Guided Waves and Transmission Lines

Free space is a fine medium for waves and a terrible one for infrastructure: an antenna's power dilutes as $1/r^2$ , and a private message broadcast to the hemisphere is neither private nor efficient. The alternative — piping electromagnetic waves along metal rails — is as old as the telegraph and as current as the coaxial jumper on a quantum computer, and it changes the physics in one decisive way. Confinement quantizes. Force a wave to respect boundary conditions on conducting walls and the free continuum of plane waves collapses to a discrete menu of modes, each with its own transverse field pattern and its own rulebook for propagation. This article builds the menu. The construction is a single clean maneuver: assume the guided form $e^{i(kz - \omega t)}$ along the axis, and Maxwell's equations reorganize themselves — all four transverse field components become slaves to the two longitudinal ones, $E_z$ and $B_z$ , which alone satisfy a two-dimensional eigenvalue problem. The menu then writes itself: TE modes ( $E_z = 0$ ), TM modes ( $B_z = 0$ ), and the aristocrat, TEM ( $E_z = B_z = 0$ ) — a mode whose cross-section is literally a frozen electrostatic field sliding down the line at the speed of light in the filling, with no cutoff, no dispersion, and a theorem attached: it exists only between two or more conductors. A hollow pipe cannot carry it, which is why hollow waveguides refuse everything below a cutoff frequency, and why the coaxial cable — two conductors, TEM, DC to gigahertz — wired the twentieth century. The TEM story culminates in the telegrapher's equations and the characteristic impedance $Z_0 = \sqrt{L'/C'}$ , derived here from the fields with the lovely theorem $L'C' = \mu\varepsilon$ that makes every TEM line, whatever its shape, a light-speed delay line. The rectangular guide's explicit modes and the cutoff physics get their own articles next.

Key takeaways

Guided-wave ansatz: $\mathbf{E}, \mathbf{B} \propto e^{i(kz-\omega t)}$ in a uniform structure along $z$ . Maxwell then slaves all transverse components to $(E_z, B_z)$ , which obey the 2D Helmholtz eigenproblem $(\nabla_\perp^2 + \gamma^2)\{E_z, B_z\} = 0$ with $\gamma^2 = \omega^2/c^2 - k^2$ .
The mode taxonomy: TE ( $E_z = 0$ , wall condition $\partial_n B_z = 0$ ), TM ( $B_z = 0$ , wall condition $E_z = 0$ ), TEM ( $E_z = B_z = 0$ , forcing $\gamma = 0$ , $k = \omega/c$ ).
Walls quantize $\gamma$ into a discrete ladder; each TE/TM mode inherits a cutoff $\omega_c = c\gamma$ — real propagation above, evanescence below. The ladder's explicit values (rectangular guide) and the dispersion physics belong to the next two articles, by name.
The no-TEM theorem: a TEM cross-section is a 2D electrostatics problem, and a single closed conductor is one equipotential — uniqueness forces zero field. TEM needs two conductors: coax, twin-lead, microstrip.
TEM lines are circuit-izable: the fields deliver the telegrapher's equations, wave speed $v = 1/\sqrt{L'C'}$ with the geometry-independent theorem $L'C' = \mu\varepsilon$ , and characteristic impedance $Z_0 = \sqrt{L'/C'}$ (coax: $Z_0 = \tfrac{60\ \Omega}{\sqrt{\varepsilon_r}}\ln\tfrac{b}{a}$ — 50 Ω is a geometry).
KEY IDEA: one axial ansatz reduces Maxwell to a transverse eigenvalue problem whose boundary conditions quantize the modes — TE and TM inherit cutoffs from their nonzero transverse eigenvalue, while TEM, the two-conductor zero mode, propagates dispersionlessly at the medium's light speed and behaves as a circuit with impedance $\sqrt{L'/C'}$ .

The core insight

Separate the field into a longitudinal factor $e^{i(kz-\omega t)}$ and a transverse boundary-value problem: the wall conditions quantize the transverse eigenvalue $\gamma$ , giving TE and TM modes cutoff frequencies $\omega_c = c\gamma$ , while the $\gamma = 0$ TEM mode — possible only with two or more conductors — propagates without cutoff or dispersion and reduces to the telegrapher's equations with $Z_0 = \sqrt{L'/C'}$ .

Where this lives today

The TEM circuit picture is the operating system of high-speed electronics. Every trace on a modern motherboard is designed as a transmission line — impedance-controlled to ±10%, length-matched so parallel bits arrive together (at 5 ns/m equivalent, millimeters matter at gigabit rates), terminated to kill reflections; signal-integrity engineering is the telegrapher's equations applied at industrial scale. The same discipline runs the world's instrumentation (50 Ω end to end), cable broadband (75 Ω, DOCSIS pushing 10 Gb/s through 1960s-topology coax), and the cryogenic wiring of quantum computers, where dozens of coaxial TEM lines per qubit chip carry microwave pulses to millikelvin stages — with the added constraint that each line's thermal conduction budget is as engineered as its $Z_0$ . Hollow-guide physics holds the high ground where power and loss rule: radar transmitters, satellite feeds, particle-accelerator RF distribution (kilometers of pressurized rectangular guide at megawatt levels), and the 5G/6G push into millimeter waves has revived waveguide-on-chip in the form of substrate-integrated waveguides. The taxonomy itself keeps paying dividends in new substrates: silicon photonics guides light in TE/TM-like modes of dielectric ridges (the no-TEM theorem does not bind dielectric guides — no conductors — but the same longitudinal-slaving reduction organizes their modes), and the "N conductors, N−1 TEM modes" counting reappears verbatim in the multi-conductor buses and differential pairs of every serial link standard.

Origins · 1 minOliver Heaviside and how this was discovered

Oliver Heaviside

1850–1925 · Camden Town and Torquay

Heaviside — telegrapher by trade, self-taught after leaving school at sixteen, nephew-in-law of Wheatstone — is the reason this article's circuit half exists in usable form. In the 1870s–80s, working alone and unpaid in a room his family worried about, he recast Maxwell's twenty quaternion-flavored equations into the four vector equations everyone now learns (with Gibbs, independently), invented the operational calculus that engineers still use as Laplace-transform shorthand, and derived the telegrapher's equations in their modern form — including the distortionless line condition that showed, against the massed opinion of the cable industry, that adding inductance would cure long-distance telephone distortion. The industry's chief engineer (W. H. Preece) had staked his authority on the opposite; Heaviside's loading-coil insight was patented by others (Pupin, profitably) while he lived on charity and grew eccentric — signing letters "W.O.R.M.", painting his nails, furnishing his home with granite blocks. He coined "impedance," "inductance," "conductance," "permeability"; predicted with Kennelly the ionospheric layer that bears his name (confirmed 1924, the year before his death); and wrote the three-volume Electromagnetic Theory whose blunt aphorisms ("Shall I refuse my dinner because I do not fully understand the process of digestion?") still circulate. The 50-ohm world runs on the mathematics of a man the establishment declined to employ.

History

1855

— Kelvin's telegraph theory: the RC (diffusive) cable equation for the Atlantic cable project

1876

— Heaviside adds inductance to the cable analysis: the full telegrapher's equations

1887

— Heaviside's distortionless-line condition; the loading-coil insight the industry rejects, then adopts

1893

— Thomson (J.J.) and Rayleigh discuss waves in hollow tubes; Rayleigh proves the mode structure and cutoffs in 1897

1933

— Southworth at Bell Labs transmits signals through hollow waveguide, reviving Rayleigh's "curiosity"

1940

s — Radar war-effort industrializes rectangular guide; the WR standard series and 50-ohm coax conventions solidify

1965

— Wheeler's microstrip analysis brings TEM-line design to printed circuit boards

2020

s — Impedance-controlled TEM lines wire cryogenic quantum processors and 100-gigabit serial links alike

The physical picture

Why does confinement create a menu? Because walls impose standing-wave conditions transversely while leaving the axis free. A guided mode is a drumhead pattern (transverse) married to a traveling wave (axial): the drumhead's eigenvalue problem — Laplacian plus boundary conditions, exactly the machinery the E2 separation-of-variables articles built — has discrete solutions, and each solution is a mode. The marriage contract is the constraint $\gamma^2 + k^2 = \omega^2/c^2$ : the total "wavenumber budget" $\omega/c$ is split between transverse structure ( $\gamma$ , fixed by geometry) and axial travel ( $k$ , whatever is left). A mode with a big transverse appetite leaves little for travel; when the frequency drops to $\omega_c = c\gamma$ the budget just covers the drumhead and nothing remains for propagation — the cutoff, whose full physics (evanescence, dispersion, the strange velocities) is the chapter-closer's subject. The same budget explains the zigzag picture worth carrying everywhere: a TE or TM mode is a plane wave bouncing between the walls at an angle, its transverse standing structure built from the up- and down-going components, its axial phase pattern from their common drift. High frequency: shallow bounce angle, nearly free flight. Approaching cutoff: the bounce steepens toward perpendicular — the wave rattles across the guide, going nowhere. (The cutoff-and-dispersion article makes this quantitative; here it earns the taxonomy its intuition.) And the exception that runs the world: TEM. Set both longitudinal components to zero and the constraint collapses to $\gamma = 0$ , $k = \omega/c$ exactly — no transverse budget spent at all. The price is structural: the transverse field must then be simultaneously curl-free and divergence-free in the cross-section — a two-dimensional electrostatic field — and electrostatics in a region bounded by a single equipotential is identically zero (Derivation B, leaning on the E2 uniqueness theorem by name). Give it two conductors, though, and every electrostatic two-conductor field you have ever solved — coax, parallel plates, twin lead — becomes a wave, frozen in cross-section, sliding at the speed of light in the dielectric, at every frequency down to DC. That mode is why cable television, Ethernet-over-twisted-pair, and the probes on every oscilloscope exist; its circuit avatar — voltage and current waves governed by the telegrapher's equations — is Derivation C.

Uniform structure along $z$ (any cross-section — the power of this section is its geometry-blindness), fields $\mathbf{E} = \mathbf{E}_0(x,y)\,e^{i(kz - \omega t)}$ and likewise $\mathbf{B}$ . Each Cartesian component satisfies the vacuum wave equation (the chapter opener's machinery), which the ansatz converts to the transverse Helmholtz problem

\left(\nabla_\perp^2 + \gamma^2\right)\begin{pmatrix}E_z \\ B_z\end{pmatrix} = 0, \qquad \gamma^2 \equiv \frac{\omega^2}{c^2} - k^2, \qquad (1)

where: $\nabla_\perp^2$ is the cross-sectional Laplacian and $\gamma$ — the transverse eigenvalue — is the quantity the walls will quantize. Read (1) as the wavenumber budget: $\omega^2/c^2$ total, $\gamma^2$ spent on transverse structure, $k^2$ left for travel. The remarkable economy (Derivation A): the four transverse components need not be solved for at all —

E_x = \frac{i}{\gamma^2}\left(k\,\partial_x E_z + \omega\,\partial_y B_z\right), \qquad E_y = \frac{i}{\gamma^2}\left(k\,\partial_y E_z - \omega\,\partial_x B_z\right), \qquad (2)

where: (with the two magnetic twins) the entire transverse field is algebraically slaved to the longitudinal pair. Two scalar functions instead of six: the ansatz has done to waveguides what the potential formulation will do to radiation. The $1/\gamma^2$ is a loaded gun — the moment $\gamma = 0$ , (2) becomes $0/0$ and the TEM case secedes into its own (richer) law. The taxonomy follows from which longitudinal component survives: TE ( $E_z = 0$ ; drive everything from $B_z$ , wall condition $\partial_n B_z = 0$ ), TM ( $B_z = 0$ ; drive from $E_z$ , wall condition $E_z = 0$ — both conditions descend from the PEC package of the electrodynamic-boundary-conditions article, recalled by name), and TEM (both zero). For TE and TM the walls make (1) a genuine eigenvalue problem with a discrete positive spectrum $\gamma_1 < \gamma_2 < \cdots$ , and each mode inherits

k(\omega) = \frac{1}{c}\sqrt{\omega^2 - \omega_c^2}, \qquad \omega_c = c\,\gamma \qquad \text{(one cutoff per mode)}, \qquad (3)

where: propagation is real above $\omega_c$ and the mode is evanescent below — the structural origin of the hollow guide's high-pass character. This article stops at the structure deliberately: the explicit eigenvalue ladder of the rectangular guide is the next article's boundary-value solve (Rectangular Waveguides), and everything (3) implies — the strange velocities, $v_pv_g = c^2$ , below-cutoff physics — is the chapter closer's (Cutoff Frequencies and Dispersion in Waveguides), both by name. For the TEM mode, $\gamma = 0$ forces $k = \omega/c$ — vacuum dispersion, no cutoff — and Derivations B and C settle who may have it and what it costs:

\frac{\partial V}{\partial z} = -L'\frac{\partial I}{\partial t}, \qquad \frac{\partial I}{\partial z} = -C'\frac{\partial V}{\partial t}, \qquad v = \frac{1}{\sqrt{L'C'}} = \frac{c}{\sqrt{\varepsilon_r}}, \qquad Z_0 = \sqrt{\frac{L'}{C'}}, \qquad (4)

where: $V$ and $I$ are honest voltages and currents (uniquely defined only because the TEM cross-section is electrostatic — path-independence of $\int\mathbf{E}\cdot d\boldsymbol{\ell}$ in the cross-sectional plane), $L'$ and $C'$ are the static per-length inductance and capacitance of the two-conductor geometry, and the wave speed is geometry-independent by the theorem $L'C' = \mu\varepsilon$ (Derivation C's centerpiece). The characteristic impedance — the $V/I$ ratio a traveling wave maintains — is the one number a cable presents to the circuits at its ends; for coax,

Z_0^{\rm coax} = \frac{1}{2\pi}\sqrt{\frac{\mu}{\varepsilon}}\,\ln\frac{b}{a} = \frac{59.96\ \Omega}{\sqrt{\varepsilon_r}}\,\ln\frac{b}{a}, \qquad (5)

where: $a, b$ are inner and outer radii: impedance is a geometry. Fifty ohms is $b/a = 3.49$ in polyethylene ( $\varepsilon_r = 2.25$ ); 75 Ω (video) is $b/a = 6.53$ . Worked example 1 builds a real cable from (5) and checks the theorem's promises against the datasheet.

Figure

The TEM line as circuit — the frozen coax cross-section slides at the dielectric light speed, and the telegrapher ladder is its exact shadow

Derivation AThe reduction — six components become two

The goal is the master structure: equations (1)–(2) and the taxonomy they force, valid for any cylindrical geometry. The strategy is to feed the axial ansatz to the two curl equations and watch the transverse components become algebra. This derivation is the article's spine — everything after it is special cases — so the bookkeeping is shown honestly for one representative pair.

Step 1. The ansatz kills two derivatives. With $\partial_z \to ik$ and $\partial_t \to -i\omega$ , Faraday's and Ampère–Maxwell's curls become six algebraic-in-z,t component equations. Write the two $x$ - $y$ pairs that involve the transverse components; from Faraday's $x$ -component and Ampère's $y$ -component:

\partial_y E_z - ik\,E_y = i\omega B_x, \qquad \partial_y B_z - ik\,B_y = -\frac{i\omega}{c^2}E_x \quad(\text{plus two mirror equations}).

Step 2. Solve the linear system. Four equations, four transverse unknowns $(E_x, E_y, B_x, B_y)$ , with $\partial_x E_z$ , $\partial_y E_z$ , $\partial_x B_z$ , $\partial_y B_z$ as sources. Eliminating pairwise (multiply one equation by $k$ , its partner by $\omega/c^2$ ... the algebra is mechanical) yields (2) and its magnetic twins, all carrying the factor $i/\gamma^2$ with $\gamma^2 = \omega^2/c^2 - k^2$ . Why slavery was inevitable: the ansatz leaves only two "dynamical" directions (the transverse plane), and Maxwell's curls are first-order — once the longitudinal profiles are chosen, the transverse fields are determined the way $\mathbf{B}$ was determined from $\mathbf{E}$ in the completion routine. The waveguide problem is the completion routine with boundary conditions.

Step 3. Close on the longitudinal pair. Substituting the slaved components back into the remaining Maxwell equations (or, faster: each Cartesian component of both fields satisfies the wave equation, and the ansatz splits $\nabla^2 = \nabla_\perp^2 + \partial_z^2$ ) gives (1): two decoupled scalar Helmholtz problems for $E_z$ and $B_z$ . Decoupled in the bulk — the walls couple them in general geometry, but for the high-symmetry cross-sections of practice (rectangle, circle) the boundary conditions split cleanly and modes are purely TE or TM. Why the wall conditions differ by family: the PEC package requires tangential E and normal B to vanish; $E_z$ is always tangential to the wall, giving TM its Dirichlet condition $E_z = 0$ ; for TE the requirement on the slaved transverse fields translates, via (2), into the Neumann condition $\partial_n B_z = 0$ — the next article audits this translation explicitly for the rectangle.

Step 4. Read the taxonomy from the singular case. For TE/TM the eigenproblem (1) with nontrivial boundary conditions has positive discrete spectrum ( $\gamma^2 > 0$ — a Sturm–Liouville fact: the drumhead has no nontrivial zero-frequency pattern), hence real cutoffs by (3). The residual case $\gamma^2 = 0$ makes (2) singular: transverse fields can no longer be slaved and must be solved directly — with both longitudinal components zero, that direct problem is the subject of Derivation B. The taxonomy is thus exhaustive by construction: every guided solution has $E_z, B_z$ solving (1), and the trichotomy TE/TM/TEM merely records which of them vanish. One subtlety worth flagging: the Neumann constant eigenfunction — a uniform $B_z$ — is excluded from the TE spectrum because it produces zero transverse field through (2), yielding no physical mode. This is the "no TE $_{00}$ " fact the next article inherits from the Sturm–Liouville positivity of the drumhead problem.

Go deeper · Derivation B · 2 minThe no-TEM theorem — why a hollow pipe has a cutoff at all

Derivation BThe no-TEM theorem — why a hollow pipe has a cutoff at all

The goal is the taxonomy's most consequential theorem: a TEM mode ( $E_z = B_z = 0$ ) cannot exist inside a single hollow conductor, so everything a hollow pipe carries pays a cutoff. The strategy: show the TEM cross-section is a two-dimensional electrostatics problem and let the E2 uniqueness theorem finish it. Three moves, each recalling machinery this track already owns — the theorem is a harvest, not a new planting.

Step 1. TEM forces vacuum dispersion. With $E_z = B_z = 0$ , equations (2) read $0/\gamma^2$ : for nonzero transverse fields the only escape is $\gamma^2 = 0$ , i.e. $k = \omega/c$ exactly. (Physically: nothing transverse is "standing," so the whole budget travels.)

Step 2. The cross-section is electrostatic. Write Faraday's $z$ -component for the ansatz: $(\nabla_\perp\times\mathbf{E}_\perp)_z = i\omega B_z = 0$ — the transverse field is curl-free in the cross-sectional plane. Gauss in the charge-free interior with $\partial_z E_z = 0$ : $\nabla_\perp\cdot\mathbf{E}_\perp = 0$ — divergence-free too. Curl-free allows a potential, $\mathbf{E}_\perp = -\nabla_\perp\phi$ (the E0 scalar-potentials theorem, recalled by name); divergence-free makes it harmonic:

\nabla_\perp^2\phi = 0 \quad\text{in the cross-section}.

At each instant, each cross-sectional slice of a TEM wave is a solution of the two-dimensional Laplace equation — the frozen-electrostatics slogan, now literal.

Step 3. Uniqueness pulls the trigger. On a perfect conductor tangential E vanishes, so every conductor boundary is an equipotential of $\phi$ . A single hollow conductor makes the entire boundary of the (simply connected) interior one equipotential, and the E2 uniqueness theorem — a harmonic function is determined by its boundary values, recalled by name from Uniqueness Theorems — forces $\phi =$ const throughout: $\mathbf{E}_\perp = 0$ . No TEM mode. Why two conductors repair it: with an inner and an outer conductor the boundary has two components that may sit at different potentials, the interior is no longer simply bounded by one equipotential, and every two-conductor electrostatic solution — coax's radial field, the parallel-plate field, twin lead's dipole-like pattern — promotes directly to a TEM wave. The mode count follows the conductor count: N conductors support N−1 independent TEM modes (one potential difference each).

Step 4. The consequence worth italicizing: since TE and TM modes all carry $\gamma > 0$ (Derivation A, Step 4), a hollow pipe propagates nothing below its lowest cutoff $c\gamma_1$ . A waveguide is a high-pass filter by theorem, not by engineering choice — DC power and low-frequency signals categorically require a second conductor. This single result explains the coexistence of two transmission technologies: hollow guides where power handling and low loss at microwave frequencies dominate, and two-conductor lines wherever bandwidth must reach down toward DC.

Go deeper · Derivation C · 2 minTEM lines as circuits — the telegrapher's equations, the LC theorem, and the coax worked end to end

Derivation CTEM lines as circuits — the telegrapher's equations, the LC theorem, and the coax worked end to end

The goal is the bridge from fields to engineering: derive the telegrapher's equations (4) from the TEM field structure, prove the geometry-independent theorem $L'C' = \mu\varepsilon$ , and cash both on the coaxial cable — arriving at the $Z_0$ formula (5) that RF engineering treats as an axiom. The strategy: because the TEM cross-section is electrostatic (Derivation B), voltage and current are legitimate wave variables — a fact that fails for TE/TM modes and is the entire reason circuit theory survives at radio frequency.

Step 1. Legitimate $V$ and $I$ . In each cross-sectional plane, $\mathbf{E}_\perp$ is curl-free, so $V(z,t) = -\int_1^2 \mathbf{E}_\perp\cdot d\boldsymbol{\ell}$ (inner to outer conductor) is path-independent within the plane — an honest voltage. Ampère around the inner conductor in the plane defines $I(z,t)$ unambiguously. Why this fails for TE/TM: with longitudinal fields present, the in-plane loop integral of E is fed by $\partial_t B_z$ — "voltage" becomes path-dependent, and hollow-guide practice speaks of mode impedances defined by convention instead. TEM's circuit language is a theorem, not an approximation.

Step 2. Faraday delivers the first telegrapher equation. Take a thin rectangular loop of axial extent $dz$ spanning the two conductors: the EMF around it is $V(z+dz) - V(z)$ plus end contributions, and the flux through it is (flux per length) $\times dz = L'I\,dz$ by the definition of the static inductance per length. Faraday:

\frac{\partial V}{\partial z} = -L'\,\frac{\partial I}{\partial t}.

Step 3. Charge conservation delivers the second. The conductors carry charge per length $\lambda = C'V$ (static capacitance definition — legal because the cross-section is the static problem). Continuity on a slice $dz$ of the inner conductor: $\partial_z I = -\partial_t\lambda$ , giving $\partial_z I = -C'\,\partial_t V$ . Cross-differentiate the pair: both $V$ and $I$ obey the 1D wave equation with speed $v = 1/\sqrt{L'C'}$ , and the traveling-wave ratio is $V/I = \sqrt{L'/C'} \equiv Z_0$ .

Step 4. The $L'C' = \mu\varepsilon$ theorem. For any two-conductor geometry: the static capacitance solves $\nabla_\perp^2\phi = 0$ with the conductor boundary values, and the static inductance solves the same boundary-value problem for the stream function of $\mathbf{B}_\perp$ — the two 2D problems are conjugate harmonic partners, and their product of "geometry factors" cancels identically:

L'C' = \mu\varepsilon \qquad\Longrightarrow\qquad v = \frac{1}{\sqrt{\mu\varepsilon}} = \frac{c}{\sqrt{\varepsilon_r}},

independent of shape. Why this had to happen: Derivation B showed the TEM wave is the free wave of the filling medium wearing an electrostatic cross-section; its speed was never the geometry's to choose. The geometry keeps exactly one degree of freedom — the ratio $L'/C'$ , i.e. $Z_0$ . One neat corollary: measure a cable's delay (5.0 ns/m below) and you have measured its dielectric constant.

Step 5. Coax, end to end. Static results from the E-track, recalled: $C' = 2\pi\varepsilon/\ln(b/a)$ (cylindrical capacitor) and $L' = (\mu/2\pi)\ln(b/a)$ (coax flux per current). Product: $\mu\varepsilon$ ✓ (theorem audited). Ratio:

Z_0 = \sqrt{\frac{L'}{C'}} = \frac{1}{2\pi}\sqrt{\frac{\mu}{\varepsilon}}\ln\frac{b}{a} = \frac{59.96\ \Omega}{\sqrt{\varepsilon_r}}\,\ln\frac{b}{a},

which is (5). RG-58 ( $a = 0.45$ mm, $b = 1.48$ mm, polyethylene $\varepsilon_r = 2.25$ ): $L' = 238$ nH/m, $C' = 105$ pF/m, $Z_0 = 47$ Ω, $v = 2.0\times10^8$ m/s = 0.67c, delay 5.0 ns/m — the datasheet, reproduced from two electrostatics formulas. Why 50 Ω rules the world: for air-filled coax, minimum loss sits near 77 Ω and maximum power-handling near 30 Ω; 50 Ω is the historical compromise (and 75 Ω, near the loss optimum, went to video where power never mattered).

Limiting cases

Case	Result	Meaning
$\gamma = 0$ , two conductors	TEM: $k = \omega/c$ , no cutoff	frozen electrostatics at light speed; DC welcome
$\gamma = 0$ , one hollow conductor	$\mathbf{E}_\perp \equiv 0$	no-TEM theorem: hollow pipes are high-pass
TE/TM, $\omega \gg \omega_c$	$k \to \omega/c$	zigzag flattens; mode forgets its walls
TE/TM, $\omega \to \omega_c^+$	$k \to 0$	budget exhausted transversely — closer article's story
$\omega < \omega_c$	$k$ imaginary	evanescent; attenuators — closer article's story
$N$ conductors	$N - 1$ TEM modes	coax 1, twin lead 1, three-wire 2...
filled line, any shape	$v = c/\sqrt{\varepsilon_r}$	$L'C' = \mu\varepsilon$ : geometry sets only $Z_0$
matched load $Z_L = Z_0$	no reflected wave	the line looks infinite; maximum power transfer

Common trapVoltage on a waveguide is a convention; voltage on a TEM line is physics — and impedance is not resistance

Two conflations cost real debugging hours. First: the clean circuit variables of (4) exist because the TEM cross-section is electrostatic; in a hollow guide, "voltage" depends on the path you integrate along (Faraday feeds the loop with $\partial_t B_z$ ), and the several "waveguide impedances" in the literature are useful conventions that disagree with each other by geometry factors. Quoting a hollow guide's impedance without naming the convention is a specification bug. Second: $Z_0 = \sqrt{L'/C'}$ is real for a lossless line — 50 Ω — but it is not a resistor: no power is dissipated in it. It is the ratio a traveling wave maintains between $V$ and $I$ ; an infinite (or matched) cable draws current like a 50 Ω resistor forever, yet every joule is in flight down the line, not heating anything. The distinction becomes operational at every connector: a load equal to $Z_0$ absorbs the wave (no echo), while open or short ends reflect totally — with opposite signs — which is why unterminated high-speed lines ring and why worked example 2's fault-finder works. And keep the jurisdiction fence: everything circuit-like here is TEM-only; the moment frequency climbs enough to excite the line's first non-TEM mode (coax too has one, near where the mean circumference matches a wavelength), the pure circuit description quietly expires.

Worked examples

1 · Building a 50-ohm cable from statics (confidence builder)

Design a polyethylene-filled coax ( $\varepsilon_r = 2.25$ ) with $Z_0 = 50\ \Omega$ around an inner conductor of radius $a = 0.45$ mm, and predict its full datasheet: $L'$ , $C'$ , wave speed, and delay per meter.

Step 1. Geometry from (5): $\ln(b/a) = Z_0\sqrt{\varepsilon_r}/59.96 = 50\times1.5/59.96 = 1.251$ , so $b/a = e^{1.251} = \mathbf{3.49}$ : outer radius $b = 1.57$ mm... with the standard rounding of real RG-58 stock, $b \approx 1.5$ mm. (For 75 Ω video cable the same formula demands $b/a = 6.53$ — fatter jacket for the same core, which you can verify by eye against any TV cable.)

Step 2. The per-length constants, from the recalled statics: $L' = (\mu_0/2\pi)\ln(b/a) = 2\times10^{-7}\times1.251 = \mathbf{250\ nH/m}$ ; $C' = 2\pi\varepsilon_0\varepsilon_r/\ln(b/a) = \mathbf{100\ pF/m}$ .

Step 3. Audit the theorem: $L'C' = 250\times10^{-9}\times100\times10^{-12} = 2.50\times10^{-17}\ \text{s}^2/\text{m}^2 = \mu_0\varepsilon_0\varepsilon_r$ ✓. Speed: $v = 1/\sqrt{L'C'} = \mathbf{2.0\times10^8\ m/s}$ — the "66% velocity factor" on the datasheet is $1/\sqrt{2.25}$ , pure dielectric constant, no geometry.

Step 4. Delay: $1/v = \mathbf{5.0\ ns/m}$ . Every RF bench memorizes this number — it is why a 20 m cable run costs 100 ns of latency, why oscilloscope deskewing exists, and why example 2's fault-finder can read distance off a clock.

2 · Time-domain reflectometry: finding the break with a pulse (exam level)

A technician connects a fast pulse generator and oscilloscope to one end of a long RG-58 run buried in a wall. The launched step reflects, returning inverted after 187 ns. (a) Where is the fault and what kind is it? (b) What would an open (unplugged) end have shown? (c) Why must the generator itself present 50 Ω?

Step 1. Distance: the echo's round trip took 187 ns at $v = 2.0\times10^8$ m/s, so the fault sits at $z = \tfrac{1}{2}\,v\,t = \tfrac{1}{2}\times2.0\times10^8\times187\times10^{-9} = \mathbf{18.7\ m}$ down the cable.

Step 2. Type: reflection inverted means the reflection coefficient $\Gamma = (Z_L - Z_0)/(Z_L + Z_0)$ is negative — the termination impedance is below 50 Ω, and a full-amplitude inversion means $Z_L \approx 0$ : a short circuit (crushed cable, pinched shield). The sign logic is the string-on-a-wall analogy of the normal-incidence article, recalled: heavy wall (low impedance for the voltage wave) flips the pulse.

Step 3. An open end gives $Z_L \to \infty$ , $\Gamma = +1$ : an upright echo at the same delay. Between the extremes, a partial mismatch returns a scaled echo — $\Gamma = \pm 1/3$ for a 25 Ω or 100 Ω splice — and a matched 50 Ω termination returns nothing at all: the line looks infinite. TDR instruments are exactly this experiment industrialized: every impedance blemish along a cable (kink, water ingress, cheap connector) paints its signed echo at its distance, a sonar built from the telegrapher's equations.

Step 4. The generator's 50 Ω output matters because the echo, arriving back at the source, sees the generator as its termination. Matched source: the echo is absorbed on arrival, one clean edge per feature. Mismatched source: the echo re-reflects and rattles — multiple round trips, a staircase of ghosts, and an unreadable trace. Impedance discipline at both ends is why every instrument input, cable, and connector in an RF lab agrees on one number; the alternative is a corridor of funhouse mirrors.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §9.5 (Guided Waves) · Jackson, Classical Electrodynamics, 3rd ed., Ch. 8 (Waveguides, Resonant Cavities, and Optical Fibers) · Zangwill, Modern Electrodynamics, Ch. 19 (Waveguides and Transmission Lines) · Pozar, Microwave Engineering, 4th ed., Ch. 2-3 This article owns the framework (reduction, taxonomy, no-TEM theorem) and the complete TEM/transmission-line story (telegrapher's equations from fields, the $L'C' = \mu\varepsilon$ theorem, coax end to end). The rectangular guide's explicit boundary-value problem — including the rectangular cutoff $f_{10} = c/2a$ , the WR-90 numeric example ( $\approx 6.56$ GHz), and the $a = 2$ cm example ( $7.5$ GHz) — is treated in Rectangular Waveguides; the dispersion and velocity package — including $v_{\text{ph}}v_{\text{gr}} = c^2$ and evanescence below cutoff — is treated in Cutoff Frequencies and Dispersion in Waveguides. Key numbers used in this article: 50 Ω PE coax $b/a = 3.49$ (75 Ω: $b/a = 6.53$ ); RG-58 $L' = 238$ nH/m, $C' = 105$ pF/m, $Z_0 = 47$ Ω, $v = 2.0\times10^8$ m/s (0.67c), delay 5.0 ns/m; $L'C' = 2.50\times10^{-17}\ \text{s}^2/\text{m}^2 = \mu_0\varepsilon_0\varepsilon_r$ ✓; TDR example: 187 ns → 18.7 m.

Builds on: The Electromagnetic Wave Equation · Monochromatic Plane Waves in Vacuum · Complex Notation for Electromagnetic Waves · Reflection and Transmission at Oblique Incidence · Energy Density and the Poynting Vector

Leads to: Rectangular Waveguides · Cutoff Frequencies and Dispersion in Waveguides · Intensity and Time-Averaged Quantities · Momentum and Radiation Pressure

What does this article own versus the two waveguide articles after it?

This article is the framework: the axial ansatz, the reduction of Maxwell to the transverse Helmholtz problem with everything slaved to $(E_z, B_z)$ , the TE/TM/TEM taxonomy with its wall conditions, the no-TEM theorem, and the complete TEM/transmission-line story (telegrapher's equations, $L'C' = \mu\varepsilon$ , characteristic impedance, coax). Rectangular Waveguides takes the framework into one geometry and solves it: the TE $_{mn}$ /TM $_{mn}$ eigenfunctions, the explicit cutoff ladder, the TE₁₀ field map and single-mode band. Cutoff Frequencies and Dispersion in Waveguides then spends the eigenvalues: the hyperbolic dispersion relation, phase and group velocities with $v_pv_g = c^2$ , evanescence below cutoff, and pulse propagation in guides. Rule of thumb: "why modes exist and what TEM is" — here; "what the modes of a real box look like" — next; "how fast and how far" — the closer.

Why can circuit concepts like voltage and impedance be trusted on a TEM line at gigahertz frequencies when circuit theory supposedly fails at high frequency?

Because the TEM mode is precisely the loophole. Circuit theory's high-frequency failure has a specific cause: when fields have longitudinal structure, $\oint\mathbf{E}\cdot d\boldsymbol{\ell}$ around an in-plane loop is fed by changing flux and "voltage" becomes path-dependent. Derivation B showed the TEM cross-section is curl-free in the plane — so the potential difference between the two conductors is well-defined in every slice, the charge and current per length obey static relationships, and the exact Maxwell dynamics compresses without loss into the telegrapher's equations. Circuit theory doesn't merely approximate a TEM line; it is exactly equivalent to Maxwell for that mode (lossless, single-mode regime). The failure returns exactly when the premise does: above the frequency where the line's first non-TEM mode can propagate (for RG-58, tens of GHz, where the circumference approaches a wavelength), or at discontinuities (connectors, bends) whose local fringing fields are not TEM — which board designers model as lumped parasitics bolted onto the transmission-line backbone.

Do dielectric guides like optical fibers obey the no-TEM theorem — they have no conductors at all?

The theorem's premise fails — no conducting boundaries — so its conclusion doesn't bind, but the conclusion survives anyway in modified form: an optical fiber also has no true TEM mode, for a different reason. Dielectric guiding confines by total internal reflection (the Brewster/TIR article's physics), and the confined solutions of the same longitudinal-slaving reduction turn out to be TE, TM, and — new in dielectric guides — hybrid modes (HE/EH) with both $E_z$ and $B_z$ nonzero, because the dielectric boundary conditions couple the two Helmholtz problems that metal walls kept separate. The fiber's fundamental HE₁₁ mode is TEM-like (nearly transverse, no cutoff — it propagates at any frequency, though ever more weakly confined), which is why single-mode fiber exists without a second conductor. The framework of Derivation A carries over intact; only the boundary conditions — and hence the taxonomy's fine print — change. Griffiths stops at metal walls; Jackson Ch. 8 carries the dielectric case.

Where does the energy actually flow on a TEM line — in the wires or between them?

Between them, overwhelmingly. The Poynting vector $\mathbf{E}\times\mathbf{H}$ of the TEM mode lives in the dielectric: for coax, E is radial and H azimuthal, so $\mathbf{S}$ points axially in the insulation, and integrating it over the annulus reproduces exactly $P = VI$ . The conductors carry current but almost no energy flux — inside an ideal conductor $\mathbf{E} = 0$ , hence $\mathbf{S} = 0$ ; in a real one, a tiny radial Poynting component dips into the metal to pay the $I^2R$ toll (the static-fields Poynting article's wire audit, and the skin-depth article's surface-resistance accounting, both recalled by name — this is the same physics at the third telling). The practical morals: the dielectric's quality (loss tangent) matters as much as the copper's, since the power lives in it; "the signal" can be legitimately pictured as the field configuration gliding between the rails at $c/\sqrt{\varepsilon_r}$ while electrons merely shuffle; and a cable's power rating is set by dielectric breakdown and dielectric heating, not by anything resembling electron traffic congestion.

Why is a matched load reflectionless — what physically absorbs the wave?

Match the boundary condition and there is nothing left over to reflect. A traveling wave on the line arrives at the load maintaining $V/I = Z_0$ ; a load resistor $R = Z_0$ demands exactly that same ratio. The incident wave alone satisfies the load's constraint perfectly, so the "solve for the reflected wave that patches the mismatch" step — the Fresnel logic of this chapter, in one dimension — returns zero. Physically, the resistor dissipates power at exactly the rate the wave delivers it: $P = V^2/Z_0$ , converting the guided energy to heat with no residue. Any other $R$ under- or over-draws current relative to the wave's ratio, and the discrepancy propagates back as the echo $\Gamma = (R - Z_0)/(R + Z_0)$ — the same mismatch algebra as the normal-incidence article's $(n_1 - n_2)/(n_1 + n_2)$ , because it is the same algebra: impedances junctioned in one dimension. This is also why "50 Ω" pervades labs: it is not that 50 Ω is electrically magical, but that agreement is — any single standard makes every cable-instrument junction reflectionless.

What limits how much power a transmission line or waveguide can carry?

Three ceilings, met in order depending on the hardware. (1) Dielectric breakdown: the peak E-field anywhere must stay below the insulation's breakdown strength (air ~3 MV/m, polyethylene ~20 MV/m); in coax the maximum sits at the inner conductor surface, which is why high-power coax is fat and sometimes pressurized with insulating gas. Hollow guides, with no inner conductor and larger cross-sections, hold the high-power crown — megawatt radar pulses travel in pressurized rectangular guide. (2) Thermal: average power deposits $I^2R_s$ losses in the walls (skin-depth article's accounting) and loss-tangent heating in the dielectric; cooling sets the continuous rating, usually far below the breakdown-limited pulse rating. (3) Multipaction and corona in vacuum/space hardware: secondary-electron resonances at microwave frequencies can ignite discharges well below DC breakdown thresholds — a dominant constraint in satellite feeds. The design pattern is always the same: compute the field map (this chapter), find the hot spot, and derate — one more place where the "frozen electrostatics" picture of TEM pays: the breakdown analysis of a cable is its electrostatic problem.