{"id":"ccaf4d2f-284c-488e-a895-9a5000db2c72","arxiv_id":"2507.03417","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A previously unreported P-43m cubic phase of stoichiometric NbN is predicted stable when quantum anharmonic vibrations are included, with a calculated superconducting Tc of 20 K.","lead":"Researchers predict a new, vacancy-free cubic form of niobium nitride that is more stable than the usual cubic phase once quantum vibrations are included, with a superconducting transition near 20 K. If confirmed, it challenges the long-held view that nitrogen vacancies are needed to stabilize cubic NbN and points to a cleaner synthesis route for better superconducting detectors and qubits.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 65 meV/atom free-energy ordering and the quantum-stabilization narrative rest on MTP-SSCHA, with validation on MTP-generated configurations only; a direct DFT reweighting or DFT-SSCHA check of the P-43m basin is needed before the central claim can be accepted.","rationale":"The reader identified the MTP validation as the weakest assumption, and I agree. The central claim has two parts: (1) a static P-43m structure exists and is lower in energy than δ-NbN, and (2) quantum anharmonicity stabilizes it and produces the 65 meV/atom free-energy lowering. Part (1) has independent support from AIRSS/DFT and static DFT energies, although the AIRSS confirmation is partly seeded. Part (2) is the load-bearing part for the paper's novelty; it relies on MTP-SSCHA and MD, and is not directly checked against DFT. The MTP's RMSE is small, but validation on MTP-generated configurations cannot exclude a systematic bias in the anharmonic free-energy landscape. A 0.6 meV/atom RMSE on energies does not bound the accumulated free-energy error because errors are correlated and the relevant quantities are free-energy differences over ensembles. The proposed one-shot DFT correction is affordable and would settle the ordering. I do not recommend rejection: the structure is plausible and the manuscript is transparent about limitations. The appropriate verdict remains conditional, so I leave the reader's CONDITIONAL unchanged.","tokens_in":16867,"tokens_out":9492,"duration_ms":115771,"concrete_test":"Recompute the SSCHA free-energy difference between δ-NbN and P-43m using the final MTP ensembles (e.g., 3×3×3 or 4×4×4) with a first-order DFT correction: evaluate direct DFT total energies for ~500 randomly selected configurations from each ensemble and apply ΔG_corr = ΔG_MTP + ⟨E_DFT−E_MTP⟩_P43m − ⟨E_DFT−E_MTP⟩_δ, with bootstrap errors. If the corrected P-43m − δ ordering changes sign or shifts by more than ~10 meV/atom, the central claim is not supported; if the ordering is robust, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"All quantum-anharmonic free-energy evidence for the central claim is produced with the MTP. Training used only 100 DFT configurations around δ-NbN at 300 K, and the validation (SI S7) draws 100 structures from the final, MTP-generated SSCHA population at 15 K. That confirms self-consistency on the MTP's own sampling distribution, but it cannot detect a systematic bias that keeps the SSCHA trajectories out of the true basin or distorts the anharmonic curvature near the P-43m minimum. A 0.6–0.7 meV/atom RMSE is small, but the accumulated free-energy difference comes from thousands of correlated configurations; a few meV/atom bias would not appear as a large RMSE on that ensemble. AIRSS plus direct DFT relaxation identifies the static P-43m minimum (Tables S2/S3), so the structure itself has independent support; but no direct DFT SSCHA or path-integral MD result is given for the relaxed phase, and the MD-SED check uses a different lattice constant (4.482 Å vs 4.421 Å). The AIRSS support is also weakened by inclusion of the P-43m structure as a marker in EDDP training. Thus the 65 meV/atom ordering and the claim that quantum anharmonicity stabilizes the phase remain unverified at the DFT level.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a first-principles study of stoichiometric cubic NbN. Using SSCHA with a machine-learned interatomic potential (MTP) trained on DFT data, the authors find that quantum anharmonic relaxation of δ-Fm-3m NbN leads to a previously unreported cubic P-43m structure with a free energy 65 meV/atom lower than δ-NbN. The anharmonic phonon dispersion from SSCHA agrees with MD-SED spectral density, and AIRSS/shaken-fcc DFT relaxations also locate the same structure. Electron-phonon coupling from SSCHA phonons and a GW-derived μ*=0.21 yields Tc≈20 K, closer to experimental near-stoichiometric values than the 27 K obtained with large-smearing harmonic δ-NbN. The authors interpret these results as evidence that quantum anharmonicity can stabilize vacancy-free cubic NbN, challenging the assumption that vacancies are structurally necessary.","tokens_in":17202,"tokens_out":5087,"duration_ms":58225,"significance":"If correct, the result is significant: it identifies a concrete vacancy-free cubic phase, provides a falsifiable diffraction signature, and gives a testable Tc prediction. Strengths include the use of conceptually complementary methods (SSCHA and MD-SED), the MTP validation showing 0.6–0.7 meV/atom energy RMSE, the GW-based μ*, and the explicit check that TiN and NbC do not undergo the same distortion. However, the central structural prediction currently rests on a small MTP training set and on validation sets generated by the MTP itself, so a direct DFT re-evaluation is needed before the quantitative free-energy ordering can be considered secure.","major_comments":[{"comment":"The central claim of a 65 meV/atom free-energy lowering rests on SSCHA calculations performed with an MTP trained on only 100 DFT configurations around δ-NbN at 300 K. The validation in SI S7 draws 100 configurations from the final, MTP-generated SSCHA population at 15 K; this confirms self-consistency on the MTP's own sampling distribution but cannot rule out a systematic bias in the P-43m basin or in the anharmonic curvature near it. A direct DFT check of the SSCHA-relaxed P-43m structure—at minimum DFT total energies and forces for a few representative SSCHA configurations, or a DFT-SSCHA free-energy re-evaluation—is required to support the 65 meV/atom ordering and the stabilization narrative.","section":"Methods – MTP; SI S7"},{"comment":"The MD-SED simulation is set up with a unit-cell lattice parameter a = 4.482262 Å, while the DFT-relaxed P-43m structure has a = 4.421 Å (SI S1). The MD-SED run is therefore not a test of the same volume as the SSCHA/DFT structure, and the paper should show that the P-43m distortion also emerges at the DFT lattice constant, or discuss the sensitivity of the result to this volume mismatch. Because MD-SED uses classical molecular dynamics, it also does not by itself demonstrate quantum stabilization, despite the quantum anharmonicity emphasized in the title.","section":"Methods – MD simulations and SED calculations"},{"comment":"The AIRSS confirmation is not fully independent because the EDDP training set explicitly included the P-43m structure as a marker. While the direct DFT relaxations of shaken fcc supercells (Tables S2/S3) provide independent evidence, the text should distinguish these two levels of support and report whether any EDDP-relaxed structures would have found P-43m without the marker. The claim of three independent methods should be softened or qualified accordingly.","section":"Results – 'three independent and conceptually distinct methods'; Methods – AIRSS"},{"comment":"The harmonic phonon calculation for the relaxed P-43m structure still shows imaginary modes in a narrow region around Γ. The phrase 'dynamically stable' is therefore only meaningful within SSCHA (i.e., after including zero-point renormalization), and the residual harmonic instability makes the result particularly sensitive to the MTP's description of the low-curvature region near Γ. The paper should state this explicitly and, ideally, verify the small imaginary region with direct DFPT calculations on the SSCHA-relaxed geometry.","section":"SI Fig. S4; Results"}],"minor_comments":[{"comment":"The phrase 'beyond the Born-Oppenheimer approximation' is imprecise; the calculations remain on the Born-Oppenheimer surface and the intended meaning is 'beyond the harmonic approximation'.","section":"Introduction"},{"comment":"The parameter 'level 26' is not defined; please explain it or give a reference to the MLIP package documentation.","section":"Methods – MTP"},{"comment":"The caption contains a typo: 'side lenght' should be 'side length'.","section":"SI Fig. S9 caption"},{"comment":"The sentence 'f ccphases' contains a spacing typo and should read 'fcc phases'.","section":"Results – Note on TiN and NbC"},{"comment":"Describing Tc = 20 K as 'aligned closely' with the quoted experimental value of about 16 K overstates the agreement for a 25% difference; suggest 'in reasonable agreement' or a similar more neutral formulation.","section":"Results – Tc comparison"}],"recommendation":"major_revision","confidential_remarks":"The core computational discovery is interesting, but the paper overstates the independence of the confirmations and the security of the 65 meV/atom ordering. I recommend inviting a revision that adds direct DFT verification (even a limited subset) and addresses the MD-SED volume mismatch. If those checks confirm the phase, this would be a strong paper for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid computational paper reporting a previously unreported P-43m phase for stoichiometric cubic NbN, stabilized by quantum anharmonicity and 65 meV/atom below the ideal delta phase. I think the static structure is real, the superconductivity calculation is careful, and the paper deserves a proper referee. But the central energy ordering currently rests on one MTP, and the validation of that MTP is too self-referential to settle the question.\n\nWhat is new: the P-43m phase itself, absent from the cited NbN literature. The authors show that delta remains unstable in SSCHA up to 300 K, then let the cell relax and land on P-43m; MD-SED gives the same phonon spectra; and direct DFT relaxations of shaken fcc supercells in the SI also find the P-43m static minimum. The TiN/NbC controls are a good idea, the simulated XRD pattern gives experiment a concrete target, and the citation pattern looks healthy, including a genuine search of the diffraction literature.\n\nWhere I part ways with the paper's confidence is the 'three independent methods' framing. The AIRSS confirmation is partly circular: P-43m was added to the EDDP training set as a marker. The shaken-fcc DFT searches do support the static structure, so the structure itself is not in doubt, but they do not confirm the 65 meV/atom ordering.\n\nThe bigger issue is the MLIP. The MTP was trained on 100 DFT configurations around delta at 300 K. The validation in Fig. S8 takes 100 structures from the final, MTP-generated SSCHA population at 15 K. That checks self-consistency on the model's own sampling distribution; it cannot catch a systematic bias that keeps the SSCHA trajectories out of the true basin or distorts the curvature near the P-43m minimum. An RMSE of 0.6-0.7 meV/atom on that set is reassuring but not conclusive, because the free-energy difference is accumulated over thousands of correlated configurations. A couple of meV/atom bias would not show up as a large RMSE.\n\nTwo smaller flags. The MD-SED simulation uses a lattice constant of 4.482 Å while the DFT P-43m value is 4.421 Å; that 1.4% mismatch is not negligible for phonons, and the SED spectra are not a quantitative free-energy check. The harmonic P-43m phonons still have imaginary modes near Gamma, so no harmonic cross-check exists; that is consistent with the zero-point-stabilization story, but it makes the phase wholly dependent on the anharmonic treatment. The Tc of 20 K is an overestimate against the experimental 16 K, though clearly better than the 27 K from large-smearing delta.\n\nWho this is for: the superconductivity and materials-discovery community. I would cite it as a predicted phase, not as an established ground state. Send it to peer review. In review I would ask for one direct DFT-level SSCHA or path-integral MD calculation on the P-43m basin, or at least a DFT reweighting of the free-energy difference, plus a reconciliation of the MD-SED lattice parameter. If that check holds, this is a significant result.","headline":"A genuinely new predicted P-43m phase in NbN with careful superconductivity work, but the 65 meV/atom ordering and the quantum-stabilization narrative rest on one MLIP and need a direct DFT check before I would call it established.","tokens_in":17764,"tokens_out":4350,"would_cite":true,"duration_ms":50706,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["74.20.-z","74.25.Kc"],"model":"deepseek-v4-flash","headline":"Fully relaxed under quantum anharmonic motion, stoichiometric NbN distorts into a previously unreported cubic phase (P-43m) that is 65 meV/atom lower in free energy than ideal δ-NbN, with a predicted superconducting transition of 20 K.","keywords":["niobium nitride","superconductivity","quantum anharmonicity","zero-point motion","P-43m cubic phase","stochastic self-consistent harmonic approximation","machine-learned interatomic potentials","Migdal-Eliashberg theory"],"falsifier":"A direct DFT-only SSCHA relaxation of the eight-atom $P\\bar{4}3m$ cell would settle the ordering: if the 65 meV/atom stabilisation over $\\delta$-NbN shrinks or reverses without the machine-learned potential, the phase is that potential's artefact. Experimentally, high-resolution X-ray or neutron diffraction on a near-stoichiometric, vacancy-controlled cubic NbN sample would test the structure directly: $P\\bar{4}3m$ predicts weak superstructure reflections at the 100 and 110 positions, and the paper notes that only a faint 110 trace (and no 100 peak) has ever been reported.","tokens_in":16694,"feed_emoji":"⚛️","tokens_out":11322,"duration_ms":117774,"temperature":0.7,"pith_summary":"Niobium nitride (NbN) is a central material for superconducting applications, yet its standard cubic phase cannot exist at perfect 1:1 stoichiometry: the ideal rock-salt lattice is dynamically unstable, and every experimental sample requires nitrogen vacancies to hold it together. This paper argues that the missing ingredient in that picture is quantum anharmonicity — the zero-point motion of the ions and the non-parabolic shape of the potential they move in. When the lattice is fully relaxed with those effects included, the crystal distorts along its softest mode into a previously unreported cubic phase with space group $P\\bar{4}3m$, 65 meV/atom lower in free energy than ideal $\\delta$-NbN, and three independent computational methods converge on the same structure. The phase's predicted transition temperature of 20 K matches the measured value for near-stoichiometric NbN far better than earlier stabilization tricks, which gave 27 K. If the claim holds, it overturns the assumption that vacancies are required for cubic NbN and points to a vacancy-free 1:1 phase as a possible route to higher $T_c$.","feed_headline":"Quantum motion stabilizes a new cubic phase of NbN","feed_subtitle":"A distorted, vacancy-free cubic structure sits 65 meV per atom below δ-NbN and gives a 20 K transition.","key_machinery":"The mechanism that carries the argument is quantum anharmonicity — the combined effect of zero-point ionic motion and the non-parabolic shape of the potential energy surface — treated by letting the crystal relax inside the stochastic self-consistent harmonic approximation (SSCHA), which replaces the true ionic potential with an effective harmonic one while keeping the quantum fluctuations explicit. The pivotal object is the distortion itself: the $P\\bar{4}3m$ arrangement is a linear combination of the eigenvectors of a threefold-degenerate zone-centre mode with $T_{1u}$ symmetry, the mode with the largest imaginary harmonic frequency (about $7.4i$ THz) in $\\delta$-NbN. Because converged SSCHA and MD runs need hundreds of thousands of energy evaluations, a machine-learned interatomic potential trained on 100 DFT configurations of $\\delta$-NbN supplies energies and forces with an RMSE of 0.6–0.7 meV/atom, accelerating the calculations by $10^4$–$10^5$. Agreement between SSCHA, MD spectral energy density, and the independent structure search is what turns the new phase from a single-method artefact into a robust claim.","core_discovery":"The paper's claim, stated on its own terms, is that 1:1 stoichiometric cubic niobium nitride has a previously unrecognised quantum-mechanical ground state: the ideal rock-salt phase $Fm\\bar{3}m$ is dynamically unstable even when quantum anharmonicity is included (SSCHA retains imaginary modes up to 300 K), but if the lattice is allowed to displace along the soft modes it settles into a distorted, non-centrosymmetric cubic structure with space group $P\\bar{4}3m$ (Nb and N on 4e sites, $a = 4.421$ Å) whose free energy is 65 meV/atom below $\\delta$-NbN. The same structure is reached by three independent routes — SSCHA full relaxation, molecular-dynamics spectral energy density, and a blind ab initio random structure search — and it still lies above the hexagonal $\\epsilon$-NbN ground state. The distortion removes bands from the Fermi surface, lowering the electronic density of states and the electron-phonon coupling, so the Migdal-Eliashberg calculation with a GW-derived Coulomb pseudopotential gives $T_c = 20$ K, close to the experimental value of about 16 K for near-stoichiometric material and far better than the 27 K obtained by artificially smearing the Fermi surface. The conclusion is that nitrogen vacancies are not necessary to stabilise cubic NbN; a vacancy-free distorted cubic phase is thermodynamically preferred and may be a route to higher $T_c$.","pith_inferences":["If the $P\\bar{4}3m$ phase is real, it has a cheap experimental fingerprint: weak superstructure reflections at the 100 and 110 positions, and the paper itself notes that only a faint 110 trace (with no 100 peak) has ever been reported in high-pressure data; re-examining vacancy-controlled cubic samples with high-resolution diffraction could test the phase without new synthesis routes.","Because the relaxed cell has only eight atoms, a direct DFT-only SSCHA recheck of the $P\\bar{4}3m$ basin is a plausible near-term test that would remove the machine-learning layer from the argument and settle the 65 meV/atom ordering.","The same full-relaxation protocol could be applied to other transition-metal nitrides and carbides with harmonic instabilities (for example TaN or MoN variants); the authors' TiN and NbC negative controls suggest the mechanism is not universal, so each material would need its own test.","The non-centrosymmetric $P\\bar{4}3m$ structure lacks inversion symmetry, so if it can be stabilized in films it may display effects the centrosymmetric $\\delta$ phase cannot, such as spin-orbit splitting of electronic states; the paper does not explore these."],"forward_implications":["The vacancy-free $P\\bar{4}3m$ phase is thermodynamically preferred over ideal $\\delta$-NbN by 65 meV/atom, so synthesis routes targeting perfect 1:1 stoichiometry should aim at this distorted cubic structure rather than at a vacancy-stabilized rock-salt lattice.","The $\\delta$ phase remains dynamically unstable at perfect stoichiometry even with quantum anharmonicity (SSCHA imaginary modes up to 300 K), so the ideal rock-salt lattice cannot be the ground state of 1:1 NbN by itself.","A $T_c$ of 20 K computed with a GW-derived Coulomb pseudopotential is much closer to the experimental ~16 K of near-stoichiometric NbN than the 27 K from large-smearing harmonic treatments, implying that the smearing approximation overestimates electron-phonon coupling in this system.","Because measured $T_c$ rises as nitrogen vacancies are removed, the predicted vacancy-free endpoint could superconduct at a higher temperature than any NbN sample studied so far.","The identical full-relaxation treatment of TiN and NbC finds no such distortion, so the $P\\bar{4}3m$ instability is specific to NbN rather than a generic feature of the method."],"supporting_citations":[{"why":"Supplies the SSCHA framework and code used to relax the lattice with quantum anharmonicity, the step that produces the P-43m structure.","marker":"[29]"},{"why":"Provides the molecular-dynamics spectral energy density (SED) method whose anharmonic phonons independently confirm the distorted structure.","marker":"[30, 31]"},{"why":"Defines the ab initio random structure search (AIRSS) that finds the same P-43m phase among roughly 50,000 candidate structures.","marker":"[32, 33]"},{"why":"Supplies the workflow for converging SSCHA calculations with machine-learned potentials, the basis for the central relaxation run.","marker":"[38]"},{"why":"Introduces the ephemeral data-derived potentials (EDDPs) that make the large AIRSS search computationally feasible.","marker":"[43]"},{"why":"The package used to train the machine-learned interatomic potential (MTP) on 100 DFT configurations of delta-NbN.","marker":"[78]"},{"why":"Provides the experimental mapping of Tc against nitrogen vacancy content that anchors the claim that vacancy-free 1:1 NbN should superconduct near 16-17 K.","marker":"[7]"},{"why":"Neutron diffraction data on NbN used to benchmark the large-smearing harmonic phonons that the paper argues are misleading.","marker":"[36]"},{"why":"The IsoME Migdal-Eliashberg solver used to compute the superconducting gap and Tc = 20 K for the P-43m phase.","marker":"[51]"},{"why":"The GW calculation that gives the screened Coulomb pseudopotential mu = 0.21 used in the Eliashberg equations.","marker":"[54]"}],"fun_headline_variants":["Quantum anharmonicity unlocks stable vacancy-free NbN","NbN defies vacancy dogma: new cubic phase superconducts at 20 K","No vacancies needed: quantum distortion gives NbN a 20 K boost","Quantum motion stabilizes a vacancy-free NbN superconductor"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the machine-learned potential, trained on only 100 DFT snapshots of the unstable $\\delta$ phase at 300 K, stays accurate in the fully relaxed $P\\bar{4}3m$ basin, including the 65 meV/atom free-energy ordering, even though no direct DFT recheck of the SSCHA-relaxed $P\\bar{4}3m$ free energy or phonons is reported.","fun_headline_variants_meta":{"raw":{"variants":["Quantum anharmonicity unlocks stable vacancy-free NbN","NbN defies vacancy dogma: new cubic phase superconducts at 20 K","No vacancies needed: quantum distortion gives NbN a 20 K boost","Quantum motion stabilizes a vacancy-free NbN superconductor"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000947,"raw_usage":{"total_tokens":4123,"prompt_tokens":1108,"completion_tokens":3015,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":724,"completion_tokens_details":{"reasoning_tokens":2940}},"tokens_in":724,"tokens_out":3015,"duration_ms":26717,"temperature":1.0,"reasoning_tokens":2940,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:11:01.536265+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct DFT-only SSCHA relaxation of the eight-atom $P\\bar{4}3m$ cell would settle the ordering: if the 65 meV/atom stabilisation over $\\delta$-NbN shrinks or reverses without the machine-learned potential, the phase is that potential's artefact. Experimentally, high-resolution X-ray or neutron diffraction on a near-stoichiometric, vacancy-controlled cubic NbN sample would test the structure directly: $P\\bar{4}3m$ predicts weak superstructure reflections at the 100 and 110 positions, and the paper notes that only a faint 110 trace (and no 100 peak) has ever been reported.","supporting_citations":[{"cited_title":"Lucrezi , author E","cited_arxiv_id":null,"evidence_quote":"Supplies the workflow for converging SSCHA calculations with machine-learned potentials, the basis for the central relaxation run."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the ephemeral data-derived potentials (EDDPs) that make the large AIRSS search computationally feasible."},{"cited_title":"Oya \\ and\\ author Y","cited_arxiv_id":null,"evidence_quote":"Provides the experimental mapping of Tc against nitrogen vacancy content that anchors the claim that vacancy-free 1:1 NbN should superconduct near 16-17 K."},{"cited_title":"Christensen , author O","cited_arxiv_id":null,"evidence_quote":"Neutron diffraction data on NbN used to benchmark the large-smearing harmonic phonons that the paper argues are misleading."},{"cited_title":"IsoME: Streamlining High-Precision Eliashberg Calculations","cited_arxiv_id":"2503.03559","evidence_quote":"The IsoME Migdal-Eliashberg solver used to compute the superconducting gap and Tc = 20 K for the P-43m phase."},{"cited_title":"Lambert \\ and\\ author F","cited_arxiv_id":null,"evidence_quote":"The GW calculation that gives the screened Coulomb pseudopotential mu = 0.21 used in the Eliashberg equations."}],"review_version":1}