Applications of Gauss's Law and Symmetry

Gauss's law is always true and almost never useful. That sounds like an insult, but it is the single most important thing to understand about it: the law $\oint\mathbf{E}\cdot\mathrm{d}\mathbf{a}=Q_{\mathrm{enc}}/\varepsilon_0$ relates one number (the flux) to one number (the enclosed charge), while the field it contains is a whole vector function of position. One equation cannot determine infinitely many unknowns — unless something else first collapses those unknowns down to a single number. That something is symmetry. This article is the payoff of the whole flux apparatus: the three — and, honestly, only three — charge geometries where symmetry does the collapsing. Spherical symmetry forces $\mathbf{E}$ to be radial and depend only on $r$ ; cylindrical symmetry forces it perpendicular to the axis, depending only on $s$ ; planar symmetry forces it perpendicular to the plane. In each case there is a Gaussian surface on which the flux integral degenerates into $E$ times an area, and a calculation that would take a page of Coulomb integrals — the field inside a charged ball, say — finishes in three lines. You also get results that are genuinely startling: a shell with zero field everywhere inside it, and an infinite sheet whose field never weakens with distance. Master the trio as patterns, not formulas: state the symmetry, choose the surface, evaluate the flux, count the charge. That four-step ritual, and the discipline to recognize when it does not apply, is what this article installs — and it is the fastest field-computing tool you will own until the potential arrives in the next few articles.

Key takeaways

Gauss's law (from Gauss's Law in Integral Form) is universally valid, but it yields $\mathbf{E}$ explicitly only when symmetry lets you pull $E$ out of the flux integral — one equation can only deliver one unknown.
Exactly three symmetry classes work: spherical (concentric Gaussian sphere), cylindrical (coaxial Gaussian cylinder), planar (straddling pillbox).
The results to know cold: outside any spherically symmetric $Q$ , $E=\frac{1}{4\pi\varepsilon_0}\frac{Q}{r^2}$ (point-charge equivalent); inside a uniform ball, $E\propto r$ ; inside a uniform shell, $E=0$ exactly; infinite line, $E=\frac{\lambda}{2\pi\varepsilon_0 s}$ ; infinite sheet, $E=\frac{\sigma}{2\varepsilon_0}$ , independent of distance.
The falloffs tell a story: $1/r^2$ , $1/s$ , constant — each dimension of charge extent you add removes one power of distance decay.
Symmetric pieces superpose: two oppositely charged sheets give $E=\sigma/\varepsilon_0$ between and zero outside — the parallel-plate capacitor in one line.
Key idea: symmetry tells you the direction of $\mathbf{E}$ and what it depends on; the right Gaussian surface then makes $\oint\mathbf{E}\cdot\mathrm{d}\mathbf{a}$ trivially equal to $E$ times an area, turning a hard integral into one-line algebra.

Where this lives today

The symmetry patterns of this article are live engineering formulas. Every coaxial cable spec sheet — the ones behind your building's internet — quotes a capacitance per meter and a characteristic impedance that are the cylindrical Gauss pattern with a dielectric constant; every parallel-plate structure, from MEMS accelerometers to the gate of a transistor, runs on the two-sheet superposition. High-voltage engineers size bushings and cable insulation by the $1/s$ field of the cylindrical pattern, because the field is fiercest at the inner conductor's surface, exactly where equation (3) says it must be. The sphere pattern is a workhorse of physics beyond wires. Nuclear physicists model a nucleus as a uniformly charged ball: its interior field grows linearly and its electrostatic self-energy scales as $Q^2/R$ , an input to the liquid-drop model that sets where fission becomes energetically favorable. Astrophysicists use the identical mathematics with gravity — the shell theorem is why a star's interior dynamics feel only the mass beneath — and in 2026 the same Gaussian-surface logic runs inside every finite-element field solver's validation suite: before trusting a mesh on an asymmetric geometry, you check it reproduces sphere, cylinder, and plane to machine precision, because those are the cases with exact answers. This trio is the eternal test bench.

The physical picture

Think of $\mathbf{E}$ as a flow of field lines. Gauss's law says the net number of lines piercing any closed surface counts the charge inside — that much is true for any surface around any charge. The magic happens when the charge is arranged so symmetrically that the lines have no choice but to distribute themselves evenly over some natural surface: then "lines per unit area" — the field strength — is the same everywhere on that surface, and counting becomes division. The whole method is a conspiracy between the geometry of the charge and the geometry of the surface you draw.

Figure

The symmetry trio — each geometry dictates its field pattern and its Gaussian surface

The symmetry argument that fixes the field's direction is just as important as the law itself, and it is a proof, not a plausibility. For a point charge, rotational symmetry forbids any tangential component — if $\mathbf{E}$ tilted, which way would it tilt? Every candidate direction is related to another by a rotation that leaves the charge alone, so the only survivor is radial. For an infinite plane the field cannot point sideways (no direction in the plane is preferred) and must mirror itself on crossing, so it points straight out on both sides. These "which way would it tilt?" arguments are doing the real work; Gauss's law merely cashes them in. Two consequences deserve early astonishment. The field inside a uniform spherical shell is exactly zero — any interior Gaussian sphere encloses nothing, and symmetry forbids the loophole of hidden canceling fields. This is the electrostatic twin of Newton's gravitational shell theorem, and both rest on the $1/r^2$ law. And an infinite sheet's field never weakens with distance: the lines leave perpendicular and simply have nowhere to spread. Far from being exotic, that is the working approximation inside every capacitor.

The three patterns, worked in full

Everything runs on the law established in Gauss's Law in Integral Form, quoted here once and then used relentlessly:

\oint_{\partial V} \mathbf{E}\cdot \mathrm{d}\mathbf{a} = \frac{Q_{\mathrm{enc}}}{\varepsilon_0} \qquad (1)

where: $\partial V$ — any closed surface you care to draw; $\mathrm{d}\mathbf{a}$ — its outward area element; $Q_{\mathrm{enc}}$ — the charge inside it; $\varepsilon_0 = 8.854\times10^{-12}\ \mathrm{F/m}$ . The law is indifferent to where the enclosed charge sits and to all the charge outside. Its usefulness hinges on one further step: if symmetry makes $\mathbf{E}$ perpendicular to the surface with constant magnitude on a piece of area $A$ (and tangent to the rest), the integral collapses,

\oint \mathbf{E}\cdot \mathrm{d}\mathbf{a} = E\,A \qquad (2)

where: $A$ — the area of the piece the field pierces squarely. That single factorization is the entire method; each geometry below is just a different choice of the surface that makes it legal. The results, before the honest work:

E_{\text{sphere,out}}=\frac{1}{4\pi\varepsilon_0}\frac{Q}{r^2}, \qquad E_{\text{line}}=\frac{\lambda}{2\pi\varepsilon_0 s}, \qquad E_{\text{sheet}}=\frac{\sigma}{2\varepsilon_0} \qquad (3)

where: $\lambda$ and $\sigma$ — charge per unit length and per unit area. Read the exponents as a story about spreading room: a point charge's lines dilute over spheres ( $1/r^2$ ), a line's over cylinders ( $1/s$ ), a plane's not at all. Each dimension of source extent removes one power of decay.

Derivation AField of a uniformly charged solid sphere inside and out

The goal is the complete field of a ball of radius $R$ carrying uniform density $\rho$ , total charge $Q=\tfrac{4}{3}\pi R^3\rho$ — the benchmark spherical-symmetry calculation, containing the shell theorem and the point-charge equivalence as corollaries. The strategy is the four-step ritual: argue the direction by symmetry, choose a concentric Gaussian sphere, factor the flux, count the enclosed charge — once with the sphere outside the ball, once inside. Watch where the two cases differ: it is only in $Q_{\mathrm{enc}}$ .

Step 1. Symmetry. Every rotation about the center maps the charge distribution to itself, so it must map $\mathbf{E}$ to itself too: the field can have no tangential component, and its magnitude can depend on nothing but $r$ . Hence $\mathbf{E}=E(r)\,\hat{\mathbf{r}}$ . Why this is airtight: a tangential component at any point would single out a direction on the sphere through that point, and rotating about the axis through the point and the center would produce a different field from the same charge — contradiction.

Step 2. Choose the surface and factor the flux. On a concentric Gaussian sphere of radius $r$ , the field is everywhere parallel to $\mathrm{d}\mathbf{a}$ with the same magnitude $E(r)$ , so equation (2) applies with the full sphere area:

\oint\mathbf{E}\cdot\mathrm{d}\mathbf{a} = E(r)\,4\pi r^2

Step 3. Outside ( $r\ge R$ ): the surface swallows the whole ball, $Q_{\mathrm{enc}}=Q$ . Gauss's law gives $E\,4\pi r^2=Q/\varepsilon_0$ , so

E(r)=\frac{1}{4\pi\varepsilon_0}\frac{Q}{r^2} \qquad (r\ge R)

The exterior field is identical to that of a point charge $Q$ at the center — the distribution's radius has vanished from the answer. Why this matters beyond the sphere: the argument used only spherical symmetry, never uniformity, so any onion-layered $\rho(r)$ — a planet, a nucleus, an ion cloud — looks from outside like a point holding its total charge.

Step 4. Inside ( $r<R$ ): the Gaussian sphere encloses only the charge within radius $r$ , namely $Q_{\mathrm{enc}}=\rho\cdot\tfrac{4}{3}\pi r^3 = Q\,r^3/R^3$ . Then $E\,4\pi r^2 = Qr^3/(\varepsilon_0 R^3)$ , so

E(r)=\frac{1}{4\pi\varepsilon_0}\frac{Q\,r}{R^3} = \frac{\rho\,r}{3\varepsilon_0} \qquad (r\le R)

The field grows linearly from zero at the center: deeper in, less charge is underneath you. Why the charge outside radius $r$ contributes nothing: it forms a nest of uniform shells, and each shell's interior field vanishes — the shell theorem, which is Step 4 applied to a shell where $Q_{\mathrm{enc}}=0$ .

Step 5. Stitch and check. At $r=R$ both branches give $E=\tfrac{1}{4\pi\varepsilon_0}Q/R^2$ : continuous, as it must be, since there is no surface charge layer here (a genuine $\sigma$ would make $E$ jump — that story is owned by Boundary Conditions on E and V). The interior result also follows from direct integration of Coulomb's law over shells, and both branches meet at $r=R$ without a jump.

Figure

The payoff profile — E of a uniform ball rises linearly then falls as one over r squared

Go deeper · Derivation B · 1 minField of an infinite line charge

Derivation BField of an infinite line charge

The goal is the field at a perpendicular distance $s$ from an infinite straight line carrying uniform charge $\lambda$ per unit length — the cylindrical-symmetry pattern, and the engine behind every coaxial-cable formula. The strategy: symmetry kills all but the radial component, a coaxial Gaussian cylinder factors the flux through its curved wall while its flat caps contribute nothing, and the arbitrary cylinder length must cancel — watching it cancel is the built-in sanity check.

Step 1. Symmetry. Sliding the infinite line along itself and rotating about it leave the charge unchanged, so neither an axial nor an azimuthal field component can survive, and the magnitude can depend only on the perpendicular distance: $\mathbf{E}=E(s)\,\hat{\mathbf{s}}$ . Why the axial component dies specifically: reflection through any plane perpendicular to the line maps the charge to itself but reverses an axial component, forcing it to equal its own negative.

Step 2. Choose a coaxial cylinder of radius $s$ and length $L$ . On the curved wall, $\mathbf{E}$ is parallel to $\mathrm{d}\mathbf{a}$ with constant magnitude, contributing $E(s)\cdot 2\pi s L$ . On the two end caps, $\mathbf{E}$ lies in the cap plane — perpendicular to $\mathrm{d}\mathbf{a}$ — contributing zero:

\oint\mathbf{E}\cdot\mathrm{d}\mathbf{a} = E(s)\,(2\pi s L) + 0

Step 3. Enclosed charge: the cylinder threads a length $L$ of line, so $Q_{\mathrm{enc}}=\lambda L$ .

Step 4. Solve. $E(s)\,2\pi sL = \lambda L/\varepsilon_0$ , and $L$ cancels:

E(s) = \frac{\lambda}{2\pi\varepsilon_0\,s}

Why the cancellation had to happen: an infinite line has no feature to set a scale along its length, so no physical answer could depend on the bookkeeping length $L$ ; if it fails to cancel, the Gaussian surface was drawn wrong. The $1/s$ falloff is slower than a point charge's $1/r^2$ — a line's field lines spread over cylinders, whose area grows only as the first power of distance. The same result follows by integrating Coulomb's law along the line, and the $1/s$ dependence is the $L\to\infty$ limit of the finite-segment answer.

Go deeper · Derivation C · 1 minField of an infinite plane of charge

Derivation CField of an infinite plane of charge

The goal is the field of an infinite plane sheet carrying uniform surface charge $\sigma$ — the planar pattern, whose punchline (a field that ignores distance) powers the parallel-plate capacitor. The strategy: symmetry makes the field purely normal with mirror-image values on the two sides, and a pillbox straddling the sheet catches flux through both faces — the factor of 2 that students lose is manufactured honestly here.

Step 1. Symmetry. No direction within the plane is preferred, so $\mathbf{E}$ has no sideways component; reflection through the sheet swaps the two sides, so the field points away from the sheet with equal magnitude $E$ at equal distances on either side (for $\sigma>0$ ). Translation parallel to the plane forbids any dependence on the sideways coordinates.

Step 2. Choose a pillbox of face area $A$ , faces parallel to the sheet at equal heights above and below. Flux: each face is pierced squarely by outgoing field, contributing $EA$ apiece; the side wall is skimmed tangentially, contributing zero:

\oint\mathbf{E}\cdot\mathrm{d}\mathbf{a} = EA + EA = 2EA

Why both faces count positively: on the top face the field and the outward normal both point up; on the bottom face both point down. Outward flux is positive on each — nothing cancels.

Step 3. Enclosed charge: the pillbox clips a disk of sheet of area $A$ , so $Q_{\mathrm{enc}}=\sigma A$ .

Step 4. Solve. $2EA=\sigma A/\varepsilon_0$ , the area cancels, and

E = \frac{\sigma}{2\varepsilon_0}

independent of the distance to the sheet. Why no weakening with distance: the lines leave perpendicular and parallel to each other, and parallel lines never dilute — there is no extra area for them to spread into, unlike the sphere's $4\pi r^2$ or the cylinder's $2\pi s L$ .

Step 5. Superpose two sheets, $+\sigma$ and $-\sigma$ , a distance $d$ apart. Between them the two contributions point the same way and add to $\sigma/\varepsilon_0$ ; outside they oppose and cancel to zero. That is the parallel-plate capacitor's field, obtained without a single integral. The two-sheet superposition reproduces the capacitor field $\sigma/\varepsilon_0$ and, with $V=Ed$ , the standard $C=\varepsilon_0 A/d$ .

Which pattern answers which problem

you face	symmetry	Gaussian surface	result
point charge, charged shell, ball, onion $\rho(r)$	spherical	concentric sphere	$E=\frac{1}{4\pi\varepsilon_0}\frac{Q_{\mathrm{enc}}(r)}{r^2}$
infinite line, wire, long cylinder, coax	cylindrical	coaxial cylinder	$E=\frac{\lambda_{\mathrm{enc}}}{2\pi\varepsilon_0 s}$
infinite sheet, slab, capacitor plate	planar	straddling pillbox	$E=\frac{\sigma}{2\varepsilon_0}$ per sheet
finite rod, disk, ring, cube, blob	none usable	—	Coulomb integral or the potential (The Electrostatic Potential)
conductor surfaces, charged interfaces	local pillbox	see Boundary Conditions on E and V	normal $E$ jumps by $\sigma/\varepsilon_0$

Limiting cases

Case	Result	Meaning
far from any localized charge	$E \to \frac{1}{4\pi\varepsilon_0}\frac{Q}{r^2}$	everything looks like a point charge from far away; the sphere result is exact at all $r$ , not just far
inside a uniform shell	$E=0$ everywhere	shell theorem: interior Gaussian sphere encloses nothing
center of a uniform ball	$E=\frac{\rho r}{3\varepsilon_0}\to 0$	no charge beneath you, no field; grows linearly outward
infinite sheet, any distance	$E=\frac{\sigma}{2\varepsilon_0}$	parallel lines never dilute; valid near real plates until edge distances compare with plate size
line charge, $s\to\infty$ vs point	$1/s$ vs $1/r^2$	extended sources decay more slowly; each added source dimension removes one power
two parallel sheets $\pm\sigma$	$\sigma/\varepsilon_0$ inside, $0$ outside	superposition of Gauss-solved pieces — the capacitor field

Common trapGauss's law is always true, so I can always use it to find E

The most common misuse of the law in all of electrostatics. Equation (1) holds for a cube of charge, a finite rod, a charged coffee mug — and is useless for all of them, because no Gaussian surface exists on which their $E$ has constant magnitude and known direction. The law then gives one true equation containing an unknown function under an integral, and one equation cannot deliver a function. Validity is not invertibility: you may only solve for $E$ after a symmetry argument has reduced the unknowns to a single number on your surface. No symmetry, no factorization — fall back on Coulomb superposition or the potential. And beware the classic exam trap: a finite rod or disk has edges, its field tilts, and the infinite-line and infinite-sheet formulas apply only in the close-up regime where the edges are effectively at infinity.

Worked examples

1 · A charged sphere, three radii (confidence builder)

A solid sphere of radius $R=10\ \mathrm{cm}$ carries $Q=50\ \mathrm{nC}$ spread uniformly through its volume. Find $E$ at $r=5$ cm, at the surface, and at $r=20$ cm. ( $k=1/4\pi\varepsilon_0=8.99\times10^9\ \mathrm{N\,m^2/C^2}$ .)

Step 1. Constant first: $kQ = 8.99\times10^9 \times 50\times10^{-9} = 449.5\ \mathrm{V\,m}$ .

Step 2. Surface, $r=R=0.10$ m — the peak of the profile figure: $E=kQ/R^2 = 449.5/0.010 = \boxed{4.5\times10^4\ \mathrm{V/m}}$ .

Step 3. Inside, $r=0.05$ m — use the linear branch $E=kQr/R^3 = 449.5\times0.05/0.0010 = \boxed{2.2\times10^4\ \mathrm{V/m}}$ . Exactly half the surface value, as $E\propto r$ demands at half the radius.

Step 4. Outside, $r=0.20$ m — point-charge equivalent: $E=kQ/r^2 = 449.5/0.040 = \boxed{1.1\times10^4\ \mathrm{V/m}}$ . Doubling the distance from the surface value's radius quartered the field: $4.5\times10^4/4 = 1.1\times10^4$ ✓.

2 · Designing a coaxial cable (exam level)

A coaxial cable has an inner conductor of radius $a=1.0\ \mathrm{mm}$ and an outer shield of inner radius $b=5.0\ \mathrm{mm}$ , vacuum between. The inner conductor carries $\lambda=+20\ \mathrm{nC/m}$ , the shield $-20\ \mathrm{nC/m}$ . Find (a) $E$ at $s=2.0$ mm, (b) the potential difference between the conductors, (c) the capacitance per unit length. (d) What is $E$ outside the shield?

Step 1. Between the conductors, a coaxial Gaussian cylinder of radius $s$ encloses only the inner line: the line-charge pattern applies with $\lambda_{\mathrm{enc}}=+\lambda$ . The prefactor: $\lambda/2\pi\varepsilon_0 = 2k\lambda = 2\times8.99\times10^9\times20\times10^{-9} = 359.6\ \mathrm{V}$ .

Step 2. (a) $E(2.0\ \mathrm{mm}) = 359.6/0.0020 = \boxed{1.8\times10^5\ \mathrm{V/m}}$ — a strong field from a modest charge, which is why cable insulation ratings matter.

Step 3. (b) Integrate the field from $a$ to $b$ (anticipating The Electrostatic Potential): $V = \int_a^b E\,\mathrm{d}s = \frac{\lambda}{2\pi\varepsilon_0}\ln\frac{b}{a} = 359.6\times\ln 5 = 359.6\times1.609 = \boxed{579\ \mathrm{V}}$ .

Step 4. (c) $C/L = \lambda/V = 20\times10^{-9}/578.8 = \boxed{35\ \mathrm{pF/m}}$ — equivalently $2\pi\varepsilon_0/\ln(b/a) = 5.563\times10^{-11}/1.609 = 3.46\times10^{-11}\ \mathrm{F/m}$ ✓. (Real RG-58 reads ~100 pF/m: a polyethylene dielectric and a tighter geometry raise it — the physics is this formula times $\varepsilon_r$ .)

Step 5. (d) Outside the shield a Gaussian cylinder encloses $+\lambda-\lambda=0$ , so $E=\boxed{0}$ : the cable is electrostatically invisible, which is the point of the shield — and a superposition-of-patterns argument identical in spirit to the two-sheet capacitor.

Origins · 1 mincarl friedrich gauss and how this was discovered

carl friedrich gauss

1777–1855 · Göttingen, Kingdom of Hanover

Gauss was already a legend — the greatest mathematician of his age, author of the Disquisitiones Arithmeticae at 24, calculator of Ceres' orbit from a handful of observations — before he touched electromagnetism seriously. In the 1830s, past fifty, he built a magnetic observatory at Göttingen with the young Wilhelm Weber, organized the world's first international scientific data-collection network (the Magnetischer Verein, mapping Earth's magnetic field), and put electricity and magnetism on an absolute system of units for the first time. The divergence theorem that underwrites his law he had proven in celestial-mechanics work decades earlier; the flux-charge relation itself he set down in an 1835 manuscript that, characteristically, he never bothered to publish — it surfaced only in his collected works in 1867, by which time Maxwell had already enshrined it as the first of the field equations. His personal motto, pauca sed matura ("few, but ripe"), explains both the delay and why everything bearing his name still holds.

History

1785

— Coulomb establishes the inverse-square force law with his torsion balance, the experimental root of everything Gauss's law encodes.

1813

— Gauss proves versions of the divergence theorem in work on the attraction of ellipsoids, forging the flux tool his law will need.

1835

— Gauss writes down the flux–charge relation in an unpublished manuscript (printed posthumously in 1867).

1843

— Faraday's ice-pail experiment demonstrates charge enclosure and induction directly — Gauss's law as tabletop fact.

1865

— Maxwell places the law, in differential form, as the first of his field equations, where it remains.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §2.2.3 (Examples 2.1, 2.3, 2.5) · Purcell & Morin, Electricity and Magnetism, 3rd ed., ch. 1–2 · Jackson, Classical Electrodynamics, 3rd ed., §1.3–1.6 · Zangwill, Modern Electrodynamics, ch. 3 · Feynman Lectures, Vol. II, Ch. 5 Each boxed result follows from evaluating $\oint\mathbf{E}\cdot\mathrm{d}\mathbf{a}$ on the stated Gaussian surface and equating to $Q_{\mathrm{enc}}/\varepsilon_0$ ; continuity at the boundary of the solid sphere and the $1/L$ cancellation for the line charge serve as built-in consistency checks. The numerical constant $\varepsilon_0=8.854\times10^{-12}\,\mathrm{F/m}$ is CODATA-standard. Capacitance formulas $C=\varepsilon_0 A/d$ and $C/L=2\pi\varepsilon_0/\ln(b/a)$ are standard results. Worked numbers: $kQ=449.5$ V·m gives $4.5\times10^4$ , $2.2\times10^4$ , $1.1\times10^4$ V/m at $r=R$ , $R/2$ , $2R$ ; coax with $2k\lambda=359.6$ V gives $1.8\times10^5$ V/m at 2 mm, $V=359.6\ln5=579$ V, $C/L=34.6$ pF/m.

Builds on: Gauss's Law in Integral Form · Electric Flux · The Electric Field · Continuous Charge Distributions · Superposition Principle for Electric Fields Leads to: Gauss's Law in Differential Form · The Electrostatic Potential · Poisson's Equation · The Curl of the Electrostatic Field

How do I know which Gaussian surface to pick?

You do not pick it — the symmetry picks it, and if no symmetry is present there is nothing to pick. The surface must be one on which every patch has $\mathbf{E}$ either squarely perpendicular (constant magnitude) or exactly tangent (zero flux). Spherical charge → concentric sphere; infinitely long cylindrical charge → coaxial cylinder; infinite planar charge → straddling pillbox. That is the complete menu. If you find yourself inventing a creative surface for, say, a charged cube, stop: no surface exists on which the cube's field is uniform, and the cleverness is better spent setting up a Coulomb integral or a numerical solution.

Why is the field inside a uniform shell exactly zero rather than just small?

Two ingredients conspire. Symmetry first: inside the shell the field must still be radial, $E(r)\hat{\mathbf{r}}$ , by the same rotation argument as everywhere else. Gauss second: a concentric sphere of any interior radius encloses zero charge, so $E(r)\cdot4\pi r^2=0$ , forcing $E(r)=0$ at every interior point. Geometrically, near-wall charge pulls harder but there is less of it in view; far-wall charge is weaker but more plentiful — and for exactly a $1/r^2$ law the two effects cancel to all orders. This is also a precision test of Coulomb's law: laboratory searches for fields inside charged shells bound deviations from the inverse square at the $10^{-16}$ level.

The infinite sheet's field never decreases with distance — how can that be physical?

For a truly infinite sheet, walk away and the geometry looks unchanged — there is no scale to measure your distance against, so nothing for $E$ to depend on. The field lines leave perpendicular, mutually parallel, and parallel lines never dilute. The idealization earns its keep as a near-field approximation: at distances small compared with a real plate's size, the edges are effectively at infinity and $E\approx\sigma/2\varepsilon_0$ holds beautifully — that is precisely the regime inside a capacitor. Back off to distances large compared with the plate, and the same plate resolves into a blob of total charge $Q$ with the point-charge $1/r^2$ tail. The two regimes stitch smoothly through the exact (and messier) finite-disk answer.

When exactly does the enclosed charge change the answer, given that E in Gauss's law is made by all charges?

Keep the two roles separate. The $\mathbf{E}$ appearing in the flux integral is the total field, sourced by everything inside and out. But the net flux through a closed surface sees only $Q_{\mathrm{enc}}$ : outside charge sends field lines in one side and out the other, contributing zero net. So in the coax example, the field between the conductors is determined by the enclosed $+\lambda$ only — not because the shield's charge makes no field of its own, but because a cylindrically symmetric shell of charge contributes nothing inside itself (shell theorem, cylindrical edition). Symmetry plus enclosure is what lets you forget the outside world legally.

Why is the field just outside a conductor sigma over epsilon-zero instead of sigma over two epsilon-zero?

Both formulas are correct — they answer different questions. An isolated sheet of charge sends flux out of both faces, so each side carries $\sigma/2\varepsilon_0$ . At a conductor's surface, the material beneath enforces $\mathbf{E}=0$ inside, so a pillbox catches flux through its outer face only: $EA=\sigma A/\varepsilon_0$ , giving $\sigma/\varepsilon_0$ . Physically, the rest of the conductor's surface charge supplies an extra $\sigma/2\varepsilon_0$ that cancels the interior field and doubles the exterior one. The clean local statement — the normal component of $\mathbf{E}$ always jumps by $\sigma/\varepsilon_0$ across any charged surface — is developed in Boundary Conditions on E and V, and the conductor story in Conductors in Electrostatic Equilibrium.

Can I use these formulas for a finite rod or a finite disk if I am close enough?

Yes — that is their honest experimental meaning, and the error is controllable. Close to the middle of a rod of length $L$ (at $s\ll L$ ), the far ends subtend tiny angles and contribute negligibly: the field is the infinite-line $\lambda/2\pi\varepsilon_0 s$ plus corrections of order $(s/L)^2$ . Same for a disk versus the infinite sheet. The formulas fail near the edges and at distances comparable to the object's size, where the field acquires components the symmetry argument forbade for the infinite version. Rule of thumb: infinite-source formulas are near-field approximations, point-charge formulas are far-field approximations, and the crossover region belongs to honest integration.

Is the shell theorem the same one as in gravity?

Mathematically identical — both forces are inverse-square, and Gauss's law is a statement about inverse-square flux, indifferent to what is doing the attracting. Newton needed his shell theorem (proved geometrically in the Principia, with considerable effort) to justify treating planets as points; electrostatics inherits it in three lines of flux algebra, which is a fair measure of how much the field-and-flux language buys. One physical difference: mass is all one sign, so gravitational "conductors" and screening do not exist, whereas mobile charges of both signs let conductors expel fields entirely. Inside a uniform shell you feel no gravity from the shell and no electric field from its charge — for exactly the same $1/r^2$ reason.