Hi Bernd,
In my simulation I was getting Large negative compositions in both Liquid and Gamma_prime phase towards the end of the simulations. I have sent you the information showing this though email. What is the root cause that you think for such behavior.
Simulation of solidification of Gamma prime alloy during SLM process
Re: Simulation of solidification of Gamma prime alloy during SLM process
Hi Chamara,
Negative compositions are a frequent problem in complex multicomponent systems, and there different factors which can lead to that. According to my experiences, the most frequent factors are:
a) MICRESS uses extrapolation methods for redistribution of elements between phases, in order to avoid the necessity of calculating quasi-equilibrium in each time step (which would lead to extremely high calculation times). This extrapolation does not (and also cannot) respect the solubility limits of the phases. Negative compositions which are produced by a too far extrapolation can in principle be avoided by more frequent relinearisation. However, in many cases, this is not possible without increasing simulation times too much. The use of composition limits ("limits"), which are connected with penalty terms on the driving force, can alternatively be used to avoid the problem.
b) In case of stoichiometric phases, updates of the quasi-equilibrium can lead to shifted compositions. Because no solubility is assumed, this can easily lead to negative compositions in the other phases. This problem typically does not appear if local relinearisation (i.e. in each grid point separately) is used, because the mass balance of each element then must be fulfilled while calculating quasi-equilibrium. However, as local relinearisation is expensive, global relinearisation schemes need to be used in many cases. Then, the mass balance is fulfilled only on the global scale, while negative composition can easily occur locally. The problem can be aggravated if inconsistent relinearisation schemes are combined (e.g. globalF for phases 0/1 and global for 0/3).
A typical case here is MC-phase (FCC_L12#3) which must be assumed to be stoichiometric for numerical reasons but which can drastically change composition with temperature. I cannot see from the results you sent me whether this factor contributes to your problem.
c) If the same phase can exist with different composition sets, or as ordered and disordered phase like in case of γ/γ', switching between the different sets or states can occur. This is especially bad if on top of all it is not happening consistently: If e.g. γ' switches to γ when quasi-equilibrium of liquid/γ' is calculated, but stays ordered in the γ/γ'-equilibrium, inconsistencies are especially strong an will also lead to negative compositions. This, I think is the case in your simulation, as can be seen in the .TabLin output which you sent me: Just compare the compositions of the typical γ- and γ'-forming elements like Al or Cr in the different phase interactions.
The key issue here is to avoid "switching" at all costs. This can be achieved in MICRESS by different means:
- absolute composition limits ("limits"): If there is no overlap of compositions occurring in both sets or states for one of the elements, then such an absolute limit can avoid switching. This is very helpful e.g. in the case of fcc/MC-carbide (FCC_L12#1, FCC_L12#3), where the carbon composition can be used to separate the two composition sets. Unfortunately, this is not as easy for fcc/γ'.
- relative composition limits ("criterion_higher"/"criterion_lower"): In case of the distinction of γ/γ' it is easier to separate the two states by comparison of the contents of an element in both phases. For the γ/γ'-equilibrium we can use e.g. the Al content:
"criterion_higher 3 1 6" can assure that the composition of Al (component 6) must always be higher in γ' (phase 3) than in fcc (phase 1). In the same way, in the liquid/γ' interaction, Cr should always be higher in liquid than in γ': "criterion_higher 0 3 1". This was often helpful for me.
- in the next MICRESS release, there will be an option which directly checks for order/disorder of the phases by comparing the site fractions of the elements in the first and second sublattice of the FCC_L12 phase:
"ordered 3 1 2", and
"disordered 1 1 2"
This is the most effective solution to the problem, if the phases are modeled with an order/disorder transition like it is the case for FCC_L12 in the TCNI databases.
Bernd
Negative compositions are a frequent problem in complex multicomponent systems, and there different factors which can lead to that. According to my experiences, the most frequent factors are:
a) MICRESS uses extrapolation methods for redistribution of elements between phases, in order to avoid the necessity of calculating quasi-equilibrium in each time step (which would lead to extremely high calculation times). This extrapolation does not (and also cannot) respect the solubility limits of the phases. Negative compositions which are produced by a too far extrapolation can in principle be avoided by more frequent relinearisation. However, in many cases, this is not possible without increasing simulation times too much. The use of composition limits ("limits"), which are connected with penalty terms on the driving force, can alternatively be used to avoid the problem.
b) In case of stoichiometric phases, updates of the quasi-equilibrium can lead to shifted compositions. Because no solubility is assumed, this can easily lead to negative compositions in the other phases. This problem typically does not appear if local relinearisation (i.e. in each grid point separately) is used, because the mass balance of each element then must be fulfilled while calculating quasi-equilibrium. However, as local relinearisation is expensive, global relinearisation schemes need to be used in many cases. Then, the mass balance is fulfilled only on the global scale, while negative composition can easily occur locally. The problem can be aggravated if inconsistent relinearisation schemes are combined (e.g. globalF for phases 0/1 and global for 0/3).
A typical case here is MC-phase (FCC_L12#3) which must be assumed to be stoichiometric for numerical reasons but which can drastically change composition with temperature. I cannot see from the results you sent me whether this factor contributes to your problem.
c) If the same phase can exist with different composition sets, or as ordered and disordered phase like in case of γ/γ', switching between the different sets or states can occur. This is especially bad if on top of all it is not happening consistently: If e.g. γ' switches to γ when quasi-equilibrium of liquid/γ' is calculated, but stays ordered in the γ/γ'-equilibrium, inconsistencies are especially strong an will also lead to negative compositions. This, I think is the case in your simulation, as can be seen in the .TabLin output which you sent me: Just compare the compositions of the typical γ- and γ'-forming elements like Al or Cr in the different phase interactions.
The key issue here is to avoid "switching" at all costs. This can be achieved in MICRESS by different means:
- absolute composition limits ("limits"): If there is no overlap of compositions occurring in both sets or states for one of the elements, then such an absolute limit can avoid switching. This is very helpful e.g. in the case of fcc/MC-carbide (FCC_L12#1, FCC_L12#3), where the carbon composition can be used to separate the two composition sets. Unfortunately, this is not as easy for fcc/γ'.
- relative composition limits ("criterion_higher"/"criterion_lower"): In case of the distinction of γ/γ' it is easier to separate the two states by comparison of the contents of an element in both phases. For the γ/γ'-equilibrium we can use e.g. the Al content:
"criterion_higher 3 1 6" can assure that the composition of Al (component 6) must always be higher in γ' (phase 3) than in fcc (phase 1). In the same way, in the liquid/γ' interaction, Cr should always be higher in liquid than in γ': "criterion_higher 0 3 1". This was often helpful for me.
- in the next MICRESS release, there will be an option which directly checks for order/disorder of the phases by comparing the site fractions of the elements in the first and second sublattice of the FCC_L12 phase:
"ordered 3 1 2", and
"disordered 1 1 2"
This is the most effective solution to the problem, if the phases are modeled with an order/disorder transition like it is the case for FCC_L12 in the TCNI databases.
Bernd
Re: Simulation of solidification of Gamma prime alloy during SLM process
Hi Bernd,
Many thanks for the detailed explanation.
can both the "criterion_higher 3 1 6" and "criterion_higher 0 3 1" be used at the same time?
Out of curiosity, when is the next release of MICRESS is planned.
Many thanks for the detailed explanation.
can both the "criterion_higher 3 1 6" and "criterion_higher 0 3 1" be used at the same time?
Out of curiosity, when is the next release of MICRESS is planned.
Re: Simulation of solidification of Gamma prime alloy during SLM process
Hi Chamara,
Of course, you can use both options can be used simultaneously. You just add the corresponding lines in the inputs section for stoichiometric conditions.
Our next release is planned towards the end of this year, but there is still no fixed date. If you already got any beta-version of MICRESS from us after June last year, this should already contain the "ordered"/"disordered" options.
Bernd
Of course, you can use both options can be used simultaneously. You just add the corresponding lines in the inputs section for stoichiometric conditions.
Our next release is planned towards the end of this year, but there is still no fixed date. If you already got any beta-version of MICRESS from us after June last year, this should already contain the "ordered"/"disordered" options.
Bernd
Re: Simulation of solidification of Gamma prime alloy during SLM process
Using the "criteria_higher" option the negative composition in Liquid and Gamma_prime (FCC_L12#2) phase seems to solved the problem to a certian extent. Although I see some elements is reaching negative compositions in some areas during certain time interval.
Al - In liquid
Hf, C, B ,Zr in Gamma_prime
When I check the places using .phas output, the negative areas mostly appear at the interface locations (-1).
In addition now the negative commotions seems to appear in Gamma phase (FCC_L12).
I also get quite a few error messages like
Thermo-calc error 1611 MICRESS error 20 phase 0/3
Thermo-calc error 1611 MICRESS error 16 phase 0/1
Thermo-calc error 1611 MICRESS error 18 phase 1/1
and of-course some "trying hard phase..." errors
here
1 - Gamma (FCC_L12)
2 - MC (FCC_L12#3)
3 - Gamma_prime (FCC_L12#2)
Al - In liquid
Hf, C, B ,Zr in Gamma_prime
When I check the places using .phas output, the negative areas mostly appear at the interface locations (-1).
In addition now the negative commotions seems to appear in Gamma phase (FCC_L12).
I also get quite a few error messages like
Thermo-calc error 1611 MICRESS error 20 phase 0/3
Thermo-calc error 1611 MICRESS error 16 phase 0/1
Thermo-calc error 1611 MICRESS error 18 phase 1/1
and of-course some "trying hard phase..." errors
here
1 - Gamma (FCC_L12)
2 - MC (FCC_L12#3)
3 - Gamma_prime (FCC_L12#2)
Re: Simulation of solidification of Gamma prime alloy during SLM process
Hi Chamara,
It seems that the use of "criterion_higher" in this case does not completely prevent switching, for example because switching of γ to γ' in the 0/1-interfaces is still not controlled. It would be interesting to see whether the "ordered" option works better, or whether there are other reasons for your errors. I will send you a recent beta version per mail so that you might check.
Bernd
It seems that the use of "criterion_higher" in this case does not completely prevent switching, for example because switching of γ to γ' in the 0/1-interfaces is still not controlled. It would be interesting to see whether the "ordered" option works better, or whether there are other reasons for your errors. I will send you a recent beta version per mail so that you might check.
Bernd
Re: Simulation of solidification of Gamma prime alloy during SLM process
Hi Bernd,
Is this error also related to what you have discussed above in the thread.
Is this error also related to what you have discussed above in the thread.
- Attachments
-
- error 1610.PNG (8.57 KiB) Viewed 28238 times
Re: Simulation of solidification of Gamma prime alloy during SLM process
Hi Chamara,
The error number 15 shows that there are numerical difficulties while calculating the linearisation parameters of the fcc/MC interface (solvus slopes). It is difficult to guess what could be the reason. It could be linked to an order-disorder transition (i.e. fcc would switch to γ'). Then, when using the "ordered" keyword for phase 1, an error number 40100 would appear. It could also be that MC switches to fcc. Then, probably the 30201 error would appear (indicating that the compositions of phase 1 and 2 are ending up identical), and it would make sense to use "Limits 2 1" (assuming C is component no 1). Then, error=10201 would indicate when a given lower limit of C in MC (e.g. 30 at%) has been passed.
However, the error can also be due to other reasons...
Bernd
The error number 15 shows that there are numerical difficulties while calculating the linearisation parameters of the fcc/MC interface (solvus slopes). It is difficult to guess what could be the reason. It could be linked to an order-disorder transition (i.e. fcc would switch to γ'). Then, when using the "ordered" keyword for phase 1, an error number 40100 would appear. It could also be that MC switches to fcc. Then, probably the 30201 error would appear (indicating that the compositions of phase 1 and 2 are ending up identical), and it would make sense to use "Limits 2 1" (assuming C is component no 1). Then, error=10201 would indicate when a given lower limit of C in MC (e.g. 30 at%) has been passed.
However, the error can also be due to other reasons...
Bernd
Re: Simulation of solidification of Gamma prime alloy during SLM process
can it also comes due to inconsistency when defining global relinerisation options. e.g. globalF for phases 0/1 and global for 0/3
also is it okay to define all the options together like bellow so that switching of phases wont happen. Or defining everything makes things complecated
2 1-11
limits 2 9
30
100
interaction
criterion_higher 3 1 6
criterion_higher 0 3 1
ordered 3 1 2
diff_comp_sets 1 2 3
no_more_stoichio
here
1 - fcc
2 - MC
3 - Gamma_prime
also is it okay to define all the options together like bellow so that switching of phases wont happen. Or defining everything makes things complecated
2 1-11
limits 2 9
30
100
interaction
criterion_higher 3 1 6
criterion_higher 0 3 1
ordered 3 1 2
diff_comp_sets 1 2 3
no_more_stoichio
here
1 - fcc
2 - MC
3 - Gamma_prime
Re: Simulation of solidification of Gamma prime alloy during SLM process
Yes Chamara,
you can define it like that without problems to be expected. I just would prefer "interaction" to be replaced by "diagonal", which is the correct keyword in Version 6.4, but which does the same...
You are also right that global relinearisation options, and especially combining different ones, can cause problems. Only with local relinearisation the composition balance is locally correct, but you know how slow that is...
Bernd
you can define it like that without problems to be expected. I just would prefer "interaction" to be replaced by "diagonal", which is the correct keyword in Version 6.4, but which does the same...
You are also right that global relinearisation options, and especially combining different ones, can cause problems. Only with local relinearisation the composition balance is locally correct, but you know how slow that is...
Bernd