{"id":"5df3117c-1040-471d-8ba5-81f34f6a2da6","arxiv_id":"2504.17871","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The current-induced conductive state of Ca2RuO4 is driven by current-created d3 electronic defects inside a persistent Mott gap, not by thermal gap closure.","lead":"Researchers used in-operando X-ray scattering to show that passing a current through the Mott insulator Ca2RuO4 creates new correlated electronic states that conduct while the insulating gap persists. The finding distinguishes the current-driven metal-like state from the heated one and points to a defect-based mechanism relevant for low-power electronics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The d3-defect mechanism is not independently constrained: the x↔I mapping is assumed linear and d3-only, so the agreement in Fig. 4h partly validates the assumption rather than the mechanism.","rationale":"Good-faith reading: the paper's central claim is that current-induced IMT in Ca2RuO4 arises from current-generated d3 electronic defects in a d4 host, not from gap collapse. The experiment is well-executed: in-operando O K-edge RIXS with a small device, fixed geometry, reversible recovery at I5 = 0, and a thermal control that clearly distinguishes the current state from Joule-heated metal. The observed energy-selective suppression and absence of low-energy continuum are real and support a non-thermal, gap-preserving conductive state. The vulnerable point is the conversion of that spectroscopic signature into a specific microscopic population. Section 3 introduces Itotal(x) = (1−x)Id4 + xId3 with x assumed proportional to current and d3 assumed dominant. The theoretical curves in Fig. 4h then show linear suppression of A, B, D, and the experimental curves in Fig. 3g also show linear suppression; but because the x-axis of the model is x and of the experiment is I, the comparison requires the very relation it claims to validate. No independent measurement (e.g., valence-sensitive XAS, transport-derived carrier density, or spatial imaging) fixes x or the d3/d5/d6 balance. The S3.2 plaquette argument depends on unmeasured intersite v and is not fit to data. The C/C′ discrepancy is acknowledged and attributed to temperature without a finite-T calculation. These gaps do not invalidate the experiment, but they mean the 'demonstrated' language in the Conclusions overstates the evidence. The reader's weakest_assumption identifies this same step; I agree. A conditional verdict is right: the manuscript should be accepted with the requirement that the x↔I mapping and d3-only assumption be tested, or the claims softened to 'consistent with.' My recommendation is UNCHANGED relative to the reader's CONDITIONAL.","tokens_in":24991,"tokens_out":6275,"duration_ms":72590,"concrete_test":"Refit the published W_j(I) data with a two-species model W_j(I) = (1−s(I))W_j^d4 + s_3(I)W_j^d3 + s_5(I)W_j^d5, where s_3 + s_5 = s and s(I), s_3/s_5 are free, and recompute the d3/d5 cross-sections at 300 K with thermal broadening instead of the T = 0 curves in Fig. 4h. If the best fit requires s_5 ≈ 0 and reproduces the unchanged C/C′ modes without added ad hoc broadening, the d3-only linear-defect interpretation is supported; if a d5 fraction or a nonlinear s(I) fits equally well or better, the central claim lacks quantitative support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is Section 3's Itotal(x) = (1−x)Id4 + xId3, with x taken to be a d3-only concentration that scales linearly with d.c. current. The agreement between the linear suppression of modes A, B, D in Fig. 4h and the measured W(I) in Fig. 3g does not validate this mapping, because the experimental axis is current while the theoretical axis is x; choosing x ∝ I plus a single scale makes any monotone suppression curve match qualitatively. The data are equally compatible with a d5-dominated population (the d5 multiplet also lacks A/B-type modes), with a mixture of d3 and d5, or with a growing filament volume fraction that leaves the homogeneous-defect picture untested. The only stated reason for d3 dominance is the plaquette energy estimate in S3.2, which depends on an unmeasured intersite Coulomb parameter v and is not compared to data. The one quantitative discrepancy—the predicted small suppression of C/C′ that is not observed—is dismissed as a T = 0 artifact without a finite-temperature calculation. Thus the central conclusion that current creates predominantly d3 correlated defects with concentration proportional to current rests on an assumed, rather than demonstrated, relation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports in-operando O K-edge resonant inelastic x-ray scattering (RIXS) on a micrometer-sized Ca2RuO4 device while a d.c. current drives an insulator-to-metal transition (IMT). The data show an energy-selective suppression of the spectral weight of modes A, B, and D that grows with current, while Hund's modes C and C' remain essentially unchanged; the low-energy weight does not develop the continuum seen in the thermally induced metallic state, and the spectrum fully recovers when the current is removed. The authors reproduce the suppression pattern with exact-diagonalization RIXS calculations based on a weighted average Itotal(x) = (1−x)Id4 + xId3, where x is a current-induced concentration of d3 defects, and conclude that the conductive non-equilibrium steady state is caused by correlated electronic defects that coexist with a persistent Mott gap.","tokens_in":25283,"tokens_out":6601,"duration_ms":68671,"significance":"The experimental dataset is of high quality and directly addresses a long-standing question: what microscopic electronic changes accompany a current-induced IMT in a Mott insulator. The energy-selective suppression, the thermal-vs-current comparison, and the full reversibility are clean observations that will be useful to the community regardless of the final interpretation. If the d3-defect mechanism were established, the paper would provide a new microscopic route to current-driven IMT distinct from Mott-gap collapse, with implications for resistive switching and orbitronic/spintronic devices. However, the mechanistic claim is currently supported only by a qualitative model whose free parameters and assumptions (x, linear x–I relation, d3-only population, homogeneous distribution) are not independently constrained. The value of the work is therefore primarily experimental and phenomenological; the theoretical demonstration of the defect mechanism still needs substantial strengthening.","major_comments":[{"comment":"The mapping between the calculated defect concentration x and the applied current is assumed, not derived. Because the theoretical curves in Fig. 4h are plotted against x and the experimental data in Fig. 3g are plotted against I, the comparison necessarily includes a free proportionality constant that converts x to I. The linear-in-x prediction does constrain the relative slopes of modes A, B, and D, and the observed suppression is consistent with that constraint; however, the Conclusions claim that the calculation 'validates the assumptions that the d3 concentration is predominant and linearly increases with the current' overstates what a two-axis comparison with one free scale can establish. An independent determination of x (for example, from XAS line-shape analysis or from transport modeling) is needed to make the d3 concentration a demonstrated outcome rather than a model input.","section":"§3, Eq. Itotal(x) = (1−x)Id4 + xId3"},{"comment":"The calculation uses Itotal(x) = (1−x)Id4 + xId3, and the text states that d6 does not contribute to O K-edge RIXS and that d5 is not included in the main calculation. Yet the multiplet diagram in Fig. 4c shows that d5, like d3, lacks the low-energy A-type magnetic mode and has a different B-mode set, so the observed suppression of A and B is equally compatible with a d5-dominated defect population or a d3/d5 mixture. The only quantitative argument for d3 dominance is the plaquette energy estimate in S3.2 (Eqs. 9–11), which depends on the unmeasured intersite Coulomb parameter v and is not compared with any experimental quantity. A calculation that compares d3-only, d5-only, and mixed populations against the same dataset is necessary before 'the d3 concentration is predominant' can be regarded as demonstrated.","section":"§3, d5/d6 neglect"},{"comment":"The calculation predicts a small suppression of the C and C' spectral weights, while the experiment shows no resolvable change; the manuscript attributes this discrepancy to the T = 0 calculation and thermal broadening of C and C' at 300 K without performing or quoting a finite-temperature calculation. Since the claimed agreement in Fig. 4h is central to the mechanistic conclusion, this acknowledged mismatch should be backed by an explicit finite-T simulation or by a quantitative demonstration that room-temperature broadening is sufficient to wash out the predicted suppression.","section":"Fig. 4h, C/C' discrepancy"},{"comment":"The main text (Section 2) notes that the current-driven transition is 'known to be spatially inhomogeneous' (Refs. 35, 39), but the model of Section 3 assumes defects 'homogeneously distributed across the sample' (Fig. 4 caption). A filamentary or phase-separated conductive state within the 10×3 µm² probed volume would produce the same averaged RIXS suppression as a homogeneous population of defects, because the RIXS signal is an average over the beamspot. The current data therefore cannot distinguish a growing filament volume fraction from an increasing homogeneous defect density, and the latter is required for the microscopic interpretation in terms of Itotal(x).","section":"§2 (spatial inhomogeneity) vs §3 (homogeneous model)"}],"minor_comments":[{"comment":"S2.2 states that mode A at 2 mA required a ~20% larger FWHM, which appears to contradict the main-text statement that no line-shape change is identified; this should be clarified.","section":"S2.2"},{"comment":"The 'linear suppression' shading in Fig. 3g is not accompanied by a fitted line or residuals; adding the linear fit with slopes and uncertainties would make the 'proportional to current' claim quantitative.","section":"Fig. 3g"},{"comment":"The headings in S3.4 and S3.5 contain 'Wannerization' rather than 'Wannierization'.","section":"S3.4, S3.5"},{"comment":"The insets in Fig. 2 are too small to see the claimed differences in low-energy spectral weight; difference spectra or a zoomed panel would help the reader verify the comparison.","section":"Fig. 2"},{"comment":"A data-availability statement for the raw RIXS spectra and fitting parameters would improve reproducibility.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know about this paper: the experiment is the real contribution, and the mechanism is a good hypothesis that the authors overstate as a demonstration. They measure O K-edge RIXS on a micrometer-scale Ca2RuO4 device while driving d.c. current, and they see an energy-selective suppression of modes A, B, and D, linear in current, with full recovery when the current is off. The control against the thermally-driven IMT is clean and rules out simple Joule heating. That is new and solid.\n\nThe model is where I part company. They take the RIXS cross-section as Itotal(x) = (1−x)Id4 + xId3, with x the concentration of d3 defects that they assume scales linearly with current. The agreement in Fig. 4h is qualitative: any monotone suppression curve can be made to look linear in current by choosing the x↔I mapping, so the fit does not validate the mechanism. The d5 configuration is not excluded—it also lacks the A and B modes—and the plaquette energetics in S3.2 that supposedly favor d3 depends on an unmeasured intersite Coulomb parameter v and is not compared to data. The one quantitative mismatch, the predicted suppression of C/C' that is not observed, is dismissed as a T=0 artifact without a finite-temperature calculation. So the central claim—that current creates predominantly d3 correlated defects—is an assumption, not a demonstrated result.\n\nI want to be fair: the data are good, and the idea of correlated charge defects coexisting with a persistent Mott gap is a worthy hypothesis. But the paper's language in the abstract and conclusions ('we demonstrate') goes beyond what the evidence supports.\n\nI would send this to peer review rather than desk reject. The experimental observation deserves referee time, and the model is worth serious discussion even if it needs to be reframed as a hypothesis with independent constraints on x. The ideal referee report would ask for a softened conclusion, a comparison of d3 vs d5 mixtures, and either a finite-T calculation or a dropped claim about the C/C' modes.\n\nWho's it for: the ruthenate/RIXS community and anyone working on current-driven phase transitions in Mott insulators. I would not cite it as proof of the defect mechanism, but I might cite it for the in-operando methodology.\n\nRegards.","headline":"A well-executed in-operando RIXS study with a clean observation, but the d3-defect mechanism is a plausible hypothesis presented as a demonstrated conclusion.","tokens_in":25875,"tokens_out":2371,"would_cite":true,"duration_ms":23599,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.30.+h","78.70.Ck","71.27.+a"],"model":"deepseek-v4-flash","headline":"The paper claims that applying a d.c. current across Ca2RuO4 creates electronic defects at ruthenium sites—d3/d5 charge configurations—whose concentration rises with current and turns the material conductive while the Mott gap remains open.","keywords":["current-driven insulator-to-metal transition","Ca2RuO4","resonant inelastic x-ray scattering","Mott insulator","non-equilibrium steady state","correlated electronic defects","d3 charge defects","persistent Mott gap"],"falsifier":"Spatially resolved RIXS or nano-infrared imaging during current flow could show whether the spectral suppression is uniform or localized in conductive filaments; a Ru L-edge RIXS or XAS measurement could quantify the d5 population directly. If the energy-selective suppression tracks only filamentary regions, or if d5 states appear at concentrations comparable to d3, the claimed homogeneous electronic-defect mechanism would be contradicted.","tokens_in":24790,"feed_emoji":"⚡","tokens_out":5946,"duration_ms":53920,"temperature":0.7,"pith_summary":"This paper aims to settle what happens electronically when an electric current switches the Mott insulator Ca2RuO4 into a conducting state. Using in-operando resonant inelastic x-ray scattering on a micrometer-sized device at fixed temperature, it finds that rising current suppresses the RIXS intensity in an energy-selective way, and that the same suppression does not occur when the transition is driven by heating. The authors argue, with cluster calculations, that the current creates correlated d3/d5 charge defects at ruthenium sites whose concentration grows with current, and that these defects conduct while the Mott gap persists. If correct, the current-driven transition is a defect-mediated electronic transition rather than a gap-collapse metal transition.","feed_headline":"Current makes Ca2RuO4 conductive without closing its Mott gap","feed_subtitle":"In-operando RIXS traces how current-created d3/d5 defects, not heat, drive the metal-like state.","key_machinery":"The load-bearing object is the linear-mixing model for the RIXS cross section, $I_{\\rm total}(x) = (1-x)I_{d^4} + x I_{d^3}$, where $x$ is the concentration of current-induced d3 defects and $I_{d^4}$, $I_{d^3}$ are the calculated spectra of defect-free and defect-containing Ru-O clusters. It converts a set of spectroscopic observations—energy-selective suppression of modes A, B, and D without peak shifts—into a quantitative statement about defect concentration, assuming $x$ scales with applied current. The supporting machinery is in-operando O K-edge RIXS on a ~10×12×1 µm flake, with spectra at both apical and planar oxygen resonances, alongside exact-diagonalization cross-section calculations that include spin-orbit coupling, Hund's coupling, crystal-field splittings, and Ru-O hybridization. The model's key output is the prediction that modes A, B, and D lose weight linearly with $x$, while modes C and C′ remain nearly unchanged.","core_discovery":"The central claim is that the current-driven insulator-to-metal transition in Ca2RuO4 is not the usual Mott-gap collapse. In a device held at 300 K, the RIXS spectra show no peak shifts and no low-energy particle-hole continuum; instead, the spectral weight of certain excitations—magnetic spin-orbital modes at about 80 meV, the spin-orbit exciton around 350 meV, and the high-energy dd excitations near 3.5 eV—falls linearly with current, while Hund's-exchange modes barely change. The paper attributes this energy-selective suppression to the formation of correlated electronic states, predominantly d3 (one hole per Ru site) alongside the d4 host, with d5 and d6 also possible but less populated. Calculations of the O K-edge RIXS cross section using exact diagonalization, combined as $I_{\\rm total}(x) = (1-x) I_{d^4} + x I_{d^3}$, reproduce the observed linear suppression with defect concentration x, which the paper identifies with current. The conclusion is that a conductive non-equilibrium steady state coexists with a persistent Mott gap, with an electronic rather than structural origin.","pith_inferences":["A natural extension the paper does not develop is that the same linear-mixing formula could be fit to RIXS data with both d3 and d5 fractions free, which would test whether d5 contributes measurably at higher currents.","If the defect population is in fact homogeneous, transport across the device should obey a smooth scaling of resistance with spectral-weight suppression; spatially resolved measurements could check whether conduction instead runs through filaments.","The energy-selective suppression pattern might serve as a spectroscopic fingerprint for identifying current-induced defect-mediated conduction in other correlated oxides, not just ruthenates.","The paper's plaquette energy arguments imply that tuning octahedral distortions or intersite Coulomb interactions—for instance by strain—could change which defect charge states dominate, offering a design lever for such transitions."],"forward_implications":["The current-driven conducting state in Ca2RuO4 is electronically distinct from the high-temperature metallic phase, so Joule heating does not explain the transition.","A persistent Mott gap can coexist with high conductivity when current-created correlated defects provide in-gap charge carriers.","The defect concentration, tracked through the RIXS spectral weight, is a directly measurable order parameter for the non-equilibrium steady state.","In-operando RIXS can be applied to other electrically driven phase transitions to distinguish electronic-defect mechanisms from gap closure or structural transitions.","Switching devices based on such materials would rely on controlling the density of electronic defects rather than on heating or structural collapse."],"supporting_citations":[{"why":"Establishes the current-stabilized metal in Ca2RuO4, the phenomenon the paper studies in-operando.","marker":"[14]"},{"why":"Shows the current-stabilized state has a unique crystal structure distinct from the thermal metal, used to compare current- and temperature-driven transitions.","marker":"[18]"},{"why":"Provides ARPES evidence that the current-driven transition occurs without Mott breakdown, the prior claim this paper's RIXS data build on.","marker":"[23]"},{"why":"Supports the persistence of the Mott gap in the current-driven conductive state, a key experimental anchor.","marker":"[24]"},{"why":"Identifies the spin-orbital excitations of Ca2RuO4 measured by RIXS, the basis for assigning modes A, B, C, C′, and D.","marker":"[30]"},{"why":"Provides the non-equilibrium steady state context for current-biased Ca2RuO4.","marker":"[34]"},{"why":"Interprets spin-orbit-induced orbital excitations in ruthenates, used to understand the low-energy RIXS spectra.","marker":"[37]"},{"why":"Addresses spatial inhomogeneity and electric-field-induced pattern formation, relevant to the assumed uniform defect distribution and octahedral distortion interfaces.","marker":"[39]"}],"fun_headline_variants":["Ca2RuO4 turns metal by current, Mott gap stays intact","Current-driven metal state in Ca2RuO4 keeps its Mott gap","RIXS reveals current makes Ca2RuO4 metal, gap persists","Metal-like Ca2RuO4 from current alone: Mott gap survives","Current creates d3 states, Ca2RuO4 conducts with gap intact"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes the RIXS changes come from a uniform population of d3 defects whose fraction grows linearly with current, with d3 the only important charge state; if conduction instead runs through filaments, comes from structural reconfiguration, or involves comparable d5 populations, the quantitative demonstration loses its footing.","fun_headline_variants_meta":{"raw":{"variants":["Ca2RuO4 turns metal by current, Mott gap stays intact","Current-driven metal state in Ca2RuO4 keeps its Mott gap","RIXS reveals current makes Ca2RuO4 metal, gap persists","Metal-like Ca2RuO4 from current alone: Mott gap survives","Current creates d3 states, Ca2RuO4 conducts with gap intact"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000179,"raw_usage":{"total_tokens":1315,"prompt_tokens":974,"completion_tokens":341,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":244}},"tokens_in":590,"tokens_out":341,"duration_ms":3938,"temperature":1.0,"reasoning_tokens":244,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:30:49.748851+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Spatially resolved RIXS or nano-infrared imaging during current flow could show whether the spectral suppression is uniform or localized in conductive filaments; a Ru L-edge RIXS or XAS measurement could quantify the d5 population directly. If the energy-selective suppression tracks only filamentary regions, or if d5 states appear at concentrations comparable to d3, the claimed homogeneous electronic-defect mechanism would be contradicted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the current-stabilized metal in Ca2RuO4, the phenomenon the paper studies in-operando."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows the current-stabilized state has a unique crystal structure distinct from the thermal metal, used to compare current- and temperature-driven transitions."},{"cited_title":"Nakatsuji and Y","cited_arxiv_id":null,"evidence_quote":"Provides ARPES evidence that the current-driven transition occurs without Mott breakdown, the prior claim this paper's RIXS data build on."},{"cited_title":"Ricc´ o, M","cited_arxiv_id":null,"evidence_quote":"Supports the persistence of the Mott gap in the current-driven conductive state, a key experimental anchor."},{"cited_title":"Curcio, C","cited_arxiv_id":null,"evidence_quote":"Identifies the spin-orbital excitations of Ca2RuO4 measured by RIXS, the basis for assigning modes A, B, C, C′, and D."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the non-equilibrium steady state context for current-biased Ca2RuO4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Interprets spin-orbit-induced orbital excitations in ruthenates, used to understand the low-energy RIXS spectra."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Addresses spatial inhomogeneity and electric-field-induced pattern formation, relevant to the assumed uniform defect distribution and octahedral distortion interfaces."}],"review_version":1}