Issue 02 · Friday, October 23, 2026Deep Dive — Advanced Materials
INFLECTION.
The Weekly Magazine of Innovation
Advanced Materials · The Third Magnet

The Magnet That Isn't
Magnetic — and Why It Matters

For a century physics said magnets come in two kinds. Three physicists just won Europe's top condensed-matter prize for proving there are three. The new one has no external field at all — and that silence may be the whole point.

→  Why "invisible" beats "strong" in the memory wars
→  The poster-child material that may be a mirage
→  Signals: a telescope, a catalyst, a cancer label
Contents
Dispatch · Editor's note2
Feature · The Third Magnet3
Feature · Three Breakthroughs4
Feature · From Lab to Market5
Against the Grain · The Contrarian6
Signals7
By the Numbers8
The Long View · Sources9
Dispatch

Judge a material by what it refuses to do.

Engineers love strength. Stronger magnets, faster switches, denser bits. This week's cover story is a quiet argument that the next leap in memory comes from the opposite virtue: a magnet that does nothing outside itself.

Here is the lens for the issue. Every ferromagnet — the kind that sticks to your fridge — broadcasts a field. In a memory chip, that broadcast is a nuisance: pack bits closer and they start talking over each other. The old fix was antiferromagnets, whose neighbouring atoms cancel perfectly and so stay silent. The price of silence was that they were also deaf: you could barely read or write them electrically.

Altermagnets break that bargain. They are silent outside — zero net magnetization — yet inside, electrons of opposite spin travel differently, so current through them comes out spin-polarized. Silent to the neighbours, articulate to the circuit.

I'd like you to hold one question while reading: when a field matures, does the winner come from making the loud thing louder, or from discovering the third option nobody thought to name? Magnetism spent more than a hundred years in a two-box taxonomy. The box that was missing was hiding in plain symmetry.

One more thing: this issue is unusually honest about doubt. The most-studied altermagnet candidate is now disputed, and page 6 tells you why. A good discovery survives its own skeptics. Read on.

— The Editors
Inflection · Issue 022
Feature · Advanced Materials

The Third
Magnet

Altermagnetism cancels like an antiferromagnet and conducts like a ferromagnet. A century of textbooks missed it, and the 2026 Europhysics Prize now calls it a third fundamental class of magnetism.

By the Inflection desk · Sources on page 9

For more than a hundred years, physics taught that collinear magnets come in exactly two flavours. Ferromagnets line every atomic spin up the same way and produce a net field. Antiferromagnets alternate the spins, up-down-up-down, and the field cancels to zero. In 2022 a team around Libor Šmejkal, Jairo Sinova and Tomas Jungwirth published a full symmetry classification showing a third pattern was possible: spins still cancel, but the crystal environment of the "up" atoms is a rotated copy of the "down" atoms, not a shifted one. That small geometric twist changes everything electrons feel.

The problem with loud magnets

Magnetic memory is already in production. STT-MRAM, built on ferromagnets, is entering mass production in advanced processors. But a ferromagnetic bit leaks a stray field, and stray fields limit how closely bits can be packed. Think of a library where every reader mutters: put the desks too close and nobody can concentrate.

The problem with silent ones

Antiferromagnets are the silent library, and they dance at terahertz frequencies, orders of magnitude faster than ferromagnets' gigahertz. But they are hard to read and control with ordinary electrical signals. A silent magnet you cannot talk to is not a memory.

The hidden loophole

The altermagnet's trick is that the cancellation is perfect in total, but not direction by direction. An electron moving along one crystal axis sees the "up" landscape differently from the "down" one — a momentum-dependent spin splitting. Send current through and it emerges spin-polarized, with no net magnetization anywhere.

The Mainz-led prize announcement describes the field's scale: hundreds of candidate materials identified from symmetry in 2022, confirmation by photoemission and transport in several of them, and hundreds of papers within a few years.

Inflection · Issue 023
Feature · Three Breakthroughs

How a theory became a device

1 · The map (2022)

The decisive step was theoretical: a complete symmetry classification that told experimentalists where to dig. It turned "is there a third magnet?" into a shopping list of candidates. Per Tech Times, more than 200 candidates are predicted from symmetry.

2 · The sighting

Direct proof requires watching electrons. Angle-resolved photoemission spectroscopy (ARPES) maps how electron energy depends on direction; a spin splitting in a zero-magnetization crystal is the altermagnet's fingerprint. Spectroscopy and transport experiments have confirmed it in several materials, with MnTe and CrSb named as experimentally confirmed in Tech Times' account; one CrSb paper reports surface Fermi arcs suggestive of Weyl physics.

"An entirely new magnetic phase had remained hidden for more than one hundred years."
— Jairo Sinova, Mainz
3 · The first memory cell

In a paper published online 30 December 2025 in Nature Communications, a team built a three-terminal magnetic memory cell using a 12-nanometre film of RuO₂ as the write channel. Current through the altermagnet generated a spin current with a tilted polarization that flipped a 250-nm magnetic tunnel junction without any external magnetic field.

The measured figures: field-free switching at a critical current density of 2.1 × 10⁶ A/cm², a tunnel magnetoresistance of 54%, and a thermal stability factor of 64 — above the roughly 60 the authors cite as needed for ten-year retention. Read and write run through separate channels, which is the architectural prize.

Why that matters

Switching a magnet normally needs a helper field or heavy-metal tricks. Here the crystal itself supplies the tilt. The material is the circuit element.

Inflection · Issue 024
Feature · From Lab to Market

So what does a silent magnet buy you?

The pitch has three legs. First, density: no stray field means bits can sit closer. Second, speed: antiferromagnet-like dynamics point toward terahertz operation — though, to be exact, no altermagnetic device speed has been reported yet. Third, simplicity: spin-polarized current without a ferromagnetic reference layer could shorten the stack.

The target is the memory wall in AI data centres, where moving bits costs more than computing on them. Tech Times frames altermagnetic memory as a possible generational shift from DRAM and NAND, and is explicit that this is a projection, not a demonstration.

The honest bottleneck is not physics but manufacturing: a stable, manufacturable altermagnet that works at room temperature, deposited on standard wafers. That is a materials-engineering problem, and materials problems are measured in years.

Field Notes · How an altermagnet works
1

Alternate the spins. Neighbouring atoms point opposite ways, so net magnetization is zero. Nothing leaks.

2

Rotate, don't shift. The "down" atoms' surroundings are a rotated copy of the "up" atoms'. Symmetry now treats spins unequally.

3

Send current. Along certain directions, up- and down-spin electrons have different energies. The current comes out spin-polarized.

Glossary
Collinear — all spins lie on one axis, up or down.
Spin splitting — electrons of opposite spin have different energy at the same momentum.
ARPES — photoemission that maps electron energy vs. direction.
MTJ / TMR — a magnetic tunnel junction; TMR is how much its resistance changes between states.
A business reading

Whoever wins here wins on process, not on discovery: thin-film deposition recipes, crystal-orientation control, and a short list of validated materials. The RuO₂ device is notable because it uses a material already familiar to fabs as a thin-film conductor — a small but real advantage over exotic crystals.

Inflection · Issue 025
The Contrarian
Against the Grain · Move 37

The best material in the story
may not exist.

The consensus read: altermagnets are real, so build memory from the first one that worked. The contrarian read: the first device was built on a material whose altermagnetism is now actively disputed — and that may be the most useful thing that could have happened.

The strategic read

RuO₂ was the early poster child. Theory predicted a very large spin splitting; transport experiments agreed; and the December 2025 device used it. But a spin- and angle-resolved photoemission study (arXiv:2409.13504) found the electronic structure matched a non-magnetic prediction, with no sign of the expected splitting, and concluded the proposed order in RuO₂ is not supported. Tech Times reports a 2026 review attributing apparent thin-film signals to strain, stoichiometry or surface effects.

Now flip the lens. If a device can show field-free switching from a film whose bulk altermagnetism is contested, then the lasting asset is not one crystal. It is the method: oriented thin films, spin-torque readout, and a measurement stack that tells real splitting from impostors. The firms that master discrimination — knowing which candidate is real — will be worth more than those who bet on a single material.

Why the consensus disagrees

Honest risks, in order of severity. One: the confirmed list is short — MnTe and CrSb against 200-plus predictions. Two: the memory cell's 54% TMR and 2.1 × 10⁶ A/cm² switching current are proofs of concept, and the source reports no switching speed or energy. Three: the leap from a 250-nm lab cell to wafer-scale yield is untouched by this evidence. Four: the commercial timeline is, in Tech Times' words, dependent on which candidates validate and scale.

Where I land: treat altermagnetism as real physics and unproven product. The surprising conclusion is that the RuO₂ dispute is a feature of a healthy field, and the right portfolio is long on characterization tools, short on any single-material story.

Inflection · Issue 026
Department

Signals

63 nm
Space
Roman telescope beats its optics target

NASA's Nancy Grace Roman Space Telescope, launched August 30, measures optical error of 63 nm RMS against a pre-launch requirement of about 90 nm — roughly 30% better. Engineers got there by nudging the secondary mirror by about one-sixteenth of a human hair. The Wide Field Instrument carries 18 detectors aimed at dark matter and dark energy; its coronagraph focus has been corrected and light-blocking tests follow.

SOURCE · ScienceDaily / NASA, Oct 17 2026 — sciencedaily.com/releases/2026/10/261016070630.htm
9,389
Energy · Climate
A catalyst's reaction map grows 60-fold

Indian Institute of Science researchers used machine learning with quantum simulations to expand the reaction network for CO₂-to-methanol on copper from 152 to 9,389 reactions across 105 surface species. The small network wrongly named formic acid the main product; the large one predicted about 40 times more CO₂ conversion and correctly identified methanol and carbon monoxide. The hidden route: hydrogen can transfer as intact H₂. Experiments were run by teams at HPCL and A*STAR.

SOURCE · Nature Communications, DOI 10.1038/s41467-026-77080-4; via ScienceDaily, sciencedaily.com/releases/2026/10/261013015758.htm
62%
Biotech
FDA widens Anktiva — without a new trial

On October 14 the FDA extended Anktiva plus BCG to BCG-unresponsive, high-grade papillary-only bladder cancer, calling it the first approved treatment for that group. It did so by extrapolating from the carcinoma-in-situ data: 77 patients, 62% complete response. The separate papillary cohort showed 58.2% disease-free survival at 12 months. The regulatory lesson: biology that is hard to tell apart under a microscope can share evidence.

SOURCE · Pharmacy Times, Oct 14 2026 — pharmacytimes.com/view/fda-approval-expands-anktiva-to-bcg-unresponsive-papillary-bladder-cancer
Inflection · Issue 027
Department

By the Numbers

100+

years magnetism sat in a two-class taxonomy

200+

altermagnet candidates predicted from symmetry (Tech Times)

2.1×10⁶

A/cm² field-free switching current in the 250-nm RuO₂ cell

64

thermal stability factor vs. ~60 needed for 10-year retention

54%

tunnel magnetoresistance of that memory cell

63 nm

Roman telescope optical error vs. ~90 nm required

9,389

reactions in the expanded CO₂-to-methanol network, up from 152

Inflection · Issue 028
The Long View

The things we name late

Three stories this week share one shape: a hidden variable, finally counted. Altermagnetism was hiding in crystal symmetry. The CO₂ catalyst was hiding in 9,000 reactions the old model ignored. Anktiva's new label rests on the fact that two bladder-cancer presentations look alike. Roman's mirror gained its margin by a nudge too small to see.

Each time, progress came from enlarging the map, not strengthening the engine. And each time the honest next question was the same: which part of the new map is real? RuO₂ is this week's reminder that a bright early result is a hypothesis wearing a lab coat.

So when your own field feels like it has settled into two boxes, ask what symmetry you have not yet tested. The third box is usually there.

Sources & Further Reading

1. ScienceDaily (JGU Mainz), Scientists who discovered a third form of magnetism win major physics prize — https://www.sciencedaily.com/releases/2026/10/261017230035.htm

2. EurekAlert, 2026 Europhysics Prize honors discovery of altermagnetism — https://www.eurekalert.org/news-releases/1137502

3. Tech Times, Third Form of Magnetism Earns Europe's Top Physics Prize — https://www.techtimes.com/articles/321612/20260726/third-form-magnetism-earns-europes-top-physics-prize-could-reshape-ai-hardware.htm

4. Guo et al., Magnetic memory driven by spin splitting torque…, Nature Communications 17, 1309 — https://www.nature.com/articles/s41467-025-68065-w

5. Liu et al., Absence of altermagnetic spin splitting character in rutile oxide RuO₂ — https://arxiv.org/abs/2409.13504

6. Lu et al., Observation of surface Fermi arcs in altermagnetic Weyl semimetal CrSb — https://www.arxiv.org/abs/2407.13497

7. ScienceDaily/NASA, Roman Space Telescope exceeds expectations — https://sciencedaily.com/releases/2026/10/261016070630.htm

8. ScienceDaily/IISc, AI reveals hidden chemical shortcut for CO₂ to fuel — https://sciencedaily.com/releases/2026/10/261013015758.htm

9. Pharmacy Times, FDA approval expands Anktiva — https://www.pharmacytimes.com/view/fda-approval-expands-anktiva-to-bcg-unresponsive-papillary-bladder-cancer

Inflection · Issue 029
INFLECTION.
Next Issue · Friday, October 30
The Lens

What is everyone in your field optimizing — that a third option could make irrelevant?

Researched, written & designed with Claude.
Typeset in Poppins & Lora on the Anthropic palette.