{"id":"3fa78bac-193b-40d8-95af-1a00a86094a9","arxiv_id":"2412.15157","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Centimetre-scale thin-film lithium niobate coupons were transfer-printed onto silicon nitride, yielding a push-pull modulator with Vπ = 3.2 V and modulation up to 35 GHz.","lead":"Centimetre-long strips of thin-film lithium niobate were stamped onto silicon nitride photonic chips using micro-transfer printing, a scale not previously achieved. The resulting push-pull modulator switches light with a half-wave voltage of 3.2 V and works up to at least 35 GHz, bringing active electro-optic functions closer to standard CMOS-compatible photonics.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 0.9 dB/cm propagation-loss value rests on a cut-back fit whose slope can be strongly biased by coupon-to-coupon transition-loss variation; with ±0.8 dB/cm uncertainty the 'low-loss' claim is not yet secure.","rationale":"I read the paper as claiming a process advance plus a photonic-performance demonstration: centimetre-long LN coupons can be micro-transfer-printed onto SiN, with low propagation loss, low transition loss, and a working push-pull modulator. The printing and modulator parts are supported by direct evidence: coupon images, wavelength sweeps with >31 dB extinction, a V-pi of 3.2 V from an overdrive-based extraction, and a de-embedded S21 response to 35 GHz. The weakest link is the loss extraction. The reader's weakest-assumption analysis correctly identifies the cut-back fit as assuming identical mode profiles and transition losses across all coupons. I agree with that diagnosis and make it more concrete: because every coupon adds two transitions, the fit slope is extremely sensitive to small per-coupon transition-loss variations, and the reported 0.9 +/- 0.8 dB/cm already spans a range that does not distinguish a genuinely low-loss process from a mediocre one. This does not overturn the central claim, but it means the headline loss number should be treated as an order-of-magnitude estimate, not a precise figure. A Monte Carlo sensitivity analysis and the inclusion/exclusion tests described above would settle whether the number is robust. Since the reader already issued a CONDITIONAL verdict, my stress-test does not change that verdict; it reinforces the condition that raw data and uncertainty quantification be provided.","tokens_in":9589,"tokens_out":5886,"duration_ms":53806,"concrete_test":"Run a Monte Carlo sensitivity analysis on the Fig. 5.b cut-back data: use all raw per-structure transmission values rather than the length-averaged points, draw per-facet transition losses from a N(1.8, 0.2) dB distribution, refit the linear slope 10,000 times, and report the distribution of inferred dB/cm. Separately refit after (i) including the excluded damaged coupon and (ii) dropping each coupon length in turn. If the 95% confidence interval of the slope overlaps zero, or if any single-coupon exclusion shifts the slope by more than 0.5 dB/cm, the abstract's 'approximately 0.9 dB/cm' should be replaced by an upper-bound statement such as 'below 2 dB/cm.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim has two parts: centimetre-scale transfer printing is demonstrated directly, but the photonic benefit is quantified by the propagation loss. In Section 3, Fig. 5.b, the loss is extracted from transmission through SiN waveguides covered by printed LN coupons of different lengths. This cut-back analysis is valid only if the sole variable between structures is the LN length. That requires identical hybrid mode profiles, identical SiN waveguides and grating couplers, and identical transition loss at every LN-SiN interface. Coupon-to-coupon variations in the LN-SiN gap, film thickness, or facet quality change the mode overlap and hence the per-facet transition loss. Each coupon contributes two transitions, so a transition-loss scatter of ±0.2 dB/facet (the stated uncertainty) injects ±0.4 dB into every structure's excess loss; for the 1 mm coupon this is roughly four times the expected 0.09 dB of LN propagation loss. The slope of the linear fit is therefore highly sensitive to which coupons are included. The paper already excludes one damaged structure, and the reported uncertainty interval (0.1-1.7 dB/cm) does not rule out a propagation loss near 2 dB/cm. The transfer-printing demonstration stands, but the 'preserving low optical loss' part of the central claim is not established at the precision implied by 'approximately 0.9 dB/cm.'","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the micro-transfer printing of centimetre-long (up to 1 cm) thin-film lithium niobate coupons onto a silicon nitride photonic platform. The authors characterize the optical losses using cut-back and coupon-chain test structures, reporting a propagation loss of approximately 0.9 dB/cm and a transition loss of 1.8 ± 0.2 dB per facet. They then integrate 1-cm-long coupons into a push-pull Mach-Zehnder modulator, extracting a Vπ of 3.2 V (VπL = 3.2 V.cm) from low-speed measurements and observing electro-optic response up to 35 GHz. The central claim is that micro-transfer printing can now place centimetre-scale TFLN on SiN without sacrificing the electro-optic performance of LN, enabling efficient modulators and nonlinear devices on CMOS-compatible platforms.","tokens_in":9783,"tokens_out":4717,"duration_ms":34074,"significance":"If the reported results are secure, this work represents a meaningful advance in heterogeneous integration, extending micro-transfer printed LN devices from millimetre to centimetre scale and demonstrating a VπL comparable to wafer-bonded LN-on-SiN modulators. The use of resist tethers and the absence of pre-processing on the target wafer are practical advantages that improve the CMOS-compatibility of the approach. The paper's strengths include a direct demonstration of printing 1-cm-long coupons, the integration of these coupons into functional modulator structures, and the explicit measurement of high-speed electro-optic modulation. However, the loss characterization and the Vπ extraction have uncertainties that need to be addressed before the quantitative claims are fully established.","major_comments":[{"comment":"The cut-back extraction of 0.9 ± 0.8 dB/cm propagation loss rests on the assumption that the only variable between structures is the LN length. The paper itself reports a transition-loss scatter of ±0.2 dB per facet (Fig. 5.a) and excludes one damaged structure from the fit. Because each coupon contributes two transitions, a ±0.2 dB/facet variation injects ±0.4 dB per structure, which is roughly four times the expected 0.09 dB propagation loss for the 1-mm coupon. The slope of the linear fit is therefore highly sensitive to coupon-to-coupon variations in facet quality, gap, or film thickness, and the reported interval (0.1–1.7 dB/cm) does not securely support the 'low-loss' claim. Please provide the per-structure data and a quantitative sensitivity analysis of the fit to transition-loss scatter.","section":"Section 3, Fig. 5.b"},{"comment":"The half-wave voltage Vπ = 3.2 V is extracted from a single 100-Hz measurement with no repeated traces, no error bar, and no explicit algorithm for reading Vπ from the normalized transmission-voltage curve. The agreement with a simulated VπL of 3.2 V.cm is presented as confirmation, but that simulation uses an overlap factor Γ = 0.42 as an input; thus the comparison is a consistency check, not an independent validation. Please report the measurement uncertainty and details of the extraction method.","section":"Section 4, Fig. 7.d"},{"comment":"The claim of modulation up to at least 35 GHz depends on a de-embedding procedure that is not described in the text; the reader is only told that a commercial modulator (Fujitsu FTM7937EZ) replaces the DUT and probes to extract the setup contribution. Without the reference modulator's frequency response and the de-embedding equation, it is not possible to assess how much of the measured response is due to the DUT. Please provide the de-embedding details and the reference calibration data.","section":"Section 4, Fig. 8.b"}],"minor_comments":[{"comment":"The second and third process steps are both labeled '(b)' in the figure; please renumber the panels consistently.","section":"Section 2, Fig. 1"},{"comment":"The determination of the 16.9 dB reference loss at 1550 nm from the no-coupon chain is not described; please state how this value was obtained from the wavelength sweeps.","section":"Section 3"},{"comment":"The discussion states 'up to 1 dB/cm' additional loss, while the abstract and conclusion report 'around 0.9 dB/cm' and the fitted value is 0.9 ± 0.8 dB/cm; please make these statements consistent and quote the uncertainty.","section":"Section 5"},{"comment":"Reference [26] is listed with '(n.d.)'; please complete the citation with full bibliographic information.","section":"Table 1, Ref. [26]"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid experimental demonstration of centimetre-scale micro-transfer printing of TFLN on SiN, and the modulator results are promising. However, the loss extraction in Section 3 is the main quantitative pillar of the 'low-loss' claim, and it is currently too fragile because of the large per-facet transition-loss uncertainty relative to the short-coupon propagation loss. The Vπ measurement is also preliminary. A revision that adds per-coupon data, a sensitivity analysis, and clearer measurement methodology would substantially strengthen the manuscript. The scope fits well with the journal; the novelty with respect to prior µTP LN work is the centimetre length scale and the resulting VπL value."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nRead the Niels et al. paper on centimetre-scale micro-transfer printing of thin-film LN on SiN. The headline is that they print 1 cm long LN coupons and get a push-pull modulator with Vπ = 3.2 V, the lowest reported for transfer-printed LN and consistent with a simulated VπL of 3.2 V.cm. That part is real and extends prior millimetre-scale work (refs 19,25,26) in a meaningful way. The resist-tether process without pillars is also a sensible step toward fab adoption.\n\nWhat to watch before forming a view: the 'around 0.9 dB/cm' propagation loss is not as solid as the abstract implies. The cut-back fit in Fig. 5b has five coupon lengths, one damaged sample excluded, and assumes identical transition loss at every LN-SiN interface. The paper reports transition loss of 1.8 ± 0.2 dB/facet, so each coupon carries a ±0.4 dB uncertainty from its two facets. For the 1 mm coupon, that is about four times the expected 0.09 dB of LN propagation loss, so the slope of the fit is highly sensitive to which coupons are included and to small variations in the printed interface. The uncertainty interval 0.1-1.7 dB/cm shows the value is not tightly constrained. The authors themselves qualify the loss in the discussion, attributing it to surface roughness from encapsulation removal.\n\nThe modulator data are cleaner but still single-device: Vπ is read from one low-speed waveform without an error bar, and the 35 GHz response is de-embedded with a commercial modulator reference. That is acceptable for a proof of concept, not for a metrology claim. The agreement with simulation uses Γ = 0.42 from a mode simulation, so it is a consistency check rather than a fitted prediction. Mild circularity, not a fatal one.\n\nThe paper is honest: it tables prior transfer-printed LN work, states that the design is not optimised, and makes the data available on request. The citation pattern is fine, with ref [26] giving the 6.5 mm coupon context.\n\nThis is a solid process demonstration with a promising modulator result. The loss number needs more support before being quoted as 0.9 dB/cm. I would send it to peer review and ask the authors to either release the cut-back data, add error bars on the slope, or use a loss measurement that does not hinge on identical transition losses (e.g., resonator Q or multiple coupons per length).\n\nYours,","headline":"A genuine centimetre-scale transfer-printing advance and a working 3.2 V modulator, but the 0.9 dB/cm loss claim rests on a fragile cut-back fit.","tokens_in":10472,"tokens_out":3178,"would_cite":true,"duration_ms":26098,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["42.82.-m","42.79.Hp"],"model":"deepseek-v4-flash","headline":"Micro-transfer printing can now place centimetre-long thin-film lithium niobate onto silicon nitride, preserving 0.9 dB/cm loss and giving a 3.2 V modulator at 35 GHz.","keywords":["micro-transfer printing","thin-film lithium niobate","silicon nitride","heterogeneous integration","Mach-Zehnder modulator","electro-optic modulation","propagation loss","CMOS-compatible photonics"],"falsifier":"Measure the propagation loss using a method that does not assume identical coupling per coupon, for example by comparing the loaded quality factors of ring resonators that share the same printed LN coupling section but have different resonator lengths, or by repeating the cut-back with multiple fresh coupons of each length; if the slope changes beyond the stated ±0.8 dB/cm uncertainty or the 1 cm single coupon deviates from the chain of 1 mm coupons, the hybrid-mode-overlap assumption fails.","tokens_in":128,"feed_emoji":"⚡","tokens_out":6315,"duration_ms":95647,"temperature":0.7,"pith_summary":"This paper claims that micro-transfer printing can pick up and place centimetre-long slabs of thin-film lithium niobate onto a silicon nitride photonic chip without wrecking the material's optical or electro-optic properties. The authors demonstrate propagation loss around 0.9 dB/cm and coupling loss of 1.8 dB per facet for printed coupons, and they build a 1 cm push-pull Mach-Zehnder modulator with Vπ = 3.2 V and a flat response up to at least 35 GHz. If correct, this removes the length bottleneck that previously limited transfer-printed lithium niobate to millimetre-scale devices, opening a path to low-voltage modulators and efficient nonlinear devices on mature CMOS-compatible platforms.","feed_headline":"Centimetre-scale lithium niobate printed onto silicon photonics","feed_subtitle":"A 1 cm push-pull modulator on SiN hits Vπ of 3.2 V and works to 35 GHz, with 0.9 dB/cm loss.","key_machinery":"The enabling mechanism is micro-transfer printing with resist tethers: LN coupons are suspended by photoresist over an undercut oxide, picked up with a PDMS stamp retracted at 500 mm/s, and released on the target at 4 µm/s, requiring no recess etching or pillars on the target wafer. The photonic design uses a hybrid mode in which the SiN waveguide guides 32% of the light and the overlying LN slab guides 62%, so the LN contributes electro-optic tuning without needing etched LN waveguides.","core_discovery":"The paper reports that micro-transfer printing can pick up and place thin-film lithium niobate coupons up to 1 cm long onto a silicon nitride chip, and that these printed coupons retain useful optical and electro-optic performance: propagation loss about 0.9 ± 0.8 dB/cm at 1550 nm, transition loss 1.8 ± 0.2 dB per facet, and, in a 1 cm push-pull Mach-Zehnder modulator, Vπ = 3.2 V (VπL = 3.2 V.cm) with a flat electro-optic response up to at least 35 GHz. The authors state that this is the lowest Vπ yet reported for a micro-transfer-printed modulator and that the 35 GHz response shows the printing does not degrade the high-speed properties of LN.","pith_inferences":["If the loss can be lowered by improving the encapsulation-removal step, printed cm-scale LN could approach the loss and voltage figures of monolithic LNOI while riding on mature SiN foundry platforms.","Because the length bottleneck is gone, the same printing process should make high-efficiency periodically poled LN frequency converters on SiN, where conversion efficiency scales with the square of the interaction length.","A natural testable extension is to shrink the electrode gap or increase the LN fraction of the hybrid mode; the paper's own simulation method predicts a correspondingly smaller VπL.","The coupon-chain approach also hints that multiple different active materials could be co-printed on one SiN target, but that goes beyond what this work demonstrates."],"forward_implications":["Centimetre-long printed LN makes low-voltage modulators and long-interaction nonlinear devices on SiN practical without wafer bonding.","The measured VπL of 3.2 V.cm puts micro-transfer printing on par with die-to-wafer bonding (2.9 V.cm) for LN-on-SiN modulators.","The flat electro-optic response up to 35 GHz means the printed interface itself does not set a speed limit for data-communication use.","Using resist tethers rather than LN tethers and avoiding recess etching makes the process more compatible with standard semiconductor fabs.","Printing coupons in opposite poling directions enables push-pull configurations and opens electrode designs such as GSSG for differential drive."],"supporting_citations":[{"why":"Supplies the micro-transfer printing method and its advantages, which the entire fabrication approach builds on.","marker":"[18]"},{"why":"Previous high-speed LN-on-SiN modulator by micro-transfer printing with Vπ = 14.8 V, the main benchmark extended to centimetre length here.","marker":"[25]"},{"why":"Closest prior micro-transfer-printed LN-on-Si modulator with a 6.5 mm coupon and Vπ = 5 V, which this work surpasses in length.","marker":"[26]"},{"why":"Die-to-wafer bonded LN-on-SiN modulator with VπL = 2.9 V.cm, the performance target for the push-pull design.","marker":"[15]"},{"why":"Provides the simulation method and Γ = 0.42 used to predict the measured VπL of 3.2 V.cm.","marker":"[16]"},{"why":"Developed the reliable micro-transfer printing method for LN coupons that this work adapts to centimetre scale.","marker":"[19]"}],"fun_headline_variants":["Printed cm-scale lithium niobate on SiN: Vπ 3.2V, 35GHz","Micro-transfer printing brings 1cm TFLN to SiN: 0.9dB/cm loss","35GHz electro-optic response from printed lithium niobate modulator","Heterogeneous integration: cm-long TFLN on SiN by micro-transfer printing","1cm printed LN on SiN: low loss, low Vπ, high speed"],"cache_read_input_tokens":12416,"weakest_assumption_plain":"The 0.9 dB/cm loss value rests on the assumption that every printed coupon has the same hybrid mode profile, so only the coupon length varies between cut-back structures; if the LN-to-SiN gap, film thickness, or facet quality differs from coupon to coupon, the linear fit is biased.","fun_headline_variants_meta":{"raw":{"variants":["Printed cm-scale lithium niobate on SiN: Vπ 3.2V, 35GHz","Micro-transfer printing brings 1cm TFLN to SiN: 0.9dB/cm loss","35GHz electro-optic response from printed lithium niobate modulator","Heterogeneous integration: cm-long TFLN on SiN by micro-transfer printing","1cm printed LN on SiN: low loss, low Vπ, high speed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00072,"raw_usage":{"total_tokens":3211,"prompt_tokens":902,"completion_tokens":2309,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":2193}},"tokens_in":518,"tokens_out":2309,"duration_ms":15946,"temperature":1.0,"reasoning_tokens":2193,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:34:14.639708+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the propagation loss using a method that does not assume identical coupling per coupon, for example by comparing the loaded quality factors of ring resonators that share the same printed LN coupling section but have different resonator lengths, or by repeating the cut-back with multiple fresh coupons of each length; if the slope changes beyond the stated ±0.8 dB/cm uncertainty or the 1 cm single coupon deviates from the chain of 1 mm coupons, the hybrid-mode-overlap assumption fails.","supporting_citations":[{"cited_title":"Present and future of micro -transfer printing for heterogeneous photonic integrated circuits,","cited_arxiv_id":null,"evidence_quote":"Supplies the micro-transfer printing method and its advantages, which the entire fabrication approach builds on."},{"cited_title":"Heterogeneous integration of a high -speed lithium niobate modulator on silicon nitride using micro -transfer printing,","cited_arxiv_id":null,"evidence_quote":"Previous high-speed LN-on-SiN modulator by micro-transfer printing with Vπ = 14.8 V, the main benchmark extended to centimetre length here."},{"cited_title":"Micro- transfer printed thin-film lithium niobate modulator for heterogeneous Si photonic platform,","cited_arxiv_id":null,"evidence_quote":"Closest prior micro-transfer-printed LN-on-Si modulator with a 6.5 mm coupon and Vπ = 5 V, which this work surpasses in length."},{"cited_title":"High - Performance Electro-Optic Modulator on Silicon Nitride Platform with Heterogeneous Integration of Lithium Niobate,","cited_arxiv_id":null,"evidence_quote":"Die-to-wafer bonded LN-on-SiN modulator with VπL = 2.9 V.cm, the performance target for the push-pull design."},{"cited_title":"A heterogeneously integrated silicon photonic/lithium niobate travelling wave electro-optic modulator,","cited_arxiv_id":null,"evidence_quote":"Provides the simulation method and Γ = 0.42 used to predict the measured VπL of 3.2 V.cm."},{"cited_title":"Reliable micro -transfer printing method for heterogeneous integration of lithium niobate and semiconductor thin films,","cited_arxiv_id":null,"evidence_quote":"Developed the reliable micro-transfer printing method for LN coupons that this work adapts to centimetre scale."}],"review_version":1}