Contactless Electrode Geometry

• Contactless electrode geometry is defined by indirect coupling through conductive bodies, insulators, or gaps, influencing sensitivity, localization, and electrostatic effects.
• It employs advanced methodologies such as Laplacian kernel ridge regression in tactile sensors and admittivity field modeling in EIT to achieve precise measurement mapping.
• Design optimization involves balancing geometric trade-offs—like electrode spacing, gap shaping, and field distribution—to meet performance targets in applications from MEMS devices to high-voltage multipliers.

Contactless electrode geometry denotes a class of electrode arrangements in which coupling to the measured or actuated system is mediated through a continuous conductive body, an insulating access layer, a finite gap, a boundary-admittivity field, or electrostatic induction rather than a directly exposed, one-site/one-contact interface. In the current literature, this includes embedded electrodes in continuous piezoresistive elastomers, vertical photonic probes that use the chip backside as a common electrode, quadrupolar probes operated in capacitive contact above a surface, electrodeless and geometry-adaptive EIT formulations, nanogap redox sensors whose performance is set by diffusion geometry, curved-gap MEMS capacitors, and Cavallo multipliers that accumulate high voltage by repeated induction and charge transfer [(Piacenza et al., 2018); (Melati et al., 2014); (Settimi et al., 2011); (Dardé et al., 2021); (Hyvönen et al., 2016); (Joshi et al., 2019); (Ashok et al., 17 Aug 2025); (Blatnik et al., 11 Apr 2026)].

1. Architectural forms of contactless coupling

The literature does not use a single physical mechanism. Instead, contactless electrode geometry appears in several recurring forms: embedded terminals inside a continuous bulk conductor; vertical top-to-backside conduction paths accessed capacitively through an insulator; probes lifted above a sample in capacitive-contact mode; inverse models that replace exact electrode sets by extended boundary segments and an admittivity field; finite-gap nanogap or MEMS capacitor layouts; and induction-based high-voltage multipliers in which charge is transported by a movable shuttle electrode rather than by a continuous feedthrough [(Piacenza et al., 2018); (Melati et al., 2014); (Settimi et al., 2011); (Dardé et al., 2021); (Joshi et al., 2019); (Ashok et al., 17 Aug 2025); (Blatnik et al., 11 Apr 2026)].

Architecture	Geometric realization	Reported consequence
Continuous piezoresistive body	Four embedded electrodes in a single CNT–PDMS volume	Six unique pairwise measurements from four wires
Vertical CLIPP	One top metal strip plus chip backside common electrode	Reduced footprint and simplified electrical connectivity
Contactless-capable quadrupolar probe	Wenner or dipole-dipole array operated with $h>0$	Capacitive coupling to soils and concretes
Electrodeless EIT	Extended electrodes $E_m$ with boundary admittivity $\zeta$	Simultaneous reconstruction of interior conductivity and effective contact location
Curved MEMS capacitor	Planar movable electrode with curved fixed electrodes	Sensitivity changed without changing proof-mass or footprint
Cavallo multiplier	A, B, C, and ground electrodes with movable shuttle B	High voltage built up by repeated electrostatic induction

A central implication is that “contactless” does not mean “geometry-free.” In every cited system, geometry remains the primary control variable for sensitivity, localization, identifiability, current amplification, or breakdown risk. What changes is the mode by which geometry acts: through overlapping current paths, access capacitance, boundary admittivity, diffusion length, capacitance gradient, or capacitance ratios.

2. Embedded electrodes in continuous conductive media

A particularly explicit contactless geometry is the tactile sensor of "Contact Localization through Spatially Overlapping Piezoresistive Signals" (Piacenza et al., 2018). The sensor is a single continuous volume of piezoresistive elastomer rather than an array of isolated taxels. In the proof-of-concept device, the sensing body is a rectangular block of PDMS filled with MWCNTs at the percolation threshold, with four embedded electrodes routed through side channels and brought out as four wires. The effective sensing area is $10\ \mathrm{mm}\times 16\ \mathrm{mm}$, and the four electrodes generate six unique terminal pairs, $\binom{4}{2}=6$. Because the body is continuous, indentation perturbs the conductive network and changes the resistance measured between every electrode pair (Piacenza et al., 2018).

The geometry is contactless in a specific sense: the electrodes are not directly exposed to the indentation region. Instead, the device uses $30\ \mathrm{mm}$ side channels filled with a more conductive CNT–PDMS mixture at $2.5\ \mathrm{wt.}\%$, while the central sensing volume is $0.4\ \mathrm{wt.}\%$ CNT–PDMS. Copper wires are embedded at the ends of the side channels. This mechanically isolates the wire contacts from direct indentation, so the measured changes predominantly arise from deformation of the central sensing volume rather than disturbance of the terminal contacts (Piacenza et al., 2018).

The inverse problem is formulated directly from data. For each indentation, the feature vector is

$\mathbf{r}_i=(r_i^1,\dots,r_i^6),$

and the full measurement tuple is

$\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$

The learned map is

$E_m$0

The authors first used linear regression and then Laplacian kernel ridge regression to compensate for the nonlinear bias of the linear model, especially the tendency to bias predictions toward the center (Piacenza et al., 2018).

The quantitative result is notable because it derives from a deliberately simple geometry. Using 216 training indentations and 60 test indentations, all at $E_m$1 depth, the reported median localization errors were $E_m$2 for a center predictor, $E_m$3 for a random predictor, $E_m$4 for linear regression, and $E_m$5 for Laplacian ridge regression, with mean error $E_m$6 and standard deviation $E_m$7. The paper therefore reports submillimeter median localization accuracy on a $E_m$8 sensor using only four electrodes and six pairwise measurements (Piacenza et al., 2018).

This geometry differs structurally from taxel arrays and from EIT-like boundary-only schemes. Compared with a conventional $E_m$9 taxel array, which needs at least one wire per taxel and often two, the continuous-body layout uses four wires while extracting six pairwise channels. Compared with boundary-electrode EIT, the electrodes are embedded within the piezoresistive volume, and the inverse map is learned empirically rather than derived from an analytic conductivity model. The same paper also states a scaling law that recurs in later contactless designs: the number of pairwise channels grows quadratically with electrode count, with $\zeta$0, $\zeta$1, and $\zeta$2 pairs (Piacenza et al., 2018).

3. Vertical, capacitive, and standoff geometries

The vertical CLIPP for InP photonic integrated circuits implements a different contactless geometry (Melati et al., 2014). Instead of a coplanar two-electrode layout, it uses a single top metal strip above the waveguide and the backside of the chip as a common electrode. A SiN isolation layer separates the top strip from the waveguide, providing capacitive access without direct electrical contact. The realized waveguide is a $\zeta$3-wide rib with a $\zeta$4-high InGaAsP core, $\zeta$5 etch depth, and an InP substrate thicker than $\zeta$6. The substrate is semi-insulating Fe-doped InP with carrier concentration roughly in the range $\zeta$7–$\zeta$8, and its finite residual conductivity allows the backside to function as a large-area common electrode without backside metallization (Melati et al., 2014).

The equivalent model contains the access capacitance $\zeta$9, waveguide conductance $10\ \mathrm{mm}\times 16\ \mathrm{mm}$0, substrate conductance $10\ \mathrm{mm}\times 16\ \mathrm{mm}$1, and parasitic capacitance $10\ \mathrm{mm}\times 16\ \mathrm{mm}$2. The total conductance is written as

$10\ \mathrm{mm}\times 16\ \mathrm{mm}$3

with $10\ \mathrm{mm}\times 16\ \mathrm{mm}$4 the top-electrode length as the main design parameter. The measurement chain applies sinusoidal excitation $10\ \mathrm{mm}\times 16\ \mathrm{mm}$5 to the bottom electrode, senses current at the top electrode with a TIA, and reconstructs admittance $10\ \mathrm{mm}\times 16\ \mathrm{mm}$6 through lock-in demodulation, with $10\ \mathrm{mm}\times 16\ \mathrm{mm}$7. The resistive sensing region is reported around $10\ \mathrm{mm}\times 16\ \mathrm{mm}$8. The measured light-induced conductance follows $10\ \mathrm{mm}\times 16\ \mathrm{mm}$9 with $\binom{4}{2}=6$0, reaches approximately $\binom{4}{2}=6$1 at $\binom{4}{2}=6$2, and yields a resolution of about $\binom{4}{2}=6$3 with minimum detectable power about $\binom{4}{2}=6$4 (Melati et al., 2014).

A standoff variant appears in the RESPER prototype, which was designed as a quadrupolar, contactless-capable probe configured as Wenner or dipole-dipole arrays (Settimi et al., 2011). In its capacitive-contact mode, the probe is lifted above the surface, with normalized height ratio

$\binom{4}{2}=6$5

so $\binom{4}{2}=6$6 denotes capacitive contact and $\binom{4}{2}=6$7 galvanic contact. For the dipole-dipole family,

$\binom{4}{2}=6$8

with $\binom{4}{2}=6$9 and design configurations $30\ \mathrm{mm}$0, giving $30\ \mathrm{mm}$1, $30\ \mathrm{mm}$2, $30\ \mathrm{mm}$3, $30\ \mathrm{mm}$4, $30\ \mathrm{mm}$5, and $30\ \mathrm{mm}$6. The geometric factor is

$30\ \mathrm{mm}$7

with $30\ \mathrm{mm}$8. The paper reports that $30\ \mathrm{mm}$9 was theoretically attractive but could not be implemented because the required $2.5\ \mathrm{wt.}\%$0 spring-shaft holes could not be drilled within the available plate spacing (Settimi et al., 2011).

These systems illustrate two distinct meanings of contactlessness. In CLIPP, the electrode is geometrically present but electrically isolated from the optical waveguide by SiN; in RESPER, the probe is physically separated from the sample and couples electrostatically through free space or air gap. In both cases, electrode placement, path length, and parasitic coupling remain decisive.

4. Geometry as an unknown in EIT

Electrical impedance tomography provides a more abstract reformulation of contactless electrode geometry. In the electrodeless electrode model, electrodes are not defined as exact geometric subsets $2.5\ \mathrm{wt.}\%$1; instead, the model introduces larger, approximately known extended electrodes $2.5\ \mathrm{wt.}\%$2 and a boundary admittivity field $2.5\ \mathrm{wt.}\%$3 that determines where current actually enters and leaves the body (Dardé et al., 2021). The forward model is

$2.5\ \mathrm{wt.}\%$4

$2.5\ \mathrm{wt.}\%$5

$2.5\ \mathrm{wt.}\%$6

The admissible set requires $2.5\ \mathrm{wt.}\%$7, and the inversion reconstructs both interior conductivity and boundary admittivity through a MAP/Gauss–Newton formulation. Two parametrizations are used: a piecewise-linear representation on $2.5\ \mathrm{wt.}\%$8 and a parametric hat representation with variable height, location, and width on each extended electrode (Dardé et al., 2021).

The conceptual shift is important. The contact is no longer hard-coded as a sharp electrode boundary; instead, electrode location and strength are encoded analytically through $2.5\ \mathrm{wt.}\%$9. The paper reports that this formulation is robust to uncertain electrode geometry in two-dimensional water tank data, that standard CEM degrades under electrode misplacement, and that the PH parametrization often localizes contacts better than the PL parametrization (Dardé et al., 2021).

A related but distinct strategy appears in "Compensation for geometric modeling errors by electrode movement in electrical impedance tomography" (Hyvönen et al., 2016). That work retains the complete electrode model but estimates electrode locations and sizes as part of the inverse problem. In 2D disk geometry, the parameter vector is

$0.4\ \mathrm{wt.}\%$0

while in a 3D cylindrical geometry it becomes

$0.4\ \mathrm{wt.}\%$1

The Fréchet derivative of the measurement map with respect to electrode deformation is sampled through

$0.4\ \mathrm{wt.}\%$2

The theoretical explanation is approximate conformal invariance of the CEM in 2D for small electrodes; the paper states that the approach is effective in 2D, including experimental water tank data, but does not in general yield a functional algorithm in 3D (Hyvönen et al., 2016).

These EIT results address a common misconception. “Electrodeless” does not eliminate the boundary model; it replaces explicit geometric certainty by a distributed boundary parameterization. Likewise, allowing electrodes to move does not provide a universal correction for geometry mismatch: the cited 2D results are supported by conformal flexibility, whereas the 3D case is explicitly reported as limited (Dardé et al., 2021, Hyvönen et al., 2016).

5. Curvature, gap engineering, and nonuniform fields

In capacitive MEMS accelerometers, contactless electrode geometry appears as controlled gap shaping. "Design and Analysis of Curved Electrode Configurations for Enhanced Sensitivity in 1-Axis MEMS Accelerometers" studies a planar movable electrode against six fixed-electrode profiles: biconvex, biconcave, concavo-convex, convexo-concave, plano-convex, and plano-concave (Ashok et al., 17 Aug 2025). The output relation is

$0.4\ \mathrm{wt.}\%$3

with displacement

$0.4\ \mathrm{wt.}\%$4

and differential capacitances

$0.4\ \mathrm{wt.}\%$5

For a convex arc,

$0.4\ \mathrm{wt.}\%$6

whereas a concave arc uses

$0.4\ \mathrm{wt.}\%$7

The paper concludes that biconvex electrodes yield the greatest sensitivity improvement, with sensitivity monotonically increasing with arc length; concave and plano-concave designs reduce performance; and concavo-convex and convexo-concave profiles introduce polarity inversion in the output voltage. Analytical and COMSOL results agree with deviation less than $0.4\ \mathrm{wt.}\%$8 across all six configurations, and the reported improvement is obtained without changing the proof-mass dimensions or device footprint (Ashok et al., 17 Aug 2025).

Nanogap electrochemical biosensors provide an analogous result in reaction–diffusion form rather than in electrostatics (Joshi et al., 2019). The key geometric variable in 1D is the gap width $0.4\ \mathrm{wt.}\%$9, and the diffusion-limited steady-state current density for two planar electrodes is

$\mathbf{r}_i=(r_i^1,\dots,r_i^6),$0

The paper maps reaction–diffusion onto an equivalent RC network, with transport resistor

$\mathbf{r}_i=(r_i^1,\dots,r_i^6),$1

and node capacitance

$\mathbf{r}_i=(r_i^1,\dots,r_i^6),$2

In 2D, it compares layouts with varying area, fixed area but different arrangement, and planar nanostructuring on a single surface. The reported design rules are that smaller gap gives stronger amplification and lower detection limit, layout matters even at fixed area, current in 2D is not proportional to total area, and planar nanostructuring is fabrication-friendly but increases settling time. The paper reports about $\mathbf{r}_i=(r_i^1,\dots,r_i^6),$3 settling time for a conventional parallel nanogap scheme and about $\mathbf{r}_i=(r_i^1,\dots,r_i^6),$4 for planar scheme C (Joshi et al., 2019).

Taken together, these papers show that the operative design variable is often the spatial distribution of gap rather than the nominal area of an electrode. A plausible implication is that contactless performance optimization frequently reduces to shaping the local coupling kernel—electrostatic, diffusive, or capacitive—rather than merely enlarging electrodes.

6. Induction geometries, scaling laws, and limits

A high-voltage realization of contactless electrode geometry is the Cavallo multiplier designed for a cryogenic nEDM experiment (Blatnik et al., 11 Apr 2026). The active set comprises A, B, and C electrodes plus grounding structures. Electrode A is connected to an external input of about $\mathbf{r}_i=(r_i^1,\dots,r_i^6),$5, B is a movable shuttle, and C is the high-voltage output electrode. The cycle is: B sits near A and is grounded; A induces charge on B through $\mathbf{r}_i=(r_i^1,\dots,r_i^6),$6; B disconnects from ground and moves to C; B touches C and charge redistributes; then B returns to A. The output therefore grows by repeated electrostatic induction and charge transfer rather than by a continuous high-voltage feedthrough (Blatnik et al., 11 Apr 2026).

The gain per cycle is expressed as

$\mathbf{r}_i=(r_i^1,\dots,r_i^6),$7

The design objective was to achieve gain $\mathbf{r}_i=(r_i^1,\dots,r_i^6),$8, target $\mathbf{r}_i=(r_i^1,\dots,r_i^6),$9, and keep peak surface fields below about $\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$0. The final axisymmetric geometry uses smoothly joined hyperbolic-tangent parametric curves, including

$\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$1

The reported final design achieves mathematical gain $\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$2, reaches $\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$3 in about $\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$4 cycles, and yields a maximum electric field of $\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$5, with the highest fields confined to small areas (Blatnik et al., 11 Apr 2026).

Several limitations recur across otherwise disparate contactless geometries. In the piezoresistive tactile sensor, not all pairwise channels are equally sensitive, and edge regions are harder to localize than central ones (Piacenza et al., 2018). In electrodeless and geometry-adaptive EIT, robustness to uncertain contact geometry is strong in 2D but not general in 3D (Dardé et al., 2021, Hyvönen et al., 2016). In curved MEMS capacitors, convexity and concavity are not interchangeable: biconvex improves sensitivity, while concave-based designs reduce it (Ashok et al., 17 Aug 2025). In nanogap redox sensors, smaller gaps improve amplification but the more strongly two-dimensional planar layouts settle more slowly (Joshi et al., 2019). In the Cavallo multiplier, higher gain, lower breakdown probability, and cryogenic mechanical constraints form an explicit three-way compromise (Blatnik et al., 11 Apr 2026).

The general design lesson is therefore restrictive rather than universal. Contactless electrode geometry is not a single topology but a family of geometry-driven compromises. The common pattern is that the contact region is displaced, distributed, or replaced—embedded in bulk, mediated by an insulator, lifted off the sample, encoded by $\Phi_i=(x_i,y_i,d_i,r_i^1,\dots,r_i^6).$6, spread over a nonuniform gap, or realized through a shuttle electrode—but the governing performance still depends on how field lines, current paths, diffusion trajectories, or capacitance ratios are shaped by the electrode layout.