Pith. sign in

REVIEW 4 major objections 5 minor 12 references

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.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

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.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A credible first-of-kind integration demo that would be much stronger with a water-only baseline and full EIDORS disclosure. the 4 major comments →

arxiv 2508.20431 v1 pith:ANHEBOCI submitted 2025-08-28 physics.ins-det eess.SP

Electrical Impedance Tomography with an Integrated Picoliter-Volume Subtractive Microfluidic Chamber in 65 nm CMOS

Antonio Victor Machado de Oliveira , Debjit Sarkar , Ali Hajimiri This is my paper
classification physics.ins-det eess.SP
keywords electrical impedance tomographyCMOSmicrofluidicssubtractive fabricationinterconnect layerspicoliter chamberconductivity imagingimpedance image reconstruction
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

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.

Core claim

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

What carries the argument

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.

Load-bearing premise

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

What would settle it

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.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • 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.

Where Pith is reading between the lines

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

  • 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.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

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).

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 (4)
  1. [Sec. IV, Fig. 6] 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.
  2. [Sec. IV, electrode characterization] 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.
  3. [Sec. IV, Fig. 6 and reconstruction parameters] 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.
  4. [Sec. IV, EIDORS setup] 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.
minor comments (5)
  1. [Title and Abstract] The title contains 'V olume' with an intervening space; this should be corrected to 'Volume'.
  2. [Sec. II, Fig. 3 caption] 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.
  3. [Sec. IV, Fig. 7] 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.
  4. [Sec. IV, Fig. 8] 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.
  5. [References] 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.

Circularity Check

0 steps flagged

No circularity found; the reconstruction pipeline is an external inverse solver and the self-citations supply fabrication precedent, not the target result.

full rationale

The paper's central claim is an experimental demonstration: EIT with fully integrated on-chip microfluidics in 65 nm CMOS. The reconstruction chain is standard EIT: boundary voltages are measured from a 16-electrode array and fed to EIDORS, an external open-source solver (ref. 12), which inverts the Laplace-equation forward model. No parameter is fitted to the data and then renamed as a prediction; the conductivity maps are direct outputs of an independent solver. The self-citations (refs. 9–11) support the subtractive fabrication method used to create the chamber and electrodes, but that method is prior published work, not the EIT result, and no uniqueness theorem or ansatz is imported from those citations to force the reconstruction. The acknowledged limitations—ill-posedness, sensitivity to impurities, temperature/fluid-flow drift, SNR variation—are validation concerns, not circularity: they do not make the reconstructed image equal to the input voltages by construction. No quoted step reduces to its own input, so there is no circular derivation.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The central claim rests on four domain assumptions, the most fragile being that measured voltages reflect the bulk conductivity distribution rather than electrode interface effects. Three hand-chosen numerical settings shape the displayed reconstructions; none are disclosed. No invented physical entities are introduced.

free parameters (3)
  • EIDORS regularization scheme and weight = not reported
    The ill-posed inverse problem requires regularization (cited ref. 8); the chosen scheme strongly shapes the Fig. 6 maps, and it is not stated.
  • Stimulation electrode pattern and LO frequency = 50 kHz (Fig. 6); 1-100 kHz (Fig. 8)
    Which electrode pairs are driven and at what frequency change the measured voltage sets; the switching pattern is not specified.
  • Display conductivity range for reconstructed maps = not reported
    The colormap normalization in Fig. 6 determines how convincingly spheres appear as low-conductivity areas; the scale is not disclosed.
axioms (4)
  • standard math Boundary voltages obey the Laplace-equation forward model with known electrode geometry.
    Foundation of the EIDORS reconstruction invoked in Sec. III; standard EIT modeling assumption.
  • domain assumption Glass microspheres have substantially lower conductivity than the surrounding water.
    Stated in Sec. IV ('the microspheres maintain a low conductivity versus frequency'); required for the low-conductivity regions to be attributed to the spheres.
  • domain assumption Measured voltage differences are dominated by the chamber's conductivity distribution, with negligible or invariant electrode-electrolyte interface impedance.
    Unstated but load-bearing; no electrode impedance or calibration data are reported in Sec. IV.
  • domain assumption The wet-etched chamber is fluid-tight and water-filled, and all 16 electrodes survive the alternating Al-etch and EDTA/H2O2 processing.
    Required for the Sec. IV measurements to reflect the intended sample; individual electrode survival is not characterized.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Electrical Impedance Tomography with an Integrated Picoliter-Volume Subtractive Microfluidic Chamber in 65 nm CMOS." pith.science (2026). https://pith.science/paper/ANHEBOCI

@misc{pith2026250820431,
  author       = {Pith},
  title        = {Pith review of: Electrical Impedance Tomography with an Integrated Picoliter-Volume Subtractive Microfluidic Chamber in 65 nm CMOS},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ANHEBOCI}},
  note         = {Machine review of arXiv:2508.20431}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Electrical impedance tomography with fully integrated microfluidics and electronics is presented for the first time in a CMOS chip. Chambers and electrodes are fabricated in the interconnect layers of a 65 nm CMOS chip through post-processing, enabling picoliter-volumes to be processed and imaged. Tomography maps are reconstructed by reading out voltages from a 16-element electrode array and processing the data off-chip, and sources of variation in reconstruction are discussed. The EIT system presented in this work serves as a proof-of-concept towards using CMOS as a platform for co-integrated microfluidics and electronics.

Figures

Figures reproduced from arXiv: 2508.20431 by Ali Hajimiri, Antonio Victor Machado de Oliveira, Debjit Sarkar.

Figure 1
Figure 1. Figure 1: Illustration of field lines and equipotential curves for a pair of [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A standard CMOS chip is designed with passivation [PITH_FULL_IMAGE:figures/full_fig_p001_2.png] view at source ↗
Figure 2
Figure 2. Figure 2: Simplified illustration of the subtractive method used to create [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Micrograph of a 20 µm by 20 µm microfluidic chamber (a) before [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figure 7
Figure 7. Figure 7: Histograms of amplitude variation over time for electrodes measured [PITH_FULL_IMAGE:figures/full_fig_p003_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Tomography map reconstructed for the sample measured in Fig. 6b [PITH_FULL_IMAGE:figures/full_fig_p003_8.png] view at source ↗
Figure 6
Figure 6. Figure 6: Micrographs of samples and their corresponding reconstructed [PITH_FULL_IMAGE:figures/full_fig_p003_6.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

  1. [1]

    B. Sun, Y . Zhao, H. Zhong, L. Ma and H. Wang, ”A New Method for Electrical Impedance Tomography With Incomplete Electrode Array,” in IEEE Sensors Journal, vol. 23, no. 18, pp. 21610-21619, 2023

    work page 2023

  2. [2]

    Z. Chen, Y . Yang and P. -O. Bagnaninchi, ”Hybrid Learning-Based Cell Aggregate Imaging With Miniature Electrical Impedance Tomography,” in IEEE Transactions on Instrumentation and Measurement, vol. 70, pp. 1-10, 2021

    work page 2021

  3. [3]

    Y . Yang, J. Jia, S. Smith, N. Jamil, W. Gamal and P. -O. Bagnaninchi, ”A Miniature Electrical Impedance Tomography Sensor and 3-D Image Reconstruction for Cell Imaging,” in IEEE Sensors Journal, vol. 17, no. 2, pp. 514-523, 2017

    work page 2017

  4. [4]

    Electrical impedance tomog- raphy system based on active electrodes

    Gaggero PO, Adler A, Brunner J, Seitz P. Electrical impedance tomog- raphy system based on active electrodes. Physiol Meas. 33(5):831-47, 2012

    work page 2012

  5. [5]

    H. Dong, R. Bingnan Liu, L. Micklem, S. Peisan E, F. Giorgio-Serchi and Y . Yang, ”Data-Efficient Tactile Sensing With Electrical Impedance Tomography,” in IEEE Sensors Journal, vol. 25, no. 11, pp. 19724- 19733, 2025

    work page 2025

  6. [6]

    Electrical impedance tomography for real-time and label-free cellular viability assays of 3D tumour spheroids,

    H. Wu, Y . Yang, P. O. Bagnaninchi, and J. Jia, “Electrical impedance tomography for real-time and label-free cellular viability assays of 3D tumour spheroids,” Analyst, vol. 143, no. 17, pp. 4189–4198, 2018

    work page 2018

  7. [7]

    Y . Yang, H. Wu, J. Jia and P. -O. Bagnaninchi, ”Scaffold-Based 3-D Cell Culture Imaging Using a Miniature Electrical Impedance Tomography Sensor,” in IEEE Sensors Journal, vol. 19, no. 20, pp. 9071-9080, 2019

    work page 2019

  8. [8]

    Image reconstruction in elec- trical impedance tomography based on structure-aware sparse Bayesian learning,

    S. Liu, J. Jia, Y . D. Zhang, and Y . Yang, “Image reconstruction in elec- trical impedance tomography based on structure-aware sparse Bayesian learning,” IEEE Trans. Med. Imag., vol. 37, no. 9, pp. 2090–2102, Sep. 2018

    work page 2090

  9. [9]

    Subtractive photonics in bulk CMOS,

    C. Ives, D. Sarkar, and A. Hajimiri, “Subtractive photonics in bulk CMOS,” IEEE Journal of Solid-State Circuits, vol. 58, no. 11, pp. 3030–3043, 2023

    work page 2023

  10. [10]

    Fatemi, C

    R. Fatemi, C. Ives, A. Khachaturian, and A. Hajimiri, ”Subtractive photonics,” Opt. Express 29, 877-893, 2021

    work page 2021

  11. [11]

    Sarkar and A

    D. Sarkar and A. Hajimiri, ”Visible spectrum waveguiding in bulk CMOS,” Opt. Express 31, 42365-42372, 2023

    work page 2023

  12. [12]

    Adler and W

    A. Adler and W. R. B. Lionheart, ”Uses and abuses of EIDORS: An extensible software base for EIT”, Physiol. Meas. 27:S25-S42, 2006

    work page 2006

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.