AlSi10Mg - Nucleation of Diamond and Mg2Si at high cooling rates

dendritic solidification, eutectics, peritectics,....
Post Reply
jolinho
Posts: 1
Joined: Mon Apr 26, 2021 1:09 pm
anti_bot: 333

AlSi10Mg - Nucleation of Diamond and Mg2Si at high cooling rates

Post by jolinho » Tue Jun 22, 2021 10:52 am

Hi everyone,

I set up a simulation for AlSi10Mg alloy. I want to simulate the dendritic growth of the FCC (in my simulation: fraction 1) and the Diamond (fraction 2) and Mg2Si-phase (fraction 3) with high cooling rates as they occur during laser powder bed fusion. Below I describe my simulation set-up.

My simulation domain is 200x1x200 and I simulate with a grid resolution of 0.01 µm. I use a rectangular grain as starting grain. The used Thermo-Calc-file and the input-file of my simulation are attached.

I started to optimize the numerical parameter to set up a simulation that works with the huge cooling rate and without the nucleation of the Diamond and Mg2Si phase as described in:

https://board.micress.de/viewtopic.php? ... ates#p3197

After that, I tried to nucleate the Diamond phase at the interface and the Mg2Si-phase in the bulk. I was able to nucleate the two phases. I enabled "moving frame" to move my simulation box and speed up my simulation.

The dendrites look realistic at the beginning. The driving force of the dendrite tip is constant over time. The Diamond phase nucleates at the interface with higher simulation time:
AlSi10Mg_N_mg_004b.PNG
AlSi10Mg_N_mg_004b.PNG (567.82 KiB) Viewed 907 times

In the end of the simulation, where I expect complete solidification, there is still liquid present. I am not sure if the temperature profile is correct. I expected dendrites with a eutectic structure in bulk. I think that the simulation is not valid yet regarding the numerical or nucleation parameters:
AlSi10Mg_N_mg_004b_t0.001.PNG
AlSi10Mg_N_mg_004b_t0.001.PNG (757.08 KiB) Viewed 907 times

In the complete output when "out moving frame" is active there is also seen liquid between the dendrites. Is this output a stitching of the single pictures of the moving frame or is the area outside of the moving frame also simulated?
AlSi10Mg_N_mg_004bG_t0.001.PNG
AlSi10Mg_N_mg_004bG_t0.001.PNG (167.86 KiB) Viewed 907 times

I tried to increase my simulation box without a moving frame to simulate the growth of the Diamond- and Mg2Si-phase, but it led to a very long simulation time.

Thanks for your help!

Johannes
Attachments
AlSiMg.GES5
(38.38 KiB) Downloaded 117 times
AlSi10Mg_N_mg_004b_in.txt
(34.1 KiB) Downloaded 137 times

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: AlSi10Mg - Nucleation of Diamond and Mg2Si at high cooling rates

Post by Bernd » Tue Jun 22, 2021 6:20 pm

Hi Johannes,

Welcome to the MICRESS Forum and thank you for showing us your setup for simulation of LPBF of AlSiMg. This is a complex process regarding the extreme thermal conditions and the fast melting and solidification process, and you already did a quite good job to get there.

When setting up a simulation in this field, the first step should be to decide on the length and time scales and on grid resolution. In principle, this data is determined primarily by the laser parameters, and, in consequence, by the thermal parameters which you apply as boundary conditions to MICRESS. Essentially, these data decide whether or not you will be able to define your simulation domain large enough to get full solidification (and thus precipitation of Si, Mg2Si) inside this domain, or even the full layer thickness. Therefore it makes sense to intensely review these data before setting up the simulation, because changing afterwards would make it necessary to also revise all these decisions.

With respect to the thermal parameters (gradient, cooling rate) which you have chosen, it appears that your choice makes it rather difficult for MICRESS: The cooling rate is about 100 times higher than the gradient (in units of cm), leading to an untypically high value of 100 cm/s for the front velocity. I don't know whether this just was your first estimate, or whether these thermal data come from temperature field simulations. In the latter case, please note that it may be very difficult to correctly extract these values, which have been obtained on a much larger simulation grid and which cannot correctly reflect the bending of the temperature field through the effect of latent heat close to the solidification front. E.g., assuming a higher temperature gradient instead would greatly simplify the problem for MICRESS from a numerical point of view...

One of the consequences of the extreme conditions is that (after some time) the dendrites in your simulation get coarser as undercooling increases: They just are not able to grow fast enough at the given grid resolution. The reason for that is that the "mob_corr" functionality prevents numerical trapping artifacts by reducing the effective interface mobility (see here). For too large spacing, this kinetic contribution gets large in comparison to curvature contributions (making dendrites getting coarser), and eventually reduces the growth velocity so much that undercooling increases constantly. You can e.g. check the .driv output to visualize this behaviour.
By the way, "mob_corr" has been optimized for the default averaging parameter of the driving force of 0.5. I would therefore not recommend using 0.95 for keeping the interfaces stable, but instead use a stabilisation term.

In the same way, the problem that the mushy zone does not fit inside your simulation domain is also a consequence of the relatively small temperature gradient you have chosen. Even if I believe that by performance optimization you should be able to increase the height of your domain considerably, this advantage probably will be eaten up by the necessity to increase grid resolution. So, in your current setup, the solution to the problem could only be to split your simulation in two: a first one which creates a stationary dendrite array (consisting of only the top of the mushy zone), and a second one for the rest of the solidification process by using the "restart" functionality and switching off the moving frame. Please note that in case of "moving_frame" the rows of grid cells which move out at the bottom are just collected in the *G.* output files but not further modified.

Finally, I would like to remark that you still are using the old version 6.4 of MICRESS from early 2018. Meanwhile we have improved many aspect which are especially relevant for additive processes in Version 7.0. Even more important, Version 7.1, which is scheduled to be released later this year, represents a "dedicated" Version for Additive Manufacturing as it will provide several improvements with respect to handling thermodynamic data, diffusion data and thin interface correction at high temperature gradients, making it possible to simulate entire additive layers. Furthermore, a new thermal model is included which is based on a 1d-cylinder approximation and which will allow direct linking to laser parameters.

Bernd

Post Reply