6
$\begingroup$

When running single point equilibrium calculations on an array of compositions I come across some compositions (for specific examples see below) that return the error "ERROR 1611 IN QTHISS: TOO MANY ITERATIONS". For most such compositions if I increase the "max number of grid points" parameter I no longer get the error and instead get the result of the calculation (eq. phases, fractions, etc.). However, there are some compositions such that even when I increase the "max number of grid points" to something like 16,000,000 (which makes the calculation use a lot of RAM and take orders of magnitude longer to run) they still give this error.

I have a question:

What is the reason some compositions fail to calculate? Are they just at a part of the free energy landscape that makes it difficult for the minimizer?

An example of such a composition is Ni0.05Ti0.32Nb0.59Al0.04 (atomic frac.) at 1323K using the TCHEA5.0 database. It always gives an error up to a max grid points of 20,000,000.

A few notes:

  • It seems that the calculations are affected by prior calculations (which makes sense because one calculation will change the state of the "System" which will then alter how subsequent calculations are performed). For instance, running Ni0.05Ti0.33Nb0.58Al0.04 and then Ni0.05Ti0.32Nb0.59Al0.04 (with the conditions above) gives a result, but running Ni0.05Ti0.32Nb0.59Al0.04 only gives an error.
  • The persistent errors seems to occur for compositions where there are 2 or more composition sets of a given phase in equilibrium (e.g., at a composition where there is BCC_B2#1-BCC_B2#2 or FCC_A1#2-FCC_A1#3). For example, there are 3 BCC phases in equilibrium for Ni0.05Ti0.32Nb0.59Al0.04 and Ni0.05Ti0.32Nb0.59Al0.04.

I know that such errors may seem like a minor issue to some, but for me its a show-stopper so I would very much appreciate some help.

$\endgroup$

1 Answer 1

-1
$\begingroup$

If you use Thermo-Calc and database provided by TCSAB, which I assume is encrypted, there is very little help you can get from anyone expect from TCSAB itself. But there are some free High Entropy Alloy (HEA) databases around and I may be able to point at some and then you can use other software such as OpenCalphad which is also free or MetCalc or Pandat (which are commercial and may have theor own HEA alloy databases). I am not sure how good TC-Python is for multicomponent systems.
The calculation problem is usually that the difference in Gibbs energy between the verious phases is very small in certain composition and T regions and there are ordering transformations in FCC and BCC which complicate finding the most stable state.

The question is also how much one can trust the result. All multicomponent calculations are based on model parameters determined from experiments in binary and ternary alloys. A few multicomponent experimental data may be used to improve such extrapolations. There are many useful papers in the Calphad journal which can be helpful. It the calculation works at 1300K but not 1325 you may use the result at 1300K, the uncertainty is probably more than 25K. Or you can calculate at 1350K you can interpolate manually. Or vary the composition a little and look for experimental results in similar alloys.

You can find OC at opencalphad.org or at sundmanbo/opencalphad at Github and you are welcome to contact me by email. I am retired but still trying to promote good use of thermodynamics and modeling.

$\endgroup$
1
  • $\begingroup$ @sgp45 if detailed answers by some of the most knowledgeable experts in the field, are going to get downvoted, it seems like your 100-point bounty won't attract a lot of effort from others. I hope I can encourage you to reconsider your voting practices. I have upvoted 6000+ times on this site and downvoted a grand total of 0 times. If you don't like an answer, you can always leave a comment or click "flag" and ask the community to peer-review it! $\endgroup$ Oct 13, 2023 at 20:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .