Cylindrical Polar Coordinates

Give a problem an axis of symmetry — a charged wire, a coaxial cable, a pipe, a waveguide, a spinning fluid, a drumhead — and Cartesian coordinates fight you at every step, because the natural boundaries are circles and the natural fields depend only on how far you stand from the axis. Cylindrical polar coordinates $(\rho,\phi,z)$ are the cure: keep the Cartesian height $z$ , but replace the $(x,y)$ plane with plane polar coordinates, so a point is fixed by its perpendicular distance $\rho$ from the axis, the azimuthal bearing $\phi$ around it, and its altitude $z$ . It is the simplest curvilinear system beyond Cartesian, and the one you reach for first whenever geometry has a preferred direction. The price of that convenience is a single, far-reaching subtlety: the coordinate basis is no longer nailed to space. Walk around the axis and the radial and azimuthal unit vectors swing to stay perpendicular to the coordinate surfaces, so $\hat{\mathbf{e}}_\rho$ and $\hat{\mathbf{e}}_\phi$ have nonzero derivatives, and a correct gradient, divergence, or curl must account for that rotation. The bookkeeping is captured entirely by the orthogonal scale factors $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ : that lone nontrivial factor $h_\phi=\rho$ is the source of every $1/\rho$ prefactor, every $\partial_\rho(\rho\,\cdot)$ grouping, and the $\rho^{-2}\partial_\phi^2$ term in the Laplacian. The payoff is dramatic. Under axial symmetry the operators collapse to one-dimensional radial expressions, Laplace's and Helmholtz's equations separate cleanly, the natural volume and surface elements make Gauss's and Stokes' theorems almost trivial, and the radial equation that survives is Bessel's — whose regular solution, selected by finiteness on the axis, is the cylinder function governing everything from waveguide modes to a struck drum.

Key takeaways

Cylindrical polar coordinates $(\rho,\phi,z)$ fix a point by perpendicular distance $\rho\ge 0$ from a chosen axis, azimuth $0\le\phi<2\pi$ around it, and height $z$ along it.
The system is orthogonal with scale factors $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ , giving line element $\mathrm{d}s^2=\mathrm{d}\rho^2+\rho^2\mathrm{d}\phi^2+\mathrm{d}z^2$ and volume element $\rho\,\mathrm{d}\rho\,\mathrm{d}\phi\,\mathrm{d}z$ .
The unit vectors rotate with position: $\partial_\phi\hat{\mathbf{e}}_\rho=\hat{\mathbf{e}}_\phi$ and $\partial_\phi\hat{\mathbf{e}}_\phi=-\hat{\mathbf{e}}_\rho$ , which is what forces the extra terms in every operator.
The single factor $\rho$ propagates everywhere: divergence carries $\tfrac1\rho\partial_\rho(\rho A_\rho)$ , curl carries $\tfrac1\rho\partial_\rho(\rho A_\phi)$ , and the Laplacian carries $\tfrac1\rho\partial_\rho(\rho\,\partial_\rho f)+\tfrac1{\rho^2}\partial_\phi^2 f$ .
Under axial symmetry the operators become radial ODEs, the $\rho=0$ axis is a coordinate (not physical) singularity, and regularity there selects the Bessel function $J_m$ over the divergent $Y_m$ .
Key idea: cylindrical coordinates are orthogonal with scale factors $(1,\rho,1)$ ; drop these into the universal orthogonal-curvilinear templates and every operator follows mechanically, the lone factor $\rho$ generating all the $1/\rho$ structure.

The core insight

Cylindrical coordinates are orthogonal with scale factors $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ ; insert these into the universal orthogonal-curvilinear templates and every operator — gradient, divergence, curl, Laplacian — follows mechanically, the lone factor $\rho$ generating all the $1/\rho$ structure.

Where this lives today

Cylindrical coordinates and their Bessel radial functions are working tools across modern physics and engineering. A circular metallic waveguide propagates a TM₀₁ mode only above a cutoff wavenumber $k_c=j_{0,1}/a=2.40483/a$ set by that same first Bessel zero, and step-index optical fibres are analysed as LP modes built from $J_m$ inside the core and decaying modified-Bessel functions outside — the architecture of long-haul telecommunications rests on solving the cylindrical Helmholtz equation. Particle accelerators store fields in cylindrical RF cavities whose resonant frequencies are Bessel zeros; MRI scanners wind gradient coils whose fields expand in cylindrical harmonics; tokamak plasma columns are modelled in exactly this frame. In fluid dynamics the Taylor–Couette apparatus — fluid sheared between rotating coaxial cylinders — is a canonical testbed for the onset of turbulence, and quantised vortices in rotating Bose–Einstein condensates are cylindrically symmetric solutions with the axis as a topological defect. Numerical relativity and computational fluid dynamics adopt cylindrical grids whenever a source, jet, or accretion disk has an axis, because separating that symmetry out is the cheapest way to turn a hard PDE into a tractable radial one.

The physical picture

Picture stacking flat polar-coordinate sheets along the $z$ -axis. Inside each horizontal sheet you locate a point by radius $\rho$ and bearing $\phi$ ; the sheets are indexed by height $z$ . The surfaces of constant coordinate are then three familiar families: constant $\rho$ are coaxial cylinders, constant $\phi$ are half-planes hinged on the axis, and constant $z$ are horizontal planes — and, crucially, these three families cross at right angles everywhere, which is what makes the system orthogonal and keeps every operator formula diagonal. The local unit vectors point along increasing coordinate: $\hat{\mathbf{e}}_\rho$ straight out from the axis, $\hat{\mathbf{e}}_\phi$ tangent to the circle, $\hat{\mathbf{e}}_z$ straight up. The first figure shows this triad riding on the coordinate cage, with the one relation that drives everything: an azimuthal step of $\mathrm{d}\phi$ sweeps a physical arc of length $\rho\,\mathrm{d}\phi$ , not $\mathrm{d}\phi$ .

Figure

Cylindrical coordinates locate a point by ρ, the perpendicular distance from the z axis, φ the azimuth around it, and z the height along it. Constant-ρ surfaces are coaxial cylinders, constant-φ are half-planes, constant-z are planes, meeting at right angles. The local triad ê_ρ, ê_φ, ê_z rotates with φ, and a step dφ sweeps a physical arc ρ dφ, so the volume element is dV = ρ dρ dφ dz

Two consequences flow from that rotating triad. First, the divergence's $\tfrac1\rho\partial_\rho(\rho A_\rho)$ is nothing more mysterious than radial flux spreading over cylindrical shells whose area grows like $\rho$ — think of $\rho\,\mathrm{d}\phi$ as "real distance per radian", small near the axis and large far out. Second, the axis $\rho=0$ is special: there $\phi$ is undefined and $\hat{\mathbf{e}}_\rho,\hat{\mathbf{e}}_\phi$ are ambiguous. This is a coordinate singularity, not a physical one, but it carries teeth — demanding that a real field stay finite on the axis is a genuine boundary condition. When a symmetric problem separates, that condition is exactly what selects the well-behaved radial solution. The second figure shows the two radial solutions of the cylindrical Helmholtz equation: the regular Bessel function $J_0$ , finite on the axis, and its partner $Y_0$ , which plunges to $-\infty$ there and is thrown out.

Figure

The two independent radial solutions of the cylindrical wave equation plotted against k⊥ρ. The Bessel function J₀ stays finite on the axis and is the physical one; its partner Y₀ diverges to minus infinity as ρ → 0 and is rejected by regularity. J₀ has its first zero at k⊥ρ = 2.40483, which fixes the fundamental cutoff of a circular waveguide and the lowest mode of a circular drum

The core equations

The system is defined by keeping $z$ and writing the plane as polar coordinates. These relations, invertible away from the axis, are the map from which everything else is computed.

x=\rho\cos\phi,\quad y=\rho\sin\phi,\quad z=z \qquad (1)

where: $\rho$ — radial distance from the $z$ -axis, $\rho=\sqrt{x^2+y^2}\ge 0$ ; $\phi$ — azimuthal angle, $\phi=\operatorname{atan2}(y,x)\in[0,2\pi)$ ; $z$ — height, identical to the Cartesian $z$ . Reading it: $\rho,\phi$ are ordinary plane polar coordinates in every horizontal slice. Differentiating the position vector along each coordinate and normalising gives the local orthonormal triad. The radial and azimuthal vectors carry the angle $\phi$ explicitly, which is the algebraic fingerprint of their rotation.

\hat{\mathbf{e}}_\rho=\cos\phi\,\hat{\mathbf{x}}+\sin\phi\,\hat{\mathbf{y}},\quad \hat{\mathbf{e}}_\phi=-\sin\phi\,\hat{\mathbf{x}}+\cos\phi\,\hat{\mathbf{y}},\quad \hat{\mathbf{e}}_z=\hat{\mathbf{z}} \qquad (2)

where: $\hat{\mathbf{e}}_\rho$ — unit vector of increasing $\rho$ (radially outward); $\hat{\mathbf{e}}_\phi$ — unit vector of increasing $\phi$ (tangent to the circle); $\hat{\mathbf{x}},\hat{\mathbf{y}},\hat{\mathbf{z}}$ — the fixed Cartesian vectors. Reading it: the triad is right-handed, $\hat{\mathbf{e}}_\rho\times\hat{\mathbf{e}}_\phi=\hat{\mathbf{e}}_z$ , and orthonormal at every point. Because $\phi$ sits inside (2), the in-plane vectors have nonzero $\phi$ -derivatives. These two little relations are the entire reason the operators look busier than their Cartesian cousins.

\frac{\partial\hat{\mathbf{e}}_\rho}{\partial\phi}=\hat{\mathbf{e}}_\phi,\qquad \frac{\partial\hat{\mathbf{e}}_\phi}{\partial\phi}=-\hat{\mathbf{e}}_\rho,\qquad \text{all other }\partial_i\hat{\mathbf{e}}_j=0 \qquad (3)

where: $\partial/\partial\phi$ — derivative as the azimuth changes; only $\hat{\mathbf{e}}_\rho$ and $\hat{\mathbf{e}}_\phi$ turn, and they turn into each other. Reading it: carry $\hat{\mathbf{e}}_\rho$ once around the axis and it traces a full circle back to itself. The physical displacement built from coordinate steps exposes the scale factors directly: the coefficient multiplying each $\mathrm{d}u_i$ is the physical length per unit coordinate. Their product is the volume element.

\mathrm{d}\mathbf{r}=\mathrm{d}\rho\,\hat{\mathbf{e}}_\rho+\rho\,\mathrm{d}\phi\,\hat{\mathbf{e}}_\phi+\mathrm{d}z\,\hat{\mathbf{e}}_z,\qquad \mathrm{d}V=\rho\,\mathrm{d}\rho\,\mathrm{d}\phi\,\mathrm{d}z \qquad (4)

where: $\mathrm{d}\mathbf{r}$ — infinitesimal displacement; $\mathrm{d}V$ — volume element $=h_\rho h_\phi h_z\,\mathrm{d}\rho\,\mathrm{d}\phi\,\mathrm{d}z$ ; the scale factors $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ multiply $\mathrm{d}\rho,\mathrm{d}\phi,\mathrm{d}z$ . Reading it: only the azimuthal step is stretched, by the running factor $\rho$ . The gradient is rate of change per physical length, so each partial derivative is divided by its scale factor — the sole effect being the $1/\rho$ on the angular term.

\nabla f=\frac{\partial f}{\partial\rho}\,\hat{\mathbf{e}}_\rho+\frac{1}{\rho}\frac{\partial f}{\partial\phi}\,\hat{\mathbf{e}}_\phi+\frac{\partial f}{\partial z}\,\hat{\mathbf{e}}_z \qquad (5)

where: $f$ — a scalar field $f(\rho,\phi,z)$ ; $1/\rho$ — the $h_\phi^{-1}$ factor converting an angular rate into a physical rate. Reading it: identical to Cartesian except the azimuthal component is $\tfrac1\rho\partial_\phi f$ , not $\partial_\phi f$ . The divergence is net outward flux per unit volume. Because the cylindrical shell area grows with $\rho$ , the radial term groups $\rho A_\rho$ inside the derivative.

\nabla\cdot\mathbf{A}=\frac{1}{\rho}\frac{\partial(\rho A_\rho)}{\partial\rho}+\frac{1}{\rho}\frac{\partial A_\phi}{\partial\phi}+\frac{\partial A_z}{\partial z} \qquad (6)

where: $A_\rho,A_\phi,A_z$ — physical components of $\mathbf{A}=A_\rho\hat{\mathbf{e}}_\rho+A_\phi\hat{\mathbf{e}}_\phi+A_z\hat{\mathbf{e}}_z$ . Reading it: expanding the radial term gives $\partial_\rho A_\rho+A_\rho/\rho$ , and that extra $A_\rho/\rho$ is exactly the piece a naive Cartesian calculation drops. The curl is circulation per unit area. The $z$ -component again carries the $\rho A_\phi$ grouping, for the same shell-area reason.

\nabla\times\mathbf{A}=\left(\frac{1}{\rho}\frac{\partial A_z}{\partial\phi}-\frac{\partial A_\phi}{\partial z}\right)\hat{\mathbf{e}}_\rho+\left(\frac{\partial A_\rho}{\partial z}-\frac{\partial A_z}{\partial\rho}\right)\hat{\mathbf{e}}_\phi+\frac{1}{\rho}\left(\frac{\partial(\rho A_\phi)}{\partial\rho}-\frac{\partial A_\rho}{\partial\phi}\right)\hat{\mathbf{e}}_z \qquad (7)

where: $\nabla\times\mathbf{A}$ — the curl, measuring local circulation of $\mathbf{A}$ . Reading it: the $\hat{\mathbf{e}}_\phi$ component alone has no $1/\rho$ dressing, because a loop in the $\rho z$ -plane sees no azimuthal stretching. The Laplacian is $\nabla\cdot(\nabla f)$ , inheriting the $\rho$ grouping from the divergence and the $1/\rho$ from the gradient in one self-adjoint operator — the single most used equation in all of cylindrical physics.

\nabla^2 f=\frac{1}{\rho}\frac{\partial}{\partial\rho}\!\left(\rho\frac{\partial f}{\partial\rho}\right)+\frac{1}{\rho^2}\frac{\partial^2 f}{\partial\phi^2}+\frac{\partial^2 f}{\partial z^2} \qquad (8)

where: $\nabla^2 f$ — the scalar Laplacian; $1/\rho^2$ — the $h_\phi^{-2}$ factor on the angular second derivative. Reading it: the radial part is $f''+\tfrac1\rho f'$ , the plane-polar operator, plus the untouched $z$ -derivative.

Derivation AScale factors and the line element

The goal is to prove from first principles that cylindrical coordinates are orthogonal with scale factors $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ , because those three numbers — and nothing else — fix the volume element, the gradient, the divergence, the curl and the Laplacian through the universal orthogonal-curvilinear templates. We do it by computing the physical line element $\mathrm{d}s^2=\mathrm{d}x^2+\mathrm{d}y^2+\mathrm{d}z^2$ in the new coordinates and reading off the diagonal metric; the cancellation of the cross term is itself the statement that the basis is orthogonal.

Step 1. Differentiate the defining relations (1): $\mathrm{d}x=\cos\phi\,\mathrm{d}\rho-\rho\sin\phi\,\mathrm{d}\phi$ and $\mathrm{d}y=\sin\phi\,\mathrm{d}\rho+\rho\cos\phi\,\mathrm{d}\phi$ , while $\mathrm{d}z=\mathrm{d}z$ . Why: the metric lives in Cartesian distances, so we must express the Cartesian differentials in terms of the coordinate differentials before squaring.

Step 2. Square and add. The $\mathrm{d}\rho\,\mathrm{d}\phi$ cross terms cancel because $\sin\phi\cos\phi-\cos\phi\sin\phi=0$ , leaving $\mathrm{d}x^2+\mathrm{d}y^2=\mathrm{d}\rho^2+\rho^2\,\mathrm{d}\phi^2$ , hence $\mathrm{d}s^2=\mathrm{d}\rho^2+\rho^2\,\mathrm{d}\phi^2+\mathrm{d}z^2$ . Why: the vanishing of the cross term is precisely orthogonality — the coordinate axes meet at right angles, so the metric is diagonal with no $g_{\rho\phi}$ .

Step 3. Read off the scale factors from $g_{ii}=h_i^2$ : $h_\rho^2=1,\ h_\phi^2=\rho^2,\ h_z^2=1$ , so $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ . Why: in an orthogonal system the diagonal metric entries are the squared scale factors, so a diagonal $\mathrm{d}s^2$ hands you the $h_i$ directly.

Step 4. Cross-check against the Jacobian of (1): $\det\big(\partial(x,y,z)/\partial(\rho,\phi,z)\big)=\rho=h_\rho h_\phi h_z$ , matching the volume element $\mathrm{d}V=\rho\,\mathrm{d}\rho\,\mathrm{d}\phi\,\mathrm{d}z$ . Why: two independent routes — the metric and the change-of-variables Jacobian — must give the same volume factor, and their agreement certifies the scale factors. *The scale factors are $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ and the Jacobian is $\rho$ ; as $\rho\to$ any fixed value $h_\phi$ becomes locally constant and the Cartesian limit is recovered.*

Go deeper · Derivation B · 1 minDivergence and Laplacian from the scale factors

Derivation BDivergence and Laplacian from the scale factors

The goal is to show that once the scale factors are in hand, the divergence and Laplacian require no fresh geometric reasoning at all — they drop out of the orthogonal-curvilinear master formulas by substitution, and every $1/\rho$ and $\rho A_\rho$ grouping is generated automatically by the lone factor $h_\phi=\rho$ . This is the practical heart of the whole subject: you memorise $h_\phi=\rho$ and the templates, and never differentiate a unit vector by hand.

Step 1. Start from the Lamé divergence template $\nabla\cdot\mathbf{A}=\dfrac{1}{h_\rho h_\phi h_z}\big[\partial_\rho(h_\phi h_z A_\rho)+\partial_\phi(h_z h_\rho A_\phi)+\partial_z(h_\rho h_\phi A_z)\big]$ . Why: this is the coordinate-free "flux per unit volume" statement evaluated on a curvilinear box; it holds for any orthogonal system with the $h_i$ left symbolic.

Step 2. Insert $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ , so the prefactor is $1/\rho$ and the face factors are $h_\phi h_z=\rho$ , $h_z h_\rho=1$ , $h_\rho h_\phi=\rho$ . This gives $\nabla\cdot\mathbf{A}=\tfrac1\rho\big[\partial_\rho(\rho A_\rho)+\partial_\phi A_\phi+\partial_z(\rho A_z)\big]$ . Why: the substitution is purely mechanical, and the $\rho$ inside the radial and $z$ terms is the growing shell area entering the flux balance.

Step 3. In the $z$ term the factor $\rho$ is constant in $z$ , so $\tfrac1\rho\partial_z(\rho A_z)=\partial_z A_z$ , recovering equation (6): $\nabla\cdot\mathbf{A}=\tfrac1\rho\partial_\rho(\rho A_\rho)+\tfrac1\rho\partial_\phi A_\phi+\partial_z A_z$ . Why: only the truly $\rho$ -dependent geometry survives; the untouched $z$ -direction reverts to its flat form.

Step 4. Compose with the gradient (5) by feeding $A_\rho=\partial_\rho f$ , $A_\phi=\tfrac1\rho\partial_\phi f$ , $A_z=\partial_z f$ into the result of Step 3: $\nabla^2 f=\tfrac1\rho\partial_\rho(\rho\,\partial_\rho f)+\tfrac1\rho\partial_\phi\!\big(\tfrac1\rho\partial_\phi f\big)+\partial_z^2 f=\tfrac1\rho\partial_\rho(\rho\,\partial_\rho f)+\tfrac1{\rho^2}\partial_\phi^2 f+\partial_z^2 f$ . Why: the Laplacian is the divergence of the gradient, and the $1/\rho$ pulling through the $\phi$ -derivative is what promotes it to $1/\rho^2$ , the signature angular term of equation (8). *Why these matter: expanding $\tfrac1\rho\partial_\rho(\rho A_\rho)=\partial_\rho A_\rho+A_\rho/\rho$ makes the extra $A_\rho/\rho$ term explicit; setting all $\phi$ and $z$ derivatives to zero collapses $\nabla^2$ to the purely radial ODE $f''+\tfrac1\rho f'$ .*

Go deeper · Derivation C · 1 minThe z-component of the curl from circulation

Derivation CThe z-component of the curl from circulation

The goal is to obtain the $\hat{\mathbf{e}}_z$ component of the cylindrical curl directly from its geometric meaning — circulation per unit area — rather than by quoting the determinant, so that the origin of the $\tfrac1\rho\partial_\rho(\rho A_\phi)$ grouping becomes visible. We take an infinitesimal patch in a constant- $z$ plane bounded by $\rho,\rho+\mathrm{d}\rho$ and $\phi,\phi+\mathrm{d}\phi$ , add up $\oint\mathbf{A}\cdot\mathrm{d}\mathbf{r}$ counter-clockwise, and divide by the patch area; Stokes' theorem in the shrinking limit turns that ratio into the curl component.

Step 1. The patch area is the product of the radial edge $\mathrm{d}\rho$ and the arc $\rho\,\mathrm{d}\phi$ , so $\mathrm{d}A=\rho\,\mathrm{d}\rho\,\mathrm{d}\phi$ . Why: the azimuthal edge is stretched by $\rho$ , and this area is the denominator that will divide the circulation.

Step 2. Along the two azimuthal arcs the tangential contribution is $A_\phi$ times the arc length $\rho\,\mathrm{d}\phi$ , evaluated at $\rho$ and $\rho+\mathrm{d}\rho$ with opposite senses: $[A_\phi(\rho+\mathrm{d}\rho)(\rho+\mathrm{d}\rho)-A_\phi(\rho)\rho]\,\mathrm{d}\phi\approx\partial_\rho(\rho A_\phi)\,\mathrm{d}\rho\,\mathrm{d}\phi$ . Why: because the arc length itself scales with $\rho$ , it is the product $\rho A_\phi$ — not $A_\phi$ alone — that is differenced between the two arcs, which is exactly where the grouping comes from.

Step 3. Along the two radial edges the tangential contribution is $A_\rho$ times $\mathrm{d}\rho$ , differenced across $\phi$ : $-[A_\rho(\phi+\mathrm{d}\phi)-A_\rho(\phi)]\,\mathrm{d}\rho\approx-\partial_\phi A_\rho\,\mathrm{d}\phi\,\mathrm{d}\rho$ . Why: the radial edges are not stretched, so only $A_\rho$ appears, and the minus sign comes from traversing them in the counter-clockwise sense.

Step 4. Add the two contributions and divide by $\mathrm{d}A=\rho\,\mathrm{d}\rho\,\mathrm{d}\phi$ : $(\nabla\times\mathbf{A})_z=\tfrac1\rho\big[\partial_\rho(\rho A_\phi)-\partial_\phi A_\rho\big]$ , matching the $\hat{\mathbf{e}}_z$ term of equation (7). Why: circulation per unit area in the vanishing-patch limit is the definition of the normal curl component, now explicit in the scale factor. *Why: the $\hat{\mathbf{e}}_\rho$ and $\hat{\mathbf{e}}_\phi$ components arise identically from loops in the $\phi z$ - and $\rho z$ -planes, where the unstretched $z$ -edge removes any $1/\rho$ from the $\hat{\mathbf{e}}_\phi$ term.*

Limiting cases

Cylindrical coordinates specialise and connect to their neighbours in several instructive limits, each obtained by collapsing one coordinate or one derivative.

Case	Result	Meaning
Axial symmetry ( $\partial_\phi=0$ )	$\nabla^2 f=\tfrac1\rho\partial_\rho(\rho\,\partial_\rho f)+\partial_z^2 f$	the angular term vanishes; a PDE in three variables drops to two, and full $z$ -independence leaves a pure radial ODE
Plane polar ( $z$ ignored)	$\nabla^2 f=\tfrac1\rho\partial_\rho(\rho\,\partial_\rho f)+\tfrac1{\rho^2}\partial_\phi^2 f$	dropping $z$ recovers two-dimensional polar coordinates, the natural frame for disks and circular membranes
Local Cartesian ( $\rho\to\infty$ )	$h_\phi=\rho\to$ large, curvature terms $\to 0$	far from the axis a small patch looks flat; the $1/\rho$ terms are negligible and operators approach Cartesian form
On-axis ( $\rho\to 0$ )	$\phi$ undefined, $\hat{\mathbf{e}}_\rho,\hat{\mathbf{e}}_\phi$ ambiguous	a removable coordinate singularity; physical fields stay finite, and that finiteness is the boundary condition selecting $J_m$ over $Y_m$

Common trapTreating the unit vectors as constant

The single most common error is to carry over the Cartesian habit and treat $\hat{\mathbf{e}}_\rho,\hat{\mathbf{e}}_\phi$ as fixed. They are not: $\partial_\phi\hat{\mathbf{e}}_\rho=\hat{\mathbf{e}}_\phi$ and $\partial_\phi\hat{\mathbf{e}}_\phi=-\hat{\mathbf{e}}_\rho$ , so when you differentiate a vector field you must differentiate its basis too. Skip this and you drop the $A_\rho/\rho$ piece of the divergence and the $\rho A_\phi$ grouping of the curl, producing operators that are simply wrong — for instance a "divergence" $\partial_\rho A_\rho+\tfrac1\rho\partial_\phi A_\phi+\partial_z A_z$ that fails the flux test on the trivial field $\mathbf{A}=\rho\,\hat{\mathbf{e}}_\rho$ . The cure is to never differentiate a unit vector by hand: use equations (5)–(8) as written, since they have already absorbed the rotation of the triad into the scale-factor structure.

Worked examples

1 · The Laplacian on two test fields (confidence builder)

Apply the cylindrical Laplacian (8) to $f=\rho^2$ and to $f=\rho\cos\phi$ , and confirm each against the Cartesian answer.

Step 1. For $f=\rho^2$ only the radial term acts: $\nabla^2 f=\tfrac1\rho\partial_\rho\big(\rho\cdot 2\rho\big)=\tfrac1\rho\partial_\rho(2\rho^2)=\tfrac1\rho(4\rho)=\mathbf{4}$ . In Cartesian, $\rho^2=x^2+y^2$ , so $\nabla^2(x^2+y^2)=2+2=\mathbf{4}$ — they agree.

Step 2. For $f=\rho\cos\phi$ both the radial and angular terms contribute. Radial: $\tfrac1\rho\partial_\rho(\rho\cos\phi)=\tfrac1\rho\cos\phi$ . Angular: $\tfrac1{\rho^2}\partial_\phi^2(\rho\cos\phi)=\tfrac1{\rho^2}\big(\rho\cdot(-\cos\phi)\big)=-\tfrac1\rho\cos\phi$ .

Step 3. Add them: $\nabla^2 f=\tfrac1\rho\cos\phi-\tfrac1\rho\cos\phi=\mathbf{0}$ . And indeed $\rho\cos\phi=x$ , whose Laplacian is $0$ . The cylindrical Laplacian gives ∇²ρ² = 4 and ∇²(ρ cos φ) = 0, matching Cartesian exactly — and the second case shows the radial and angular terms cancelling, which they must for the harmonic function x.

2 · Field and capacitance of a coaxial cable (fully worked)

A coaxial cable has an inner conductor of radius $a=1.0$ mm held at $V_0=1.0$ V and a grounded outer conductor of radius $b=3.5$ mm, with vacuum between them. Using axial symmetry, find the potential, the radial field, the field at each conductor, and the capacitance per unit length.

Step 1. Between the conductors there is no charge, so the potential obeys Laplace's equation. With no $\phi$ - or $z$ -dependence, equation (8) collapses to $\tfrac1\rho\tfrac{\mathrm{d}}{\mathrm{d}\rho}\!\big(\rho\tfrac{\mathrm{d}V}{\mathrm{d}\rho}\big)=0$ , whose general solution is $V(\rho)=A\ln\rho+B$ — the hallmark logarithmic potential of cylindrical symmetry.

Step 2. Fix $A,B$ from the boundary values $V(a)=V_0$ and $V(b)=0$ . Subtracting gives $A=\dfrac{V_0}{\ln(a/b)}=-\dfrac{V_0}{\ln(b/a)}$ , so $V(\rho)=V_0\,\dfrac{\ln(b/\rho)}{\ln(b/a)}$ . With $\ln(b/a)=\ln 3.5=\mathbf{1.2528}$ .

Step 3. The field is radial, $\mathbf{E}=-\nabla V=-\tfrac{\mathrm{d}V}{\mathrm{d}\rho}\,\hat{\mathbf{e}}_\rho=\dfrac{V_0}{\rho\ln(b/a)}\,\hat{\mathbf{e}}_\rho$ . At the inner conductor $E(a)=\dfrac{1.0}{(1.0\times10^{-3})(1.2528)}=\mathbf{798\ V/m}$ , and at the outer $E(b)=\dfrac{1.0}{(3.5\times10^{-3})(1.2528)}=\mathbf{228\ V/m}$ — the field is strongest at the inner wall, which is why cables break down there.

Step 4. A Gaussian cylinder of radius $\rho$ and length $L$ gives $E(2\pi\rho L)=Q/\varepsilon_0$ , so the charge per length is $Q/L=2\pi\varepsilon_0\,\rho E=\dfrac{2\pi\varepsilon_0 V_0}{\ln(b/a)}$ , and the capacitance per length is $C/L=\dfrac{Q}{L V_0}=\dfrac{2\pi\varepsilon_0}{\ln(b/a)}=\dfrac{2\pi(8.854\times10^{-12})}{1.2528}=\mathbf{44.4\ pF/m}$ . V(ρ) = V₀ ln(b/ρ)/ln(b/a), E(a) = 798 V/m, E(b) = 228 V/m, and C/L = 44.4 pF/m — the entire result flowing from the axially symmetric Laplacian and one Gaussian cylinder.

3 · The lowest mode of a circular drum

A circular membrane of radius $a=0.30$ m carries transverse waves at speed $c=100$ m/s. Find the frequency of its fundamental (axisymmetric) mode.

Step 1. The axisymmetric displacement $u(\rho,t)=R(\rho)\cos\omega t$ obeys the wave equation, whose spatial part is the axially symmetric Helmholtz equation $\tfrac1\rho\tfrac{\mathrm{d}}{\mathrm{d}\rho}\!\big(\rho R'\big)+k^2 R=0$ with $k=\omega/c$ . This is Bessel's equation of order zero, with regular solution $R(\rho)=J_0(k\rho)$ ; the singular partner $Y_0$ is discarded by finiteness on the axis.

Step 2. The clamped rim requires $R(a)=0$ , i.e. $J_0(ka)=0$ . The lowest root is the first zero of $J_0$ , $j_{0,1}=\mathbf{2.40483}$ , so $ka=j_{0,1}$ and $k=j_{0,1}/a$ .

Step 3. Hence $\omega=ck=\dfrac{c\,j_{0,1}}{a}=\dfrac{(100)(2.40483)}{0.30}=\mathbf{801.6\ rad/s}$ , giving $f=\omega/2\pi=\mathbf{127.6\ Hz}$ . The fundamental sits at f = 127.6 Hz, set entirely by the first Bessel zero 2.40483 — and because that zero is not a whole-number multiple of anything, a drum's overtones are inharmonic, which is why a timpano sounds nothing like a plucked string.

Physics of the portrait

The radial equation that survives every axially symmetric problem is Bessel's, and its solutions — the cylinder functions — are the true payoff of cylindrical coordinates. They carry the name of the astronomer who systematised them.

Origins · 1 minFriedrich Bessel and how this was discovered

Friedrich Bessel

1784–1846 · cylinder functions and the radial solutions of axially symmetric problems

Friedrich Wilhelm Bessel began as a teenage apprentice clerk in a Bremen trading house, taught himself astronomy and mathematics by night, and rose to direct the Königsberg Observatory, where in 1838 he made the first reliable measurement of stellar parallax — the distance to 61 Cygni. The functions that bear his name arose from a concrete problem: in his 1824 memoir on planetary perturbations he expanded the eccentric anomaly of an orbiting body in a Fourier series, and the coefficients turned out to be the integrals $J_n$ that he then studied in their own right, proving their recurrence and orthogonality relations. He was not the first to meet them — Daniel Bernoulli had encountered $J_0$ in the hanging chain and Euler in the vibrating membrane — but Bessel was the first to treat the whole family systematically, which is why they are the "cylinder functions" of mathematical physics. Their defining feature is exactly the one this article needs: of the two independent solutions of the cylindrical radial equation, only $J_n$ stays finite on the axis, so regularity at $\rho=0$ picks it out uniquely, and the discrete zeros of $J_n$ set the allowed modes of every drum, waveguide, and cylindrical cavity.

History

1732

— Daniel Bernoulli meets the order-zero cylinder function in the oscillating hanging chain

1764

— Euler analyses the vibrating circular membrane and encounters integer-order cylinder functions

1824

— Bessel systematises the functions Jₙ in his study of planetary perturbations

1867

— Carl Neumann introduces the second-kind functions Yₙ, singular on the axis

1871

— Hankel combines them into the complex cylinder functions Hₙ for radiating waves

Going further

Texts. Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., Ch. 3 (Curvilinear Coordinates) and Ch. 14 (Bessel Functions) · Stone & Goldbart, Mathematics for Physics, on curvilinear coordinates and cylinder functions · Riley, Hobson & Bence, Mathematical Methods for Physics and Engineering, 3rd ed., Ch. 10 · Boas, Mathematical Methods in the Physical Sciences, 3rd ed., Ch. 6 · Jackson, Classical Electrodynamics, 3rd ed., §3.7–3.11 for the cylindrical boundary-value problems. All operator forms follow from specialising the orthogonal-curvilinear (Lamé) templates to $(h_\rho,h_\phi,h_z)=(1,\rho,1)$ . Key numerical results: $\nabla^2\rho^2=4$ and $\nabla^2(\rho\cos\phi)=0$ ; the coax relations $V=V_0\ln(b/\rho)/\ln(b/a)$ , $E(a)=798$ V/m, $E(b)=228$ V/m, $C/L=2\pi\varepsilon_0/\ln(b/a)=44.4$ pF/m for $a=1.0$ mm, $b=3.5$ mm; and the drum fundamental $f=c\,j_{0,1}/(2\pi a)=127.6$ Hz for $c=100$ m/s, $a=0.30$ m using $j_{0,1}=2.40483$ . The line-charge field $\lambda/2\pi\varepsilon_0\rho$ and wire field $\mu_0 I/2\pi\rho$ are standard electromagnetism results.

Prerequisites. Scalar and Vector Fields · The Gradient Operator · The Del Operator and Vector Differential Identities. Leads to. The Laplacian Operator · The Divergence (Gauss) Theorem · Stokes' Theorem · Volume Integrals.

Ask the tutor: Ask me to separate Laplace's or the wave equation in cylindrical coordinates for your geometry, and watch the scale factors turn it into a Bessel problem step by step.

Why is it ρ and φ here rather than r and θ?

The convention keeps cylindrical and spherical coordinates from colliding. In cylindrical coordinates $\rho$ is the perpendicular distance from the axis and $\phi$ is the azimuth measured in the plane; in spherical coordinates $r$ is the distance from the origin and $\theta$ is the polar angle from the axis. If you wrote $r$ for the cylindrical radius you would clash with the spherical $r$ , and problems that use both systems — a charged disk viewed near and far, say — would become ambiguous. Some engineering texts do write $(r,\theta,z)$ for cylindrical, so always check which radius an author means; in mathematical physics $(\rho,\phi,z)$ is standard, and $\phi$ is deliberately the same symbol as the spherical azimuth because it plays the identical role.

What exactly makes the coordinates orthogonal, and why does it matter?

Orthogonal means the three coordinate surfaces meet at right angles everywhere, equivalently that the line element $\mathrm{d}s^2$ has no cross terms like $\mathrm{d}\rho\,\mathrm{d}\phi$ . You can see it happen in the derivation: when you square and add $\mathrm{d}x$ and $\mathrm{d}y$ , the cross term carries $\sin\phi\cos\phi-\cos\phi\sin\phi=0$ and vanishes. It matters enormously for practical work, because orthogonality is what makes the metric diagonal, the volume a plain product $h_\rho h_\phi h_z$ of edge lengths, and each operator decouple into three clean terms. A non-orthogonal coordinate change would force the full machinery of the metric tensor with off-diagonal $g_{ij}$ and $\sqrt{\det g}$ in place of the simple Jacobian.

Where does the factor 1/ρ in the operators actually come from?

From the single nontrivial scale factor $h_\phi=\rho$ , and nothing else. An azimuthal step $\mathrm{d}\phi$ moves you a physical distance $\rho\,\mathrm{d}\phi$ , so any rate of change in the $\phi$ -direction must be divided by $\rho$ to be a rate per metre — that is the $1/\rho$ in the gradient. In the divergence and Laplacian the same $\rho$ appears twice over: once as the growing shell area that puts $\rho$ inside $\partial_\rho(\rho\,\cdot)$ , and once as the $1/\rho$ prefactor from the volume element. Combine them and the angular second derivative picks up $1/\rho^2$ . So if you can remember the one fact $h_\phi=\rho$ , you can reconstruct every $1/\rho$ and $1/\rho^2$ in the entire scheme.

Why does a symmetric problem separate so cleanly in these coordinates?

Because the cylindrical Laplacian is a sum of three pieces that each involve only one coordinate: a radial operator in $\rho$ , a $\partial_\phi^2$ in $\phi$ , and a $\partial_z^2$ in $z$ . When you write a trial solution as a product $R(\rho)\Phi(\phi)Z(z)$ and divide through, each variable's terms can be set equal to a constant independently, splitting one partial differential equation into three ordinary ones. The $\phi$ -equation gives $e^{im\phi}$ with $m$ an integer (so the field is single-valued around the axis), the $z$ -equation gives exponentials or sinusoids, and what remains for $R$ is Bessel's equation. Separation of variables is not a lucky trick here; it is a direct consequence of the operator being diagonal in these coordinates.

If the axis singularity is only a coordinate artefact, why does it constrain the physics?

Because the coordinate breakdown flags a place where a solution is allowed to misbehave, and you must decide by hand whether the real field does. The radial equation has two independent solutions, one finite on the axis ( $J_m$ ) and one divergent there ( $Y_m$ , or $\ln\rho$ in the static case). Mathematics permits any combination; physics usually demands a finite field on the axis, and imposing that requirement kills the divergent solution and leaves a one-parameter family — which, together with an outer boundary condition, quantises the allowed modes. So the "artefact" at $\rho=0$ is doing real work: it supplies exactly one of the two boundary conditions every second-order radial problem needs.

When should I prefer cylindrical over spherical or Cartesian coordinates?

Match the coordinate system to the symmetry of the boundaries and the source, not to the region you happen to picture. Use cylindrical coordinates when the geometry has a single preferred axis and looks the same as you slide along it or rotate about it: wires, pipes, coaxial cables, solenoids, waveguides, rotating fluids, and beams. Use spherical coordinates when the symmetry is about a point — planets, atoms, isolated charges seen from far away — and Cartesian when the boundaries are flat and axis-aligned. Picking the frame that turns your boundaries into constant-coordinate surfaces is what lets separation of variables work and collapses the operators to their simplest radial form; a mismatched frame leaves the boundary conditions tangled across several coordinates.