Dear sir or madam,
sorry to bother you, I am currently performing a phase-field simulation of Cu-base alloy. While nucleation is successfully triggered in my simulation, I am encountering an issue with the subsequent remelting and solidification process.
Specifically, the transformation from both solid to liquid and liquid to solid is extremely slow, and the crystal growth rate appears to be much lower than expected. this behavior seems unphysical for my system and suggests that some parameters may not be properly set.
I would like to ask for your advice on which parameters I should focus on to improve the solidification kinetics. I have already try to change the kinetics coefficients in the phase intereaction parts, but still can not get what I expected.
For your reference, I have attched my driving files, ges5 file and temperature field. Any feedback on them would be highly appreciated.
Thank you very much for your time and support.
Best regards,
Dai He
Question on parameters setting for Cu-base alloy solidification kinetics
Question on parameters setting for Cu-base alloy solidification kinetics
You do not have the required permissions to view the files attached to this post.
Re: Question on parameters setting for Cu-base alloy solidification kinetics
Dear Dai He,
Your problem comes from the diffusion coefficients: If you have a look at the .diff-output or at the .TabD-file, you see that all diffusion coefficients are zero. This automatically leads to a tiny interface mobility (see .mueS-output), because you request diffusion limited interface kinetics by using "mob_corr" together with a large input value of the (physical) mobility.
Could it be that there was an incompatibility of your mobility database, or you did not append a mobility database at all? If you do not have access to a mobility database suitable for your system, you should just define the diffusion coefficients in the liquid phase manually (e.g. 1.D-5 cm2/s for all elements). The diffusivity of the fcc-phase can be disregarded in a first step.
Best wishes
Bernd
Your problem comes from the diffusion coefficients: If you have a look at the .diff-output or at the .TabD-file, you see that all diffusion coefficients are zero. This automatically leads to a tiny interface mobility (see .mueS-output), because you request diffusion limited interface kinetics by using "mob_corr" together with a large input value of the (physical) mobility.
Could it be that there was an incompatibility of your mobility database, or you did not append a mobility database at all? If you do not have access to a mobility database suitable for your system, you should just define the diffusion coefficients in the liquid phase manually (e.g. 1.D-5 cm2/s for all elements). The diffusivity of the fcc-phase can be disregarded in a first step.
Best wishes
Bernd
Re: Question on parameters setting for Cu-base alloy solidification kinetics
Dear Bernd,
Thanks for your reply.
First of all, I would like to sincerely thank you for your previous support. With your guidence, I was able to successfully run the simulations considering only the FCC and liquid phases.
I am now trying to extend the model by including additional phases. However, I have encountered some difficulties. Since the MOBCU4 database which I have does not provide the data for the phases I am using, I attempted to define diffusion manually.
Unfortuantely, this led to a error in the scr file shown the error below and the other phases can not grow up although the nucleation was triggered.
trying hard phases 1 0 level: 4 zp= 1023776 error= 1
trying harder! Error = 1
trying hard phases 1 0 level: 5 zp= 1053756 error= 2
trying harder! Error = 2
trying hard phases 1 0 level: 5 zp= 1123438 error= 2
trying harder! Error = 2
trying hard phases 0 1 level: 3 zp= 428517 error= 1
trying hard phases 1 0 level: 5 zp= 440359 error= 1
Thermo-Calc error 1611 MICRESS error 14 Thermo-Calc
trying hard phases 1 0 level: 4 zp= 461263 error= 1
Thermo-Calc error 1611 MICRESS error 15 Thermo-Calc
--> Force start values again
trying hard phases 1 0 level: 4 zp= 504639 error= 1
trying harder! Error = 1
trying hard phases 0 1 level: 8 zp= 511962 error= 104
trying harder! Error = 1
At this point, I am unsure how to proceed. I would greatly appreciate your advice on the folowing:
1. How should I improve or adjust the diffusion parameters I have defined?
2. Are there any recommanded strategies for estimating or calibrating diffusion coefficients in such cases?
3. Are there any other model can be used in my cases except atc mob_corr.
Any guidance you could provide would be very valuable to me.For your reference, I have attached my codes, scr and log files
Thank you again for your support.
Best regards
Dai He
Thanks for your reply.
First of all, I would like to sincerely thank you for your previous support. With your guidence, I was able to successfully run the simulations considering only the FCC and liquid phases.
I am now trying to extend the model by including additional phases. However, I have encountered some difficulties. Since the MOBCU4 database which I have does not provide the data for the phases I am using, I attempted to define diffusion manually.
Unfortuantely, this led to a error in the scr file shown the error below and the other phases can not grow up although the nucleation was triggered.
trying hard phases 1 0 level: 4 zp= 1023776 error= 1
trying harder! Error = 1
trying hard phases 1 0 level: 5 zp= 1053756 error= 2
trying harder! Error = 2
trying hard phases 1 0 level: 5 zp= 1123438 error= 2
trying harder! Error = 2
trying hard phases 0 1 level: 3 zp= 428517 error= 1
trying hard phases 1 0 level: 5 zp= 440359 error= 1
Thermo-Calc error 1611 MICRESS error 14 Thermo-Calc
trying hard phases 1 0 level: 4 zp= 461263 error= 1
Thermo-Calc error 1611 MICRESS error 15 Thermo-Calc
--> Force start values again
trying hard phases 1 0 level: 4 zp= 504639 error= 1
trying harder! Error = 1
trying hard phases 0 1 level: 8 zp= 511962 error= 104
trying harder! Error = 1
At this point, I am unsure how to proceed. I would greatly appreciate your advice on the folowing:
1. How should I improve or adjust the diffusion parameters I have defined?
2. Are there any recommanded strategies for estimating or calibrating diffusion coefficients in such cases?
3. Are there any other model can be used in my cases except atc mob_corr.
Any guidance you could provide would be very valuable to me.For your reference, I have attached my codes, scr and log files
Thank you again for your support.
Best regards
Dai He
You do not have the required permissions to view the files attached to this post.
Re: Question on parameters setting for Cu-base alloy solidification kinetics
Dear Dai He,
This time, the problem has nothing to do with the diffusion data. The problem is that you start with a highly metastable system consisting of fcc and liquid phase only, while Laves phase should be present even at high temperatures. This leads to strange effects like the needle-like melting of the fcc. Furthermore, the liquid phase obviously runs into a miscibility gap, leading to fluctuations of the liquid composition and to morphological instabilities of the dendrites.
This problem is aggravated because you even do not allow for nucleation of Laves or Bcc phases. To be honest, I do not understand your nucleation strategy, which you have followed with the 10 seed types defined. Please note that "split_from_grain" is typically used for removing small rest liquids remaining at low temperature for numerical reasons, but it does not provide nucleation of new phases from the melt! By the way, the number of seeds of liquid into fcc (seed type 1) is highly exaggerated - in the present setup, essentially, you would not need any of them...
If your goal is the simulation of melting and re-solidification of a prior deposited Cu-Cr-Zr-Nb-alloy under SLM conditions, I would propose to start with the simulation of solidification only including all stable phases, beginning from a completely liquid initial state. By this way, you can obtain a thermodynamically stable and consistent initial microstructure which you can use as initial microstructure for the melting/re-solidification without instabilities.
Alternatively, it may also be possible to nucleate Laves phase already from liquid during melting and ahead of the growing dendrites. Then, Cr and Nb would perhaps be already depleted sufficiently to avoid demixing of the liquid.
Best wishes
Bernd
This time, the problem has nothing to do with the diffusion data. The problem is that you start with a highly metastable system consisting of fcc and liquid phase only, while Laves phase should be present even at high temperatures. This leads to strange effects like the needle-like melting of the fcc. Furthermore, the liquid phase obviously runs into a miscibility gap, leading to fluctuations of the liquid composition and to morphological instabilities of the dendrites.
This problem is aggravated because you even do not allow for nucleation of Laves or Bcc phases. To be honest, I do not understand your nucleation strategy, which you have followed with the 10 seed types defined. Please note that "split_from_grain" is typically used for removing small rest liquids remaining at low temperature for numerical reasons, but it does not provide nucleation of new phases from the melt! By the way, the number of seeds of liquid into fcc (seed type 1) is highly exaggerated - in the present setup, essentially, you would not need any of them...
If your goal is the simulation of melting and re-solidification of a prior deposited Cu-Cr-Zr-Nb-alloy under SLM conditions, I would propose to start with the simulation of solidification only including all stable phases, beginning from a completely liquid initial state. By this way, you can obtain a thermodynamically stable and consistent initial microstructure which you can use as initial microstructure for the melting/re-solidification without instabilities.
Alternatively, it may also be possible to nucleate Laves phase already from liquid during melting and ahead of the growing dendrites. Then, Cr and Nb would perhaps be already depleted sufficiently to avoid demixing of the liquid.
Best wishes
Bernd
Re: Question on parameters setting for Cu-base alloy solidification kinetics
Dear Bernd,
Thank you again for your previous reply. Following your suggestion, I have started with a simplified version of the model.
In this current setup, I only consider the Cu–Cr–Nb system and focus exclusively on bulk nucleation during the solidification, without including initial microstructure and remelting at this stage.
However, I have encountered a numerical issue in the simulations. In the BCC_B2 and C15_Laves phases, the fraction of alloying elements is relatively high. During nucleation, the enrichment of Cr and Nb from the surrounding liquid into the nuclei leads to a depletion of these elements in nearby liquid cells. As a result, the concentrations in some liquid regions drop to negative values, while in certain nucleation sites they exceed 100%. This ultimately results in the error. I have tried to follow some advises in the Forum such as change the concentration solver to ''diagonal'' but still could not fix it.
Could you please advise on how I might modify the model to avoid this issue or improve its stability?
For your reference, I have attached the driving file.
Thank you very much for your time and support.
Best regards,
Dai He
Thank you again for your previous reply. Following your suggestion, I have started with a simplified version of the model.
In this current setup, I only consider the Cu–Cr–Nb system and focus exclusively on bulk nucleation during the solidification, without including initial microstructure and remelting at this stage.
However, I have encountered a numerical issue in the simulations. In the BCC_B2 and C15_Laves phases, the fraction of alloying elements is relatively high. During nucleation, the enrichment of Cr and Nb from the surrounding liquid into the nuclei leads to a depletion of these elements in nearby liquid cells. As a result, the concentrations in some liquid regions drop to negative values, while in certain nucleation sites they exceed 100%. This ultimately results in the error. I have tried to follow some advises in the Forum such as change the concentration solver to ''diagonal'' but still could not fix it.
Could you please advise on how I might modify the model to avoid this issue or improve its stability?
For your reference, I have attached the driving file.
Thank you very much for your time and support.
Best regards,
Dai He
You do not have the required permissions to view the files attached to this post.
Re: Question on parameters setting for Cu-base alloy solidification kinetics
Dear Dai He,
In principle, it is a good idea to focus first on solidification starting from a fully liquid domain. However, in the way you do it, you force all solid phases to nucleate and grow under extremely high undercooling. This makes simulation complicated, because you essentially would need a much higher grid resolution for this purpose, and because the risk is high to enter into wrong metastable equilibria. This is probably not what you inted to do. Therefore, you should start at much higher temperatures.
The formation of negative compositions due to strong segregation is a consequence of the linear extrapolation approach of the phase diagram (see also here). To monitor such behaviour you should have a look on the compositions in the liquid (.c1pha0 and .c2pha0), where negative compositions can be seen first. Furthermore, I recommend to check that for each phase individually, because the nucleation of several phases at the same time may lead to interdependencies which make it more difficult to understand what is going on.
Given that negative phase compositions (or also those higher than 100%) are caused by the linear extrapolation method of redistribution, it is clear that this problem is strongly aggravated by high undercooling and by a too high time interval for updating linearisation data. Both is the case for you! Therefore, I would start with reducing the initial undercooling and decreasing the update interval for thermodynamic data. If after that, if there are still negative concentrations in the liquid phase which start forming around the growing particle, you may use a penalty on the driving force (Numerical Parameters/Concentration Solver) to avoid the problem, e.g. for the interface liquid/BCC_B2 and element 2:
penalty 0 2 2
A further complication is that in principle all solid phases in your system can form various composition sets (concentration ranges) which must be regarded as different phases. Please see here for more detailed information. In your case, this may most probably happen with the LAVES phase, which can either have a composition close to pure Nb or have similar amounts of Cr and Nb. You should check the .TabLin output to find out which one was selected. If only one of the two composition sets has a physical relevance for you, and TQ-calclations sometimes converge into the wrong set, then you should use "limits" to avoid that, and check whether the start compositions are properly set in the .GES5-file (see link above).
Bernd
In principle, it is a good idea to focus first on solidification starting from a fully liquid domain. However, in the way you do it, you force all solid phases to nucleate and grow under extremely high undercooling. This makes simulation complicated, because you essentially would need a much higher grid resolution for this purpose, and because the risk is high to enter into wrong metastable equilibria. This is probably not what you inted to do. Therefore, you should start at much higher temperatures.
The formation of negative compositions due to strong segregation is a consequence of the linear extrapolation approach of the phase diagram (see also here). To monitor such behaviour you should have a look on the compositions in the liquid (.c1pha0 and .c2pha0), where negative compositions can be seen first. Furthermore, I recommend to check that for each phase individually, because the nucleation of several phases at the same time may lead to interdependencies which make it more difficult to understand what is going on.
Given that negative phase compositions (or also those higher than 100%) are caused by the linear extrapolation method of redistribution, it is clear that this problem is strongly aggravated by high undercooling and by a too high time interval for updating linearisation data. Both is the case for you! Therefore, I would start with reducing the initial undercooling and decreasing the update interval for thermodynamic data. If after that, if there are still negative concentrations in the liquid phase which start forming around the growing particle, you may use a penalty on the driving force (Numerical Parameters/Concentration Solver) to avoid the problem, e.g. for the interface liquid/BCC_B2 and element 2:
penalty 0 2 2
A further complication is that in principle all solid phases in your system can form various composition sets (concentration ranges) which must be regarded as different phases. Please see here for more detailed information. In your case, this may most probably happen with the LAVES phase, which can either have a composition close to pure Nb or have similar amounts of Cr and Nb. You should check the .TabLin output to find out which one was selected. If only one of the two composition sets has a physical relevance for you, and TQ-calclations sometimes converge into the wrong set, then you should use "limits" to avoid that, and check whether the start compositions are properly set in the .GES5-file (see link above).
Bernd