moving cov_dep and co to the init WIP: Export cantera files with coverage dependent parameters make cov dict into cantera readable format for export WIP: Adding unittests to test cantera export Make sure units of coverage dependence are si
-
-
Adding in new unit tests to the solver surface test
test that coverage dependence is doing what it should be doing Richard: Fix some surface solver tests details. This is meant to represent CH3X, not HX. Doesn't make a difference to what we're testing, but better to be right.
-
-
-
-
-
Changing coverage dependence SMILES to species objects
Look for coverage_dependence attribute
-
-
-
-
Can read Coverage Dependent kinetics from Chemkin files
For now, only reactions with COV are turned into SurfaceArrhenius objects, all others are left as Arrhenius. I'm not quite sure what's happening around lines 467-473 (Emily's changes, I'm just committing them)
-
Do not write coverage dependent parameters if coverage dependence is …
…turned off save_chemkin_file has coverage_dependence flag, which is passed on to write_kinetics_entry
-
Turn on coverage dependence in the coverage-dependence solver tests.
This defaults to False if not specified as true. The fact that this wasn't noticed, tells us that the test isn't really testing that we do evaluate coverage-dependence correctly.
-
Correcting ordering of COV parameters in Chemkin files.
Should be (a, m, E).
-
Chemkin reading COV: Do parameters in order a, m, E.
also, turning the forgotten 'J/mol' into Eunits !
-
Chemkin reading COV parameters: should preserve case of species names.
If we're even going to bother writing this code we really ought to test it
-
Put coverage-dependence paramaters in order a, m, E, consistently.
Dictionaries are order-preserving, and humans' brains are, so we'll have fewer bugs if we're consistent with ourselves and other software's conventions. Chemkin and Cantera use (a, m, E) ordering. Now we do, throughout.
-
Chemkin reading COV parameters. Fixes for reading sticking coefficients.
If there's a STICK line before a COV line we can cope.
-
Chemkin COV writing: correct units for the E parameter
The E should be in the same units as Ea, which is kcal/mol. That's 4184 times smaller.
-
update __repr__ string writing
do not write if the coverage dependence dictionary is empty
-
Revert "Do not write coverage dependent parameters if coverage depend…
…ence is turned off" This reverts commit 7841f24. Had to manually tweak a few things to resolve conflicts. The point is to remove the extra code from the chemkin-writing methods. If we turn off coverage, the parameters should not be in the model. If the parameters are in the model, they should be written by chemkin.
-
Chemkin writing coverage dependence: fix bugs
- An empty dictionary shouldn't be printed (even though it's not 'None') - We need sticking coefficients cov parameters to be in the right units too! (we previously fixed from J/mol to kcal/mol in one place, but not both)
-
Chemkin writing coverage dependence: reduce code duplication
Protects us from bugs where we fix things in one place and not the other.
-
Renaming apply_coverage_depndence_to_reaction to process_coverage_dep…
…endence. This better describes what it does. Also expanded docstring. Also removed it from INSIDE the apply_kinetics_to_reaction method and removed if from and elif block, so that it is duplicated in less code (but still called in the same way, I think)
-
process_coverage_dependence now REMOVES parameters if turned off.
I'm still not very sure about the `coverage_dependence=False` settings, and what should be turned off by it.
-
SurfaceReactor: Remove unused (and mutable) default arguments.
Putting dictionaries as a default argument is a famous gotcha https://docs.python-guide.org/writing/gotchas/ I wondered if it's being done on purpose for some efficiency reason, to share memory, but decided no. a risk of error, and is not needed anyway as we never initialize the object with those parameters anyway.
-
CovDep test: report timing of the simulations.
I'm hoping to speed it up later.
-
-
CovDep test: change T, and plot coverage.
At 600 K the eqm coverage is much higher, so you get a more noticeable coverage dependence. By plotting coverage instead of concentration it's easier to see what's happening.
-
CovDep Test: make the surface reactor test slower.
Add a bunch of redundant species and reactions, to slow it down a bit, for benchmarking purposes. Simulation took 3.550e+00 seconds in rmgpy.solver.surfaceTest.SurfaceReactorCheck.test_solve_h2_coverage_dependence
-
Refactor the coverage dependence calculations in surface solver.
Main goal was to optimize everything happening inside the loops in the "residual" function (which gets called a lot). Simulation took 4.541e-01 seconds in rmgpy.solver.surfaceTest.SurfaceReactorCheck.test_solve_h2_coverage_dependence (previous run was 3.550e+00, so we're about 8x faster)
-
CovDep test: stop plotting the graphs.
These are helpful for manual inspection and helped me confirm it's working, but they're not needed for the automated test suite or continuous integration.
-
added check that covdep makes a difference
changed CH3 coverage unittest to run an additional simulation without coverage dependence. the solution arrays from each are compared.
-
updated documentation with coverage dependence
added instructions on input file construction and updating the database.
-
