Post
by Bernd » Tue Aug 26, 2025 7:29 pm
Dear Ahmet,
That is a very intersting idea. If I understood well, you want to simulate melting of a powder mixture consisting of particles types containing only of pure elements each. In principle, this could work perfectly, but there are quite a number of potential pitfalls. Therefore, it makes sense to start with a simple case with two elements and particle types, and go on with multiple components later-on.
I think, the biggest problem could be that melting of such a multi-particle mixture is far from equilibrium at the very beginning. I once simulated melting of a multicomponent braze consisting of 4 different solder powders and also experienced such problems. But at the end, I got it done!
The most straight-forward approach would be to define an initial microstructure with two different solid phases. You can easily define the initial concentrations for each phase (consisting of one element only). The phase state should be stable phase of each pure elemet before melting. Please note that it is necessary to include a small amount of all other elements (e.g. 1.E-4 wt%) within each phase and not just set them to 0! This avoids unnecessary numerical problems. Furthermore, MICRESS cannot treat the empty space between the powder particles, so the initial microstructure should be dense (but see below!).
The numerical difficulties of nucleating liquid phase at e.g. an interface between pure Al and Fe will probably be severe. It is extremely important to enter into the correct phase equilibria upon initialisation, so you will have to tweek the temperature and phase fractions for initalisation. If you do not get it running nevertheless, this could be a helpful workaround: Allow some space between the particles and fill it with an inactive phase (use a higher phase number in MICRESS which may be linked to LIQUID in the database, but without interactions with any other phase and without diffusion. The initial composition of this phase could be set to some mixture composition resembling an early melt phase. When the melting temperature of this phase is reached, you would switch this phase to real liquid (using the add_to_grain option), so that melting of the particles already starts with some liquid with already reasonable composition. This should make thing numerically much more easy without affecting the overall behaviour too much, as long as the amount of the initial melt phase is small. In my brazing simulation, I just defined the lowest-melting braze powder as such an inactive liquid and started the melting simulation by "activating" the melt in the described way.
If in the powder different elements exist as the same phase (e.g. Al-fcc/Fe-fcc), it is important to have them as different composition sets in the .GES5 file and define two distinct phases for them in MICRESS. Otherwise, thermodynamics will get crazy and numerical problems are inevitable.
A further complication is that elements of the melt can dissolve into the high-melting particles and form intermetallic phases at different places. There is no single way to find out which phases could get stable somewhere and thus need to be included in the simulation. You will perhaps need various Thermo-Calc calculations to evaluate these phases for each particle type, especially if you move on to multi-component alloys.
Please tell us about your experiences. I am keen on hearing whether my proposed approach workes for you.
Bernd
