{"id":"68289503-0ca3-4143-a6ef-b2df36d2a3e2","arxiv_id":"2501.16280","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"QDET, DMET, and spin-flip TDDFT calculations reproduce the measured 1.30 eV emission of the iron impurity in aluminum nitride, and the paper defines a convergence protocol for these strongly correlated defect states.","lead":"Researchers tested three quantum chemistry methods on an iron atom sitting in aluminum nitride, a proposed material for quantum computers. The methods reproduce the measured light emission of this defect and the paper gives a recipe for running such calculations reliably.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Concern: the claimed 0.1–0.2 eV agreement with the 1.30 eV ZPL depends on subtracting TDDFT Franck–Condon shifts from vertical energies, but the method used for that subtraction (SF-TDDFT) is also the method that produces the PL line shape; an independent check of the excited-state geometry is…","rationale":"I largely agree with the reader's assessment of the weakest assumption, but I push it one step further. The reader identifies two external inputs: the assignment of the 1.30 eV ZPL and the SF-TDDFT Franck–Condon shifts. Both are indeed load-bearing, but the paper has stronger support for the experimental assignment than the reader suggests: the SF-TDDFT/HSE PL lineshape in Fig. 5(b) matches the experimental spectrum well, and the paper notes that the PBE functional gives larger Huang–Rhys factors and poorer agreement. This lineshape agreement is a genuine success of SF-TDDFT and partially validates both the assignment and the excited-state geometry calculation. However, the central quantitative claim of the paper—that QDET and DMET reproduce the ZPL within 0.1–0.2 eV—still rests on the SF-TDDFT geometry via the Franck–Condon subtraction. Because the same method is used to produce the PL spectrum and the shift, the agreement with experiment is not an independent validation of the embedding calculations. The paper is otherwise transparent and the convergence studies are careful. Therefore I recommend CONDITIONAL acceptance: the paper should provide an independent verification of the excited-state Franck–Condon shift (or at least a sensitivity analysis of Table 1 to the shift) before the central claim is taken at face value. This is a check the authors could perform with existing tools, and it would settle whether the 0.1–0.2 eV agreement is real or an artifact of the geometry choice.","tokens_in":18042,"tokens_out":2381,"duration_ms":20385,"concrete_test":"Recompute the adiabatic excitation energy of the 4E state using an independent excited-state geometry for FeAl, e.g., optimize the 4E geometry with CASSCF/NEVPT2 or state-averaged CASSCF for the embedded cluster (as in the pDMET framework) or with a different functional/kernel in SF-TDDFT, and recompute the Franck–Condon shift. If the resulting shift changes by more than about 0.05 eV, the AEEs in Table 1, and the claimed 0.1–0.2 eV agreement with the experimental ZPL, must be revised accordingly.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central validation of the embedding methods is the comparison with experiment in Table 1. The table converts computed vertical excitation energies to adiabatic excitation energies by subtracting a Franck–Condon shift (0.08–0.11 eV) obtained exclusively from SF-TDDFT geometry relaxations of the 4E state. Thus the agreement of QDET/NEVPT2-DMET with the 1.30 eV ZPL is contingent on SF-TDDFT correctly describing the excited-state potential energy surface and its curvature. This is not an independent confirmation: the same SF-TDDFT calculation is used to generate the photoluminescence spectrum in Fig. 5(b), which is then compared with experiment to support the assignment and the geometry. If the SF-TDDFT excited-state geometry is wrong (e.g., because of incorrect treatment of the near-degenerate quartet manifold, which the paper itself identifies as the reason ∆-SCF fails), then every AEE in Table 1 shifts by the same amount and the claimed 0.1–0.2 eV agreement is not evidence for the accuracy of the embedding methods. The paper contains no benchmark of SF-TDDFT for FeAl against an independent method for the excited-state geometry; e.g., no comparison with a geometry obtained from a multireference method or from an experimental isotope shift / Huang–Rhys factor independent of SF-TDDFT. The ZPL comparison therefore rests on an unverified assumption about the excited-state geometry.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the neutral iron substitutional impurity (FeAl) in wurtzite AlN as a test case for strongly correlated transition-metal defects in insulators. Using two quantum embedding methods, QDET and DMET (with CASSCF and NEVPT2 solvers), and spin-flip TDDFT, the authors compute the 6A1 ground state and the lowest 4E excited state. They report vertical excitation energies with systematic supercell-size and active-space convergence tests, convert them to adiabatic excitation energies using Franck-Condon shifts from spin-flip TDDFT, and compare with the experimental zero-phonon line at 1.30 eV. They further compute vibrationally resolved photoluminescence spectra and show that spin-flip TDDFT with the HSE functional reproduces the measured line shape, while spin-polarized DFT does not. The central claim is that the embedding methods, when properly converged and with double-counting corrections, reliably describe this correlated defect, and that spin-flip TDDFT provides accurate excited-state geometries and spectra.","tokens_in":18299,"tokens_out":5184,"duration_ms":48314,"significance":"If the results hold, the paper provides a valuable benchmark for quantum embedding methods on a challenging system where cRPA-based approaches had previously struggled. The systematic active-space protocol for QDET and the identification of the need for larger active spaces in DMET are useful practical guidance for defect calculations. The paper's strengths include the absence of any parameter fitted to the defect's experimental excitation energy, explicit convergence checks over supercell size and active-space size, and direct comparison with an external experimental zero-phonon line and photoluminescence spectrum. The open-source implementations (pDMET and WEST) are also a plus for reproducibility. The experimental agreement within roughly 0.1-0.2 eV for NEVPT2-DMET, QDET, and hybrid TDDFT, together with the successful PL line shape, would support the usefulness of these methods for transition-metal defects in wide-gap insulators.","major_comments":[{"comment":"The abstract claims that 'both DMET and QDET accurately describe the ground state and low-lying excited states,' but Table 1 shows CAS-DMET(15e,15o) giving an adiabatic excitation energy of 1.94 eV, which is 0.64 eV above the experimental ZPL and outside the stated 0.1-0.2 eV accuracy. Only NEVPT2-DMET is accurate. The claim should be qualified to 'DMET with a post-CASSCF correction' or 'NEVPT2-DMET,' and the abstract should not imply that the bare CASSCF-based DMET used here is accurate for this defect.","section":"Abstract; Table 1"},{"comment":"The construction of the adiabatic excitation energies is not fully transparent. The text says that Franck-Condon shifts from TDDFT@PBE (0.11 eV) are used, but it also reports HSE and DDH shifts of 0.08 eV, and Table 1 lists EFC,ES only for the PBE-TDDFT row. To make the comparison reproducible, the table should report the vertical excitation energy and the functional-specific Franck-Condon shift used for every method, and the text should explicitly state whether a common shift was applied to the QDET/NEVPT2-DMET entries or whether functional-specific shifts were used.","section":"Table 1 and text near 'From the results reported in Table 1'"},{"comment":"The accuracy of every AEE in Table 1 depends on the spin-flip TDDFT excited-state geometry, since the Franck-Condon shift is obtained from that geometry. The paper should explicitly state that the agreement of the computed PL line shape with experiment in Fig. 5(b), which is governed by Huang-Rhys factors and phonon energies, provides an independent experimental check of that geometry. It should also discuss the sensitivity of the result to the functional choice: PBE gives a shift of 0.11 eV while HSE and DDH give 0.08 eV, a spread of 0.03 eV that is comparable to the claimed accuracy of the embedding methods.","section":"Figure 5(b); section 'To assess the accuracy...'"}],"minor_comments":[{"comment":"The comparison in Table 1 relies on assigning the 1.30 eV zero-phonon line to the 4E -> 6A1 transition of the isolated neutral FeAl defect. The paper cites Ref. 25 for the measurement but does not discuss the evidence for this assignment; a brief justification or a citation to a review that establishes the assignment would strengthen the validation.","section":"Introduction / assignment of experimental ZPL"},{"comment":"The caption states that the computed spectra are horizontally shifted to align the ZPL with experiment. Please state explicitly that this comparison validates the line shape and Huang-Rhys factors, not the absolute transition energy.","section":"Figure 5(b) caption"},{"comment":"The statement 'TDDFT yields photoluminescence spectra in agreement with experiments' should be qualified to 'spin-flip TDDFT with the HSE or DDH functional,' since the Supporting Information shows that PBE gives a significantly larger Huang-Rhys factor (S_total = 5.73 vs 3.24 for HSE) and a correspondingly poorer line shape.","section":"Abstract"},{"comment":"The columns EFC,GS and EFC,ES are populated only for the PBE-TDDFT row. For readability, the caption should explain that these columns apply to the row shown, or the values should be listed for each functional family.","section":"Table 1"},{"comment":"The description of the QDET active spaces as '(7e,6o), (19e,12o), and (51e,28o)' is clear, but the reader would benefit from a sentence explaining why adding occupied valence bands increases the electron count by the stated amounts; the current text implies this without giving the counting rule.","section":"Section 'We now turn to exploring larger active spaces'"}],"recommendation":"minor_revision","confidential_remarks":"The paper is a solid benchmark application and the central comparison to experiment is mostly convincing. The two issues I would press in revision are (i) sharpening the abstract's DMET claim in light of the CAS-DMET outlier and (ii) making the Franck-Condon-shift protocol in Table 1 fully explicit. The reliance on spin-flip TDDFT for the excited-state geometry is real, but the successful PL line shape agreement with experiment provides a strong external check, so I do not regard it as a blocking flaw. The paper fits the journal's scope well; the active-space convergence protocol is a useful contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the useful thing: this is a carefully executed benchmark showing that QDET and DMET can handle a transition-metal impurity in an insulator, which is harder than the vacancy centers these methods were previously tested on. The active-space convergence study is the real contribution: QDET converges with the minimal (5e,5o) space plus a few valence bands, while DMET needs a (15e,15o) space to get within shouting distance, and the paper says so plainly. That is a nontrivial, reproducible insight. The supercell-size tests, the extrapolation to the non-embedding limit for DMET, and the open code (WEST, pDMET) all back the claims. The spin-flip TDDFT photoluminescence spectrum matching the 1994 experiment, with the symmetry-breaking e-mode analysis, is also good work.\n\nThe soft spots are real but not load-bearing. The adiabatic excitation energies in Table 1 are all obtained by subtracting SF-TDDFT Franck-Condon shifts from vertical energies, and the same SF-TDDFT method produces the PL lineshape that is used to validate the geometry. The stress-test worry about circularity is partially answered: the PL lineshape comparison is an independent experimental check on the excited-state geometry, and it is not a trivial one. But nobody has benchmarked SF-TDDFT's excited-state geometry for FeAl against a multireference geometry or an experimentally derived Huang-Rhys factor. If that geometry were off, every AEE in Table 1 would shift by the same amount, and the 0.1-0.2 eV agreement would be partly an artifact. So a serious referee should ask for that check, but I would not sink the paper on it.\n\nSecond soft spot: the whole comparison rests on the assignment of the 1.30 eV ZPL to the 4E -> 6A1 transition of neutral FeAl from Baur et al. 1994. That is the standard reading, but it is a single measurement, and the paper does not hedge about it. Minor issues: the abstract says 'low-lying excited states' plural, but only the lowest 4E state is computed at the correlated level; and the supporting information is referenced but not deposited as a data file, so reproducibility is partial.\n\nBottom line: this is a competent, honest methods paper. It deserves a serious referee. I would recommend accept after the referee asks for an independent check of the SF-TDDFT geometry, or at least a sensitivity analysis with a different FC shift estimate, plus the input files. The central claim—that embedding methods with proper double counting can handle this class of defects—holds up.","headline":"A solid, honest benchmark showing QDET and DMET can treat a TM impurity in AlN; the validation rests heavily on SF-TDDFT for the excited-state geometry, which a referee should ask to be checked independently.","tokens_in":18895,"tokens_out":2475,"would_cite":true,"duration_ms":23824,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.55.-i","71.15.-m","78.55.-m"],"model":"deepseek-v4-flash","headline":"Quantum embedding methods DMET and QDET accurately describe the strongly correlated ground and excited states of an iron impurity in aluminum nitride, and spin-flip TDDFT reproduces its measured photoluminescence spectrum.","keywords":["iron impurity","aluminum nitride","spin defects","quantum embedding","density matrix embedding theory","quantum defect embedding theory","spin-flip time-dependent density functional theory","photoluminescence"],"falsifier":"Find the 1.30 eV line in a sample with independently known FeAl concentration and charge state and show it belongs to another center, or compute the $^4E$ excited-state geometry with a multireference gradient method and obtain a Franck-Condon shift outside 0.08--0.11 eV; either result would move the adiabatic energies away from the measured zero-phonon line and break the claimed agreement.","tokens_in":17804,"feed_emoji":"⚛️","tokens_out":9367,"duration_ms":82252,"temperature":0.7,"pith_summary":"Defect spin qubits in wide-gap materials live or die on their excited-state physics, and transition metal impurities are the hardest cases because their $d$-electron states are strongly correlated. This paper takes iron substituting aluminum in aluminum nitride as a benchmark and asks whether two quantum embedding methods, DMET and QDET, can describe both the ground state and the lowest excited state quantitatively. It claims they can: with carefully chosen active spaces and correct double-counting corrections, both methods give adiabatic excitation energies for the lowest $^4E$ state within roughly 0.1--0.2 eV of the measured 1.30 eV zero-phonon line. The paper also claims that spin-flip TDDFT, used for excited-state geometries, reproduces the measured photoluminescence spectrum in detail. A sympathetic reader would take the paper to be establishing that a practical, convergent protocol now exists for computing strongly correlated spin-defect excitations from first principles.","feed_headline":"Two embedding methods bracket iron's 1.30 eV line in AlN","feed_subtitle":"DMET and QDET place the lowest quartet state at 1.0–1.5 eV, and spin-flip TDDFT matches the measured photoluminescence shape.","key_machinery":"The load-bearing object is an effective Hamiltonian for a small active space of defect orbitals, $H_{\\text{eff}} = \\sum_{ij} t^{\\text{eff}}_{ij} a^\\dagger_i a_j + \\frac{1}{2} \\sum_{ijkl} v^{\\text{eff}}_{ijkl} a^\\dagger_i a^\\dagger_j a_l a_k$, diagonalized at high level while embedded in a lower-level description of the host. QDET (a Green's-function-based embedding method) builds this Hamiltonian from a $G_0W_0$ calculation and solves it with full configuration interaction, using an exact double-counting correction instead of an approximate one; DMET (a wave-function-based embedding method) builds a fragment-plus-bath space from a restricted open-shell Hartree-Fock wave function through a Schmidt decomposition and solves it with CASSCF or NEVPT2. The active-space protocol carries the argument: start from the five Fe $3d$ orbitals, then add occupied states nearest the valence band maximum for QDET, or Fe--N bonding and virtual $d$ orbitals for DMET, until excitation energies converge. Spin-flip TDDFT supplies the missing piece the embedding methods cannot provide, namely excited-state geometries, and those geometries yield the Franck-Condon shifts and Huang-Rhys spectral densities used to compare with the measured zero-phonon line and photoluminescence shape.","core_discovery":"The paper's central claim is that quantum embedding methods can handle the strongly correlated physics of a transition metal impurity in an insulator, a regime where simpler treatments fail. For iron substituting aluminum in wurtzite AlN, both DMET and QDET reproduce the high-spin $^6A_1$ ground state and place the lowest excited quartet $^4E$ at adiabatic excitation energies of about 1.0--1.5 eV, bracketing the measured zero-phonon line at 1.30 eV. The paper asserts that this agreement is not accidental: QDET needs only the five Fe $3d$ orbitals plus a few states near the valence band maximum, while DMET needs a larger active space including Fe--N bonding orbitals, and both are stable with respect to supercell size and functional choice. It further claims that spin-flip TDDFT correctly assigns the emitting state as $^4E$ and reproduces the experimental photoluminescence line shape, whereas spin-polarized DFT does not, because it cannot describe the multi-determinant nature and symmetry-breaking relaxation of the quartet.","pith_inferences":["The paper leaves implicit that its QDET recipe, starting from the five Fe $3d$ orbitals and adding states near the valence band maximum, likely transfers to other $3d$ impurities in nitride and oxide hosts, while the DMET space will depend on local bonding.","The 0.08--0.11 eV Franck-Condon correction is supplied entirely by spin-flip TDDFT; benchmarking it against a correlated excited-state geometry for a similar $^4E$ defect is the most direct test of the paper's error budget.","A screening workflow suggested but not developed here would pair QDET vertical energies with spin-flip TDDFT geometries and line shapes, reserving DMET for cases needing a wave-function embedding analysis.","Because only one measured zero-phonon line anchors the comparison, additional optical data, for example from isotopically purified or co-doped samples, would test the calculated quartet manifold more strictly than the present single line."],"forward_implications":["Transition metal impurities in insulators, previously flagged as a difficult case for embedding approaches such as cRPA, become tractable when the active space is grown systematically and double counting is handled exactly.","QDET needs a minimal or near-minimal active space for this defect, while DMET needs a larger space including Fe--N bonding orbitals; the paper's protocol tells practitioners how to converge both.","Spin-flip TDDFT can supply reliable excited-state geometries and photoluminescence line shapes for strongly correlated spin defects, while spin-polarized DFT cannot, even when its excitation energy looks acceptable.","Adiabatic excitation energies suitable for comparison with zero-phonon-line experiments can be assembled from embedding vertical energies plus TDDFT Franck-Condon shifts, with errors of order 0.1--0.2 eV.","The same division of labor, embedding for vertical energies and spin-flip TDDFT for relaxation, can be applied to other qubit-candidate defects in wide-gap hosts."],"supporting_citations":[{"why":"Supplies the experimental zero-phonon line at 1.30 eV and the photoluminescence spectrum used as the accuracy benchmark.","marker":"25"},{"why":"The cRPA study that found the FeAl defect especially sensitive to functional and double-counting choices; this is the challenge the paper addresses.","marker":"16"},{"why":"Provides the Green's function formulation of QDET and the exact double-counting correction used for the QDET calculations.","marker":"7"},{"why":"Introduces the quantum embedding theory for strongly correlated states and the active-space selection procedure the paper follows.","marker":"15"},{"why":"In-preparation work supplying the new double-counting formalism and hybridization treatment in QDET.","marker":"35"},{"why":"Develops the spin-flip TDDFT implementation used here for excited-state geometries, Franck-Condon shifts, and photoluminescence spectra.","marker":"12"},{"why":"Establishes the periodic DMET implementation with restricted open-shell Hartree-Fock and multireference solvers used for the DMET calculations.","marker":"46"},{"why":"Demonstrates the DMET approach to optical properties of defect states and the extrapolation procedure extended in this paper.","marker":"11"},{"why":"Provides the Huang-Rhys theory framework used to compute vibrationally resolved photoluminescence line shapes.","marker":"76"},{"why":"Validates the first-principles calculation of photoluminescence spectra for point defects, the method used for the line-shape comparison.","marker":"77"}],"fun_headline_variants":["Iron in AlN: two embedding methods nail the 1.30 eV line","Quantum embedding captures iron's strongly correlated state in AlN","Spin-defect physics solved: DMET and QDET bracket iron's line in AlN","New protocol for transition metal defects: iron in AlN benchmarked","Embedding methods tame iron impurity's correlated states in AlN"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The accuracy claim assumes the measured 1.30 eV spectral line is the lowest excited quartet to ground sextet transition of the isolated neutral FeAl defect, and that the spin-flip TDDFT excited-state geometry used for the vibrational correction is correct.","fun_headline_variants_meta":{"raw":{"variants":["Iron in AlN: two embedding methods nail the 1.30 eV line","Quantum embedding captures iron's strongly correlated state in AlN","Spin-defect physics solved: DMET and QDET bracket iron's line in AlN","New protocol for transition metal defects: iron in AlN benchmarked","Embedding methods tame iron impurity's correlated states in AlN"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000195,"raw_usage":{"total_tokens":1349,"prompt_tokens":932,"completion_tokens":417,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":548,"completion_tokens_details":{"reasoning_tokens":320}},"tokens_in":548,"tokens_out":417,"duration_ms":4191,"temperature":1.0,"reasoning_tokens":320,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T13:34:39.128466+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find the 1.30 eV line in a sample with independently known FeAl concentration and charge state and show it belongs to another center, or compute the $^4E$ excited-state geometry with a multireference gradient method and obtain a Franck-Condon shift outside 0.08--0.11 eV; either result would move the adiabatic energies away from the measured zero-phonon line and break the claimed agreement.","supporting_citations":[{"cited_title":"Determination of the GaN/AlN band offset via the (‐/0) acceptor level of iron","cited_arxiv_id":null,"evidence_quote":"Supplies the experimental zero-phonon line at 1.30 eV and the photoluminescence spectrum used as the accuracy benchmark."},{"cited_title":"I.; Hampel, A.; Cano, J.; R\\\"osner, M.; Dreyer, C","cited_arxiv_id":null,"evidence_quote":"The cRPA study that found the FeAl defect especially sensitive to functional and double-counting choices; this is the challenge the paper addresses."},{"cited_title":"Green's function formulation of quantum defect embedding theory","cited_arxiv_id":null,"evidence_quote":"Provides the Green's function formulation of QDET and the exact double-counting correction used for the QDET calculations."},{"cited_title":"Quantum embedding theory for strongly correlated states in materials","cited_arxiv_id":null,"evidence_quote":"Introduces the quantum embedding theory for strongly correlated states and the active-space selection procedure the paper follows."},{"cited_title":"W.-z.; Jin, Y.; Govoni, M.; Galli, G","cited_arxiv_id":null,"evidence_quote":"In-preparation work supplying the new double-counting formalism and hybridization treatment in QDET."},{"cited_title":"W.-z.; Govoni, M.; Xu, A","cited_arxiv_id":null,"evidence_quote":"Develops the spin-flip TDDFT implementation used here for excited-state geometries, Franck-Condon shifts, and photoluminescence spectra."},{"cited_title":"Q.; Hermes, M","cited_arxiv_id":null,"evidence_quote":"Establishes the periodic DMET implementation with restricted open-shell Hartree-Fock and multireference solvers used for the DMET calculations."},{"cited_title":"R.; Galli, G.; Gagliardi, L","cited_arxiv_id":null,"evidence_quote":"Demonstrates the DMET approach to optical properties of defect states and the extrapolation procedure extended in this paper."},{"cited_title":"B.; Awschalom, D","cited_arxiv_id":null,"evidence_quote":"Provides the Huang-Rhys theory framework used to compute vibrationally resolved photoluminescence line shapes."}],"review_version":1}