CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   simulation of cvd-reactor (http://www.cfd-online.com/Forums/fluent/28818-simulation-cvd-reactor.html)

Christian August 14, 2001 07:11

simulation of cvd-reactor
 
Hello Fluent-users,

I'm working on the simulation of an atmospheric cvd-reactor. So far I'm using only a simplified 2d-geometry. The carrier-fluid is hydrogen, the other reactant is trichlorosilane Si-H-Cl3. I've defined one surface reaction on the two side walls, which are at 1473 K:

H2 + SiHCl3 -> Si + 3 HCl

In my first version the surface reaction rate is defined with the Arrhenius model, in a second version with an UDF. All results for the surface deposition rates are quite reasonable, but the concentration of hydrogen in the reactor gives me a headache. The concentrations of hydrogen and trichlorosilane CLOSE to the walls in direction of the flow become smaller - as expected. Trichlorosilane and hyrogen react on the walls, silicon is deposited on the walls and HCl is created. BUT the contour plot of the concentration of hydrogen shows, that after a couple of centimeters behind the reactor entrance (infront of the reactor is an inlet channel with the same diameter and 1000K wall-temperature) the LOWEST concentration of hydrogen is in the MIDDLE between the two parallel walls and NOT on the walls, where the surface reaction takes place!!?? (concentration of SiHCl3 is decreasing from the sidewalls to the center, concentration of HCl is increasing from the sidewalls - as expected) I'm using the dilute-approx. model for mass diffusivity with polynomial coefficients - we've checked the values.

Trying to summarize the effects: the gas mixture arrives at the reactor entrance with a temperature of 1000 K and a parabolic velocity profile. The walls in the reactor are at 1473 K. The gas is heated from 1000 K to 1473K, density decreases, velocity increases. Surface reactions on the sidewalls: H2 + SiHCl3 -> Si + 3 HCl i.e. 2 molecules "disappear" from the volume, 3 HCl molecules go into the volume, 1 Si stays on the surface. Concentration of HCl increases in the volume beginnig from the sidewalls. Concentration of SiHCl3 decreases in the volume beginning from the sidewalls. HYDROGEN: on the first few centimeters in the reactor the concentration of H2 decreases beginning from the sidewalls. Then comes a little area with homogenous concentration of H2 between the sidewalls. In the following 80% of the reactor the concentration of H2 is LOWER in the center of the reactor than close to the sidewalls, where the surface reaction happens.

Does anybody understand this effect? Can anybody give me an idea what's wrong?

Thanks, Christian


Mary August 14, 2001 17:25

Re: simulation of cvd-reactor
 
Are you performing a steady state analysis?

One thing I've noticed about FLUENT is that when you begin your analysis, the control volume is not an empty volume (or area). The control volume is initialized containing the fluid you specified last in your system. If you edit your fluids you will notice it at the bottom of the list and thats the fluid that fills your system before you do 1 iteration.

There is an easy way to catch this problem. Come up with a 5th gas, make it inert and light, not a part of any reaction and not a inlet gas and add it LAST to your materials...(I hope you know what I mean by that. You can have a material specified in the analysis that is not in the system) Make certain that once you converge, that inert fluid is GONE.

If you are running a non-steady state analysis, and if you are checking the system at a time step shorter than the residence time, then you will - and should - see an inert gas contentration profile in the reactor.

Good luck

Christian August 15, 2001 06:18

Re: simulation of cvd-reactor
 
Hello Mary,

thanks for the advice! Yes, I'm performing a steady state analysis. I already had this idea with an inert gas - I've used helium, but not using it as the bulk species. I will try using it as the bulk species. thanks, Christian

mark August 21, 2001 12:27

Re: simulation of cvd-reactor
 
Christian,

I am not a fluent user, but I would check the following things:

1. Species convergence. Are all of your species converging adequately?

2. Hand-in-hand with species convergence is mass conservation (i.e. is the pressure correction eq. converging?) Does mass in = mass out? (Don't forget to include the mdot to the walls)

If either mass conservation or species convergence is lacking, then you might want to look closer at the dilute approximation. I assume by this that you are defining a reference species and the code is solving for n-1 species where n is the number of species in the system. In this case all species errors get lumped into the reference species. You might want to use something like a Multi-component diffusion or Stefan-Maxwell model. I am not exactly sure what options Fluent offers, but surely they have some model that enforces species conservation.

Hope this helps.

Christian August 22, 2001 04:46

Re: simulation of cvd-reactor
 
Mark,

thank you for your help.

concerning 1) I've set monitors on the concentrations of all species. The convergence of species is adequate.

concerning 2) the error in the mass balance is of the magnitude of 10^-9 kg

As the molecular weight of SiHCl3 is rather high, the mass fraction of SiHCl3 is about 0.9, even for low concentrations. The dilute approximation is only applicable for chemical species present at low mass fraction in a carrier fluid.

I want / have to change the diffusion model to multi-component or kinetic theory, but the necessary parameters for trichlorosilane SiHCl3 are very hard to get. Any hints where I can find these parameters?

regards, C.

mark August 23, 2001 11:34

Re: simulation of cvd-reactor
 
I assume you are talking about the Lennard-Jones parameters or an estimate of the Dij's?

The Lennard-Jones parameters sometimes aren't so easy to find. You might try searching Sandia's web site for references. They are doing quite a bit of computational simulation in molecular dynamics and you should be able to find something there.

You might also want to ask Fluent. They should be able to give you this information if they have it.

If you can't find the parameters for SiHCl3, try finding and using the parameters for a "similar" molecule.

I wish I could give you more help, but the accuracy/lack of property data seems to be a fairly global concern.

Mark

mark August 23, 2001 11:46

Re: simulation of cvd-reactor
 
Christian,

A couple of references:

1. R.A. Svehla - NASA TR R-132, 1962.

2. P. Neufeld, A. Janzen, R. Aziz - J. Chemical Physics, Vol. 57, No. 3, 1972, pp. 1100-1102.

3. R. Kee, G. Dixon-Lewis, et. al. - Sandia Report SAND86-8246, 1986.

I don't know that the 1st and 3rd are generally available, but the 2nd should be available at any good library.

Mark

Christian August 27, 2001 04:27

Re: simulation of cvd-reactor
 
Mark, thank you for the assistance! Fluent has sent me this link to http://www.symp14.nist.gov/ , but they don't have any parameters for trichlorosilane neither. The only reference in the fluent manuals is "Molecular theory of gases and liquids" by Joseph O. Hirschfelder. I will try to get the papers / books you gave me the references to. Concerning the workaround: We've used the parameters of silane and the results were already quite reasonable. regards, Christian


All times are GMT -4. The time now is 13:54.