Knight-Leveson RC Experiment

• The Knight-Leveson experiment is an RC discharge study using a vacuum capacitor that exposes deviations from the ideal single-exponential decay.
• It demonstrates a two-exponential behavior where a fast decay (α = 1/(RC)) is followed by a slower tail (β ≪ α) attributed to surface-charge dynamics and geometric effects.
• The experiment serves as both a scientific probe and a pedagogical tool, prompting refined circuit models and careful validation of measurement techniques.

Searching arXiv for the cited papers and related context. The Knight–Leveson experiment, in the context of upper-division laboratory and pedagogy circles, denotes an RC discharge experiment using a vacuum (or air) capacitor that reveals reproducible deviations from the textbook single-exponential decay. Its defining purpose is to strip away ordinary dielectric-loss mechanisms and expose an open problem: why $V(t)$ does not follow $V_0 e^{-t/(RC)}$ even in the “simplest” capacitor-resistor circuit. In the implementation described in the recent teaching-focused literature, a vacuum capacitor is placed in parallel with a precision resistor, charged to high voltage, disconnected from the source, and then monitored with high-precision voltage and current measurements; the resulting discharge displays an early-time exponential followed by a slower late-time tail, motivating model refinement, validation, and a Maxwellian account of current guidance by surface charges (Kowalski, 2024).

1. Definition and conceptual scope

In this usage, the phrase refers to an RC discharge experiment using a vacuum capacitor-resistor circuit. The experiment is framed both as a scientific probe and as a pedagogical device. Scientifically, it isolates behavior that is often hidden by ordinary dielectric-loss mechanisms. Pedagogically, it places students in contact with an unresolved discrepancy between idealized circuit theory and precision measurements.

The choice of a vacuum capacitor is central. Using a vacuum (or air) capacitor minimizes the usual internal dielectric phenomena, specifically polarization, absorption or soakage, and nonlinear loss tangents. This leaves three classes of effects emphasized in the literature: surface-charge dynamics on conductors and wires, possible non-Ohmic leakage paths external to the dielectric bulk, and geometric dependence of charge redistribution and relaxation when the plate fields are confined to the interplate region but the currents must be guided by fields outside the capacitor (Kowalski, 2024).

The main objective is therefore not merely to confirm the standard RC law, but to show that real circuits, even when made deliberately simple, can depart from textbook models. Careful measurements then force a refinement of both modeling assumptions and experimental design.

2. Apparatus, circuit topology, and measurement protocol

The reported apparatus centers on a sealed vacuum capacitor, Comet model CFMN-2800BAC/8-DE-G, with nominal $C = 2.789\ \text{nF}$. The discharge resistors are $1\%$ metal-film, $1/4\ \text{W}$, used in single-resistor configurations with $R = 47\ \text{k}\Omega$, $182\ \text{k}\Omega$, $337\ \text{k}\Omega$, and $637\ \text{k}\Omega$. A two-resistor configuration is also used, in which each plate is connected to ground through separate resistors $R_1 = 337\ \text{k}\Omega$ and $V_0 e^{-t/(RC)}$0. The source is a $V_0 e^{-t/(RC)}$1 DC power supply, disconnected by a mechanical switch, and the measurements are acquired with a Keysight 34465A digital voltmeter. Oversampling and moving-average filtering on log-scale data are used to improve precision, and a shielded metal enclosure is used to mitigate noise (Kowalski, 2024).

The experimental procedure is straightforward in circuit topology but demanding in metrology. The capacitor is charged to $V_0 e^{-t/(RC)}$2, the switch is opened to start discharge through the selected resistor network, and $V_0 e^{-t/(RC)}$3 and/or the resistor voltage are recorded so that currents $V_0 e^{-t/(RC)}$4 can be inferred. The data are then presented on logarithmic axes and in dimensionless time $V_0 e^{-t/(RC)}$5, which makes departures from a single exponential visually explicit.

Resistance	Configuration	Textbook $V_0 e^{-t/(RC)}$6 and $V_0 e^{-t/(RC)}$7
$V_0 e^{-t/(RC)}$8	Single-resistor discharge	$V_0 e^{-t/(RC)}$9; $C = 2.789\ \text{nF}$0
$C = 2.789\ \text{nF}$1	Single-resistor discharge	$C = 2.789\ \text{nF}$2; $C = 2.789\ \text{nF}$3
$C = 2.789\ \text{nF}$4	Single-resistor discharge	$C = 2.789\ \text{nF}$5; $C = 2.789\ \text{nF}$6
$C = 2.789\ \text{nF}$7	Single-resistor discharge	$C = 2.789\ \text{nF}$8; $C = 2.789\ \text{nF}$9

The paper also discusses series-capacitor experiments using dielectric film capacitors, $1\%$0, with $1\%$1 and $1\%$2, in order to probe charge transfer between isolated conductors and reveal sign changes and minima in $1\%$3 (Kowalski, 2024).

3. Departure from the textbook RC law

The standard linear, time-invariant RC model predicts

$1\%$4

and

$1\%$5

On a logarithmic plot, $1\%$6 should therefore be a straight line with slope $1\%$7.

The observed vacuum-capacitor discharges do not follow that expectation across the full time range. In the reported single-resistor measurements, the earliest part tracks the textbook slope, but a slower tail emerges at late times. The result is a clear curvature on $1\%$8 plots, rather than a single straight line. The effect strengthens at late times and is reproducible across the resistor values studied (Kowalski, 2024).

The two-resistor plate-current measurement sharpens the anomaly. On a linear scale the two plate currents appear similar, but on a logarithmic scale they differ: one side’s current almost matches a single exponential, while the other exhibits a pronounced slow tail. This indicates unequal inflow and outflow currents over time and therefore transient net charge on the capacitor or wires. The series-capacitor experiments reinforce the same point from another angle: $1\%$9 can exhibit a dip, hence non-monotonic behavior, when switching changes the initial charge configuration. The paper interprets these phenomena as evidence for multi-process relaxation and changing current direction rather than a simple one-rate decay.

A further result is that geometry matters. When the data are replotted against $1/4\ \text{W}$0, some capacitor combinations overlap while others do not. This means that $1/4\ \text{W}$1 and $1/4\ \text{W}$2 are not sufficient to parameterize the slow tail; capacitor geometry becomes an additional variable for the late-time dynamics.

4. Two-rate modeling and the “sum of charges” interpretation

Several candidate model classes are considered, but the best-supported description for the reported dataset is a two-exponential decay. In general form,

$1/4\ \text{W}$3

For the vacuum-capacitor measurements, $1/4\ \text{W}$4 is reported to suffice: $1/4\ \text{W}$5 with $1/4\ \text{W}$6 and $1/4\ \text{W}$7. The text states that these fits “typically overlap the data within the line width” (Kowalski, 2024).

The physical picture advanced is a “sum of charges” model. Charges are separated conceptually into $1/4\ \text{W}$8, whose field is confined between the plates and which produce the usual RC decay term with rate $1/4\ \text{W}$9, and $R = 47\ \text{k}\Omega$0, consisting of surface charges on the outside of the capacitor and along the circuit wires. The latter are required to guide current through the resistors in a Maxwellian view of circuits and are inferred to relax on a different timescale, $R = 47\ \text{k}\Omega$1. Because $R = 47\ \text{k}\Omega$2 and $R = 47\ \text{k}\Omega$3 decay differently, the effective capacity to hold charge at a given voltage changes over time.

This leads to the dynamic-capacitance formulation

$R = 47\ \text{k}\Omega$4

with governing equation

$R = 47\ \text{k}\Omega$5

If $R = 47\ \text{k}\Omega$6 is constant, the single exponential is recovered. With the measured two-exponential $R = 47\ \text{k}\Omega$7, however, $R = 47\ \text{k}\Omega$8 evolves from $R = 47\ \text{k}\Omega$9 at early times to a much larger effective capacitance at late times. The paper presents this as a phenomenological description whose microscopic origin is attributed to surface-charge dynamics rather than dielectric soakage, since the dielectric is vacuum (Kowalski, 2024).

Other candidate forms are discussed but not adopted as necessary for this dataset. A stretched exponential can represent a broad distribution of timescales, and voltage-dependent leakage $182\ \text{k}\Omega$0 can generate nonlinear behavior, but the reported measurements do not require those forms. Power-law tails are not reported. Field emission, residual gas ionization, and surface conduction are described as plausible in vacuum hardware at high fields, yet the emphasis of the paper remains on geometry-dependent surface-charge redistribution and Maxwellian current guidance rather than Fowler–Nordheim or gas-ionization analysis.

5. Data analysis, scaling trends, and validity controls

The data-analysis workflow begins with the textbook fit: a single exponential fitted to $182\ \text{k}\Omega$1 with slope $182\ \text{k}\Omega$2. The full analysis then fits the two-exponential form $182\ \text{k}\Omega$3, with $182\ \text{k}\Omega$4 fixed by $182\ \text{k}\Omega$5 and $182\ \text{k}\Omega$6, $182\ \text{k}\Omega$7, and $182\ \text{k}\Omega$8 estimated from voltage-versus-time data. For the series-capacitor geometry, the reported model is

$182\ \text{k}\Omega$9

The extracted parameter trends are structured rather than arbitrary. The fast-process amplitude $337\ \text{k}\Omega$0 scales as $337\ \text{k}\Omega$1, as expected from Ohmic current scaling. The slow-process amplitude $337\ \text{k}\Omega$2 also scales roughly inversely with $337\ \text{k}\Omega$3, which the paper interprets as consistent with the idea that the slow process is linked to the same current-guiding surfaces but with different dynamics and geometry. The slow rate $337\ \text{k}\Omega$4 is distinct from $337\ \text{k}\Omega$5 and is far smaller than $337\ \text{k}\Omega$6; in the plotted normalization it is shown alongside $337\ \text{k}\Omega$7, indicating that the slow tail has a time constant tens to hundreds of times $337\ \text{k}\Omega$8 (Kowalski, 2024).

The validity controls receive unusual emphasis because the experiment is designed as a modeling-and-validation exercise. Switch bounce is handled with advanced triggering, and mechanical switch capacitance is considered. Probe impedance and frequency response are checked with dual-probe comparisons at the same node to diagnose loading and compensation issues. The digital voltmeter’s high input impedance is described as minimizing loading relative to $337\ \text{k}\Omega$9. Noise is mitigated with shielding in a grounded metal enclosure, oversampling, moving-average filtering on logarithmic data, and power-line period averaging for larger-capacitance runs. The paper also acknowledges resistor temperature coefficients under varying dissipation, capacitor mechanical constriction versus voltage, and ADC nonlinearity. None of these effects is reported to invalidate the central conclusions.

No numerical $637\ \text{k}\Omega$0, AIC, or BIC values are reported. Instead, the goodness-of-fit judgment is qualitative but specific: two-exponential fits match the data to within plotting resolution, whereas single exponentials leave systematic late-time residuals that appear as curvature on logarithmic plots.

6. Pedagogical structure, practical reproduction, and extensions

The experiment is explicitly presented as a laboratory exercise in constructing new knowledge. The step-by-step flow begins with building the vacuum-capacitor RC circuit in a shielded box and selecting $637\ \text{k}\Omega$1 to set $637\ \text{k}\Omega$2 in the approximate range $637\ \text{k}\Omega$3–$637\ \text{k}\Omega$4 for $637\ \text{k}\Omega$5. The capacitor is charged to about $637\ \text{k}\Omega$6, the switch is opened cleanly, $637\ \text{k}\Omega$7 is recorded at high sampling, and $637\ \text{k}\Omega$8 is computed from resistor voltage. Students then plot $637\ \text{k}\Omega$9 versus $R_1 = 337\ \text{k}\Omega$0 and versus $R_1 = 337\ \text{k}\Omega$1, observe the late-time tail, validate the result by repeated runs and loading checks, compare one- and two-exponential models, extract $R_1 = 337\ \text{k}\Omega$2, $R_1 = 337\ \text{k}\Omega$3, $R_1 = 337\ \text{k}\Omega$4, and $R_1 = 337\ \text{k}\Omega$5, and extend the analysis to the two-plate and series-capacitor configurations (Kowalski, 2024).

Several recurrent pitfalls are identified. Linear-scale plots can hide the deviations; the literature therefore insists on logarithmic presentation. Probe loading and compensation can mimic anomalies, so the exercise emphasizes DVM-versus-oscilloscope comparisons and dual-probe tests. A further misconception is to attribute the slow tail automatically to dielectric loss. The pedagogical response is to remind students that the dielectric is vacuum and to redirect attention to surface-charge dynamics and geometry.

The reported practical recommendations for reproduction are correspondingly concrete: use a sealed vacuum capacitor such as the Comet CFMN-2800BAC/8-DE-G, choose $R_1 = 337\ \text{k}\Omega$6 in the $R_1 = 337\ \text{k}\Omega$7–$R_1 = 337\ \text{k}\Omega$8 range, charge to about $R_1 = 337\ \text{k}\Omega$9 for signal-to-noise ratio, measure with a high-input-impedance DVM, and incorporate shielding, oversampling, moving averages, and power-line averaging as needed. A particularly informative extension is the two-resistor plate-current measurement with $V_0 e^{-t/(RC)}$00, which exposes unequal plate currents and strengthens the case for multi-process relaxation. Additional extensions include systematic variation of $V_0 e^{-t/(RC)}$01 and $V_0 e^{-t/(RC)}$02, comparison of different vacuum-capacitor geometries, altered initial conditions in the series-capacitor setup, and environmental or voltage changes to probe the sensitivity of the slow process (Kowalski, 2024).

7. Terminological ambiguity and relation to Knight-shift measurements

In the laboratory context summarized above, “Knight–Leveson experiment” designates the vacuum-capacitor RC discharge and its non-exponential relaxation. That usage should be distinguished from the far more established condensed-matter meaning of “Knight shift,” which refers to the field-induced change of nuclear resonance frequency caused by hyperfine coupling to conduction-electron spins. In metallic NMR, the standard decomposition is

$V_0 e^{-t/(RC)}$03

with $V_0 e^{-t/(RC)}$04 tracking the uniform electronic spin susceptibility at the nuclear site (Baek et al., 2012).

Knight-shift measurements are used to diagnose pairing symmetry in superconductors rather than discharge dynamics in circuits. In LiFeAs, for example, the reported $V_0 e^{-t/(RC)}$05As Knight shift remains at its normal-state value over a finite temperature interval after bulk superconductivity begins and then drops abruptly at a lower temperature, defining distinct $V_0 e^{-t/(RC)}$06 and $V_0 e^{-t/(RC)}$07 and suggesting an anomalous superconducting state with a triplet-like component in a narrow field-temperature window before a singlet-dominated state emerges at lower temperature (Baek et al., 2012). In strained $V_0 e^{-t/(RC)}$08, theory has been developed to explain a substantial drop in the Knight shift across $V_0 e^{-t/(RC)}$09, including a strain-induced rotation of the $V_0 e^{-t/(RC)}$10-vector that yields anisotropic Knight-shift reductions for in-plane fields (Lindquist et al., 2019).

This suggests a recurring source of confusion: the shared word “Knight” can invite conflation of two unrelated experimental traditions. The vacuum RC experiment concerns multi-timescale energy relaxation, charge redistribution, geometry dependence, and Maxwellian surface-charge guidance in real circuits. Knight-shift experiments concern spin susceptibility, hyperfine coupling, and superconducting order-parameter symmetry. The overlap is terminological rather than methodological (Kowalski, 2024).