Dear Bernd,
Firstly I have to thank you for your constant and really constructive helps.
Regarding to the new concern of mine, I am going to investigate the diffusion procedure at the boundary of a Cr7C3 sample surrounded by Austenite during the cooling process.http://board.micress.de/viewtopic.php?f ... +diffusion looks to be very helpful. Am I right? do you have any other predefined dri which is so close my purpose?
Regards,
Omid
Diffusion at Austenite-Cr7C3 Interface
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Omid,
as you may have noticed, as Board Moderator I decided to split up your question form the former topic, because it seems your new question is no more linked to eutectic solidification. To keep the forum structure transparent to other users, it is better to start a new topic in such case. Of course, you are welcome to edit the title of this new topic, which I have chosen deliberately.
Now to your question: If you know or expect that interface diffusion is important in your case, the topic which you linked should be indeed one of the most relevant ones in this forum. Interface or grain boundary diffusion means that diffusion is increased by the interface or grain boundary as compared to diffusion in the bulk. To my understanding, this is more often the case at lower temperatures, while at high temperatures bulk diffusion is typically dominant.
There is another topic which can be important with this respect, because Cr7C3 is a type of carbide which has a broad solubility range for Cr, Fe and other metals. This means that the transformation Cr7Cr/fcc can show a nple behaviour. This fact may be important in case you are interested in growth or shrinkage kinetics of M7C3.
Bernd
as you may have noticed, as Board Moderator I decided to split up your question form the former topic, because it seems your new question is no more linked to eutectic solidification. To keep the forum structure transparent to other users, it is better to start a new topic in such case. Of course, you are welcome to edit the title of this new topic, which I have chosen deliberately.
Now to your question: If you know or expect that interface diffusion is important in your case, the topic which you linked should be indeed one of the most relevant ones in this forum. Interface or grain boundary diffusion means that diffusion is increased by the interface or grain boundary as compared to diffusion in the bulk. To my understanding, this is more often the case at lower temperatures, while at high temperatures bulk diffusion is typically dominant.
There is another topic which can be important with this respect, because Cr7C3 is a type of carbide which has a broad solubility range for Cr, Fe and other metals. This means that the transformation Cr7Cr/fcc can show a nple behaviour. This fact may be important in case you are interested in growth or shrinkage kinetics of M7C3.
Bernd
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Bernd,
Of course it's much better to organize the topics in a visible and easy to track way. Thanks for the improvement. but there is still a problem, my process is not a solidification, it's a solid-solid transition
Actually in the system it is predicted in the literature that the existing phase is M7C3, and at the low cooling rates like cooling in the weather it turns in to the M23C6, and M could be both the Fe and Cr. In which we do not know that which of them persist in our experiments. I want to develop the model based on the both elements and then by comparison with the experiments we can conclude the element. BTW we have a domain completely Austenite, and there is a circular grain made up of M7C3, when we cool it down in the water (quench) there would be no change in the grain, just the Austenite turns into martensite. But when we cool it down in the weather, the composition around the grain changes and the grain turns into the M23C6. by this description, do you have any useful predefined model? Or any other advise?
VG
Omid.
Of course it's much better to organize the topics in a visible and easy to track way. Thanks for the improvement. but there is still a problem, my process is not a solidification, it's a solid-solid transition
Actually in the system it is predicted in the literature that the existing phase is M7C3, and at the low cooling rates like cooling in the weather it turns in to the M23C6, and M could be both the Fe and Cr. In which we do not know that which of them persist in our experiments. I want to develop the model based on the both elements and then by comparison with the experiments we can conclude the element. BTW we have a domain completely Austenite, and there is a circular grain made up of M7C3, when we cool it down in the water (quench) there would be no change in the grain, just the Austenite turns into martensite. But when we cool it down in the weather, the composition around the grain changes and the grain turns into the M23C6. by this description, do you have any useful predefined model? Or any other advise?
VG
Omid.
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Omid,
no, there is no predefined setup available, and I also did not do something similar before. But from your explanation, I do not see why grain boundary diffusion should be important. So, I would first start without that, it would just make it more complicated. You can always add grain boundary diffusion afterwards, once that the simulation is running properly and in case you see a necessity.
For both carbide phases, carbon is a stoichiometric element which you should define as such (with TQ-coupling it will be automatically done if you forget...). You should not define a phase interaction between the two carbide phases ("no_phase_interaction") because there cannot be a direct transformation (both phases are stoichiometric in C with different composition!). Therefore, transformation will be indirect by diffusion through and via the austenite phase.
You need some nucleation types to put M23C6 seeds either into the austenite or at the interface between fcc and M7C3.
Do you want to use coupling to Thermocalc?
Bernd
no, there is no predefined setup available, and I also did not do something similar before. But from your explanation, I do not see why grain boundary diffusion should be important. So, I would first start without that, it would just make it more complicated. You can always add grain boundary diffusion afterwards, once that the simulation is running properly and in case you see a necessity.
For both carbide phases, carbon is a stoichiometric element which you should define as such (with TQ-coupling it will be automatically done if you forget...). You should not define a phase interaction between the two carbide phases ("no_phase_interaction") because there cannot be a direct transformation (both phases are stoichiometric in C with different composition!). Therefore, transformation will be indirect by diffusion through and via the austenite phase.
You need some nucleation types to put M23C6 seeds either into the austenite or at the interface between fcc and M7C3.
Do you want to use coupling to Thermocalc?
Bernd
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Bernd,
By the descriptions, I know more or less what to do. Sure I want to couple with Thermocalc but at the moment we have a problem with the license affair.
We are dealing with it and are waiting for the IT service's response. Afterwards we can improve the model. I will be back as soon as I get any enhancement in licensing progress.
VG,
Omid.
By the descriptions, I know more or less what to do. Sure I want to couple with Thermocalc but at the moment we have a problem with the license affair.
We are dealing with it and are waiting for the IT service's response. Afterwards we can improve the model. I will be back as soon as I get any enhancement in licensing progress.VG,
Omid.
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Bernd,
I have already developed a preliminary driving file and during the running I see some strang errors. Could you please have a look on the driving file to see what is happening?
Thermocalc error 1611 and micress error 20!
Regards,
Omid.
I have already developed a preliminary driving file and during the running I see some strang errors. Could you please have a look on the driving file to see what is happening?
Thermocalc error 1611 and micress error 20!
Regards,
Omid.
- Attachments
-
- New folder.zip
- (11.93 KiB) Downloaded 227 times
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Omid,
I found essentially 3 major issues with your input file which cause problems:
1.) Definition of stoichiometric phases:
You cannot set elements stoichiometric in all phases! Without redistribution of elements between the phases there cannot be any phase interaction/movement. It is correct and necessary to set the elements C and Cr stoichiometric in M7C3 and M23C6, but not in FCC_A1. Here, you have complete solubility, and there is no reason for a stoichiometric condition.
2.) Assignment of phases:
It is correct that you cannot avoid to keep phase 0 in the MICRESS phase list. It is per default anisotropic, and you cannot have several grains of phase 0 with different orientations.
But if you do not want to use phase 0 in the simulation, you should not define any interactions with this phase (phase interaction data) and not define a diffusion coefficient from database for phase 0. Because then you won't be asked to link phase 0 to any thermodynamic phase in the database, and you avoid the double usage of thermodynamic phases which you had (FCC_A1 and FCC_A1_2). On top of that, you define a phase interaction between the two identical phases which creates trouble in this case.
3.) Diffusion coefficients;
When using diffusion coefficients from database, you should always check whether there is mobility data available for the specified phase in the .GES5 file. For phase 0 I already mentioned that you should not define any (the corresponding keyword is "diagonal n".
For M7C3 and M23C6 there is no data available, and you should also chose "diagonal n" (or use manually defined values if you know any). Requesting non-existing data leads to unpredictable results (NAN in this case).
For FCC_A1, you can use the available data using the keyword "multi" for getting the full diffusion matrix. If you do that, you will notice that the simulation is very slow, due to the high diffusivity of carbon in FCC_A1. Given the fact that the diffusivity of carbon here is orders of magnitude faster than that of Cr (you can check in the .TabD or .diff output), it is a good approximation to use "inifinite" diffusion for carbon (keyword "diagonal i") in FCC. Then, given that there are no gradients in C composition, it is sufficient to take into account only the diagonal term for Cr in FCC ("diagonal g", which is synonymous to the old keyword "database_global").
Furthermore, I would advise you to use a smaller simulation domain for designing and testing your setup. Otherwise, you will probably loose too much time waiting for response.
Bernd
I found essentially 3 major issues with your input file which cause problems:
1.) Definition of stoichiometric phases:
You cannot set elements stoichiometric in all phases! Without redistribution of elements between the phases there cannot be any phase interaction/movement. It is correct and necessary to set the elements C and Cr stoichiometric in M7C3 and M23C6, but not in FCC_A1. Here, you have complete solubility, and there is no reason for a stoichiometric condition.
2.) Assignment of phases:
It is correct that you cannot avoid to keep phase 0 in the MICRESS phase list. It is per default anisotropic, and you cannot have several grains of phase 0 with different orientations.
But if you do not want to use phase 0 in the simulation, you should not define any interactions with this phase (phase interaction data) and not define a diffusion coefficient from database for phase 0. Because then you won't be asked to link phase 0 to any thermodynamic phase in the database, and you avoid the double usage of thermodynamic phases which you had (FCC_A1 and FCC_A1_2). On top of that, you define a phase interaction between the two identical phases which creates trouble in this case.
3.) Diffusion coefficients;
When using diffusion coefficients from database, you should always check whether there is mobility data available for the specified phase in the .GES5 file. For phase 0 I already mentioned that you should not define any (the corresponding keyword is "diagonal n".
For M7C3 and M23C6 there is no data available, and you should also chose "diagonal n" (or use manually defined values if you know any). Requesting non-existing data leads to unpredictable results (NAN in this case).
For FCC_A1, you can use the available data using the keyword "multi" for getting the full diffusion matrix. If you do that, you will notice that the simulation is very slow, due to the high diffusivity of carbon in FCC_A1. Given the fact that the diffusivity of carbon here is orders of magnitude faster than that of Cr (you can check in the .TabD or .diff output), it is a good approximation to use "inifinite" diffusion for carbon (keyword "diagonal i") in FCC. Then, given that there are no gradients in C composition, it is sufficient to take into account only the diagonal term for Cr in FCC ("diagonal g", which is synonymous to the old keyword "database_global").
Furthermore, I would advise you to use a smaller simulation domain for designing and testing your setup. Otherwise, you will probably loose too much time waiting for response.
Bernd
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Bernd,
Many thanks for the advises. I have already applied your comments but still see that driving file asks me to link the phase 0 to one of the referred phases of GES5 file and consequently asks for the further data.
Moreover, I also have no idea about Maximum T deviation! I have randomly adjusted it at 2 K and always get a message of Relinearisation in interface of grains X/Y due to a T deviation of some (a number)!
Kindly have a look at the dri file to see the error and please let me know your valued comments.
Regards,
Many thanks for the advises. I have already applied your comments but still see that driving file asks me to link the phase 0 to one of the referred phases of GES5 file and consequently asks for the further data.
Moreover, I also have no idea about Maximum T deviation! I have randomly adjusted it at 2 K and always get a message of Relinearisation in interface of grains X/Y due to a T deviation of some (a number)!
Kindly have a look at the dri file to see the error and please let me know your valued comments.
Regards,
- Attachments
-
- gamma-transition-TQ_dri.txt
- (34.79 KiB) Downloaded 249 times
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Omid,
You still defined diffusion from database for phase 0. For this reason you are still asked to link phase 0 to a phase in the database. If you use "diagonal n" in section "Diffusion Data" for phase 0 as I proposed, you will not be asked any more.
As relinearisation scheme you have chosen "automatic", which requires to define a critical temperature difference. In principle, this option is smart because it takes action only when a certain deviation is reached. But you need to chose this parameter such that it will not relinearize in each time step (very slow) or not at all (very inaccurate). I personally prefer using a constant time interval for relinearisation (e.g. "manual 0.1") instead.
Bernd
You still defined diffusion from database for phase 0. For this reason you are still asked to link phase 0 to a phase in the database. If you use "diagonal n" in section "Diffusion Data" for phase 0 as I proposed, you will not be asked any more.
As relinearisation scheme you have chosen "automatic", which requires to define a critical temperature difference. In principle, this option is smart because it takes action only when a certain deviation is reached. But you need to chose this parameter such that it will not relinearize in each time step (very slow) or not at all (very inaccurate). I personally prefer using a constant time interval for relinearisation (e.g. "manual 0.1") instead.
Bernd
Re: Diffusion at Austenite-Cr7C3 Interface
Dear Omid,
I had a closer look at your input file. There is a conceptual problem: the simulation has no control!
From a physical point of view, the phase transformations between M23C6 or M7C3 with fcc could be controlled by Cr diffusion. But then, the transformation would be extremely slow.
Another type of growth is possible if the carbides have a high solubility range for Cr so that they can grow without diffusion of Cr. This is probably the case here. Then, growth can be much faster, and is controlled by Carbon diffusion. And this is a problem, because you set C diffusion to infinite...
Within the latter case there are two subcases, depending on the redistribution kinetics of Cr: Para-Equilibrium or nple (no partition - local equilibrium). I do not know which one is more realistic in your case.
For the MICRESS simulation this means that we should not define Cr as stoichiometric in the two carbide phases because we need to take advantage of the high solubility range (although I told you the contrary above...). Furthermore, either "nple" or "para" should be used for Cr in M23C6 and M7C3. If we do not apply any redistribution control, the behavior will be intermediate, closer to "para".
Nevertheless, such a simulation is difficult, because there is already a M7C3 existing at the beginning with equilibrium Cr content. When growth starts with low Cr content (that of the fcc phase),
there will be a strong gradient inside the interface which can cause problems. Another problem is the far extrapolation of the phase diagram which is necessary within these redistribution models and which easily produces negative concentrations inside the interface.
I sent you an adapted .dri file by email which takes into account the above considerations.
Bernd
I had a closer look at your input file. There is a conceptual problem: the simulation has no control!
From a physical point of view, the phase transformations between M23C6 or M7C3 with fcc could be controlled by Cr diffusion. But then, the transformation would be extremely slow.
Another type of growth is possible if the carbides have a high solubility range for Cr so that they can grow without diffusion of Cr. This is probably the case here. Then, growth can be much faster, and is controlled by Carbon diffusion. And this is a problem, because you set C diffusion to infinite...
Within the latter case there are two subcases, depending on the redistribution kinetics of Cr: Para-Equilibrium or nple (no partition - local equilibrium). I do not know which one is more realistic in your case.
For the MICRESS simulation this means that we should not define Cr as stoichiometric in the two carbide phases because we need to take advantage of the high solubility range (although I told you the contrary above...). Furthermore, either "nple" or "para" should be used for Cr in M23C6 and M7C3. If we do not apply any redistribution control, the behavior will be intermediate, closer to "para".
Nevertheless, such a simulation is difficult, because there is already a M7C3 existing at the beginning with equilibrium Cr content. When growth starts with low Cr content (that of the fcc phase),
there will be a strong gradient inside the interface which can cause problems. Another problem is the far extrapolation of the phase diagram which is necessary within these redistribution models and which easily produces negative concentrations inside the interface.
I sent you an adapted .dri file by email which takes into account the above considerations.
Bernd