Magnetic domain walls, the narrow boundaries that separate tiny regions of opposing magnetization inside a material, were supposed to be racing across thin films at tens of thousands of kilometers per second during ultrafast laser pulses. A new study published in Nature Materials in April 2026 says they are doing nothing of the sort. Using extreme ultraviolet microscopy capable of resolving features down to about 13.5 nanometers in under 40 femtoseconds, an international team directly watched domain walls in two types of magnetic thin films and found them effectively frozen in place, even as the magnetization around them collapsed and recovered at blistering speed.
That result matters beyond the lab. Magnetic random-access memory, or MRAM, stores data in nanoscale magnetic regions. If flipping those regions at femtosecond speeds also sent domain walls careening unpredictably, engineers would face a serious obstacle: stored bits could be scrambled every time data was written. The new finding suggests that obstacle may not exist, at least in the materials tested so far.
What the experiment showed
The research team, which included theorist Johan Mentink of Radboud University, studied TbCo and Co/Pd thin films. Both materials exhibit perpendicular magnetic anisotropy, meaning their magnetization points straight up or straight down through the film, forming labyrinth-like patterns of alternating domains separated by sharp walls. That geometry made the walls easy to track frame by frame.
By hitting the films with an intense femtosecond laser pulse and then probing them with coherent extreme ultraviolet light generated through high-harmonic generation, the researchers captured real-space snapshots of the magnetic texture as it evolved. Within the experiment’s spatial uncertainty of a few nanometers, the domain walls did not budge. The magnetization inside the domains, by contrast, dropped dramatically and partially recovered on sub-picosecond timescales.
A companion preprint on arXiv, associated with the same Nature Materials article, lays out the optical setup in detail: how the probe beam, multilayer mirrors, and detection optics are synchronized with the pump pulses to freeze each magnetic snapshot. That open-access document gives other groups a roadmap for replication.
Why earlier studies told a different story
The stability result directly contradicts a line of research that had reported extreme domain-wall speeds. A 2023 study published in Physical Review Letters, available as a preprint on arXiv (2303.16131), reported speeds of roughly 66 kilometers per second. The arXiv link points to the preprint version of that paper, not the final journal publication. Those figures were not measured by watching walls move. Instead, they were inferred from changes in scattering patterns, reciprocal-space data that required model assumptions to convert into velocities.
The new direct-imaging work effectively stress-tests those assumptions. The Nature Materials authors argue that what looked like wall motion in reciprocal space likely reflected something else: transient disordering of the magnetization, local amplitude suppression, or rapid changes in domain roughness, all of which would alter scattering patterns without requiring the walls themselves to translate across the film.
Earlier experiments had already flagged the difficulty. A 2021 Nature Communications paper noted limits in resolving domain-wall broadening with high-harmonic-generation microscopy and documented irreversible changes to domain structure at the laser fluences needed for demagnetization. A foundational 2012 Nature Communications study showed that ultrafast optical pulses can manipulate spin structures inside domain walls but stopped short of claiming the walls physically moved at extreme speeds. The new imaging data fits more comfortably with those cautious earlier findings than with the 66 km/s claim.
As of May 2026, the authors of the 2023 Physical Review Letters study have not publicly revised or responded to the new results. The disagreement may ultimately reflect differences in material systems, laser fluence regimes, or measurement geometry rather than a simple right-or-wrong split. Reciprocal-space probes are sensitive to any rapid change in the spatial Fourier components of magnetization, and distinguishing coherent wall translation from other transient effects remains genuinely hard.
The physics behind the stillness
Mentink, in a Radboud University news release accompanying the paper, offered a physical explanation. Local spins can reorient almost instantaneously under the laser-induced effective fields, but dragging an entire domain wall through a disordered, defect-rich crystal lattice is a collective process limited by damping and pinning. In plain terms: flipping a single compass needle is fast, but pushing a whole fence of them sideways through mud is slow. The femtosecond laser pulse simply does not last long enough to overcome those barriers.
That interpretation is consistent with the imaging data, though it remains a theoretical framing rather than a direct measurement of the speed limit for wall motion. Pinning down that limit experimentally would require pushing the laser fluence higher or using materials with weaker pinning, neither of which has been reported yet.
What this means for magnetic memory
MRAM is already a commercial technology, used in embedded applications where its combination of speed, endurance, and non-volatility outperforms flash memory. Current MRAM cells switch magnetization on nanosecond timescales using spin-transfer torque or spin-orbit torque. Pushing write speeds toward the femtosecond regime, roughly a million times faster, has been a long-term goal but one shadowed by the worry that ultrafast switching might destabilize the domain structure holding each bit.
If domain walls genuinely stay fixed during femtosecond-scale magnetization reversal, device designers gain a significant degree of freedom. They could use shorter, more intense pulses to flip bits locally without engineering elaborate safeguards against wall creep. That could simplify cell architecture and reduce energy per write operation, since a shorter pulse deposits less total heat into the device.
But the leap from laboratory thin films to manufacturable memory arrays is long. The Nature Materials study examined continuous, unpatterned films of TbCo and Co/Pd. Real MRAM cells are patterned nanostructures with edges, constrictions, and engineered pinning sites that could either enhance or suppress wall mobility. Different alloy compositions, interface qualities, and device geometries, such as synthetic antiferromagnets or in-plane anisotropy stacks, might respond differently to the same laser excitation. No data yet show how wall stability behaves under repeated cycling, thermal stress, or simultaneous electric currents, all routine conditions in working memory chips.
The university press release frames the result as a step toward faster and smaller data storage, but that connection remains qualitative. Specific benchmarks, such as the minimum domain-wall stability required for a given bit density or read-write cycle time, have not been published alongside the imaging work.
Open questions for ultrafast spintronic device design
Follow-up experiments will likely test whether the same stability holds at higher laser fluences, in patterned nanostructures, and under realistic thermal loads. Hybrid approaches that combine extreme ultraviolet imaging with time-resolved scattering or electrical transport measurements could help reconcile the real-space and reciprocal-space views of domain dynamics, clarifying whether any residual ultrafast motion lurks below current imaging sensitivity.
There is also a practical bottleneck in the measurement technique itself. Generating coherent high-harmonic radiation, maintaining nanometer-precision alignment, and synchronizing pump-probe timing are tasks currently confined to specialized laboratories. Whether the approach can migrate to more compact, industry-compatible platforms is an open engineering question. Without that step, the technique remains a powerful research diagnostic rather than a routine tool for device production lines.
For now, the most defensible conclusion is specific and narrow: in TbCo and Co/Pd thin films under the conditions tested, domain walls are far less mobile on femtosecond timescales than earlier indirect measurements implied. That immobility, if it holds up across a wider range of materials and device-realistic conditions, could remove one of the key uncertainties standing between ultrafast magnetism research and the next generation of spintronic memory.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
