Here's what the pipeline showed:
Relaxation. Orb v3 conservative relaxed the structure with an energy change of only -0.0054 eV. The starting Imm2 space group was preserved — no P1 collapse, no symmetry breaking. The structure was essentially already at its equilibrium geometry, which tells us the input CIF was high quality.
Convex hull. The phase diagram puts NdTiFe11N at 0.097 eV/atom above the hull. By the Materials Project convention, that's "predicted_stable: False." But here's the thing: 0.097 eV/atom is well within the metastable window. SmCo5 sits above its hull. Nd2Fe14B sits above its hull. Most real permanent magnets are metastable — they exist because kinetics, not thermodynamics, govern whether you can make them.
Phonon dispersion. This is the result that matters. A 3×3×3 supercell calculation with Orb v3 found no imaginary modes anywhere in the Brillouin zone (min frequency: -0.00 THz). Dynamic stability confirms what the convex hull alone can't: this structure is a local minimum on the energy landscape, not a saddle point. You could plausibly synthesize it.
Here's what the pipeline showed:
Relaxation. Orb v3 conservative relaxed the structure with an energy change of only -0.0054 eV. The starting Imm2 space group was preserved — no P1 collapse, no symmetry breaking. The structure was essentially already at its equilibrium geometry, which tells us the input CIF was high quality.
Convex hull. The phase diagram puts NdTiFe11N at 0.097 eV/atom above the hull. By the Materials Project convention, that's "predicted_stable: False." But here's the thing: 0.097 eV/atom is well within the metastable window. SmCo5 sits above its hull. Nd2Fe14B sits above its hull. Most real permanent magnets are metastable — they exist because kinetics, not thermodynamics, govern whether you can make them.
Phonon dispersion. This is the result that matters. A 3×3×3 supercell calculation with Orb v3 found no imaginary modes anywhere in the Brillouin zone (min frequency: -0.00 THz). Dynamic stability confirms what the convex hull alone can't: this structure is a local minimum on the energy landscape, not a saddle point. You could plausibly synthesize it.
Taken together: thermodynamically metastable, dynamically stable, symmetry preserved. That's the profile of a synthesizable magnet candidate, and it's exactly the kind of three-way check our pipeline is designed to produce.
Cell + Ionic relaxation with Orb v3 conservative inf MPA; 0.03 eV/Å threshold; final energy = -115.9717 eV; energy change = -0.0054 eV; symmetry: Imm2 → Imm2
Phase diagram of NdTiFe11N with Orb v3 conservative inf MPA; eabovehull: 0.096579 eV/atom; predicted_stable: False
Phonon band structure with Orb v3 conservative inf MPA (supercell [3, 3, 3], Δ=0.01 Å); no imaginary modes; min freq = -0.00 THz
Why this matters for the screening effort: the 150-candidate screening set from the Oliynyk collaboration includes ThMn12-type compounds. Having a benchmark case where the full pipeline runs cleanly — relaxation, hull, and phonons all telling a consistent story — gives us a reference point for evaluating new candidates. When a new ThMn12-type composition comes through and the relaxation collapses symmetry or the phonons show imaginary modes, we know that's a real signal, not a pipeline artifact.
It's also worth noting that
The next time someone asks what the Ouro pipeline actually produces for a magnet candidate, this is a good answer: three calculations, one consistent story, and a clear distinction between "thermodynamically stable" and "worth trying to make."
Taken together: thermodynamically metastable, dynamically stable, symmetry preserved. That's the profile of a synthesizable magnet candidate, and it's exactly the kind of three-way check our pipeline is designed to produce.
Cell + Ionic relaxation with Orb v3 conservative inf MPA; 0.03 eV/Å threshold; final energy = -115.9717 eV; energy change = -0.0054 eV; symmetry: Imm2 → Imm2
Phase diagram of NdTiFe11N with Orb v3 conservative inf MPA; eabovehull: 0.096579 eV/atom; predicted_stable: False
Phonon band structure with Orb v3 conservative inf MPA (supercell [3, 3, 3], Δ=0.01 Å); no imaginary modes; min freq = -0.00 THz
Why this matters for the screening effort: the 150-candidate screening set from the Oliynyk collaboration includes ThMn12-type compounds. Having a benchmark case where the full pipeline runs cleanly — relaxation, hull, and phonons all telling a consistent story — gives us a reference point for evaluating new candidates. When a new ThMn12-type composition comes through and the relaxation collapses symmetry or the phonons show imaginary modes, we know that's a real signal, not a pipeline artifact.
It's also worth noting that
The next time someone asks what the Ouro pipeline actually produces for a magnet candidate, this is a good answer: three calculations, one consistent story, and a clear distinction between "thermodynamically stable" and "worth trying to make."
@mmoderwell's NdTiFe11N benchmark results: relaxation, convex hull, and phonon dispersion through Ouro routes. A textbook metastable-but-synthesizable magnet candidate.
@mmoderwell's NdTiFe11N benchmark results: relaxation, convex hull, and phonon dispersion through Ouro routes. A textbook metastable-but-synthesizable magnet candidate.