Add native RKKY interlayer coupling (ext_RKKY)#403
Open
SanthoshSivasubramani wants to merge 4 commits into
Open
Add native RKKY interlayer coupling (ext_RKKY)#403SanthoshSivasubramani wants to merge 4 commits into
SanthoshSivasubramani wants to merge 4 commits into
Conversation
Adds a bilinear RKKY interlayer coupling field term exposed to the scripting API as ext_RKKY(region1, region2, J), with J the areal coupling strength in J/m2 (J<0 antiferromagnetic). The effective field is B = J/(Msat*dz)*m_partner. The partner cell is located by scanning the z column, so the coupling spans a nonmagnetic spacer gap between layers, which ext_InterExchange (nearest neighbour only) cannot do. For adjacent layers the torque is identical to ext_InterExchange with A_inter = J*dz/2. Files: cuda/rkky.cu (kernel), cuda/rkky.go (host wrapper), cuda/rkky_wrapper.go (cuda2go generated), engine/rkky.go (ext_RKKY field term), test/rkky_native.mx3 and test/rkky_gap.mx3 (regression tests).
There was a problem hiding this comment.
Pull request overview
Adds a native RKKY (interlayer exchange) coupling term to the simulation effective field, exposed to mx3 scripts as ext_RKKY(region1, region2, J), including CUDA implementation and regression tests (adjacent layers + spacer-gap coupling).
Changes:
- Introduces a new CUDA kernel (
addrkky) plus Go host wrapper to add the RKKY effective field contribution. - Registers a new engine field term (
B_rkky) and the script functionext_RKKY(...). - Adds two mx3 regression tests validating adjacent-layer and spacer-gap coupling behavior.
Reviewed changes
Copilot reviewed 5 out of 6 changed files in this pull request and generated 5 comments.
Show a summary per file
| File | Description |
|---|---|
| engine/rkky.go | Engine-side ext_RKKY registration and B_rkky field-term wiring. |
| cuda/rkky.cu | CUDA kernel that finds partner cells along z and accumulates the RKKY field. |
| cuda/rkky.go | Host-side kernel launch wrapper and argument setup. |
| cuda/rkky_wrapper.go | Auto-generated CUDA2GO wrapper + embedded PTX for addrkky. |
| test/rkky_native.mx3 | Regression test for adjacent-layer coupling field magnitude/direction. |
| test/rkky_gap.mx3 | Regression test for coupling across a nonmagnetic spacer gap. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+39
to
+45
| func RKKY(region1, region2 int, J float64) { | ||
| rkkyPairs = append(rkkyPairs, rkkyPair{region1, region2, J}) | ||
| if !rkkyRegistered { | ||
| AddFieldTerm(B_rkky) | ||
| rkkyRegistered = true | ||
| } | ||
| } |
Comment on lines
+51
to
+69
| // locate the nearest partner-region cell in the same (x,y) column along z | ||
| int P = -1; | ||
| int best = Nz + 1; | ||
| for (int jz = 0; jz < Nz; jz++) { | ||
| if (jz == iz) { | ||
| continue; | ||
| } | ||
| int Q = idx(ix, iy, jz); | ||
| if (regions[Q] == partner) { | ||
| int d = (jz > iz) ? (jz - iz) : (iz - jz); | ||
| if (d < best) { | ||
| best = d; | ||
| P = Q; | ||
| } | ||
| } | ||
| } | ||
| if (P < 0) { | ||
| return; | ||
| } |
Comment on lines
+4
to
+5
| // Layer 0 (region 100) is +z, layer 1 (region 101) is +x, so the field in | ||
| // region 100 points along +x scaled by J/(Msat*dz). |
| m.setRegion(100, uniform(0, 0, 1)) | ||
| m.setRegion(101, uniform(1, 0, 0)) | ||
| Run(0) | ||
| ExpectV("B_rkky adjacent", B_rkky.Region(100).Average(), vector(-2.22222, 0, 0), 1e-3) |
| m.setRegion(100, uniform(0, 0, 1)) | ||
| m.setRegion(101, uniform(1, 0, 0)) | ||
| Run(0) | ||
| ExpectV("B_rkky across spacer", B_rkky.Region(100).Average(), vector(-2.22222, 0, 0), 1e-3) |
cuda/rkky.cu: locate the partner by scanning outward from iz and stopping at the first hit (O(distance) instead of O(Nz)); the nearest-partner result is unchanged. test/rkky_native.mx3, test/rkky_gap.mx3: also assert the field on region 101 (the other side of the pair), and correct a comment about the field sign for J<0.
This was referenced Jul 3, 2026
Register the RKKY field via SetEffectiveField as a built-in effective-field term instead of AddFieldTerm, so it does not pollute the user-defined B_custom and is not removed by RemoveCustomFields().
Addresses Copilot review: (1) apply the areal coupling only at the interface cell so it is independent of layer thickness (was applied to every cell, over-counting thick layers); (2) validate region ids via defRegionId; (3) overwrite an existing pair instead of appending a duplicate; (4) register the RKKY energy (E_rkky, Edens_rkky) so E_total includes it. Added test/rkky_thick.mx3 (2-cell layers: field halves, energy thickness-independent). VM-validated on L4: all tests OK.
Author
|
Thanks for the review — pushed
On the O(Nz) z-column scan: this is kept intentionally — scanning the column is what lets the coupling span a non-magnetic spacer gap, and Nz is small for layered SAF stacks, so the cost is negligible. Validated on an NVIDIA L4 (CUDA 12.3): all tests pass. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a native RKKY (interlayer exchange) coupling term, exposed to the input
script as
ext_RKKY(region1, region2, J), whereJis the areal couplingstrength in J/m2 (
J < 0for antiferromagnetic / synthetic-antiferromagnetcoupling).
Related discussion: #402.
Motivation
mumax3 has no native RKKY term today. The established approach is to repurpose
the Heisenberg exchange between two regions, as in the existing
test/rkky.mx3:This has two limitations:
coupling (J/m2) to a dimensionless exchange scale factor.
Msat=0neighbour is treated as self), so it cannot couple two magneticlayers separated by a nonmagnetic spacer, which is the usual SAF geometry.
ext_RKKYtakes the physicalJdirectly, and locates the partner cell byscanning the z column, so the coupling spans a spacer gap.
Implementation
The term is added to the effective field as
consistent with the exchange field convention. For adjacent layers this is
torque identical to
ext_InterExchangewithA_inter = J * dz / 2(the twodiffer only by a component parallel to m, which exerts no torque).
Files:
cuda/rkky.cu: kerneladdrkkycuda/rkky.go: host wrapperAddRKKYcuda/rkky_wrapper.go: cuda2go generated wrapperengine/rkky.go:ext_RKKYfield termtest/rkky_native.mx3,test/rkky_gap.mx3: regression testsValidation (NVIDIA L4, CUDA 12.3)
J/(Msat*dz)to 6 significant figures.Bproportional tom_partner).test/rkky_gap.mx3), whereext_InterExchangegives zero.ext_scaleExchange/ext_InterExchangefor adjacent layers.Both regression tests pass.
Notes
This registers the effective field term, so dynamics and field driven
relaxare correct. It does not yet register an energy density, so
E_totaldoes notinclude the RKKY contribution; that can be added as a follow up if preferred.