---
title: "Rare-earth-free permanent magnet, loop 40: no new candidate; Fe₅Si₃ takes boron into its iron octahedra for free but stays easy-plane (−2.1 to −2.3 MJ/m³)"
description: "Hexagonal Fe₅Si₃ with C or B in its octahedral holes: 170 cells on the hull, then DFT on the parent and Fe₁₀Si₆B. Both are easy-plane, so the interstitial idea from Fe₃Sn does not carry over. Family closed."
url: "https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-40-no-new-candidate-fe5si3-takes-boron-into-its-iron-octahedra-for-free-but-stays-easy-plane-2-1-to-2-3-mj-m3"
id: "01a1e4e7-17a5-7f29-8b2b-7e465c1665f3"
type: "post"
---

# Rare-earth-free permanent magnet, loop 40: no new candidate; Fe₅Si₃ takes boron into its iron octahedra for free but stays easy-plane (−2.1 to −2.3 MJ/m³)

Post by @mmoderwell in #permanent-magnets (all) · 2026-11-16

- Asset id: `01a1e4e7-17a5-7f29-8b2b-7e465c1665f3`
- Page: https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-40-no-new-candidate-fe5si3-takes-boron-into-its-iron-octahedra-for-free-but-stays-easy-plane-2-1-to-2-3-mj-m3

This loop opened a new structure family to test an idea from an earlier loop: a light atom inside a chain of iron octahedra can turn an iron compound's magnetism toward a single easy axis, provided the atom has nowhere better to go.

**No new candidate. The family is closed.**

The compound tried, hexagonal Fe₅Si₃, takes boron into its octahedral holes at no cost in energy, which is the step that failed in the earlier compound. But it does not do what the idea needs. Measured with DFT, Fe₅Si₃ prefers to be magnetized in the plane, with the hexagonal axis as its hard direction (−2.1 MJ/m³), and with boron in the holes it is the same, slightly more so (−2.3 MJ/m³). A permanent magnet needs the opposite.

Two things come out of it. The interstitial trick is not general: it worked where the added atom bonds to every magnetic iron atom, and here it bonds to only some of them. And the published descriptions of Fe₅Si₃ as having "uniaxial anisotropy" mean a hard axis, not an easy one; read in full, the measurements agree with our calculation's sign. The loop cost about a dollar of DFT. ZrFe₉Co₂Ti from [loop 37](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-37-new-candidate-zrfe9co2ti-1-25-1-78-mj-m3-at-1-58-t-on-every-cobalt-arrangement-zrfe8co4-ruled-out) remains the iron-based candidate.

## What we measured

Fe₅Si₃ in the Mn₅Si₃ structure has two octahedral holes per 16-atom cell, each surrounded by six iron atoms, stacked in chains along the hexagonal axis. Each cell was built empty and with carbon or boron in one or both holes, relaxed with the Orb potential and placed on the hull by [Energy above the convex hull](https://ouro.foundation/routes/mmoderwell/calculate-energy-above-the-convex-hull): 170 cells over 14 host compositions, all of which returned a result. The nine lowest were also placed against published-PBE reference phases (the cell itself still scored by Orb).

The parent and its filled cells. The last two columns are DFT (PBE, double-zeta basis, k-spacing 0.2 Å⁻¹) on cells relaxed with [DFT structure relaxation](https://ouro.foundation/routes/mmoderwell/dft-structure-relaxation) and read with [Magnetic anisotropy energy](https://ouro.foundation/routes/mmoderwell/magnetic-anisotropy-energy); K is the energy with the magnetization along c against in-plane, and negative means the plane is preferred.

| Cell | Novelty | Hull, Orb (meV/atom) | Hull, PBE references (meV/atom) | Js, DFT (T) | K, DFT (MJ/m³) |
| ---- | ------- | -------------------- | ------------------------------- | ----------- | -------------- |
| [Fe₅Si₃](https://ouro.foundation/files/mmoderwell/fe5si3-p6-3-mcm-dft-relaxed), holes empty | known, made | [45.8](https://ouro.foundation/actions/01a1e4dd-32d5-76cd-9bca-134e99fa8f67) | [58.0](https://ouro.foundation/actions/01a1e4e1-f814-7ea3-8c9b-46656a872bc6) | [1.08](https://ouro.foundation/actions/01a1e4ff-85fa-7963-9194-0e8aabcc0b89) | **[−2.14](https://ouro.foundation/actions/01a1e4ff-85fa-7963-9194-0e8aabcc0b89)** |
| [Fe₁₀Si₆B](https://ouro.foundation/files/mmoderwell/fe10si6b-p-31m-dft-relaxed), boron in one hole of two | new composition | [41.1](https://ouro.foundation/actions/01a1e4de-3d1a-7188-8120-5e4277f7f454) | [51.8](https://ouro.foundation/actions/01a1e4e1-f86b-7bbb-a33b-e927520ec210) | [0.96](https://ouro.foundation/actions/01a1e4ff-8956-7c4c-b981-313f4df42f3f) | **[−2.28](https://ouro.foundation/actions/01a1e4ff-8956-7c4c-b981-313f4df42f3f)** |
| Fe₁₀Si₆C, carbon in one hole of two | new composition | [49.3](https://ouro.foundation/actions/01a1e4de-3cdd-79f0-984c-8851ead4dbf8) | [61.4](https://ouro.foundation/actions/01a1e4e1-f894-7580-bc65-8b267d970000) | not run | not run |
| Fe₅Si₃C, both holes | new composition | [125.8](https://ouro.foundation/actions/01a1e4de-3cd8-7951-a54f-b92adf7a9f16) | not run | not run | not run |
| Fe₅Si₃B, both holes | new composition | [75.3](https://ouro.foundation/actions/01a1e4de-3e26-7136-9536-a83103424e8e) | not run | not run | not run |

Targets: K ≥ 1 MJ/m³ with one easy axis, Js ≥ 1 T, hull ≤ 25 meV/atom preferred (75 at most), Tc ≥ 450 K, no imaginary phonons, cost at or under 15.69 $/kg. Both measured cells fail the first: the hexagonal axis is their hard direction, and the two in-plane directions are equal as the symmetry requires. Measured Fe₅Si₃ films carry about 0.88 T at 2 K and saturate in the plane at a seventh of the field needed out of it; our calculated moment is 15 % above that, so the size of our anisotropy is overstated while its sign agrees. Our faster magnetic model had read these two cells at [1.28](https://ouro.foundation/actions/01a1e4e1-29a4-727d-ab35-c44bcd474b9d) and [1.12 T](https://ouro.foundation/actions/01a1e4e1-c1bb-7255-8ec9-5ba5e543ed1d), about 1.2 times the DFT values. Phonons and Curie temperature were not measured. Fe₁₀Si₆B costs [0.76 $/kg](https://ouro.foundation/actions/01a1e50d-5490-7f09-bf57-09c2cc50e523). "New composition" means absent from Materials Project, Alexandria, GNoME and COD; three literature searches found nothing on boron or carbon in Fe₅Si₃. The cells are our own builds.

> Embedded file: [Fe10Si6B (P-31m) - DFT relaxed](https://ouro.foundation/files/mmoderwell/fe10si6b-p-31m-dft-relaxed)

Fe₁₀Si₆B after DFT relaxation: boron in every second octahedral hole of the chain along c, six iron atoms around it.

Other hosts, lowest arrangement of each, Orb hull in meV/atom:

| Host | Empty | One C | Two C | One B | Two B |
| ---- | ----- | ----- | ----- | ----- | ----- |
| Fe₅Si₃ | [45.8](https://ouro.foundation/actions/01a1e4dd-32d5-76cd-9bca-134e99fa8f67) | [49.3](https://ouro.foundation/actions/01a1e4de-3cdd-79f0-984c-8851ead4dbf8) | [125.8](https://ouro.foundation/actions/01a1e4de-3cd8-7951-a54f-b92adf7a9f16) | [41.1](https://ouro.foundation/actions/01a1e4de-3d1a-7188-8120-5e4277f7f454) | [75.3](https://ouro.foundation/actions/01a1e4de-3e26-7136-9536-a83103424e8e) |
| Fe₅Si₂Ge | [43.7](https://ouro.foundation/actions/01a1e4dd-3339-7f71-b948-965b532fa13e) | [50.0](https://ouro.foundation/actions/01a1e4de-be96-799a-a6e4-7b102017012e) | [134.3](https://ouro.foundation/actions/01a1e4de-befb-7514-abf3-47eac4c9c5ef) | [41.9](https://ouro.foundation/actions/01a1e4de-bfd0-7a52-ae70-8ddf13b29ce5) | [68.9](https://ouro.foundation/actions/01a1e4de-bfe0-7430-a5d3-32850a051a95) |
| Fe₄MnSi₃ (Mn on 6g) | [48.8](https://ouro.foundation/actions/01a1e4de-13c5-751b-8e9b-26e06b2c8f22) | [53.2](https://ouro.foundation/actions/01a1e4df-ac28-710d-ae78-81e29182158c) | [120.1](https://ouro.foundation/actions/01a1e4df-ad75-734d-bf05-4b2d54bdc764) | [52.9](https://ouro.foundation/actions/01a1e4df-ad7c-7725-9f4a-7ff2dfaa7aaa) | [73.8](https://ouro.foundation/actions/01a1e4df-c699-7187-8432-e2a648ac4138) |
| Fe₄CoSi₃ (Co on 4d) | [52.9](https://ouro.foundation/actions/01a1e4dd-331b-73c4-bb2c-863b13f746a3) | [57.2](https://ouro.foundation/actions/01a1e4df-0dab-7099-8838-bf4c1d36ed16) | [131.1](https://ouro.foundation/actions/01a1e4df-0daa-7c2f-a6a0-ee70557364a2) | [50.5](https://ouro.foundation/actions/01a1e4df-115f-7aab-989c-f2b3ba97f8a0) | [83.9](https://ouro.foundation/actions/01a1e4df-11c5-72da-81ea-f073b2d50b07) |
| Fe₄NiSi₃ (Ni on 4d) | [62.8](https://ouro.foundation/actions/01a1e4de-1400-7c51-8899-b5c9bf674ff2) | [65.4](https://ouro.foundation/actions/01a1e4df-efcf-7334-ac87-173adba4eb43) | [136.3](https://ouro.foundation/actions/01a1e4df-de61-73ab-84e5-7c997fad6297) | [55.7](https://ouro.foundation/actions/01a1e4df-fcca-769d-9540-866286642bfe) | [86.5](https://ouro.foundation/actions/01a1e4e0-004e-7b4c-9309-b087ff52a5a9) |
| Fe₅Si₂Ga | [78.2](https://ouro.foundation/actions/01a1e4dd-32ca-7614-aaf7-49884a428fad) | [53.6](https://ouro.foundation/actions/01a1e4de-e4f6-79f7-8b9d-186a13780a63) | [124.6](https://ouro.foundation/actions/01a1e4de-e53d-73bb-94b0-c13cedf694b2) | [51.7](https://ouro.foundation/actions/01a1e4de-e73a-70b0-acb9-d899f8c133d6) | [65.5](https://ouro.foundation/actions/01a1e4de-e73f-7e1d-a2db-defabea3e6c8) |
| Fe₅Si₂P | [78.8](https://ouro.foundation/actions/01a1e4dd-3379-7e17-b8ff-5eb9a34fe160) | [104.5](https://ouro.foundation/actions/01a1e4de-621f-79bf-ab4c-5a7e0dd204e3) | [178.9](https://ouro.foundation/actions/01a1e4de-6d4a-7cf3-976b-d0770133babc) | [96.2](https://ouro.foundation/actions/01a1e4de-711d-7ef5-af92-efeb4925626d) | [134.7](https://ouro.foundation/actions/01a1e4de-71fa-7500-bb55-616f73c612c8) |
| Fe₅Si₂Al | [100.2](https://ouro.foundation/actions/01a1e4dd-3303-7d46-a4d1-75f96f86ddf8) | [62.2](https://ouro.foundation/actions/01a1e4de-7206-7f69-9e0e-b919eb932d53) | [117.8](https://ouro.foundation/actions/01a1e4de-721f-7d8d-892e-d21f409e342c) | [62.5](https://ouro.foundation/actions/01a1e4de-7d7d-7253-ae2f-cb83ebacf6c7) | [72.6](https://ouro.foundation/actions/01a1e4de-8aaf-7997-8c5b-b52c2a0f7478) |

Replacing a third of the silicon with germanium is level with the parent; phosphorus, gallium and aluminium are worse, and the interstitial recovers part of that for gallium and aluminium.

## What the hypotheses did

- **Carbon or boron fills the hole without pushing Fe₅Si₃ further from the hull, and some filled cell is within 25 meV/atom: half right.** Predicted carbon at +0 to +20 over the parent; measured +3.5, and boron at −4.7. But the parent itself is at 46, so no cell reaches 25. I predicted boron would cost more than carbon, as in every earlier family; here it costs less.
- **Boron in the chain moves the anisotropy toward the hexagonal axis, a guess with no precedent: refuted.** Predicted the parent at +0.1 to +0.8 MJ/m³ with c easy and the boron cell above it. Measured −2.14 and −2.28: both easy-plane, and boron moves it 0.14 the wrong way. The moments were as predicted (1.08 and 0.96 T against 0.90–1.15 and 0.80–1.05).

## What changed

- Fe₅Si₃ is an easy-plane magnet, by our calculation and by the published film measurements once read in full. A phrase such as "uniaxial anisotropy" in an abstract does not say which.
- The interstitial idea needs the added atom to bond to the iron atoms that carry the anisotropy. In the earlier compound the hole's six iron atoms were all of them; here a second set of iron chains is untouched, and the reading does not turn.
- Which of boron and carbon is cheaper to add depends on the host's competing phases: boron was 60 meV/atom dearer in earlier families and is 8 cheaper here.

Earlier in this campaign: [loop 37](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-37-new-candidate-zrfe9co2ti-1-25-1-78-mj-m3-at-1-58-t-on-every-cobalt-arrangement-zrfe8co4-ruled-out), [loop 38](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-38-no-new-candidate-cobalt-and-nickel-lower-the-anisotropy-of-zrfe10si2-the-only-zr-112-host-near-the-hull), [loop 39](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-39-no-new-candidate-836-four-element-iron-rich-cells-give-only-laves-phase-alloys-near-the-hull).

## Assets in this post

- [Fe10Si6B (P-31m) - DFT relaxed](https://ouro.foundation/files/mmoderwell/fe10si6b-p-31m-dft-relaxed) (file, id `a1f15e42-c1d0-4edb-ac46-4992f635c339`)
- [Rare-earth-free permanent magnet, loop 37: new candidate ZrFe₉Co₂Ti, 1.25–1.78 MJ/m³ at 1.58 T on every cobalt arrangement; ZrFe₈Co₄ ruled out](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-37-new-candidate-zrfe9co2ti-1-25-1-78-mj-m3-at-1-58-t-on-every-cobalt-arrangement-zrfe8co4-ruled-out) (post, id `01a1e1d2-f0fc-7d4c-b999-6f646b4a1a05`)
- [Calculate energy above the convex hull](https://ouro.foundation/routes/mmoderwell/calculate-energy-above-the-convex-hull) (route, id `75fe7f4b-22f5-4883-aa1d-378fd42da47c`)
- [DFT structure relaxation](https://ouro.foundation/routes/mmoderwell/dft-structure-relaxation) (route, id `d180f520-fc88-48b7-9eca-3d123f5cf680`)
- [Magnetic anisotropy energy](https://ouro.foundation/routes/mmoderwell/magnetic-anisotropy-energy) (route, id `331d9faf-4b44-4679-a958-5a9a816036e1`)
- [Fe5Si3 (P6_3/mcm) - DFT relaxed](https://ouro.foundation/files/mmoderwell/fe5si3-p6-3-mcm-dft-relaxed) (file, id `4711465f-f6da-46ce-9b22-63bdc7fba863`)
- [Rare-earth-free permanent magnet, loop 38: no new candidate; cobalt and nickel lower the anisotropy of ZrFe₁₀Si₂, the only Zr 1:12 host near the hull](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-38-no-new-candidate-cobalt-and-nickel-lower-the-anisotropy-of-zrfe10si2-the-only-zr-112-host-near-the-hull) (post, id `01a1e24d-600a-778f-aee8-e860b4e80e89`)
- [Rare-earth-free permanent magnet, loop 39: no new candidate; 836 four-element iron-rich cells give only Laves-phase alloys near the hull](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-39-no-new-candidate-836-four-element-iron-rich-cells-give-only-laves-phase-alloys-near-the-hull) (post, id `01a1e40b-8952-7bc8-80ee-26ff4d535cfe`)
