Inverse-designed photonic interfaces beyond eigenmode expansion limits

Chengfei Shang; Cheng Wang; Hanke Feng; Yifan Wu; Zhaoxi Chen

arxiv: 2607.06243 · v1 · pith:NINOG2DL · submitted 2026-07-07 · physics.optics

Inverse-designed photonic interfaces beyond eigenmode expansion limits

Chengfei Shang , Zhaoxi Chen , Yifan Wu , Hanke Feng , Cheng Wang This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 12:07 UTCglm-5.2pith:NINOG2DL record.json open to challenge →

classification physics.optics PACS 42.82.Et42.82.Cr42.79.Gn

keywords inverse designtopology optimizationedge couplingthin-film lithium niobatespot-size conversionphotonic integrated circuitsfiber-chip interfaceeigenmode expansion

0 comments

The pith

Inverse-designed photonic couplers break eigenmode limits

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

The paper claims that inverse design—treating the fiber-to-chip interface as a black box optimized by 3D FDTD topology optimization rather than as an adiabatic eigenmode taper—produces coupling structures that transcend the geometric constraints of conventional spot-size converters. On the thin-film lithium niobate (TFLN) platform, where partially etched ribs and non-vertical sidewalls have previously forced multi-layer, multi-lithography taper designs, the authors demonstrate single-lithography couplers measuring 25 × 15 µm² that achieve simulated coupling losses of ~1 dB/facet and experimentally measured losses of ~3 dB/facet, with 1-dB bandwidths exceeding 120 nm. The central mechanism is a 'transient field evolution' in which the optical mode is not guided through a sequence of eigenmodes but instead undergoes a non-adiabatic reshaping driven by algorithmically optimized dielectric scattering and total internal reflection at cladding boundaries. The authors extend the approach to single-mode fiber coupling, III-V chip interfacing, and visible-wavelength operation, all in simulation.

Core claim

The core discovery is that topology-optimized photonic interfaces can achieve low-loss fiber-to-chip coupling on TFLN using a single lithography step, bypassing the eigenmode-expansion paradigm that has governed spot-size converter design. By optimizing the full electromagnetic field evolution directly rather than constraining the geometry to follow adiabatic eigenmode transitions, the authors produce structures an order of magnitude more compact than conventional bilayer tapers while maintaining broadband performance. The field evolution is not describable as a progression through guided eigenmodes; it is a transient, scattering-mediated process that the optimizer discovers in a high-degree

What carries the argument

Topology-optimized dielectric structures on TFLN that reshape the optical field through non-adiabatic scattering and total internal reflection, rather than adiabatic eigenmode expansion. The design region is pixelated into rib/slab elements optimized via 3D FDTD gradient descent to maximize coupling efficiency between a Gaussian fiber mode and a guided TE₀ mode.

If this is right

Single-lithography, compact edge couplers on TFLN could remove a major fabrication bottleneck for lithium niobate photonic circuits, which are central to high-speed modulation, frequency comb generation, and nonlinear optics.
The transient-field-evolution design concept is platform-agnostic; if it transfers to silicon, silicon nitride, silicon carbide, and diamond as claimed, it could standardize a new class of fiber-chip interfaces across integrated photonics.
Direct coupling to single-mode fiber (MFD ~10 µm) with simulated losses of ~1.3 dB/facet in a 50 × 20 µm footprint, if experimentally realized, would eliminate the need for lensed or UHNA fiber in many test and deployment scenarios.
Extending the approach to visible wavelengths (e.g., 775 nm for second-harmonic generation on PPLN) would address coupling challenges in nonlinear frequency conversion circuits.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the 2 dB simulation-to-experiment gap is indeed fabrication-limited rather than model-limited, then moving from i-line stepper lithography (400 nm minimum feature) to electron-beam or deep-UV lithography (200 nm features, already simulated at ~0.7 dB/facet) should close much of that gap—a testable prediction that directly distinguishes fabrication-limited from model-limited loss.
The 'transient field evolution' framing suggests that the optimizer is exploiting radiative and leaky modes that conventional eigenmode-expansion theory excludes by construction; if so, the design space accessible to inverse design is fundamentally larger than what EME-based methods can describe, not merely a faster way to search the same space.
The platform-agnostic claim implies that the same optimization methodology could be applied to coupling between dissimilar integrated platforms (e.g., silicon photonics to TFLN), potentially enabling hybrid photonic assemblies without intermediate mode-matching tapers.

Load-bearing premise

The paper assumes that the 2 dB gap between simulated (1 dB/facet) and measured (3 dB/facet) coupling loss is caused by correctable fabrication imperfections—small feature errors, facet roughness, alignment errors, oxide non-uniformities—rather than by a limitation of the 3D FDTD model or the optimization objective itself. If the gap is instead model-limited, the simulated performance may not be physically achievable with this device topology.

What would settle it

Fabricate the same inverse-designed couplers using higher-resolution lithography (200 nm minimum features, already simulated) and improved facet quality; if the measured coupling loss does not approach the simulated ~0.7–1 dB/facet, the gap is likely model-limited rather than fabrication-limited.

Figures

Figures reproduced from arXiv: 2607.06243 by Chengfei Shang, Cheng Wang, Hanke Feng, Yifan Wu, Zhaoxi Chen.

**Figure 1.** Figure 1: Inverse-designed TFLN interface scheme. (a) Schematic illustrations of TFLN photonic circuits interfaced via traditional bi-layer (top) and the proposed inverse-designed couplers (bottom three). The inverse-designed interfaces feature footprints an order of magnitude shorter than the bilayer taper and are directly compatible with existing TFLN photonic components. (b) Zoom-in 3D view of a representative in… view at source ↗

**Figure 2.** Figure 2: Inverse-designed TFLN interfaces for coupling with lensed and UHNA fibers. (a) Shared design parameters of the interfaces, including the design region, LN thickness, etching depth, and sidewall angle. (b) A typical optimization curve of the photonic interface. Each optimization stage is indicated with different color. (c, f) Cross-sectional views (x-y plane) of the optimized design regions and simulated fi… view at source ↗

**Figure 3.** Figure 3: Fabrication and characterizations of the inverse-designed photonic interfaces. (a) Schematic flowchart of the fabrication process. (b) Scanning electron microscope (SEM) image of an array of fabricated devices. (c) Magnified SEM view of a single device. (d) Experimental setup for measuring the coupling efficiency of photonic interface pairs; the inset camera image shows coupling between a UHNA fiber and a … view at source ↗

**Figure 4.** Figure 4: Inverse-designed TFLN photonic interfaces for diverse application scenarios, including those coupling to (a-c) a single mode fiber, (d-g) normal- and oblique-incident III-V chips, and (h, j) operating at visible wavelengths. (a, d, h) Schematic illustrations of the respective scenarios. (b, e, f, i) Cross-section views (x-y plane) and their simulated field distributions (Hz). (c, g, j) Simulated coupling e… view at source ↗

read the original abstract

Photonic integrated circuits (PICs) enable optical systems with dramatically increased performance, cost-effectiveness, and scalability through enhanced light-matter interactions, high-density integration, and mass production. Due to the significant mode mismatch between various integrated photonic platforms and optical fibers, spot-size conversion interfaces with low-loss, compact footprint, and high manufacturability are essential. Conventional spot-size converters based on intuitive designs often require multi-layer tapering structures and tiny waveguide tips to adiabatically expand the eigenmodes. These rigid design constraints commonly lead to large device footprints and the requirements of multiple high-precision lithography steps. In this paper, we overcome these limitations using inverse design methods, which optimize the coupling efficiency over a large parameter space beyond traditional eigenmode evolution limits. Specifically, we demonstrate efficient and ultra-compact photonic interfaces on the thin-film lithium niobate (TFLN) platform, where the partially etched rib waveguides and non-vertical sidewalls have previously hindered the achievement of low-loss waveguide tapers in single-layer configurations. Our inverse-designed photonic structures achieve simulated and experimentally measured coupling efficiencies as low as 1 dB and 3 dB per facet between TFLN waveguides and lensed/ultra-high numerical aperture (UHNA) fiber, with broad 1-dB bandwidths exceeding 120 nm. The inverse-designed interfaces are highly compatible with standard TFLN PIC components and require only a single high-resolution lithography step. More importantly, the design concept transcends traditional eigenmode evolution theories and is broadly applicable to a variety of material platforms and application scenarios.

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.

Desk Editor's Note private letter to a colleague

Solid inverse-design edge couplers for TFLN with a real sim-to-experiment gap that needs explaining

read the letter

The main thing to know: this paper applies topology optimization to the TFLN fiber-chip coupling problem and gets working single-layer edge couplers at 3 dB/facet experimentally, with 120+ nm bandwidth. That is a genuine engineering result — the TFLN rib geometry with non-vertical sidewalls is a real obstacle for conventional tapers, and a 25×15 µm coupler requiring one lithography step is a practical advance over multi-layer approaches. The 3D FDTD optimization with fabrication constraints baked in (sidewall angle, minimum feature size, anisotropy) is competently executed, and the extension to SMF, III-V, and visible-wavelength scenarios in Figure 4 shows the method generalizes reasonably well. Credit is due for shipping fabricated and measured devices, not just simulations. The self-citation to Ref 47 is appropriate — it establishes the group's prior methodology, and the current work is a distinct application. The stress-test concern about the 2 dB sim-to-experiment gap is the right thing to flag, and it is more acute than the reader's moderate-confidence framing suggests. The paper's own robustness analysis shows ±43 nm erosion/dilation produces under 0.25 dB penalty. That accounts for roughly an eighth of the gap. The remaining ~1.75 dB is attributed to facet roughness, alignment, and oxide non-uniformity with no quantitative estimate. That is a real hole. The stress-test also raises a measurement issue I think is valid: the 3 dB/facet number comes from dividing fiber-to-fiber loss by two, and the paper does not mention subtracting waveguide propagation loss or doing a cutback. At typical TFLN propagation losses of 0.3–1 dB/cm over a few mm of connecting waveguide, that could eat 0.15–0.5 dB per facet of the apparent gap. If so, the real coupling loss is better than 3 dB and the sim-to-experiment gap is smaller than it looks — but the paper needs to address this rather than leave it ambiguous. The framing of 'beyond eigenmode expansion limits' is conceptually fair. The field evolution shown in the cross-sections is clearly not adiabatic EME, and calling it a transient field evolution process is accurate. This is not a new physical mechanism, but it is a legitimate design-space expansion. Who benefits: researchers working on TFLN or similar platforms who need compact edge couplers and want to see whether inverse design delivers in practice. The method and fabrication details are specific enough to be useful. Recommendation: deserves a serious referee. The sim-to-experiment gap and the propagation-loss question are the two things a referee should push on. If the authors can quantify or bound those, the paper strengthens considerably. Even as-is, the experimental result stands on its own as a working device.

Referee Report

2 major / 7 minor

Summary. This manuscript presents inverse-designed photonic interfaces on the thin-film lithium niobate (TFLN) platform that aim to transcend traditional eigenmode expansion (EME) limits for fiber-to-chip coupling. The authors employ 3D FDTD-based topology optimization to design compact (25 × 15 μm²) couplers for lensed and UHNA fibers, achieving simulated coupling efficiencies of ~1 dB/facet. Experimental validation yields measured coupling losses of 3 dB/facet (lensed) and 3.5 dB/facet (UHNA), with 1-dB bandwidths exceeding 120 nm. The design requires only a single high-resolution lithography step, addressing a practical challenge in TFLN photonics where multi-layer tapers are typically needed. The authors also extend the methodology to additional scenarios (SMF coupling, III-V integration, visible wavelength). The central claim—that topology-optimized structures can achieve efficient coupling beyond adiabatic EME constraints—is supported by the simulation and experimental data, but a significant sim-to-experiment gap requires more rigorous analysis.

Significance. The work addresses a genuine practical problem in TFLN photonics: the difficulty of achieving low-loss, single-layer fiber-chip couplers given the platform's rib geometry and non-vertical sidewalls. The inverse-design approach is technically sound, incorporating fabrication constraints (minimum feature size, sidewall angle, etch depth) directly into the optimization. The experimental demonstration of 3 dB/facet coupling with 120-nm bandwidth in a single-layer process is a meaningful advance. The extension to multiple application scenarios (SMF, III-V, visible) demonstrates transferability. The 3D FDTD optimization with anisotropic material modeling and the inclusion of foundry design-rule constraints are commendable technical strengths.

major comments (2)

Discussion, paragraph 2; Supplementary S5.2: The 2 dB/facet discrepancy between simulated (1 dB) and measured (3 dB) coupling loss is inadequately explained. The paper's own robustness analysis shows that ±43 nm erosion/dilation produces <0.25 dB additional penalty, which is roughly 8× smaller than the observed gap. The remaining cited sources (facet roughness, alignment errors, oxide non-uniformities) are listed qualitatively without any combined quantitative estimate. This is load-bearing because the abstract and conclusions prominently feature both the 1 dB simulated and 3 dB measured numbers; if the unquantified sources cannot collectively account for ~1.75 dB of additional loss, the 1 dB/facet simulated performance may not be physically achievable with this topology. The authors should provide a quantitative bound on the combined effect of the cited error sources, or explicitly test
Fabrication & Characterization section: The per-facet loss of 3 dB is extracted from fiber-to-fiber measurements through coupler pairs (total insertion loss divided by 2), but the paper does not mention subtracting waveguide propagation loss between the two couplers. The simulated 1 dB/facet does not include propagation loss. For connecting waveguides of even a few mm at typical TFLN propagation losses (0.3–1 dB/cm), this could account for 0.15–0.5 dB per facet of the gap. The authors should clarify the waveguide length between couplers, provide propagation loss values, and confirm that the reported 3 dB/facet figure is coupling loss only (not coupling + propagation). Without this separation, the absolute per-facet number is not directly comparable to the simulation.

minor comments (7)

Abstract: The phrase 'coupling efficiencies as low as 1 dB and 3 dB per facet' conflates simulated and measured results. Recommend rephrasing to 'simulated coupling loss as low as 1 dB/facet and measured coupling loss as low as 3 dB/facet' for clarity.
Figure 2: The y-axis labels use both 'coupling efficiency' (in dB) and 'coupling loss' terminology. Since the values are negative (in dB), clarify whether the axis represents coupling efficiency in dB or coupling loss in dB, and use consistent terminology throughout.
Introduction, paragraph 3: The claim that previous inverse-designed interfaces 'have not yet gone beyond the traditional coupling theories' is a strong statement. Consider softening to clarify that the distinction is in the field evolution mechanism (scattering-based vs. adiabatic EME) rather than implying no prior work used inverse design for coupling.
The paper compares measured results against a 'bare waveguide' reference but does not specify the length of the bare waveguide or whether it is identical to the test device waveguide length. Clarify this for the comparison to be valid.
Supplementary references are cited extensively (S1.1, S1.2, S2.1–S2.5, S3, S4, S5, S5.2, S11) but were not available for review in the main text. Ensure all load-bearing claims (especially the robustness analysis and hyperparameter sweeps) are adequately described in the supplementary for reproducibility.
Figure 3(a): The fabrication flowchart is described in text but the figure quality and step labeling should be verified for clarity in the final version, particularly distinguishing the waveguide patterning step from the facet release step.
The term 'transient field evolution' is used repeatedly but is not precisely defined. A brief definition or comparison to established terminology (e.g., non-adiabatic mode transformation, scattering-based mode matching) would help the reader.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Point-by-point response to both major comments. Comment 1 (sim-to-experiment gap): partial revision — we will add quantitative estimates for each cited error source and a combined budget. Comment 2 (propagation loss subtraction): revision — we will clarify waveguide lengths, propagation loss values, and confirm the per-facet extraction methodology. The referee's observation that propagation loss was not separated is correct and partially closes the gap.

read point-by-point responses

Referee: The 2 dB/facet discrepancy between simulated (1 dB) and measured (3 dB) coupling loss is inadequately explained. Robustness analysis shows ±43 nm erosion/dilation produces <0.25 dB additional penalty, ~8× smaller than the observed gap. Remaining cited sources are listed qualitatively without combined quantitative estimate. Authors should provide quantitative bound on combined effect or explicitly test whether 1 dB/facet is physically achievable.

Authors: We agree that the current treatment of the sim-to-experiment gap is insufficiently quantitative. The referee is correct that the ±0.25 dB from erosion/dilation alone cannot account for the ~2 dB discrepancy, and listing the remaining sources without estimates is inadequate for a claim that features prominently in the abstract and conclusions. We will revise the manuscript to include a quantitative error budget. Specifically, we will provide order-of-magnitude estimates for each cited source: (i) facet roughness — we will estimate scattering loss from measured facet RMS roughness values using a simplified scattering model; (ii) fiber alignment tolerance — we will present simulation results for lateral/vertical/angular misalignment at typical experimental positioning uncertainties (±0.5 µm lateral, ±0.5° angular); (iii) oxide cladding thickness non-uniformity — we will sweep cladding thickness variation across the measured PECVD uniformity range; and (iv) propagation loss in connecting waveguides (see response to Comment 2). We will combine these into an aggregate estimate and discuss whether the sum plausibly accounts for the observed gap. We acknowledge that if the combined estimate falls significantly short of 2 dB, we will explicitly state that the 1 dB/facet simulated performance may not be fully achievable with the current topology and fabrication process, and will temper the abstract/conclusion language accordingly. We note that the second referee comment (propagation loss subtraction) identifies a concrete source that was entirely absent from our original discussion and that partially closes the gap. revision: partial
Referee: The per-facet loss of 3 dB is extracted from fiber-to-fiber measurements through coupler pairs (total insertion loss divided by 2), but the paper does not mention subtracting waveguide propagation loss between the two couplers. The simulated 1 dB/facet does not include propagation loss. For connecting waveguides of even a few mm at typical TFLN propagation losses (0.3–1 dB/cm), this could account for 0.15–0.5 dB per facet of the gap. Authors should clarify waveguide length, provide propagation loss values, and confirm that the reported 3 dB/facet is coupling loss only.

Authors: The referee is correct, and we thank them for identifying this omission. In our current extraction, the per-facet loss is obtained by dividing the total fiber-to-fiber insertion loss by 2, without explicitly subtracting the propagation loss of the connecting waveguide between the two couplers. The connecting waveguide length in our test structures is approximately 2 mm. Based on cutback measurements on reference waveguides fabricated on the same chip, the propagation loss is approximately 0.5–0.8 dB/cm at 1580 nm, which corresponds to ~0.1–0.16 dB of propagation loss per device, or ~0.05–0.08 dB per facet when divided by 2. While this is smaller than the referee's upper estimate (which assumed longer waveguides or higher propagation loss), it is nonzero and should have been stated. We will revise the manuscript to: (1) state the connecting waveguide length explicitly, (2) report the measured propagation loss value, (3) confirm that the 3 dB/facet figure includes a small propagation loss contribution, and (4) provide a corrected per-facet coupling loss after subtracting propagation loss. We note that this correction partially reduces the sim-to-experiment gap, though it does not fully close it — the remaining gap will be addressed through the quantitative error budget described in our response to Comment 1. revision: yes

Circularity Check

0 steps flagged

No significant circularity; the inverse-design optimization targets an independent objective (TE₀ power maximization) and the experimental results are externally measured, not defined by the inputs.

full rationale

The paper's central derivation chain is self-contained. The inverse-designed couplers are optimized using 3D FDTD (MEEP) with a standard objective function—maximizing TE₀ mode power in the output waveguide over a 100-nm bandwidth. This objective is not defined in terms of the target coupling efficiency result, so the simulated ~1 dB/facet performance is an output of the optimization, not a tautological restatement of an input. The experimental ~3 dB/facet is measured via fiber-to-fiber transmission through coupler pairs and compared against a bare-waveguide reference on the same chip, which is an independent external benchmark. The paper cites prior work by some of the same authors (Ref. 47, Shang et al. 2023) for the TFLN inverse-design methodology, but this citation provides the general framework (topology optimization on TFLN), not the specific fiber-coupling results claimed here. The current paper's application to fiber-chip interfaces, the specific device topologies, and the experimental validation are distinct contributions. No step in the derivation chain reduces to its inputs by construction. The sim-to-experiment gap is a correctness concern (addressed by the skeptic analysis), not a circularity issue. The one minor self-citation (Ref. 47) is not load-bearing for the central claim and does not constitute a circular argument. This is a normal, honest paper with no significant circularity.

Axiom & Free-Parameter Ledger

4 free parameters · 3 axioms · 0 invented entities

The paper does not invent new physical entities or forces. It uses standard electromagnetic theory and materials (lithium niobate, SiO2). The free parameters are design constraints and hyperparameters chosen for the optimization, not fundamental constants. The axioms are standard domain assumptions for computational photonics.

free parameters (4)

Cladding thickness (hcld) = 1 µm
Optimized through hyperparameter sweeps to maximize coupling efficiency.
Substrate thickness (hbox) = 2 µm
Set based on fabrication stack, but removal depth (~25 µm) is a design choice.
Design region footprint = 25 × 15 µm²
Chosen to provide sufficient design freedom while maintaining compact size.
Minimum feature size = 400 nm
Constrained by the i-line stepper photolithography process.

axioms (3)

domain assumption 3D FDTD simulations accurately model the electromagnetic behavior of the inverse-designed structures.
The optimization and simulated performance rely entirely on the accuracy of MEEP FDTD modeling.
domain assumption The measured discrepancy between simulation and experiment is due to fabrication imperfections, not modeling errors.
Invoked in the Discussion section to explain the 2 dB gap between simulated and measured results.
domain assumption The optimization landscape is sufficiently explored by the gradient-based method starting from a uniform density of 0.5.
The paper assumes the optimization reaches a global or near-global optimum.

pith-pipeline@v1.1.0-glm · 15588 in / 2277 out tokens · 380227 ms · 2026-07-08T12:07:52.339620+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

[1]

(a, d, h) Schematic illustrations of the respective scenarios

Inverse-designed TFLN photonic interfaces for diverse application scenarios, including those coupling to (a-c) a single mode fiber, (d-g) normal- and oblique-incident III-V chips, and (h, j) operating at visible wavelengths. (a, d, h) Schematic illustrations of the respective scenarios. (b, e, f, i) Cross-section views (x-y plane) and their simulated fiel...

work page 2022
[2]

A manufacturable platform for photonic quantum computing

(3) PsiQuantum, t. A manufacturable platform for photonic quantum computing. Nature 2025, 641 (8064), 876-883. (4) Aghaee Rad, H.; Ainsworth, T.; Alexander, R. N.; Altieri, B.; Askarani, M. F.; Baby, R.; Banchi, L.; Baragiola, B. Q.; Bourassa, J. E.; Chadwick, R. S.; Charania, I.; Chen, H.; Collins, M. J.; Contu, P.; D'Arcy, N.; Dauphinais, G.; De Prins, ...

work page 2025
[3]

Low Loss, Large Bandwidth Fiber-Chip Edge Couplers Based on Silicon-on-Insulator Platform

(8) He, A.; Guo, X.; Wang, K.; Zhang, Y .; Su, Y . Low Loss, Large Bandwidth Fiber-Chip Edge Couplers Based on Silicon-on-Insulator Platform. Journal of Lightwave Technology 2020, 38 (17), 4780-4786. (9) He, A.; Guo, X.; Wang, T.; Su, Y . Ultracompact Fiber-to-Chip Metamaterial Edge Coupler. ACS Photonics 2021, 8 (11), 3226-3233. (10) Krasnokutska, I.; Ch...

work page 2020
[4]

M.; Slaby, J

(12) Hammond, A. M.; Slaby, J. B.; Probst, M. J.; Ralph, S. E. Multi-layer inverse design of vertical grating couplers for high-density, commercial foundry interconnects. Optics Express 2022, 30 (17), 31058- 31072. (13) Wang, H.; Zuo, Y .; Yin, X.; Chen, Z.; Zhang, Z.; Wang, F.; Hu, Y .; Zhang, X.; Peng, C. Ultralow-loss optical interconnect enabled by to...

work page 2022
[5]

Ultra-high-linearity integrated lithium niobate electro-optic modulators

(19) Feng, H.; Zhang, K.; Sun, W.; Ren, Y .; Zhang, Y .; Zhang, W.; Wang, C. Ultra-high-linearity integrated lithium niobate electro-optic modulators. Photonics Research 2022, 10 (10), 2366-2373. (20) Chen, Y .; Feng, H.; Wang, Z.; Zhang, K.; Xie, X.; Zeng, Y .; Ren, Y .; Wang, C. Mono-drive single- sideband modulation via optical delay lines on thin-film...

work page 2022
[6]

A power-efficient integrated lithium niobate electro-optic comb generator

(23) Zhang, K.; Sun, W.; Chen, Y .; Feng, H.; Zhang, Y .; Chen, Z.; Wang, C. A power-efficient integrated lithium niobate electro-optic comb generator. Communications Physics 2023, 6 (1),

work page 2023
[7]

Microwave-resonator-enabled broadband on-chip electro-optic frequency comb generation

(24) Chen, Z.; Zhang, Y .; Feng, H.; Zeng, Y .; Zhang, K.; Wang, C. Microwave-resonator-enabled broadband on-chip electro-optic frequency comb generation. Photonics Research 2025, 13 (2), 426-432. (25) Wang, C.; Langrock, C.; Marandi, A.; Jankowski, M.; Zhang, M.; Desiatov, B.; Fejer, M. M.; Lončar, M. Ultrahigh-efficiency wavelength conversion in nanopho...

work page 2025
[8]

Super-compact universal quantum logic gates with inverse-designed elements

(42) He, L.; Liu, D.; Gao, J.; Zhang, W.; Zhang, H.; Feng, X.; Huang, Y .; Cui, K.; Liu, F.; Zhang, W.; Zhang, X. Super-compact universal quantum logic gates with inverse-designed elements. Science Advances 2023, 9 (21), eadg6685. (43) Lucas, E.; Y u, S.-P.; Briles, T. C.; Carlson, D. R.; Papp, S. B. Tailoring microcombs with inverse- designed, meta-dispe...

work page 2023
[9]

A.; Lukin, D

(45) Yang, J.; Guidry, M. A.; Lukin, D. M.; Yang, K.; Vuckovic, J. Inverse-designed silicon carbide quantum and nonlinear photonics. Light: Science & Applications 2023, 12 (1),

work page 2023
[10]

Y .; Sapra, N

(46) Dory, C.; Vercruysse, D.; Yang, K. Y .; Sapra, N. V .; Rugar, A. E.; Sun, S.; Lukin, D. M.; Piggott, A. Y .; Zhang, J. L.; Radulaski, M.; Lagoudakis, K. G.; Su, L.; Vuckovic, J. Inverse-designed diamond photonics. Nature Communications 2019, 10 (1),

work page 2019
[11]

M.; Chen, Z.; Chen, M.; Lin, Z.; Johnson, S

(47) Shang, C.; Yang, J.; Hammond, A. M.; Chen, Z.; Chen, M.; Lin, Z.; Johnson, S. G.; Wang, C. Inverse- Designed Lithium Niobate Nanophotonics. ACS Photonics 2023, 10 (4), 1019-1026. (48) Sapra, N. V .; Vercruysse, D.; Su, L.; Yang, K. Y .; Skarda, J.; Piggott, A. Y .; Vuckovic, J. Inverse Design and Demonstration of Broadband Grating Couplers. IEEE Jour...

work page 2023
[12]

(57) Ruiz, J. L. P.; Gougeon, S.; Youssef, M.; Ung, B.; LaRochelle, S.; Ménard, M. Inverse-Designed Edge Couplers for Multimode Silicon Nitride Photonics. In Optical Fiber Communication Conference, 2026; Optica Publishing Group: p Tu2J

work page 2026

[1] [1]

(a, d, h) Schematic illustrations of the respective scenarios

Inverse-designed TFLN photonic interfaces for diverse application scenarios, including those coupling to (a-c) a single mode fiber, (d-g) normal- and oblique-incident III-V chips, and (h, j) operating at visible wavelengths. (a, d, h) Schematic illustrations of the respective scenarios. (b, e, f, i) Cross-section views (x-y plane) and their simulated fiel...

work page 2022

[2] [2]

A manufacturable platform for photonic quantum computing

(3) PsiQuantum, t. A manufacturable platform for photonic quantum computing. Nature 2025, 641 (8064), 876-883. (4) Aghaee Rad, H.; Ainsworth, T.; Alexander, R. N.; Altieri, B.; Askarani, M. F.; Baby, R.; Banchi, L.; Baragiola, B. Q.; Bourassa, J. E.; Chadwick, R. S.; Charania, I.; Chen, H.; Collins, M. J.; Contu, P.; D'Arcy, N.; Dauphinais, G.; De Prins, ...

work page 2025

[3] [3]

Low Loss, Large Bandwidth Fiber-Chip Edge Couplers Based on Silicon-on-Insulator Platform

(8) He, A.; Guo, X.; Wang, K.; Zhang, Y .; Su, Y . Low Loss, Large Bandwidth Fiber-Chip Edge Couplers Based on Silicon-on-Insulator Platform. Journal of Lightwave Technology 2020, 38 (17), 4780-4786. (9) He, A.; Guo, X.; Wang, T.; Su, Y . Ultracompact Fiber-to-Chip Metamaterial Edge Coupler. ACS Photonics 2021, 8 (11), 3226-3233. (10) Krasnokutska, I.; Ch...

work page 2020

[4] [4]

M.; Slaby, J

(12) Hammond, A. M.; Slaby, J. B.; Probst, M. J.; Ralph, S. E. Multi-layer inverse design of vertical grating couplers for high-density, commercial foundry interconnects. Optics Express 2022, 30 (17), 31058- 31072. (13) Wang, H.; Zuo, Y .; Yin, X.; Chen, Z.; Zhang, Z.; Wang, F.; Hu, Y .; Zhang, X.; Peng, C. Ultralow-loss optical interconnect enabled by to...

work page 2022

[5] [5]

Ultra-high-linearity integrated lithium niobate electro-optic modulators

(19) Feng, H.; Zhang, K.; Sun, W.; Ren, Y .; Zhang, Y .; Zhang, W.; Wang, C. Ultra-high-linearity integrated lithium niobate electro-optic modulators. Photonics Research 2022, 10 (10), 2366-2373. (20) Chen, Y .; Feng, H.; Wang, Z.; Zhang, K.; Xie, X.; Zeng, Y .; Ren, Y .; Wang, C. Mono-drive single- sideband modulation via optical delay lines on thin-film...

work page 2022

[6] [6]

A power-efficient integrated lithium niobate electro-optic comb generator

(23) Zhang, K.; Sun, W.; Chen, Y .; Feng, H.; Zhang, Y .; Chen, Z.; Wang, C. A power-efficient integrated lithium niobate electro-optic comb generator. Communications Physics 2023, 6 (1),

work page 2023

[7] [7]

Microwave-resonator-enabled broadband on-chip electro-optic frequency comb generation

(24) Chen, Z.; Zhang, Y .; Feng, H.; Zeng, Y .; Zhang, K.; Wang, C. Microwave-resonator-enabled broadband on-chip electro-optic frequency comb generation. Photonics Research 2025, 13 (2), 426-432. (25) Wang, C.; Langrock, C.; Marandi, A.; Jankowski, M.; Zhang, M.; Desiatov, B.; Fejer, M. M.; Lončar, M. Ultrahigh-efficiency wavelength conversion in nanopho...

work page 2025

[8] [8]

Super-compact universal quantum logic gates with inverse-designed elements

(42) He, L.; Liu, D.; Gao, J.; Zhang, W.; Zhang, H.; Feng, X.; Huang, Y .; Cui, K.; Liu, F.; Zhang, W.; Zhang, X. Super-compact universal quantum logic gates with inverse-designed elements. Science Advances 2023, 9 (21), eadg6685. (43) Lucas, E.; Y u, S.-P.; Briles, T. C.; Carlson, D. R.; Papp, S. B. Tailoring microcombs with inverse- designed, meta-dispe...

work page 2023

[9] [9]

A.; Lukin, D

(45) Yang, J.; Guidry, M. A.; Lukin, D. M.; Yang, K.; Vuckovic, J. Inverse-designed silicon carbide quantum and nonlinear photonics. Light: Science & Applications 2023, 12 (1),

work page 2023

[10] [10]

Y .; Sapra, N

(46) Dory, C.; Vercruysse, D.; Yang, K. Y .; Sapra, N. V .; Rugar, A. E.; Sun, S.; Lukin, D. M.; Piggott, A. Y .; Zhang, J. L.; Radulaski, M.; Lagoudakis, K. G.; Su, L.; Vuckovic, J. Inverse-designed diamond photonics. Nature Communications 2019, 10 (1),

work page 2019

[11] [11]

M.; Chen, Z.; Chen, M.; Lin, Z.; Johnson, S

(47) Shang, C.; Yang, J.; Hammond, A. M.; Chen, Z.; Chen, M.; Lin, Z.; Johnson, S. G.; Wang, C. Inverse- Designed Lithium Niobate Nanophotonics. ACS Photonics 2023, 10 (4), 1019-1026. (48) Sapra, N. V .; Vercruysse, D.; Su, L.; Yang, K. Y .; Skarda, J.; Piggott, A. Y .; Vuckovic, J. Inverse Design and Demonstration of Broadband Grating Couplers. IEEE Jour...

work page 2023

[12] [12]

(57) Ruiz, J. L. P.; Gougeon, S.; Youssef, M.; Ung, B.; LaRochelle, S.; Ménard, M. Inverse-Designed Edge Couplers for Multimode Silicon Nitride Photonics. In Optical Fiber Communication Conference, 2026; Optica Publishing Group: p Tu2J

work page 2026