{"id":"bf260400-31fb-4154-84b3-c4a6492dda8f","arxiv_id":"2506.17038","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Fixed-node and fixed-phase spin-orbit diffusion Monte Carlo predict a 2.9(1) eV fundamental gap for monolayer CrI3, consistent with a QSGW result and the highest optical reflectance peak.","lead":"Using two flavors of diffusion Monte Carlo, the authors compute the fundamental electronic gap of monolayer chromium triiodide as 2.9 ± 0.1 eV. The value settles a controversy between DFT, which gives about 1 eV, and GW, which gives 2.6 to 3.25 eV, and shows electron correlation matters more than spin-orbit coupling for this gap.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The two-point TDL extrapolation carries the 2.9(1) eV claim, yet Δqp and Δopt differ by 0.69 eV at Nf.u.=18 and 0.16 eV after extrapolation; nothing rules out a residual offset.","rationale":"The paper is careful and reproducible: pseudopotentials are cross-checked against ccECP, the PBE+U2.0 trial wavefunction is motivated by prior variational optimization, and the CIPSI natural-orbital analysis provides a physical reason to expect a single-reference description of the excitation. The deposited data allow independent reanalysis. My concern is not with the overall methodology but with the specific step that converts raw DMC supercell energies into '2.9(1) eV'. That step is the two-point finite-size extrapolation, and it is load-bearing because the central number is the average of two quantities that are assumed—not demonstrated—to meet in the thermodynamic limit. The 0.69 eV spread at Nf.u.=18, the 0.16 eV residual after extrapolation, the lack of charged-cell corrections for the quasiparticle branch, and the single-supercell FPSODMC result together mean that the ±0.1 eV error bar is closer to a statistical statement than a full systematic budget. The value itself remains plausible: it agrees with the most recent QSGW result and with the highest observed optical peak interpreted as a weakly bound excitation, so a rejection would not be warranted. The appropriate status is conditional on a larger-supercell or corrected-extrapolation check, which is exactly the reader's verdict; my read therefore does not change it.","tokens_in":17788,"tokens_out":12548,"duration_ms":134025,"concrete_test":"Compute FNDMC ground, excited, cation, and anion energies at Nf.u.=32 (4×4×1) with the same PBE+U2.0 trial wavefunction, and compare Δqp, Δopt, and their average from two-point (8,18), three-point (8,18,32), and 1/N fits. If the TDL gap shifts by more than about 0.1 eV, or if Δqp−Δopt does not shrink below about 0.1 eV, the quoted 2.9(1) eV value and its error bar are not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline value is obtained by averaging the Eq. 3 quasiparticle gap and the Eq. 4 neutral-promotion gap after a two-point thermodynamic-limit extrapolation. This average is only meaningful if both definitions converge to the same TDL value. The paper argues they must, because the promoted electron in Δopt occupies a delocalized CBM, but at the largest computed supercell (Nf.u.=18) the two definitions still differ by 0.69 eV (Δqp = 3.25(21) eV, Δopt = 2.56(10) eV), and the [8,18] two-point linear extrapolation leaves a 0.16 eV residual (2.970 eV vs 2.815 eV). The quoted ±0.1 eV error bar is smaller than this residual. A two-point fit cannot distinguish a difference that vanishes at infinity from a finite offset caused by residual excitonic character, a wrong asymptotic form, or uncorrected charged-cell finite-size effects. The paper itself states that the difference is expected to vanish with larger supercells but that they cannot afford them, and it explicitly reports energies without kinetic-, potential-, or gap-corrections. The FPSODMC result rests on an even weaker assumption: it is computed only at Nf.u.=8 and is assigned the SR finite-size scaling. These are acknowledged limitations, but they sit directly between the raw DMC data and the quoted precision.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents fixed-node (FNDMC) and fixed-phase spin-orbit (FPSODMC) diffusion Monte Carlo calculations of the fundamental gap of monolayer CrI3. The gap is evaluated from two definitions: the quasiparticle gap Δqp (Eq. 3, IP−EA) and the neutral-promotion gap Δopt (Eq. 4), computed in supercells up to 18 formula units. A two-point linear extrapolation in 1/Nf.u. using Nf.u.=8 and 18 gives Δqp=2.97 eV and Δopt=2.81 eV; their average, 2.9(1) eV, is the paper's headline result. The authors compare with optical reflectance (2.7 eV) and GW results, and use CIPSI natural-orbital analysis to argue that the VBM→CBM excitation is essentially single-reference in an appropriate orbital basis.","tokens_in":18039,"tokens_out":6666,"duration_ms":64291,"significance":"If the result is correct, it provides a high-level benchmark for a debated quantity in a magnetic 2D material and demonstrates the feasibility of DMC for such systems. The work has notable strengths: converged DFT/DMC parameters, validation of pseudopotentials against ccECPs, agreement between two independent DMC implementations, consistency with QSGW (2.9 eV), and deposition of input/output data in the Materials Data Facility. The CIPSI density-matrix analysis is a useful diagnostic for justifying single-reference trial wave functions. The main significance lies in the central value 2.9(1) eV and its implication that electron correlation, not SOC, dominates the gap renormalization.","major_comments":[{"comment":"The headline value 2.9(1) eV is obtained by averaging the extrapolated Δqp and Δopt, but at the largest computed supercell (Nf.u.=18) these two definitions still differ by 0.69 eV, and after the two-point [8,18] extrapolation they differ by 0.16 eV. The reported ±0.1 eV error bar is smaller than this residual and than the statistical errors of the largest-cell endpoints (e.g., ±0.21 eV on Δqp at Nf.u.=18). A two-point fit cannot distinguish a difference that decays to zero in the TDL from a finite offset due to residual excitonic character, incorrect asymptotic form, or uncorrected charged-cell effects. The text acknowledges that larger supercells are unaffordable; the central claim should either be supported by additional cell sizes, a more conservative error estimate, or a clearly labeled model dependence of the TDL limit.","section":"Sec. III, Fig. 2(b), Table 5"},{"comment":"The FPSODMC TDL value rests on a single supercell (Nf.u.=8) with the assumption that the SR finite-size scaling applies unchanged to the FR data. At that size the FR and SR gaps are statistically indistinguishable (Δqp: 3.49(14) vs 3.59(21); Δopt: 2.283(80) vs 2.24(10)); a few-tenths-of-an-eV shift of the FR curve would change the conclusion that SOC is a minor effect. The statement that 'the same fundamental gap was obtained within statistical error' is a null result with limited power, and it is load-bearing for the paper's second main conclusion. Please either report FR data at additional cell sizes or temper the conclusion to reflect the resolution currently available.","section":"Sec. III, Table 4"},{"comment":"The quasiparticle gap Δqp involves charged supercells (N±1), which in periodic boundary conditions are subject to image-charge finite-size errors. The authors state explicitly that they do not apply kinetic-, potential-, or gap-corrections and report raw quantities. Since the difference between Δqp and Δopt at Nf.u.=18 (0.69 eV) is the main finite-size residue that the two-point extrapolation must cure, an estimate of the charged-cell correction (e.g., via a model Coulomb interaction or by comparing to a neutral-promotion equivalent) is needed to attribute the convergence to the claimed physical mechanism. Without this, the TDL value of Δqp carries an uncontrolled systematic error.","section":"Sec. II (last paragraph) and Sec. III (Eq. 3)"}],"minor_comments":[{"comment":"The acronym 'FNMDC' appears on its first use in the Introduction, but the rest of the paper uses 'FNDMC'; please use a single acronym consistently.","section":"Introduction"},{"comment":"The y-axis label 'EDMC/Nf.u.' should be typeset as E_DMC/N_f.u. or replaced with 'DMC energy per formula unit' for clarity.","section":"Fig. 2(a)"},{"comment":"The sentence 'we take the average of the two approaches, Δqp = 2.97 eV and Δopt = 2.81 eV' is ambiguous: the average is 2.89 eV, while the listed values are the two extrapolated definitions. Please rephrase to read 'averaging the two extrapolated definitions, 2.97 eV and 2.81 eV, gives 2.89 eV.'","section":"Sec. III, after Eq. 4"},{"comment":"The energy difference for the excited state between PBE and PBE+U2.0 trial wave functions (0.0014 Ha ≈ 0.04 eV) is small but statistically significant; citing the value in eV would make the claimed significant lowering more transparent.","section":"Supplemental Information, Table 6"},{"comment":"Using the symbol '∞' for the thermodynamic-limit row is unconventional; consider renaming the row 'TDL' to avoid confusion with the 'Nf.u.' header.","section":"Supplemental Information, Tables 3–5"}],"recommendation":"major_revision","confidential_remarks":"The reader's stress-test concern about the two-point extrapolation is, on my reading, the central issue. I agree with the conditional verdict. The paper is scientifically sound in its methodology and presentation, but the quoted precision of the headline value is not supported by the finite-size data. I recommend major revision. No concerns with citation practices or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe headline: this is the first DMC fundamental gap for monolayer CrI3, and the value, 2.9(1) eV, is believable. What's new is the application of fixed-node and fixed-phase spin-orbit DMC to a magnetic 2D insulator, plus the CIPSI density-matrix analysis that explains why a single-reference trial wavefunction works. The paper also ships its data (Materials Data Facility, DOI) and uses open-source QMCPACK and Quantum ESPRESSO, so the results are reproducible.\n\nWhat it does well: the DMC calculations are carefully converged for the supercells used, with timestep checks, pseudopotential validation against ccECP, and T-moves for the nonlocal parts. The agreement between FNDMC and FPSODMC, and the rough agreement with QSGW and the 2.7 eV optical peak, give a consistent picture. The density-matrix diagnostic is a genuinely useful way to justify the trial wavefunction; the natural-orbital localization of the excitation is a nice result.\n\nThe soft spots: the quoted 2.9(1) eV depends on a two-point thermodynamic-limit extrapolation from 8 and 18 formula units, and the two gap definitions still differ by 0.69 eV at the largest cell (3.25 vs 2.56 eV). After extrapolation the residual is 0.16 eV, which is larger than the quoted ±0.1 eV statistical error. The paper acknowledges that larger supercells are unaffordable and that no kinetic, potential, or gap finite-size corrections were applied. So the headline precision is optimistic. FPSODMC is computed only at 8 formula units and assigned the SR finite-size scaling; that is a reasonable assumption, but it remains an assumption. The identification of the 2.7 eV experimental peak as the fundamental gap is also an assumption, as the paper explicitly says, so the \"agreement with experiment\" is weaker than a direct ARPES measurement would be.\n\nNone of these sink the paper. The central value is probably close to the true gap; the issue is the error bar, not the direction. The Hubbard U=2 eV in the trial wavefunction is a parameter, but it was selected in prior work to minimize the ground-state energy, so the central claim is not fitted to the experimental or GW gap. That is honest.\n\nWho this is for: anyone working on 2D magnets or benchmarking GW and DFT+U on correlated 2D materials. The paper deserves a serious referee, not a desk reject: it is a first-of-its-kind calculation, the data are deposited, and the limitations section is candid. A referee should push for a fuller error budget or, failing that, a clear statement that 2.9(1) eV is a DMC estimate with systematic uncertainty of order 0.2–0.3 eV, not a high-precision benchmark.\n\nMy recommendation: send it to peer review. With revision it becomes a solid reference point; even as-is, the community will cite it.","headline":"A careful, reproducible DMC estimate of the monolayer CrI3 gap that is probably near the right value, but the quoted 2.9(1) eV overstates precision given the two-point TDL fit and the 0.69 eV spread between gap definitions.","tokens_in":18656,"tokens_out":2603,"would_cite":true,"duration_ms":25314,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Diffusion Monte Carlo pins monolayer CrI3's fundamental gap at 2.9 ± 0.1 eV","keywords":["diffusion Monte Carlo","fundamental gap","monolayer CrI3","2D magnetic materials","electron correlation","spin-orbit coupling","quasiparticle gap","natural orbitals"],"falsifier":"An angle-resolved photoemission (ARPES) measurement of the quasiparticle gap of freestanding monolayer CrI3 that falls outside 2.8–3.0 eV, or a DMC run at Nf.u.=32 or 54 in which the neutral-promotion and quasiparticle definitions remain separated by more than ≈0.2 eV after extrapolation, would refute the 2.9(1) eV claim.","tokens_in":17563,"feed_emoji":"🧲","tokens_out":8961,"duration_ms":78259,"temperature":0.7,"pith_summary":"The paper aims to settle a long-standing disagreement about the fundamental gap of monolayer CrI3, the first experimentally realized two-dimensional material with intrinsic magnetism. Density functional theory gives gaps between about 0.8 and 1.1 eV, while GW calculations scatter from 2.59 to 3.25 eV, and there is no direct photoemission measurement of the fundamental gap. Using two flavors of diffusion Monte Carlo, the fixed-node and fixed-phase spin-orbit methods, the paper obtains 2.9 ± 0.1 eV. The same value is recovered from both the neutral-promotion and quasiparticle definitions of the gap in the thermodynamic limit, and it agrees with the highest optical reflectance peak and with a recent QSGW calculation. The paper also argues that electron correlation, not spin-orbit coupling, is what controls the gap.","feed_headline":"Quantum Monte Carlo pins CrI3 gap at 2.9 eV","feed_subtitle":"Two independent QMC flavors agree, matching the top optical peak and a QSGW result.","key_machinery":"The key objects are the two gap estimators. The quasiparticle gap $\\Delta^{\\mathrm{qp}}_f(\\Gamma)$ is the ionization potential minus the electron affinity, while the neutral-promotion gap $\\Delta^{\\mathrm{opt}}_f(\\Gamma)$ is the energy of a VBM-to-CBM excitation; in the thermodynamic limit the two are expected to coincide because the promoted electron is delocalized. The calculations use fixed-node DMC with collinear spin determinants and fixed-phase spin-orbit DMC with spinor determinants, both with a Jastrow factor, extrapolated linearly in inverse cell size. A selected configuration-interaction calculation with natural orbitals supplies the supporting analysis: the particle and hole occupations of the excitation are sharply dominated by single orbitals, which justifies single-reference trial wavefunctions.","core_discovery":"At the center of the paper is a numerical claim: the fundamental gap of monolayer CrI3 is 2.9(1) eV, obtained with fixed-node and fixed-phase spin-orbit diffusion Monte Carlo using PBE+U(2 eV) Slater-Jastrow trial wavefunctions. Two formally different estimators are computed at the Γ-point: the quasiparticle gap, defined as the difference between the ionization potential and electron affinity, and the neutral-promotion gap, defined as the energy of promoting an electron from the valence-band maximum to the conduction-band maximum. After a two-point linear extrapolation from the 8- and 18-formula-unit supercells, the quasiparticle gap is 2.97 eV and the neutral-promotion gap 2.81 eV, and their average is reported as 2.9(1) eV. Fully relativistic fixed-phase calculations at the largest affordable supercell give the same value within error bars, despite the underlying DFT gap dropping by about 0.5 eV when spin-orbit coupling is included. The authors conclude that electron correlation, not spin-orbit coupling, determines the fundamental gap, and that a single-reference trial wavefunction is adequate for this excitation.","pith_inferences":["Taken literally, the 2.9 eV quasiparticle gap and the 2.7 eV optical peak imply an exciton binding energy of roughly 0.2 eV for the VBM-to-CBM-like transition, a quantity that could be checked by two-photon or pump-probe spectroscopy.","The same fixed-phase spin-orbit DMC protocol applied to CrBr3 and CrCl3 would test whether the correlation-over-SOC hierarchy is a general feature of chromium trihalide monolayers.","The natural-orbital occupation analysis suggests a practical screening rule: when particle and hole occupations of the natural orbitals are sharply dominated by single orbitals, single-reference DMC trial wavefunctions should be trusted for the gap; this could be validated on other open-shell 2D magnets.","The roughly 0.2 eV gap between the DMC value and the 2.7 eV optical peak may partly reflect fixed-node bias that overestimates excited-state energies; larger fixed-phase spin-orbit DMC supercells would determine whether the remaining difference is physical or methodological."],"forward_implications":["The fundamental gap of monolayer CrI3 is 2.9(1) eV, sitting above DFT values and matching the QSGW result of 2.9 eV.","Electron correlation, not spin-orbit coupling, controls the gap: adding SOC shifts the underlying DFT gap by about 0.5 eV, while the DMC gap is unchanged within statistical error.","In the thermodynamic limit the neutral-promotion gap and the quasiparticle gap coincide, so DMC can use either definition to benchmark fundamental gaps in 2D magnetic insulators.","A single-reference Slater-Jastrow trial wavefunction built from PBE+U2.0 orbitals is sufficient for this excitation, as shown by the sharply peaked particle and hole occupations in the natural-orbital basis.","The result anchors the optical spectrum: the fundamental gap sits about 0.2 eV above the highest reflectance peak at 2.7 eV, consistent with that peak having little excitonic character."],"supporting_citations":[{"why":"Supplies the experimental optical reflectance spectrum and the ≈2.7 eV peak against which the DMC gap is compared.","marker":"[12]"},{"why":"Provides the argument that the neutral-promotion gap coincides with the fundamental gap in the thermodynamic limit for delocalized excitations.","marker":"[24]"},{"why":"Previous QSGW calculation giving 2.9 eV, the GW result the DMC value agrees with best; also supplies the 3.25 eV QSGW value.","marker":"[14]"},{"why":"Earlier G0W0 results (2.76 and 2.59 eV) that bracket the DMC value and motivate the GW comparison.","marker":"[13]"},{"why":"GW and optical study used to classify the experimental peaks and interpret their excitonic character.","marker":"[9]"},{"why":"Source of the DMC-optimized monolayer geometry and the PBE+U2.0 choice that underlies the trial wavefunctions.","marker":"[16]"},{"why":"Defines the fixed-node DMC method used for the scalar-relativistic gap calculations.","marker":"[28]"},{"why":"Defines the fixed-phase spin-orbit DMC method used for the fully relativistic gap calculations.","marker":"[29-33]"}],"fun_headline_variants":["DMC puts CrI3 gap at 2.9 eV","Two QMC estimators agree: CrI3 gap 2.9 eV","Spin-orbit minor, correlation key for CrI3 gap","CrI3 gap from DMC matches optical peak"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the neutral electron promotion from the valence-band maximum to the conduction-band minimum, computed in finite supercells, equals the true fundamental gap once extrapolated to infinite size; at the largest supercell the two definitions still differ by 0.69 eV, and only the two-point extrapolation brings them to 0.16 eV of each other.","fun_headline_variants_meta":{"raw":{"variants":["DMC puts CrI3 gap at 2.9 eV","Two QMC estimators agree: CrI3 gap 2.9 eV","Spin-orbit minor, correlation key for CrI3 gap","CrI3 gap from DMC matches optical peak"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000624,"raw_usage":{"total_tokens":2900,"prompt_tokens":970,"completion_tokens":1930,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":586,"completion_tokens_details":{"reasoning_tokens":1869}},"tokens_in":586,"tokens_out":1930,"duration_ms":13920,"temperature":1.0,"reasoning_tokens":1869,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:13:46.159559+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An angle-resolved photoemission (ARPES) measurement of the quasiparticle gap of freestanding monolayer CrI3 that falls outside 2.8–3.0 eV, or a DMC run at Nf.u.=32 or 54 in which the neutral-promotion and quasiparticle definitions remain separated by more than ≈0.2 eV after extrapolation, would refute the 2.9(1) eV claim.","supporting_citations":[{"cited_title":"Seyler et al","cited_arxiv_id":null,"evidence_quote":"Supplies the experimental optical reflectance spectrum and the ≈2.7 eV peak against which the DMC gap is compared."},{"cited_title":"Dubecký , author F","cited_arxiv_id":null,"evidence_quote":"Provides the argument that the neutral-promotion gap coincides with the fundamental gap in the thermodynamic limit for delocalized excitations."},{"cited_title":"Acharya , author D","cited_arxiv_id":null,"evidence_quote":"Previous QSGW calculation giving 2.9 eV, the GW result the DMC value agrees with best; also supplies the 3.25 eV QSGW value."},{"cited_title":"Wu et al","cited_arxiv_id":null,"evidence_quote":"GW and optical study used to classify the experimental peaks and interpret their excitonic character."},{"cited_title":"Staros , author G","cited_arxiv_id":null,"evidence_quote":"Source of the DMC-optimized monolayer geometry and the PBE+U2.0 choice that underlies the trial wavefunctions."}],"review_version":2}