{"id":"d1600ae7-8417-4203-a770-ac036cbcd1ff","arxiv_id":"1908.07161","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Photoexcitation of bismuth weakens the nearest-neighbor interatomic force by up to about 50%, driving the softening of the A1g optical mode and transverse acoustic modes.","lead":"Using femtosecond x-ray pulses, researchers extracted the forces between atoms in laser-excited bismuth from the way the atoms vibrate. They found the dimer bonds that hold bismuth's Peierls-distorted structure together weaken dramatically, which explains the partial reversal of the distortion.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim rests on a restricted force-constant model (3 of 21 matrices, fixed eigenvectors, frozen DFPT shells); a synthetic recovery test is needed to show the 50% nearest-neighbor softening is identifiable.","rationale":"The reader identified the fixed-eigenvector assumption as the weakest point, and that is indeed one facet of the problem. My stress test broadens it to the full restricted model space: only three of twenty-one force matrices are varied, the omitted shells are frozen at ground-state DFPT values, and the reduced chi-square of ~300 signals that the model is not statistically complete. The paper itself reports a disagreement with DFPT for the ninth-nearest-neighbor force, which shows that the frozen-shell assumption is not harmless. A synthetic-data recovery test would settle whether the large apparent softening of lambda_1 is an identifiable physical quantity or an artifact of the constrained model. I do not see grounds to reject the paper: the measurement, the time-domain method, and the qualitative conclusion are credible, and the authors are explicit about their assumptions. The concern is quantitative and model-dependent, which supports the reader's conditional verdict rather than changing it.","tokens_in":10555,"tokens_out":9522,"duration_ms":109678,"concrete_test":"Run the published fitting pipeline on synthetic data generated from a realistic excited-state force model (for example, constrained DFT of ref. 17, or the experimental best fit with eigenvectors rotated by 10-20 degrees and/or perturbed outer shells), then refit with (i) all 21 force matrices free under L1 regularization and (ii) eigenvectors of the three largest matrices free. If the recovered nearest-neighbor eigenvalue remains about 50% of the ground-state value and dominates the A1g and TA softening, the claim survives; if it shifts by more than about 20% or another shell takes over, the headline 'primarily due to...' is model-dependent and should be softened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is the identifiability of the fitted force constants. The central claim—that the A1g and TA softenings are primarily caused by a ~50% reduction of the first nearest-neighbor eigenvalue (Fig. 2(a))—is obtained from Eq. (4) by varying only three force matrices, fixing their eigenvectors to the ground-state bond frame, and holding all other shells at their ground-state DFPT values. With a reduced chi-square of about 300, the model does not reproduce the data at the level of the reported frequency uncertainties, and the paper attributes the excess to geometry systematics. But a large chi-square also opens the possibility that the fitted lambda_1 is absorbing model error from the frozen shells or from eigenvector rotation. The paper itself notes disagreement with constrained DFT for the ninth-nearest-neighbor force (Fig. 3(b)), so the frozen-shell assumption is not innocuous; adding IFCs one at a time (Fig. 5(c)) does not rule out correlated changes across shells. If the true excited-state force changes live partly in omitted shells or in rotated eigenvectors, the 'bonding-direction' eigenvalue reported in Fig. 2(a) is a projection artifact rather than a measured force.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports time-resolved x-ray diffuse scattering measurements on photoexcited bismuth at laser fluences of 2.5–8.2 mJ/cm2. Oscillations in the diffuse scattering at twice the phonon frequency are used to extract the momentum-resolved excited-state phonon dispersion, which is then fit with a Born–von Karman model containing 21 independent force matrices. The authors fix 18 matrices to ground-state DFPT values, assume that photoexcitation changes only the eigenvalues (not the eigenvectors) of the force matrices, and fit the eigenvalues of the first-, second-, and ninth-nearest-neighbor force matrices, with an additional constraint from the A1g zone-center mode. Their central finding is that the first nearest-neighbor eigenvalue along the bonding direction is softened to roughly 50% of its ground-state value at 8 mJ/cm2 and that this softening dominates the observed A1g and transverse-acoustic mode softening. They also compare their reconstructed forces with constrained DFT calculations and note a disagreement for the ninth-nearest-neighbor force.","tokens_in":10831,"tokens_out":6461,"duration_ms":60345,"significance":"If correct, this is a significant experimental advance: it is one of the first direct, momentum-resolved measurements of nonequilibrium interatomic forces, moving beyond zone-center optical-mode probes and providing a microscopic picture of the photoinduced partial reversal of the Peierls distortion in bismuth. The manuscript is clearly written and unusually candid about its assumptions and limitations (fixed eigenvectors, reduced chi-square of about 300, geometry-only error bars). The paper also includes convergence tests for the number of fitted force matrices, which is good practice. The main significance risk is that the key quantitative claim—the ~50% nearest-neighbor softening—is the output of a heavily constrained fit whose identifiability has not been demonstrated; the fit quality and error treatment need strengthening before the central claim is fully established.","major_comments":[{"comment":"The fit reaches a reduced chi-square of about 300, which the text attributes to geometry systematics and model incompleteness. This is a load-bearing issue because the central claim is a ~50% reduction of the first nearest-neighbor eigenvalue (Fig. 2(a)). The error bars in Fig. 2 are obtained by varying the crystal alignment, so they do not include model error. The plateau in Fig. 5(c) with increasing number of adjustable IFCs is evidence against adding individual shells, but it does not rule out correlated changes across several omitted shells. I request a synthetic recovery test: generate synthetic frequency maps from models with, e.g., changes in the 4th–5th shells or with rotated eigenvectors, and show that the constrained three-matrix fit recovers the true first-shell eigenvalue within quoted uncertainties. Without such a test, the 50% figure is not established as an identifiable parameter of the data.","section":"Supplement: Fitting Routine for Interatomic Forces; Eq. (4)"},{"comment":"The assumption that the photoexcitation changes only the eigenvalues, not the eigenvectors, of each force matrix is stated explicitly, and it underpins the interpretation of the fitted first eigenvalue as the force along the bonding direction. If the bond directions themselves rotate or the force-matrix eigenvectors change in the excited state, the fitted eigenvalue is a mixture of ground-state components and the extracted bonding-direction softening is biased. The paper does not provide evidence for this assumption, such as a DFPT calculation of the excited-state force matrices showing small eigenvector rotation, or a synthetic test where the fitting is applied to data generated with rotated eigenvectors. This point is not a circularity but a correctness-risk that should be addressed before the central claim can be taken at face value.","section":"Main text, after Eq. (2)"},{"comment":"The observed diffuse-scattering frequency is the second harmonic of the phonon frequency, and each pixel is assigned to a phonon branch by the maximum computed TDS intensity. The text states that this assumes similar excitation amplitude relative to equilibrium for all phonon modes. If the actual excitation amplitudes are mode-dependent, some pixels will be assigned to the wrong branch, and the fitted force eigenvalues will be biased. The paper should test the sensitivity of the fitted forces to branch assignment, for example by repeating the fit with a different TDS weighting threshold or by excluding pixels near branch crossings.","section":"Supplement: Fitting Routine for Phonon Frequencies / branch assignment"}],"minor_comments":[{"comment":"The title contains a stray space: 'Phot oexcited' should be 'Photoexcited'.","section":"Title"},{"comment":"In the sentence 'a 30% reduction in freuency', 'freuency' should be 'frequency'.","section":"Main text, first paragraph of results"},{"comment":"The acronym 'DPFT' appears in the Fig. 3(b) caption and once in the main text; it should be 'DFPT' for consistency.","section":"Fig. 3 caption and main text"},{"comment":"The sentence 'The second harmonic of the LO and TO phonon branches were not observed' should use 'was' instead of 'were', since the subject is 'the second harmonic'.","section":"Main text, paragraph on observed branches"},{"comment":"Equation (4) defines the chi-square with a prefactor 1/(N−nX), but the symbols N and nX are not defined explicitly on first use; please define them in the text or equation.","section":"Supplement: Fitting Routine for Interatomic Forces"}],"recommendation":"major_revision","confidential_remarks":"This is a well-executed experimental paper that fits the journal's scope. I recommend major revision with a focus on demonstrating that the constrained three-matrix fit can recover the reported nearest-neighbor softening from synthetic data generated with plausible model error, rather than requiring new experiments. The paper's candid discussion of limitations is a strength, but the central quantitative claim needs stronger identifiability support."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Mike,\n\nThis paper is the first to extract excited-state interatomic forces in a photoexcited solid across the Brillouin zone, using time-resolved x-ray diffuse scattering in bismuth. The method is a natural extension of equilibrium force-constant fitting to a transient state, and the key qualitative conclusion—that weakening of the nearest-neighbor bonding force drives the partial reversal of the Peierls distortion—is likely correct. The fluence-dependent data and the comparison to DFT add weight.\n\nThe soft spots are where the quantitative claims outrun the model. The reduced chi-square of about 300 means the fit does not reproduce the measured frequencies within their statistical errors; the authors attribute this to geometry systematics, but it also means the fitted force constants are absorbing model error. The error bars shown are only the spread from varying the crystal alignment, not from model uncertainty. More importantly, the assumption that only the eigenvalues, not the eigenvectors, of the force matrices change is strong. If the excited-state bond directions rotate, or if the frozen shells change in a correlated way, the reported ~50% softening of the nearest-neighbor eigenvalue would be a projection artifact rather than a measured force. The paper's own Fig. 3(b) shows the ninth-neighbor force disagreeing with DFT, so the frozen-shell assumption is not innocent. Adding adjustable forces one at a time (Fig. 5) does not rule out correlated changes across shells.\n\nI don't think these problems are fatal. The qualitative picture is robust and the authors are transparent about the limitations. But the specific numbers—especially the extrapolation to 16 mJ/cm2—should be treated as model-dependent. What would strengthen the paper: release the code and data, run a synthetic recovery test to show the 50% softening is identifiable under the model, and test sensitivity to relaxing the fixed-eigenvector assumption.\n\nThis is a serious experimental paper that deserves a real refereeing, not a desk reject. I'd recommend conditional acceptance: the method and qualitative result are valuable, but the quantitative claims need to be framed more cautiously unless the robustness checks come through.","headline":"First experimental transient interatomic forces in bismuth: plausible qualitative result, model-dependent numbers that need robustness checks.","tokens_in":11408,"tokens_out":2822,"would_cite":true,"duration_ms":28922,"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":"This paper reports the first experimental determination of the photoexcited interatomic forces in bismuth, showing that the softening of the A1g and transverse acoustic modes comes primarily from weakening of the nearest-neighbor dimer…","keywords":["photoexcited bismuth","nonequilibrium interatomic forces","time-resolved x-ray diffuse scattering","phonon dispersion","Born-von Karman model","Peierls distortion","A1g mode softening","femtosecond X-ray scattering"],"falsifier":"Extend the measurement to fluences above 8 mJ/cm2 and include a momentum range where the LO and TO branches are observable: the model predicts the nearest-neighbor bond eigenvalue continues falling linearly toward zero near 16 mJ/cm2 while the A1g and transverse acoustic softenings remain proportional to it. Observing a plateau in the bond force, or dispersion changes that require rotating the force-matrix eigenvectors, would settle the claim.","tokens_in":10377,"feed_emoji":"🔬","tokens_out":9074,"duration_ms":89478,"temperature":0.7,"pith_summary":"Photoexcited bismuth is a test case for how light changes the forces between atoms, because its ground state is a Peierls-distorted crystal whose dimer bonds are known to weaken under excitation. This paper measures the transient interatomic forces directly rather than inferring them from a single mode: it fits the photoexcited phonon dispersion, mapped by femtosecond time-resolved x-ray diffuse scattering across the Brillouin zone, to a fifteen-nearest-neighbor Born-von Karman force-constant model. The result is that the observed softening of the A1g optical mode and the transverse acoustic modes comes primarily from a weakening of the nearest-neighbor forces along the bonding direction; at 8 mJ/cm2 that bond force drops to about half its ground-state value and the A1g mode falls from 2.95 to 2 THz. If correct, this ties the long-studied partial reversal of bismuth's Peierls distortion to a specific microscopic force change and demonstrates a way to measure nonequilibrium forces in other light-driven materials.","feed_headline":"Laser pulses cut bismuth's dimer bond force in half","feed_subtitle":"X-ray scattering tracks the softening to the nearest-neighbor bond, driving the Peierls distortion's partial reversal.","key_machinery":"The load-bearing object is the excited-state phonon dispersion extracted from time-resolved x-ray diffuse scattering, converted to forces through a Born-von Karman model: a fixed set of pair-wise interatomic force matrices whose Fourier transform gives the dynamical matrix, with phonon frequencies as the square roots of its eigenvalues. The fit uses the observation that diffuse-scattering intensity oscillations appear at twice the phonon frequency, so each detector pixel yields a phonon frequency at a known wavevector, and branch assignment is done by computing the thermal diffuse scattering intensity. The central constraint is that photoexcitation changes only the eigenvalues of the force matrices, never their eigenvectors, so the adjustable parameters are force eigenvalues per symmetry-inequivalent atom pair, and only the three largest force matrices actually need to be varied. The zone-center A1g frequency is added to the least-squares objective as a constraint, analogous to including a Raman frequency in a ground-state dispersion fit.","core_discovery":"On the paper's own terms, the discovery is that photoexcitation renormalizes bismuth's interatomic force landscape in a highly local way: the largest eigenvalue of the first-nearest-neighbor force matrix, the restoring force along the dimer bond, is weakened to almost 50% of its ground-state value at the highest fluence studied, 8 mJ/cm2. This single force change dominates both the zone-center A1g mode softening and the softening of the transverse acoustic branches that the diffuse-scattering signal is most sensitive to. The second- and ninth-nearest-neighbor forces change only weakly or not at all within uncertainty, so the observed partial reversal of the Peierls distortion is attributed to the near-neighbor dimer bond rather than to a general lattice softening. A linear extrapolation of the bond force to zero lands near 16 mJ/cm2, close to the fluence at which theory expects the electronically driven Peierls distortion to vanish. The measured dispersion softening is stronger than constrained density-functional predictions, especially for the TA mode near the L point.","pith_inferences":["The same diffuse-scattering and force-matrix fitting route should transfer to other Peierls and charge-density-wave systems as long as a sizable fraction of the Brillouin zone shows resolvable frequency shifts; the requirement is many q-resolved frequencies, not just a zone-center mode.","If time-resolved measurements can isolate phonon polarizations or detect off-specular scattering, the fixed-eigenvector assumption could be tested directly; any observed rotation of the bond directions would mean the reported bond-force eigenvalue is a projected rather than a true measure.","The near-50% bond softening at moderate fluence implies strong anharmonicity in the excited-state potential, so measuring fluence-dependent higher-order force constants, such as mode coupling and three-phonon decay, could check whether the harmonic Born-von Karman picture holds deeper into the nonequilibrium state.","The discrepancy with density-functional predictions suggests a benchmark: comparing measured excited-state force matrices against hot-carrier calculations could identify which part of the electronic-structure approximation, such as carrier screening or exchange-correlation, needs revision."],"forward_implications":["The A1g mode's 2.95-to-2 THz softening at 8 mJ/cm2 is dominated by the nearest-neighbor bond force, so time-resolved measurements of that mode can serve as a proxy for the bond force in bismuth.","Extrapolating the measured bond-force drop to zero implies the Peierls-distorted structure should become unstable near 16 mJ/cm2, matching the fluence range where the symmetric high-symmetry phase is expected.","The observed acoustic softening is more pronounced than constrained density-functional predictions, meaning the microscopic force renormalization in the hot-carrier state is stronger than current excited-state calculations capture.","Adding more than three adjustable force matrices does not significantly improve the fit, indicating that the dominant nonequilibrium response is localized on the nearest-neighbor dimer bond."],"supporting_citations":[{"why":"Supplies the Born-von Karman force-constant model and dynamical-matrix relation that turn phonon frequencies into interatomic forces.","marker":"[7]"},{"why":"Established time-resolved x-ray diffuse scattering as a probe of transient phonon dispersion; this is the measurement method the paper extends to forces.","marker":"[13]"},{"why":"Provides the experimental configuration and two-phonon squeezing analysis for bismuth that the present data collection follows.","marker":"[16]"},{"why":"Gives constrained density-functional predictions of excited-state forces and dispersion that the measured softening is compared against and found to exceed.","marker":"[17]"},{"why":"Predicts the electronically driven Peierls distortion in bismuth and the fluence scale for its vanishing, used to interpret the extrapolated zero of the bond force.","marker":"[18]"},{"why":"Supplies the independently estimated fluence at which the excited-state lattice becomes symmetric, anchoring the extrapolation to roughly 16 mJ/cm2.","marker":"[21]"},{"why":"Provides the linear-prediction fitting algorithm used to extract damped oscillatory phonon frequencies from each detector pixel.","marker":"[26]"},{"why":"Gives the thermal diffuse scattering intensity calculation used to assign each observed frequency to a phonon branch.","marker":"[29]"}],"fun_headline_variants":["Laser pulse halves bismuth's dimer bond force","Photoexcitation halves bismuth's nearest-neighbor force","Dimer bond force in bismuth cut in half by laser","Laser light weakens bismuth dimer bond by half","X-ray scattering reveals bismuth's bond force halves"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fit assumes that laser excitation leaves the directions of the chemical bonds unchanged, changing only the strength of the restoring forces along those directions; if the bond directions rotate or the force-matrix eigenvectors renormalize, the reported bonding-direction force is a biased measure.","fun_headline_variants_meta":{"raw":{"variants":["Laser pulse halves bismuth's dimer bond force","Photoexcitation halves bismuth's nearest-neighbor force","Dimer bond force in bismuth cut in half by laser","Laser light weakens bismuth dimer bond by half","X-ray scattering reveals bismuth's bond force halves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000646,"raw_usage":{"total_tokens":2925,"prompt_tokens":861,"completion_tokens":2064,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":477,"completion_tokens_details":{"reasoning_tokens":1979}},"tokens_in":477,"tokens_out":2064,"duration_ms":15139,"temperature":1.0,"reasoning_tokens":1979,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:24:08.498935+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extend the measurement to fluences above 8 mJ/cm2 and include a momentum range where the LO and TO branches are observable: the model predicts the nearest-neighbor bond eigenvalue continues falling linearly toward zero near 16 mJ/cm2 while the A1g and transverse acoustic softenings remain proportional to it. Observing a plateau in the bond force, or dispersion changes that require rotating the force-matrix eigenvectors, would settle the claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Born-von Karman force-constant model and dynamical-matrix relation that turn phonon frequencies into interatomic forces."},{"cited_title":"Inelastic x-ray scattering from phonons,","cited_arxiv_id":null,"evidence_quote":"Established time-resolved x-ray diffuse scattering as a probe of transient phonon dispersion; this is the measurement method the paper extends to forces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the experimental configuration and two-phonon squeezing analysis for bismuth that the present data collection follows."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives constrained density-functional predictions of excited-state forces and dispersion that the measured softening is compared against and found to exceed."},{"cited_title":"Murray, S","cited_arxiv_id":null,"evidence_quote":"Predicts the electronically driven Peierls distortion in bismuth and the fluence scale for its vanishing, used to interpret the extrapolated zero of the bond force."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the independently estimated fluence at which the excited-state lattice becomes symmetric, anchoring the extrapolation to roughly 16 mJ/cm2."},{"cited_title":"Herrmann, S","cited_arxiv_id":null,"evidence_quote":"Provides the linear-prediction fitting algorithm used to extract damped oscillatory phonon frequencies from each detector pixel."},{"cited_title":"Henighan, M","cited_arxiv_id":null,"evidence_quote":"Gives the thermal diffuse scattering intensity calculation used to assign each observed frequency to a phonon branch."}],"review_version":1}