{"id":"ff77e27e-06f1-4992-8df8-62d5d4ab73ee","arxiv_id":"2508.20431","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"First demonstration of electrical impedance tomography with a fully integrated on-chip picoliter microfluidic chamber and 16-electrode array in 65 nm CMOS, validated against optical micrographs of glass microspheres.","lead":"A single CMOS chip now contains both the fluid chamber and the electronics needed to image it: a 225-picoliter microfluidic cavity with 16 electrodes, fabricated by etching channels into the chip's own wiring layers. It is a proof of concept for lab-on-chip impedance imaging, where small liquid samples can be processed and measured on the same die.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No empty-chamber baseline or electrode calibration supports Fig. 6; EIDORS could be reconstructing per-electrode mismatch as conductivity.","rationale":"The Reader's weakest assumption identifies the same load-bearing premise: the measured boundary voltages are treated as encoding the bulk conductivity distribution, with electrode non-idealities and parasitic effects negligible or uniform. My read agrees and sharpens it: the paper has no control experiment that would distinguish a genuine conductivity-contrast signal from per-electrode mismatch or inverse-solver artifacts. The SNR figures in Fig. 7 are raw voltage-domain statistics, not image-domain validation, so they do not establish spatial fidelity. The paper's own statement in Sec. IV that 'inaccuracies are largely attributed to the ill-posed nature of the EIT problem' is a limitation that would be quantified by an error metric, but none is provided. This concern is not about internal inconsistency; the fabrication and circuit description are plausible, and the EIDORS-based pipeline is standard. It is about the sufficiency of evidence for the central demonstration. Since the Reader already returned CONDITIONAL and my concern is the same condition with a concrete missing control, no verdict change is needed; the suggested water-only baseline test would convert the condition into either acceptance or rejection of the Fig. 6 interpretation.","tokens_in":3750,"tokens_out":3984,"duration_ms":48898,"concrete_test":"Acquire a water-only (no microsphere) dataset using the exact same electrode configuration and EIDORS pipeline, and compute the reconstructed conductivity map. If the baseline map contains localized low-conductivity regions exceeding, say, 3x the max contrast of a forward-simulated homogeneous chamber with the same noise level, then per-electrode or parasitic artifacts are not negligible and the Fig. 6 correlation is inconclusive. Report the same colormap normalization and solver settings as Fig. 6. A null baseline map that is flat (contrast below threshold) would remove the concern.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim depends on the Fig. 6 conductivity maps being governed by the microspheres' bulk conductivity. For that to hold, electrode-electrolyte interface impedance, parasitic capacitances, and etch-induced electrode variability must be negligible or uniform (Sec. IV). The paper reports no electrode impedance characterization, no water-only null reconstruction, no quantitative position/resolution/contrast error, and no EIDORS mesh/electrode-model/regularization settings. The measured SNR (32-52 dB, Fig. 7) shows time variation in raw voltages attributed to 'temperature and fluid flow'; such drift could be common-mode or could map into smooth low-conductivity regions if the inverse solver is uncalibrated. Without a baseline, the apparent low-conductivity spots in Fig. 6 could be artifacts of per-electrode gain/phase mismatch or channel crosstalk, not the spheres. The 'strong correlation' is thus a visual claim on unreported colormap normalization, not a demonstrated EIT reconstruction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a 65 nm CMOS chip in which a ~225 pL microfluidic chamber and 16 electrodes are formed in the interconnect layers by subtractive wet etching, with on-chip LO generation, multiplexing, and voltage readout. Electrical impedance tomography is performed by injecting an LO signal across pairs of electrodes and measuring the resulting voltages at the other electrodes; the data are transferred off-chip and processed with EIDORS. The central claim is that reconstructed conductivity maps for glass microspheres in a water-filled chamber visually track the microsphere positions (Fig. 6), constituting the first demonstration of fully integrated on-chip microfluidics with EIT. The paper also reports an SNR of 32--52 dB (Fig. 7) and shows reconstruction robustness between 1 kHz and 100 kHz (Fig. 8).","tokens_in":3921,"tokens_out":2603,"duration_ms":33645,"significance":"If the central claim is valid, this is a significant proof-of-concept: it would show that a standard CMOS process can host both picoliter-volume sample confinement and the electronics needed for EIT, without post-processing additions such as silicon or plastic microfluidics. The subtractive-etch fabrication route is already published for photonics by the same group (refs. 9--11), so the new contribution is the application to integrated EIT. The paper is commendably explicit about using EIDORS as an external, standard solver and about the system-level components. However, the evidence supporting the central tomography claim is currently visual and lacks the necessary controls, so the significance is conditional on closing that validation gap.","major_comments":[{"comment":"The load-bearing evidence for the central claim is the visual 'strong correlation' between microsphere positions and reconstructed low-conductivity regions. No water-only (empty chamber) baseline reconstruction is reported, so the reader cannot distinguish genuine bulk-conductivity contrast from artifacts caused by per-electrode gain/phase mismatch, channel crosstalk, parasitic capacitances, or nonuniform electrode-electrolyte interface impedance. Because EIDORS solves an ill-posed inverse problem, an uncalibrated reconstruction can map such systematic measurement errors into smooth low-conductivity areas. A baseline reconstruction under identical conditions with only deionized water is essential; without it, Fig. 6 cannot support the claimed correlation.","section":"Sec. IV, Fig. 6"},{"comment":"The forward model used by EIDORS assumes that the measured boundary voltages are governed by the bulk conductivity distribution. This requires the electrode-electrolyte interface impedances and any fabrication-induced per-electrode variability to be negligible or uniform. The paper does not report any electrode impedance spectroscopy, calibration with known-conductivity solutions, or comparison of measured voltages against a simulated homogeneous-chamber model. A simple test would be to fill the chamber with two or three solutions of known conductivity and compare the recorded boundary voltages with EIDORS predictions; this would establish that the measurement chain is sensitive to bulk conductivity rather than to electrode non-idealities.","section":"Sec. IV, electrode characterization"},{"comment":"No quantitative metric is provided for the claimed correlation. The reconstructed maps in Fig. 6 display a colormap whose normalization and range are not specified, and there is no position error, contrast-to-noise ratio, or resolution measure. A quantitative comparison between microsphere centroids from the optical micrographs and the minima of the reconstructed low-conductivity regions (e.g., a table of centroid distances or a correlation coefficient) would make the claim falsifiable and not merely visual. Without such a metric, the statement 'strong correlation' is not supported by the data as presented.","section":"Sec. IV, Fig. 6 and reconstruction parameters"},{"comment":"The EIDORS reconstruction is not reproducible as described. The paper does not state the electrode model (point vs. complete electrode model), the finite-element mesh resolution, the regularization method, the regularization hyperparameter, or the data normalization. These choices materially affect the output image, especially in a severely ill-posed problem with only 16 electrodes. Reporting this information is necessary both for reproducibility and for assessing whether the regularization itself could create spurious low-conductivity regions. This is a major omission for a measurement-only paper whose central result is an image.","section":"Sec. IV, EIDORS setup"}],"minor_comments":[{"comment":"The title contains 'V olume' with an intervening space; this should be corrected to 'Volume'.","section":"Title and Abstract"},{"comment":"Fig. 3 describes test structures with 20 µm × 20 µm chambers and capacities of 4.9 pL and 3.3 pL, while the actual EIT chamber in Sec. III is 150 µm × 150 µm and ~225 pL. Please clarify that Fig. 3 shows separate test structures, not the chamber used for the EIT measurements, to avoid reader confusion.","section":"Sec. II, Fig. 3 caption"},{"comment":"The SNR histograms should have labeled axes and error bars or a statement on the number of samples per bin. The text attributes the histogram shape to 'shifts in local temperature and fluid flow' without evidence; at minimum, show the raw amplitude time series or a spectral analysis to support this attribution.","section":"Sec. IV, Fig. 7"},{"comment":"The colormap scale and normalization are omitted from the reconstructed maps in Fig. 8 as well as Fig. 6. If the display range is not fixed, comparing maps across frequencies is difficult. Please specify whether the color scale is fixed per subfigure or globally.","section":"Sec. IV, Fig. 8"},{"comment":"Ref. [12] is the correct EIDORS reference, but a version number or software release would aid reproducibility. Some reference entries contain double spaces and inconsistent punctuation; a final copyedit would help.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper presents an attractive hardware platform and the fabrication route is credible given the group's prior work. The missing baseline and calibration measurements are not merely cosmetic: they are the difference between 'EIT demonstrated' and 'an uncalibrated inverse solver applied to a CMOS chip.' I view this as repairable within the scope of the manuscript, because the chip and measurement setup still exist and additional measurements can be performed. The editor may wish to remind the authors that a water-only baseline and a simple electrode-impedance check are standard practice for EIT validation and should be prioritized in the revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a genuine first demonstration of EIT with the microfluidic chamber and electrodes co-fabricated in the same CMOS interconnect stack. The subtractive-etch route from the group's photonics work is applied to a new domain, and the paper does not oversell it. The central claim—that reconstructed low-conductivity regions track glass microspheres—is consistent with the data shown. SNR of 32–52 dB and frequency robustness across 1–100 kHz are real evidence that the measurement chain works. I would not accuse the authors of circular reasoning or fitting constants; the inversion is external EIDORS, and the self-citations are the fabrication recipe, which is legitimate.\n\nThe soft spots are real but proportionate. The biggest one is the missing empty-chamber baseline. Without a water-only reconstruction, per-electrode gain/phase mismatch, etch damage, or parasitic capacitance could masquerade as smooth low-conductivity regions. The paper also reports no electrode impedance characterization, no EIDORS mesh or regularization settings, and no quantitative position or resolution error metric. So the \"strong correlation\" in Fig. 6 is currently a visual claim on unreported colormap normalization. This is the load-bearing weakness, and it is a standard control that the authors can plausibly add. The paper acknowledges sources of variation but attributes SNR drift to temperature and fluid flow without testing that attribution; that is a minor issue.\n\nI think the stress-test note is correct to flag the missing baseline, but I would not overstate the risk: the frequency robustness and reasonable SNR make a pure artifact story less likely. Still, the condition for full acceptance is exactly what the reader said—add the water-only control, disclose the EIDORS configuration, and report an error metric.\n\nWho is this for: people working on CMOS lab-on-chip, integrated sensors, or EIT hardware. It is a proof-of-concept, not a solved clinical tool. I would engage with it, and a serious referee in microfluidics or impedance sensing should see it. It deserves revision, not desk rejection.","headline":"A credible first-of-kind integration demo that would be much stronger with a water-only baseline and full EIDORS disclosure.","tokens_in":4510,"tokens_out":1637,"would_cite":true,"duration_ms":21097,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Electrical impedance tomography is demonstrated for the first time with a microfluidic chamber etched directly into a CMOS chip's wiring layers, imaging glass microspheres in 225 picoliters of water.","keywords":["electrical impedance tomography","CMOS","microfluidics","subtractive fabrication","interconnect layers","picoliter chamber","conductivity imaging","impedance image reconstruction"],"falsifier":"Repeat the exact same measurement with the chamber filled only with water and no microspheres. If the same reconstruction software returns localized low-conductivity patches similar to those in Fig. 6, the correlation is an inverse-solver artifact rather than tomography. A second decisive check is to measure the impedance of each electrode in isolation: if the per-electrode spread predicts where low-conductivity regions appear, electrode non-uniformity — not the sample — is what is being imaged.","tokens_in":3578,"feed_emoji":"🔬","tokens_out":8263,"duration_ms":84301,"temperature":0.7,"pith_summary":"Electrical impedance tomography is normally an external scanning technique; this paper tries to shrink it into a single CMOS chip with the sample chamber built into the wiring layers. The authors wet-etch selected metal and barrier layers out of a 65 nm chip's interconnect stack, leaving dielectric-walled chambers with electrodes facing them from multiple metal layers, then fill a ~225 pL chamber with water and glass microspheres and drive and read the 16 electrodes with on-chip circuitry. Reconstructed conductivity maps, computed off-chip with open-source EIT software, place low-conductivity spots where the micrographs show the spheres. If that correlation is genuine, the subtractive etch route already used for on-chip photonics becomes a way to make co-integrated microfluidic-and-electronic chips without bonding on any plastic or silicon fluidics, which would matter for label-free imaging of picoliter biological samples.","feed_headline":"Chip wiring layers become a picoliter impedance imager","feed_subtitle":"A wet etch carves a 225-pL chamber and 16 electrodes from the interconnect metal, then boundary voltages map glass beads in water.","key_machinery":"Subtractive fabrication in the CMOS interconnect stack: alternating wet etches remove chosen copper/aluminum and barrier metals so that inter-metal dielectrics remain as chamber walls, channels, and exposed electrode surfaces, with no added fluidic layer. The electrical engine is a 16-electrode array connected through digitally reconfigurable multiplexers to on-chip stimulation and readout circuits, and the imaging engine is off-chip EIT reconstruction that converts boundary-voltage measurements into a conductivity map.","core_discovery":"The paper asserts the first EIT system whose microfluidic chamber and electronics are fabricated together in one CMOS die. The chamber, roughly 150 µm by 150 µm in lateral extent and about 10 µm deep (≈225 pL), is created by alternating wet etches — Al Etch Type A for copper/aluminum and an EDTA/H2O2 mix for Ta/TaN or Ti/TiN barriers — that hollow out the interconnect metal while leaving inter-metal dielectrics as the chamber walls. A 16-electrode array spans the chamber floor at 66% fill factor, with electrodes also on several metal layers below; on-chip multiplexers reconfigure the electrodes as stimulator, ground, sense output, or floating, and a LO derived from the digital clock drives s","pith_inferences":["The correlation in Fig. 6 is qualitative; a quantitative phantom study with known conductivity targets and an error metric would be the natural next step the paper does not itself supply.","The same etch could carve channels, mixers, or traps connecting the chamber to the chip's photonic and electronic areas, making the paper's closing vision of biology-plus-electronics-plus-photonics on one die testable.","Because the etch is a post-processing step, chip-to-chip uniformity may vary; per-chip electrode impedance calibration could turn this proof-of-concept into a quantitative bioimpedance instrument.","The frequency-dependent measurement capability, though only used to confirm sphere-like behavior here, points toward dielectric-spectroscopy imaging that distinguishes cell types by their impedance signature rather than just locating insulating objects."],"forward_implications":["Single-chip fluidics-and-electronics: the subtractive route builds the sample container from the foundry's own interconnect layers, so no bonding, plastic, or silicon fluidic parts are needed.","Picoliter-scale sample confinement: at ~225 pL the integrated chamber holds volumes an order of magnitude smaller than typical miniature EIT sensors, and the process already demonstrates sub-chambers down to ~20 fL.","Dense electrode access: because electrodes are defined by the metal stack, 16 electrodes cover 66% of the chamber floor and additional metal layers face the chamber from the sides, naturally suited to arrays.","Frequency-flexible operation: with the LO running 500 Hz–100 kHz and similar maps at 1 kHz and 100 kHz for the insulating spheres, the platform could support frequency-difference imaging.","A direct corollary of the claimed first: subsequent integrated EIT designs can now assume the chamber does not have to be external, freeing the packaging to be optimized for the electronics."],"supporting_citations":[{"why":"Transfers the subtractive interconnect-etching method from photonics to microfluidics; supplies the core fabrication idea.","marker":"[9]"},{"why":"Provides the wet-etch chemistry (Al Etch Type A and EDTA/H2O2) that removes metal and barrier layers to form the chamber.","marker":"[10]"},{"why":"Supplies the open-source reconstruction software that turns the measured boundary voltages into conductivity maps.","marker":"[12]"},{"why":"Motivates EIT as a non-invasive label-free imaging technique that this work miniaturizes.","marker":"[1]"},{"why":"Supports the claim that smaller EIT volumes increase sensitivity to imperfections and impurities, used to explain reconstruction inaccuracies.","marker":"[2]"},{"why":"Shows the inter-metal dielectrics are suitable for interfacing with biological samples, justifying the material choice for the chamber.","marker":"[11]"}],"fun_headline_variants":["Picoliter EIT chip carves chamber from wiring metal","First CMOS EIT merges microfluidics and electrodes on-die","16-electrode CMOS imager maps beads in 225-pL chamber","Wiring metal becomes microfluidic chamber for tomographic chip"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The conductivity maps are only meaningful if the voltages at the 16 electrodes are actually set by the water-and-sphere conductivity distribution under the usual Laplace-equation forward model; the paper reports no electrode-interface impedance measurements, no empty-chamber baseline, and no error metric, so if etch damage, bubbles, or electrode-to-electrode variability — rather than the glass spheres — drive the voltage pattern, the 'low conductivity' patches could be artifa","fun_headline_variants_meta":{"raw":{"variants":["Picoliter EIT chip carves chamber from wiring metal","First CMOS EIT merges microfluidics and electrodes on-die","16-electrode CMOS imager maps beads in 225-pL chamber","Wiring metal becomes microfluidic chamber for tomographic chip"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000605,"raw_usage":{"total_tokens":2612,"prompt_tokens":652,"completion_tokens":1960,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":1886}},"tokens_in":396,"tokens_out":1960,"duration_ms":15853,"temperature":1.0,"reasoning_tokens":1886,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T15:06:17.995828+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the exact same measurement with the chamber filled only with water and no microspheres. If the same reconstruction software returns localized low-conductivity patches similar to those in Fig. 6, the correlation is an inverse-solver artifact rather than tomography. A second decisive check is to measure the impedance of each electrode in isolation: if the per-electrode spread predicts where low-conductivity regions appear, electrode non-uniformity — not the sample — is what is being imaged.","supporting_citations":[{"cited_title":"Subtractive photonics in bulk CMOS,","cited_arxiv_id":null,"evidence_quote":"Transfers the subtractive interconnect-etching method from photonics to microfluidics; supplies the core fabrication idea."},{"cited_title":"Fatemi, C","cited_arxiv_id":null,"evidence_quote":"Provides the wet-etch chemistry (Al Etch Type A and EDTA/H2O2) that removes metal and barrier layers to form the chamber."},{"cited_title":"Adler and W","cited_arxiv_id":null,"evidence_quote":"Supplies the open-source reconstruction software that turns the measured boundary voltages into conductivity maps."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Motivates EIT as a non-invasive label-free imaging technique that this work miniaturizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supports the claim that smaller EIT volumes increase sensitivity to imperfections and impurities, used to explain reconstruction inaccuracies."},{"cited_title":"Sarkar and A","cited_arxiv_id":null,"evidence_quote":"Shows the inter-metal dielectrics are suitable for interfacing with biological samples, justifying the material choice for the chamber."}],"review_version":1}