How do dipolar surface twists behave in the van der Waals magnet Fe3GaTe2?

A twist confined to the surface, not the bulk

A skyrmion is a vortex-like spin texture. In a magnet of finite thickness it can run through the whole sample as a three-dimensional skyrmion string, so that surface and interior behave as one object. The paper's central object is different: a dipolar surface twist is a spin twist that sits on the surface of a stripe domain or a skyrmion string and does not extend coherently through the thickness.

The authors reach this picture through simulation first. They compare strings at a high magnetic quality factor f = Ku/Kd, where anisotropy energy dominates demagnetization energy, with strings at a low f ≈ 0.23, where demagnetization energy prevails. The low-f case produces spindle-like strings whose spins rotate from out-of-plane in the middle to in-plane near the surfaces, generating an extra Bloch-type core that shrinks toward the surface. That surface feature is the DST.

Because Fe3GaTe2 lacks strong Dzyaloshinskii–Moriya interaction, the twists can take either clockwise or counter-clockwise helicity at degenerate energy. This is why the authors call them dipolar: their formation is governed mainly by dipolar interactions tied to the low magnetic quality factor, and micromagnetic simulations with anisotropic exchange coupling show the twists persist across a wide range of interlayer coupling strengths.

Solid dots distinguish twists from skyrmion rings

The identification rests on Lorentz transmission electron microscopy (L-TEM), which images deflected electrons to reveal magnetic contrast integrated through the sample thickness. Conventional skyrmion strings give a ring-like contrast. Unconventional spindle-like strings instead show an additional solid black dot, and line profiles confirm the different intensity peaks.

The solid black or white dot is the signature the authors assign to a DST, with the two colors corresponding to opposite helicities. This matters because it separates DSTs from trivial magnetic bubbles, which in centrosymmetric systems typically show half-black/half-white contrast from Bloch lines. The DST contrast survives projection through the thickness because the top and bottom surface spin configurations do not simultaneously satisfy lateral alignment and opposite helicity, leaving a finite projected in-plane magnetization.

The experimental match comes from Fe3GaTe2 lamellae of 71, 95, and 123 nm thickness, imaged at 290 K and zero field. High-density dot-like textures about 65 nm across appear alongside stripe domains, and the same dot contrasts appear in mechanically exfoliated flakes, indicating the textures do not depend on focused-ion-beam processing. As thickness increases, stripe contrast becomes more pronounced and evolves into a maze-like pattern at 123 nm.

Fractional charge makes counting the practical measure

Conventional skyrmions carry integer topological charge. The DSTs stabilized above stripe domains instead carry fractional charge that varies with the open boundary, so the authors describe them by N, the number of DSTs, with N = 1 denoting a single fractional topological charge. This is a bookkeeping choice, not a claim that each twist is a conventional skyrmion.

Under field swapping the twists assemble into composite objects. The paper reports aggregates of up to 13 DSTs and ring-like structures, with sizes growing from about 65 nm to over 700 nm as N increases. Clusters and multi-N objects are distinguished in defocused images: clusters show randomly stacked twists of opposite helicity at irregular spacing, whereas multi-N objects show a regular alternating arrangement.

Size comes with fragility. Stability decreases at higher N, and these high-N metastable objects often collapse abruptly under slight field variations, making them sensitive to magnetic and thermal fluctuations. The field-dependent density data and hysteresis curves track which configurations survive over increasing and decreasing field sweeps, so the composite states should be read as field-history-dependent rather than as a single equilibrium ground state.

Current produces scintillation or ordered motion

The dynamics were probed in an electrical microdevice with Fe3GaTe2 lamella, Pt electrodes, and a Si chip, using 40 ns current pulses at 10 Hz. Because DSTs resolve best at 0° tilt while bulk Néel skyrmions need a finite tilt, the two were studied separately: DSTs at 0° and bulk skyrmions at 15°.

At j = 1.37 × 10^11 A/m² and zero field, over a 62 s period, the authors observed nucleation, annihilation, and helicity switching of DSTs, which they call a scintillating mode. The positions of the twists stayed essentially unchanged, indicating local dynamics rather than translation. Raising j to 2.18 × 10^11 A/m², the threshold for driving bulk stripe domains, made DSTs hard to track because translation and scintillation coexisted.

Bulk skyrmion lattices created by field cooling moved in an ordered manner at zero field, with a tracked skyrmion following an almost linear trajectory and a negligible skyrmion Hall effect. Applying 80 mT to remove skyrmion–skyrmion interactions resolved a small Hall angle of about 4°, consistent with near balance between the Gilbert damping α and non-adiabatic coefficient β. The paper summarizes the current-density-dependent velocities of DSTs and bulk skyrmions as two distinct dynamical regimes.

Pinning turns translation into local reconstruction

The authors explain the split with micromagnetic simulations using Fe3GaTe2 parameters, a 1 ns current pulse, zero field, and pinning potentials of Kpinning = 2Ku, 5Ku, 7.5Ku, and 10Ku. Under weak pinning (2Ku), DSTs depin and translate. Under stronger pinning (5Ku and above), translation is suppressed and the spin torque instead drives localized reconstruction, producing helicity reversal through transient spin configurations and additional topological defects.

This is a simulation-based mechanism, and the authors note the simulated current density of 8 × 10^11 A/m² exceeds experimental values because thermal effects were not included. Helicity reversal occurs only when the injected torque exceeds the energy barrier for reconstruction; otherwise the original helicity recovers after the pulse.

The paper argues Joule heating alone cannot explain scintillation. The nearly fixed DST positions argue against heating above Tc, finite-element simulations place sample temperature below Tc during pulses, and temperature-dependent L-TEM shows heating weakens DST contrast without reversing it. The authors therefore attribute the dynamics mainly to spin-transfer torque acting within a local pinning landscape, with thermal effects possibly lowering local barriers. Candidate pinning sources include intrinsic defects such as Fe vacancies and interstitial Ga or Te atoms, plus an amorphous surface layer from focused-ion-beam processing.

Surface and bulk dynamics are decoupled here

The imaging evidence is thickness-integrated, and the authors state that L-TEM limits depth-resolved interpretation. Reconstructed induction maps are used qualitatively to support assignments of projected spin configurations and helicities, while topological charges are calculated from micromagnetic spin configurations rather than measured directly. The surface-versus-bulk distinction therefore rests on the combination of thickness- and tilt-dependent contrast, simulation, and the differing current responses, not on a direct depth-resolved measurement.

Within that scope, the paper's claim is specific: DSTs are surface-localized topological states whose dynamics are decoupled from bulk counterparts in Fe3GaTe2 at room temperature and zero field. That decoupling is what allows one sample to host two current-driven behaviors at once, and it is the reason the twists are proposed as a separate degree of freedom rather than a variant of the bulk skyrmion.

The current-density threshold is the key comparison

For readers weighing this result, the most decision-relevant number is not the maximum aggregate size but the current-density split. At 1.37 × 10^11 A/m² the DSTs scintillate in place; at 2.18 × 10^11 A/m², the threshold for bulk stripe motion, the two behaviors coexist and DST tracking becomes difficult. The practical distinction between surface-localized and bulk dynamics therefore depends on staying below the bulk-driving threshold.

The second comparison is stability versus N. Single fractional-charge twists and small composites persist at room temperature and zero field, while high-N objects up to 13 twists are metastable and collapse under slight field changes. Any use of the composite states would have to specify the field-sweeping history that produced them, because the paper's field-swapping protocol, not a field-free equilibrium, generates the diversity.

Sources used in this answer