pith. sign in

arxiv: 2604.13450 · v1 · submitted 2026-04-15 · ⚛️ physics.optics

Distributed Coherent Optical Computing via Injection-Locked Photonic Networks

Shenghan Gao , Kathy L\"udge , Francesco Da Ros , Nathan Youngblood This is my paper

Pith reviewed 2026-05-10 12:59 UTC · model grok-4.3

classification ⚛️ physics.optics
keywords optical injection lockingcoherent photonic computingdistributed computingsemiconductor lasersphase stabilityrate equation modelreal-time processingbalanced detection
0
0 comments X

The pith

Optical injection locking synchronizes remote lasers to enable real-time coherent photonic computations without electrical conversions.

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

The paper aims to show that injecting light from a remote modulated laser into a local slave laser can lock their phases despite fiber-induced drifts, allowing direct optical interference for operations like matrix multiplications. This avoids the latency and power costs of converting signals to electrical or digital domains at each step. Using rate-equation models of semiconductor lasers, the work maps stable locking regions by varying injection power, frequency detuning, and modulation depth. End-to-end simulations with balanced detection and integration demonstrate that lower injection ratios reduce unwanted modulation transfer and raise accuracy. If the approach holds, it opens pathways for distributed photonic processors that operate continuously over optical links.

Core claim

Coherent photonic computing requires phase stability between interfering paths, a challenge for remote data sources due to environmental variations in fiber. We propose optical injection locking as a strategy to enable distributed, real-time coherent optical processing without optical-to-electrical or analog-to-digital conversions. Simulations with a semiconductor laser rate-equation model show that higher injection powers broaden the locking margin but introduce relaxation oscillations and amplitude-phase mixing, while lower powers produce a narrower yet more predictable window stable under large modulation. Symbol-sequence simulations confirm that lower injection ratios suppress residual远程

What carries the argument

Optical injection locking, in which light from a remote master laser is coupled into a slave laser to force frequency and phase synchronization, thereby preserving coherence for direct optical linear operations.

Load-bearing premise

The semiconductor laser rate-equation model plus balanced detection and temporal integration accurately represent real laser dynamics and produce reliable computational accuracy under the simulated modulation conditions.

What would settle it

A laboratory test measuring phase error and computation accuracy in a fiber-linked injection-locked laser pair under controlled temperature or vibration variations would show whether locking remains stable enough for low-error coherent operations.

Figures

Figures reproduced from arXiv: 2604.13450 by Francesco Da Ros, Kathy L\"udge, Nathan Youngblood, Shenghan Gao.

Figure 1
Figure 1. Figure 1: Overview of distributed computing strategies and proposed architecture. a) Conventional digital computing approaches require multiple conversion penalties at O-E and A-D conversion boundaries when networking distributed computing nodes. b) Incoherent optical architectures are able to perform real-time processing of data from a remote source, but encoding is limited to positive real-valued numbers and by th… view at source ↗
Figure 2
Figure 2. Figure 2: Mapping locking range for various detuning and injection ratios. a) Conceptual overview of parameter sweep and simulated analysis used to extract locking region from laser rate equations. b) Relative intensity swing of injected laser under varying injection ratios and detunings for α = 0 and α = 5. Boundaries where ∆I/I¯ ≤ 0.10 is indicated in red. c) Two-dimensional locking-region maps for α = 0 and α = 5… view at source ↗
Figure 3
Figure 3. Figure 3: Injection-ratio sweep at ∆ν = 0 with shallow modulation (ma = 0.1%, mϕ = 0.1%). a) Illustration of simulation setup to extract frequency dependent transfer curves for amplitude (and phase) modulation. Panels b) and d) show AM→AM transfer-function heatmaps over injection ratio (−60 to −20 dB) and fmod, with color indicating 20 log10 |H|, for α = 0 and α = 5, respectively. c) AM→AM and PM→PM transfer functio… view at source ↗
Figure 4
Figure 4. Figure 4: Detuning-resolved transfer-function heatmaps under modulated injection (ma = 25% and mϕ = 25%). a) Conceptual overview of simulation model. Panels b–c) correspond to injection ratio −20 dB for α = 0 and α = 5, and panels d–e) correspond to injection ratio −50 dB for α = 0 and α = 5. Because the inferred locking window is much narrower at −50 dB, panels d–e) zoom the detuning axis to ±1 GHz; the full ±10 GH… view at source ↗
Figure 5
Figure 5. Figure 5: Modulation-depth sweep at injection ratio −50 dB and ∆ν = 0. a) α = 0: AM→AM and PM→PM transfer functions for multiple modulation depths. b) α = 5: AM→AM, PM→PM, and PM-driven cross-channel responses, where localized distortion can appear near the RO neighborhood at high modulation depth (red dashed guide). c) α = 0, fmod = 2 GHz: PM→PM phase waveform in time domain for multiple depths. d) α = 5, fmod = 2 … view at source ↗
Figure 6
Figure 6. Figure 6: Numerical evaluation of distributed coherent processing through photoelectric multiply-accumulate operations. a) Conceptual illustration of simulated system used to evaluate distributed, coherent optical vector-vector multiplication. Plot shows injected laser field settling dynamics after start of OIL with a CW signal (0–10 ns, unmodulated). b) Modulated remote laser (top) and modulated injected laser fiel… view at source ↗
read the original abstract

Coherent photonic computing uses both the phase and amplitude of light to implement linear operations such as dot products and matrix multiplication but requires phase stability between the interfering paths. This poses a challenge for such strategies when optical data is generated at a remote source due to environmental phase variations in fiber. Conventional approaches to distributed computing rely on optical-to-electrical conversion and buffering, limiting truly real-time and distributed computation. Here, we propose a new strategy via optical injection locking to enable distributed, real-time coherent optical processing without unnecessary conversions in the optical-to-electrical or analog-to-digital domains. Using a semiconductor laser rate-equation model, we explore the conditions required for stable operation by sweeping the power injection ratio, frequency detuning, and modulation conditions of the remote and injected lasers. Our results indicate that higher injection powers broaden the locking margin but more readily exhibit frequency-selective features associated with relaxation oscillations and increased amplitude-phase mixing, whereas lower injection powers yield a narrower, but more predictable operating window which remains stable under large modulation depth. End-to-end symbol-sequence simulations with balanced detection and temporal integration further confirm that reducing the injection ratio suppresses residual remote-modulation components in the injected laser output and improves computational accuracy. Overall, our study provides guidance and design trade-offs for remote coherent detection and distributed coherent photonic computing enabled by injection locking.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes optical injection locking as a strategy for distributed, real-time coherent photonic computing that avoids optical-to-electrical conversions. Using standard single-mode semiconductor laser rate equations, the authors sweep injection power ratio, frequency detuning, and modulation depth to identify stable locking windows, then perform end-to-end symbol-sequence simulations with balanced detection and temporal integration to show that lower injection ratios suppress residual remote-modulation sidebands and improve computational accuracy.

Significance. If the simulated trade-offs hold in experiment, the work would provide a concrete design route for phase-stable remote coherent operations, which is relevant to scalable photonic linear algebra and distributed sensing. The parameter sweeps and end-to-end simulations constitute reproducible numerical evidence for the claimed stability-accuracy trade-off.

major comments (2)
  1. [simulation methods / rate-equation section] The rate-equation model (described in the simulation methods) omits spontaneous-emission noise, thermal frequency drifts, fiber dispersion, and possible multi-mode or gain-saturation dynamics under large modulation depths. These omissions are load-bearing for the central claim because the predicted suppression of remote-modulation sidebands and the locking-margin behavior rest directly on the deterministic dynamics of the model; inclusion of noise or dispersion could alter the reported accuracy gains and the preference for low injection ratios.
  2. [results / end-to-end simulations] No direct comparison is presented to alternative phase-stabilization techniques (e.g., optical phase-locked loops or pilot-tone methods). The end-to-end accuracy results therefore cannot be benchmarked against existing approaches, weakening the claim that injection locking offers a superior route for distributed coherent computation.
minor comments (2)
  1. Figure captions should explicitly state the integration window duration and the symbol rate used in the balanced-detection simulations so that the reported accuracy numbers can be reproduced.
  2. The abstract states that 'lower injection powers yield a narrower, but more predictable operating window'; the corresponding figures should include error bands or multiple noise realizations to quantify the claimed predictability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and positive assessment of the work's potential relevance. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [simulation methods / rate-equation section] The rate-equation model (described in the simulation methods) omits spontaneous-emission noise, thermal frequency drifts, fiber dispersion, and possible multi-mode or gain-saturation dynamics under large modulation depths. These omissions are load-bearing for the central claim because the predicted suppression of remote-modulation sidebands and the locking-margin behavior rest directly on the deterministic dynamics of the model; inclusion of noise or dispersion could alter the reported accuracy gains and the preference for low injection ratios.

    Authors: We agree that the model employs deterministic rate equations and omits spontaneous emission, thermal drifts, dispersion, and possible multi-mode effects. This simplification isolates the injection-locking mechanism and the resulting sideband suppression, which originates from the phase-locking dynamics rather than stochastic processes. The preference for lower injection ratios follows from reduced amplitude-phase coupling in the locked state, a deterministic feature. In the revised manuscript we have added an explicit limitations paragraph in the Discussion section that acknowledges these omissions and states that while absolute accuracy numbers may shift under noise, the qualitative stability-accuracy trade-off is expected to persist. Stochastic extensions are noted as future work. revision: partial

  2. Referee: [results / end-to-end simulations] No direct comparison is presented to alternative phase-stabilization techniques (e.g., optical phase-locked loops or pilot-tone methods). The end-to-end accuracy results therefore cannot be benchmarked against existing approaches, weakening the claim that injection locking offers a superior route for distributed coherent computation.

    Authors: The manuscript presents injection locking as a new strategy that avoids O/E conversion for real-time distributed operation; it does not claim superiority. We have added a concise comparative paragraph in the revised Introduction that outlines conceptual differences: injection locking is passive and requires no active feedback loop (unlike OPLLs, which introduce latency and electronics), while pilot tones consume optical power. The end-to-end simulations quantify performance within the injection-locking framework. A quantitative benchmark against the alternatives would require implementing equivalent OPLL and pilot-tone models in the same simulation environment, which is outside the present scope focused on injection locking. revision: yes

Circularity Check

0 steps flagged

No circularity: results from forward simulation of rate equations

full rationale

The paper derives its claims exclusively from numerical integration of standard single-mode semiconductor laser rate equations, sweeping injection ratio, detuning, and modulation depth to observe locking margins and residual modulation suppression. End-to-end symbol-sequence simulations with balanced detection and temporal integration are likewise forward runs of the same model. No parameters are fitted to target accuracy metrics, no self-citations supply load-bearing uniqueness theorems or ansatzes, and no prediction is definitionally equivalent to an input. The chain is therefore self-contained physical modeling rather than a closed loop.

Axiom & Free-Parameter Ledger

3 free parameters · 1 axioms · 0 invented entities

The central claim rests on the validity of the semiconductor laser rate-equation model and the assumption that simulation outcomes translate to physical accuracy.

free parameters (3)
  • power injection ratio
    Swept as a key variable to explore locking margin and stability.
  • frequency detuning
    Swept parameter in the model to determine locking conditions.
  • modulation depth and conditions
    Varied to test stability under remote modulation.
axioms (1)
  • domain assumption Semiconductor laser rate-equation model accurately represents the dynamics of injection-locked lasers under the studied conditions.
    Invoked to explore stable operation by sweeping parameters.

pith-pipeline@v0.9.0 · 5537 in / 1230 out tokens · 24379 ms · 2026-05-10T12:59:26.036265+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    Programmable photonic circuits,

    W. Bogaerts, D. P ´erez, J. Capmany, D. A. B. Miller, J. Poon, D. Englund, F. Morichetti, and A. Melloni, “Programmable photonic circuits,”Nature, vol. 586, pp. 207–216, Oct. 2020

    work page 2020

  2. [2]

    Broadcast and Weight: An Integrated Network for Scalable Photonic Spike Processing,

    A. N. Tait, M. A. Nahmias, B. J. Shastri, and P. R. Prucnal, “Broadcast and Weight: An Integrated Network for Scalable Photonic Spike Processing,” Journal of Lightwave Technology, vol. 32, no. 21, pp. 3427–3439, Nov. 2014

    work page 2014

  3. [3]

    Parallel convolutional processing using an integrated photonic tensor core,

    J. Feldmann, N. Youngblood, M. Karpovet al., “Parallel convolutional processing using an integrated photonic tensor core,”Nature, vol. 589, pp. 52–58, 2021

    work page 2021

  4. [4]

    Neuromorphic photonic networks using silicon photonic weight banks,

    A. N. Tait, T. Ferreira de Lima, E. Zhou, A. X. Wu, M. A. Nahmias, B. J. Shastri, and P. R. Prucnal, “Neuromorphic photonic networks using silicon photonic weight banks,”Scientific Reports, vol. 7, p. 7430, Aug. 2017

    work page 2017

  5. [5]

    Deep learning with coherent nanophotonic circuits,

    Y . Shen, N. C. Harris, S. Skirlo, M. Prabhu, T. Baehr-Jones, M. Hochberg, X. Sun, S. Zhao, H. Larochelle, D. Englund, and M. Solja ˇci´c, “Deep learning with coherent nanophotonic circuits,”Nature Photonics, vol. 11, no. 7, pp. 441–446, Jul. 2017

    work page 2017

  6. [6]

    Large-Scale Optical Neural Networks Based on Photoelectric Multiplication,

    R. Hamerly, L. Bernstein, A. Sludds, M. Solja ˇci´c, and D. Englund, “Large-Scale Optical Neural Networks Based on Photoelectric Multiplication,”Physical Review X, vol. 9, p. 021032, 2019

    work page 2019

  7. [7]

    11 TOPS photonic convolutional accelerator for optical neural networks,

    X. Xuet al., “11 TOPS photonic convolutional accelerator for optical neural networks,”Nature, vol. 589, pp. 44–51, 2021

    work page 2021

  8. [8]

    Netcast: Low-Power Edge Computing with WDM-defined Optical Neural Networks,

    R. Hamerly, A. Sludds, S. Bandyopadhyay, Z. Chen, Z. Zhong, L. Bernstein, and D. Englund, “Netcast: Low-Power Edge Computing with WDM-defined Optical Neural Networks,”Journal of Lightwave Technology, vol. 42, no. 22, pp. 7795–7806, 2024

    work page 2024

  9. [9]

    Delocalized photonic deep learning on the internet’s edge,

    A. Sludds, S. Bandyopadhyay, Z. Chen, Z. Zhong, J. Cochrane, L. Bernstein, D. Bunandar, P. B. Dixon, S. A. Hamilton, M. Streshinsky, A. Novack, T. Baehr-Jones, M. Hochberg, M. Ghobadi, R. Hamerly, and D. Englund, “Delocalized photonic deep learning on the internet’s edge,”Science, vol. 378, no. 6617, pp. 270–276, Oct. 2022

    work page 2022

  10. [10]

    Realization of an integrated coherent photonic platform for scalable matrix operations,

    S. Rahimi Kari, N. A. Nobile, D. Pantin, V . Shah, and N. Youngblood, “Realization of an integrated coherent photonic platform for scalable matrix operations,”Optica, vol. 11, no. 4, pp. 542–551, Apr. 2024

    work page 2024

  11. [11]

    Quantization-aware Photonic Homodyne computing for Accelerated Artificial Intelligence and Scientific Simulation,

    L. Zhou, K. Xue, A. Fallah, L. Liu, C.-H. Lee, K. Kwon, C. Cheung, Y . Li, Y . Yu, Y .-J. Lee, S. Zhao, R. Hamerly, E. Waks, D. Englund, C. Sideris, M. Yu, and Z. Chen, “Quantization-aware Photonic Homodyne computing for Accelerated Artificial Intelligence and Scientific Simulation,” arXiv:2602.08269, 2026

    work page arXiv 2026

  12. [12]

    Deep learning with coherent VCSEL neural networks,

    Z. Chen, A. Sludds, R. Davis III, I. Christen, L. Bernstein, L. Ateshian, T. Heuser, N. Heermeier, J. A. Lott, S. Reitzenstein, R. Hamerly, and D. Englund, “Deep learning with coherent VCSEL neural networks,”Nature Photonics, vol. 17, pp. 723–730, Aug. 2023

    work page 2023

  13. [13]

    An optical neural chip for implementing complex-valued neural network,

    H. Zhang, M. Gu, X. D. Jiang, J. Thompson, H. Cai, S. Paesani, R. Santagati, A. Laing, Y . Zhang, M. H. Yung, Y . Z. Shi, F. K. Muhammad, G. Q. Lo, X. S. Luo, B. Dong, D. L. Kwong, L. C. Kwek, and A. Q. Liu, “An optical neural chip for implementing complex-valued neural network,”Nature Communications, vol. 12, p. 457, Jan. 2021

    work page 2021

  14. [14]

    Optical Injection Locking: From Principle to Applications,

    Z. Liu and R. Slav ´ık, “Optical Injection Locking: From Principle to Applications,”Journal of Lightwave Technology, vol. 38, no. 1, pp. 43–59, Jan. 2020

    work page 2020

  15. [15]

    Homodyne OFDM with Optical Injection Locking for Carrier Recovery,

    Z. Liu, J.-Y . Kim, D. S. Wu, D. J. Richardson, and R. Slav ´ık, “Homodyne OFDM with Optical Injection Locking for Carrier Recovery,”Journal of Lightwave Technology, vol. 33, no. 1, pp. 34–43, Jan. 2015

    work page 2015

  16. [16]

    Injection Locking Properties of a Semiconductor Laser,

    R. Lang, “Injection Locking Properties of a Semiconductor Laser,”IEEE Journal of Quantum Electronics, vol. QE-18, no. 6, pp. 976–983, Jun. 1982

    work page 1982

  17. [17]

    Locking Conditions and Stability Properties for a Semiconductor Laser with External Light Injection,

    F. Mogensen, H. Olesen, and G. Jacobsen, “Locking Conditions and Stability Properties for a Semiconductor Laser with External Light Injection,”IEEE Journal of Quantum Electronics, vol. QE-21, no. 7, pp. 784–793, Jul. 1985

    work page 1985

  18. [18]

    A unifying view of bifurcations in a semiconductor laser subject to optical injection,

    S. Wieczorek, B. Krauskopf, and D. Lenstra, “A unifying view of bifurcations in a semiconductor laser subject to optical injection,”Optics Communications, vol. 172, pp. 279–295, 1999

    work page 1999

  19. [19]

    Frequency Response Enhancement of Optical Injection-Locked Lasers,

    E. K. Lau, H.-K. Sung, and M. C. Wu, “Frequency Response Enhancement of Optical Injection-Locked Lasers,”IEEE Journal of Quantum Electronics, vol. 44, no. 1, pp. 90–99, Jan. 2008

    work page 2008

  20. [20]

    A Study of Locking Phenomena in Oscillators,

    R. Adler, “A Study of Locking Phenomena in Oscillators,”Proceedings of the IRE, vol. 34, no. 6, pp. 351–357, Jun. 1946

    work page 1946

  21. [21]

    Locking Range and Stability of Injection Locked 1.54µm InGaAsP Semiconductor Lasers,

    C. H. Henry, N. A. Olsson, and N. K. Dutta, “Locking Range and Stability of Injection Locked 1.54µm InGaAsP Semiconductor Lasers,”IEEE Journal of Quantum Electronics, vol. QE-21, no. 8, pp. 1152–1156, Aug. 1985

    work page 1985

  22. [22]

    Time-Multiplexed Reservoir Computing with Quantum-Dot Lasers: Impact of Charge-Carrier Scattering Timescale,

    H. Dong, L. Jaurigue, and K. L ¨udge, “Time-Multiplexed Reservoir Computing with Quantum-Dot Lasers: Impact of Charge-Carrier Scattering Timescale,” Physica Status Solidi RRL, vol. 19, p. 2400433, 2025

    work page 2025

  23. [23]

    The influence of timescales and data injection schemes for reservoir computing using spin-VCSELs,

    L. M ¨uhlnickel, J. A. Jaurigue, L. C. Jaurigueet al., “The influence of timescales and data injection schemes for reservoir computing using spin-VCSELs,” Communications Physics, vol. 7, p. 370, 2024

    work page 2024

  24. [24]

    Optically injected quantum dot lasers: impact of nonlinear carrier lifetimes on frequency-locking dynamics,

    J. Pausch, C. Otto, E. Tylaite, N. Majer, E. Sch ¨oll, and K. L ¨udge, “Optically injected quantum dot lasers: impact of nonlinear carrier lifetimes on frequency-locking dynamics,”New Journal of Physics, vol. 14, p. 053018, 2012

    work page 2012

  25. [25]

    Amplitude Noise Limiting Amplifier for Phase Encoded Signals Using Injection Locking in Semiconductor Lasers,

    A. Fragkos, A. Bogris, D. Syvridis, and R. Phelan, “Amplitude Noise Limiting Amplifier for Phase Encoded Signals Using Injection Locking in Semiconductor Lasers,”Journal of Lightwave Technology, vol. 30, no. 5, pp. 764–771, Mar. 2012. 13

    work page 2012

  26. [26]

    Bounded phase phenomena in the optically injected laser,

    B. Kelleher, D. Goulding, B. Baselga Pascual, S. P. Hegarty, and G. Huyet, “Bounded phase phenomena in the optically injected laser,”Physical Review E, vol. 85, p. 046212, 2012

    work page 2012

  27. [27]

    Bandwidth Enhancement by Master Modulation of Optical Injection-Locked Lasers,

    E. K. Lau, L. J. Wong, X. Zhao, Y .-K. Chen, C. J. Chang-Hasnain, and M. C. Wu, “Bandwidth Enhancement by Master Modulation of Optical Injection-Locked Lasers,”Journal of Lightwave Technology, vol. 26, no. 15, pp. 2584–2593, Aug. 2008

    work page 2008

  28. [28]

    Clock and data recovery of an extended Manchester code for PAM-n,

    H. Chung, S.-b. Jeon, D.-w. Lee, H.-j. Park, J.-h. Kim, and S.-h. Sohn, “Clock and data recovery of an extended Manchester code for PAM-n,” inProc. Int. Conf. ICT Convergence (ICTC), 2012, pp. 436–440

    work page 2012