Discussion Forum

Hello everyone,

I would like to begin this thread by saying that (a) this forum is amazing as I have been learning so much in the last days and (b) I have just started using COMSOL and I am a noob at FEA in general, so please be merciful in your comments :)

That being said, I am trying to simulate laser melting of metals but, as a first simple approach, what I am interested in is to obtain a relatively accurate temperature distribution in the melt pool that forms when a laser scans over a slab of metal.

I started my model taking inspiration from the laser_heating_Si wafer tutorial that is available online: I created a parallelepiped of material (thickness x width x length) and a heat source with a gaussian profile. I am using heat transfer module with phase change, because I want to take into account the latent heat of melting of my material. I set the initial temperature of my slab to be room T (293 K) and I impose the same constant temperature to the sides and the bottom of the slab to be maintained throughout the simulation. For what concerns the top surface, there are both heat input from the laser (it scans along "length" direction at midpoint of '"width") and heat output due to the surface's emissivity.

When I run my model I obtain a nice melt pool that moves from one end to the other of the slab, the problem is that the size of this melt pool is quite small in terms of width and depth (given the parameters that I input and my knowledge of expected width and depth values from experimental data) and its peak temperature is INSANELY high.
For instance, in the case of a 55 µm laser spot size scanning at 100 W power and 2 m/s speed, I get peak temperatures as high as ~15000 K where, from common experiments, I should only be able to melt the alloy (T melting ~ 1673 K). We are talking about a discrepancy of 1 order of magnitude here.

Originally, I also experienced a singular issue where I observed local cooling below 293 K (even negative temperatures!) of the plate in front of the laser beam right before heating. I fixed this by refining the mesh (after reading it on the forum and making sure that the element size was smaller than the typical thermal diffusivity distance (x=sqrt(D*t)).
Although the negative temperature problem was fixed by mesh refinement, the peak temperatures remain in this crazy high region.

Given my poor familiarity with COMSOL, at the moment I am out of ideas on what the causes of this issue could be, mostly because it seems to me that my boundary and initial conditions make sense. Anyone who could shed some light on this will have my sincerest gratitude!

Bonus question: in some cases, when performing SLM of metals, the temperature of the molten metal can even reach and surpass the boiling point (~3200 K for steel) with subsequent vaporization (and relative latent heat). How could I model this into COMSOL? In other words, can I set the heat transfer module on my domain to have three distinct phases (solid-->liquid-->vapor) and thus model both the melting latent heat and the vaporizing latent heat?

Thank you in advance,
Umberto