Demag Kernel incorrect when using multiple cells in the Z direction in thin films

[marcrovi]

Hello,

I've been trying to simulate multilayered systems and therefore I am setting the geometry and setting regions to different layers and I can only do that by putting multiple cells in the Z direction. The system I am using is the following

---

CellSize:=5.e-9
NumCells:=512
ZNumCells:= 6
SetGridSize(NumCells, NumCells, ZNumCells)
SetCellSize(CellSize, CellSize, 6.e-9)
SETPBC(6,6,0)

//Material
Msat=1000e3
Aex=1e-11
alpha = 0.007
anisC1=vector(1,1,0)
anisC2=vector(-1,1,0)
Kc1=5000
Kc2=0

---

In this system the demag kernel is:
B_demag = (-5.375485731444011e-05, 7.689423121822377e-05, -0.06891722977161407)

For this material and since its a thin film (36nm), I expect to have a demag in plane and I have an easy axis in the x-direction.
However, the demag values are not correct for an extended thin film (it is mostly out of plane...)
If I use only one cell in the Z direction of 36 nm:

---

CellSize:=5.e-9
NumCells:=512
ZNumCells:= 1
SetGridSize(NumCells, NumCells, ZNumCells)
SetCellSize(CellSize, CellSize, 36.e-9)
SETPBC(6,6,0)
setgeom(universe())

---

The B_demag = (0.0011605947511270642, 2.4958247274442513e-11, 1.287635509328311e-08)

I would have expected both to be the same, since its exactly the same system just one has 1 cell of 36nm, while the other system is 6 cells of 6nm each. I see that using only 2 cells also works, but for 3+ it fails and gives an out of plane component.

The rest of the code is completely equal in each case:

---

m =uniform(1,0,0)
relax()
FixDt = 1e-11;
// SOME FMR CODE HERE

---

Is this a problem on the mumax3 kernel calculation? I have seen other problems with it when changing the initial state of m.

Also, the simulations I want to do is to see the dipolar coupling between layers or different materials and therefore if the demag is not correct I wont get good results. :(

Thank you for any response or fix.

Marc

[JonathanMaes]

When running the code on mumax3.10, I get

• ZNumCells = 6 → B_demag = (-0.0002259404088060061, 3.312126686920898e-13, 1.5901631863120731e-18)
• ZNumCells = 1 → B_demag = ( 0.00105499557685107, -1.3783979738946073e-05, 0)

which is in-plane as expected, for both cases.

I used this input script derived from your message:

CellSize := 5.e-9
NumCells := 512
ZNumCells := 6 // Or 1
SetGridSize(NumCells, NumCells, ZNumCells)
SetCellSize(CellSize, CellSize, 6.e-9) // Or 36.e-9
SETPBC(6, 6, 0)

//Material
Msat = 1000e3
Aex = 1e-11
alpha = 0.007
anisC1 = vector(1, 1, 0)
anisC2 = vector(-1, 1, 0)
Kc1 = 5000
Kc2 = 0

m = uniform(1, 0, 0)
relax()
FixDt = 1e-11
print(B_demag)

Could you share the exact, complete input script you used which gives you the unexpected result?