A research team led by Liang Liu at Shanghai Jiao Tong University has demonstrated that ordinary platinum — the heavy metal already used in nearly every spintronic research stack on the planet — can be converted into a high-performance quantum spin filter simply by changing the orientation of its crystal surface, bypassing a decade-long symmetry constraint that had seemed to demand exotic or difficult-to-manufacture materials. The results, published Thursday in Physical Review X, set a record for out-of-plane spin Hall conductivity and delivered the deterministic, field-free switching of perpendicular magnetic memory bits that the spintronic memory industry has been working toward since the mid-2010s.
For engineers designing spin-orbit torque magnetic random-access memory (SOT-MRAM) — the leading candidate to replace DRAM in high-speed, high-endurance applications — the finding matters because it removes what had been one of the most cited materials obstacles to commercialization: the need for an external magnetic field to write a memory bit, or the need to replace platinum with a harder-to-manufacture exotic compound to get around that requirement. An overview of the SOT-MRAM industry’s development trajectory shows that this field-free problem has defined the commercial roadmap for years.
What Platinum Could and Couldn’t Do
The spin Hall effect has been central to spintronics for more than two decades. When electric current flows through a heavy metal such as platinum, the strong spin-orbit coupling in the material deflects electrons with opposite spins in opposite directions, generating a transverse flow of spin — a spin current — that can exert a torque on an adjacent magnetic layer and flip its magnetization. Doing that electrically, without a current flowing through the delicate magnetic tunnel junction at the heart of the memory cell, is the key advantage of SOT-MRAM over its predecessor: it separates read and write paths, enabling sub-nanosecond switching and near-unlimited write endurance. The challenge of field-free perpendicular switching has been the field’s central engineering problem throughout this period.
The catch is geometry. The memory cells with the highest density and best thermal stability store bits as magnetization pointing perpendicular to the chip surface — up or down. But the spin current that conventional platinum generates under the spin Hall effect is polarized in-plane, parallel to the chip surface. A spin current polarized in-plane cannot deterministically flip a perpendicularly magnetized bit without help — specifically, without an external magnetic field that breaks the symmetry the switching requires. That external field blocks standalone memory cells from being built with standard platinum.
Researchers have spent years working around this constraint. One route used intrinsically low-symmetry materials: tungsten diselenide (WTe₂), the Weyl semimetal TaIrTe₄, and the noncollinear antiferromagnet Mn₃Sn each generates out-of-plane spin polarization because their crystal structures break the relevant symmetry at the bulk level. Another route, explored by Liang Liu’s own group in 2021, demonstrated field-free switching using WTe₂ as the spin source. Each of those materials, however, came with manufacturing headaches: instability in air, incompatibility with standard physical vapor deposition equipment, or simply insufficient spin conductivity at room temperature.
Reorienting Platinum’s Surface
The Shanghai team’s solution was to change not what platinum is made of, but how it is grown. By depositing platinum films in so-called (n10) orientations — with n equal to 2, 3, or 4, corresponding to the (210), (310), and (410) crystal faces — the researchers created surfaces that are atomically asymmetric. The standard (001) platinum surface has a mirror symmetry that locks the spin Hall effect to in-plane polarization. The (n10) surfaces lack that mirror symmetry, and the breaking of it is what makes the new physics possible.
At the asymmetric surface, a local electric field — described in the paper as a Rashba-Edelstein effective field — points perpendicular to the chip plane. As conduction electrons flow through the platinum and approach this surface, the surface acts as a quantum gate: it preferentially transmits electrons whose spin is aligned with the local perpendicular field while reflecting or blocking the others. The bulk platinum generates spin current in many polarization directions simultaneously; the surface filters out everything except the out-of-plane component, dramatically amplifying that component’s share of what actually gets injected into the adjacent magnetic layer.
The mechanism — which the team calls noncollinear spin-orbit filtering — works by the same principle as a polarizing lens in photography. An unpolarized light beam contains photons oscillating in every plane; the filter passes only those vibrating in one direction. Here, the filter selects for out-of-plane spin rather than for optical polarization, and the filtering is governed by the quantum mechanics of spin-orbit coupling at an asymmetric interface rather than by classical optics. A companion theoretical preprint from the same group provides first-principles modeling of spin filtering covering how momentum-resolved spin Berry-curvature texture in the bulk gets selectively transmitted at the interface when the interfacial spin-orbit coupling has a finite out-of-plane component.
Record Numbers and What They Mean
The (210)-oriented platinum films produced an out-of-plane spin Hall conductivity of 0.75 × 10⁵ (ħ/2e) Ω⁻¹ m⁻¹ at room temperature. The paper states this surpasses all previously reported values from approaches specifically developed to provide out-of-plane spin polarization — including the exotic-material routes. To put the comparison in terms of the field: prior orientation-dependent work on platinum, such as a 2022 UC San Diego study that grew films in (200), (220), and (111) orientations, documented a 54% enhancement in charge-to-spin conversion efficiency compared to polycrystalline films, but only for in-plane spin. The Shanghai team has now extended this orientation-engineering principle specifically to the out-of-plane direction, which is where the field-free perpendicular switching problem lives.
The device tests went as far as demonstrating that the oriented platinum films reliably switched adjacent cobalt magnetic layers at room temperature without any external magnet, and that the switching was deterministic — always in the intended direction, not random. Crucially, the paper shows that changing the surface crystal point group symmetry from C₁ᵥ (which enables the out-of-plane component) to C₄ᵥ (the conventional, in-plane-only geometry) turns the effect on and off. That tunability matters for engineering: it means the out-of-plane capability is not an accident of the specific sample but a direct consequence of a fabrication choice that can be deliberately dialed in or out.
Why Platinum, Why Now
Intel and Samsung have both announced active SOT-MRAM development programs, and the field treats the external-field requirement as the principal remaining barrier to standalone memory cells. The exotic-materials route has been pursued for several years as a way to generate out-of-plane spin current without a field, but it brings supply-chain friction: platinum is already deployed in spintronic research stacks worldwide, is well-characterized, is stable in air, and is routinely deposited in research and industrial fabs. The surface-engineering trick demonstrated here requires only that the deposition be done onto an appropriately oriented substrate — something achievable with the epitaxial growth tools already in use.
A competing approach published in July 2025 achieved field-free switching using a platinum-titanium alloy (Pt₇₅Ti₂₅), with the alloying providing the electric asymmetry needed to generate out-of-plane torque. The Liu team’s approach is distinct: rather than altering platinum’s composition, it alters its surface orientation. Both are CMOS-compatible; the question of which integrates more cleanly into back-end-of-line processes is a question for the next generation of device-level demonstrations.
A General Principle for Heavy Metal Engineering
The paper’s final implication extends beyond platinum. The authors describe noncollinear spin-orbit filtering as a framework applicable, in principle, to any conventional high-symmetry heavy metal — tungsten, tantalum, gold — whose surface can be terminated in an appropriately asymmetric orientation. If that generalization holds up under independent testing, the field of spin-source engineering may shift focus from the search for new exotic bulk materials toward the deliberate control of surface terminations in familiar metals. That would represent a meaningful change in how the problem is framed: instead of asking “what exotic material has the right bulk symmetry?”, researchers would ask “what surface orientation unlocks the right interface physics in a metal we already know how to use?”
The paper was received by Physical Review X in November 2025, revised in May 2026, and accepted in late June, indicating that peer review was thorough and that the experimental claims survived scrutiny. The affiliated institutions span Shanghai Jiao Tong University, the Chinese Academy of Sciences’ Institute of Solid State Physics, Shanxi Normal University, Tongji University, the National University of Singapore, ShanghaiTech University, and Hefei National Laboratory. The full text is available under a CC BY 4.0 open-access license.
Independent reproduction of the record spin Hall conductivity figure, and full demonstrations in complete SOT-MRAM device stacks rather than test structures, remain the natural next milestones. The crystal surface, it turns out, had been hiding a spin filter the whole time.
Frequently Asked Questions
What is SOT-MRAM and why does the external-field requirement matter?
Spin-orbit torque magnetic random-access memory (SOT-MRAM) is a next-generation non-volatile memory technology that writes bits by injecting a spin current from an adjacent heavy-metal layer into a magnetic tunnel junction, flipping the magnetic bit electrically without running the write current through the junction itself. This separation of read and write paths allows sub-nanosecond switching speeds and dramatically improved write endurance compared to both flash memory and its predecessor, spin-transfer torque MRAM (STT-MRAM). The external-field problem arises because the most scalable, high-density memory bits use perpendicular magnetic anisotropy — their magnetization points up or down, perpendicular to the chip surface — but conventional spin currents from platinum are polarized in-plane. To flip a perpendicular bit with an in-plane spin current, a symmetry-breaking external magnetic field has historically been required. A memory chip that needs an applied magnetic field to write a bit cannot be miniaturized for practical use, which is why field-free switching has been the field’s central engineering goal for nearly a decade.
How exactly does the platinum surface act as a quantum spin filter?
In conventional platinum, the bulk spin Hall effect generates spin currents in several polarization directions simultaneously. Under the standard (001) crystal orientation, the mirror symmetry of the surface averages out any net out-of-plane component, leaving only in-plane spin current reaching the adjacent magnetic layer. In the (n10)-oriented films, the surface lacks that mirror symmetry. A local Rashba-Edelstein effective field — generated by the broken inversion symmetry at the asymmetric interface — points perpendicular to the film plane. As electrons arrive at this surface, their transmission probability depends on whether their spin aligns with that perpendicular field: aligned electrons pass through preferentially, others are reflected. The result is that the spin current delivered to the magnetic layer is enriched in out-of-plane polarization, even though the bulk platinum itself was not generating more of it. The surface selectively passes what it needs and discards the rest — the definition of a filter.
Could this surface-orientation technique be applied to metals other than platinum?
The paper argues that noncollinear spin-orbit filtering is a general framework: the mechanism depends on the interface having a finite out-of-plane Rashba-Edelstein component, not on any property unique to platinum. The same approach should in principle apply to any heavy metal with strong spin-orbit coupling — tungsten, tantalum, gold — if its surface can be grown in a similarly asymmetric low-symmetry orientation. Whether the effect is equally large in other metals depends on each material’s band structure and spin Berry-curvature texture. Testing that generalization across the standard toolkit of spintronic heavy metals is a natural direction for follow-up work.
When could this reach commercial memory chips?
The demonstration is at the thin-film and test-structure stage; the next milestones are independent reproduction of the record spin Hall conductivity figure by other research groups, and demonstration of complete SOT-MRAM device stacks — full magnetic tunnel junction structures with platinum write layers in the new orientation — at switching energies compatible with the power budgets of commercial memory. After that, integration with standard back-end-of-line CMOS processes and wafer-scale uniformity of the epitaxially oriented films would need to be validated. Given that the exotic-materials route and the competing platinum-titanium alloy approach are at similar stages, a realistic commercial timeline for any field-free SOT-MRAM technology likely runs on the scale of several years, contingent on these integration milestones.