Chu–Harrington Limit (CHL) Explained

• Chu–Harrington Limit is a theoretical framework that establishes lower bounds on the radiation quality factor Q and fractional bandwidth for electrically small, passive antennas.
• It employs canonical asymptotic formulas (e.g., Q ~ (ka)⁻³ + (ka)⁻¹) for dominant dipolar modes, linking the size of the enclosing sphere to performance limits.
• Recent analyses extend CHL to two-dimensional radiators, plasmonic dimers, and naturally resonant emitters while refining spherical transmission-line models and directivity tradeoffs.

Searching arXiv for recent and directly relevant papers on the Chu–Harrington limit and related updates. The Chu–Harrington limit (CHL) denotes a family of size-constrained bounds on the radiation quality factor $Q$ of passive radiators whose sources are confined within a smallest enclosing sphere of radius $a$, and, by extension, on the fractional bandwidth of electrically small single-resonant devices. In the classical spherical formulation, the exterior field is expanded in TE/TM spherical modes with free-space wavenumber $k=\omega/c$, and the bound is expressed in terms of the electrical size $ka$. For the dominant dipolar modes and $ka \ll 1$, the canonical forms are $$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$ and $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$. Recent work shows that this classical interpretation remains accurate in the electrically small regime, while later analyses refine its meaning for two-dimensional radiators, moderate $ka$, directivity-constrained radiation, singular plasmonic resonators, and naturally resonant emitters (Pfeiffer et al., 2024, Pascale et al., 2021, Latypov, 28 Apr 2026).

1. Classical definition and mathematical setting

Chu’s historical construction treats each spherical TE/TM mode of a source enclosed within $r \le a$ by means of an exact external ladder circuit, and defines modal $Q$ as the energy stored in reactive elements divided by radiated power. Harrington later extended the framework to directivity–bandwidth tradeoffs for bounded radiators. In the notation used in recent re-examinations, $a$0, $a$1 is the radius of the smallest sphere enclosing the radiator, $a$2 is the spherical harmonic order, and $a$3 is the directivity. Depending on convention, $a$4 is written either as $a$5 in a fields-based form or as $a$6 in a single-resonance convention (Pfeiffer et al., 2024).

The standard CHL assumptions are specific. The radiator is linear, passive, time-invariant, and typically reciprocal; all sources and materials are contained within a sphere of radius $a$7; the exterior region is homogeneous, isotropic, and lossless free space; and the fields outside the sphere are represented by outgoing spherical waves. Under those assumptions, the lower bound applies to the radiation-limited quality factor $a$8, not directly to input impedance, matching-network complexity, or peak cross-section. Collin and Rothschild re-derived the classical values via field integrals that subtract radiative energy, so the circuit-based and field-based formulations coincide in the electrically small regime (Pfeiffer et al., 2024).

2. Canonical small-antenna formulas and Harrington’s directivity criterion

For the lowest-order spherical modes, the widely cited electrically small asymptotics are

$a$9

For a combined self-resonant TE+TM dipole with equal powers and matched phases, the best order-wise scaling is commonly quoted as $k=\omega/c$0. In single-resonant high-$k=\omega/c$1 cases, the bandwidth proxy $k=\omega/c$2 is used, whereas for multi-resonant behavior or $k=\omega/c$3 that relation is not precise (Pfeiffer et al., 2024).

Harrington’s contribution addressed directivity under a bandwidth constraint. With the first $k=\omega/c$4 spherical harmonics, using both TE and TM families phased constructively, the maximum “normal directivity” is

$k=\omega/c$5

With the identification $k=\omega/c$6, this becomes

$k=\omega/c$7

Within Harrington’s assumptions, $k=\omega/c$8 implies “high $k=\omega/c$9.” The classical picture therefore couples enclosure size, modal content, and attainable bandwidth: strong directivity growth requires access to higher spherical orders, but those higher orders are accompanied by larger reactive-energy burdens in the electrically small regime (Pfeiffer et al., 2024).

3. Two-dimensional CHL-type bounds for translationally invariant radiators

For translationally invariant objects, the 2021 analysis of singular plasmonic resonators derives two-dimensional Chu-type bounds directly rather than importing the three-dimensional spherical result. In the quasi-static limit for electric-kind radiators, with longitudinal currents and zero curl, the radiation $ka$0 can be written as the functional

$ka$1

where $ka$2 is the cross-section boundary, $ka$3 is the surface charge density per unit length, $ka$4, and $ka$5 with $ka$6 the radius of the smallest enclosing circle (Pascale et al., 2021).

Minimizing the quasi-static functional subject to fixed dipole moments yields geometry-dependent lower bounds in terms of polarizabilities. For the electric kind,

$ka$7

where $ka$8 is the largest eigenvalue of the scaled two-dimensional electric polarizability tensor. For the magnetic kind,

$ka$9

where $ka \ll 1$0 is the scalar two-dimensional magnetic polarizability. For a circular cylinder, the electric and magnetic bounds coincide: $ka \ll 1$1 For a dimer of two circular cylinders in the touching limit, the minimum normalized value becomes

$ka \ll 1$2

These expressions make explicit that, in two dimensions, the minimum $ka \ll 1$3 scales as $ka \ll 1$4 rather than $ka \ll 1$5. The paper attributes that difference to the logarithmic Green’s function and the altered stored-energy–radiation balance in two-dimensional electrodynamics (Pascale et al., 2021).

4. Singular plasmonic dimers and the loss of the simple $ka \ll 1$6-bandwidth correspondence

The singular-plasmonics case study is a translationally invariant dimer of two identical circular cylinders of radius $ka \ll 1$7, separated by an edge-edge gap $ka \ll 1$8, and described by a Drude permittivity

$ka \ll 1$9

Under uniform plane-wave excitation polarized along the dimer axis, the electrostatic modal expansion is organized by eigenpermittivities

$$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$0

with

$$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$1

All eigenvalues accumulate at $$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$2. Under the Drude map, the bright-mode resonance frequencies are

$$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$3

As the gap shrinks, the bright resonances redshift and cluster ever more densely near $$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$4 (Pascale et al., 2021).

The $$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$5-th bright mode has normalized dipole moment

$$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$6

and radiation quality factor

$$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$7

In the touching limit, the dipole moment shrinks and

$$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$8

so the radiation $$Q_{\min}^{\mathrm{TM1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$9 diverges rather than decreases. With Drude loss, the dissipation quality factor is

$$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$0

The associated critical-coupling locus is

$$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$1

For fixed loss, increasing $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$2 or decreasing $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$3 drives modes from radiation-limited to dissipation-limited behavior (Pascale et al., 2021).

The central result is that there is no contradiction with CHL. When peaks are isolated and under-damped, the fractional bandwidth satisfies $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$4, and for low-order well-separated peaks the inverse FWHM agrees with the analytic $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$5. Near touching, however, the discrete spectrum turns into a dense continuum around the accumulation point. The broadband response over any finite interval then results from the superposition of many plasmon resonances, each individually obeying the two-dimensional Chu bound and in fact possessing large $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$6. A common misconception is that the broadband scattering of singular dimers implies an anomalously low single-mode $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$7; the paper shows instead that the link between overall bandwidth and any one resonance’s $$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$8 fails once the peaks cease to be well separated (Pascale et al., 2021).

5. Updated spherical theory beyond the classical Chu model

A 2024 reformulation reexamines spherical radiation through an equivalent one-dimensional transmission-line model that exactly models the fields. For each outward TM$$Q_{\min}^{\mathrm{TE1}} \gtrsim (ka)^{-3} + (ka)^{-1}$$9 mode, the radial problem is represented by a lossless tapered waveguide with propagation constant and characteristic impedance

$ka$0

This immediately yields a precise cutoff condition,

$ka$1

Below that radius, $ka$2 is imaginary and the mode is evanescent; above it, $ka$3 is real and the mode propagates. The same analysis introduces a position-dependent reflection coefficient $ka$4 from the outward wave impedance and the local characteristic impedance, and uses it to partition energy in each spherical shell into radiated and stored components (Pfeiffer et al., 2024).

The new energy partition is

$ka$5

Below cutoff, $ka$6 is purely imaginary, $ka$7, $ka$8, and all energy is stored. This leads to modal $ka$9 values that agree with Chu and Collin in the small-$r \le a$0 regime but deviate as $r \le a$1 approaches or exceeds the mode cutoff. For $r \le a$2, the updated small-$r \le a$3 expansion matches Chu to the first two leading orders, yet at $r \le a$4 the new $r \le a$5 is approximately $r \le a$6 smaller than Chu’s $r \le a$7 value (Pfeiffer et al., 2024).

The same framework revises Harrington’s directivity analysis. For equal TE/TM pairs of order $r \le a$8, with amplitudes $r \le a$9 phased to add in the main beam,

$Q$0

Lagrange-multiplier optimization at fixed $Q$1 yields

$Q$2

Unlike Harrington’s truncation at $Q$3, this optimum excites all spherical harmonics, including $Q$4, with smoothly decaying weights. Quantitatively, for $Q$5 the maximum directivity is within a factor $Q$6 of Harrington’s $Q$7 over $Q$8; for $Q$9 at $a$00, $a$01 dB, corresponding to $a$02 aperture efficiency; and for large $a$03 with $a$04, the maximum aperture efficiency remains above unity under the updated $a$05, whereas Chu-based modal $a$06 would predict a decay toward zero (Pfeiffer et al., 2024).

6. Scattering-channel bounds, superscattering, and what CHL does not bound

CHL is frequently conflated with peak-scattering bounds, but the two are distinct. For three-dimensional free-space scattering by a passive object, the optical theorem and S-matrix unitarity impose a per-partial-wave limit

$a$07

For the dipolar channel $a$08,

$a$09

This is a cap on peak scattering in one multipolar channel at one frequency. By contrast, CHL constrains minimum radiation $a$10 and therefore bandwidth. A radiator can exceed the single-channel dipolar peak by combining multiple multipolar channels constructively and still fully satisfy CHL (Mikhailovskaya et al., 2022).

The flat split-ring superscatterer demonstrates that distinction explicitly. The reported device is a two-dimensional array of near-field coupled split-ring resonators optimized over a 19-dimensional search space. A genetically optimized six-element configuration, operated in the $a$11–$a$12 GHz range, exhibits a measured peak total scattering cross-section of about $a$13 at the optimized orientation, exceeding the single-channel dipolar bound at the measured frequency by approximately $a$14. The enhancement is produced by engineered overlap of multiple resonances, including MD, ED, EQ, MQ, and octupolar terms. The paper proposes an empirical “super-radiant criterion”

$a$15

together with a practical device-specific form,

$a$16

and reports Monte Carlo evidence that randomly distributed resonant dipoles do not exceed that normalized ceiling (Mikhailovskaya et al., 2022).

The result is consistent with CHL for two separate reasons. First, CHL does not constrain peak $a$17 at a single frequency. Second, the aggregate array is not electrically small in Chu’s sense: with a practical circumscribing radius $a$18–$a$19 mm at $a$20 GHz, $a$21–$a$22, so the classical lower bound $a$23 is weak. Even for a single constituent CSRR, with $a$24 mm and $a$25 near $a$26 GHz, the simulated $a$27 remains comfortably above the Chu lower bound of approximately $a$28–$a$29. A common misconception is therefore that surpassing a single-channel scattering limit constitutes a CHL violation; the data instead show that scattering-channel bounds and CHL regulate different observables (Mikhailovskaya et al., 2022).

7. Naturally resonant emitters, efficiency bounds, and broader scope

The CHL also extends to naturally resonant electrically small emitters under its standard assumptions. In that setting, $a$30 is written as

$a$31

and near a single Lorentzian resonance the bandwidth quality factor satisfies

$a$32

For dipolar radiators,

$a$33

with the deep-subwavelength approximation $a$34. Since $a$35 and $a$36 for radiation efficiency $a$37, one obtains the efficiency bound

$a$38

In the same limit, the paper rewrites this as a bound on radiated power density per unit input power and unit volume,

$a$39

with $a$40 (Latypov, 28 Apr 2026).

This motivates a figure of merit

$a$41

for which $a$42 is the CHL-attainable ideal of a passive, single-resonance, lossless emitter. The same framework yields CHL-dictated constraints on atomic emitters. Using the transform-limited inequality

$a$43

with excited-state lifetime $a$44, CHL implies

$a$45

With the spontaneous-emission relation

$a$46

one obtains

$a$47

These results rest on the observation that natural resonance mechanisms change the internal realization of the source but not the exterior electromagnetic boundary-value problem on which CHL is based (Latypov, 28 Apr 2026).

The scope of the limit is therefore broad but not unlimited. It remains applicable to passive, linear, time-invariant free-space emitters bounded by a sphere, regardless of whether resonance is achieved by conventional matching, plasmonic localization, mechanical strain, or quantum transitions. It ceases to apply in its standard form when the assumptions themselves are altered: increasing the effective bounding sphere through arrays or distributed apertures, introducing active or non-Foster circuitry, using parametric or nonlinear time variance, or coupling strongly to non-free-space propagating channels. Across these cases, the consistent theme is that CHL fixes a minimum radiation $a$48 for the allowable external modal content; apparent exceptions arise when the relevant observable is not $a$49, when many resonances overlap, or when the enclosure and channel assumptions are changed rather than violated (Latypov, 28 Apr 2026).