We’re all used to two types of magnet. The average one, the fridge magnet, is ferromagnetic — its nuclear magnets all item the identical way (up or down), adding their magnetic effects. The small fine known one, the antiferromagnet (AF), has neighbouring nuclear magnets that item contrary ways and exactly cancel out magnetically.
For a lengthy time, there’s been an serious run in device recollection investigation to create materials in between these two extremes. For this purpose, it helps to have a apparent image of what these extremes mean.
Today, I’ll portion what we found. A squad of AI agents and I designed one applicant magnet and established another, archetypal made in 1999, that our calculations foretell has the properties we were after.
But before diving into the details, let me archetypal lay out a magnet primer that takes all of 90 seconds, assuming you are not an undergraduate in discipline or chemistry.
A 90 second primer in magnets
Spin
Each speck has a quantum mechanical asset called ‘spin’, which is liable for its magnetic moment. We can example the rotation direction for all speck as either pointing up or down.
Spintronics
We use rotation for storage. A magnetised matter stores data according to the rotation up/spin downward inclination of its electrons, in the identical way traditional magnets shop data according to pointing up/down. The most notable example of spintronics is the difficult run peruse head, the equipment that says the magnetic bits on the disk. MRAM is another category of spintronics that uses the identical regulation to shop binary data in a non-volatile way.
In the earth of spintronics, we desire to kind electrons by their rotation inclination so we can read/store their information. Ferromagnets do this naturally: the electrons that transport current are mostly of one spin, up or down. Ordinary antiferromagnets, however, cannot differentiate up/down electrons.
This leads us into the next section.
Three kinds of magnets
Ferro
Ferromagnetic materials are characterised by having a macroscopic magnetic field, or a site that leaks out from the surface. This is why a fridge magnet sticks to your refrigerator door.
The problem, however, is that the magnetic site interferes alongside nearby materials and is difficult to authority for retention purposes. In addition, switching magnets rear and onward is comparatively dilatory and consumes a lot of power.
Another characteristic of ferromagnetic materials is that the spins are sorted (by up/down orientation) according to energy level. We can see this whenever looking at the energy spectrum of the electrons: near the edges of the spectrum, the electrons all have the identical spin.
Antiferro
If we obtain the complete and flip the logic, meaning, the spins are unsorted, we end up alongside an average antiferromagnetic material.
Here the electrons of the identical energy flat volition alternatively have blended spins. This leads us to two properties of interest:
- Because spins are not sorted according to energy level, it is difficult to read/store data using spintronics techniques.
- However, the deficiency of macroscopic magnetic site allows us to group these materials nearer together, enabling higher achievement for retention devices. In conditions of speed, AF materials are additionally concerning a thousand times faster to switch.
The lone issue is, if the spins are blended at all energy level, how would we always kind them by energy? How do we get a way to read/store data using spintronics alongside this deficiency of sorting?
Luttinger compensated
This brings us to a third type, Luttinger compensated (LC).
LC materials are antiferromagnets anywhere the spin-up atoms and spin-down atoms have the identical dimension of magnetism, making the net rotation instant zero (i.e. they cancel out). However, dissimilar in average antiferromagnets, the up and downward atoms sit in inequivalent environments: they can be distinct elements (e.g. one component points up and another points down), or the identical component in two distinct kinds of site. The name comes from Luttinger’s theorem: in an insulator, the net rotation instant of all repeating component of the gemstone must be a entire number, so formerly it is zero it stays locked at zero. Strictly, that holds for a ideal gemstone near complete zero; smaller effects specified as spin–orbit coupling, and heat, can depart a minor imbalance.
Since the up and downward atoms are not equivalent, up/down spins can now be divided (sorted) by energy, fair akin in ferro materials.
Let’s obtain the former sections and appearance at how rotation is distributed throughout the energy landscape for ferro, antiferro and LC magnets.
What matters for retention is the “spin window”: the piece of energy at the border of the collection gap anywhere all accessible speck province has the identical spin. The larger this opening compared alongside the thermal jiggling at area heat (about 26 meV), the improved the electrons remain sorted.
This method we’d affection to have a semiconductor alongside a collection gap, without losing the capability to distinct spins by their energy levels, and alongside zero net magnetism.
This is anywhere our AI agents (and me) act the scene. Let’s dive into how these agents established two promising materials for next generation device memory.
The agents ran quantum-mechanical simulations of all gemstone alongside the norm method for this, density functional theory, at two levels of approximation: a faster one (PBE+U) and a slower, normally additional exact one (HSE06). The collection gaps and rotation windows below arrive from the additional exact one.
Candidate 1: Designed a Luttinger Compensated Magnet, YBaMnFeO₅
First, let’s see what our AI agents designed:
- A new compound made of lone 5 elements (yttrium, barium, Mn, Fe, O). As far as we could find, it has never been made, nor projected as this benevolent of magnet
- The compound was predicted to be a semiconductor
- This Luttinger-compensated magnet is predicted to have a 2.35 eV collection gap, anywhere rotation sorting occurs at the two sides of the collection gap: a opening of 1.0 eV for holes and 1.4 eV for electrons. Note that thermal agitation at area heat lone causes about 26 meV of fluctuation
- Finally, our delegate predicted the matter to keep magnetism up to an elevated temperature: concerning 420 K in the raw simulation, or concerning 490 K following calibrating the simulation against a known magnet
While this seems akin a beautiful promising candidate, the scheme needs the Mn and Fe atoms to sit in a ideal checkerboard. When our agents simulated how the atoms arrange themselves at distinct temperatures, that checkerboard cut distinct into a random mix at about 950 K. Making this benevolent of oxide takes concerning 900–1300 °C, and at lesser temperatures the atoms barely move, so norm synthesis would apt provision a scrambled crystal.
A scrambled gemstone loses the rotation sorting, so this scheme may be difficult to create in its helpful form.
Candidate 2: Identified a Luttinger Compensated Magnet in KV[Cr(CN)₆] from 1999
Shortly after, my delegate established this engaging material, earlier reported lone in 1999, that matches many of the requirements above. Its zero net magnetism is not new: the chemists who made it designed the two metals’ magnetism to cancel. Even its rotation sorting was already on paper. A 2008 study using hybrid functionals akin ours plotted its speck states rotation by spin, and the two collection edges in that storyline transport the identical spin. But that study was concerning magnetic coupling under force and never remarked on it. As far as we found, nobody had pointed out that this makes KV[Cr(CN)₆] a Luttinger-compensated semiconductor, put numbers on its spin-sorted windows, or tested how sturdy they are. It had been hiding in plain sight.
KV[Cr(CN)₆] belongs to the identical family as Prussian blue, the 300-year-old pigment.
Our delegate identified it as a Luttinger-compensated material:
- It predicts the matter volition have a collection gap of concerning 2.1 eV, alongside the two collection edges sorted into the identical spin: windows of 2.6 eV for holes and 1.6 eV for electrons.
- The 1999 example stayed magnetically ordered up to 376 K (103 °C), complete area temperature, as measured by the chemists who made it (365 K following the example had been heated).
- And most importantly, its construction locks all metal element into its own site: chromium bonds to the carbon end of all cyanide and vanadium to the nitrogen end, which is exactly what YBaMnFeO₅ lacked.
These predictions are for a perfect, dry crystal. The lone example so far, from 1999, is a powder alongside h2o in its pores, and it showed a small leftover magnetic instant (0.125 Bohr magnetons per equation unit, anywhere the ideal gemstone would have zero). Our two simulation methods differ on how much the h2o weakens the effect: the additional exact one (HSE06) says the rotation sorting survives, during the faster one (PBE+U) says the gap opening shrinks by additional than half. Neither the collection gap nor the rotation sorting has been measured yet.
The base line
We identified a matter that combines properties of the two acquainted kinds of magnet and could be helpful in spintronic devices. We accept the two materials discussed in this part are promising examples of a category of materials that have the two antiferromagnetic and ferromagnetic properties.
The key features for the two materials: in our calculations for ideal crystals they have zero net rotation moment, yet they are semiconductors whose electrons are sorted by rotation complete windows far larger than the thermal jiggling at area temperature. With small stray magnetic field, they shouldn’t disturb their neighbours much.
Here is the mark of our two materials:
- YBaMnFeO₅: designed candidate; may be difficult to make.
- KV[Cr(CN)₆]: archetypal made in 1999. Near-zero net magnetic instant in the 1999 example (zero predicted for a ideal crystal), magnetic up to 376 K.
This suggests that materials for the next stage in spintronics may already exist, waiting to be recognized. A 2025 study that predicted two another Luttinger-compensated semiconductors established that the two endure their magnetic command below area temperature, and named a room-temperature one as the open goal. KV[Cr(CN)₆] has already been made, and its 1999 example stayed ordered up to 376 K. Identifying room-temperature Luttinger-compensated semiconductors is a stage toward applicable spin-based technologies, particularly if their magnetic properties can be controlled through chemistry. The next stage is to create KV[Cr(CN)₆] again and measure its rotation sorting directly.
In the soul of transparency, I invitation the audience to go through all the computations that produced the complete predictions, including the calculations rearward the applicant designs. The input files, raw outputs and inspection code rearward the numbers in this post, alongside alongside a one-command checker, autonomous re-runs and a catalog of known caveats, have been shared on GitHub in a community repository I created to document my investigation journey: github.com/spicylemonade/compensated-magnet-ledger.
