{"id":"88380c01-82b6-493f-962f-8de36f7b8a26","arxiv_id":"2607.17537","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"First-coordination-sphere multireference density matrix embedding reproduces CASSCF spin-phonon relaxation rates within an order of magnitude for five molecular magnets and, for the first time, enables such rates for a molecular crystal.","lead":"An embedding method treats only the magnetic atom's nearest neighbors with expensive quantum chemistry while handling the rest cheaply, and still reproduces full-calculation spin relaxation rates within about a factor of ten. It also enables the first multireference spin-phonon relaxation calculation for a molecular crystal—a step toward designing spin qubits and molecular magnets in realistic materials.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing step-size/error analysis for the finite-difference CFP derivatives leaves the one periodic result—Crystal-I—without a reference check; this is the weakest load-bearing point.","rationale":"The reader's weakest assumption and mine coincide: the finite-difference CFP derivatives are the unexamined link between the embedding electronic structure and the relaxation rates. The paper is an honest method demonstration with independent benchmarks—non-embedded CASSCF for molecules and experimental rates—so there is no circularity. Equilibrium KD energies are not sensitive to derivative errors, and rates depend on derivatives squared, making this genuinely load-bearing. However, the molecular parity plots and RMSEs provide some direct evidence that the derivatives are reasonable, and the stated accuracy target (within one order of magnitude) is forgiving. The periodic Crystal-I case is the most exposed because it has no non-embedded reference and the environment effect is small, so the periodic demonstration would not detect a displacement-dependent embedding bias. This supports keeping the verdict CONDITIONAL rather than moving to ACCEPT or REJECT; the concern is real but addressable by a straightforward convergence check.","tokens_in":14947,"tokens_out":4137,"duration_ms":40007,"concrete_test":"Recompute Complex-I (and, if feasible, Crystal-I) spin-phonon rates using displacement steps δ = 0.005, 0.01, 0.02, and 0.04 Å with otherwise identical electronic-structure and phonon settings. Compare ∂B_l^m/∂R_i and τ(T) over 2–100 K. If τ changes by more than a factor of 2 across δ, the derivative protocol is not converged. Then run Crystal-I at δ = 0.005 and 0.02 Å; if the periodic ∂B_l^m/∂R_i changes by more than ~20% between steps, the one periodic result is numerically under-determined and the claim of periodic rates should be downgraded.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central comparison for the molecular complexes is internally solid: parity plots of ∂B_l^m/∂R_i and τ(T) against non-embedded CASSCF are genuine independent benchmarks, and the Discussion honestly limits the claim to weakly correlated systems. The load-bearing weakness is the derivative step. Equations (6) and (8) make all rates quadratic in the spin-phonon couplings, which in turn are built from ∂B_l^m/∂R_i via Eq. (10). The Computational Methods section reports only ±0.01/0.02 Å finite differences for molecules and ±0.01 Å for the crystal, with no step-size convergence study, no estimate of fitting noise in extracting B_l^m from the SOC Hamiltonian, and no error propagation into τ. For the molecular complexes, the parity plots indirectly bound derivative errors, so this is an accuracy limitation rather than a fatal flaw. For Crystal-I, however, there is no non-embedded reference; the statement that the periodic derivatives are 'smooth' (Results, Molecular Crystal) is not a check of accuracy. Moreover, the crystal environment barely changes the rates (Fig. 10), so this single periodic demonstration does not test the embedding in a regime where the environment matters. A systematic DMET bias that grew with nuclear displacement—through the embedding potential or bath localization—would corrupt Orbach and Raman rates while leaving equilibrium KD energies looking reasonable.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an embedded multireference workflow for computing spin–phonon relaxation rates, combining CAS-DMET with state-interaction SO coupling and open quantum system theory. The method is benchmarked on three Co(II) and two Dy(III) single-molecule magnets using fragments of increasing size (f1–f3), and is applied to a Co-based molecular crystal under periodic boundary conditions using only the first coordination shell as the impurity. The central claims are that (i) the smallest fragment f1 reproduces non-embedded CASSCF relaxation times within about an order of magnitude across all molecular complexes, and (ii) periodic CAS-DMET can compute spin–phonon rates for a crystal with a correlated problem of only 259 basis functions out of 2569. The comparison metrics are Kramers-doublet energies, crystal-field parameter derivatives, and relaxation times as a function of temperature.","tokens_in":15164,"tokens_out":4623,"duration_ms":41622,"significance":"If the claims hold, this is a substantial methodological advance: it offers a practical route to multireference spin–phonon calculations for molecular crystals, a regime previously inaccessible to conventional CASSCF. The work benefits from genuine external benchmarks: for the five molecular complexes the embedded results are compared against non-embedded CASSCF, not merely against the embedding itself, and the parity plots of ∂B_l^m/∂R_i with decreasing RMSE as the fragment grows provide a clear convergence signal. The authors are also explicitly honest about the limitations of the study, including the restriction to weakly correlated systems and the known deviations from experiment in the low-temperature Raman regime. The code is made available. The main weakness is that the periodic calculation, which is the most novel component, lacks an independent reference and relies on finite-difference derivatives without step-size convergence or error analysis.","major_comments":[{"comment":"The spin–phonon couplings are obtained by finite differences of fitted B_l^m with ±0.01/0.02 Å for molecules and ±0.01 Å for Crystal-I, but the paper reports no step-size convergence study, no estimate of the fitting noise in B_l^m, and no error propagation into τ. Since Eqs. (6) and (8) make both Orbach and Raman rates quadratic in ∂B_l^m/∂R_i, a systematic bias of the DMET derivative as a function of displacement would directly corrupt the rates while leaving equilibrium KD energies unaffected. For the molecular complexes, the parity plots against non-embedded CASSCF provide an indirect bound on derivative error; for Crystal-I there is no such reference, and the statement that the periodic derivatives are 'smooth' is not a check of accuracy. Please add a step-size convergence test (e.g., 0.005, 0.01, 0.02 Å) and an estimate of the uncertainty in ∂B_l^m/∂R_i from the fit to the SOC Hami","section":"Computational Methods — 'Spin-phonon coupling matrix' (Eqs. 4, 6, 8, 10)"},{"comment":"The abstract states that 'across all systems, treating only the first coordination sphere ... reproduces spin relaxation rates in good agreement with non-embedded CASSCF calculations.' This is not strictly supported for Crystal-I, which has no non-embedded periodic CASSCF reference; the comparison is to the isolated-molecule CASSCF result and to experiment. Moreover, Fig. 10 shows that the crystal environment changes the rates only modestly, so this single periodic demonstration does not test the embedding in a regime where the environment materially affects relaxation. The claim would be appropriately qualified, or the authors should provide a second periodic test where environment effects are significant.","section":"Abstract and Results — 'Molecular Crystal'"},{"comment":"The paper states that 'errors of more than ≈20 cm−1 in crystal-field parameter derivatives propagate directly into the relaxation times' but does not quantify this propagation. Because the rates are quadratic in the coupling matrix elements, a relative error in ∂B_l^m/∂R_i translates into roughly twice that relative error in Γ and τ. Reporting percentage RMSEs of the derivatives and the resulting uncertainty in τ — at least for the most practical f1 fragmentation — would make the order-of-magnitude claim more quantitative and would also enable a concrete assessment of whether f1 is sufficiently converged for predictive use.","section":"Eqs. (6) and (8); Discussion and Conclusions"}],"minor_comments":[{"comment":"The caption says 'for Complex II' in panel (b), but the surrounding section concerns Complex-III; the caption should read 'Complex III'.","section":"Figure 7 caption"},{"comment":"The abstract's '10–68% of the total basis functions' mixes the molecular f1 fractions (26–68%) with the crystal f1 fraction (10%). Since the crystal is not compared to non-embedded CASSCF, please separate the two statements or clarify the comparison basis.","section":"Abstract / Table 1"},{"comment":"The sentence 'The CAS-DMET calculations (across 6N, N being the number of atoms in a system, geometries for all five complexes)' is unclear. The meaning of '6N' and how geometries were selected should be spelled out.","section":"Computational Methods"},{"comment":"The phrase 'errors of more than ≈20 cm−1 in crystal-field parameter derivatives' is dimensionally inconsistent with the RMSE values given earlier (cm−1 Å−1). Clarify whether the intended quantity is an error in the derivative or an integrated energy error.","section":"Discussion and Conclusions"}],"recommendation":"major_revision","confidential_remarks":"The molecular benchmarks are convincing and the paper is a good fit for the journal. The main concern is that the periodic proof-of-concept is not yet validated: the derivative step-size issue is load-bearing for the central 'crystal' claim, and the chosen system shows almost no environmental effect, so the periodic demonstration does not yet establish that CAS-DMET captures environment-induced changes in spin–phonon relaxation. I would support publication after a step-size/error analysis is added and the claims are appropriately qualified."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis paper is worth your time. It does something genuinely new: it couples CAS-DMET to the existing ab initio spin-phonon workflow and shows, across five Co/Dy complexes, that a first-coordination-sphere embedding reproduces non-embedded CASSCF relaxation rates within about an order of magnitude. That is a useful result, not a trivial one. The fragment-size protocol (f1/f2/f3) is systematic, the parity plots of crystal-field parameter derivatives with reported RMSEs are credible, and the discussion is honest about the systems being weakly correlated and still tractable without embedding.\n\nThe benchmarks are independent — non-embedded CASSCF and experiment — so there is no circularity problem. The authors also made the interface code available on GitHub, which is good practice.\n\nThe soft spots are real but mostly addressable. First, the spin-phonon couplings are finite-difference derivatives of fitted crystal-field parameters (0.01/0.02 Å steps), and there is no step-size convergence study or error estimate. For the molecular complexes, the agreement in the parity plots indirectly bounds the error, so this is an accuracy limitation, not a fatal flaw. Second, the periodic demonstration — the only place where the derivative step is unvalidated — is a single crystal in which the environment barely changes the rates. That does not test the method where embedding matters most. Third, the abstract says “quantitative” but the actual criterion is “within an order of magnitude”; that overstates it, and there is also a small numerical discrepancy (10–68% vs 26–68% basis reduction) between the abstract and the Discussion.\n\nNone of this sinks the paper. The central claim — that CAS-DMET can reproduce CASSCF spin-phonon rates at a fraction of the cost — is supported by the evidence. I would send this to peer review. The authors should be asked to add a step-size convergence test and propagate errors into the rates, to either validate or down-claim the periodic result, and to tone down “quantitative” in the abstract to match what they actually show.\n\nFor a reader in molecular magnetism or spin-phonon theory, this is a useful contribution and worth citing.","headline":"A solid, genuinely new method demonstration: CAS-DMET reproduces CASSCF spin-phonon rates within an order of magnitude across five complexes, with a real but addressable gap in derivative step-size validation for the periodic case.","tokens_in":15787,"tokens_out":2687,"would_cite":true,"duration_ms":22739,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper tries to establish that multireference density matrix embedding (CAS-DMET) reproduces conventional CASSCF spin-phonon relaxation rates for magnetic molecules and a molecular crystal using only the first coordination sphere of the","keywords":["multireference density matrix embedding","spin-phonon relaxation","CASSCF","single-molecule magnets","molecular crystals","Kramers doublets","crystal-field parameters","periodic boundary conditions"],"falsifier":"For any of the five complexes, recompute the f1 crystal-field parameter derivatives at several finite-difference step sizes (e.g., 0.005, 0.01, 0.02, 0.04 Å). If the resulting ∂B_l^m/∂R_i values move by more than the RMSE observed against the non-embedded CASSCF derivatives, or if the relaxation times shift by more than an order of magnitude, the derivative pipeline is not converged. For the crystal, where no non-embedded reference exists, the analogous check is to compare periodic f1 derivatives with f2/f3 periodic derivatives or with a large cluster model.","tokens_in":14723,"feed_emoji":"🧲","tokens_out":10277,"duration_ms":76321,"temperature":0.7,"pith_summary":"The paper argues that spin-phonon relaxation calculations, which usually require expensive multireference methods on the whole molecule, can be done with a wave-function-in-wave-function embedding: only the magnetic center and its first coordination shell are treated with a complete-active-space self-consistent field (CASSCF) solver, while the rest is captured by a mean-field bath. Across three cobalt and two dysprosium single-molecule magnets, this first-shell fragment reproduces non-embedded CASSCF relaxation times within about one order of magnitude while cutting the correlated basis to 26–68% of the total. A periodic version of the same workflow treats a cobalt-based molecular crystal with an embedded active space of 259 of 2569 basis functions and produces relaxation rates close to both the isolated-molecule calculation and experiment. If correct, this extends quantitative multireference spin-phonon relaxation from isolated molecules to molecular crystals, and points to a practical route for systems where full CASSCF is intractable.","feed_headline":"First-shell embedding reproduces spin-phonon rates at 10-68% cost","feed_subtitle":"Metal's first coordination shell alone matches full multireference spin-relaxation rates in five magnets and a crystal.","key_machinery":"The central object is the CAS-DMET embedding—density matrix embedding theory solved with a CASSCF (complete active space self-consistent field) impurity solver. The full system is Schmidt-decomposed around a chosen fragment, producing a small set of bath orbitals entangled with the fragment, and the resulting impurity Hamiltonian is solved with CASSCF. For all complexes the fragment is defined by coordination shells (f1, f2, f3), and the paper shows the f1 fragment suffices for relaxation rates. The spin-phonon signal itself is carried by derivatives of the crystal-field Hamiltonian H_S = sum_m,l B_l^m O_l^m, where O_l^m are tesseral operators; these derivatives are obtained with finite diff","core_discovery":"The central claim is that the spin-phonon coupling matrix—the derivatives of the crystal-field parameters B_l^m with respect to nuclear displacements—is accurately captured when only the metal ion and its first coordination sphere form the DMET fragment. For the five benchmark systems, the resulting Orbach and Raman relaxation times stay within one order of magnitude of the non-embedded CASSCF reference over the whole temperature range studied, even though the correlated problem shrinks to 26–68% of the full basis. For the molecular crystal, periodic CAS-DMET yields relaxation times that match both the isolated-molecule CASSCF result and experiment, with an impurity of 259 orbitals drawn fro","pith_inferences":["A practical screening pipeline for molecular qubits and magnets could be built on this: computing a candidate's spin-phonon rates inside its real crystalline environment at a small fraction of the cost of full multireference calculations.","The finite-difference step size and the displacement fidelity of the DMET bath are not tested in the paper; checking convergence with respect to step size, or comparing periodic f1 derivatives with larger fragments, would be the most direct way to harden the protocol.","The slow KD convergence seen for one dysprosium complex (Complex-V) suggests the 'first shell is enough' rule will not hold universally; systems with significant second-sphere electronic influence will likely need larger fragments.","Extending the same embedding to higher-level solvers such as NEVPT2 or multiconfiguration pair-density functional theory, which the paper lists as future work, would show whether the derivative quality—and hence rate accuracy—improves beyond CASSCF."],"forward_implications":["Spin-phonon relaxation rates become computable for molecular crystals with multireference accuracy, a regime where conventional periodic CASSCF is impractical.","For the molecules studied, first-coordination-sphere embedding is a controlled approximation: rates stay within one order of magnitude of full CASSCF while the correlated basis drops to 26–68% of the total.","Larger fragments systematically improve the agreement, so the embedding offers a tunable accuracy/cost trade-off rather than an all-or-nothing choice.","The crystal calculation indicates that in this cobalt system the environment beyond the first coordination shell has little influence on relaxation, suggesting spin-phonon relaxation is largely local in such compounds.","Errors in high-lying Kramers doublets (doubly degenerate spin-orbit levels), which can reach roughly 90 cm^-1 for the smallest fragments, propagate into low-temperature Raman rates, making the whole spin-orbit manifold the quantity to monitor for convergence."],"fun_headline_variants":["Spin-phonon rates with just first shell: 10-68% cost","First coordination sphere nails spin-phonon rates","Crystal spin relaxation: 259 orbitals beat full basis","DMET cuts spin-phonon computation to 10-68%","First-shell embedding matches CASSCF spin-phonon rates"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole relaxation signal reduces to numerical derivatives of fitted crystal-field parameters obtained by re-running the DMET embedding at ±0.01/±0.02 Å displaced geometries, and the paper does not test whether that finite-difference derivative is converged or whether the embedding potential remains unbiased under displacement—so a systematic displacement-dependent error in the DMET bath would flow directly into every Orbach and Raman rate without being visible in equilibri","fun_headline_variants_meta":{"raw":{"variants":["Spin-phonon rates with just first shell: 10-68% cost","First coordination sphere nails spin-phonon rates","Crystal spin relaxation: 259 orbitals beat full basis","DMET cuts spin-phonon computation to 10-68%","First-shell embedding matches CASSCF spin-phonon rates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000232,"raw_usage":{"total_tokens":1316,"prompt_tokens":721,"completion_tokens":595,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":465,"completion_tokens_details":{"reasoning_tokens":507}},"tokens_in":465,"tokens_out":595,"duration_ms":5235,"temperature":1.0,"reasoning_tokens":507,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T17:41:29.510096+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For any of the five complexes, recompute the f1 crystal-field parameter derivatives at several finite-difference step sizes (e.g., 0.005, 0.01, 0.02, 0.04 Å). If the resulting ∂B_l^m/∂R_i values move by more than the RMSE observed against the non-embedded CASSCF derivatives, or if the relaxation times shift by more than an order of magnitude, the derivative pipeline is not converged. For the crystal, where no non-embedded reference exists, the analogous check is to compare periodic f1 derivatives with f2/f3 periodic derivatives or with a large cluster model.","supporting_citations":[],"review_version":1}