REVIEW 3 major objections 5 minor 24 references
Zero-change foundry compatible silicon photonics MEMS optical switch
T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read Zero-change foundry MEMS switch reaches 30 dB extinction at 20 nW
desk verdict A credible zero-change foundry MEMS switch, but the process-maturity claim rests on one release and uncalibrated traces. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central mechanism is a vertical adiabatic coupler whose coupling is set by the vertical gap between two waveguides; one waveguide sits on a released SOI cantilever and the other sits in the fixed slab. Applying a voltage between the BEOL metal 2 layer and the doped cantilever pulls the cantilever upward, increasing the gap and suppressing the optical coupling. The timed wet release, removing roughly 500 nm of passivation, coating with $\sim$25 nm of alumina, etching in a buffered oxide etchant for about 155 minutes, then supercritical drying, is what converts a standard foundry die into a working MEMS device. Squeeze-film air damping, modeled as $C = \mu L w^3 / h^3$, sets the observed $
What would settle it
Run the same release process on a die with reference waveguides that have no movable cantilever; if their insertion loss changes by more than about 0.1 dB after release, or if a device whose cantilever is mechanically blocked still shows 30 dB extinction, the high extinction is not caused by the designed MEMS deflection.
Extended reading notes
Core claim
The paper's core demonstration is an electrostatic MEMS switch element whose top electrode is the foundry's back-end-of-line (BEOL) metal layer and whose movable ground is a doped silicon-on-insulator (SOI) cantilever. Etch holes let a timed wet release remove the inter-metal dielectric and the buried oxide so the cantilever can deflect upward by roughly 700 nm at 35 V. The cantilever carries part of an adiabatic coupler; upward motion decouples the waveguides and sends the input to the through port instead of the output port. In the C band the authors measure over 30 dB extinction between ON and OFF states at about 32 V, insertion loss below 1.5 dB (including about 0.6 dB of propagation los
Load-bearing premise
The result stands or falls on whether the timed wet release reliably frees the cantilever without stiction, damaged anchors, or added waveguide loss, so that the measured extinction and insertion loss really come from the intended MEMS motion.
Editorial extensions
If this is right
- MEMS switch elements can be co-fabricated with standard silicon photonic components on the same foundry wafer, enabling switch arrays without a custom process stack.
- Because each path crosses only one switch element in a cross-bar layout, the cumulative insertion loss does not grow with radix the way cascaded Mach-Zehnder meshes do.
- The broadband C/L-band response supports fat-pipe switching of many wavelengths at once, rather than wavelength-selective routing.
- At under 20 nW static power per element, large arrays become feasible from a power-budget standpoint, though the roughly 32 V actuation still needs driver electronics.
- Squeeze-film damping can be engineered out by perforation and cantilever aspect ratio, or by vacuum sealing, pointing toward microsecond-scale switching in later designs.
Reading between the lines
- The paper leaves implicit that the same BEOL-metal actuation could be repurposed for tunable phase shifters or attenuators; the switch element's analog transmission curve already gives controllable attenuation over a wide range.
- A normally-OFF variant that harnesses built-in metal stress rather than fighting it would make circuit testing and control easier; the paper mentions this direction but does not demonstrate it.
- One testable extension is to carry the same release recipe across multiple foundry runs with different passivation and oxide thicknesses; the timed-etch margin is the main variable to re-tune.
- With vacuum sealing, the squeeze-film damping that limits the current roughly 600 microsecond response should drop away, leaving the near-300 kHz mechanical resonance as the speed limit.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports a silicon-photonics MEMS optical switch unit cell fabricated through an AIM Photonics MPW run with no changes to the foundry front-end, followed by a BEOL post-processing release (timed wet etch in ALPAD 639 and supercritical drying). The central claims are: more than 30 dB extinction ratio at about 32 V, insertion loss below 1.5 dB (including about 0.6 dB propagation loss), broadband operation across the C/L band, static power consumption below 20 nW, and switching dynamics around 600 µs attributed to squeeze-film damping. The authors also discuss a modified cross-bar architecture for scaling and propose shaped-waveform driving to accelerate switching.
Significance. If the headline numbers are reproducible, this is a valuable step toward integrating MEMS actuation into a standard commercial silicon-photonics flow: the switch is broadband, low-loss, high-extinction, and consumes nanowatts of static power. The strengths of the paper are that the main optical and power claims appear to be direct measurements rather than fitted extractions, and the design is simple enough to be assessed from the text and figures. The main significance, however, depends on process maturity and measurement reliability: the device count, calibration baseline, and release repeatability are not documented, so the current manuscript supports a proof-of-concept rather than a robust process claim.
major comments (3)
- [§II, Figs. 3(a), 3(b), and Fig. 4(a)] The headline extinction ratio (>30 dB) and insertion loss (<1.5 dB) are presented as single representative transmission curves with no device count, no error bars, and no reference-waveguide calibration. The statement that the insertion loss 'also includes ~0.6 dB of propagation loss' does not quantify grating-coupler loss or any release-induced change in waveguide loss. Without a straight-waveguide or unreleased control on the same die, the absolute insertion loss cannot be attributed to the switch element rather than to the measurement baseline. In addition, Fig. 4(a) shows 100-cycle repeatability on the same released device, not release-to-release repeatability. Please add reference-normalized spectra, report the number of devices tested, and provide mean/standard deviation for extinction, insertion loss, and control-waveguide loss.
- [§II, release-process paragraph] The 'zero-change foundry-compatible process' claim is at bottom a process-maturity claim, but the evidence is a single timed wet release described as 'estimated to be around 155 minutes.' The SEM image and the 100-cycle electrical repeatability do not establish that the release recipe reliably frees cantilevers without stiction, anchor damage, or etch-induced passive-loss change. Please report the number of released dies/devices, the yield of functional switches, observed failure modes, and any wafer-to-wafer or run-to-run variation. This is load-bearing for the central claim because the reported extinction and insertion loss are measured on one released structure.
- [§II, Fig. 4(c)] The static power claim (<20 nW) is one of the headline results, but the measurement procedure is not described. Fig. 4(c) shows a power curve versus voltage, yet there is no explanation of how power was derived (e.g., voltage-current product at the actuation electrode), what leakage path is being measured, or what the measurement precision is at the nanowatt level. Please specify the measurement setup, the instrumentation accuracy, and the number of measurements; otherwise the 20 nW figure cannot be independently assessed.
minor comments (5)
- [General / supplementary references] The text repeatedly refers to Supplementary Sections 1, 2, 3, 5, 6, and 8, but no supplementary material is included with the manuscript. Verify that these references are available to reviewers or state the relevant details in the main text.
- [Abstract and Fig. 3(b)] The abstract says operation is in the C-band, while the text and Fig. 3(b) show broadband switching across the C and L bands (1500–1650 nm). Harmonize the terminology.
- [§II, damping discussion] The analytical squeeze-film damping expression is not numbered and the symbols (µ, ρ, ωm, W, h) are defined only partially in the running text. Add an equation number and a consistent definition table so the 'agrees well with the fit' statement can be checked.
- [Fig. 3(c) / Fig. 3(b)] The simulated insertion loss is stated as <0.25 dB per adiabatic coupler, while the measured switch loss includes a pair of couplers plus propagation/bend/transition losses. A short sentence reconciling these numbers (e.g., the expected total from passive components) would improve clarity.
- [Fig. 4(b)] The pull-in phenomenon is shown as a ramp up/down curve, but the figure does not label the ramp directions in the plot itself. Add labels or a caption description of the hysteresis/destruction cycle.
Circularity Check
No significant circularity: headline optical, power, and repeatability claims are direct measurements; the only model-data comparison (squeezed-film damping) is an ancillary parameter cross-check, not a derived prediction.
full rationale
The paper's central claims (more than 30 dB extinction ratio, insertion loss <1.5 dB, static power ~20 nW, 100-cycle repeatability) are reported as direct measurements of the fabricated device, not as outputs of a fitted model. The optical and mechanical simulations are independent design calculations, not fits to the measured transmission curves. The only passage involving a fitted parameter is in the switching-dynamics section: 'The analytically derived damping coefficient agrees well with the fit to the experimental data.' This is a parameter cross-check between a physics-based formula and an extracted value; it is not a prediction derived from the same data, and it does not feed into the headline extinction/loss/power claims. Citations to prior work by the same group (e.g., [11,14,16]) provide background and design antecedents but are not used to forbid alternatives or to justify the present measured results. The absence of reference-waveguide calibration and the single-device traces are experimental-robustness concerns, not circularity. No load-bearing step reduces to its own input by construction.
Assumptions & free parameters
free parameters (1)
- squeezed-film damping coefficient / normalized damping ratio =
not listed numerically; matched to measured ~600 us transient
assumptions (4)
- domain assumption The AIM Photonics BEOL stack (Metal 1, Metal 2, vias, passivation) can serve as the fixed electrode and the doped SOI cantilever as the movable electrode with a nominal ~3 um gap.
- domain assumption The timed wet release removes BOX and inter-metal dielectric uniformly without stiction, anchor damage, or etch-induced optical loss.
- domain assumption The measured optical transmission change is caused by the designed adiabatic-coupler decoupling, not by release-induced loss changes or measurement drift.
- domain assumption The squeezed-film damping model C = mu L w^3 / h^3 from Pandey and Pratap [22] applies to the released cantilever in air.
Cite this review
Pith. "Pith review of Zero-change foundry compatible silicon photonics MEMS optical switch." pith.science (2026). https://pith.science/paper/OHNAAYKB
@misc{pith2026260803146,
author = {Pith},
title = {Pith review of: Zero-change foundry compatible silicon photonics MEMS optical switch},
year = {2026},
howpublished = {\url{https://pith.science/paper/OHNAAYKB}},
note = {Machine review of arXiv:2608.03146}
}
read the original abstract
Large-scale photonic switches are emerging as essential devices for energy-efficient optical interconnect in data centers and AI/ML clusters as a key enabler for high-bandwidth and low-latency connectivity. Combining micro-electro-mechanical (MEMS) based mechanical reconfigurability with silicon photonic integrated circuits can enable a large-scale, low-loss, programmable platform required for large-scale optical circuit switches. We demonstrate a broadband silicon photonics MEMS switch with more than 30 dB extinction ratio operating in C-band using a zero-change foundry-compatible process and Back-end-of-Line (BEOL) post-processing. The optical switch element exhibits an insertion loss of less than 1.5 dB with a low static power consumption of approx 20 nW at maximum actuation voltage. Our results illustrate that MEMS-based silicon photonics modulators and phase shifters can be used alongside standard silicon photonics components seamlessly in scenarios where performance in terms of footprint, extinction ratio, broad bandwidth, and low-loss operation is of paramount importance.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
Weiyang Wang, Moein Khazraee, Zhizhen Zhong, Manya Ghobadi, Zhihao Jia, Dheevatsa Mudigere, Ying Zhang, and Anthony Kewitsch, “{TopoOpt}: Co-optimizing network topology and parallelization strategy for distributed training jobs,” in20th USENIX Symposium on Networked Systems Design and Implementation (NSDI 23)(2023) pp. 739–767
work page 2023
-
[2]
Norm Jouppi, George Kurian, Sheng Li, Peter Ma, Rahul Nagarajan, Lifeng Nai, Nishant Patil, Suvinay Subramanian, Andy Swing, Brian Towles,et al., “Tpu v4: An optically reconfigurable supercomputer for machine learning with hardware support for embeddings,” inProceedings of the 50th annual international symposium on computer architecture(2023) pp. 1–14
work page 2023
-
[3]
Mission apollo: Landing optical circuit switching at datacenter scale,
Ryohei Urata, Hong Liu, Kevin Yasumura, Erji Mao, Jill Berger, Xiang Zhou, Cedric Lam, Roy Bannon, Darren Hutchinson, Daniel Nelson,et al., “Mission apollo: Landing optical circuit switching at datacenter scale,” arXiv preprint arXiv:2208.10041 (2022)
arXiv 2022
-
[4]
1100 x 1100 port mems-based optical crossconnect with 4-db maximum loss,
J Kim, CJ Nuzman, B Kumar, DF Lieuwen, JS Kraus, A Weiss, CP Lichtenwalner, AR Papazian, RE Frahm, NR Basavanhally,et al., “1100 x 1100 port mems-based optical crossconnect with 4-db maximum loss,” IEEE Photonics Technology Letters15, 1537–1539 (2003)
work page 2003
-
[5]
32×32 silicon electro-optic switch with built-in monitors and balanced- status units,
Lei Qiao, Weijie Tang, and Tao Chu, “32×32 silicon electro-optic switch with built-in monitors and balanced- status units,” Scientific Reports7, 42306 (2017)
work page 2017
-
[6]
Nicolas Dupuis, Benjamin G Lee, Alexander V Rylyakov, Daniel M Kuchta, Christian W Baks, Jason S Orcutt, Douglas M Gill, William MJ Green, and Clint L Schow, “Design and fabrication of low-insertion-loss and low-crosstalk broadband mach–zehnder silicon photonic switches,” Journal of Lightwave Technology33, 3597– 3606 (2015)
work page 2015
-
[7]
32×32 silicon photonic switch,
Dritan Celo, Dominic J Goodwill, Jia Jiang, Patrick Dumais, Chunshu Zhang, Fei Zhao, Xin Tu, Chunhui Zhang, Shengyong Yan, Jifang He,et al., “32×32 silicon photonic switch,” in2016 21st OptoElectronics and Communications Conference (OECC) held jointly with 2016 International Conference on Photonics in Switching (PS)(IEEE, 2016) pp. 1–3
work page 2016
-
[8]
Large-scale silicon photonics switch based on 45-nm cmos technology,
Kazuhiro Ikeda, Keijiro Suzuki, Ryotaro Konoike, Shu Namiki, and Hitoshi Kawashima, “Large-scale silicon photonics switch based on 45-nm cmos technology,” Optics Communications466, 125677 (2020)
work page 2020
Show all 24 references
-
[9]
Integrated silicon photonic mems,
Niels Quack, Alain Yuji Takabayashi, Hamed Sattari, Pierre Edinger, Gaehun Jo, Simon J Bleiker, Carlos Errando-Herranz, Kristinn B Gylfason, Frank Niklaus, Umar Khan,et al., “Integrated silicon photonic mems,” Microsystems & Nanoengineering9, 27 (2023)
2023
-
[10]
Mems for photonic integrated circuits,
Carlos Errando-Herranz, Alain Yuji Takabayashi, Pierre Edinger, Hamed Sattari, Kristinn B Gylfason, and Niels Quack, “Mems for photonic integrated circuits,” IEEE Journal of Selected Topics in Quantum Electronics26, 1–16 (2019)
2019
-
[11]
Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers,
Tae Joon Seok, Niels Quack, Sangyoon Han, Richard S Muller, and Ming C Wu, “Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers,” Optica3, 64–70 (2016)
2016
-
[12]
Programmable photonic arrays based on microelectromechanical elements with femtowatt-level standby power consumption,
Dong Uk Kim, Young Jae Park, Do Yun Kim, Youngjae Jeong, Min Gi Lim, Myung Seok Hong, Man Jae Her, Yoonhyuk Rah, Dong Ju Choi, Sangyoon Han,et al., “Programmable photonic arrays based on microelectromechanical elements with femtowatt-level standby power consumption,” Nature Ph...
2023
-
[13]
Silicon photonic microelectromechanical switch using lateral adiabatic waveguide couplers,
Takumi Nagai and Kazuhiro Hane, “Silicon photonic microelectromechanical switch using lateral adiabatic waveguide couplers,” Optics Express26, 33906–33917 (2018)
2018
-
[14]
Wafer-scale silicon photonic switches beyond die size limit,
Tae Joon Seok, Kyungmok Kwon, Johannes Henriksson, Jianheng Luo, and Ming C Wu, “Wafer-scale silicon photonic switches beyond die size limit,” Optica6, 490– 494 (2019)
2019
-
[15]
Large-scale silicon photonic mems switch with flip-chip integrated cmos drivers,
“Large-scale silicon photonic mems switch with flip-chip integrated cmos drivers,” inIET Conference Proceedings CP839, Vol. 2023 (IET, 2023) pp. 1390–1393
2023
-
[16]
32×32 silicon photonic mems switch with gap- adjustable directional couplers fabricated in commercial cmos foundry,
Sangyoon Han, Jeremy Beguelin, Lane Ochikubo, John Jacobs, Tae Joon Seok, Kyoungsik Yu, Niels Quack, Chang-Kyu Kim, Richard S Muller, and Ming C Wu, “32×32 silicon photonic mems switch with gap- adjustable directional couplers fabricated in commercial cmos foundry,” journal of...
2021
-
[17]
Integrated photonic mems switch for visible light,
Alperen Govdeli, Hong Chen, Saeed S Azadeh, John N Straguzzi, Hongyao Chua, Guo-Qiang Lo, Joyce KS Poon, and Wesley D Sacher, “Integrated photonic mems switch for visible light,” Optics Express33, 650–664 (2025)
2025
-
[18]
Broadband compact single-pole double-throw silicon photonic mems switch,
Alain Yuji Takabayashi, Hamed Sattari, Pierre Edinger, Peter Verheyen, Kristinn B Gylfason, Wim Bogaerts, and Niels Quack, “Broadband compact single-pole double-throw silicon photonic mems switch,” Journal of Microelectromechanical Systems30, 322–329 (2021)
2021
-
[19]
Silicon photonic microelectromechanical phase shifters for scalable programmable photonics,
Pierre Edinger, Alain Yuji Takabayashi, Carlos Errando- Herranz, Umar Khan, Hamed Sattari, Peter Verheyen, Wim Bogaerts, Niels Quack, and Kristinn B Gylfason, “Silicon photonic microelectromechanical phase shifters for scalable programmable photonics,” Optics Letters46, 5671–5...
2021
-
[20]
Vacuum-sealed silicon photonic mems tunable ring resonator with an independent control over coupling and phase,
Pierre Edinger, Gaehun Jo, Chris Phong Van Nguyen, Alain Yuji Takabayashi, Carlos Errando-Herranz, Cleitus Antony, Giuseppe Talli, Peter Verheyen, Umar Khan, Simon J Bleiker,et al., “Vacuum-sealed silicon photonic mems tunable ring resonator with an independent control over co...
2023
-
[21]
Parallel convolutional processing using an integrated photonic tensor core,
Johannes Feldmann, Nathan Youngblood, Maxim Karpov, Helge Gehring, Xuan Li, Maik Stappers, 6 Manuel Le Gallo, Xin Fu, Anton Lukashchuk, Arslan S Raja,et al., “Parallel convolutional processing using an integrated photonic tensor core,” Nature589, 52–58 (2021)
2021
-
[22]
Effect of flexural modes on squeeze film damping in mems cantilever resonators,
Ashok Kumar Pandey and Rudra Pratap, “Effect of flexural modes on squeeze film damping in mems cantilever resonators,” Journal of Micromechanics and Microengineering17, 2475 (2007)
2007
-
[23]
Wafer-level hermetically sealed silicon photonic mems,
Gaehun Jo, Pierre Edinger, Simon J Bleiker, Xiaojing Wang, Alain Yuji Takabayashi, Hamed Sattari, Niels Quack, Moises Jezzini, Jun Su Lee, Peter Verheyen, et al., “Wafer-level hermetically sealed silicon photonic mems,” Photonics Research10, A14–A21 (2022)
2022
-
[24]
Preshaping command inputs to reduce system vibration,
Neil C Singer and Warren P Seering, “Preshaping command inputs to reduce system vibration,” Journal of dynamic systems, measurement, and control112, 76–82 (1990)
1990
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.