{"id":"bb8e33a2-a32b-4892-b80b-2d5b4012fc4c","arxiv_id":"2506.20379","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Intrinsic silicon's flexoelectric coefficient rises from 15.2 nC/m to 1.8 μC/m between 223 and 473 K, and the Arrhenius slope is claimed to reproduce the silicon bandgap.","lead":"This paper measures how the flexoelectric response of silicon changes with temperature and doping. It finds that intrinsic silicon's flexoelectric coefficient grows by about two orders of magnitude on warming and attributes this to thermally excited charge carriers.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Data self-inconsistency: stated 15.2 nC/m at 223 K and 1.8 µC/m at 473 K give E_g≈0.6–0.7 eV, not 1.12 eV; the claimed Eq. (4) fit requires an unstated, narrow temperature window.","rationale":"I read the paper as making two claims: a qualitative one (thermally enhanced flexoelectricity in intrinsic Si, mediated by free carriers) and a quantitative one (the Arrhenius slope reproduces the Si bandgap via Eq. 4). The qualitative claim is plausible and well supported by the correlation among flexoelectric coefficient, capacitance, and carrier concentration. The quantitative claim is the load-bearing part: the paper states that the measured slopes 'consistently yield' E_g ≈ 1.12 eV and uses this as evidence that intrinsic carrier excitation is the origin of the effect. My concern is internal and does not depend on disputing the barrier-layer model in Eq. (1): the paper's own end-point values imply a much weaker temperature dependence than E_g=1.12 eV predicts. Including the T^(3/4) prefactor makes the discrepancy even larger. The printed Eq. (4) has a dimensional error, which adds ambiguity about exactly what regression was performed. The reader's weakest assumption focused on the validity of Eq. (1); I agree that is a legitimate concern, but the data inconsistency is more decisive because it attacks the central fitted result even if Eq. (1) is accepted. The paper could still be salvageable if the fit was deliberately restricted to a pre-saturation window and the full data are made available; hence I would keep the verdict conditional but with an explicit requirement to report the raw (T, μ) data, the fit range, and the regression residuals.","tokens_in":8708,"tokens_out":11714,"duration_ms":138067,"concrete_test":"Digitize the (T^-1, ln μ) points for at least one intrinsic Si sample from Fig. 3a; perform a weighted least-squares fit on the full 223–473 K range and on a sliding temperature window. Require the reported E_g=1.12 eV to be the best-fit over a pre-specified range with acceptable residuals. If the full-range slope corresponds to ~0.6–0.7 eV and only a narrow segment gives 1.12 eV, the central claim fails as stated.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central quantitative claim—that the Arrhenius slope of ln μ13^eff versus T^-1 yields E_g=1.12 eV (Eq. 4, Fig. 3b)—is contradicted by the numbers reported in the abstract and main text. With μ(223 K)=15.2 nC/m and μ(473 K)=1.8 μC/m, the end-to-end ratio is 118, so ln(μ2/μ1)/Δ(T^-1) ≈ 2014 K, i.e. E_g ≈ 0.69 eV. Including the T^(3/4) prefactor from N_C N_V lowers this to ≈0.61 eV. Using the room-temperature value 29.1 nC/m from Fig. 1c together with the 223 K point gives E_g ≈ 0.13 eV. The 1.12 eV line therefore cannot describe the data over the stated range; it must come from a restricted temperature sub-range. The manuscript never states the fit range, the number of points, or how the saturated high-temperature branch is handled. Eq. (4) as printed is also dimensionally wrong: the derivative of −E_g/(4k_BT) with respect to T^-1 is not −E_g/(4k_BT); the slope should be −E_g/(4k_B) (a constant), plus a small T^(3/4)-prefactor term. This typo obscures what was actually fitted. Unless the fit range and raw data are supplied, the headline 'fitted E_g ≈ 1.12 eV' is not supported by the paper's own displayed numbers.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports measurements of the effective flexoelectric coefficient μ_13^eff in intrinsic and heavily As-doped silicon beams over 223–473 K using three-point bending with a static preload and an oscillatory force. The authors find that heavily doped Si has μ_13^eff ≈ 2.6 μC/m, nearly independent of temperature, whereas intrinsic Si grows from ~15.2 nC/m at 223 K to ~1.8 μC/m at 473 K, with saturation above ~150 °C. They attribute the response to a Schottky-barrier (barrier-layer) mechanism, in which μ_13^eff ∝ N^{1/2}. Combining this with the intrinsic carrier density leads to Eq. (4), d ln μ/d(T^{-1}) ≈ -E_g/(4k_B T), and the authors claim that Arrhenius fits yield E_g ≈ 1.12 eV for Si and ≈0.66 eV for Ge, matching the known bandgaps. The paper concludes that thermally excited carriers control flexoelectricity in intrinsic first-generation semiconductors.","tokens_in":9142,"tokens_out":10464,"duration_ms":111095,"significance":"If the central quantitative claim is correct, the paper would establish a simple relationship between the temperature dependence of flexoelectricity and the semiconductor bandgap, and it would help explain the orders-of-magnitude discrepancy between earlier Si flexoelectric measurements and first-principles calculations. The qualitative picture is well supported by the parallel temperature dependences of flexoelectricity, interfacial capacitance, and carrier concentration, as well as by the contrast between doped and intrinsic samples. The paper also makes a falsifiable prediction—that the Arrhenius slope is controlled by E_g/(4k_B)—and tests it on Ge. However, as detailed in the major comments, the paper's own reported endpoint values are inconsistent with the claimed fitted bandgap, and the fitting procedure is not documented. The reliance on a model from the authors' own earlier work (refs. [23,32]) is a mild circularity concern, but it is mitigated by the fact that the temperature dependence in Eq. (4) arises mainly from the standard intrinsic carrier density. Overall, the data set and conceptual link are valuable, but the headline quantitative claim is currently unsupported.","major_comments":[{"comment":"Using the two endpoint values quoted in the abstract (μ=15.2 nC/m at 223 K and μ=1.8 μC/m at 473 K), the ratio is 118, and the corresponding Arrhenius slope is ln(118)/(1/223−1/473)≈2.0×10^3 K. If this slope is identified with E_g/(4k_B), the implied bandgap is ≈0.69 eV, not 1.12 eV; including the T^{3/4} prefactor from N_C N_V only raises this to ≈0.78 eV. The room-temperature value of 29.1 nC/m (Fig. 1c) together with the 223-K point gives an apparent E_g of only ≈0.2 eV. Therefore the statement that the slopes \"consistently yield an E_g of ~1.12 eV\" cannot be true over the full 223–473 K range. The fit must have been restricted to a narrower, unstated temperature sub-range, or the quoted endpoint values are misreported. The manuscript must state the exact fitting range, the number of data points, how the saturation above ~150 °C is handled, and show the raw Arrhenius plots with the fitted line. Without this, the headline \"fitted E_g ≈ 1.12 eV\" is not supported by the paper's own displayed numbers; the same requirement applies to the Ge data in Fig. 3b.","section":"Abstract, Fig. 2a, Fig. 3b"},{"comment":"Eq. (4) is dimensionally inconsistent as printed. Since d/d(T^{-1})[-E_g/(4k_B T)] = -E_g/(4k_B), the right-hand side should be a constant in kelvin, not -E_g/(4k_B T). Moreover, the T^{3/4} temperature dependence of (N_C N_V)^{1/4} contributes an additional term of -3T/4 to the slope, which is about 7% of the E_g term at 300 K and is not negligible if a quantitative bandgap is claimed. The authors should provide the correctly derived expression and state exactly which functional form was used in the fit.","section":"Eq. (4)"},{"comment":"Eq. (3) is obtained from Eq. (1) by inserting Eq. (2), but this requires that the barrier height, the deformation potential, the dielectric constant, and the sample thickness are all temperature-independent over 223–473 K. The text acknowledges this only in passing (\"neglecting the weak temperature independence\") and refers to the Supplemental Material. Because the extracted E_g is the sole quantitative evidence for the central claim, a sensitivity estimate is needed: for example, how much would a typical ∂ε_r/∂T or a 0.1 eV change in barrier height with temperature shift the fitted E_g? If those effects are comparable to ±0.1 eV, the claimed match to 1.12 eV would not be meaningful.","section":"Eq. (3) and Section 9 of the Supplemental Material"}],"minor_comments":[{"comment":"Refs. [27] and [36] are the same reference (Schiaffino et al., Phys. Rev. B 99, 085107 (2019)), and Refs. [31] and [46] are also duplicates (Li et al., Appl. Phys. Lett. 124, 062904 (2024)).","section":"References"},{"comment":"\"Femi level\" should read \"Fermi level.\"","section":"Fig. 1d"},{"comment":"\"Neglecting the weak temperature independence\" should read \"neglecting the weak temperature dependence.\"","section":"Before Eq. (4)"},{"comment":"The same symbol φ is used for both the Schottky barrier height and the surface deformation potential, which makes the definition of Eq. (1) ambiguous; distinct symbols (e.g., φ_B and φ_D) should be used.","section":"After Eq. (1)"},{"comment":"The statement that the flexoelectric coefficient \"get saturated\" (Fig. 2a) should be rephrased for grammatical correctness.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The central issue is not the concept but the lack of documentation of the fitting procedure. I recommend requiring the authors to provide the raw Arrhenius data with the fitting range, the number of points, and the exact fitting function, including whether the T^{3/4} prefactor was included. The duplicate references and the dimensional typo in Eq. (4) suggest the manuscript also needs careful proofreading before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The key experimental observation is likely real and worth knowing: over 223–473 K, the effective flexoelectric coefficient of intrinsic silicon rises by two orders of magnitude while heavily doped silicon stays flat, and germanium shows the same trend. The correlation with measured carrier concentration and capacitance is strong evidence that free carriers in the Schottky barrier layer mediate the effect. That is a genuinely useful result for anyone working on Si-compatible electromechanics.\n\nThe paper also does some things well. They check for static-force artifacts, measure carrier concentration directly rather than assuming it, include Ge as a universality test, and connect to an existing barrier-layer model rather than inventing a new one. The saturation at high temperature is acknowledged and discussed with plausible mechanisms.\n\nNow the soft spots, in order of severity. First, the abstract quotes 15.2 nC/m at 223 K and 1.8 μC/m at 473 K. End-to-end, that ratio gives an Arrhenius slope of about 0.69 eV, not the claimed 1.12 eV. The 1.12 eV fit must come from a restricted temperature window, but the paper never states the fit range, the number of points, or the error bars. Since the central quantitative punchline is the extraction of the bandgap, this is a load-bearing problem, not a cosmetic one. Second, Eq. (4) is dimensionally wrong as printed: the derivative of -E_g/(4 k_B T) with respect to T^-1 is a constant, -E_g/(4 k_B), not -E_g/(4 k_B T). That typo obscures what was actually fitted. Third, Eq. (1) is taken from prior work, including the same group's paper [32], and the authors assume temperature-independent barrier height, deformation potential, and dielectric constant without direct justification. That assumption is plausible but should be stated as a limitation.\n\nThe stress-test note slightly overstates the case: the displayed numbers do not contradict the qualitative picture, but they do mean the bandgap claim is currently unsupported. A referee can fix this by demanding the raw data, fit ranges, and error bars, and by making the authors correct Eq. (4).\n\nWho this is for: people working on flexoelectricity in semiconductors, especially silicon-based devices. It deserves a serious referee, not a desk reject, because the observation is new, the experiment is nontrivial, and the quantitative issue is fixable with better reporting. My recommendation: send it out, but make the referee dig into the fitting procedure before publication.","headline":"New temperature-dependent flexoelectric data in Si/Ge that are probably real, but the headline bandgap extraction is not supported by the paper's own numbers.","tokens_in":9605,"tokens_out":2303,"would_cite":true,"duration_ms":25054,"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":"Thermally excited free carriers, not the bulk lattice response, set the temperature-dependent flexoelectricity of intrinsic silicon, with a slope governed by the bandgap.","keywords":["flexoelectricity","silicon","Schottky barrier","intrinsic carrier excitation","bandgap","germanium","temperature dependence","flexoelectronics"],"falsifier":"Repeat the measurement on intrinsic silicon with contacts that do not form a Schottky barrier, for example using ohmic heavily doped surface layers or electrodes with matched work function. If the effective coefficient still rises by two orders of magnitude with temperature and its $\\ln(\\mu)$ versus $T^{-1}$ slope still equals $-E_g/(4k_B)$, the proposed barrier mechanism is not the origin; if the enhancement disappears, the barrier layer is required.","tokens_in":8536,"feed_emoji":"⚡","tokens_out":6691,"duration_ms":68796,"temperature":0.7,"pith_summary":"Flexoelectricity in silicon—polarization generated by a strain gradient—can be far larger than first-principles predictions once the metal contacts form a Schottky barrier layer. This paper reports that the effective flexoelectric coefficient of heavily arsenic-doped silicon is about 2.6 µC/m and nearly temperature-independent, while in intrinsic silicon it rises from 15.2 nC/m to 1.8 µC/m as temperature increases from 223 K to 473 K. The authors argue that the difference is controlled by the free-carrier population: donor ionization in doped silicon barely changes with temperature, whereas intrinsic excitation in undoped silicon grows exponentially. They derive a quantitative relationship, $d\\ln(\\mu_{13}^{\\mathrm{eff}})/d(T^{-1}) \\approx -E_g/(4k_B T)$, and show that the fitted slope reproduces the ~1.12 eV bandgap of silicon and the ~0.66 eV bandgap of germanium. If correct, this makes the bandgap the key parameter governing thermal flexoelectric response in first-generation semiconductors and offers a way to tune flexoelectric devices by temperature or doping.","feed_headline":"Silicon flexoelectricity jumps 100x with heat, tracking bandgap","feed_subtitle":"The measured slope of silicon's flexoelectric coefficient matches the bandgap over four kT, in silicon and germanium.","key_machinery":"The load-bearing machinery is the barrier-layer model of semiconductor flexoelectricity, given as $\\mu_{13}^{\\mathrm{eff}} = \\varphi t (N \\varepsilon_0 \\varepsilon_r q / 2\\phi)^{1/2}$, where $N$ is carrier concentration, $\\phi$ the Schottky barrier height, $\\varphi$ the surface deformation potential, and $t$ the beam thickness. Substituting the intrinsic carrier concentration $N = (N_C N_V)^{1/2} e^{-E_g/(2k_B T)}$ and differentiating with respect to $T^{-1}$ produces the distilled result Eq. (4), $d\\ln(\\mu_{13}^{\\mathrm{eff}})/d(T^{-1}) \\approx -E_g/(4k_B T)$. This identity is what lets the paper read the semiconductor bandgap directly from the temperature slope of the flexoelectric coefficient, and it is the mechanism through which doping and temperature become interchangeable tuning knobs for the effective response.","core_discovery":"The central discovery is that the effective flexoelectric coefficient $\\mu_{13}^{\\mathrm{eff}}$ in Au/Si/Au beams tracks the semiconductor's free-carrier concentration, not the bulk flexoelectric response. Bending modifies the Schottky barrier layer at the gold–silicon interface through the surface deformation potential, converting strain gradients into interfacial polarization changes that dominate the measured signal. Combining the barrier-layer expression with the intrinsic carrier density formula gives $\\mu_{13}^{\\mathrm{eff}} \\propto e^{-E_g/(4k_B T)}$, which differentiates to the concise relation $d\\ln(\\mu_{13}^{\\mathrm{eff}})/d(T^{-1}) \\approx -E_g/(4k_B T)$. Fits on multiple intrinsic silicon samples yield $E_g \\approx 1.12$ eV, matching the room-temperature silicon bandgap, and the same procedure on germanium yields $\\approx 0.66$ eV. The paper therefore claims that thermal excitation of free carriers, rather than lattice or surface flexoelectricity, is responsible for the strong temperature dependence, and that heavy doping suppresses it because arsenic donor ionization is nearly temperature-independent.","pith_inferences":["If Eq. (4) is general, the same Arrhenius slope should appear in any semiconductor where intrinsic carriers dominate the Schottky-barrier response, so the relationship can be tested across compound semiconductors and should reproduce each material's bandgap at the relevant temperature.","At higher temperatures the nominally temperature-independent prefactor in Eq. (3) will contribute its own terms to the slope; measurable deviations from the pure $-E_g/(4k_B T)$ line could expose the temperature dependence of barrier height or deformation potential.","Illumination of intrinsic silicon should shift free-carrier concentration without changing temperature, so a photoflexoelectric analog is a direct, testable extension: the coefficient should rise with photocarrier density in the same manner as with thermal excitation."],"forward_implications":["Doping or temperature can tune the effective flexoelectric coefficient of silicon by about two orders of magnitude, providing a design lever for silicon-based electromechanical sensors, energy harvesters, and flexotronic devices.","Equation (4) lets a flexoelectric bending measurement serve as a bandgap probe for semiconductors whose carrier density is dominated by intrinsic excitation.","In heavily doped silicon, the effective coefficient stays nearly constant across the measured temperature range, which is useful for devices requiring stable electromechanical response.","The observed saturation of intrinsic silicon's coefficient near 150 °C bounds the usable temperature range and points to a carrier-related ceiling, such as tunneling through a narrowed barrier or dielectric degradation, rather than exhaustion of carriers."],"supporting_citations":[{"why":"Establishes the barrier-layer mechanism and the large flexoelectric-like enhancement in oxide semiconductors that the paper applies to silicon.","marker":"[22]"},{"why":"Supplies the barrier-layer expression for the effective flexoelectric coefficient and reports flexoelectric saturation in halide perovskites.","marker":"[23]"},{"why":"Prior study using the same barrier-layer model for the flexoelectric-like response of Nb-doped SrTiO3, providing the form of Eq. (1).","marker":"[32]"},{"why":"Textbook source of the intrinsic carrier concentration formula in Eq. (2) and the silicon and germanium bandgap values used for comparison.","marker":"[40]"},{"why":"Provides the ~0.05 eV ionization energy of arsenic in silicon used to explain why heavily doped silicon remains temperature-insensitive.","marker":"[51]"},{"why":"Earlier direct measurement of silicon flexoelectricity whose 40–80 nC/m values frame the doping-enhancement comparison.","marker":"[30]"},{"why":"First-principles calculation giving the ~1 nC/m bulk flexoelectric coefficient that the measured values exceed.","marker":"[26]"},{"why":"First-principles calculation of lattice-mediated bulk flexoelectricity providing the theoretical baseline for silicon.","marker":"[28]"},{"why":"Computes the surface contribution to silicon flexoelectricity and gives an effective value of ~6 V, used to separate surface effects from the barrier-layer mechanism.","marker":"[29]"}],"fun_headline_variants":["Heat-driven flexoelectricity in silicon tied to bandgap","Carrier excitation boosts silicon flexoelectricity 100-fold","Silicon flexoelectricity follows bandgap as temperature rises","Thermal carriers set silicon flexoelectric response","Flexoelectricity in silicon: thermal excitation key"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the measured flexoelectric signal comes almost entirely from the Schottky barrier layer at the gold–silicon contacts, and that barrier height, deformation potential, dielectric constant, and sample thickness are effectively temperature-independent over the measured range; if those conditions fail, the fitted slope connecting to the bandgap is not meaningful.","fun_headline_variants_meta":{"raw":{"variants":["Heat-driven flexoelectricity in silicon tied to bandgap","Carrier excitation boosts silicon flexoelectricity 100-fold","Silicon flexoelectricity follows bandgap as temperature rises","Thermal carriers set silicon flexoelectric response","Flexoelectricity in silicon: thermal excitation key"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000378,"raw_usage":{"total_tokens":2036,"prompt_tokens":997,"completion_tokens":1039,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":613,"completion_tokens_details":{"reasoning_tokens":963}},"tokens_in":613,"tokens_out":1039,"duration_ms":9053,"temperature":1.0,"reasoning_tokens":963,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:49:51.446250+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the measurement on intrinsic silicon with contacts that do not form a Schottky barrier, for example using ohmic heavily doped surface layers or electrodes with matched work function. If the effective coefficient still rises by two orders of magnitude with temperature and its $\\ln(\\mu)$ versus $T^{-1}$ slope still equals $-E_g/(4k_B)$, the proposed barrier mechanism is not the origin; if the enhancement disappears, the barrier layer is required.","supporting_citations":[{"cited_title":"Narvaez, F","cited_arxiv_id":null,"evidence_quote":"Establishes the barrier-layer mechanism and the large flexoelectric-like enhancement in oxide semiconductors that the paper applies to silicon."},{"cited_title":"Shu et al., Photoflexoelectric effect in halide perovskites, Nat","cited_arxiv_id":null,"evidence_quote":"Supplies the barrier-layer expression for the effective flexoelectric coefficient and reports flexoelectric saturation in halide perovskites."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Prior study using the same barrier-layer model for the flexoelectric-like response of Nb-doped SrTiO3, providing the form of Eq. (1)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Textbook source of the intrinsic carrier concentration formula in Eq. (2) and the silicon and germanium bandgap values used for comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the ~0.05 eV ionization energy of arsenic in silicon used to explain why heavily doped silicon remains temperature-insensitive."},{"cited_title":"Narvaez Morales, Flexoelectricity in single crystals (Universitat Autònoma de Barcelona, 2016)","cited_arxiv_id":null,"evidence_quote":"Earlier direct measurement of silicon flexoelectricity whose 40–80 nC/m values frame the doping-enhancement comparison."},{"cited_title":"Hong and D","cited_arxiv_id":null,"evidence_quote":"First-principles calculation giving the ~1 nC/m bulk flexoelectric coefficient that the measured values exceed."},{"cited_title":"Royo and M","cited_arxiv_id":null,"evidence_quote":"First-principles calculation of lattice-mediated bulk flexoelectricity providing the theoretical baseline for silicon."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Computes the surface contribution to silicon flexoelectricity and gives an effective value of ~6 V, used to separate surface effects from the barrier-layer mechanism."}],"review_version":1}