The memory problem no one talks about
Every computer you own has a quiet bargain running underneath everything you do. Processors are fast but forgetful: turn off the power and their memory vanishes. Flash storage remembers everything without power, but it is slow. Between those two extremes sits magnetic memory — the kind that holds data without power and can still be read and written quickly.
Today's magnetic memory, known as MRAM, is built on ferromagnets: materials whose tiny atomic magnets all point in the same direction. That alignment is exactly what makes ferromagnets useful for storing a 1 or a 0 — and exactly what causes trouble. Every aligned magnet leaks a magnetic field into its surroundings. Pack the memory cells too close together and they start interfering with each other. Hold a strong magnet nearby and you can scramble the data. Magnetism keeps surprising scientists at every scale — NASA's Parker Solar Probe recently caught a hidden magnetic flip in the solar wind, a reminder of how much this force still shapes physics. For decades, engineers have worked around these limits because nothing better existed.
Physicists, meanwhile, have been eyeing a different class of materials: antiferromagnets. These materials have magnetic structure but almost no external magnetic field — they do not leak, they cannot be disturbed by stray fields, and in theory they can switch states thousands of times faster than ferromagnets. The catch? Because they barely interact with the outside world, writing information into them with an electric current seemed nearly impossible. You cannot easily grab what you cannot see.
On October 21, 2026, a team led by Professor Pedram Khalili at Northwestern University announced that this catch may finally be gone. In a paper published in Nature, the researchers describe a new physical effect — octupole-driven spin-transfer torque, or OTT — that lets an ordinary electric current both write and read information in antiferromagnetic materials, at room temperature, in nanoscale devices.
What antiferromagnets actually are
To understand the discovery, it helps to picture what is happening inside a magnetic material at the atomic scale. Every atom behaves like a tiny compass needle — physicists call this its "spin." In a ferromagnet, like the iron in an ordinary fridge magnet, all of those needles point in the same direction. Their effects add up, producing the strong magnetic field you can feel.
Antiferromagnets arrange their atomic needles so that they cancel each other out. In the simplest version, neighboring needles point in exactly opposite directions, like a chessboard of up and down arrows. The Northwestern team's work focuses on a more exotic arrangement found in a material called platinum-manganese (PtMn₃): here the spins cancel out in groups of three, locked at angles of about 120 degrees to each other — a triangular, "noncollinear" pattern. The result: the material is magnetic on the inside but nearly invisible magnetically on the outside.
That invisibility is a superpower. An antiferromagnetic memory cell would not disturb its neighbors, so cells could be packed far more densely than today's MRAM allows. It would shrug off external magnetic fields that would corrupt a ferromagnet's data. And the physics of antiferromagnets suggests they can be switched thousands of times faster — potentially at terahertz speeds, the regime where future communications technology wants to operate. The challenge was never the material's potential. It was control: how do you write a bit into something that refuses to respond to magnetic fields at all?
The discovery: octupole-driven spin-transfer torque
The answer, according to Khalili's team — including postdoctoral scholars Jaimin Kang and Mohammad Hamdi — is a current that works on a subtler property of the material. In ordinary MRAM, writing works through spin-transfer torque: electrons whose spins are aligned in one direction slam into a magnetic layer and push its magnetization around, like water pushing a paddle wheel. But that technique fundamentally relies on a net magnetization to push against. Antiferromagnets have almost none.
OTT flips the logic. The noncollinear arrangement of spins in PtMn₃ gives rise to a more complex ordering property called a magnetic octupole — think of it less like a compass needle and more like a handedness, or chirality, of the triangular spin pattern. The researchers found that an electric current flowing through the material naturally splits into unequal spin currents between the material's sublattices. Even though the current as a whole carries no net spin, those unequal internal components can push on the octupole order and flip it from one stable state to another.
In other words, the current does not need a net magnetization to grab onto. It grabs the octupole instead. The team calls this "octupole-driven spin-transfer torque," and the accompanying theory shows it is a reciprocal of the ordinary spin-transfer torque physicists already understand — a genuinely new physical mechanism, not just a tweak of the old one. Because the octupole states are stable, they can represent bits: one chirality is a 0, the flipped chirality is a 1. It caps a remarkable month for quantum magnetism: physicists have also finally observed Bethe strings, a 95-year-old prediction about how spins behave in magnetic matter.
The device: a tunnel junction with no ferromagnets at all
Discovering an effect is one thing; building a memory device around it is another. The Northwestern team did both. They constructed nanoscale devices from two layers of the antiferromagnetic material separated by a thin insulating barrier of magnesium oxide (MgO) — a structure they call an all-antiferromagnetic tunnel junction, or AATJ, written as PtMn₃|MgO|PtMn₃.
Here is how it works. To write a bit, a pulse of electric current passes through the device; OTT flips the octupole order of one of the antiferromagnetic layers. To read the bit back, the device exploits a quantum effect called tunneling magnetoresistance (TMR): electrons tunnel through the thin barrier, and the electrical resistance they encounter depends on whether the octupole orders of the two layers point the same way or in opposite directions. High or low resistance — that difference is the stored bit, readable with a simple electrical measurement.
The headline number: the team observed a tunneling magnetoresistance of 363% at room temperature. That means the resistance in one state is more than four times the resistance in the other — a large, easily detected contrast, comparable in size to the best conventional ferromagnet-based tunnel junctions. Crucially, both the write (current-induced switching) and the read (large TMR) happen at room temperature, in the same all-antiferromagnet device, with no cryogenic cooling required. Experiments that only work near absolute zero rarely escape the laboratory. Room temperature is where computing lives.
Why silicon compatibility matters
There is a second, quieter breakthrough embedded in the paper. The team grew these devices on conventional silicon wafers using sputter deposition — an established, widely used technique in semiconductor manufacturing. Khalili notes that this should make it possible to deposit the junctions on the large wafer sizes that advanced chip production requires, using tools that already exist in semiconductor foundries.
That detail matters enormously. Many exotic materials discoveries stall because they cannot be made with the tools the chip industry owns. A memory technology that grows on plain silicon, with standard deposition equipment, can in principle be integrated alongside the processors and logic chips it would serve. The collaboration list hints at where the industry is watching: Western Digital, a major storage company, participated in the work, alongside researchers from several US and Italian universities.
This is also why the discovery was announced as a platform rather than a product. The reported evidence is specific: room-temperature switching and 363% TMR in nanoscale PtMn₃|MgO|PtMn₃ junctions, plus a theoretical model of the torque. The pathways to "deeply scaled magnetic memory" and room-temperature terahertz electronics are proposed directions for further development — plausible and exciting, but not yet demonstrated applications.
What this could enable
If OTT-based memory matures, three advantages stand out.
First, speed. Antiferromagnets respond to torques far faster than ferromagnets because their internal magnetic fields are enormous — once perturbed, the spins snap into the new state with extraordinary quickness. Terahertz-frequency operation is the long-term prize: memory and oscillators running thousands of times faster than today's gigahertz electronics.
Second, reliability and density. A memory with no stray fields cannot be corrupted by a nearby magnet and does not corrupt its neighbors. Cells can be packed closer together, growing storage capacity without pushing each cell to atomic dimensions.
Third, energy efficiency. Moving a spin with a current pulse is inherently more efficient than generating magnetic fields with wires, and antiferromagnetic switching promises to need less energy per bit written. In an era when AI data centers are straining power grids, memory that sips rather than gulps electricity matters at planetary scale.
There is also a conceptual shift worth noting. The work shows that electrical control of magnetism does not require a net magnetization at all — the thing engineers have relied on since the invention of the compass. A "spin-neutral" current can do the job through the octupole channel. That reframes what materials can be used for spintronics, the field of electronics that exploits electron spin rather than just electron charge.
Honest caveats
Good science reporting needs the other side of the ledger. The devices demonstrated are nanoscale research structures, not memory chips. Endurance (how many write cycles a cell survives), switching speed in real circuits, uniformity across billions of cells, and integration cost all remain open questions. The 363% TMR is a record for this material class, but commercial MRAM benefits from decades of optimization. And while the physics is demonstrated, nobody has yet built an OTT-based memory array, let alone shipped one.
It is also worth keeping this discovery in its own lane. Around the same time, physicists have been chasing an entirely different phenomenon — the "ghost of superconductivity," strange traces of zero-resistance electron pairing detected in the material UTe₂. Superconductivity is about electricity flowing without resistance; OTT is about storing information in magnetism. Both could reshape future electronics, but they are different physics, different materials, and different stories.
The bottom line
The Northwestern team's discovery gives antiferromagnetic spintronics something it has lacked since the field began: a practical way to write and read information using nothing but electric current, at room temperature, on silicon. OTT — a torque that acts on the material's octupole order instead of its magnetization — turns the invisibility of antiferromagnets from a bug into a feature. Faster, denser, more robust, and more efficient memory is the prize; the work ahead is turning a laboratory demonstration into a technology.