{"id":"01285976-ee3f-4a07-b1e5-e94914157119","arxiv_id":"2507.20866","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Resonant tunnelling diode neurons are demonstrated, experimentally and numerically, for time-series edge detection, Iris classification in a photonic spiking extreme learning machine, and tunable optical fading memory.","lead":"This paper shows that resonant tunnelling diode devices can act as light-triggered artificial neurons, detecting rising edges in data, classifying patterns in a small photonic neural network, and storing spikes in a tunable optical memory. It is a proof-of-concept for ultrafast, low-power neuromorphic hardware built from compact optoelectronic components.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Experimental edge-detection claim rests on a single threshold-tuned Mackey-Glass run, with no held-out data or control showing the dual-modulation scheme generalizes.","rationale":"The paper plausibly demonstrates a qualitative proof of concept: a real telecom-wavelength RTD can be biased to produce spikes that correlate with rising edges in the Mackey-Glass time series, and the numerical model is grounded in prior RTD work. I do not dispute that the experimental demonstration exists. However, the strongest quantitative experimental claim—16 of 20 detections with no false positives—is supported by a single realization in which the detection threshold and the device's spiking threshold are chosen on the same data. There is no held-out evaluation, no repeated trial statistics, and no control to show that the dual photonic-electronic modulation is the cause of the apparent edge selectivity. The reader's weakest assumption concerned transferability of the lumped-circuit and laser models to GHz-scale and multi-device simulations; that is a real concern for the numerical ELM and memory results. But the experimental edge-detection claim is more load-bearing for the paper's headline, because it is the only hardware validation and the basis for claiming an ultrafast event detector. My recommended verdict is unchanged: the paper should be CONDITIONAL, with the condition being additional validation of the experimental edge-detection result under fixed parameters and against a control condition.","tokens_in":17463,"tokens_out":5186,"duration_ms":70048,"concrete_test":"Repeat the §3.1 experiment on at least one held-out Mackey-Glass segment (ideally five fresh segments) with all encoding, bias, and threshold parameters fixed from the first segment, and report the detection rate and false positives on the held-out data. In addition, run a control in which the delayed electrical modulation is replaced by a time-shuffled copy of y(t−τ) while the optical input is unchanged; if performance collapses or is not significantly above the shuffled control, the 16/20 result should be treated as a tuned demonstration rather than a validated edge detector.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central hardware claim is the experimental rising-edge detection in §3.1 (Figs. 5-6): an RTD neuron detects 16 of 20 Mackey-Glass edges with no false positives. The evidential value of this result depends on how the target edges and the RTD spiking threshold were set. The targets are defined as time steps where the delayed difference Dτ = y(t) − y(t−τ) exceeds 0.32, and the RTD's threshold is then implemented empirically by choosing optical pulse amplitudes and voltage modulation levels on the same time series (text near Fig. 4). With only one MG segment, no repeated trials, no held-out segment, and no baseline condition, the reported 16/20 and zero false positives could reflect threshold and encoding choices tuned to this particular trace rather than a robust event-detection capability. The paper also does not state the criterion used to classify a feature in the DC-filtered voltage trace as a spike, so the binary spike/no-spike series is under-specified. A control experiment with a scrambled delayed electrical signal would reveal whether the dual-modulation scheme genuinely uses the delayed difference or whether the output mainly tracks raw optical amplitude. This concern is load-bearing because if the experimental edge-detection result is not reproducible under fixed parameters, the paper's only direct hardware evidence for an ultrafast RTD event detector is weakened.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes and numerically/experimentally investigates spiking resonant tunnelling diode (RTD) neurons for neuromorphic photonic processing. It reports (i) an experimental demonstration of rising-edge detection in a Mackey-Glass time series using dual photonic-electronic modulation of a single RTD (16/20 edges, no false positives), (ii) numerical simulations of a 20-neuron RTD-based extreme learning machine for Iris classification (up to 96.5% accuracy), and (iii) numerical simulations of a 10-neuron RTD-laser fading memory cell with attenuation-tunable memory depth. The paper uses a lumped-circuit RTD model from prior work and standard laser rate equations.","tokens_in":17860,"tokens_out":4452,"duration_ms":46067,"significance":"The work is potentially significant because it demonstrates that a single RTD can perform a spiking computation on real telecom-wavelength optical input, and it proposes scalable architectures (ELM, memory cell) that exploit RTD's speed. Strengths include use of a previously fitted model (Table 1), a data DOI, and a clear physical description of the dual-modulation idea. However, the experimental evidence is a single trace with a post-hoc threshold, and the numerical results lack reproducibility details (random matrix seed, laser parameters). If the experimental result is confirmed with controls and the simulations are made reproducible, the paper would constitute a useful step toward ultrafast photonic spiking hardware. As it stands, the evidence supports a proof-of-concept rather than the strong performance claims in the abstract.","major_comments":[{"comment":"The experimental rising-edge detection is a single demonstration on one Mackey-Glass segment. The delayed-difference threshold (0.32) is identified by inspecting the target edges on that same trace, and the optical/electrical encoding parameters (2 mW average power, -0.3 V modulation) are tuned on the same segment. With no held-out segment, no repeated trials, and no control condition (e.g., a scrambled delayed signal), the 16/20 detection rate with zero false positives cannot be distinguished from threshold fitting. Please report the spike classification criterion (how a transient in the DC-filtered trace is scored as a spike), repeat the experiment on a fresh MG segment with fixed parameters, and add a control that disrupts the delayed-difference relation.","section":"§3.1, Figs. 5–6"},{"comment":"The numerical edge-detection also selects the threshold (0.45) from the same time series. The sentence 'Once the RTD spiking threshold was correctly tuned, the event-based rising edge detection system was able to correctly identify all target features' is circular if the tuning uses the known target locations. Please specify a threshold-selection rule that does not use the target labels, or reframe the demonstration as a parameter exploration rather than a blind detection.","section":"§3.1, Fig. 4"},{"comment":"The ELM classification results are not reproducible as reported. The random input weight matrix (Nf x N) is neither specified nor fixed by a seed; the accuracy curves in Fig. 8(c,d) show single runs; and no error bars or repeated random realizations are given. The claim of 'consistently above 93%' is not supported without a spread of results. Please provide the random matrix or a reproducible seed, report mean ± std over multiple realizations, and compare against a linear baseline on the raw four features.","section":"§3.2.1, Figs. 7–8"},{"comment":"The laser rate-equation parameters (J, η, γm, γl, γnr, τph, N0) are not listed, although the RTD parameters are. The memory cell also assumes that the all-to-all optical coupling is equivalent to each neuron receiving the mean of all outputs through a single attenuator; this is an ansatz that should be justified or relaxed. Without parameter values and a coupling-model check, the quantitative memory-depth curves (Fig. 10c) are not reproducible.","section":"§3.2.2, Eqs. (4)–(5) and Fig. 10"},{"comment":"The bias distribution B_n = 0.9 − 0.185 n^2 is introduced without derivation or sensitivity analysis. Because the memory depth is controlled by the competition between bias thresholds and feedback attenuation, the claim of tunability should be supported by showing the dependence of memory depth on the bias distribution, not only on attenuation. Also, define how memory depth is measured (number of cycles until no neuron spikes, or something else) and how the 131 ps spike activation delay in the autaptic case is computed.","section":"§3.2.2, Fig. 10"}],"minor_comments":[{"comment":"The title contains a typo: 'R es-' should be 'Res-', and the author name has an odd spacing 'Jos´ e F igueiredo'.","section":"Title and author list"},{"comment":"The sentence 'A bias point in the peak was chosen... shown in Figure 7 a)' should refer to Figure 1(a), not Figure 7.","section":"§5, Experimental Methods"},{"comment":"The caption says 'Pulse widths of 20 ps are separated by 500 ps,' but the main text says '2 ns-long (negative) square pulses were used every 200 ns.' Please clarify the actual experimental pulse parameters.","section":"§3.1, Fig. 5"},{"comment":"Units are not specified for the parameters in Table 1; state that all values are in SI units or list units per parameter.","section":"Table 1"},{"comment":"In the Heaviside step function, T is not defined in the text; specify that T corresponds to the RTD spike threshold set by bias and optical amplitude.","section":"Eq. (7)"},{"comment":"There is a typo: 'we use use here' should be 'we use here'.","section":"§3.2.1"}],"recommendation":"major_revision","confidential_remarks":"The experimental edge-detection claim is the key novelty of the paper. I would encourage the editor to require the control experiments and spike-detection criterion before acceptance. The numerical sections need reproducibility details (random matrix seed, laser parameters, sensitivity analyses). The paper fits the journal's scope, and there are no concerns about authorship or citation practices."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new material is the application of an established RTD spiking model to three tasks: experimental rising-edge detection in the Mackey-Glass series (16/20 edges, no false positives), a numerically simulated 20-neuron spiking ELM that reaches 96.5% on Iris, and a 10-neuron coupled memory cell with attenuation-tuned fading depth. The spiking model and excitability map come from prior work [22,18]; the applications are new, and the experimental edge-detection is a real telecom-wavelength demonstration with a data DOI. That counts as a legitimate proof of concept.\n\nThe stress-test concern is right on the main experimental claim. The detection threshold (0.32 for the delayed difference) is set after inspecting the same trace, the spike/no-spike classification criterion is not stated, and there is one segment with no repeats and no held-out data. A scrambled-delay control would meaningfully strengthen the claim that the dual modulation, not just the raw optical amplitude, drives the output. The paper is honest in calling it a proof of concept, but the 16/20 number should not be read as a predictive rate.\n\nThe ELM section is quantitatively the thinnest. The random mixing matrix is never specified (no seed, no distribution), there are no error bars or repeated-draw statistics, and no baseline comparison to a standard classifier on the same four features. The node-significance algorithm comes from the authors' own prior paper; reusing it is not circular, but the 93% with six nodes needs to be shown to be robust to the random matrix.\n\nThe memory simulation is the cleanest of the three: the bias formula is given, the attenuation sweep produces a sensible memory-depth curve, and no hidden fitting is apparent. It is a numerical prediction, so the usual caveat about the lumped model at GHz rates applies, but the design logic is straightforward.\n\nBottom line: this is a useful proof-of-concept paper, not a quantitative benchmark. The hardware result is worth referee time, but the authors should be asked for a held-out segment, a stated spike-detection criterion, a scrambled-encoding control, and some repeat runs. I would accept it for peer review and push for those additions before publication.","headline":"Useful RTD proof-of-concept with a real experimental demo, but the quantitative claims need held-out tests and controls.","tokens_in":18370,"tokens_out":2895,"would_cite":true,"duration_ms":32509,"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":"A resonant tunnelling diode, biased at its negative differential resistance region, can act as an ultrafast event detector and, in simulation, as the neuron of photonic classifiers and tunable optical memories.","keywords":["resonant tunnelling diode","neuromorphic photonics","photonic spiking neural network","edge detection","extreme learning machine","fading memory","Mackey-Glass time series","optoelectronic spiking neuron"],"falsifier":"Repeat the experiment at the simulated speeds: drive a photo-detecting RTD with 20 picosecond optical pulses at biases across its peak and valley, and compare the measured spike/no-spike boundary with the model's threshold map; then run the chaotic time-series edge-detection with 500 picosecond pulse separation and the ten-step delay. If the boundary, the detected edges, or the roughly 300 picosecond refractory period disagrees with the model, the gigahertz-rate claims do not transfer to hardware.","tokens_in":17296,"feed_emoji":"⚡","tokens_out":8139,"duration_ms":91216,"temperature":0.7,"pith_summary":"This paper argues that a single light-sensitive resonant tunnelling diode (RTD) can serve as a multi-modal, optical-and-electronic spiking neuron that detects fast rising edges in time-series data. The authors demonstrate this experimentally on a standard chaotic time series: the diode fired at 16 of 20 target edges and produced no false positives. They then use numerical simulation to extend the same RTD neuron to larger architectures: 20 uncoupled RTDs form a photonic spiking extreme learning machine that classifies a benchmark flower dataset with up to 96.5% accuracy, and 10 coupled RTD-laser neurons form a fading memory whose storage time is set by optical attenuation. If accurate, RTDs would provide a single hardware platform for event detection, classification, and tunable short-term memory at gigahertz rates in telecom wavelength bands.","feed_headline":"RTD neuron detects 16 of 20 rising edges, no false alarms","feed_subtitle":"Single telecom-wavelength diode also powers a 96.5% photonic classifier and tunable optical memory in simulation.","key_machinery":"The load-bearing object is the RTD spiking neuron modelled as a lumped circuit, with capacitance, inductance, resistance, and the diode's nonlinear current-voltage relation, coupled to two-level laser rate equations to form an RTD-laser neuron. Its defining mechanism is threshold-controlled excitability: near the peak or valley of the current-voltage curve a small perturbation triggers an all-or-nothing spike, and the required perturbation amplitude moves with bias voltage. That single mechanism does triple duty: dual photonic-electronic modulation computes the difference between a raw signal and its delayed copy as a rising-edge detector; the same threshold acts as a binary Heaviside activation in the extreme learning machine; and a spread of bias thresholds across ten coupled devices, combined with attenuated feedback, sets how many cycles a stored spike survives.","core_discovery":"On its own terms, the paper's central discovery is that the excitability of an RTD, rooted in its steep N-shaped current-voltage curve with a negative differential resistance region, can be exploited as a controllable threshold for spiking, and that this threshold can be modulated simultaneously by light and by voltage. Encoding a raw time series in optical pulse amplitude and its delayed copy in bias voltage makes the diode fire precisely when the delayed difference exceeds threshold, so the device computes a temporal derivative in hardware. The same thresholding behaviour, read across a spatial array, implements a Heaviside activation for a photonic extreme learning machine; and when RTD outputs are reinjected through an attenuating all-to-all feedback loop, the number of regenerated cycles becomes a tunable memory depth. The experimental edge-detection result, 16 of 20 edges with no false positives, is the paper's direct evidence for the single-neuron claims; the network results are numerical.","pith_inferences":["Editorial extension: the dual-modulation edge detector could be integrated into a photonic circuit with a fixed optical delay line, making the temporal-difference operation passive and potentially faster than the electronic delay used in the experiment.","Editorial extension: the binary 'node significance' training result suggests that a hardware RTD array could be reconfigured with simple switches rather than analogue weight memories, since only a handful of binary output weights are needed.","Editorial extension: the attenuation-tuned fading memory cell could serve as a short-term buffer in a reservoir computer, although the paper itself only demonstrates storage and tuning, not readout of stored information into a downstream task.","Editorial extension: the experimental demonstration encoded each time step with 200 ns-long pulses while the simulations use 20 ps pulses; pushing the experiment toward the simulated gigahertz rates would reveal whether parasitic electrical effects, rather than the diode physics, set the practical speed limit."],"forward_implications":["A single RTD neuron can act as an ultrafast, tunable alarm or change-detector for time-series data, with its sensitivity set by either optical power or bias voltage.","Because the edge detector operates in both the optical and electronic domains, it can accept a raw optical signal and produce a spiking electronic output without a separate analogue-to-digital conversion step.","A 20-neuron RTD array can classify a benchmark dataset at gigahertz rates while training only the output layer, with peak simulated accuracy of 96.5%.","A coupled network of ten RTD-laser neurons can store a spike for a controllable number of feedback cycles, with the memory depth set simply by changing optical attenuation.","The same RTD hardware can be reconfigured between processing and memory roles because both rely on the same voltage-tunable spiking threshold."],"supporting_citations":[{"why":"It supplies the lumped-circuit RTD model, the fitted current-voltage parameters, and the earlier demonstration of optically and electronically triggered excitability that all numerical results in this paper build on.","marker":"[22]"},{"why":"It describes the experimental photo-detecting RTD device, its heterostructure, and its peak-region spiking behaviour used in the edge-detection experiment.","marker":"[23]"},{"why":"It establishes the excitable spiking regimes of RTDs and the bias conditions near the peak and valley used to trigger spikes.","marker":"[18]"},{"why":"It provides the analysis of RTD-laser spiking dynamics and the transitions in and out of the negative differential resistance region used to justify the coupled neuron model.","marker":"[19]"},{"why":"It provides the two-level laser rate equations used to convert RTD electrical spikes into optical output in the RTD-laser and memory simulations.","marker":"[25]"},{"why":"It supplies the chaotic time series used as the benchmark for rising-edge detection in both simulation and experiment.","marker":"[27]"},{"why":"It defines the extreme learning machine paradigm that the 20-RTD array implements.","marker":"[28]"},{"why":"It supplies the flower dataset classification benchmark used to test the photonic spiking extreme learning machine.","marker":"[34]"},{"why":"It introduces the node significance algorithm used for binary-weight training of the RTD array.","marker":"[35]"},{"why":"It reports the autaptic regenerative spiking memory with delayed feedback that the ten-neuron fading memory cell extends.","marker":"[37]"}],"fun_headline_variants":["RTD neuron's spike threshold computes temporal derivative in hardware","Tunable optical spike memory from coupled RTD neurons","Photonic RTD neuron detects edges, powers classifier and memory","Single RTD neuron: edge detection, classification, tunable memory","Coupled RTD neurons create adjustable photonic spiking memory"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The numerical results rest on the assumption that the fitted lumped-circuit RTD model and the laser rate equations remain quantitatively correct at the simulated gigahertz rates and for ten to twenty coupled devices; if real high-speed spiking dynamics diverge from those equations, the 96.5% classification accuracy and the memory-depth curves are predictions about a model rather than about hardware.","fun_headline_variants_meta":{"raw":{"variants":["RTD neuron's spike threshold computes temporal derivative in hardware","Tunable optical spike memory from coupled RTD neurons","Photonic RTD neuron detects edges, powers classifier and memory","Single RTD neuron: edge detection, classification, tunable memory","Coupled RTD neurons create adjustable photonic spiking memory"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000319,"raw_usage":{"total_tokens":1830,"prompt_tokens":1002,"completion_tokens":828,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":618,"completion_tokens_details":{"reasoning_tokens":743}},"tokens_in":618,"tokens_out":828,"duration_ms":9716,"temperature":1.0,"reasoning_tokens":743,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T13:11:51.300690+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the experiment at the simulated speeds: drive a photo-detecting RTD with 20 picosecond optical pulses at biases across its peak and valley, and compare the measured spike/no-spike boundary with the model's threshold map; then run the chaotic time-series edge-detection with 500 picosecond pulse separation and the ten-step delay. If the boundary, the detected edges, or the roughly 300 picosecond refractory period disagrees with the model, the gigahertz-rate claims do not transfer to hardware.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the chaotic time series used as the benchmark for rising-edge detection in both simulation and experiment."},{"cited_title":"Huang, Q.-Y","cited_arxiv_id":null,"evidence_quote":"It defines the extreme learning machine paradigm that the 20-RTD array implements."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the flower dataset classification benchmark used to test the photonic spiking extreme learning machine."},{"cited_title":"Owen-Newns, J","cited_arxiv_id":null,"evidence_quote":"It introduces the node significance algorithm used for binary-weight training of the RTD array."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It reports the autaptic regenerative spiking memory with delayed feedback that the ten-neuron fading memory cell extends."}],"review_version":1}