{"id":"509c36c2-a350-4e7e-a110-5d8d1ae32198","arxiv_id":"2505.05565","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Hubbard excitons in Sr2IrO4 and Sr3Ir2O7 are stable only below the temperature where short-range antiferromagnetic correlations develop, indicating spin-mediated binding.","lead":"This paper reports that excitons in two iridate Mott insulators only form when short-range antiferromagnetic order is present, pointing to a spin-based binding mechanism rather than ordinary Coulomb attraction. The result offers a new way to control excitons through magnetism and tests ideas relevant to high-temperature superconductivity.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The disappearance of the 1 THz line above TN in Sr3Ir2O7 is asserted without a quantitative bound; a red-shifted or broadened Lorentzian could mimic the observed positive Drude-like response, weakening the correlation-binding inference.","rationale":"The reader's weakest assumption was that the 1 THz Lorentzian in Sr3Ir2O7 is an intra-excitonic transition and that its disappearance above TN reflects loss of the bound state rather than broadening, a fitting artifact, or screening. My concern sharpens that point: the disappearance itself is not quantitatively bounded, because no Drude-Lorentz fit with an alternative broad or red-shifted oscillator is reported for T>TN, and the positive 'metallic' response is compatible with such an oscillator located below the low-frequency cutoff or heavily broadened. This is the most load-bearing issue because if a broadened or red-shifted HE persists above TN, the central correlation-binding conclusion loses its key negative control. I agree with the reader's conditional verdict. The paper has independent supporting evidence—the Drude-to-Lorentz spectral weight transfer, the persistence of the response in Sr2IrO4, and the fluence independence—but the absence claim in Sr3Ir2O7 is the linchpin and needs a quantitative model comparison before the mechanism identification can be considered definitive.","tokens_in":15914,"tokens_out":7432,"duration_ms":85960,"concrete_test":"Re-fit the T>TN (e.g., 300 K) Sr3Ir2O7 Δσ1(ω) and Δσ2(ω) spectra with a model that includes the Drude term plus one additional Lorentz oscillator whose center frequency is free over 0.1-1.5 THz and whose width is free; compare via reduced χ² or AIC against the Drude-only model and report a 95% confidence upper bound on the Lorentzian spectral weight at 1 THz. If the broadened or red-shifted oscillator is statistically favored, or if its allowed spectral weight is comparable to the 80 K HE spectral weight, the disappearance claim is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central inference is that Hubbard excitons in Sr3Ir2O7 are stable only below the temperature at which short-range antiferromagnetic correlations develop. This rests on the claim that the 1 THz intra-excitonic Lorentzian vanishes above TN. That claim is not quantitatively established. In the Methods, the authors state that for T>TN 'the Drude model did not fit the data' and therefore 'we did not plot a fit to a Drude-Lorentz model' with data at or above TN; the absence of the HE mode is inferred from a positive, featureless Δσ1 and Δσ2 response. However, a Lorentzian whose center frequency falls below the ~0.8 THz low-frequency cutoff of the THz probe would produce positive and slowly varying Δσ1 and Δσ2 across the measured band, closely resembling the reported 'metallic' response. A heavily broadened oscillator centered near 1 THz could similarly hide its negative σ2 lobe below 1 THz within the noise and leave a positive tail. The frequency-integrated ΔEmax traces in Fig. 3e do not resolve this ambiguity, because anchoring to the peak of the static THz field is sensitive to Drude weight, carrier mobility, and lifetime changes, not solely to Lorentzian spectral weight. Without a model comparison that includes a broadened or red-shifted Lorentzian and reports an upper bound on its spectral weight, the data do not exclude the possibility that the intra-excitonic transition persists above TN in broadened or red-shifted form.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports time-resolved THz spectroscopy measurements on the square-lattice iridates Sr2IrO4 and Sr3Ir2O7, aiming to establish that Hubbard excitons (holon-doublon bound states) are stabilized by antiferromagnetic spin correlations rather than by Coulomb attraction. The authors observe a transient Drude-to-Lorentz crossover in Sr3Ir2O7 below TN, assign the emergent ~1 THz mode to an intra-excitonic transition, and compare its temperature dependence with that of Sr2IrO4. They find that the excitonic response persists up to 300 K in Sr2IrO4 but disappears above TN in Sr3Ir2O7, which they attribute to the rapid loss of short-range AFM correlations in the bilayer compound. Additional fluence-dependent measurements show no detectable redshift or broadening up to 4 mJ/cm2, which the authors interpret as evidence against primarily Coulomb-mediated binding. The main conclusion is that HEs exist only below the temperature scale at which short-range AFM correlations develop.","tokens_in":16249,"tokens_out":3817,"duration_ms":43087,"significance":"If the conclusions hold, this would be an important experimental step toward establishing spin-mediated exciton binding in a solid-state Mott insulator, a mechanism long discussed in the context of strongly correlated systems. The comparative choice of Sr2IrO4 (2D Heisenberg) and Sr3Ir2O7 (3D Ising) is well motivated, and the Drude-to-Lorentz crossover in Fig. 2c is convincing evidence for the transient formation of a bound state in Sr3Ir2O7. The fluence robustness of the intra-excitonic line is a valuable and potentially discriminating observation. The manuscript is clearly written and the experiments are described in sufficient detail to be reproduced. However, the central temperature-dependence claim currently rests on an absence claim that is not quantitatively established, and the frequency-integrated data used to support the correlation-length correspondence are not specific to the Lorentzian spectral weight.","major_comments":[{"comment":"The central inference that HEs disappear above TN in Sr3Ir2O7 rests on the claim that the Lorentzian component becomes undetectable for T>TN. The Methods state that for T>TN the Drude model did not fit the data and therefore no Drude-Lorentz fit is plotted, while Fig. S4 shows only positive, featureless Δσ1 and Δσ2. This absence claim is not quantitatively established: a Lorentzian whose center frequency is redshifted below the ~0.8 THz low-frequency cutoff, or whose linewidth is heavily broadened around 1 THz, would produce positive, slowly varying Δσ1 and Δσ2 across the measured band and could mimic the reported metallic response. I request a model comparison that includes a broadened and/or red-shifted Lorentzian together with Drude-like terms for the T>TN data, and an upper bound on the HE spectral weight as a function of temperature. Without such a bound, the data do not exclude persistence of the intra-excitonic transition above TN in broadened or red-shifted form, so the paper's central conclusion is not yet fully supported.","section":"Methods I.C; Figs. 3c,d and S4"},{"comment":"The frequency-integrated ΔEmax traces in Fig. 3e are used to argue that the excitonic response tracks the AFM correlation length. However, anchoring the EOS gate to the peak of the static THz field makes the measured amplitude sensitive to Drude weight, carrier mobility, and carrier lifetime in addition to Lorentzian spectral weight. The sharp superlinear increase near TN in Sr3Ir2O7 could therefore reflect changes in the transient metallic response rather than the onset of HE oscillator strength. The authors should show that the temperature dependence of ΔEmax is reproduced by the Lorentzian component of the frequency-resolved fits, or otherwise separate the Drude and Lorentz contributions in the temperature-dependent data.","section":"Methods I.A and Fig. 3e"},{"comment":"The comparison between the instantaneous DC conductivity at 80 K (~220 Ω−1cm−1, extrapolated from the Drude component) and the equilibrium DC conductivity at 300 K (<50 Ω−1cm−1) is used to argue that the enhanced equilibrium conductivity above TN cannot explain the loss of the HE. This comparison is suggestive but not conclusive: the extrapolated Drude DC value carries the uncertainty of the Drude-Lorentz decomposition, and the relevant screening comparison is with the actual transient dielectric environment in the paramagnetic phase, not with the equilibrium state at 300 K. Reporting the confidence interval of the extrapolated conductivity and, if possible, a measurement of the transient conductivity at T>TN would strengthen the argument.","section":"Fig. 4a and the paragraph beginning 'Although the Mott gap...'"}],"minor_comments":[{"comment":"Reference [12] is assigned to two different works (C. S. Chiu et al., Science 365, 251 (2019) and O. Mehio et al., Nature Physics 19, 1876 (2023)); the duplicate numbering should be corrected.","section":"References"},{"comment":"There is a typo in 'laregely unaffected'; it should read 'largely unaffected'.","section":"Introduction, last paragraph"},{"comment":"The phrase 'experimetnal extraction' should read 'experimental extraction'.","section":"Supplemental Methods I.B.1"},{"comment":"The caption describes panels (c) and (d), but the figure contains panels (b) and (c); the panel references should be corrected.","section":"Fig. 4 caption"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Omar — quick take on arXiv:2505.05565. The genuinely new thing here is the comparison between Sr2IrO4 and Sr3Ir2O7 across their respective Néel temperatures. Same experimental platform, same probe, and the two materials differ in exactly the way that lets you test whether short-range AFM correlations are what hold the 1 THz intra-excitonic mode together. Sr2IrO4 shows the Lorentzian at 300 K (T/TN = 1.3); Sr3Ir2O7 loses it above its TN. That difference is striking and not present in the earlier Nature Physics paper. The fluence robustness is also a nice addition.\n\nCredit where due: the Drude-to-Lorentz crossover in Fig. 2c is convincing, the fitting is careful about phonon overlap, and the paper is honest about the competing models for Sr3Ir2O7's magnon spectrum. The screening-robustness argument, though indirect, is reasonable.\n\nThe soft spot is exactly what the stress test flags. The claim that the HE 'can only exist below where short-range AFM correlations develop' rests on the absence of a Lorentzian above TN in Sr3Ir2O7, and that absence is not quantitatively bounded. The Methods say the Drude model did not fit the data, so they didn't plot a fit. But a heavily broadened or red-shifted Lorentzian centered below the ~0.8 THz cutoff would yield positive Δσ1 and Δσ2 across the band, looking just like the 'metallic' response. Fig. S4 shows data at several delays, which helps, but still no upper limit on a possible HE spectral weight. The ΔEmax traces in Fig. 3e don't resolve it either, since they anchor to the peak of the static field. This is not fatal — the comparative logic is sound — but it is the load-bearing joint and it needs a quantitative treatment. A two-component fit with and without a broadened HE, reporting an upper bound on its spectral weight, would settle it.\n\nMinor: the photo-carrier density estimate is rough (they assume a0 = 0.4 nm as the smallest Bohr radius), and the fluence study uses only two fluence points per material. Both are minor.\n\nBottom line: this deserves a serious referee. The central claim is likely right, and the experimental design is genuinely clever. Fix the absence claim and it becomes a strong paper. Yes, send to peer review. I'd cite it if I worked in this area.","headline":"A smart comparative tr-TDTS study that makes a strong case for spin-mediated exciton binding, but the key absence claim above TN needs a quantitative bound before the mechanism is nailed.","tokens_in":16792,"tokens_out":2270,"would_cite":true,"duration_ms":23876,"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":"Time-resolved THz spectroscopy shows that the intra-excitonic resonance in two iridate Mott insulators appears only at temperatures where short-range antiferromagnetic correlations exist, implicating spin exchange as the binding agent.","keywords":["Hubbard excitons","antiferromagnetic correlations","Mott insulators","time-resolved THz spectroscopy","holon-doublon pairs","Sr2IrO4","Sr3Ir2O7","spin-mediated binding"],"falsifier":"One decisive test would be to run the same tr-TDTS measurement on a quasi-2D Mott insulator with negligible antiferromagnetic exchange (for example a non-magnetic Mott insulator): if a similar intra-excitonic resonance appeared and survived above any magnetic ordering temperature, the spin-binding claim would be falsified. Alternatively, a high-resolution temperature scan of the 1 THz peak in Sr3Ir2O7 that shows the peak persisting but broadening asymmetrically above TN would indicate the apparent disappearance is a lineshape artifact rather than true dissociation.","tokens_in":15683,"feed_emoji":"🧲","tokens_out":12592,"duration_ms":109858,"temperature":0.7,"pith_summary":"The paper sets out to show that excitons in two-dimensional Mott insulators can be bound by antiferromagnetic spin correlations rather than by the Coulomb interaction alone. Using time-resolved THz spectroscopy on the square-lattice iridates Sr2IrO4 and Sr3Ir2O7, the authors find that a transient intra-excitonic resonance appears only in the temperature range where short-range antiferromagnetic order exists. Because the two compounds have different magnetic critical behavior – 2D Heisenberg in Sr2IrO4, where short-range order persists well above the ordering temperature, and 3D Ising in Sr3Ir2O7, where it collapses at the transition – the correlation between the resonance and the spin environment can be isolated. The excitons also survive photodoping up to densities where Coulomb-bound excitons would have already undergone a Mott transition, which the authors take as further evidence against Coulomb binding as the dominant mechanism. If correct, the result establishes spin-bound Hubbard excitons as real solid-state excitations and opens a route to controlling excitons through magnetic degrees of freedom.","feed_headline":"Spin correlations, not Coulomb force, bind excitons in two iridates","feed_subtitle":"The intra-excitonic peak disappears above the Néel temperature in Sr3Ir2O7 but survives in Sr2IrO4.","key_machinery":"The central object is the Hubbard exciton: a bound state of a holon (empty site) and a doublon (doubly occupied site) in a Mott insulator, stabilized because the pair moving coherently through an antiferromagnet confines the string of flipped spins between them to a finite length. The discriminating tool is the comparative temperature dependence in two materials from the same Ruddlesden–Popper iridate family that share almost identical electronic structure but belong to different magnetic universality classes – 2D Heisenberg (Sr2IrO4) vs 3D Ising (Sr3Ir2O7). The 1–1.5 THz intra-excitonic resonance, tracked by time-resolved THz spectroscopy, serves as the readout; its presence or absence tracks the spin correlation length rather than the thermodynamic order parameter.","core_discovery":"The central claim is that the transient 1 THz (in Sr3Ir2O7) and 1.5 THz (in Sr2IrO4) Lorentzian features observed in the photo-induced THz conductivity are intra-excitonic transitions of Hubbard excitons – bound holon–doublon pairs – whose binding is mediated by the antiferromagnetic exchange interaction. In Sr3Ir2O7, the resonance disappears sharply above the Néel temperature TN = 285 K, tracking the rapid loss of short-range antiferromagnetic correlations; in Sr2IrO4, it persists up to 300 K (T/TN = 1.3) because short-range Heisenberg correlations survive far above TN = 230 K. The authors further show that the excitonic peak does not shift or broaden up to photodoping densities near the expected excitonic Mott transition, and that the photo-excited state at low temperature is more conductive than the equilibrium paramagnetic state above TN, ruling out simple conductivity screening as the cause of the temperature dependence. They conclude that the excitons can only exist where short-range antiferromagnetic correlations develop, so the binding is spin-mediated rather than predominantly Coulombic.","pith_inferences":["A testable extension would be to tune the magnetic exchange in a single compound (e.g., via pressure, chemical doping, or magnetic field) and verify that the intra-excitonic peak frequency and its temperature onset shift accordingly; the paper does not report such a continuous tuning.","If spin-mediated binding dominates, then the same mechanism should bind like-charged carriers (holon–holon or doublon–doublon pairs); the authors note this as a possible route to Cooper pairing, but the current data only address the oppositely charged pair.","The fluence-invariance result suggests that the exciton binding energy is set by the exchange scale rather than by the Coulomb scale, so in materials with larger exchange the excitonic Mott transition density should be correspondingly higher; this could be checked across the iridate family or in other 2D Mott insulators such as the cuprates.","A natural follow-up is to measure the full time-resolved spectrum at the highest fluences and look for the predicted excitonic Mott transition at densities beyond 1e14 cm-2, where the peak would eventually shift or dissociate; the paper reports no sign up to 4 mJ/cm2, leaving the location of that transition open."],"forward_implications":["If spin-exchange binding is the operative mechanism, the intra-excitonic transition frequency in any such Mott insulator should scale with the zone-boundary magnon energy (the exchange scale), a ratio the authors already show holds between Sr2IrO4 and Sr3Ir2O7.","The persistence of the exciton at densities near the predicted excitonic Mott transition implies that spin-bound excitons are far more stable to screening than Coulomb-bound excitons, so magnetically ordered or correlated hosts could support exciton condensates or fluids at higher densities.","The temperature dependence of the excitonic response provides a new, non-equilibrium probe of short-range antiferromagnetic correlations: the appearance threshold of the intra-excitonic peak marks the onset of the correlation length sufficient for binding.","Because the two compounds share nearly identical electronic structure, the contrast in excitonic behavior is attributable to the difference in magnetic universality classes, making the pair a controlled test bed for spin-mediated binding."],"supporting_citations":[{"why":"Supplies the prior tr-TDTS observation of a Hubbard exciton fluid in Sr2IrO4 and the Drude-to-Lorentz analysis method used as the baseline.","marker":"[12]"},{"why":"Defines the transient THz signature of intra-excitonic transitions – a Lorentzian in the real conductivity with dispersive imaginary part – used to identify the resonance.","marker":"[34]"},{"why":"Documents the screening-induced redshift and broadening of Coulomb-bound excitons and the excitonic Mott transition criterion, the contrast for the fluence-invariance result.","marker":"[35]"},{"why":"Provides measured spin correlation lengths of Sr2IrO4 showing that short-range antiferromagnetic correlations persist well above TN.","marker":"[26]"},{"why":"Establishes the 3D Ising universality class of Sr3Ir2O7, so that short-range correlations collapse near TN.","marker":"[32]"},{"why":"Supplies the magnon spectrum and zone-boundary exchange energy of Sr3Ir2O7 used in the ratio comparison and in arguing short-range correlations die out above TN.","marker":"[28]"},{"why":"Provides the magnon spectrum and zone-boundary exchange energy of Sr2IrO4 used to set the expected spin-exchange scale.","marker":"[24]"},{"why":"Supplies the equilibrium DC conductivity of Sr3Ir2O7 at 300 K, the comparison value that rules out a metallic-screening explanation.","marker":"[6]"}],"fun_headline_variants":["Spin, not Coulomb, binds excitons in iridates","Spin correlations bind excitons in iridates","Magnetic bonds, not charge, hold excitons in iridates","Excitons need spin correlations to bind in iridates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the ~1 THz Lorentzian in Sr3Ir2O7 is genuinely an intra-excitonic transition of a holon–doublon pair, and that its disappearance above the Néel temperature reflects dissociation of the bound state rather than a broadening, fitting artifact, or screening effect from increased conductivity.","fun_headline_variants_meta":{"raw":{"variants":["Spin, not Coulomb, binds excitons in iridates","Spin correlations bind excitons in iridates","Magnetic bonds, not charge, hold excitons in iridates","Excitons need spin correlations to bind in iridates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002228,"raw_usage":{"total_tokens":8613,"prompt_tokens":936,"completion_tokens":7677,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":7607}},"tokens_in":552,"tokens_out":7677,"duration_ms":56516,"temperature":1.0,"reasoning_tokens":7607,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:02:11.956984+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"One decisive test would be to run the same tr-TDTS measurement on a quasi-2D Mott insulator with negligible antiferromagnetic exchange (for example a non-magnetic Mott insulator): if a similar intra-excitonic resonance appeared and survived above any magnetic ordering temperature, the spin-binding claim would be falsified. Alternatively, a high-resolution temperature scan of the 1 THz peak in Sr3Ir2O7 that shows the peak persisting but broadening asymmetrically above TN would indicate the apparent disappearance is a lineshape artifact rather than true dissociation.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the prior tr-TDTS observation of a Hubbard exciton fluid in Sr2IrO4 and the Drude-to-Lorentz analysis method used as the baseline."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the transient THz signature of intra-excitonic transitions – a Lorentzian in the real conductivity with dispersive imaginary part – used to identify the resonance."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the screening-induced redshift and broadening of Coulomb-bound excitons and the excitonic Mott transition criterion, the contrast for the fluence-invariance result."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides measured spin correlation lengths of Sr2IrO4 showing that short-range antiferromagnetic correlations persist well above TN."},{"cited_title":"Suwa, S.-S","cited_arxiv_id":null,"evidence_quote":"Establishes the 3D Ising universality class of Sr3Ir2O7, so that short-range correlations collapse near TN."},{"cited_title":"Gretarsson, N","cited_arxiv_id":null,"evidence_quote":"Supplies the magnon spectrum and zone-boundary exchange energy of Sr3Ir2O7 used in the ratio comparison and in arguing short-range correlations die out above TN."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the magnon spectrum and zone-boundary exchange energy of Sr2IrO4 used to set the expected spin-exchange scale."},{"cited_title":"Lenarˇ ciˇ c and P","cited_arxiv_id":null,"evidence_quote":"Supplies the equilibrium DC conductivity of Sr3Ir2O7 at 300 K, the comparison value that rules out a metallic-screening explanation."}],"review_version":1}