{"id":"0d513e9a-e9dd-4cb2-97d1-9444ee64f57f","arxiv_id":"2506.15540","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"MnBi4Te7 single crystals show a field-cooling-induced exchange bias with a training effect that also shifts the anomalous Hall loops, attributed by the authors to Mn-Bi antisite defects.","lead":"Exchange bias, a magnetic memory effect, appears in single crystals of the layered magnet MnBi4Te7 and also shifts the material's electrical Hall response. The authors propose that atomic-scale Mn-Bi defects create the magnetic interfaces responsible, but the causal link is inferred rather than directly tested.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Surface STM defect counts do not establish the bulk defect-induced exchange-bias mechanism: no bulk defect quantification or control crystal is presented, so the central causal claim is not settled.","rationale":"The reader's weakest assumption identifies the same central problem: the paper infers a bulk, defect-induced exchange bias from surface STM statistics without a control crystal or bulk defect characterization. I agree that the field-cooled M-H and Hall measurements plausibly show an exchange-bias-like shift, so the observational core is worth reporting. However, the abstract overstates causality, and the paper does not exclude natural phase coexistence or measurement-history effects. In addition, the training-effect analysis in Table I has a near-degeneracy: at 2 K both decay constants are 0.76, and at 6 K they are 0.425 and 0.42, so the individual amplitudes Af and Ar are not separately identified; the reported sign reversal from -264 Oe to +306 Oe should not be interpreted as physical until the fit is reframed with pf=pr (or with full covariance information). This reinforces the CONDITIONAL verdict rather than changing it, so no adjustment is needed. The concrete experimental check on bulk defect concentration is the decisive test for the paper's headline causal claim.","tokens_in":13318,"tokens_out":7773,"duration_ms":86742,"concrete_test":"Compare field-cooled HEB and training curves on at least two MnBi4Te7 crystals whose bulk Mn_Bi concentrations differ measurably (quantified by single-crystal diffuse scattering, EXAFS, or atomic-pair distribution function fitting), using identical measurement protocols; if HEB does not scale with bulk defect density or persists in a low-defect crystal, the defect-causality claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that intrinsic Mn_Bi antisite defects create the FM/AFM interface responsible for the observed exchange bias rests entirely on STM imaging of the cleaved surface (Sec. 3.1, Fig. 1g,h), which finds ~5–7% Mn_Bi coverage on septuple-layer terraces. No bulk-sensitive probe (single-crystal diffuse scattering, EXAFS, neutron/PDF) quantifies the interior Mn_Bi concentration that would pin AFM domains, and no control crystal with a different defect density is measured. Without such a correlation, the field-cooled M-H and Hall shifts in Sec. 3.2.2 are equally compatible with the natural FM/AFM coexistence of the SL/QL superlattice, with surface termination effects, or with minor-loop/history artifacts. The abstract's causal wording ('intrinsic Mn-Bi antisite defects induce strong interlayer exchange coupling, giving rise to ... exchange bias') is therefore stronger than the evidence. A secondary internal issue compounds this: the training-effect fit of Eq. (2) and Table I reports individual frozen/rotatable amplitudes with equal or nearly equal decay constants (0.76 vs 0.76 at 2 K; 0.425 vs 0.42 at 6 K), making the individual amplitudes and their sign reversal unidentifiable without reported uncertainties.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a combined STM, magnetization, AC susceptibility, longitudinal resistivity, and Hall effect study of MnBi4Te7 single crystals. It identifies Mn–Bi antisite defects on the septuple-layer surface with approximately 5–7% coverage, observes exchange-bias-like shifts in field-cooled magnetization loops at 2 K and 6 K with a training effect, and finds a corresponding shift in the anomalous Hall loops at 6 K. The authors attribute the exchange bias to the intrinsic Mn–Bi antisite defects and use a phase-fraction percolation model to describe the temperature- and field-dependent resistivity. The central causal claim is that the defects create a ferromagnetic/antiferromagnetic interface giving rise to exchange bias.","tokens_in":13658,"tokens_out":5099,"duration_ms":48311,"significance":"If the causal link between atomic-scale defects and exchange bias were established, this would be a valuable contribution to the MnBi2Te4(Bi2Te3)n family, connecting microscopic disorder to macroscopic transport and suggesting a route toward interface engineering without chemical doping. The paper includes useful data: STM images with defect identification, AC susceptibility showing phase coexistence, field-cooled M-H and Hall loops exhibiting exchange-bias-like shifts, and a resistivity model with multiple regimes. However, the key mechanistic claim is not demonstrated by the present evidence, and the training-effect fit suffers from parameter identifiability problems. The manuscript is potentially publishable after significant revision.","major_comments":[{"comment":"The abstract's statement that \"intrinsic Mn–Bi antisite defects induce strong interlayer exchange coupling, giving rise to a robust exchange bias\" is not supported by the presented evidence. The STM measurements quantify defects only on the cleaved surface, and no bulk-sensitive probe (such as single-crystal diffuse scattering, EXAFS, neutron/PDF) or control crystal with a different defect density is provided to show that these defects are present in the interior and that they act as the pinning centers producing a ferromagnetic/antiferromagnetic interface. The observed field-cooled M-H and Hall shifts in §3.2.2 could equally arise from the intrinsic FM/AFM coexistence of the SL/QL superlattice, from surface termination effects, or from minor-loop/history artifacts. Please rephrase the causal claim to a correlation (e.g., \"the EB is associated with coexisting FM/AFM regions, and surface antisite defects may contribute\") or supply additional bulk/control measurements that directly test the proposed mechanism.","section":"§3.1, §3.2.2, Fig. 1(g,h) and Fig. 3"},{"comment":"The frozen/rotatable spin relaxation model reports nearly identical decay constants at each temperature (pf = 0.76 and pr = 0.76 at 2 K; pf = 0.425 and pr = 0.42 at 6 K). With such a near-degeneracy, the individual amplitudes Af and Ar cannot be reliably separated without reported uncertainties or a correlation matrix. The sign reversal of Af between −264 Oe (2 K) and +306 Oe (6 K), which is highlighted in the abstract and conclusion, is therefore not statistically supported. Please provide fit uncertainties, residuals, or a discussion of the parameter correlation to justify the physical interpretation of these amplitudes.","section":"§3.2.2, Eq. (2) and Table I"},{"comment":"The claim that the phase-fraction percolation model \"captures the distinct scattering mechanisms\" is not well supported because the fitting procedure is not described (e.g., number of free parameters, constraints, initial guesses, goodness-of-fit criteria, uncertainties), and several fitted parameters in Table II vary non-monotonically with field or take unphysical signs (notably ρ4.5 remains negative at all fields, and ρ2 drops by more than an order of magnitude between 0 T and 0.5 T). As presented, the model is not falsifiable and the parameter values cannot be assessed. The authors should provide a detailed fitting protocol, uncertainties, and a discussion of the physical constraints expected for these coefficients.","section":"§3.3, Eq. (6) and Table II"}],"minor_comments":[{"comment":"The figure caption labels panels (g) and (h) as large-scale topography and zoom-in, but the text refers to \"Fig. 1(h)\" for the zoom-in while also citing \"Fig. 1(g)\" for defect estimation; please clarify the panel assignments consistently.","section":"Figure 1 and text"},{"comment":"The text states that the density of states is suppressed \"within the energy range from −214 meV to −375 meV\"; this ordering is confusing because the lower bound is more negative than the upper bound. Please rewrite as \"between −375 meV and −214 meV\" or use absolute values.","section":"§3.1, Fig. 1(i)"},{"comment":"The sentence \"Coercive fields for the positive and negative branches of M-H loops are shown in Fig. 3(b) & 3(d) at 2 K and 6 K\" references Fig. 3(d) for the coercive field plot, but Fig. 3(d) is an M-H loop; the correct reference should probably be Fig. 3(e).","section":"§3.2.2, Fig. 3"},{"comment":"In the text accompanying Eq. (8), the activation energy appears as \"D/kmeVT\" whereas the displayed equation uses \"D/kBT\"; please correct the typo and define the symbols consistently (kB is presumably the Boltzmann constant).","section":"Eq. (8) and surrounding text"},{"comment":"The fitted parameters in Table I lack units and uncertainties; the exchange-bias fields and amplitudes are presumably in Oersted, but this should be stated explicitly, and error bars should be reported for all fitted values.","section":"Table I"},{"comment":"The manuscript uses \"spin-freezing temperature,\" \"blocking temperature,\" and \"Tf\" interchangeably; please use one consistent term and define its physical meaning in this context.","section":"§3.2.1"}],"recommendation":"major_revision","confidential_remarks":"The central novelty of this paper is the proposed defect-mediated exchange-bias mechanism. While the magnetization and Hall data show exchange-bias-like signatures and the training effect is analyzed, the causal link to Mn–Bi antisite defects rests entirely on surface STM measurements and is not directly tested. If the authors cannot provide bulk defect quantification or a control sample, the conclusions must be significantly toned down to a correlational statement. The training-effect fitting degeneracy also needs to be addressed, as the current analysis does not support the sign-reversal interpretation. These concerns are fixable within substantial revision, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nQuick take on arXiv:2506.15540. The paper reports exchange-bias-like shifts in field-cooled M-H loops of bulk MnBi4Te7 single crystals at 2 K and 6 K, with training effects and a corresponding shift of the anomalous Hall loop at 6 K. That combination is new for this compound, and the raw magnetization and Hall data look honestly acquired. AC susceptibility showing frequency dependence between 6 and 10 K supports genuine FM-AFM coexistence. The magnetotransport phase-fraction model is routine but applied carefully.\n\nThe soft spot is the central causal claim. The abstract says intrinsic Mn-Bi antisite defects induce the exchange bias, but the evidence is STM of the cleaved surface showing 5–7% Mn_Bi coverage. No bulk-sensitive defect quantification and no control crystal with different defect density are presented. Without that, the field-cooled shifts are equally compatible with the natural SL/QL superlattice or with surface termination and minor-loop history effects. The abstract overstates the defect link. Also, the training-effect fit has an identifiability problem: at 2 K the frozen and rotatable decay constants are both 0.76, and at 6 K they are 0.425 and 0.42, so the individual amplitudes and the sign reversal are not reliable without reported uncertainties. The resistivity fit has many free parameters, but that part is descriptive and less concerning.\n\nMinor issues include confused figure labels (the text swaps Fig. 1(f) and (h)) and missing units in Table I. These are fixable.\n\nWho is this for? People working on the MBT family and exchange bias in bulk crystals. The paper doesn't close the defect-causality question, but it gives a clear phenomenology worth explaining. I'd send it to peer review; a good referee can push for a more cautious abstract and for bulk defect evidence or an explicit statement that the mechanism is unresolved. Verdict: conditional, with major revision.","headline":"A solid observational report of exchange bias and training in bulk MnBi4Te7, but the defect-causality claim outruns the evidence.","tokens_in":14225,"tokens_out":2429,"would_cite":false,"duration_ms":21304,"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":"The paper argues that native Mn-on-Bi antisite defects in MnBi4Te7 act as built-in FM/AFM interfaces that shift both magnetization and anomalous Hall hysteresis loops.","keywords":["MnBi4Te7","exchange bias","anomalous Hall effect","Mn-Bi antisite defects","training effect","FM-AFM phase coexistence","magnetic topological insulator","scanning tunneling microscopy"],"falsifier":"Count Mn-on-Bi antisites on a cross-section through the same crystal (for example by scanning transmission electron microscopy) and compare with the 5-7% surface STM value; if the bulk density is much lower, surface defects cannot anchor the bulk interface. In addition, compare ZFC and FC hysteresis loops at 2 K and 6 K: if a zero-field-cooled loop already shows the same shifted anomalous Hall and magnetization hysteresis, or if varying the cooling field does not change $H_{EB}$ as a pinned AFM interface would require, the defect-pinning mechanism fails.","tokens_in":13073,"feed_emoji":"🧲","tokens_out":7476,"duration_ms":69676,"temperature":0.7,"pith_summary":"MnBi4Te7 is a layered magnetic topological insulator in which ferromagnetic and antiferromagnetic blocks naturally alternate. The paper argues that native Mn-on-Bi antisite defects, imaged by STM at 5-7% surface coverage, act as built-in pinning centers that couple these blocks, producing a real exchange-bias shift in the magnetization hysteresis and in the anomalous Hall effect. If correct, this makes the intrinsic disorder of the crystal a tunable magnetic interface: warming from 2 K to 6 K converts an asymmetric loop shift into a symmetric one, and the frozen-spin contribution to training reverses sign, signaling a reconfiguration of interfacial spins. The result matters because it connects atomic-scale defects to macroscopic Hall transport in the bulk material, pointing toward stabilizing quantum anomalous Hall behavior by interface engineering without artificial heterostructures.","feed_headline":"Native defects give MnBi4Te7 a built-in exchange bias","feed_subtitle":"The same atomic-scale disorder shifts magnetic and anomalous Hall hysteresis, coupling microscopic defects to topological transport.","key_machinery":"The load-bearing mechanism is a built-in ferromagnet/antiferromagnet interface formed by intrinsic Mn-on-Bi antisite defects between the septuple and quintuple layers. The quantitative handles are the exchange-bias field $H_{EB} = (H_{C+} + H_{C-})/2$, the training-effect decomposition $H_{EB,n} = H_{EB,\\infty} + A_f e^{-n/p_f} + A_r e^{-n/p_r}$ separating frozen and rotatable interfacial spins, and the field-cooling protocol that sets the interface. This machinery lets the authors connect surface imaging (defect density) to bulk magnetization (loop shift and training) and to transport (shifted anomalous Hall loop, phase-fraction resistivity).","core_discovery":"The paper's central claim is that MnBi4Te7 single crystals exhibit intrinsic, defect-mediated exchange bias. STM shows Mn-on-Bi antisite defects at roughly 5-7% coverage on septuple-layer terraces; the authors identify these as the source of a strong interlayer exchange coupling that creates coexisting FM/AFM interfaces. After field cooling, the hysteresis loop shifts by an exchange-bias field $H_{EB} = (H_{C+} + H_{C-})/2$, and the same shift appears as a horizontal displacement of the anomalous Hall hysteresis, showing that the interface imprints itself on electronic transport. Between 2 K and 6 K the loop evolves from asymmetric to symmetric, and a two-component training-effect analysis gives relaxation amplitudes that flip from about -264 Oe to +306 Oe, which the authors interpret as frozen interfacial spins giving way to rotatable spins. Finally, a phase-fraction resistivity model separates electron-electron, spin-fluctuation, and electron-phonon scattering in the FM, AFM, and PM regimes, tying the same phase coexistence to transport.","pith_inferences":["A direct test follows from crystal growth: samples grown to deliberately vary the Mn-on-Bi antisite density should show a monotonic change in $H_{EB}$ and in the anomalous Hall shift; if they do not, the defects are spectators rather than pinning centers.","The sign reversal in training relaxation amplitude between 2 K and 6 K may be a general fingerprint of competing frozen and rotatable spins in any van der Waals magnet with FM/AFM coexistence, so the same two-exponential analysis on other MBT-family members could reveal whether this is universal.","Because the exchange bias appears in the Hall signal, the magnetic state of the surface/bulk interface could be read out electronically; this hints at a device-relevant way to probe interfacial spin reconfiguration without magnetization measurements."],"forward_implications":["Hall-effect measurements on MnBi4Te7 at low temperature must treat the anomalous Hall hysteresis as shifted by magnetic history; the exchange-bias displacement is part of the electronic response, not noise.","Temperature becomes a control knob: cycling through 2 K and 6 K switches the interfacial spin configuration between frozen-dominated and rotatable-dominated, changing both the loop shape and the training behavior.","Resistivity data can be decomposed into FM, AFM, and PM phase fractions, so transport measurements alone can track the phase coexistence that underlies the exchange bias.","If the defect interface is this effective in the bulk, interface engineering (adjusting defect density or stacking) could stabilize quantum anomalous Hall transport in MBT-family crystals without chemical doping."],"supporting_citations":[{"why":"Establishes the MBT-family magnetic ground states and the AFM-to-FM progression that classifies MnBi4Te7.","marker":"[9]"},{"why":"Supplies the synthesis route and the low saturation/spin-flop fields that make MnBi4Te7 favorable for low-field studies.","marker":"[21]"},{"why":"Provides the STM identification of Mn-on-Bi antisite defects and the band-gap reference used to confirm surface quality.","marker":"[23]"},{"why":"Reports FM-AFM coexistence in thin flakes of MnBi4Te7, the prior observation this work extends to bulk and to transport.","marker":"[29]"},{"why":"Gives the empirical power-law training-effect formula used to fit H_EB versus cycle number.","marker":"[31]"},{"why":"Gives the defect-density calculation method used to estimate the 5-7% surface coverage.","marker":"[37]"},{"why":"Supplies the frozen/rotatable two-spin relaxation model that yields the -264 Oe to +306 Oe amplitude reversal.","marker":"[39]"},{"why":"Underlies the training-effect measurement protocol for field-cooled loops.","marker":"[41]"},{"why":"Provides the percolation phase-fraction resistivity model used to fit rho_xx in FM, AFM, and PM regimes.","marker":"[43]"}],"fun_headline_variants":["Defect-induced exchange bias reshapes Hall effect in MnBi4Te7","MnBi4Te7's antisite defects drive exchange bias and Hall shift","Intrinsic defects set exchange bias and move Hall loops in MnBi4Te7","Exchange bias in MnBi4Te7 flips with temperature, affecting Hall"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument needs the 5-7% Mn-on-Bi defect density observed on the cleaved surface to represent the bulk crystal, and those defects to be the pinning centers that set up the FM/AFM interface; no defect-free control crystal or alternate cooling protocol rules out phase separation, surface termination, or minor-loop artifacts.","fun_headline_variants_meta":{"raw":{"variants":["Defect-induced exchange bias reshapes Hall effect in MnBi4Te7","MnBi4Te7's antisite defects drive exchange bias and Hall shift","Intrinsic defects set exchange bias and move Hall loops in MnBi4Te7","Exchange bias in MnBi4Te7 flips with temperature, affecting Hall"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001003,"raw_usage":{"total_tokens":4271,"prompt_tokens":1003,"completion_tokens":3268,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":619,"completion_tokens_details":{"reasoning_tokens":3184}},"tokens_in":619,"tokens_out":3268,"duration_ms":24050,"temperature":1.0,"reasoning_tokens":3184,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:33:17.585769+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Count Mn-on-Bi antisites on a cross-section through the same crystal (for example by scanning transmission electron microscopy) and compare with the 5-7% surface STM value; if the bulk density is much lower, surface defects cannot anchor the bulk interface. In addition, compare ZFC and FC hysteresis loops at 2 K and 6 K: if a zero-field-cooled loop already shows the same shifted anomalous Hall and magnetization hysteresis, or if varying the cooling field does not change $H_{EB}$ as a pinned AFM interface would require, the defect-pinning mechanism fails.","supporting_citations":[{"cited_title":"Chang, W","cited_arxiv_id":null,"evidence_quote":"Establishes the MBT-family magnetic ground states and the AFM-to-FM progression that classifies MnBi4Te7."},{"cited_title":"Ovchinnikov, X","cited_arxiv_id":null,"evidence_quote":"Supplies the synthesis route and the low saturation/spin-flop fields that make MnBi4Te7 favorable for low-field studies."},{"cited_title":"Garnica, M","cited_arxiv_id":null,"evidence_quote":"Reports FM-AFM coexistence in thin flakes of MnBi4Te7, the prior observation this work extends to bulk and to transport."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the empirical power-law training-effect formula used to fit H_EB versus cycle number."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the defect-density calculation method used to estimate the 5-7% surface coverage."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the frozen/rotatable two-spin relaxation model that yields the -264 Oe to +306 Oe amplitude reversal."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Underlies the training-effect measurement protocol for field-cooled loops."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the percolation phase-fraction resistivity model used to fit rho_xx in FM, AFM, and PM regimes."}],"review_version":2}