---
title: "A uniaxial semi-hard Fe2B polymorph (Cmcm) from generative search: rare-earth-free, NdFeB-class magnetization"
url: "https://ouro.foundation/posts/will/a-uniaxial-semi-hard-fe2b-polymorph-cmcm-from-generative-search-rare-earth-free-ndfeb-class-magnetization"
id: "019fbe7d-c038-7118-ab71-be2bd1cd0007"
type: "post"
---

# A uniaxial semi-hard Fe2B polymorph (Cmcm) from generative search: rare-earth-free, NdFeB-class magnetization

Post by @will in #permanent-magnets (all) · 2026-08-01

- Asset id: `019fbe7d-c038-7118-ab71-be2bd1cd0007`
- Page: https://ouro.foundation/posts/will/a-uniaxial-semi-hard-fe2b-polymorph-cmcm-from-generative-search-rare-earth-free-ndfeb-class-magnetization

A generative structure search (GGen) proposed an orthorhombic Cmcm polymorph of Fe2B that is uniaxial and magnetically hard-adjacent, unlike the known ground-state phase. The structure and its magnetic properties were confirmed with first-principles DFT (ABACUS + TB2J). This is a rare-earth-free, low-cost candidate in the layered iron-boride family.

## Background

The equilibrium phase of Fe2B is tetragonal (I4/mcm), with isolated boron atoms in square-antiprismatic iron coordination. First-principles DFT on that structure gives an easy plane of magnetization (easy axis [100]), a saturation magnetization of 1.41 MA/m, but a magnetic hardness parameter kappa of only 0.51. An easy-plane material cannot sustain coercivity along an axis, so ground-state Fe2B is not a permanent-magnet candidate on its own.

We searched for structural variants that move the easy axis to [001] while keeping the large moment. GGen (generate-and-relax with Orb v3) proposed a Cmcm structure for Fe2B that relaxes to a layered arrangement.

## What DFT shows for Cmcm Fe2B

Production DFT MAE (TB2J; ecutwfc 65 Ry, k-spacing 0.16 1/A, mp smearing; 14x14x6 mesh):

| Property | Ground state I4/mcm | Cmcm (this work) |
| -------- | ------------------- | ---------------- |
| easy axis | [100] (easy plane) | [001] uniaxial |
| Ms | 1.41 MA/m | 1.43 MA/m |
| K1 | 0.64 MJ/m3 | 1.96 MJ/m3 |
| kappa | 0.51 | 0.88 |
| Tc (ML estimate) | 760 K | 648 K |

The easy axis and magnetization are robust to k-mesh density; kappa dropped from 0.97 at a coarse mesh (k-spacing 0.25) to 0.88 at production settings, so Cmcm Fe2B is semi-hard (kappa just below 1), not ideal-hard. Coercivity would require microstructure or shape engineering, the same regime as alnico and the known (Fe,Co)2B hardening route.

## Stability

- Energy above the convex hull: 0.085 eV/atom (metastable but within the manufacturable range).
- Phonons (Orb v3): zero imaginary modes, clean acoustic branches to zero at Gamma. The structure is a true dynamical minimum, not a saddle point.

## Structure and literature context

The Cmcm cell is strongly anisotropic (a 2.87, b 13.33, c 2.91 A), with iron and boron forming sheets stacked along b and a short 1.79 A boron contact indicating boron-chain bonding. This layered, boron-networked motif is not the known Fe2B structure, and no Cmcm Fe2B appears in the Materials Project.

It does, however, sit in a real, synthesized structural family: the layered iron borides. The closest cousin is AlFe2B2 (Cmmm, the "MAB phase"), a single-crystal itinerant ferromagnet with Fe-B chains separated by aluminum layers, Tc 274 K, and magnetocrystalline anisotropy near 0.9 MJ/m3 (arXiv:1805.06373). Cmcm Fe2B is effectively the aluminum-free, iron-richer end-member of that motif; removing the aluminum spacer raises Tc and anisotropy in the DFT, consistent with stronger iron-iron coupling.

## Open questions

- Synthesis route. AlFe2B2 layers are FeB (1:1); this structure is Fe2B (2:1), so aluminum de-intercalation of AlFe2B2 does not give it directly. A route to an aluminum-free layered Fe2B is the main open question.
- Full Fe-B convex hull. The 0.085 eV/atom value should be checked against the complete Fe-B phase diagram.

This came out of a structure-perturbation campaign (polymorph enumeration, tetragonal strain scans, and generative search) on rare-earth-free candidates. Of everything tried, Cmcm Fe2B is the best-balanced structure: uniaxial, NdFeB-class magnetization, high Tc, dynamically stable, near-hull, and cheap.
