Quantum Capacitance Readout

• Quantum capacitance readout is a dispersive measurement technique where a quantum system’s state-dependent differential capacitance alters the resonance characteristics of an RF or microwave circuit.
• It relies on measuring the curvature of eigenenergy with respect to control parameters, distinguishing between geometric, quantum, and tunneling capacitance contributions.
• Practical implementations focus on mitigating parasitic capacitance, optimizing impedance matching, and reducing amplifier noise to enhance sensitivity in various quantum device platforms.

Searching arXiv for recent and foundational papers on quantum capacitance readout. Quantum capacitance readout is a dispersive measurement technique in which a quantum system contributes a state-dependent differential capacitance to an electrical node, so that the resonance frequency, phase, or amplitude of a coupled RF or microwave resonator becomes a proxy for the system’s state. In the adiabatic limit, this capacitance is tied to the curvature of the relevant eigenenergy with respect to a charge-coupled control parameter; in broader coupled-system treatments it is the large-detuning limit of the same resonator frequency shift that, at smaller detuning, is described as dispersive cavity pull from virtual transitions (Park et al., 2020). At the microscopic level, the same quantity can be written as a static charge susceptibility, $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$, which combines with geometric capacitance through a series relation $1/C_{\mathrm{eff}}=1/C+1/\eta$ in the semiclassical limit (Mukherjee et al., 2010).

1. Theoretical basis and definitions

Quantum capacitance readout rests on the distinction between geometric capacitance and a state-dependent reactive response of the electronic system. In a many-body formulation, the local charge is $\partial E/\partial V$, and the quantum contribution is the static susceptibility $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$; under the approximations identified in the microscopic derivation, the measured capacitance becomes the series combination

$$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$

with $C$ the geometric contribution and $\eta$ the quantum contribution (Mukherjee et al., 2010). In the few-body gate-voltage formulation, the same idea appears as an energy Hessian: for a single control voltage, $C_k(V_0)\equiv E_k''(V_0)$, and for several gate voltages it generalizes to the quantum-capacitance matrix

$$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$

The measured response along a drive direction $\boldsymbol v$ is then the quadratic form $1/C_{\mathrm{eff}}=1/C+1/\eta$0 (Secchi et al., 2022).

The relation between “quantum capacitance readout” and “dispersive readout” is not a relation between two unrelated mechanisms. A general cQED treatment shows that the resonator shift induced by a quantum circuit contains both a curvature contribution and virtual-transition terms, and that in the adiabatic limit $1/C_{\mathrm{eff}}=1/C+1/\eta$1 it reduces to

$1/C_{\mathrm{eff}}=1/C+1/\eta$2

so that the readout is naturally described as an effective capacitance or inductance of the circuit state (Park et al., 2020). At smaller detuning, the same coupled Hamiltonian yields the usual dispersive cavity pull, and the full perturbative expression is required.

A separate but closely related distinction is between quantum capacitance and tunneling capacitance. In the double-quantum-dot theory of charge and spin qubits, the gate-seen differential capacitance is written as

$1/C_{\mathrm{eff}}=1/C+1/\eta$3

where the parametric term $1/C_{\mathrm{eff}}=1/C+1/\eta$4 contains a quantum-capacitance contribution from adiabatic band curvature and a tunneling-capacitance contribution from population redistribution during the RF cycle (Mizuta et al., 2016). This distinction is central experimentally: near a coherent avoided crossing, band curvature dominates; near a reservoir-coupled transition, the reactive signal can instead be governed by rapid occupation dynamics.

2. Resonant transduction and measured observables

The circuit-level transduction mechanism is standard RF reflectometry. A device gate, lead, or auxiliary sensor is embedded in an $1/C_{\mathrm{eff}}=1/C+1/\eta$5 resonator, an RF carrier is sent down a $1/C_{\mathrm{eff}}=1/C+1/\eta$6 line, and the reflected signal is monitored through

$1/C_{\mathrm{eff}}=1/C+1/\eta$7

with $1/C_{\mathrm{eff}}=1/C+1/\eta$8. A small capacitance perturbation changes the resonator impedance and shifts the resonance; near resonance this appears as a measurable change in reflected phase or amplitude (Schupp et al., 2018). In one-port matching analyses the same logic is expressed through $1/C_{\mathrm{eff}}=1/C+1/\eta$9 at perfect match, with capacitance changes mapped into $\partial E/\partial V$0 through the resonator slope (Ares et al., 2015).

Phase-sensitive formulations make the transduction especially transparent. For an accumulation-mode silicon gate sensor, the small-signal phase response is

$\partial E/\partial V$1

where $\partial E/\partial V$2 is the resonator quality factor, $\partial E/\partial V$3 the parasitic capacitance to ground, and $\partial E/\partial V$4 the gate-connected capacitance change (Rossi et al., 2016). For a single Cooper-pair box coupled to a lumped resonator, the reflected phase shift at fixed probe frequency is

$\partial E/\partial V$5

so that larger $\partial E/\partial V$6, larger state-dependent $\partial E/\partial V$7, and smaller total resonator capacitance all increase phase contrast (Persson et al., 2010).

What is actually measured depends on the implementation. The observable may be reflected phase $\partial E/\partial V$8, demodulated $\partial E/\partial V$9, or a scalar such as $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$0 in time-domain reflectometry (Schupp et al., 2018). In all cases, the target quantity is a small reactive change in the device admittance. The practical consequence is that parasitic capacitance, impedance matching, and the noise temperature of the first gain stage are not peripheral details; they set whether a given $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$1 is visible at all.

3. Canonical implementations

Several device classes now define the experimental landscape.

Platform	Dominant capacitive mechanism	Representative result
Single Cooper-pair box	Quantum capacitance of avoided crossing	Single-shot parity measurement in approximately $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$2 ns (Persson et al., 2010)
Accumulation-mode Si MOS dot	Tunneling capacitance competing with $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$3	$$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$4; background $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$5 at $$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$6 (Rossi et al., 2016)
GaAs double quantum dot with SQUID preamp	Capacitance-sensitive RF reflectometry	$$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$7 (Schupp et al., 2018)
Surface electrons on liquid helium	Quantum capacitance of driven Rydberg transition	$$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$8 (Jennings et al., 14 Apr 2025)
Silicon gate-based sensing of open/closed channels	Gate susceptance from tunneling or quantum capacitance	$$\eta=-\partial Q/\partial V=-\partial^2E/\partial V^2$$9 at $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$0 MHz and $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$1 at $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$2 MHz (Ahmed et al., 2018)

The superconducting single Cooper-pair box is a canonical early example. In the two-level regime, its level splitting is $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$3, and the state-dependent quantum capacitance is

$$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$4

Because the sign reverses between ground and excited states, the reflected phase directly resolves parity-dependent switching, and resonators near $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$5–$$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$6 MHz with $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$7 enabled single-shot parity measurement in about $$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$8–$$\frac{1}{C_{\mathrm{eff}}}=\frac{1}{C}+\frac{1}{\eta},$$9 ns (Persson et al., 2010).

In semiconductor quantum dots, the same gate-based architecture often measures a total differential capacitance rather than a pure adiabatic quantum capacitance. In an accumulation-mode silicon MOS dot, the total sensed capacitance is

$C$0

where $C$1 is the desired tunneling capacitance and $C$2 is the detector gate’s own MOS capacitance. The central experimental result is that once the sensing gate exceeds threshold, approximately $C$3, $C$4 rises rapidly and masks the dot signal; at $C$5, the RF swing alone produces $C$6, already much larger than the single-electron tunneling capacitance of $C$7 (Rossi et al., 2016). This establishes a basic design rule for silicon readout: operating a sensing gate above accumulation threshold converts the resonator into a probe of local MOS electrostatics rather than a selective probe of dot-related susceptibility.

Gate-based sensing is not limited to isolated charge transitions. Dispersive gate sensing of a quantum point contact shows that open one-dimensional channels also produce measurable gate-connected quantum capacitance, allowing Van Hove singularities of a ballistic 1D system to be resolved and revealing localized charge pockets not easily accessed with DC transport (Jarratt et al., 2019). In accumulation-mode Si/SiGe devices, RF reflectometry was engineered around the opposite problem—large gate-to-2DEG parasitic capacitance—and two mitigation strategies, including a split accumulation gate, enabled charge readout fidelity above $C$8 with less than $C$9 integration time, specifically $\eta$0 at $\eta$1 (Liu et al., 2020).

The platform diversity now extends beyond superconducting and semiconductor nanostructures. For surface electrons on liquid helium, resonant microwave excitation of the $\eta$2 Rydberg transition changes the image charge on nearby electrodes by $\eta$3, with $\eta$4 and $\eta$5, and the resulting state-dependent capacitance is measured by RF reflectometry combined with microwave frequency modulation (Jennings et al., 14 Apr 2025). This implementation explicitly identifies the measured quantity with the curvature of the dressed two-level spectrum and the detuning derivative of the image-charge expectation value.

4. Sensitivity engineering and amplification

Raw device physics does not by itself determine readout performance. Matching, resonator design, and the first cryogenic amplifier dominate the measured sensitivity.

A clear benchmark for circuit optimization is the varactor-tuned one-port reflectometry network developed for quantum dots. By tuning to perfect impedance matching in situ, that circuit reached a capacitance sensitivity of $\eta$6 with bandwidth above $\eta$7 at a maximum source-drain bias of $\eta$8 root-mean-square (Ares et al., 2015). The same work showed why sensitivity alone can mislead for qubit readout: although the naive estimate from $\eta$9 and an assumed $C_k(V_0)\equiv E_k''(V_0)$0 suggests a very short measurement time, a more qubit-relevant metric is $C_k(V_0)\equiv E_k''(V_0)$1, the sensitivity to oscillating induced charge at a given RF bias.

The strongest low-frequency semiconductor result in the set is the insertion of an ultra-low-noise SQUID amplifier as the first cryogenic gain stage of a reflectometry chain. Operating near $C_k(V_0)\equiv E_k''(V_0)$2 MHz, the SQUID preamplifier provided gain about $C_k(V_0)\equiv E_k''(V_0)$3 dB and a system noise temperature $C_k(V_0)\equiv E_k''(V_0)$4 mK, quoted as below $C_k(V_0)\equiv E_k''(V_0)$5 mK, which enabled a record capacitance sensitivity

$C_k(V_0)\equiv E_k''(V_0)$6

at $C_k(V_0)\equiv E_k''(V_0)$7 dBm (Schupp et al., 2018). In the same setup, a $C_k(V_0)\equiv E_k''(V_0)$8 charge stability diagram was recorded in $C_k(V_0)\equiv E_k''(V_0)$9 ms, the signal-to-noise ratio at $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$0 was about $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$1, and the extrapolated integration time for $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$2 was $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$3. The authors also emphasized a platform-specific point of principle: for singlet-triplet qubits, low operating frequency is desirable because if the readout frequency becomes comparable to the interdot tunnel rate, the quantum capacitance is suppressed.

Semiconductor quantum-capacitance devices can also serve as amplifiers rather than only as sensed elements. The quantum-capacitance parametric amplifier built from a GaAs/AlGaAs 2DEG uses a gate-tunable quantum capacitance $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$4 as the nonlinear reactive element in a pumped tank circuit, with gain $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$5, center frequency $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$6, and intrinsic noise temperature $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$7 (Kass et al., 2023). Its 1 dB compression point is $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$8 at $$C_{k;ij}\equiv \frac{\partial^2 E_k}{\partial V_i\partial V_j}.$$9 gain, it remains operable at $\boldsymbol v$0, and it sustains $\boldsymbol v$1 gain at in-plane magnetic fields of $\boldsymbol v$2, $\boldsymbol v$3, and $\boldsymbol v$4. That performance is not quantum-limited, but it directly addresses a practical bottleneck of semiconductor-qubit readout: substantial cryogenic gain at UHF frequencies with low dissipation and magnetic-field compatibility.

Taken together, these results fix the main engineering priorities. Good impedance matching maximizes conversion of $\boldsymbol v$5 into reflected signal. Low parasitic capacitance preserves both frequency pull and matching freedom. The first amplifier must add as little noise as possible, but must also tolerate the RF power needed to overcome downstream noise without compression. The readout frequency must remain low enough for the targeted charge susceptibility to remain adiabatic. These are not implementation details; they are the dominant constraints on whether a nominal quantum-capacitance signal is experimentally useful.

5. Multigate, multilayer, and correlated-electron extensions

Quantum capacitance readout becomes substantially richer once the system depends on several control voltages. The multi-voltage formalism replaces the scalar capacitance by a Hessian $\boldsymbol v$6, and the measured signal depends on the direction of the applied RF oscillation in gate-voltage space: $\boldsymbol v$7 Because the matrix is symmetric, it has principal axes, and the optimal drive direction is the eigenvector with the largest-magnitude eigenvalue (Secchi et al., 2022). In Hubbard-model quantum-dot arrays this formalism identifies the boundaries between charge stability regions as the high-susceptibility loci and provides a direct procedure for optimizing spin and charge discrimination. It also shows that one entire common-mode direction can be a zero-eigenvalue direction, so that not every gate combination is useful for dispersive readout.

In layered conductors and atomically thin capacitors, the meaning of “quantum capacitance” itself changes. For double-layer graphene, the relevant quantity is not the DOS of a single isolated sheet but a collective two-layer response expressed through derivatives of the layer chemical potentials. The total capacitance retains a series form,

$\boldsymbol v$8

but $\boldsymbol v$9 contains the coupled thermodynamic stiffness of both layers (Parhizgar et al., 2017). Near the neutrality point, the paper reports that the quantum capacitance behaves like $1/C_{\mathrm{eff}}=1/C+1/\eta$00, with $1/C_{\mathrm{eff}}=1/C+1/\eta$01 decreasing as the density imbalance increases, and also reports approximately linear dependence on gate voltage over a small bias window. This is already a warning against transplanting single-layer DOS intuition into coupled-layer readout.

A more general warning appears in the theory of cross quantum capacitance. For two coupled electron liquids separated by atomic-scale distances, the exact capacitance depends on both intralayer and interlayer polarizabilities: $1/C_{\mathrm{eff}}=1/C+1/\eta$02 The interlayer terms $1/C_{\mathrm{eff}}=1/C+1/\eta$03 define the cross quantum capacitance, and the paper shows that this contribution can either increase or decrease the total capacitance; a non-monotonic dependence of total capacitance on plate separation would be an unambiguous manifestation (Berthod et al., 2021). This places a firm limit on a common simplification: in ultrathin capacitors or van der Waals heterostructures, subtracting geometric capacitance does not generally isolate a single-layer DOS.

The same DOS-to-capacitance transduction can even appear in metallic junctions. In a nanogap between Pt or Au electrodes, the capacitance saturates at short distance rather than following the classical increase, reaching a quantum-capacitance limit before contact formation, and then shows tunneling-induced leakage at still smaller separation (Ara et al., 7 Oct 2025). Because the quantum contribution depends on surface DOS, the same setup resolves a change induced by toluene adsorption on Au. For readout language, this is a useful limiting case: once the geometric capacitance is made sufficiently large, even a metal can become quantum-capacitance limited.

6. Majorana applications, limitations, and misconceptions

Topological devices have made quantum capacitance readout a parity-sensitive tool rather than only a charge-sensitive one. In the auxiliary-dot scheme for Majorana devices, parity-resolved quantum-capacitance traces versus auxiliary-dot energy yield two figures of merit simultaneously: the relative positions of the even- and odd-sector QC maxima determine the ground-state splitting, while the relative magnitudes determine the Majorana polarization, i.e. the local overlap quality of the Majorana bound states (Dourado et al., 19 Jun 2026). The same work gives analytic low-energy expressions and validates them against microscopic QD-based Kitaev-chain and qubit models. In that setting, QC is not merely a readout of parity; it is a parity-preserving diagnostic of device quality.

That usefulness comes with an important controversy. In a disordered semiconductor-superconductor nanowire connected end to end through a quantum dot and threaded by flux, parity-dependent $1/C_{\mathrm{eff}}=1/C+1/\eta$04-periodic quantum-capacitance oscillations do occur, but theory shows that the same phenomenology can arise not only from topological, well-separated Majorana zero modes, but also from partially separated Majorana modes with appreciable overlap and from topologically trivial quasi-Majorana or Andreev bound states (Stanescu et al., 29 May 2025). The result is specific and narrow: flux-dependent fermion-parity-sensitive QC oscillations are a real parity-sensitive observable, but they are not by themselves a unique signature of topological protection.

A second family of misconceptions concerns what the resonator is actually measuring. In silicon accumulation-mode dots, operation above the accumulation threshold causes the sensing gate’s own $1/C_{\mathrm{eff}}=1/C+1/\eta$05 to dominate the response, so a large background can be mistaken for an intrinsic dot signal if the threshold physics is not identified explicitly (Rossi et al., 2016). In accumulation-mode Si/SiGe reflectometry, large gate-to-2DEG capacitance can create RF leakage paths that destroy matching unless the on-chip and off-chip environment is designed around them (Liu et al., 2020). In low-noise reflectometry chains, dynamic range limits matter as much as noise temperature: once the first cryogenic amplifier enters compression or saturation, sensitivity degrades even though the device may nominally be “better matched” (Schupp et al., 2018). And in metallic or ultrashort-gap systems, tunneling leakage converts a purely reactive signal into a mixed reactive-dissipative one (Ara et al., 7 Oct 2025).

The mature view that emerges is therefore narrow but robust. Quantum capacitance readout is not a single microscopic mechanism but a family of dispersive measurements in which a quantum system contributes a small, state-dependent reactive admittance to a resonant circuit. Depending on the platform, that admittance may be dominated by energy-band curvature, tunneling susceptibility, interlayer polarizability, or image-charge redistribution. The central experimental problem is always the same: isolate the desired reactive channel from parasitic capacitance, dissipative loading, and amplifier noise strongly enough that the state-dependent $1/C_{\mathrm{eff}}=1/C+1/\eta$06 remains interpretable.