Pith. sign in

REVIEW 4 major objections 5 minor 3 references

High-efficiency silicon LED with ultra-wideband emission from visible to infrared at room temperature

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read Femtosecond laser annealing turns standard 220 nm silicon-on-insulator into a room-temperature LED emitting from 600 to 1650 nm, with external quantum efficiency above 0.26% and output power density above 20 W/cm².

desk verdict New observation, broken efficiency claims: the FLA-induced broadband emission is worth a look, but the EQE and power-density numbers rest on a wrong definition and untraceable calibration. read the letter →

arxiv 2508.07352 v1 pith:K7T324TI submitted 2025-08-10 physics.optics physics.app-ph

classification physics.opticsphysics.app-ph
keywords siliconphotonicslight-emittingdiodefemtosecondlaserannealingbroadbandluminescencesilicon-on-insulatorexternalquantumefficiencyelectroluminescencenear-infraredemission
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Silicon is an indirect-bandgap semiconductor, so its light emission is normally weak and narrow. The paper claims that a brief femtosecond-laser anneal of a standard 220 nm silicon-on-insulator wafer transforms it into a room-temperature light-emitting diode that emits continuously from 600 to 1650 nm, with an external quantum efficiency above 0.26% (measured over 900 to 1650 nm) and an output power density above 20 W/cm². These figures are several orders of magnitude higher than previous silicon-based LEDs, and the process is CMOS-compatible. If true, this gives silicon photonics a native, integrable broadband light source for communication wavelengths, gas sensing, optical coherence tomography, and on-chip inspection without III-V materials or external lasers.

What carries the argument

Femtosecond laser annealing (FLA) is the central enabler: 515 nm, roughly 290 fs pulses at 100 kHz, 300 nJ, focused through a 10×/0.26 NA objective on the 220 nm SOI device layer after ion implantation and rapid thermal annealing. It creates the modified silicon that emits and absorbs across 600–1650 nm. The EQE numbers are carried by a relative calibration: the sample's photoluminescence or electroluminescence spectrum is compared by spectral area with a calibrated commercial near-infrared reference LED (L12509-0155G) attenuated to a similar signal level, so the absolute EQE depends on the reference's known output and on equal collection efficiency for sample and reference.

What would settle it

Measure the absolute emission of the FLA-treated sample in an integrating sphere with a calibrated detector spanning 900–1650 nm under 532 nm excitation, and compare the resulting EQE with the value obtained by the spectral-area-ratio calibration to the reference LED. A discrepancy beyond the combined uncertainties would falsify the >0.26% EQE claim. A second check: replace the reference LED with an independently calibrated broadband source and see whether the ratio method reproduces the same EQE.

Watch

Extended reading notes

Core claim

The central claim is that femtosecond laser annealing of a standard 220 nm silicon-on-insulator layer produces a modified silicon state whose room-temperature luminescence is both ultra-broadband and efficient. Photoluminescence spans at least 600–1650 nm, with the drop below 1000 nm and above 1600 nm attributed to the InGaAs detector, and electroluminescence from a forward-biased horizontal PIN junction covers 700–1650 nm. The authors report a photoluminescence EQE close to 0.3% and an LED EQE above 0.26% in the 900–1650 nm window, an output power above 28 μW at 11.4 mA, and an output power density above 20 W/cm². Transient absorption shows infrared absorption with induced bleaching near 15

Load-bearing premise

The EQE result rests on the assumption that the sample and the reference LED are collected and detected with identical efficiency across 900–1650 nm, and that the reference's absolute power is known; if the spectral-area calibration is off, every quoted EQE shifts.

Editorial extensions

If this is right

  • If the EQE and power-density numbers hold, a standard CMOS process can produce a native silicon LED that emits across the telecom O-E-S-C bands, removing the need for III-V bonded sources.
  • An integrable broadband source spanning 600–2200 nm would directly support on-chip absorption spectroscopy, gas sensing, optical coherence tomography, and chip-defect inspection with one device.
  • The unsaturated EQE slope with increasing pump power and drive current implies that higher efficiency is available before saturation, so the reported 0.26% is not necessarily the ceiling.
  • Because the LED is driven by a forward-biased PIN junction rather than avalanche breakdown, it avoids high-voltage operation and is more compatible with low-power integrated electronics.
  • Since the quoted EQE is computed only in the 900–1650 nm window while visible emission is also observed, the full-spectrum efficiency would be higher than the reported values.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • One extension the authors do not pursue is patterning FLA at lithographic resolution to create arrays of independently addressable broadband emitters next to silicon photodetectors on the same chip; our inference is that this could enable wavelength-integrated lab-on-chip sensing without external light sources.
  • The log-log slopes of integrated PL intensity versus pump power (5.40 and 9.62) are far above the usual value of 1; if those slopes reflect true carrier statistics rather than measurement artifacts, the emission mechanism is not conventional single-exciton recombination, and the EQE may depend on excitation level.
  • The induced bleaching around 1500–1550 nm hints at Pauli blocking or population inversion; our extension is that electrically injected devices at higher current density should be examined for net optical gain, which would be a step toward a silicon-based superluminescent diode or laser.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper reports that femtosecond laser annealing (FLA) of 220 nm silicon-on-insulator (SOI) produces broadband luminescence from 600–1650 nm at room temperature. Photoluminescence (PL) and electroluminescence (EL) are measured, and the authors claim an external quantum efficiency (EQE) exceeding 0.26%, an output power of 28 μW, and an optical power density above 20 W/cm². They attribute the emission to efficient radiative recombination and compare their device with prior silicon-based LEDs. The central quantitative claims rest on a calibration against a Hamamatsu reference LED and on an EQE definition given as P_out/P_in.

Significance. If the reported efficiency and power density were correct, this would be a significant advance for CMOS-compatible silicon light sources, potentially impacting optical interconnects, sensing, and infrared applications. The paper's strengths include the demonstration of broadband PL and EL from standard SOI after FLA, a CMOS-compatible fabrication route, transient absorption characterization, and a comparison table with prior work. However, the central quantitative claims are undermined by a dimensional error in the definition of EQE and by an untraceable absolute calibration. These issues are load-bearing because the headline efficiency, power, and "orders of magnitude" comparisons depend directly on them.

major comments (4)
  1. [Methods, 'Extraction of EQE' (unnumbered equation)] The equation EQE = P_out/P_in defines a power-conversion efficiency, not an external quantum efficiency. EQE for an LED/PL process is the ratio of emitted photon flux to injected carrier (or absorbed photon) flux. Because the emission spans 600–1650 nm, photon energies vary by a factor of ~2.75, so the power ratio differs from the photon-flux ratio by a spectrum-dependent factor. The reported "EQE > 0.26%" is therefore not established as a quantum efficiency. If the authors intend wall-plug efficiency, the label and all comparisons to EQE values in Table 1 are mislabeled.
  2. [Methods, 'Extraction of EQE'] The absolute calibration against the Hamamatsu L12509-0155G LED is not documented sufficiently: no reference spectrum, no absolute radiant flux or NIST-traceability statement, no attenuation factor, and no demonstration that the collection/detection efficiency is identical for the packaged reference LED and the bare silicon sample. The reported 28 μW output power and the derived EQE depend entirely on this calibration. Without a reproducible absolute power scale, the central quantitative claims are unsupported.
  3. [Table 1 and Abstract] The abstract claims performance "several orders of magnitude higher than other silicon-based LEDs," but Table 1 lists Green (2001) with EQE = 0.55% and power = 180 μW, both higher than the present values (0.26%, 28 μW). The claim is defensible only for optical power density (20,000 vs <0.2 mW/cm²). The text should explicitly state which metric is being compared and correct the misleading generalization.
  4. [Fig. 1b and accompanying text] The log-log slopes of 5.40 and 9.62 for integrated PL intensity versus excitation power are interpreted as evidence of "ultra-high internal quantum efficiency" because they exceed 1. This is not a valid diagnostic. Superlinear slopes can arise from excitation-dependent carrier dynamics, nonlinear recombination pathways, or measurement artifacts; they do not directly measure IQE. The claim of minimal non-radiative recombination is unsupported by these data.
minor comments (5)
  1. [Introduction, p. 3] Typo: "which will greatly increases greatly increases energy loss" should read "which greatly increases energy loss."
  2. [Methods, 'Extraction of EQE'] Typo: "we adopte" should be "we adopt."
  3. [Results, Fig. 3 and Table 1] The active emission area used to compute 20 W/cm² is not given. From 28 μW and 20 W/cm², the area would be ~1.4×10⁻⁶ cm² (140 μm²); this should be stated explicitly and justified.
  4. [Fig. 1d] The fluorescence lifetime is mentioned but no lifetime values or fitting model are provided. Please report the measured lifetimes and the instrument response function.
  5. [General] Several references are cited in the Methods (e.g., 30–32) without clear context in the main text; please ensure all references are introduced where relevant.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: EQE is referenced to an external Hamamatsu LED and spectra/lifetimes are direct measurements.

full rationale

The paper's central claims are experimental measurements: broadband PL/EL spectra, EQE calibrated against an external Hamamatsu reference LED, and output power derived from a spectral-area ratio. No fitted parameter is renamed as a prediction; the reference LED is an external standard, not an output of the model. The slope values and lifetimes are direct observations. No self-citations are used as load-bearing justification. The calibration method relies on assumptions about collection efficiency and traceability, but that is a correctness/measurement concern, not circularity. Therefore score 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim depends on a non-standard EQE definition and on the interpretation of PL slopes. The calibration uses an external reference LED, but its accuracy is not demonstrated. No new physical entities are introduced.

free parameters (2)
  • PL log-log slope (low power) = 5.40
    Fitted to the PL integral intensity versus pump power in Fig. 1b, used to claim ultra-high internal quantum efficiency.
  • PL log-log slope (high power) = 9.62
    Fitted to the PL integral intensity versus pump power in Fig. 1b at higher powers, also interpreted as evidence of high IQE.
assumptions (4)
  • ad hoc to paper EQE is defined as the ratio of output power to input power and is treated as a quantum efficiency.
    The Methods section writes EQE = P_out/P_in. This is a power conversion efficiency, not a photon-number quantum efficiency, and no conversion is shown.
  • domain assumption The Hamamatsu L12509-0155G LED provides an accurate absolute power calibration standard for the measurement system.
    The EQE calculation relies on comparing spectral areas with the Hamamatsu LED, but no calibration traceability or spectral coverage details are provided.
  • ad hoc to paper A log-log slope greater than 1 for PL intensity versus pump power indicates ultra-high internal quantum efficiency.
    The text interprets slopes 5.40 and 9.62 as evidence of minimal non-radiative recombination, a non-standard interpretation.
  • domain assumption The observed broadband emission is intrinsic to the FLA-treated silicon and not from parasitic sources or measurement artifacts.
    No control PL or EL spectra of unannealed silicon are shown, so the attribution of the emission to the FLA treatment is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of High-efficiency silicon LED with ultra-wideband emission from visible to infrared at room temperature." pith.science (2026). https://pith.science/paper/K7T324TI

@misc{pith2026250807352,
  author       = {Pith},
  title        = {Pith review of: High-efficiency silicon LED with ultra-wideband emission from visible to infrared at room temperature},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K7T324TI}},
  note         = {Machine review of arXiv:2508.07352}
}
read the original abstract

The primary challenge in silicon photonics is achieving efficient luminescence in the communication band, crucial for its large-scale application. Despite significant efforts, silicon light sources still suffer from low efficiency and limited emission wavelengths. We addressed this by achieving broadband luminescence from 600-1650 nm through femtosecond laser annealing of 220nm standard SOI, resulting in an external quantum efficiency exceeding 0.26% and an output optical power density greater than 20 W/cm2, several orders of magnitude higher than other silicon-based LEDs in performance. The broadband LED has potential applications in optical inspection, gas sensing, optical coherence tomography, optical communication, and more.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

  1. [1]

    At present, silicon photonics systems require the integration of external light sources, which will increase the complexity and cost of the system1

    INTRODUCTION Silicon photonics enable the miniaturization and mass production of optical systems to meet the increasing demand across various applications. At present, silicon photonics systems require the integration of external light sources, which will increase the complexity and cost of the system1. As technology develops towards high integration and ...

  2. [2]

    1a) and observe ultra -wideband luminescence in the range of 900-1650 nm

    RESULTS We measure the power -dependent photoluminescence (PL) spectra of silicon after FLA a t room temperature (Fig . 1a) and observe ultra -wideband luminescence in the range of 900-1650 nm . The fluorescence intensity significantly increases with rising excitation power . The attenuation of the PL signal below 1000 nm and above 1600 nm is due to the l...

  3. [3]

    TA reveals the carrier relaxation process, confirming the absorption extension

    CONCLUSIONS We achieve ultra-broadband PL on silicon from 600 to 1650 nm (532 nm excitation) with an EQE close to 0.3%. TA reveals the carrier relaxation process, confirming the absorption extension. Subsequently, we fabricate a 900-1650 nm ultra-broadband LED on 220 nm standard SOI, fully compatible with CMOS processes, with emission wavelengths compatib...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.