Error when linearisation!

dendritic solidification, eutectics, peritectics,....
CharMIC
Posts: 211
Joined: Mon Aug 21, 2017 5:01 pm
anti_bot: 333

Re: Error when linearisation!

Post by CharMIC » Fri Mar 30, 2018 8:57 am

Hej Bernd,

I was referring to this

"Thermo-Calc automatically creates new composition sets if global minimization is enabled"

BR
Chamara

CharMIC
Posts: 211
Joined: Mon Aug 21, 2017 5:01 pm
anti_bot: 333

Re: Error when linearisation!

Post by CharMIC » Fri Mar 30, 2018 9:48 am

Bernd wrote:Hi Chamara,

When different composition sets of a phase like Laves are coming into play, it is important to first have a look on phase constitution:

Laves_Constitution.GIF

As you can see in the Thermo-Calc System Definer, C14_LAVES has two sublattices which accept all your elements but C.

In order to get a feeling which lattice constitutions could be stable, you then should make a Thermo-Calc equilibrium calculation close to your local conditions in MICRESS (e.g. interdendritic composition and T=800K):

Laves_Composition_Sets.GIF

In case of a single equilibrium calculation, you can evaluate the phase constitution of C14_LAVES in the Table Renderer. Here, it is important to suppress other phases which are metastable (like delta, sigma, etc.). In this (arbitrarily chosen) case you see that Thermo-Calc automatically has chosen two composition sets with arbitrary numbers. The constitution shows different sublattice site fractions for the elements in the different composition sets.

Thermo-Calc automatically creates new composition sets if global minimization is enabled, which is not possible (and also not wanted) in MICRESS. Therefore, it is necessary to create and correctly define those composition sets already when creating the .ges5 file.

Now, if problems with demixing of the phase compositions exist, like in your case, when just the "normal" C14_LAVES has been used, the user should proceed in the following order:

1.) Correctly define the default major components for the phase according to your knowledge or expectation

When you create the .ges5 file, just before saving it while you are still inside the GES module, use AMEND_PHASE_DESCRIPTION and specify the phase name ("C14_LAVES"). Now, chose to amend MAJOR_CONSTITUENT, which allows you specifying the major constituent(s) for each sublattice. In case the C14_LAVES#3 is the one you prefer, you should specify Fe on the first and Nb and Mo on the second sublattice. This will assure that the start values for calculation of equilibria are set accordingly. You can control this choice through the MICRESS output "Start Composition for iteration of quasi-equilibrium" which is written to the screen output and .log file during initialisation.

2.) If setting the major constituents is not sufficient to prevent demixing, you should use limits or relative criteria in MICRESS for not allowing to enter into another composition set. Especially the limits should be used with care and only for one element which is the best to distinguish the composition sets. Please have in mind that limits are specified in at% in MICRESS.

3.) If you still have problems with demixing, then the other composition set(s) which cannot be prevented from forming seem to be necessary and should be included into the MICRESS simulation. Then you have to go back to Thermo-Calc and add the composition set to the .ges5 file (if not already present). Use AMEND_PHASE_DESCRIPTION with the suboption COMPOSITION_SET to create a new composition set according to your expectations.

Bernd
Hej Bernd,

I did some simulation after adjusting the .GES5 file as you have mentioned in point 1. When I look at the .TabLin out put now I don't see sudden changes in the Nb and Mo composition compared with the previous case. I have attached the file also with this.
But I still get demixing problem (Thermo-calc error 1611 MICRESS error 18,16,15,14 phase 1/2)

I have not set up any limits or criteria in this simulation. Since Nb wt% don't change in the .TabLin file, could we say that it is not switching to another composition set? or I am wrong here?

For your reference after adjusting the .GES5 file, here is the starting composition for LAVES

Start Composition for iteration of quasi-equilibrium
****************************************************
NI in C14_LAVES: 0.8558491
AL in C14_LAVES: 0.7503150
CR in C14_LAVES: 0.7714218
FE in C14_LAVES: 62.90079
MO in C14_LAVES: 16.73225
NB in C14_LAVES: 17.11242
TI in C14_LAVES: 0.8769559

BR
Chamara
Attachments
LMD_HighHI_ThermalCycles_4.TabLin
(88.3 KiB) Downloaded 166 times

Bernd
Posts: 1179
Joined: Mon Jun 23, 2008 9:29 pm

Re: Error when linearisation!

Post by Bernd » Fri Mar 30, 2018 1:34 pm

Hej Chamara,

It looks as if you were successful with step 1 already! There are no signs of demixing anymore. The error messages which you mentioned are not specifically related to a composition set problem, so they can have any other reason.

To your former comment: When creating .ges5 files, you typically don't enter the POLY3 module, thus the option for global minimization cannot be set and/or does not apply. Composition sets are created because of AMEND_PHASE_DESCRIPTION commands which are directly part of the .tdb database file
(indeed, if before saving the .ges5 file you would really enter poly3 and do some equilibrium calculations, the created composition sets would also be included in the .ges5 file. However, we do not want that but rather define them explicitly...).

Bernd

CharMIC
Posts: 211
Joined: Mon Aug 21, 2017 5:01 pm
anti_bot: 333

Re: Error when linearisation!

Post by CharMIC » Fri Mar 30, 2018 1:54 pm

Hej Bernd,

Many thanks again fro the reply :)

How should I proceed to figure out the cause of error massages. I get them quite frequently.
Attachments
12.JPG
12.JPG (124.42 KiB) Viewed 2626 times

Bernd
Posts: 1179
Joined: Mon Jun 23, 2008 9:29 pm

Re: Error when linearisation!

Post by Bernd » Fri Mar 30, 2018 2:25 pm

At which temperature does this happen?
Do you use "mob_corr"?
How long is the diffusion length compared to interface thickness?

As I said, the error messages are not very specific...

Bernd

CharMIC
Posts: 211
Joined: Mon Aug 21, 2017 5:01 pm
anti_bot: 333

Re: Error when linearisation!

Post by CharMIC » Fri Mar 30, 2018 2:44 pm

Hej Bernd,

I started to get the error massage around 1028K
Yes, I used "mob_corr"

My interface thickness is 2.5 cells (0.125µm)
Cell size 0.05µm
Domain size 30µm * 20µm
I have initial seed set up at 8µm apart since experimentally measured PDAS is 8µm.
I uses "database global" for phase diagram of 1(gamma) and 2(LAVES) without no relinearisation mode.

If you need I can email the dri file to you

BR
Chamara

Bernd
Posts: 1179
Joined: Mon Jun 23, 2008 9:29 pm

Re: Error when linearisation!

Post by Bernd » Fri Mar 30, 2018 3:06 pm

A further question: Did you use stoichiometric condition for all elements in LAVES? This is strongly recommended because Ni is not the major component in this phase.

Bernd

CharMIC
Posts: 211
Joined: Mon Aug 21, 2017 5:01 pm
anti_bot: 333

Re: Error when linearisation!

Post by CharMIC » Fri Mar 30, 2018 3:30 pm

I used only "interaction" and "no_stoichio".

So Should I used stoichiometric condition for the simulation where I modified the .GES5 file as follows?

2 0 1 2 3 4 5 6

Here first number "2" is for LAVES phase. I am not sure of using the "0" (Ni) element in this description

In addition to that what is the difference between "no_stoichio" vs "no_more__stoichio"

Upto now I mainly used the option "no_stoichio"

BR
Chamara

Bernd
Posts: 1179
Joined: Mon Jun 23, 2008 9:29 pm

Re: Error when linearisation!

Post by Bernd » Fri Mar 30, 2018 7:08 pm

Hej Chamara,

When defining stoichiometric conditions, it is allowed to include element 0 (matrix component), but (at least at present) it has no effect. This is different for limits and "criterion_higher" / "criterion_lower", which can also be used for the matrix element.

Thus, you should use

2 1 2 3 4 5 6

or

2 1-6

"no_stoichio" and "no_more_stoichio" have exactly the same meaning - the alternatives only exist in order to fulfill semantics for the case yo have or you haven't defined any stoichiometric phases...

Bernd

CharMIC
Posts: 211
Joined: Mon Aug 21, 2017 5:01 pm
anti_bot: 333

Re: Error when linearisation!

Post by CharMIC » Sun Apr 01, 2018 5:26 pm

Hej Bernd,

Thanks aging your help on this problem. Really appropriate it. :)

I ran the simulation with stoichiometric conditions as you suggested with Amend .GES file.

Under phase diagram section, I used the following

2 1-6
interaction
no_more_stoichio

This time I got the following errors at lower temperature (Around 750K, previously I got errors around 1028K). What I observed in .TabLin file was sudden shift in the Nb wt%. It was like previous, it tries to form Mo rich Laves. So I before I run simulation with "limits", I wanted to know is the error massage related to demixing or not?

In addition to that, I see low Mo content in the LAVES than matrix. But in the experimental conditions you see higher Mo content in Laves than Matrix (I have posted experimental data from an article before in this thread). I am not sure how to fix this. Get a stable simulation and at the same time having higher Mo content in LAVES than matrix. Is this possible?

BR
Chamara
Attachments
12.JPG
12.JPG (79.53 KiB) Viewed 2586 times

Post Reply