{"id":"d67d35bb-d5ea-41c7-9042-f42bfd500fa1","arxiv_id":"2505.18438","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":7,"one_line_summary":"The authors report photon emission gain in reverse-biased Er-doped Si LEDs and claim the gain equals the electroluminescence-region width divided by the electron mean free path.","lead":"Er-doped silicon LEDs under reverse bias are reported to emit photons with a differential gain of several photons per injected electron via impact excitation, with IQE reaching 1.84% at 78 K. The paper claims the gain equals the electroluminescence width divided by the electron mean free path (G = Lex/ell), a relation that, if true, would strengthen silicon-based communication-band light sources.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed per-electron gain is a differential slope of a voltage-nonlinear curve; it can exceed unity even with IQE < 1 because the voltage sweep jointly increases current and excitation probability.","rationale":"The reader's weakest-assumption analysis correctly identifies the central flaw: the paper interprets a differential slope of Phi versus I_R/q as a per-electron gain, but this interpretation is invalid when both Phi and I_R are monotone functions of the same sweep variable V and I_R is much more nonlinear than Phi. The paper's own maximum IQE of 1.84% is itself the clearest internal evidence: an average of less than 0.02 photons per injected electron is compatible with a local differential slope greater than 1 only if the slope is not a per-electron yield. My independent check of the derivative identity confirms the reader's concern, and the additional circularity in fitting ell from the same data strengthens the rejection. I see no reason to adjust the reader's REJECT verdict. The device engineering and careful power calibration are useful, but they do not rescue the central claim of photon emission gain.","tokens_in":6739,"tokens_out":2447,"duration_ms":21447,"concrete_test":"Using the authors' fitted eq. (3) and their empirical IR(V) from Figure 3(a), compute Phi(V) = IQE(V) * IR(V)/q for each temperature. Then numerically evaluate G(V) = dPhi/d(IR/q) = IQE + (IR/q) * (dIQE/dV) / (d(IR/q)/dV) across the low-bias range where gain is quoted. If max G(V) > 1 while max IQE(V) < 1 for the same fits, the claimed photon emission gain is a derivative artifact of the joint voltage/current sweep and does not require multiple excitations per electron. A complementary but not necessary check is to measure photon counts with the bias voltage held fixed and compare single-electron counting statistics.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that a superlinear photon-flux versus electron-flux slope at low reverse bias proves multiple photons emitted per injected electron, i.e., G = dPhi/d(IR/q) > 1. This inference is not valid under the paper's own model. The measured photon flux is Phi = IQE(V) * (IR(V)/q), where IQE itself depends on reverse voltage through We = We0 n sqrt(qE ell - Delta) and through Lex(V). The reverse current is fitted as IR = I0(exp(V/V0) - b), an exponential in V. Differentiating gives dPhi/d(IR/q) = IQE(V) + (IR/q) * (dIQE/dV) / (d(IR/q)/dV). Near the impact-excitation threshold, dIQE/dV is positive and steep while d(IR/q)/dV is large but finite, so the second term can exceed 1 even if IQE(V) <= 1 everywhere. The paper's own maximum average IQE is 1.84%, i.e., 0.0184, so a differential slope greater than 1 does not establish that any electron emitted multiple photons; it only shows that the additional electrons injected at higher bias arrive when the excitation probability and emission-region width are larger. The later comparison G ≈ Lex/ell is not independent because ell is extracted from the same IQE fits used to construct Phi, making the agreement partly circular. Thus the load-bearing premise—that the measured slope is a per-electron gain—is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports photon emission gain in Er-doped silicon LEDs operated under reverse bias, defined as the emission of multiple photons per injected electron. The authors measure the reverse-bias current-voltage characteristic and the absolute near-infrared photon flux at several temperatures, fit their impact-excitation model [Eq. (3)] to the internal quantum efficiency, extract parameters including the electron mean free path ℓ, and compare a low-bias differential slope G = dΦ/d(I_R/q) with L_ex/ℓ. They report an internal quantum efficiency of 1.84% at 78 K and claim that this validates the gain mechanism G = L_ex/ℓ, with an additional check on a previously published device.","tokens_in":7033,"tokens_out":5618,"duration_ms":48248,"significance":"If the gain claim were established, it would be of clear interest for silicon photonics, because it would imply that a single hot electron can excite multiple Er ions in a wide depletion region and thereby increase the efficiency of communication-band silicon light sources. The paper contains useful experimental ingredients: calibrated absolute power measurement, KPFM characterization of the junction, temperature-dependent electroluminescence data, and parameter extraction that is compared with literature values. However, the central claim is not supported by the presented analysis. The differential-slope gain is confounded by the voltage sweep, and the comparison with L_ex/ℓ is partly circular because ℓ is extracted from the same IQE fits. These are load-bearing problems with the main conclusion, not presentation issues.","major_comments":[{"comment":"The gain is defined as the differential slope dΦ/d(I_R/q) of a voltage-swept measurement. Under the paper's own model, Φ(V) = IQE(V)·I_R(V)/q, with IQE given by Eq. (3), W_ex ∝ n·sqrt(qEℓ − Δ), and I_R fitted as I_R = I_0(exp(V/V_0) − b). Differentiating gives dΦ/d(I_R/q) = IQE(V) + (I_R/q)·(dIQE/dV)/(d(I_R/q)/dV). Near the impact-excitation threshold, dIQE/dV is positive and steep while the denominator is finite, so the second term can exceed unity even when IQE(V) is everywhere below unity. The paper's reported peak IQE is 1.84% (Fig. 3(c)), i.e., 0.0184 photons per injected electron; a differential slope greater than one is therefore fully compatible with every electron emitting far less than one photon on average. The superlinear initial rise in Fig. 3(b) only shows that the additional electrons injected at higher reverse bias arrive when the per-electron excitation probability is larger; it does not demonstrate multiple photons per injected electron.","section":"Fig. 3(b) and the paragraph beginning 'Finally, let us revisit Figure 3(b)'"},{"comment":"The comparison G ≈ L_ex/ℓ is not an independent validation. The electron mean free path ℓ is obtained by fitting Eq. (3) to the IQE(V) data shown in Fig. 3(c), and the same fits and fitted ℓ are used to construct the photon-flux curve whose low-bias slope defines G; L_ex is obtained from Eq. (1) using V_min also extracted from the same data set. The agreement in Fig. 3(h) is therefore a consistency test between a measured slope and a ratio of fitted parameters from the same data, not a prediction that could falsify the model. The previous-device check based on ref. [15] is less circular, but it again relies on a mean free path extracted from the same impact-excitation theory and does not remove the voltage-sweep confound described above.","section":"Eq. (3) and Fig. 3(h)"},{"comment":"The analysis assumes that the total reverse current I_R consists of electrons that traverse the electroluminescence region and contribute to impact excitation, with the injection electron rate taken as I_R/q. In a reverse-biased Si pn junction at these temperatures, the reverse current is generation current arising throughout the depletion region, so carriers are generated at different positions; an electron generated near the n-side will not traverse the entire high-field region, and holes also contribute to the measured current. This assumption can bias both the extracted ℓ and the low-bias slope assigned to per-electron gain. The manuscript should either justify this identification quantitatively from the device geometry and generation profile, or model the position-dependent contribution of the current to impact excitation.","section":"Eq. (3) and the treatment of reverse current"}],"minor_comments":[{"comment":"There are unresolved 'Error! Reference source not found.' placeholders in the captions and body text, including after 'Figure 1(a)' and 'Figure 2(e)'; these must be repaired before any resubmission.","section":"Throughout the manuscript"},{"comment":"The claim that the IQE of 1.84% is a 'record' is not supported by a systematic comparison with prior reverse-bias Er-doped Si LEDs at the same wavelength and temperature; the word 'record' should be removed or substantiated with a literature table.","section":"Fig. 3(c) and abstract"},{"comment":"The notation for the electroluminescence width is inconsistent: L_ex, L_exp, and L_expn appear in the same passage; each subscript should be defined and used consistently.","section":"Eq. (1) and surrounding text"},{"comment":"Eq. (2) is referred to as 'eq. (2)' but is not shown with an equation number; number it explicitly, and define every symbol in Eq. (3), including the cross-sectional area A_c and the conversion factor used to go from photon power to photon flux.","section":"Eqs. (2) and (3)"},{"comment":"The empirical fit I_R = I_0(exp(V/V_0) − b) is introduced without reporting the fitted parameters I_0, V_0, and b or their uncertainties; these values are needed to evaluate the derivative argument in the gain analysis.","section":"Fig. 3(a) and the reverse-current fit"},{"comment":"The extracted mean free path of 185 nm at 78 K is about two orders of magnitude larger than the ~4 nm value quoted from ref. [15] for a highly doped device; the manuscript should discuss whether this difference is consistent with the doping and temperature dependence of electron transport in Si.","section":"Fig. 3(f) and the extracted mean free path"}],"recommendation":"reject","confidential_remarks":"The central inference of the paper is the identification of the differential slope dΦ/d(I_R/q) with a per-electron photon gain. Because the measurement is voltage-swept and the paper's own model makes IQE voltage-dependent, a slope greater than unity does not establish multiple photons per electron, and the reported peak IQE of 0.0184 makes this interpretation particularly implausible. The G ≈ L_ex/ℓ comparison is also circular because ℓ comes from the same fits. These issues cannot be repaired by local revision of the text; the data as presented do not support the headline claim. If the authors can provide a clean test—for example, a measurement at fixed excitation energy with controlled single-electron injection, or a demonstration that the observed slope is not reproduced by the model's own IQE(V) curve—the work might be reconsidered, but the current manuscript is not suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: this paper has a real device result—a reverse-biased Er-doped Si LED with a wide depletion region, reaching an IQE of 1.84% at 78 K—but the central claim of photon emission gain (multiple photons per injected electron) is almost certainly an artifact of measuring a differential slope against voltage-tuned current.\n\nWhat is actually new and solid: the authors engineered a low-doped p/n structure that widens the electroluminescence region, building on their own impact-excitation theory. The experimental characterization is careful: KPFM, EL imaging, absolute power calibration against a commercial LED and an integrating sphere, and IQE versus voltage fits that yield physically reasonable parameters (mean free path 115–185 nm, optically active Er ~1e20 cm^-3, activation energy ~170 meV). That part deserves credit and is worth reporting.\n\nThe soft spot is fatal for the gain claim. The gain is defined as dPhi/d(I_R/q) extracted from a reverse-voltage sweep. But the photon flux is Phi = IQE(V) * I_R(V)/q, and I_R is fitted as a exponential in V while the excitation probability has a square-root threshold. Differentiating gives dPhi/d(I_R/q) = IQE + (I_R/q)*(dIQE/dV)/(dI_R/dV). Near the impact threshold, dIQE/dV is large and positive, so this slope can exceed 1 even if every electron emits at most one photon. The paper's own peak average IQE is 1.84%, which is consistent with no multiple emission per electron. Comparing the slope to L_ex/ell is partly circular because ell is extracted from the same IQE data; the earlier-device check uses the same slope methodology. So the load-bearing premise—that the measured slope is a per-electron gain—is unsupported.\n\nI would still send this to peer review rather than desk reject. The device work is substantive, the model fits are clean, and the flaw is fixable: a reanalysis using their own model would show the superlinear slope is explained by field-dependent excitation, or a fixed-voltage current-modulation experiment could test true gain. The paper needs major revision, and the gain language should be dropped unless such evidence is provided. It is a good reading-group paper as a cautionary tale about differential slopes in nonlinear systems.","headline":"Useful device engineering and a clean empirical model, but the 'photon emission gain' claim is a differential-slope artifact that does not survive contact with the paper's own equations.","tokens_in":7636,"tokens_out":2977,"would_cite":false,"duration_ms":26327,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["78.60.Fi","85.60.Jb"],"model":"deepseek-v4-flash","headline":"Er-doped silicon LEDs can emit multiple photons per injected electron.","keywords":["erbium-doped silicon","impact excitation","photon emission gain","internal quantum efficiency","reverse-biased LED","electroluminescence","silicon photonics","1.54 µm emission"],"falsifier":"Measure the low-flux differential slope $d\\Phi/dI$ on a diode whose electroluminescence width is independently imaged to be shorter than the electron mean free path; the claimed relation $G = L_{\\mathrm{ex}}/\\ell$ predicts a slope below one, whereas an interpretation artifact from the exponential reverse-current fit could still produce a slope above one near threshold.","tokens_in":1696,"feed_emoji":"💡","tokens_out":6229,"duration_ms":120110,"temperature":0.7,"pith_summary":"This paper tries to establish that a single hot electron traversing the depletion region of a reverse-biased Er-doped silicon diode can excite several Er ions, so the diode emits multiple photons per injected electron. The authors model this photon emission gain as $G = L_{\\mathrm{ex}}/\\ell$, the ratio of the electroluminescence region width to the electron mean free path, and report that the measured gain matches this ratio from 300 K down to 78 K. They also report an internal quantum efficiency (photons created per injected electron) of 1.84% at 78 K, about twenty times the room-temperature value. If the claim holds, silicon photonics gains a more efficient, electrically pumped route to communication-band light near 1.54 µm.","feed_headline":"Er-doped Si LEDs emit several photons per injected electron","feed_subtitle":"Reverse bias lets one hot electron hit many Er ions, lifting internal quantum efficiency to 1.84% at 78 K.","key_machinery":"The load-bearing object is the impact-excitation rate $W_{\\mathrm{ex}} \\simeq W_{\\mathrm{ex}0}\\, n \\sqrt{qE\\ell - \\Delta}$, which sets how many Er ions are excited per unit volume per second when electrons with mean free path $\\ell$ are accelerated by electric field $E$. Combined with the steady-state balance $W_{\\mathrm{ex}}(N_{tt}-N_{Er}) = W_d N_{Er}$ and the width $L_{\\mathrm{ex}}$ of the region where $qE\\ell \\geq \\Delta$, this yields the internal quantum efficiency expression used to fit the data. The central identity $G = L_{\\mathrm{ex}}/\\ell$ counts how many mean-free-path segments fit inside the emission region, each segment being one opportunity for a re-accelerated electron to excite another Er ion; the paper tests this identity against the measured superlinear slope of photon flux versus electron flux.","core_discovery":"The central discovery claimed is that reverse-biased Er/O/B co-doped Si PN junctions exhibit photon emission gain: the photon flux rises superlinearly with injected electron flux, and the low-flux slope exceeds one. The authors attribute this to multiple impact excitations: an electron accelerated by the depletion-region field loses energy to one Er ion, is re-accelerated, and collides again, so a single electron can excite several Er ions when the electroluminescence region is longer than the mean free path. Quantitatively, the gain is claimed to equal $G = L_{\\mathrm{ex}}/\\ell$; the measured gain and this ratio agree at all temperatures studied, and the same relation reproduces a gain of about 2.8 in the authors' earlier narrower-junction device with $L_{\\mathrm{ex}}/\\ell = 10\\,\\mathrm{nm}/4\\,\\mathrm{nm}$. The paper further claims an internal quantum efficiency of 1.84% at 78 K, obtained after correcting for the 2.3% escape probability of photons from bulk silicon.","pith_inferences":["If $G = L_{\\mathrm{ex}}/\\ell$ holds across devices, the photon-number statistics of this LED should show a random number of photons per injected electron; a superconducting-nanowire single-photon detector paired with a low-noise current source could test that directly.","The inverse analogy to photoconductive gain suggests a timing trade-off: more gain means later photons, so pulse-to-pulse jitter may grow with $L_{\\mathrm{ex}}$, which would matter for clocked quantum communication.","A continuous test would be to grade the doping so that $L_{\\mathrm{ex}}$ changes monotonically within one device; the predicted gain-versus-voltage curve could then be checked without fabricating many samples.","The claim that the optically active Er concentration is temperature-independent at about $10^{20}\\,\\mathrm{cm}^{-3}$ could be cross-checked by independent absorption or pump-probe measurements on the same diodes."],"forward_implications":["A single electron can produce more than one 1.54 µm photon when the emission region is wider than the electron mean free path, so reverse-biased Er:Si LEDs need not be limited to one photon per electron.","The gain is set by the ratio of electroluminescence width to mean free path, so widening the depletion region and lowering carrier concentration directly increase photons per electron.","Cooling from room temperature to 78 K suppresses nonradiative Er relaxation and raises the internal quantum efficiency by about a factor of twenty.","The extracted design parameters — optically active Er concentration near $10^{20}\\,\\mathrm{cm}^{-3}$ and mean free path rising from about 115 nm to 185 nm as temperature falls — give concrete targets for future device engineering.","Because the emission is electrically pumped and at the communication band, the authors argue the mechanism is a foundation for silicon-based on-chip lasers and quantum light sources."],"supporting_citations":[{"why":"Supplies the analytical impact-excitation theory, including the $W_{\\mathrm{ex}}$ formula and the fitting framework the gain claim builds on.","marker":"[15]"},{"why":"Provides the photoconductive-gain analogue (multiple carriers per photon) that frames the multiple-photons-per-electron concept.","marker":"[20]"},{"why":"Supports the interpretation of the reverse current as thermal generation via generation-recombination centers, needed to convert current into electron flux.","marker":"[21]"},{"why":"Gives the 2.3% light-escape factor used to convert external quantum efficiency into internal quantum efficiency.","marker":"[22]"},{"why":"Establishes the Er/O co-doping and energy-transfer context, including the degenerate n-type doping that motivated the B-compensated design.","marker":"[8]"},{"why":"Provides first-principles electron mean-free-path spectra in silicon that anchor the extracted mean-free-path values.","marker":"[19]"},{"why":"Supplies the radiative and nonradiative Er relaxation rates and activation energy used in fitting the temperature dependence.","marker":"[23-25]"}],"fun_headline_variants":["Impact excitation yields multiple photons per electron in Er Si LEDs","Reverse-biased Er LEDs: one electron, many photons","Er-doped Si LEDs show photon gain via impact excitation","Single electron triggers multiple Er emissions in Si LED","Photon gain achieved in Er:Si LEDs by hot-electron impact"],"cache_read_input_tokens":9600,"weakest_assumption_plain":"The gain claim rests on interpreting the measured slope of photon flux versus reverse current as photons per electron that actually undergo impact excitation, which assumes the reverse current is dominated by exactly those electrons.","fun_headline_variants_meta":{"raw":{"variants":["Impact excitation yields multiple photons per electron in Er Si LEDs","Reverse-biased Er LEDs: one electron, many photons","Er-doped Si LEDs show photon gain via impact excitation","Single electron triggers multiple Er emissions in Si LED","Photon gain achieved in Er:Si LEDs by hot-electron impact"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000182,"raw_usage":{"total_tokens":1304,"prompt_tokens":934,"completion_tokens":370,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":550,"completion_tokens_details":{"reasoning_tokens":290}},"tokens_in":550,"tokens_out":370,"duration_ms":3756,"temperature":1.0,"reasoning_tokens":290,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T14:32:18.400461+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the low-flux differential slope $d\\Phi/dI$ on a diode whose electroluminescence width is independently imaged to be shorter than the electron mean free path; the claimed relation $G = L_{\\mathrm{ex}}/\\ell$ predicts a slope below one, whereas an interpretation artifact from the exponential reverse-current fit could still produce a slope above one near threshold.","supporting_citations":[{"cited_title":"Ramíırez, F","cited_arxiv_id":null,"evidence_quote":"Supplies the analytical impact-excitation theory, including the $W_{\\mathrm{ex}}$ formula and the fitting framework the gain claim builds on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the photoconductive-gain analogue (multiple carriers per photon) that frames the multiple-photons-per-electron concept."},{"cited_title":"Obreja, A.C","cited_arxiv_id":null,"evidence_quote":"Supports the interpretation of the reverse current as thermal generation via generation-recombination centers, needed to convert current into electron flux."},{"cited_title":"Priolo, G","cited_arxiv_id":null,"evidence_quote":"Gives the 2.3% light-escape factor used to convert external quantum efficiency into internal quantum efficiency."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the Er/O co-doping and energy-transfer context, including the degenerate n-type doping that motivated the B-compensated design."},{"cited_title":"Jantsch, G","cited_arxiv_id":null,"evidence_quote":"Provides first-principles electron mean-free-path spectra in silicon that anchor the extracted mean-free-path values."}],"review_version":1}