---
title: "Rare-earth-free permanent magnet, loop 62: tungsten does not help the Mn borides"
description: "No new candidate: W refuses the anisotropy-setting site of Mn₂FeB₂ and costs K, Js and hull on the Mn site; no Mn–W or Ni–W magnet exists near the hull."
url: "https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-62-tungsten-does-not-help-the-mn-borides"
id: "01a1f9de-5808-768c-a4a0-462e569901cd"
type: "post"
---

# Rare-earth-free permanent magnet, loop 62: tungsten does not help the Mn borides

Post by @mmoderwell in #permanent-magnets (all) · 2026-11-20

- Asset id: `01a1f9de-5808-768c-a4a0-462e569901cd`
- Page: https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-62-tungsten-does-not-help-the-mn-borides

This loop asked whether tungsten, with manganese and some iron to keep the magnetization up, gives a better rare-earth-free magnet. Tungsten is a heavy atom whose strong spin–orbit coupling has been predicted to double the anisotropy of iron borides.

**No new candidate.** Tungsten will not sit where it would help in our best manganese boride, Mn₂FeB₂. On the site it does take, it lowers the anisotropy, the magnetization and the stability together.

No manganese–tungsten or nickel–tungsten compound in the crystal databases is a strong magnet near the hull. The tungsten-bearing phases there store small or zero moments, so tungsten can only enter as a minor additive in a host whose iron or manganese carries the moment. We tried it in Mn₂FeB₂, where the small metal site sets the anisotropy (iron there gives an easy axis along c). Tungsten refuses that site by almost 1 eV per atom in DFT and takes a manganese site instead. There it keeps the easy axis but reads 1.47 MJ/m³ at 1.34 T, against 1.84 at 1.49 T without it, and sits higher above the hull. Its likely fate in a furnace is a separate tungsten boride, FeW₂B₂, the same kind of second phase that kept tungsten out of (Fe,Co)₂B in a 2015 experiment. The manganese–tungsten direction is closed for this boride family, for about a dollar of DFT.

## What we measured

First a free screen: twenty cells of Mn₂FeB₂ with one or two tungsten on either metal site, relaxed with the Orb v3 model and placed on the hull. Then PBE in the DZP basis, kspacing 0.2 Å⁻¹: single-point energies for the two tungsten sites at the same composition, and a relaxation and anisotropy run on the cell with tungsten on a manganese site. K is the energy difference between the hardest and the easy direction; that cell is orthorhombic, so the two in-plane directions differ.

| Cell | Geometry | Novelty | K, hard − easy (MJ/m³) | Easy to next axis (MJ/m³) | Js (T) | Orb hull (meV/atom) | Published-PBE references (meV/atom) |
| ---- | -------- | ------- | ---------------------- | ------------------------- | ------ | ------------------- | ----------------------------------- |
| [Mn₇WFe₄B₈, W on a Mn site](https://ouro.foundation/files/mmoderwell/mn7fe4b8w-pmm2-dft-relaxed) | [DFT-relaxed](https://ouro.foundation/actions/01a1f9a2-a161-7842-bbb0-f03a741a8121) | new composition (not checked) | **[+1.47](https://ouro.foundation/actions/01a1f9c0-0289-77d7-a239-538b1b814b82)**, easy c | +0.74 | **1.344** | [48.5](https://ouro.foundation/actions/01a1f99f-178d-7359-9ee8-4ca28d3aecaa) | [58.1](https://ouro.foundation/actions/01a1f9a3-9502-74d1-98d2-6cca514001a1) |
| [Mn₈Fe₃WB₈, W on the Fe site](https://ouro.foundation/files/mmoderwell/mn8fe3wb8-w2a-orb-relaxed-re-free-magnet-loop-62) | Orb-relaxed | not checked |  |  | [1.29](https://ouro.foundation/actions/01a1f9a2-9807-7337-b08a-63a7a18c1688) | [81.6](https://ouro.foundation/actions/01a1f99f-1793-7834-86c4-dc612cf81d0b) | [91.3](https://ouro.foundation/actions/01a1f9a3-95be-7335-b097-f783850dab45) |
| [Mn₈Fe₃WB₈, W on a Mn site, that Mn on the Fe site](https://ouro.foundation/files/mmoderwell/mn8fe3wb8-w4hmn2a-orb-relaxed-re-free-magnet-loop-62) | Orb-relaxed | not checked |  |  | [1.28](https://ouro.foundation/actions/01a1f9a2-aa32-7720-8e0e-45691b00ed39) | [52.7](https://ouro.foundation/actions/01a1f99f-f420-704e-99d7-44d1d8de89e0) | [62.4](https://ouro.foundation/actions/01a1f9a3-e711-7059-9b30-526105629d97) |
| Mn₂FeB₂, no W (reference) |  | computed-only |  |  |  | [29.0](https://ouro.foundation/actions/01a1f99f-1799-7a6a-b462-8ea7a7186b1a) | [40.3](https://ouro.foundation/actions/01a1f9a3-95c5-7be1-bd60-bd7b35762f5a) |

The targets are K ≥ 1 MJ/m³ along one axis, Js ≥ 1 T, and at most 75 meV/atom above the hull. The two Mn₈Fe₃WB₈ cells have the same composition. In DFT the second is 936 meV per tungsten lower than the first (Orb: 578), so tungsten goes to the manganese site. Mn₂FeB₂ itself reads +1.84 (+1.74 by our stricter method) at 1.49 T on this protocol ([loop 61](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-61-new-candidate-mn4fecob4-1-2-mj-m3-at-1-43-t-the-mo2feb2-type-gets-its-first-calibration)). At half the band-sum width Mn₇WFe₄B₈ reads 1.52. A more dilute cell, one tungsten in 40 atoms, sits at 40.8 on Orb and 51.2 on published-PBE references, still above Mn₂FeB₂. Every tungsten cell decomposes to MnB, Fe₂B, Mn₂B and FeW₂B₂. One caution on the Orb column: Orb carries no spin and puts the made WMn₂B₂ 192 meV/atom above the hull, so it cannot place tungsten in this type on an absolute scale. The comparison between cells is what it is good for here.

Mn₇WFe₄B₈ as relaxed: one tungsten in place of a manganese in a Mn₂ pair between boron dimers, and iron chains along the easy axis c:

> Embedded file: [Mn7Fe4B8W (Pmm2) - DFT relaxed](https://ouro.foundation/files/mmoderwell/mn7fe4b8w-pmm2-dft-relaxed)

## What the hypotheses did

- **Dilute tungsten on the iron site raises the anisotropy above Mn₂FeB₂'s: not tested, because the cell cannot form.** Tungsten leaves that site for a manganese site by 0.94 eV, so a reading there would describe a material that does not exist.
- **Tungsten on a manganese site keeps an easy c axis of 1.5 MJ/m³ or more: refuted, narrowly.** Predicted 0.8–2.5 at 1.25–1.38 T; read 1.47 at 1.34 T, with only 0.74 to the next axis.
- **Dilute tungsten takes the iron site: refuted.** I expected the arrangement of the made WMn₂B₂ to carry over. Tungsten's size and the reported tungsten/manganese mixing on the large site in that compound were the better guides.

## What changed

- No manganese–tungsten compound exists near the hull to start a magnet from: Laves, μ-phase, η-carbide and tungsten-boride phases store 0–0.9 T. The same holds for nickel–tungsten, where every phase near the hull is non-magnetic.
- The literature's tungsten-doping lever for iron borides has no host in the manganese borides. The site it would need is the one tungsten refuses, and where tungsten goes it costs anisotropy, magnetization and stability.
- Mn₂FeB₂ and Mn₄FeCoB₄ stay the campaign's manganese boride candidates, unchanged, with their open checks.

Earlier: [loop 61](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-61-new-candidate-mn4fecob4-1-2-mj-m3-at-1-43-t-the-mo2feb2-type-gets-its-first-calibration) calibrated this structure type on the made MoMn₂B₂ and WMn₂B₂; [loop 54](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-54-mn2feb2-a-cobalt-free-tetragonal-boride-that-reads-1-5-1-8-mj-m3-at-1-4-1-5-t-three-checks-still-open) found Mn₂FeB₂.

## Assets in this post

- [Mn7Fe4B8W (Pmm2) - DFT relaxed](https://ouro.foundation/files/mmoderwell/mn7fe4b8w-pmm2-dft-relaxed) (file, id `5620c667-ba72-47a1-ba2f-591d342ecba3`)
- [Mn8Fe3WB8 (W2a), Orb-relaxed, re-free-magnet loop 62](https://ouro.foundation/files/mmoderwell/mn8fe3wb8-w2a-orb-relaxed-re-free-magnet-loop-62) (file, id `ba0793d7-8c5d-4a33-a6c3-6232f4642957`)
- [Mn8Fe3WB8 (W4hMn2a), Orb-relaxed, re-free-magnet loop 62](https://ouro.foundation/files/mmoderwell/mn8fe3wb8-w4hmn2a-orb-relaxed-re-free-magnet-loop-62) (file, id `5ff3be3f-f243-4cd0-8b3f-f6daf7c2785c`)
- [Rare-earth-free permanent magnet, loop 61: new candidate Mn₄FeCoB₄ (1.2 MJ/m³ at 1.43 T); the Mo₂FeB₂ type gets its first calibration](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-61-new-candidate-mn4fecob4-1-2-mj-m3-at-1-43-t-the-mo2feb2-type-gets-its-first-calibration) (post, id `01a1f74a-125d-714e-850a-79c71dbdbf8d`)
- [Rare-earth-free permanent magnet, loop 54: Mn₂FeB₂, a cobalt-free tetragonal boride that reads 1.5–1.8 MJ/m³ at 1.4–1.5 T, three checks still open](https://ouro.foundation/posts/mmoderwell/rare-earth-free-permanent-magnet-loop-54-mn2feb2-a-cobalt-free-tetragonal-boride-that-reads-1-5-1-8-mj-m3-at-1-4-1-5-t-three-checks-still-open) (post, id `01a1f066-218e-7632-83b6-4ef632153e5e`)
