A current-driven magnetic skyrmion doesn't move straight down a racetrack. Its dynamics follow the Thiele equation, the rigid-texture reduction of the Landau-Lifshitz-Gilbert equation:
G (ẑ × v) + αD v = F(J)
G = 4πQ (gyrocoupling vector, Q = topological charge, ±1)
D (dissipative tensor, from the texture's shape)
α (Gilbert damping)
F(J) (spin-transfer-torque force, ∝ current density J)
Writing v = (vx, vy) and solving that 2×2 linear system directly (this 2D engine re-derives it independently of the 3D sibling, as a numeric cross-check — both agree exactly):
vx = F·αD / (G² + (αD)²)
vy = −F·G / (G² + (αD)²)
⇒ tan(θSkH) = |vy/vx| = G / (αD)
In a single ferromagnetic layer this sideways drift is fixed by Q and α alone — raising the current only makes the skyrmion hit the track edge and annihilate faster, it cannot fix the angle. This is the central obstacle to real skyrmion racetrack memory.
The synthetic antiferromagnetic (SAF) fix: stack two ferromagnetic layers with opposite topological charge (Qtop = +1, Qbottom = −1) coupled through an RKKY interlayer exchange. Because G ∝ Q, the two layers' Magnus forces point in opposite transverse directions, so their average drift is already near zero even before any locking — but without coupling each sub-lattice still drifts to its own edge independently and can still annihilate there. The vector panel below plots both layers' velocity so you can see this: watch vtop and vbot tilt in mirrored directions, and the pale composite arrow show what the bit's centre-of-mass actually does. Raising k pulls ytop and ybot toward each other (a spring-like interlayer term), keeping the pair bound as one rigid composite that rides straight down the middle rather than splitting apart edge-to-edge. This mechanism was proposed by Zhang, Ezawa & Zhou (2016) and is one of the leading real strategies for making skyrmion racetrack memory viable.
- SAF bilayer / Single layer — toggle whether the bottom layer and its coupling exist at all.
- Current density J — sets the spin-transfer-torque driving force; higher J means faster longitudinal motion (and, without coupling, faster drift into the edge).
- Gilbert damping α — larger α shrinks each layer's own Hall angle, but only the SAF coupling removes it entirely.
- Interlayer coupling k — the RKKY exchange strength binding the two layers; at k = 0 they drift apart independently, at high k they move as one straight-line composite skyrmion.
- Track width — narrower tracks give the skyrmion less transverse room before it annihilates at the edge, making the Hall effect's cost visible faster.
- Drag the racetrack panel — pan the camera along the track by hand; it eases back to auto-follow a little over a second after you let go.