{"id":"53df7013-0e41-4a46-9efb-31b4960144e7","arxiv_id":"2501.11615","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Time-resolved THz-SNOM spectra show that InP photoconductivity decay is governed by electron diffusion and surface band-bending drift, while GaAs behaves as a simple exponentially decaying photoexcited layer.","lead":"Using terahertz near-field microscopy, the authors measured how photoexcited electrons decay in GaAs and InP wafers on picosecond timescales. In InP, diffusion and surface band bending push electrons away from the surface, an effect not seen in GaAs.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"InP band-bending claim may be an artifact of neglecting continuous ambipolar diffusion: with D=10 cm2/s, carriers spread ~1 µm in 600 ps, yet the model fixes the top layer at 350 nm and only expands it at 10 ns.","rationale":"The reader's weakest assumption focuses on the effective tip length L in the finite-dipole model, and that is a real concern: L varies by a factor of 2-5 across harmonics and samples, indicating the forward model is imperfect. But the finite-dipole L issue is primarily a calibration problem; it could in principle be resolved by better tip modeling or by treating L as a nuisance parameter. The more load-bearing weakness for the specific InP claim is the physical profile model used to interpret the data. The paper's novelty is the inference that band-bending drifts electrons away from the InP surface. To make that inference, one must show that ordinary ambipolar diffusion of a shallowly excited carrier profile cannot explain the observed faster decay of the near-field signal relative to the integrated far-field signal. The manuscript does not do this: it fixes the top layer at 350 nm for delays below 1 ns and only lets it expand at 10 ns, even though the diffusion length at intermediate delays is already comparable to or larger than the absorption depth. Thus the comparison in Fig. 13 conflates diffusion with drift, and the fitted n_exc decay could be largely a diffusion effect. This is an internal inconsistency in the modeling strategy, not merely a disagreement with external consensus. The paper deserves credit for the sensitivity and error-ellipse analysis, for the honest reporting of L discrepancies, and for the GaAs consistency checks; those are useful independent contributions. Nevertheless, the central InP claim should not be accepted as quantitative until a diffusion-inclusive profile is fitted and shown to require an extra drift term. The reader's CONDITIONAL verdict remains appropriate, so I mark the verdict UNCHANGED, but the condition should be extended to explicitly require this diffusion-inclusive re-analysis.","tokens_in":14900,"tokens_out":10855,"duration_ms":125376,"concrete_test":"Re-analyze the InP data with a time-dependent depth profile obtained by numerically solving the 1D ambipolar diffusion equation (D = 10 cm2/s, initial exponential profile with 350-nm decay length, stationary background fitted from the negative-delay spectra) and the same finite-dipole forward model (Eqs. 1-2) with L fixed to one value for all harmonics. Fit only the injected density (and possibly D as a check) to the measured S(2) spectra and the far-field transmittance. If the diffusion-only model reproduces the observed faster near-field decay within experimental error, the band-bending drift term is not required and the abstract's central claim is unsupported; if the fit degrades significantly and requires an additional surface-drift term, the claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim about InP requires distinguishing drift due to band-bending from ordinary diffusion. In Sec. 5 the photoexcited region is modeled as a two-layer box: the top layer thickness is fixed at dexc = 350 nm for all pump-probe delays below 1 ns and only allowed to expand to d10ns = 1.4 µm at 10 ns (Fig. 12). But with the stated ambipolar diffusion coefficient D = 10 cm2/s [19], the diffusion length in 600 ps is sqrt(2Dt) ≈ 1.1 µm, i.e., already three times the assumed top-layer thickness; by 800 ps it is ~1.3 µm, close to the fitted 10-ns value. A continuously diffusing shallow profile would reduce the near-surface density and make the SNOM signal S(2) decay faster than the far-field integrated density even in the complete absence of band bending. The observation that S(2) decays faster than far-field transmittance (Sec. 5, Fig. 13) is therefore not by itself 'strong evidence of a force pushing electrons away from the surface'; it is the expected signature of diffusion of a layer initially only 350 nm deep. The paper's own GaAs justification for neglecting diffusion (Sec. 4.1, verified insensitivity to layer thickness) is not repeated for InP, where the diffusion length is larger relative to the absorption depth. The additional inconsistencies (n_long = 3.3 vs 5.2×10^16 cm^-3 for 2nd vs 3rd harmonic; L = 0.37 µm in InP vs 0.87/1.9 µm in GaAs) show the simplified model is overparameterized, so the retrieved top-layer density decay, the direct evidence for drift, is not robust.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports THz scattering-type SNOM measurements of photoexcited GaAs and InP wafers, combined with a finite-dipole model to extract complex photoconductivity spectra at ultrafast pump-probe delays. It develops three normalization strategies (S(m), X(m,m'), and a self-referenced harmonic ratio), analyzes their sensitivity and parameter-retrieval accuracy via error ellipses, and applies them to the two semiconductors. For GaAs, the retrieved electron density and scattering time are consistent between the 2nd and 3rd harmonics. For InP, the near-field signal S(2) decays faster than the far-field transmittance, which the authors interpret as electron diffusion away from the surface combined with drift due to surface band-bending, supported by a two-layer model with a long-lived carrier population.","tokens_in":15287,"tokens_out":5237,"duration_ms":55258,"significance":"If the InP interpretation is correct, the paper demonstrates that THz-SNOM spectra can access depth-dependent carrier transport that far-field THz spectroscopy averages over, and it provides a useful general framework for analyzing such measurements. The sensitivity analysis with error ellipses and the critical discussion of effective tip length are valuable methodological contributions. The GaAs results are robust and give confidence in the experimental and fitting procedures. However, the central InP claim is not yet quantitatively established, because the simplified two-layer model does not exclude ordinary ambipolar diffusion as the sole cause of the faster near-field decay, and the model parameters show inconsistencies across harmonics and samples.","major_comments":[{"comment":"The two-layer model fixes the top-layer thickness at dexc = 350 nm for all delays shorter than 1 ns and only allows it to expand to 1.4 µm at 10 ns. With the stated ambipolar diffusion coefficient D = 10 cm^2/s [19], the diffusion length sqrt(2Dt) reaches about 1.1 µm at 600 ps, already three times the assumed top-layer thickness. A continuously diffusing shallow profile would reduce the near-surface density and make S(2) decay faster than the far-field integrated density even in the complete absence of band bending; the faster decay of S(2) compared with the far-field transmittance (Fig. 13) is therefore not by itself strong evidence of a force pushing electrons away from the surface. The authors should either include continuous ambipolar diffusion in the model or explicitly demonstrate that the early-time decay cannot be explained by diffusion alone.","section":"Sec. 5, Fig. 12"},{"comment":"The retrieved long-lived density differs by a factor of 1.6 between the two harmonics (n_long = 3.3×10^16 cm^-3 for m=2 and 5.2×10^16 cm^-3 for m=3), and the effective tip length L differs between harmonics (0.87 µm vs 1.9 µm in GaAs) and between samples (0.37 µm in InP). The paper acknowledges that the two-layer homogeneous-density model cannot reproduce the complex spatial profile after diffusion and drift (p. 21). Because the top-layer density decay is the direct evidence for the drift claim, the inconsistency of the model parameters across harmonics and samples leaves the quantitative support for band-bending-driven drift unconvincing; a more physical model with shared parameters or an explicit treatment of the spatial profile is needed.","section":"Sec. 5, p. 21"},{"comment":"For GaAs the authors verify that increasing the assumed photoexcited layer thickness from 750 nm to 1000 nm changes the fitted material parameters by less than 5%, which justifies neglecting diffusion there. No such verification is reported for InP, where the absorption depth is only ~350 nm and the ambipolar diffusion coefficient is 10 cm^2/s, so the diffusion length is large relative to the layer thickness. The absence of this control makes the InP interpretation fragile: the faster near-field decay could be an artifact of the fixed-thickness approximation rather than a signature of drift. A sensitivity test varying dexc with delay, or a direct diffusion-model fit, should be provided.","section":"Sec. 4.1 vs Sec. 5"}],"minor_comments":[{"comment":"The manuscript contains many diacritic and spelling artifacts (e.g., 'Cžech Academy of Sciences', 'Terahertž', 'polarižability'); these should be corrected for a polished final version.","section":"Throughout"},{"comment":"The caption of Fig. 8 refers to 'Eq. (5)' for the normalized scattered signal S(2), but the definition is given in Eq. (2); please check all cross-references.","section":"Figure captions, e.g., Fig. 8"},{"comment":"Reference [13] is cited as 'Optica 15, 8550 (2007)', but the correct source is Optics Express 15, 8550 (2007); please correct the journal name.","section":"Reference [13]"},{"comment":"In the final paragraph, the phrase 'was observed was observed' is duplicated; please remove the repetition.","section":"Sec. 6"}],"recommendation":"major_revision","confidential_remarks":"The paper is likely from a well-established group and the experimental THz-SNOM data appear carefully collected. The methodological part is solid and publishable. The main risk is overinterpretation of the InP data: the evidence for band-bending-driven drift is not conclusive because the two-layer model neglects continuous ambipolar diffusion and because the effective tip length and long-lived density are not consistent across harmonics. The authors should be given the opportunity to revise with a more rigorous transport model or at least a quantitative demonstration that diffusion alone cannot explain the data. If such a revision is not possible, the claims about drift should be softened accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper deserves a close read, but the headline result about InP is shakier than the authors let on. The new material is genuinely useful: the error-ellipse analysis (Section 3.3) shows when a THz-SNOM measurement can separate carrier density from layer thickness, and the comparison of normalization strategies (S, self-referenced, X) is a practical contribution. The GaAs experiments are careful, with the two-harmonic consistency providing some confidence in the extraction.\n\nThe soft spot is the InP interpretation. The claim that band-bending drives electrons away from the surface rests on the observation that the near-field signal decays faster than the far-field transmittance. But the model fixes the photoexcited top layer at 350 nm for all delays below 1 ns, only expanding it at 10 ns. With the stated ambipolar diffusion coefficient D=10 cm2/s, the diffusion length at 600 ps is already ~1.1 µm, three times the fixed layer thickness. A continuously diffusing profile would reduce the near-surface density and produce exactly the observed differential decay, with no band-bending needed. The paper checks insensitivity to layer thickness for GaAs (where diffusion is slower relative to the absorption depth) but not for InP, where the effect is more severe. That is a load-bearing omission.\n\nThe effective tip length L also varies by a factor of 2-5 between harmonics and samples (0.87 vs 1.9 µm in GaAs, 0.37 µm in InP). The authors are candid that L is an uncertain instrumental parameter, but the variation suggests the model is absorbing real physics. The retrieved n_long differs by 60% between harmonics, again suggesting overparameterization.\n\nNone of this is fatal to the paper's methodological core. The sensitivity analysis is a real contribution, and the GaAs results are likely sound. But the InP band-bending claim needs either a proper diffusion model or a direct test against a diffusion-only simulation before it can be taken seriously.\n\nRecommendation: send to peer review, but the referee should push hard on the InP modeling. The authors need to demonstrate that their two-layer box is not just reproducing the expected signature of ordinary diffusion.","headline":"The InP band-bending claim is undercut by a diffusion modeling omission; the methodology is still worth engaging.","tokens_in":15869,"tokens_out":2626,"would_cite":true,"duration_ms":26060,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"THz scanning near-field microscopy can tell whether photoexcited electrons recombine or are driven away from the surface.","keywords":["THz-SNOM","near-field microscopy","photoconductivity","GaAs","InP","band bending","carrier diffusion","time-resolved THz spectroscopy"],"falsifier":"Measure the InP near-surface electron density with THz-SNOM after removing the surface band bending, for example by chemical passivation or a different surface termination: if the fast drop in top-layer density persists, the band-bending drift explanation is wrong, while if it disappears the explanation is supported.","tokens_in":14634,"feed_emoji":"🔬","tokens_out":7281,"duration_ms":72736,"temperature":0.7,"pith_summary":"The paper measures ultrafast THz conductivity spectra of GaAs and InP with a scattering-type THz scanning near-field microscope, following the local response for hundreds of picoseconds after a laser pulse. In GaAs the near-field spectra retrieve an electron density that decays exponentially with a time constant of about 46 ps and an electron scattering time that decreases with density. In InP the near-surface electron density decays within about 600 ps while the total electron population seen by far-field THz transmission stays essentially constant, which the authors take as evidence that electrons are diffusing away from the photoexcited region and being repelled from the surface by band bending. The paper also derives general strategies for analyzing THz-SNOM signals and maps the parameter regions where carrier density and photoexcited-layer thickness can be retrieved independently.","feed_headline":"THz near-field probe sees electrons driven off InP's surface","feed_subtitle":"GaAs decays simply; InP's near-surface density falls faster than total carriers, a sign of band-bending drift.","key_machinery":"The central object is the normalized scattered signal $S^{(m)} = s_{\\rm exc}^{(m)}/s_{\\rm gnd}^{(m)}$, modelled by the finite-dipole expression $s^{(m)} \\propto \\alpha^{(m)}(1+r)^2$, where $\\alpha^{(m)}$ is the effective tip polarizability and $r$ the far-field p-polarized reflectance of the sample; the ratio removes the incident pulse and instrumental response. The depth-profiling mechanism is harmonic demodulation: signals demodulated at higher harmonics of the tip tapping frequency respond to thinner photoexcited layers, so measuring the second and third harmonics gives some depth resolution. For InP, the authors use a two-layer model (a thin top layer and a deep stationary layer) to fit the spectra and extract both densities and layer thicknesses, with the far-field reflectance interference term providing the sensitivity needed to separate the two parameters.","core_discovery":"The central claim, stated in the abstract, is that in InP the THz photoconductivity decay is controlled mainly by electron diffusion away from the photoexcited area and by drift due to band bending at the surface, whereas in GaAs the decay is a simpler recombination-limited process. The evidence is a two-layer fit to the normalized near-field spectra: a thin top layer of photoexcited electrons whose density collapses to nearly zero within 10 ns, over a deeper, long-lived population at $3$ to $5\\times10^{16}\\,\\mathrm{cm^{-3}}$ extending about $6\\,\\mu\\mathrm{m}$ into the crystal. Because far-field transmission shows the total electron density barely decaying, surface recombination is ruled out, leaving outward drift as the explanation. Alongside this, the paper shows that the depth selectivity of higher tapping harmonics can be used to distinguish surface-near conductivity from bulk conductivity, and that the independent retrieval of density and thickness is possible mainly where the far-field reflectance interference term contributes significantly.","pith_inferences":["If the band-bending interpretation is right, passivating or otherwise changing the InP surface should suppress the fast collapse of the top-layer electron density while leaving the far-field decay nearly unchanged; this is a directly testable prediction.","The fitted effective tip length differs between harmonics and between GaAs and InP, which suggests the finite-dipole model is absorbing some unmodeled near-field physics; a more physical tip model could alter the quantitative drift rates, though not necessarily the qualitative InP conclusion.","The harmonic-depth-selectivity idea could be applied to other layered or nanostructured semiconductors, such as heterostructures with buried two-dimensional electron gases, to extract depth-resolved carrier dynamics without sectioning the sample."],"forward_implications":["THz-SNOM spectra can expose carrier transport along the surface normal that far-field THz spectroscopy averages over.","Higher tapping harmonics can act as a crude depth scan, since each harmonic is sensitive to a different thickness of the photoexcited layer.","In InP, band-bending drift can dominate the near-surface THz response within tens of picoseconds, so surface treatment must be considered when interpreting near-field photoconductivity.","For thick photoexcited layers, the interference of the THz beam inside the layer enables simultaneous retrieval of carrier density and layer thickness, while for thin layers only the sheet conductivity can be reliably obtained.","The harmonic-ratio signal $X^{(2,3)}$ cancels the reflectance and can be used on structured surfaces, but with about four times worse accuracy than the normalized single-harmonic signal."],"supporting_citations":[{"why":"Supplies the finite-dipole model and g factor used to fit all the SNOM spectra.","marker":"[13]"},{"why":"Extends the finite-dipole polarizability calculation to layered samples, used for the photoexcited layer on a substrate.","marker":"[14]"},{"why":"Earlier THz study of GaAs nanobars that attributed rapid electron escape from the surface to band bending, the precedent for the InP interpretation.","marker":"[15]"},{"why":"Review of time-resolved THz spectroscopy of carrier dynamics that defines the far-field methodology and baseline.","marker":"[1]"},{"why":"Provides the optical absorption depths used to set the initial photoexcited-layer thickness in GaAs and InP.","marker":"[16]"},{"why":"Supplies the ambipolar diffusion coefficient for GaAs used to justify neglecting diffusion on the measured time scale.","marker":"[18]"},{"why":"Supplies the InP ambipolar diffusion coefficient used to estimate the diffusion length of long-lived carriers.","marker":"[19]"},{"why":"Supports the role of electron transport and band bending in InP photoconductivity.","marker":"[20]"}],"fun_headline_variants":["InP's THz conductivity drop reveals electron drift, not just decay","Near-field THz shows InP electrons fleeing band bending","InP photoconductivity fade driven by drift, not recombination","THz near-field reveals InP's surface drift as decay driver","InP's THz decay shows drift, unlike GaAs's simpler recombination"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusions rest on the finite-dipole model being accurate enough that its fitted effective tip length is a harmless stand-in for the real tip; the same physical tip gives different fitted lengths for different harmonics and for the two semiconductors, so this assumption is not fully secure.","fun_headline_variants_meta":{"raw":{"variants":["InP's THz conductivity drop reveals electron drift, not just decay","Near-field THz shows InP electrons fleeing band bending","InP photoconductivity fade driven by drift, not recombination","THz near-field reveals InP's surface drift as decay driver","InP's THz decay shows drift, unlike GaAs's simpler recombination"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000989,"raw_usage":{"total_tokens":4150,"prompt_tokens":860,"completion_tokens":3290,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":3198}},"tokens_in":476,"tokens_out":3290,"duration_ms":27399,"temperature":1.0,"reasoning_tokens":3198,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:03:04.067629+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the InP near-surface electron density with THz-SNOM after removing the surface band bending, for example by chemical passivation or a different surface termination: if the fast drop in top-layer density persists, the band-bending drift explanation is wrong, while if it disappears the explanation is supported.","supporting_citations":[{"cited_title":"Cvitkovic, N","cited_arxiv_id":null,"evidence_quote":"Supplies the finite-dipole model and g factor used to fit all the SNOM spectra."},{"cited_title":"Hauer, A","cited_arxiv_id":null,"evidence_quote":"Extends the finite-dipole polarizability calculation to layered samples, used for the photoexcited layer on a substrate."},{"cited_title":"Pushkarev, H","cited_arxiv_id":null,"evidence_quote":"Earlier THz study of GaAs nanobars that attributed rapid electron escape from the surface to band bending, the precedent for the InP interpretation."},{"cited_title":"Ulbricht, E","cited_arxiv_id":null,"evidence_quote":"Review of time-resolved THz spectroscopy of carrier dynamics that defines the far-field methodology and baseline."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the optical absorption depths used to set the initial photoexcited-layer thickness in GaAs and InP."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the ambipolar diffusion coefficient for GaAs used to justify neglecting diffusion on the measured time scale."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the InP ambipolar diffusion coefficient used to estimate the diffusion length of long-lived carriers."},{"cited_title":"Lloyd-Hughes, S","cited_arxiv_id":null,"evidence_quote":"Supports the role of electron transport and band bending in InP photoconductivity."}],"review_version":1}