{"id":"edf36c26-00bd-4739-81a4-6ef775bb8e86","arxiv_id":"2608.07041","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In twisted MoTe2, optically created magnetic domains in Chern insulators relax orders of magnitude slower than in metals, via thermal melting rather than edge shrinking.","lead":"Researchers used a laser pulse to flip a small patch of spins in a twisted MoTe2 ferromagnet and watched it relax back. They found that Chern insulating states relax far more slowly than metallic states, and by a different mechanism.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'melting via thermal activation' mechanism is never directly tested: no temperature dependence of τ_relax is reported, so the central mechanism rests on an untested Arrhenius assumption.","rationale":"I read the paper in good faith. The experimental core is solid: the slower relaxation near ICI/FCI is reproduced on multiple devices and spots, the pump-power and pump-size controls are thorough, and the plateau in metallic traces followed by its absence in Chern insulators is a convincing qualitative signature of different relaxation pathways. The zero-field doughnut-beam domain-writing experiment provides additional, independent support for the distinction. What is less secure is the mechanistic interpretation. The abstract and central claim assert that Chern domains 'melt via thermal activation,' but no temperature-dependent relaxation data are presented. The theory section invokes a negative spin stiffness κ2<0 obtained from a simplified mean-field Hartree-Fock treatment (Methods Sec. 14) and then extends the same mechanism to the fractional Chern insulator by analogy, explicitly noting the mean-field treatment cannot capture FCI physics. The reader's weakest assumption identified the model dependence and the lack of direct domain-wall resolution; my concern is sharper and complementary: even granting the model, the 'thermal activation' label is an interpretation that the paper never tests directly. An Arrhenius measurement would settle whether the slow homogeneous decay is genuinely thermal or is instead due to an athermal process. Because the experimental observations themselves are not in question, the appropriate verdict remains conditional rather than rejection: the paper should be accepted only if the thermal-activation mechanism is either directly verified by temperature dependence or the claims are softened to describe an athermal 'homogeneous melting' pathway. My concrete test addresses this directly without requiring sub-diffraction imaging.","tokens_in":31458,"tokens_out":9952,"duration_ms":99025,"concrete_test":"Measure τ_relax(T) at fixed ν = −1 and ν = −2/3 and fixed B/Bc = 1.2 (and, as a control, at B/Bc = 3.0) for T = 1.6, 2.1, 2.5, 3.4, and 5 K, the temperatures already used in Extended Data Fig. 14f. Fit ln τ_relax vs 1/T. If the slope gives a positive activation energy Δvac consistent with the mean-field potential and the data are linear in 1/T, the thermal-melting claim is directly supported. If τ_relax changes by less than ~2× over this T range, the mechanism is not thermal activation and the central claim must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is not just that Chern domains relax slowly, but that they 'melt via thermal activation' while metals shrink. The only evidence that the ICI/FCI decay is thermally activated is the exponential time traces and a mean-field model with a fitted negative stiffness κ2<0; the paper reports no measurement of τ_relax as a function of temperature. The authors already vary T from 1.6 to 5 K to measure Bc(T) (Extended Data Fig. 14f), so the experiment is feasible. If the decay is thermal activation over a barrier Δvac, τ_relax should follow Arrhenius, ln τ ∝ Δvac/kBT. A flat or weakly T-dependent τ would implicate an athermal process (quantum tunneling, disorder creep, or optical-induced relaxation) and would invalidate the 'thermal activation' part of the mechanism even if the negative-stiffness model is correct. This concern is load-bearing because it targets the mechanism, not just the timescale, and it does not depend on resolving the domain-wall structure. The FCI extension, which the theory explicitly cannot capture (Methods Sec. 14), makes the lack of an Arrhenius test even more consequential.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports time- and space-resolved magneto-optical experiments on twisted MoTe2 bilayers. A focused circularly polarized pump pulse creates a local spin-valley domain opposite to the external field, and a weak cross-polarized probe monitors its decay. The authors find that near both the integer Chern insulator at ν≈−1 and the fractional Chern insulators at ν≈−2/3 and −3/5, relaxation is orders of magnitude slower than in ferromagnetic metals at low B/Bc, and the decay traces remain nearly exponential and independent of the initial domain size. In metallic phases, by contrast, the decay shows an initial plateau whose duration grows with the pump-beam diameter. The paper interprets this as Chern-insulator domains relaxing by homogeneous thermally activated 'melting,' while metallic domains collapse by perimeter shrinking. A time-dependent Ginzburg–Landau (model A) simulation, with the potential and stiffness taken from a mean-field Hartree–Fock treatment of an extended Haldane model, reproduces the plateau/no-plateau dichotomy and the threshold behavior of domain writing with a doughnut beam.","tokens_in":31695,"tokens_out":5412,"duration_ms":55316,"significance":"If the microscopic mechanism is correct, the paper establishes a qualitatively new phenomenon: band topology and strong correlations change the far-from-equilibrium relaxation pathway of ferromagnetic domains, with much longer spin lifetimes in Chern insulators. The experimental execution is careful and explicit: multiple devices and spots, probe-power non-perturbation checks, power- and size-dependent controls, and consistency across B/Bc. The theoretical work provides a concrete mean-field picture (negative spin stiffness and sharp domain walls near the ICI) that is falsifiable. The main weakness is that the central 'thermal activation' mechanism is inferred rather than directly tested, and the fractional Chern insulator part is explicitly outside the microscopic theory. These gaps are load-bearing for the mechanism claim, although they do not undermine the experimental phenomenology itself.","major_comments":[{"comment":"The central mechanism, thermally activated melting, is not tested by temperature dependence. All relaxation times τ_relax are reported at the base temperature T≈1.6 K, while the model's thermal rate is Γ_thermal∼Γ exp(−Δvac/kBT). The authors already vary T from 1.6 K to 5 K to measure Bc(T) (Extended Data Fig. 14f), so a temperature sweep of τ_relax at fixed ν and B/Bc is experimentally feasible. An Arrhenius dependence ln τ ∝ Δvac/kBT would support thermal activation; a flat or weakly T-dependent τ would implicate an athermal channel (disorder creep, quantum tunneling, or light-induced relaxation) and would invalidate the 'thermal activation' part of the mechanism even if the negative-stiffness model is otherwise correct. Please either provide this measurement or explicitly soften the mechanism claim to 'consistent with thermal activation.'","section":"Out-of-equilibrium spin-valley dynamics; Methods Secs. 14–15"},{"comment":"The mechanism for the fractional Chern insulator is asserted by analogy, not derived. Methods Sec. 14 states that the mean-field treatment 'cannot capture fractional Chern insulators directly' and that a similar domain-wall structure is 'expected.' Since the abstract and main text claim the melting mechanism applies 'in the vicinity of both integer and fractional Chern insulating states,' the FCI part of the mechanism is not backed by the microscopic calculation. The experimental traces at ν≈−2/3 and −3/5 are valid observations, but the theoretical explanation for them is an extrapolation. Please provide an FCI-capable treatment or explicitly restrict the microscopic mechanism claim to the ICI and present the FCI behavior as an empirical similarity.","section":"Methods Sec. 14; main-text claim around Fig. 4"},{"comment":"The 'melting' interpretation is inferred, not directly observed. The time-resolved observable is the average magnetization in a central diffraction-limited spot; internal domain nucleation is never imaged in time, and Methods Sec. 14 acknowledges that the spatial resolution cannot resolve the domain walls. The plateau in the metallic case is strong evidence for perimeter shrinking, but the absence of a plateau in the ICI could also arise from other spatially distributed decay channels, such as disorder-dominated local relaxation or a radially varying initial preparation profile, rather than uniquely from thermal nucleation of internal domains. A quantitative likelihood argument distinguishing these alternatives, or a spatially resolved time series showing homogeneous decay, would materially strengthen the central mechanism claim.","section":"Mechanisms of Spin-Valley Relaxation; Fig. 3"},{"comment":"The paper states that the filling-factor dependence of τ_relax 'correlates with the height of the energy barrier Δvac,' but Δvac is a theoretical quantity computed from the mean-field model, not independently measured. Because the same model is used to interpret the timescales, the correlation is not an independent experimental confirmation. Please state explicitly that Δvac in Fig. 2g is a model output, and consider showing the comparison between the model's predicted τ_relax(B/Bc,ν) and the experimental values, with the unknown scale Γ stated as such.","section":"Methods Sec. 16; Fig. 2g"}],"minor_comments":[{"comment":"The text refers to a higher-order gradient term 'κ4|∇m|2,' but the differential equation contains κ4∂x^4 m; a term of the form κ4|∇m|2 would be redundant with κ2|∇m|2. Please correct the notation (likely κ4|∇²m|² or the explicit fourth-order term).","section":"Methods Sec. 14, Eq. (10)"},{"comment":"The shrinking velocity is first written as v_B = Γ Δf κ2/σ with σ ∝ √κ2, and later as v_B ∝ B_z √κ2; the field dependence of Δf should be stated so that the two expressions are transparently consistent.","section":"Methods Sec. 16"},{"comment":"The probe-pulse duration used for the relaxation measurements is not specified. Since the extraction of τ_relax assumes that the probe window is long enough to reach steady state, please give the typical pulse durations and state how the exponential fits are affected in the slow-relaxation (low B/Bc) regime.","section":"Fig. 2g; Methods Sec. 8"},{"comment":"The caption uses 'Expanded Beam' and 'Diffraction limited Beam' but does not specify which panel labels (a–d) correspond to which beam configuration beyond the text; a direct mapping would improve readability.","section":"Extended Data Fig. 11 caption"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is an impressive experimental study with strong internal controls, and the reported phenomenology is likely to be of broad interest. In my view the decisive issue is the absence of a temperature-dependent relaxation measurement: without it, the headline 'thermal activation' mechanism is an assumption rather than a tested conclusion. If the authors can supply that measurement or appropriately restrict the claim, the paper would be suitable for publication; I do not see grounds for rejection at this stage."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth a serious referee. The experiment is the strong part: it clearly shows that optically created spin domains in t-MoTe2 relax by different routes in ferromagnetic metals versus nearby Chern insulators, with the metal showing a size-dependent plateau (domain shrinking) that is absent in the ICI/FCI, and with orders-of-magnitude longer relaxation times in the Chern cases. Multiple devices, spots, power checks, and non-perturbative probe tests back this up. The doughnut-beam writing experiment, where metallic domains appear only above a threshold power while Chern domains grow smoothly, is a nice independent confirmation and matches the Allen-Cahn expectation.\n\nThe soft spot is the mechanism label. The abstract says Chern domains 'melt via thermal activation,' but the paper never measures tau_relax as a function of temperature. The exponential time traces and a mean-field model with a negative fitted stiffness kappa2 < 0 are suggestive, but the authors already vary T from 1.6 to 5 K to measure Bc(T) (Extended Data Fig. 14f), so an Arrhenius test is within reach. If tau_relax turns out to be only weakly T-dependent, an athermal process (disorder creep, optical-induced relaxation, or tunneling) would be in play, and the 'thermal activation' part of the mechanism would need revising, even if the shrinking-versus-homogeneous distinction survives. This is a load-bearing oversight for the abstract's central claim, though not for the core experimental findings.\n\nTwo lesser points. The FCI case is assumed by analogy because the mean-field treatment does not capture FCI physics; the authors say this in Methods Sec. 14. And the domain walls are not directly resolved; they also say this. Neither is fatal, but they do mean the proposed microscopic explanation (negative stiffness, sharp walls from chiral edge modes) is currently an interpretation, not a demonstrated cause.\n\nWho gets value: anyone working on moire Chern insulators, optical control of magnetism, or nonequilibrium collective spin dynamics. The experimental core is solid and the phenomenon is new. Send it to peer review; ask the authors to either measure tau_relax(T) or scale back the 'thermal activation' wording to 'activated-like relaxation.'","headline":"Solid experimental paper on domain relaxation in t-MoTe2; the shrinking-vs-melting distinction holds up, but 'thermal activation' is asserted, not directly measured.","tokens_in":32257,"tokens_out":2119,"would_cite":true,"duration_ms":19951,"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":"Optically written magnetic domains in Chern insulators decay by thermal melting in place, not by shrinking, giving spin-valley relaxation times orders of magnitude longer than in ferromagnetic metals.","keywords":["twisted MoTe2 bilayers","spin-valley dynamics","Chern insulator","fractional Chern insulator","magnetic domain relaxation","negative spin stiffness","domain walls","optical spin control"],"falsifier":"Direct imaging of a magnetic domain wall near ν = −1 with sub-100-nm resolution (for example, a scanning magnetometry probe) would settle the mechanism: if the wall is broad and tanh-like, or if relaxation traces acquire an initial plateau when the initial domain is enlarged, then shrinking is not suppressed and the negative-stiffness mechanism fails.","tokens_in":31277,"feed_emoji":"🧲","tokens_out":5527,"duration_ms":52821,"temperature":0.7,"pith_summary":"The paper asks how topology and strong correlations change the way a ferromagnet relaxes after being knocked out of equilibrium. It claims that in twisted MoTe2 bilayers, a laser-written magnetic domain embedded in an integer or fractional Chern insulating state decays by a thermal melting process, meaning local spin flips over an energy barrier, whereas in a nearby ferromagnetic metal the same domain shrinks inward from its edge. The difference in mechanism shows up as a size-dependent plateau in the metallic relaxation traces that is absent in the insulators, and as orders-of-magnitude longer spin lifetimes in the Chern states. If the claim is right, the topological character of a band determines not just static order but the very pathway by which collective spin order decays far from equilibrium.","feed_headline":"Chern ferromagnets melt in place; metals shrink","feed_subtitle":"Near integer and fractional Chern states, optically written spin domains relax orders of magnitude slower.","key_machinery":"The central object is the magnetic domain wall, studied through stochastic time-dependent Ginzburg-Landau / Model A dynamics for a non-conserved Ising order parameter, with free energy f[m] = V[m] + κ2|∇m|^2 + ... . The potential V[m] and the stiffness κ2 come from a mean-field lattice model of the two topmost hole bands of twisted MoTe2. In the ferromagnetic metal κ2 is positive, giving wide tanh-shaped walls; near the Chern insulator a self-consistent Hartree-Fock calculation yields κ2 < 0 with a stabilizing κ4 > 0, giving very sharp walls attributed to topologically protected chiral edge modes. Sharp walls suppress domain shrinking, leaving thermal activation over the barrier Δvac between false and true vacuum as the dominant relaxation channel.","core_discovery":"The paper reports that a focused circularly polarized light pulse can create a metastable micrometer-sized magnetic domain in twisted MoTe2 bilayers, polarized against the external magnetic field, and that the domain's decay is qualitatively different depending on whether the host is a ferromagnetic metal or a Chern insulator. In the metal, the domain collapses by shrinking inward from its perimeter, producing a decay trace with a plateau whose duration grows with initial domain size. In the integer and fractional Chern insulators, the domain instead decays by thermally activated local spin flips that melt it in place, producing size-independent exponential traces and relaxation times orders of magnitude longer. The authors trace this to the domain-wall structure: metallic walls are broad tanh-like walls with positive spin stiffness, whereas near the Chern insulator the spin stiffness is negative and the walls are sharp, suppressing domain-edge motion.","pith_inferences":["If the negative stiffness is a general consequence of chiral edge states on domain walls, the same melting-versus-shrinking distinction should appear in other flat-band Chern ferromagnets, not only in the MoTe2 family; this is a testable prediction for other moiré systems.","The measured relaxation time could serve as a non-invasive probe of the activation barrier Δvac, since τrelax ∼ Γ−1 exp(Δvac/kBT), so sweeping the magnetic field would map the barrier directly even where it cannot be resolved optically.","Sub-diffraction imaging of the domain-wall profile, which the paper notes its optical maps cannot resolve, would be the direct test of whether near-ν = −1 walls are genuinely sharp and negative-stiffness in origin.","For fractional Chern insulators the mechanism is assumed by analogy because the mean-field treatment cannot capture fractionalization; a future theory including fractionalized edge modes would show whether the protection is stronger or weaker than at the integer state."],"forward_implications":["At magnetic fields just above the coercive field, optically written spin domains in integer and fractional Chern insulators outlive metallic domains by orders of magnitude because the shrinking channel is suppressed.","Relaxation traces in Chern insulators remain exponential and independent of initial domain size, whereas metallic traces show an initial plateau whose duration scales with the initial domain diameter.","At high fields, roughly B/Bc > 2–3, the energy barrier Δvac is closed and all phases relax on similar tens-of-microseconds timescales set by intervalley spin processes.","At zero field, small seed bubbles are stable in Chern insulators but unstable in ferromagnetic metals, producing an abrupt power threshold for writing metallic domains and a smooth power dependence for Chern insulating domains."],"supporting_citations":[{"why":"Predicts sharp domain walls and a negative lowest-order spin stiffness in flat-band Chern ferromagnets, the theoretical foundation of the melting-versus-shrinking mechanism.","marker":"[12]"},{"why":"Supplies the extended Haldane model and hopping parameters whose mean-field solution yields the free-energy potential and domain-wall profiles.","marker":"[62]"},{"why":"Defines the stochastic Model A dynamics used to simulate non-conserved magnetization relaxation of magnetic domains.","marker":"[48]"},{"why":"Provides the Allen-Cahn coarsening law for zero-field shrinking of metallic domains used in the bubble-stability analysis.","marker":"[49]"},{"why":"Establishes that resonant attractive-polaron excitation with circularly polarized light flips the spin-valley state in twisted MoTe2, enabling optical domain creation.","marker":"[13]"},{"why":"Demonstrates optical control over the spin-valley state and topological Chern number in moiré materials, supporting the local spin-flip writing mechanism.","marker":"[14]"},{"why":"Shows optical switching of a moiré Chern ferromagnet, corroborating that spin orientation proceeds through local spin-flip events.","marker":"[15]"},{"why":"Reports fractional quantum anomalous Hall states in twisted MoTe2 and identifies the fractional Chern insulating fillings used in the experiments.","marker":"[2]"}],"fun_headline_variants":["Spin domains melt in Chern bands, shrink in metals","Topology flips ferromagnet decay: melt vs shrink","Chern insulators melt ferromagnet domains; metals shrink","Out-of-equilibrium ferromagnets: Chern melts, metal shrinks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sharp-wall, negative-stiffness picture of the Chern domain comes from a mean-field lattice calculation and is not directly resolved in the experiment; the fractional Chern case is presumed to behave likewise by analogy, because the mean-field treatment cannot describe fractionalization.","fun_headline_variants_meta":{"raw":{"variants":["Spin domains melt in Chern bands, shrink in metals","Topology flips ferromagnet decay: melt vs shrink","Chern insulators melt ferromagnet domains; metals shrink","Out-of-equilibrium ferromagnets: Chern melts, metal shrinks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000243,"raw_usage":{"total_tokens":1504,"prompt_tokens":895,"completion_tokens":609,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":511,"completion_tokens_details":{"reasoning_tokens":537}},"tokens_in":511,"tokens_out":609,"duration_ms":5767,"temperature":1.0,"reasoning_tokens":537,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T15:52:33.518787+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Direct imaging of a magnetic domain wall near ν = −1 with sub-100-nm resolution (for example, a scanning magnetometry probe) would settle the mechanism: if the wall is broad and tanh-like, or if relaxation traces acquire an initial plateau when the initial domain is enlarged, then shrinking is not suppressed and the negative-stiffness mechanism fails.","supporting_citations":[{"cited_title":"4b, we first prepare the device at a given filling factor and orient the hole spins upwards using a positive magnetic fieldB= 0.2 T","cited_arxiv_id":null,"evidence_quote":"Predicts sharp domain walls and a negative lowest-order spin stiffness in flat-band Chern ferromagnets, the theoretical foundation of the melting-versus-shrinking mechanism."},{"cited_title":"valley noise","cited_arxiv_id":null,"evidence_quote":"Supplies the extended Haldane model and hopping parameters whose mean-field solution yields the free-energy potential and domain-wall profiles."},{"cited_title":"Regnault and B","cited_arxiv_id":null,"evidence_quote":"Defines the stochastic Model A dynamics used to simulate non-conserved magnetization relaxation of magnetic domains."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Allen-Cahn coarsening law for zero-field shrinking of metallic domains used in the bubble-stability analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that resonant attractive-polaron excitation with circularly polarized light flips the spin-valley state in twisted MoTe2, enabling optical domain creation."},{"cited_title":"As argued in Ref","cited_arxiv_id":null,"evidence_quote":"Demonstrates optical control over the spin-valley state and topological Chern number in moiré materials, supporting the local spin-flip writing mechanism."},{"cited_title":"At zero magnetic field, the free energy potential has two equivalent minima at ⟨Sz⟩=±1","cited_arxiv_id":null,"evidence_quote":"Shows optical switching of a moiré Chern ferromagnet, corroborating that spin orientation proceeds through local spin-flip events."},{"cited_title":"Exfoliation of hBN and graphite was carried out under ambient conditions, whereas MoTe2 monolayers were obtained inside an N 2- filled glovebox to preserve their high quality","cited_arxiv_id":null,"evidence_quote":"Reports fractional quantum anomalous Hall states in twisted MoTe2 and identifies the fractional Chern insulating fillings used in the experiments."}],"review_version":1}