Re: [sparta-users] [EXTERNAL] Validation chemical reactions in a box

[Marc V. M. <mar...@kr...>]

Hi Michael,

Thank you for the valuable feedback again, it really helped to
better understand the simulations. I decided to run some further tests to
validate the chemical reaction models but I encountered some results which
raised a few questions. I've tested three different cases, mainly a close
box simulation, a hypersonic flow around a 2D cylinder and a hypersonic
flow around the Orion Crew module. The results are cross-checked with other
DSMC codes for the sake of comparison.

*Closed box*
*Description*: box with specular walls filled with static molecular oxygen
(O2) particles. Comparing results of temperature and species over time with
the analytical results, the QK reaction model is used.

*Problem*: the amount of dissociation is much lower than expected (see
"closed_box_fraction.png"). Checking the translational, rotational and
vibrational temperatures, see "closed_box_temperature.png" (note: T
particles is from "compute temp", T grid cells from "compute grid temp", T
rotational from "compute grid trot", T vibrational from "compute grid
tvib", Tvib/grid from "compute Tvib/grid" averaged over all cells.), it is
found that the vibrational temperature surpases the thermal temperature
which to my understanding should not happen. Furthermore, at the start of
the simulation, the thermal temperature is equal to the value that was
input (20,000K) but the rotational and vibrational temperatures start at
0K. Instead, I would expect all three temperatures to be set to the value
in "temp" when defining the mixture as explained in the manual.

*Questions*:

   1. When setting the temperature of the mixture, are the translational,
   rotational and vibrational temperatures also set to this temperature?
   2. What is the relationship between the rotational, vibrational and
   translational temperatures?
   3. For this case, the translational temperature could be taken as the T
   particles or T grid cells? Why is Thermal/grid not the same if there is no
   freestream velocity?
   4. What is the difference between computing the vibrational temperature
   as "compute grid tvib" and "compute Tvib/grid"?

*Hypersonic flow 2D Cylinder*
*Description*: hypersonic flow over a cylinder in 2D with and without
reactions, results over the stagnation line are compared with dsmcFoam
(uses QK model) and MONACO (uses TCE model), the QK and TCE reaction models
are tested in SPARTA. The collide settings were set to "collide vss air
air.vss", "collide_modify rotate smooth vibrate smooth". The "relax
variable" in collide was also tested without difference. For reactions, the
settings were "react qk air.tce" or  "react tce air.tce".

*Problem*: when considering reactions with the QK and TCE models, the
species fraction, translational temperature (from "compute thermal/grid"),
rotational temperature (from "compute grid trot") and velocity over the
stagnation line are as expected. However, the vibrational temperature (from
"compute grid tvib") is unexpectedly higher with the QK model and lower
with the TCE model. On the other hand, when simulating without the
reactions, the translational, rotational and vibrational temperatures are
all much different than expected, especially for the vibrational
temperature which is incredibly low.

*Questions*:

   1. Why is the vibrational temperature so low when reactions are
   deactivated?
   2. Could this be the cause for the deviation in the other temperatures?
   3. Why is the vibrational temperature so large with the QK model?

*Images*: results are plotted over the stagnation line

   - "2D_cylinder_no_reactions_num_density.png" number density for
   simulation without reactions
   - "2D_cylinder_no_reactions_temperature.png" translational,
   rotational and vibrational temperatures for simulation without reactions
   - "2D_cylinder_no_reactions_velocity.png" velocity for simulation
   without reactions
   - "2D_cylinder_reactions_num_density.png" number density for simulation
   with reactions
   - "2D_cylinder_reactions_temperature.png"  translational, rotational and
   vibrational temperatures for simulation with reactions
   - "2D_cylinder_reactions_velocity.png" velocity for simulation with
   reactions


*Hypersonic flow Orion Crew Module*
*Description*: hypersonic flow over the Orion Crew module at 105 km with
and without reactions, the QK and TCE reaction models are used. The results
are compared against results from dsmcFoam code (uses QK model) with the
same parameters. Same settings for collide and react from the 2D cylinder
case are used.

*Problem*: overall, all flow quantities give the expected results when
using a reacting flow with the exception of an overpredicted formation of
NO from the QK model and a low vibrational temperature with the TCE model.
See figures in "ORION_velocity.png", "ORION_temp.png", "ORION_pressure.png"
and "ORION_frac.png". However, in the simulations without reactions, the
vibrational temperature is lower than expected, similarly to the 2D
cylinder test case.

*Question: *why is the vibrational temperature so low when deactivating
reactions, as in the 2D cylinder case?

Thanks in advance for the help,
Marc

El mar, 24 sept 2024 a las 5:16, Gallis, Michael A. (<ma...@sa...>)
escribió:

> Marc
>
>
>
> I am afraid I don’t understand what you are doing and why.
>
>
>
> When the code sets up the initial domain, everything is initialized at
> equilibrium. So,  there is no need to run to achieve equilibration. If the
> temperature is high enough reactions will occur.
>
>
>
> Having the products same as the reactants and with zero energy release
> means that for every collision you will just reshuffle the equilibrium
> distribution function and just observe/tally the equilibrium reaction
> rate.
>
>
>
> You can count reactions either through the stats line or look at the end
> of the run where the code logs how many of reach reactions took place.
>
>
>
> I hope that helps.
>
>
>
> Michael.
>
>
>
>
>
>
>
> *From: *Marc Vives Massana <mar...@kr...>
> *Date: *Monday, September 23, 2024 at 6:28 AM
> *To: *Gallis, Michael A. <ma...@sa...>
> *Cc: *spa...@li... <
> spa...@li...>
> *Subject: *Re: [EXTERNAL] [sparta-users] Validation chemical reactions in
> a box
>
> Dear Michael,
>
>
>
> I've tried to implement what you suggested but when simulating only the
> reaction O2+O2 -> O2 + O2 with zero energy release, apparently, no
> reactions take place. The way I've implemented it now is to first simulate
> the dissociation of oxygen with its corresponding reactions until it
> reaches equilibrium, then, i only simulate O2+O2 -> O2 + O2 with zero
> energy release.
>
>
>
> Furthermore, I now check the number of reactions from the log file but is
> there a command in SPARTA to count the number of a specific reaction?
>
>
>
> Best,
>
> Marc
>
>
>
> El vie, 20 sept 2024 a las 17:15, Gallis, Michael A. (<ma...@sa...>)
> escribió:
>
> Marc
>
>
>
> The equilibrium DSMC simulations maintain the gas at equilibrium, so there
> is no flow, transport or a transient. As said, all the code does is a
> numerical, Monte Carlo integration of the reaction cross section over the
> velocity distribution function.
>
>
>
> You can easily calculate the reaction rate at equilibrium, for any
> reaction. You can just suspend product creation and energy release. The gas
> will stay at equilibrium and you just tally the reactions for any
> particular temperature.
>
>
>
> So for example for O2+O2 -> O2 + O +O  just simulate : O2+O2 -> O2 + O2
> with zero energy release.
>
>
>
> Michael
>
>
>
>
>
> *From: *Marc Vives Massana <mar...@kr...>
> *Date: *Thursday, September 19, 2024 at 6:25 AM
> *To: *Gallis, Michael A. <ma...@sa...>
> *Cc: *spa...@li... <
> spa...@li...>
> *Subject: *Re: [EXTERNAL] [sparta-users] Validation chemical reactions in
> a box
>
> Dear Michael,
>
>
>
> Thank you again for the clarifications! After reading more about what you
> suggested, I've realised that what I actually want to do is to compare the
> analytical QK model expression for reaction rates at equilibrium against
> the results from the DSMC simulation. This would effectively verify the set
> up and then compare the results with the experimental data you provided in
> your papers to validate for example.
>
>
>
> Do you know a way to compute the reaction rates in equilibrium when using
> the QK model? Furthermore, is it possible to obtain an analytical
> expression for the transient properties in a "closed box" when using the QK
> model.
>
>
>
> Thanks in advance,
>
> Marc
>
>
>
> El mié, 18 sept 2024 a las 0:51, Gallis, Michael A. (<ma...@sa...>)
> escribió:
>
> Marc
>
>
>
> For the “closed box” cases like the one you are simulating the code works
> as a Monte Carlo integrator of the reaction probability over the velocity
> distribution function. For the TCE model the reaction probability is
> extracted from the reaction probability using an identical albeit
> analytical integration. So, if the code is not reproducing analytical
> results or is not in agreement with other codes, is because you are using
> different reaction rates. All these are clearly described in Bird 1994,
> Chapter 5. Make sure you are familiar this before you attempt any
> code-to-code comparisons.
>
>
>
> The reaction rates that come with Sparta are taken from Park (1990).
> However, other authors have proposed different reaction rate sets. Thus,
> they are not unique nor are they supposed to agree with any other set.
>
>
>
> For surface chemistry I think the NASA/Ames folk would be better
> positioned to answer your questions. None of the test cases in the 2015
> tutorial included surface chemistry or gas phase chemistry validation.
>
>
>
> Michael
>
>
>
>
>
>
>
>
>
>
>
> *From: *Marc Vives Massana <mar...@kr...>
> *Date: *Tuesday, September 17, 2024 at 6:18 AM
> *To: *Gallis, Michael A. <ma...@sa...>
> *Cc: *spa...@li... <
> spa...@li...>
> *Subject: *Re: [EXTERNAL] [sparta-users] Validation chemical reactions in
> a box
>
> You don't often get email from mar...@kr.... Learn why this
> is important <https://aka.ms/LearnAboutSenderIdentification>
>
> Dear Michael,
>
>
>
> Thanks for the feedback. The comparison I'm trying to do is not only
> against other codes but also analytical data. Therefore, I took the
> reaction parameters described in sparta/examples/chem example which I
> believe is a similar case. My next question would be where the Arrhenius
> activation energy Ea, prefactor A, and exponent b, used in the example case
> are obtained from. Is a general database stating these parameters for all
> chemical reactions at different temperatures publicly available?
>
>
>
> The next cases that I wanted to replicate are the validated test cases
> that were used to validate SPARTA and presented in the slides from the
> short course in 2015 available from the website. To better understand the
> chemical reactions taking place on the surface and to have a reference, is
> it possible to obtain the cases set up that were used for those
> validations?  This would validate the chemical reactions on the surface
> which requires some input data for the reactions that I'm
> struggling to find. Do you know where I can find reliable data for the
> reactions taking place on the surface?
>
>
>
> Thanks in advance,
>
> Marc
>
>
>
> El jue, 12 sept 2024 a las 16:41, Gallis, Michael A. (<ma...@sa...>)
> escribió:
>
> Marc
>
>
>
> This is a very interesting comparison you are trying to do. However,
> before you make a code-to-code comparison you need to make sure both codes
> are using the same molecular parameters, relaxation rates and most
> importantly reaction rates.
>
>
>
> Michael
>
>
>
> *From: *Marc Vives Massana <mar...@kr...>
> *Date: *Thursday, September 12, 2024 at 6:42 AM
> *To: *spa...@li... <
> spa...@li...>
> *Subject: *[EXTERNAL] [sparta-users] Validation chemical reactions in a
> box
>
> Some people who received this message don't often get email from
> mar...@kr.... Learn why this is important
> <https://aka.ms/LearnAboutSenderIdentification>
>
> Dear SPARTA users,
>
>
>
> I've been working on a validation case for the chemical reactions but the
> results are not as good as expected.
>
>
>
> The case is based on the work done by Scanlon et al. (see:
> https://doi.org/10.2514/1.J053370) who were validating dsmcFoam. The case
> basically consists of a box of side length 1e-5 m with gas at 20.000 K and
> 0.0063 atm. The boundaries are specular and adiabatic walls. Initially,
> there is only O2 present in the box and only the dissociation reactions are
> considered:
>
> O2 + O2 --> O + O + O2
> O2 + O --> O + O + O
>
>
>
> I tested the TCE and QK models and the results for the temperature and
> species fractions are not so good, see the figures below. Therefore, I
> kindly ask if someone could help me in implementing the chemical reactions
> or finding some validated cases with SPARTA so I could take a look at them.
>
>
>
> Thanks in advance,
>
> Marc Vives
>
>
>
> I'll also leave the input file in case it helps:
>
>
>
> ## INITIALISATION ##
>
> seed 12345
> dimension 3
>
> global gridcut 1e-5 comm/sort yes
>
> global mem/limit 1024
>
> units si
>
> boundary        r r r
>
>
> # PROBLEM DEFINITION ##
>
> variable l equal 1e-5/2
> create_box      -$l      $l       -$l      $l       -$l      $l
> create_grid     10 10 10       block * * *
>
> global          nrho 2.3118406573249943e+22 fnum 463
>
> species         air.species O2 O
> mixture         air O2 O vstream 0.0 0.0 0.0 temp 20000
> mixture         air O2 frac 1.0
> mixture         air O frac 0.0
>
> create_particles air n 0
>
> ## SETTINGS ##
>
> collide vss air air.vss
> react tce air.tce
>
> timestep        1e-9
>
> ## RUN ##
>
> compute counts count O2 O
> compute temp temp
> fix output print 1 "$(step) $(c_temp) $(c_counts[1]) $(c_counts[2])" file
> output screen no
>
> stats_style step np ncoll nreact c_temp
>
> stats 1000
>
> run 10000
>
>
>
>