Experimental x Simulation
30 views
Skip to first unread message
Angelo Junior
Aug 23, 2025, 12:40:38 PMAug 23
to mumax2
Hello Mumax Community,
I have a question about replicating experimental measurements in simulations to compare simulation vs. experimental results. For example, I read a paper comparing experimental results for Co/Pt. They used a multilayer structure of Co (0.6 nm) / Pt (2 nm) × 6 disk. In the paper, they also replicated the same parameters in the simulations, but they did not make the exact simulation parameters explicit.
How should I replicate this multilayer in the simulation? Should I set the Co thickness to 0.6 nm, or should I consider the effective thickness of the whole multilayer, which would be 15.6 nm?
Could someone explain in detail how researchers usually handle this?
Josh Lauzier
Aug 24, 2025, 4:58:51 PMAug 24
to mumax2
Hi,
It depends on what exactly you're trying to simulate. Using an effective medium approximation will work for a lot of cases, as long as you know what you're doing. In general, I would say it is safer to explicitly model the multilayer when possible. That way, you don't have to worry about whether you're scaling things properly, or if the EMA will miss certain details that get lost in the approximation. But it is not always possible, or if possible it will be very slow.
EMA can be very good, especially when the simulation would be too large to realistically simulate otherwise. There is a mumax tutorial paper with a very nice example involving a large multilayer stack, which you can find here. It walks through the process in detail. Note however, the simplifying assumptions they make (in that case, that the magnetization doesn't vary much in the z-direction). If the magnetization changes were very rich and varied a lot in all directions on small scales, those assumptions wouldn't work.
EMA might be a good idea in your case. For those thicknesses, there is no nice cellsize you can choose (you need to pick a common divisor. The closest one would be 0.2 nm). You would need 78 layers, and even then 0.2 nm is below where micromagnetics is strictly valid, you're well into atomistic sizes at that point.
Best,
Josh L.
It depends on what exactly you're trying to simulate. Using an effective medium approximation will work for a lot of cases, as long as you know what you're doing. In general, I would say it is safer to explicitly model the multilayer when possible. That way, you don't have to worry about whether you're scaling things properly, or if the EMA will miss certain details that get lost in the approximation. But it is not always possible, or if possible it will be very slow.
EMA can be very good, especially when the simulation would be too large to realistically simulate otherwise. There is a mumax tutorial paper with a very nice example involving a large multilayer stack, which you can find here. It walks through the process in detail. Note however, the simplifying assumptions they make (in that case, that the magnetization doesn't vary much in the z-direction). If the magnetization changes were very rich and varied a lot in all directions on small scales, those assumptions wouldn't work.
EMA might be a good idea in your case. For those thicknesses, there is no nice cellsize you can choose (you need to pick a common divisor. The closest one would be 0.2 nm). You would need 78 layers, and even then 0.2 nm is below where micromagnetics is strictly valid, you're well into atomistic sizes at that point.
Best,
Josh L.
Reply all
Reply to author
Forward
0 new messages