High-Tech Crystal Learns to Waste Less Power

Inside a thin film of strontium iridate, barely 27 nanometres thick, electrons do something that most metals cannot. They remember which way they’re spinning. Not in any mystical sense; the crystal’s geometry forces it, the atoms arranged in a pattern that links each electron’s momentum to its spin as surely as a screw thread links rotation to direction. Pull current through the material and a flood of spin-polarised electrons pours out sideways. The effect is, depending on your familiarity with spintronics, either obvious or baffling. But its consequences are concrete: you can flip a magnetic bit with a fraction of the current that conventional memory devices require.

That matters because data storage runs hot. Every write operation in a standard magnetic memory device involves pushing electrons through a material until enough torque accumulates to flip a magnetic domain, and the threshold currents are punishing. In a world adding data centres by the megawatt, the physics of how you switch a magnet is no longer just an academic problem.

The team at the Ningbo Institute of Materials Technology and Engineering, part of the Chinese Academy of Sciences, approached the problem from an unusual angle. Most research into so-called spin-orbit torque materials has been a kind of organised prospecting: grow a topological semimetal, measure its efficiency, discard what fails, repeat. What the Ningbo group did instead was start with symmetry. Specifically, they looked for crystal structures where the rules of geometry make topological behaviour unavoidable rather than accidental.

The distinction is worth sitting with for a moment. Ordinary topological semimetals develop their exotic band structures through what physicists call accidental band inversion, a particular configuration that happens to produce protected electronic states. Non-symmorphic symmetry is different: it involves fractional translations coupled with rotational operations, and it forces electron bands to cross at specific points in momentum space regardless of other details. The crossing cannot be removed by small perturbations; it is, in a sense, written into the lattice itself. Think of it as the difference between a bridge that happens to be strong and one whose geometry makes collapse structurally impossible.

A Guardian Built Into the Crystal

Hexagonal strontium iridate has that geometry. Grown on a substrate of strontium titanate oriented along the (111) direction, SrIrO3 adopts a hexagonal phase characterised by both corner-shared and face-sharing oxygen octahedra, and the resulting crystal symmetry group contains glide planes and screw axes. Those non-symmorphic elements enforce a four-fold band degeneracy at the Brillouin zone boundary, producing robust 3D Dirac points near the Fermi energy. The researchers confirmed this directly using angle-resolved photoemission spectroscopy performed in situ at the Swiss Light Source, watching the band structure emerge without ever breaking vacuum. Linear dispersion along all three crystallographic directions at the A point made the three-dimensional character of the Dirac cone unambiguous.

What they found alongside those bulk Dirac points was, in some ways, equally important. The surface of the material hosts distinct two-dimensional states whose energy dispersion does not shift as the photon energy changes, a tell-tale sign of genuine surface localisation rather than bulk projection. These surface states are spin-momentum locked, meaning the direction of spin is tied to the direction of travel. That locking produces a second mechanism for generating spin current, the so-called Edelstein effect, on top of the bulk spin Hall contribution from the large Berry curvature around the Dirac points.

The two mechanisms working in concert yield a spin-orbit torque efficiency of roughly 2.26, a dimensionless ratio comparing spin current generated to charge current applied. To put that in context: platinum, the workhorse of spintronic device research for more than a decade, typically manages something between 0.1 and 0.5. Tungsten can reach perhaps 0.3 to 0.4 on a good day. The hexagonal iridate’s figure is not merely an incremental improvement; it sits outside the range that most heavy-metal candidates cluster within. Comparisons against other topological semimetals in the published literature suggest this represents a record, or close to it.

The Limits and What Comes Next

There are caveats, as there usually are. The electrical conductivity of the thin films is modest enough that SrIrO3 may be operating in a “bad metal” regime where the spin Hall conductivity is partially suppressed relative to what a more conductive sample would deliver. Bulk single crystals of hexagonal SrIrO3 show higher conductivity, suggesting that films optimised for electron mean free path could push the efficiency figures higher still. The growth temperatures involved are also incompatible with back-end-of-line silicon processing, the standard manufacturing constraint for integrating new materials with existing chip fabrication. Possible workarounds involve growing films on sacrificial layers that can then be transferred to silicon substrates, an approach that other oxide perovskites have demonstrated.

What makes the work arguably more significant than any single efficiency number is the design logic underlying it. Non-symmorphic symmetry is common among the 230 possible crystal space groups, and it protects a menagerie of exotic fermion types beyond ordinary Dirac points, including Möbius fermions and hourglass fermions, each with distinctive band structures. The researchers’ demonstration that deliberately engineering crystal symmetry can reliably produce large spin-orbit torque performance opens a map of unexplored territory. If hexagonal SrIrO3 is the proof of principle, the search can now proceed through non-symmorphic materials systematically rather than by chance.

The underlying computational problem driving all of this is not going away. AI training runs consume electricity on a scale that would have seemed implausible a decade ago, and the pressure on memory bandwidth and write efficiency only intensifies as model sizes grow. Spintronic memory, in principle, combines non-volatility with endurance and switching speed that charge-based devices struggle to match. The remaining obstacle has always been write current. Whether the particular answer lies in strontium iridate specifically or in some cousin material waiting to be identified, the approach of locking topological protection into the crystal structure rather than hoping it emerges suggests that the energy arithmetic of future memory might look quite different from today’s.

DOI: 10.1093/nsr/nwag077

Frequently Asked Questions

Why does it matter so much how much current you need to flip a magnetic bit?

Write current is the main reason spintronic memory still struggles to compete with flash storage at scale. High currents generate heat, degrade materials over time through electromigration, and make it harder to pack memory cells densely without thermal interference between neighbours. Cutting the switching current density by even a factor of two has compounding benefits across speed, endurance, and chip area, which is why materials that achieve this are taken seriously even at early lab-scale stages.

What actually is spin-orbit torque and how does it switch a magnet?

When current flows through certain materials, electrons with different spins get deflected in opposite directions, a phenomenon called the spin Hall effect. That separation generates a pure spin current perpendicular to the charge flow, and when that spin current is absorbed by an adjacent magnetic layer, it exerts a torque that can rotate the magnetisation. The efficiency of that process, how much spin current you get per unit of charge, is what the strontium iridate result pushes to record-high levels.

Is the topological protection in this material really permanent, or could manufacturing defects break it?

That is precisely what makes the non-symmorphic approach appealing over conventional topological semimetals. In systems based on accidental band inversion, disorder or impurities can gradually suppress the topological features. Here the Dirac points are enforced by the crystal’s symmetry group itself, so as long as the basic lattice structure is intact the protected band crossings cannot be removed by small perturbations. It does not make the material defect-immune, but it makes the topological properties considerably more robust under realistic film-growth conditions.

Could this material actually end up in consumer devices, or is it too exotic to manufacture?

The honest answer is that integration with silicon is a genuine obstacle right now. The growth temperatures needed for high-quality hexagonal SrIrO3 films are incompatible with the back-end-of-line process used to add memory to finished chips. Researchers have suggested routes around this, including growing the films on sacrificial oxide layers and then transferring them to silicon substrates, a technique that has worked for other perovskite materials. Whether the performance advantage is large enough to justify that engineering overhead is a question the next few years of device work will have to answer.

Why focus on crystal symmetry rather than just testing more materials?

Conventional materials search in spintronics has largely been empirical: grow a candidate, measure it, move on. That approach has found useful materials, but it is slow and tends to cluster around families of compounds that researchers already know how to make. Using non-symmorphic symmetry as a design criterion inverts the logic; instead of asking what a material happens to do, you ask what geometry compels certain electronic behaviour, then identify candidates with that geometry. With more than 1600 non-symmorphic space groups among the 230 total, and many exotic fermion types still largely unexplored for spintronics, the researchers argue the approach opens a much larger search space than trial-and-error prospecting.