{"id":"bacc1322-f2a8-4cdf-b187-53642d3b452b","arxiv_id":"2607.20287","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The resistance of the semiconductor Eu5In2As6 remembers the magnetic-field history it has experienced, an effect that starts near 30 K in the paramagnetic phase rather than in the magnetically ordered state.","lead":"Eu5In2As6 shows a magnetic-field history memory in its electrical resistance that appears at 30 K, well above its 16-K antiferromagnetic ordering, with no matching magnetization memory. The effect is reproducible across samples and could open a new class of field-history memory effects in magnetic semiconductors.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Paramagnetic-phase claim lacks field-dependent neutron search; a field-induced magnetic order could host the MRM.","rationale":"The reader correctly identified several weaknesses, especially the lock-in detection limit and the symmetry-model contradiction. However, the single most load-bearing concern is that the paramagnetic nature of the phase at the fields where MRM is observed is not experimentally established. The paper's neutron data are zero-field (except for a sub-TN check at 15 K), while the memory effect is measured under fields that could in principle induce or modify magnetic order. If field-induced long-range order exists, the MRM would be explained by conventional ordered-phase mechanisms, and the headline novelty collapses. This is a missing control, not a disagreement with the data. The proposed test—field-dependent neutron scattering above TN—would directly settle whether the phenomenon truly occurs in the paramagnetic phase. I therefore recommend keeping the CONDITIONAL verdict, with this specific measurement as a condition for acceptance. The reader's concern about unmeasured low-field ZFC data is real but less decisive: the hysteresis loops at 18–25 K are measured within the lock-in range, so the memory effect itself is not solely an artifact of the detection limit.","tokens_in":19713,"tokens_out":5894,"duration_ms":65381,"concrete_test":"Perform neutron diffraction on Eu5In2As6 at T = 18, 20, 25, 30 K while sweeping magnetic field from 0 to 9 T (both increasing and decreasing). Monitor candidate wavevectors including Q = (0,0,1/2) and (1,0,0), and, ideally, survey a volume of reciprocal space at each field to detect any superlattice peak. A positive detection of field-induced magnetic scattering (or a field-history-dependent peak intensity) would invalidate the paramagnetic-phase claim; a null result would support it.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that MRM occurs in the paramagnetic phase (T between ~16 K and ~30 K). The evidence for paramagnetism comes from neutron diffraction measured at zero field (Fig. S8c) and from M(H) at 10 K and 20 K (Fig. S8f). However, the MRM itself is defined by field-history-dependent resistivity measured at finite fields (2–9 T). The manuscript never checks whether a magnetic Bragg peak develops under the same finite fields at T > 16 K. Fig. S8d shows field sweeps of the (0,0,1/2) peak only at 15 K, below TN, and finds no hysteresis but does not address whether the peak appears or sharpens at 18–30 K under field. If a field-induced AFM or spin-density-wave order appears above 16 K, then the resistivity hysteresis could be a conventional field-induced domain effect, and the claim that MRM occurs in a paramagnetic phase (with no long-range order) would be false. The paper explicitly states 'we did not find any other magnetic wave vectors, so there are no FM or AFM orders at T > 16 K' — but this is only supported by zero-field data. This gap is load-bearing because the title and the novelty rest on the absence of magnetic order in the temperature window where the MRM is observed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a magnetoresistive memory (MRM) in the magnetic semiconductor Eu5In2As6. Specifically, below about 30 K — approximately twice the antiferromagnetic transition temperature TN = 16 K — the resistivity depends on the history of an applied magnetic field: ZFC and field-trained ρ(T) curves separate by up to four orders of magnitude, ρ(H) up- and down-sweeps form a reproducible hysteresis loop, and the resistivity relaxes on minute timescales. The authors find no corresponding hysteresis in magnetization, no anomaly in specific heat at 30 K, and no long-range magnetic order above 16 K in zero-field neutron diffraction. They interpret the effect as arising either from magnetic polarons or from a hidden time-reversal-odd order with domain-wall scattering, and suggest the effect may enable memory applications.","tokens_in":20036,"tokens_out":7814,"duration_ms":74071,"significance":"If the central claim is established, this is a significant result: MRM has previously been confined to two strongly correlated metallic systems — perovskite manganites and pyrochlore iridates — and in both cases only in magnetically ordered phases. Demonstrating MRM in a low-carrier semiconductor and in a regime with no detected long-range magnetic order would broaden the phenomenology and motivate searches in related Zintl phases. The paper's strengths are the reproducibility across several crystals, the small angle dependence of the effect, and the explicit controls: current-independence excludes Joule heating, frequency-independent AC susceptibility and time-independent magnetization exclude spin-glass and conventional ferromagnetic hysteresis, and the zero-field neutron data place TN and TN′ well below the MRM onset. These controls make the basic field-history-dependent resistivity effect credible. The main limitations are the absence of field-dependent neutron scattering above TN and the unmeasured low-field untrained state; both are load-bearing for the 'paramagnetic phase' and 'orders of magnitude' claims.","major_comments":[{"comment":"The paramagnetic-phase claim is not established under the same conditions where MRM is measured. Neutron diffraction shows magnetic Bragg peaks only in zero-field temperature scans [Q=(0,0,1/2) and Q=(1,0,0) versus T] and in field sweeps performed at 15 K, i.e. below TN. No search is reported for field-induced magnetic order at T = 18–30 K under the 2–9 T fields where the MRM is observed. The statement 'we did not find any other magnetic wave vectors, so there are no FM or AFM orders at T>16 K' is therefore only a zero-field statement. M(H) at 20 K (Fig. S8f) rules out net ferromagnetic hysteresis but not an antiferromagnetic or SDW order. Because the abstract and title rest on MRM occurring 'well within the paramagnetic phase,' field-dependent neutron scattering (or an equivalent bulk probe) at a representative T>16 K in the MRM field range is needed, or the claims must be explicitly re","section":"Fig. S8 / Methods (Neutron diffraction)"},{"comment":"The quantitative 'orders of magnitude' memory at zero or low field is not directly measured. The Fig. 1b caption states that the untrained ZFC resistivity 'exceeded the lock-in detection limit and was not measurable below H = 1.5 T,' yet the Results describe 'ρ(H=0)' as three orders of magnitude larger in the ZFC curve than in subsequent field sweeps, and Fig. 1a/b draw the green ZFC curves down to H = 0. The MRM contrast at 2 T in Fig. 2a is measured and does support a real memory effect, but the zero-field and low-field claims should be labeled as extrapolated or inferred rather than presented as data, and the figures should mark the unmeasured region clearly.","section":"Fig. 1b caption / Results, pp. 4–5"},{"comment":"The proposed hidden-order model is internally inconsistent with the reported angle independence. The section argues that a time-reversal-odd Ag order parameter coupled as F ∼ b χ HxHyHz can explain the absence of magnetization hysteresis, but then states that 'this coupling indicates that a strong angle-dependence in the hysteresis of magnetoresistance, which is not observed in experiments.' Since the near angle-independence of the MRM is used earlier as evidence that the effect is intrinsic, this contradiction needs to be resolved: either identify a different symmetry-allowed coupling that is angle-independent, or explicitly withdraw the hidden-order explanation and rely on the polaronic/short-range correlated scenario.","section":"Supplementary Sec. 10 (Theoretical Discussion)"}],"minor_comments":[{"comment":"The green ZFC resistivity curve in the right column appears continuous to H = 0, despite the same figure/caption noting the untrained state is unmeasurable below 1.5 T. Use dashed curves or open symbols to distinguish extrapolated regions.","section":"Fig. 1a"},{"comment":"The stretched-exponential form is written as ρ = ρ0[1 − exp(−τ/t)^β]; the standard form is exp[−(t/τ)^β]. Please correct the equation and provide the fitted τ, β, and residuals for the data in Fig. 3a, since only τ = 1.5 min is quoted.","section":"Fig. 3 / Time Dependence"},{"comment":"The data availability statement says data 'will be made available online after publication'; a persistent repository link or DOI should be included for the review process, especially given the heavy reliance on supplementary figures.","section":"Methods / Data availability"},{"comment":"The sentence 'Non-overlapping upsweep and downsweep curves indicate a broken TRS' is stronger than the evidence: hysteresis plus slow relaxation can produce non-overlapping curves in a non-equilibrium state without equilibrium time-reversal-symmetry breaking. Consider rewording to 'history-dependent' or 'effectively broken time-reversal symmetry in the memory state.'","section":"Results, p. 6"},{"comment":"Minor typographical issues: 'New Y ork' in the NIST affiliation, 'leading the observed hysteresis' should be 'leading to', and the reference formatting is inconsistent (e.g., ref. 38 uses lowercase initials).","section":"Affiliations / General"}],"recommendation":"major_revision","confidential_remarks":"The paper reports a credible and reproducible transport phenomenon, and the machine-readable? no — but the experimental controls are thoughtful. The key technical gap is the missing field-dependent neutron search above TN; if the authors can supply those data or convincingly reformulate the claim as 'MRM in the absence of detected zero-field magnetic order,' the paper would be publishable. The hidden-order discussion in the Supplement should also be reconciled with the angle-independence before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is worth a serious look: it reports a reproducible magnetoresistive memory in Eu5In2As6 at temperatures up to ~30 K, well above its 16 K AFM transition, and in a weakly doped semiconductor. That is genuinely new — the two prior MRM families (manganites, pyrochlore iridates) show the effect only below magnetic order. The transport data look careful: multiple samples, angle sweeps, current-dependence controls against Joule heating, AC susceptibility against spin-glass behavior, and a sister compound (Eu5In2Sb6) that shows none of it. The relaxation and training reproducibility are presented clearly.\n\nWhere I'd push back is the title claim that the effect lives in the paramagnetic phase. The neutron data that rule out long-range order above 16 K were taken at zero field. The MRM is measured at 2–9 T. The manuscript never checks whether a magnetic Bragg peak appears or sharpens at 18–30 K under the same fields. If field-induced AFM/SDW order appears there, the memory could be a conventional domain effect and the \"paramagnetic\" framing collapses. That is a load-bearing gap, and a referee should ask for field-dependent neutron diffraction (or another local probe) at T > 16 K before the claim is accepted.\n\nThe symmetry model in the supplement has a similar problem: the proposed Ag hidden order that couples to HxHyHz predicts strong angle dependence, which the experiments do not show. The authors acknowledge this inconsistency but do not resolve it. Since the polaronic short-range scenario remains a viable alternative, the central observation survives, but the hidden-order discussion is currently more speculative than the paper's conclusions suggest.\n\nA minor point: the ZFC resistivity is unmeasurable below 1.5 T, so the most dramatic part of the memory (the low-field resistance spike) is not directly observed. The effect is still visible at higher fields and in the temperature sweeps, so I'm not calling it fatal, just a limitation to state explicitly.\n\nOverall: the core observation looks real and reproducible, the significance is high if the paramagnetic-phase claim holds, and the paper deserves peer review. I'd send it out, with the neutron-field question as the main issue.","headline":"A reproducible MRM above T_N in a doped semiconductor is a real new observation, but the paramagnetic-phase claim needs field-dependent neutron data before it can be accepted.","tokens_in":20579,"tokens_out":3784,"would_cite":true,"duration_ms":33800,"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":"Eu5In2As6 shows magnetoresistive memory starting at twice its antiferromagnetic transition temperature, in the paramagnetic phase.","keywords":["magnetoresistive memory","Eu5In2As6","paramagnetic phase","hidden order","magnetic polarons","resistivity relaxation","colossal magnetoresistance","magnetoelastic coupling"],"falsifier":"Measure the untrained zero-field resistivity with a high-impedance four-probe bridge that can resolve values above the lock-in limit below 1.5 T, and repeat the time-dependence measurements while monitoring the sample thermometer. If the order-of-magnitude ZFC-versus-trained splitting disappears, or the relaxation time matches the cryostat thermal time constant, the intrinsic-memory claim fails. Alternatively, field-trained neutron or resonant X-ray scattering at 30 K could detect the proposed hidden order or rule it out.","tokens_in":19627,"feed_emoji":"🧲","tokens_out":6326,"duration_ms":54392,"temperature":0.7,"pith_summary":"The paper sets out to show that the magnetic semiconductor Eu5In2As6 stores magnetic-field history in its electrical resistance while it is still paramagnetic, more than a factor of two above its antiferromagnetic transition at 16 K. Below about 30 K, zero-field-cooled crystals have much higher resistance than field-trained ones, the field upsweep and downsweep traces fail to overlap, and the resistance relaxes on minute timescales. Neither neutron diffraction, magnetization, nor specific heat finds long-range magnetic order, magnetic hysteresis, or a thermodynamic anomaly in that window, so the memory cannot be blamed on conventional magnetic domains. The paper argues that the effect must come from field-trainable scattering objects—either domain walls of a hidden electronic order or short-range magnetic polarons—and that it turns on together with a resistivity minimum and an accelerated lattice contraction.","feed_headline":"Resistance remembers field history at twice the ordering temperature","feed_subtitle":"History-dependent resistance appears in a paramagnetic semiconductor, with no long-range magnetic order.","key_machinery":"The load-bearing experimental object is the ratio of untrained to trained resistivity as a function of temperature: it is unity above 30 K and rises by orders of magnitude below 30 K, acting as the memory's effective order parameter even though no conventional order appears. The protocol of a training field (the field applied while cooling before measurement) versus a measuring field (the field during the resistivity ramp) defines the distinct memory states. A stretched-exponential fit to the fixed-field resistivity relaxation, with timescales of minutes, is interpreted as retarded growth or proliferation of scattering domain walls. The absence of frequency-dependent AC susceptibility and of","core_discovery":"The central discovery is a magnetoresistive memory (MRM) in Eu5In2As6: below roughly 30 K the resistivity depends on the magnetic-field history, not just the instantaneous field, even though antiferromagnetic order sets in at 16 K and no long-range order exists above it. The zero-field-cooled resistivity exceeds the trained-state resistivity by up to four orders of magnitude at 2 K, the two curves merge above 30 K, the hysteresis loop between field upsweep and downsweep is reversible and reproducible across samples, and the resistivity relaxes on a 1.5–5 minute scale at fixed field. This memory has no counterpart in magnetization, specific heat, or neutron diffraction, and it does not appear","pith_inferences":["Editorial extension: the quickest falsification is to measure the untrained zero-field resistivity below 1.5 T with a high-impedance bridge; the paper admits that this region exceeded its lock-in detection limit, so the most dramatic memory signal is still unmeasured.","Editorial extension: because carrier-poor samples (below ~5 × 10¹⁷ cm⁻³) show no MRM, doping or gating could tune or raise the 30 K onset, potentially extending the effect toward higher temperatures.","Editorial extension: the supplement's symmetry analysis identifies a time-reversal-odd, fully symmetric order parameter that couples to the field only through the triple product HxHyHz; angle-resolved magnetostriction near 30 K could test whether such a multipolar hidden order is actually present.","Editorial extension: the two proposed mechanisms, polaron percolation and hidden-order domain walls, predict different field-history dependence of diffuse neutron scattering, so a field-trained neutron study above 16 K could separate them."],"forward_implications":["MRM no longer appears tied to strongly correlated metals with metal-insulator transitions: Eu5In2As6 is a weakly correlated semiconductor with a 45 meV gap and a carrier density of 10^16–10^18 cm⁻³.","The memory and the resistivity minimum appear together and are both absent in Eu5In2Sb6, implicating hybridization between Eu f-states and extrinsic carriers as the microscopic trigger.","Because magnetization and neutron diffraction show no hysteresis, the resistive memory is decoupled from the bulk magnetic moment; a memory device built on it would be read electrically, with no moving magnetic domains.","The same 30 K onset across five crystals with different carrier densities points to an intrinsic electronic transition rather than a sample-specific inhomogeneity.","A lattice contraction begins at 30 K, so any successful explanation must couple the electronic order or polarons to the lattice, not only to spins."],"fun_headline_variants":["Field-history memory appears in a paramagnetic phase","Resistance remembers field without long-range order","Magnetoresistive memory works at twice the ordering temperature","Paramagnetic semiconductor stores field history in resistance","No magnetic order? Resistance still remembers field"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The central claim stands or falls on the assumption that the history-dependent resistivity and its minute-scale relaxation are intrinsic to Eu5In2As6 and not caused by measurement limits, contact settling, or slow thermal equilibration; the most dramatic untrained-state signal below 1.5 T exceeded the lock-in detection limit and was not directly measured.","fun_headline_variants_meta":{"raw":{"variants":["Field-history memory appears in a paramagnetic phase","Resistance remembers field without long-range order","Magnetoresistive memory works at twice the ordering temperature","Paramagnetic semiconductor stores field history in resistance","No magnetic order? Resistance still remembers field"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000383,"raw_usage":{"total_tokens":1865,"prompt_tokens":742,"completion_tokens":1123,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":486,"completion_tokens_details":{"reasoning_tokens":1052}},"tokens_in":486,"tokens_out":1123,"duration_ms":9944,"temperature":1.0,"reasoning_tokens":1052,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T10:15:30.167614+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the untrained zero-field resistivity with a high-impedance four-probe bridge that can resolve values above the lock-in limit below 1.5 T, and repeat the time-dependence measurements while monitoring the sample thermometer. If the order-of-magnitude ZFC-versus-trained splitting disappears, or the relaxation time matches the cryostat thermal time constant, the intrinsic-memory claim fails. Alternatively, field-trained neutron or resonant X-ray scattering at 30 K could detect the proposed hidden order or rule it out.","supporting_citations":[],"review_version":1}