Visual walkthrough — Bessel's equation and Bessel functions (intro, physical relevance)

Step 1 — Why a circle is not a string

WHAT. A guitar string is a line: every point sits the same distance from its neighbours. A drumhead is a disc: as you walk outward from the centre, each new "ring" is bigger than the last.

WHY. This single geometric fact is the whole story. On a line, a wave's energy stays in a fixed strip. On a disc, the same energy has to smear over an ever-growing circumference — the further out you go, the longer the ring the wave must cover. That forced spreading is what will eventually make our wave weaken, and it is the physical seed of every term we are about to derive.

PICTURE. Look at the figure. On the left, the ripples on the string (red) are evenly tall. On the right, the ripples on the pond spread onto ever-larger rings (blue), so the same "amount of wiggle" must get thinner.

Now that $r$ has a name, the "growing circumference" of Step 1 is precisely $2\pi r$: at distance $r$ from the centre the ring has that length, and it grows without bound as $r$ increases. Hold that fact — it returns in Step 6.

Step 2 — The law the drum obeys

WHAT. Physics gives one equation for a steadily-vibrating membrane, the Helmholtz equation in polar coordinates. Before we write it, meet its last new symbol:

With every symbol ($u,r,\theta ,k$) now defined, the law reads: $\underset{\text{curvature outward}}{\underbrace{u_{rr}}}+\underset{\text{the circle tax}}{\underbrace{\frac{1}{r}{u}_r}}+\underset{\text{curvature around}}{\underbrace{\frac{1}{r^2}{u}_{\theta \theta }}}+\underset{\text{springiness}}{\underbrace{k^{2}u}}=0.$

WHY each term is here.

• $u_{rr}$ — how the surface bends as you move out. Bending stores tension energy.
• $\frac{1}{r}{u}_r$ — the "circle tax." The $1/r$ is exactly the growing-ring effect from Step 1; it is not in the string equation. Watch this term — it becomes the star.
• $\frac{1}{r^2}{u}_{\theta \theta }$ — bending as you go around; the $1/r^2$ scales the angular ruler to the circle's size.
• $k^{2}u$ — the restoring "springiness"; without it the surface would never oscillate.

PICTURE. Each term is drawn as an arrow acting at one sample point on the drum: outward curvature, the sideways circle-tax pull, angular curvature, and the spring pushing back to flat.

Step 3 — Split the problem in two: radius vs. angle

WHAT. We guess the answer is a product of a radius-only piece and an angle-only piece: $u(r,\theta )=R(r)\Theta (\theta ).$

WHY. A round drum has no favourite direction, so how the wave depends on distance ($R$) and how it depends on direction ($\Theta$) ought to be independent stories. Multiplying two one-variable functions is the simplest way to say "these don't interfere." This trick is Separation of variables.

Substitute and multiply through by $\frac{r^2}{R\Theta }$ so each symbol lands on the side it belongs to: $\underset{\text{depends only on }r}{\underbrace{\frac{r^2{R}^{\prime \prime }+r{R}^{\prime }}{R}+r^2{k}^2}}=\underset{\text{depends only on }\theta }{\underbrace{-\frac{{\Theta }^{\prime \prime }}{\Theta }}}.$ Here $R^{\prime \prime }=\frac{d^{2}R}{d{r}^2}$ and ${\Theta }^{\prime \prime }=\frac{d^2\Theta }{d{\theta }^2}$ — just the curvatures of each single-variable piece.

PICTURE. The drum is peeled into two independent gauges: a radial slider ($R$ along a spoke) and an angular dial ($\Theta$ around the rim).

Step 4 — Both sides must equal the same constant — and why it must be positive

WHAT. The left side changes only when $r$ changes; the right only when $\theta$ changes. If two things stay equal while you wiggle their inputs independently, neither can actually be moving — they are a constant. Call that constant $\lambda$: $-\frac{{\Theta }^{\prime \prime }}{\Theta }=\lambda ⟹{\Theta }^{\prime \prime }+\lambda \Theta =0.$

WHY the constant must be positive. We need $\Theta$ to be a repeating wave around the rim — going once around must come back to the same height. Look at the three cases:

• If $\lambda <0$, write $\lambda =-{\mu }^2$: then ${\Theta }^{\prime \prime }={\mu }^2\Theta$, whose solutions are $e^{\pm \mu \theta }$ — exponentials that grow or shrink, never returning to the start. A drum has no seam, so this is impossible.
• If $\lambda =0$: then ${\Theta }^{\prime \prime }=0$, giving $\Theta =a+b\theta$, which only repeats if $b=0$ (a boring constant).
• If $\lambda >0$, write $\lambda ={\nu }^2$ (a positive number is a square): then ${\Theta }^{\prime \prime }+{\nu }^2\Theta =0$ has the oscillating solutions $\cos \nu \theta ,\sin \nu \theta$ — exactly the periodic shapes a rim needs.

So we are forced to pick $\lambda ={\nu }^2>0$, and the angular equation becomes ${\Theta }^{\prime \prime }+{\nu }^2\Theta =0$.

WHY $\nu$ then becomes the integer $n$. So far $\nu$ is any nonnegative real number ($\nu =\sqrt{\lambda }\ge 0$). But single-valuedness demands $\Theta (\theta +2\pi )=\Theta (\theta )$: after one full lap the wave must line up with itself. For $\cos \nu \theta$ that lining-up happens only when $\nu$ is a whole number. From now on we therefore write $\nu =n$, an integer $0,1,2,\dots$ — the rename is not magic; periodicity picks it out.

PICTURE. The rim carries $n$ evenly spaced up/down lobes. $n=0$ is a uniform breathing ring; $n=1$ has one high side and one low side; $n=2$ has four alternating lobes.

Step 5 — One clean stretch of the ruler: $x=kr$

WHAT. The stray $k$'s are ugly. Rename the radial ruler: let $x=kr$, so $r=x/k$. We now rewrite each $r$-derivative as an $x$-derivative using the chain rule, step by step: $R^{\prime }(r)=\frac{dR}{dr}=\frac{dR}{dx}\cdot \frac{dx}{dr}=\frac{dR}{dx}\cdot k=k\frac{dR}{dx},$ because $\frac{dx}{dr}=\frac{d}{dr}(kr)=k$. Differentiate once more, and each new $\frac{d}{dr}$ again brings down a factor $k$: $R^{\prime \prime }(r)=\frac{d}{dr}\left(k\frac{dR}{dx}\right)=k\cdot \frac{d}{dx}\left(k\frac{dR}{dx}\right)=k^2\frac{d^{2}R}{d{x}^2}.$

WHY / the cancellation, made explicit. Substitute $r=x/k$, $R^{\prime }=k\frac{dR}{dx}$, $R^{\prime \prime }=k^2\frac{d^{2}R}{d{x}^2}$ into $r^2{R}^{\prime \prime }+r{R}^{\prime }+(r^2{k}^2-{\nu }^2)R=0$: $\frac{x^2}{k^2}\cdot k^2\frac{d^{2}R}{d{x}^2}+\frac{x}{k}\cdot k\frac{dR}{dx}+\left(\frac{x^2}{k^2}{k}^2-{\nu }^2\right)R=0.$ Term by term the $k$'s cancel — $\frac{x^2}{k^2}\cdot k^2=x^2$, $\frac{x}{k}\cdot k=x$, $\frac{x^2}{k^2}{k}^2=x^2$ — leaving $\boxed{x^2\frac{d^{2}R}{d{x}^2}+x\frac{dR}{dx}+(x^2-{\nu }^2)R=0}$

This is Bessel's equation — the parent's headline result, now earned. Read the three coefficients:

• $x^2{R}^{\prime \prime }$ — the outward-curvature term (from $u_{rr}$),
• $x{R}^{\prime }$ — the circle-tax survivor (from that $\frac{1}{r}{u}_r$; the extra $x$ is the growing ring),
• $(x^2-{\nu }^2)R$ — springiness $x^2$ minus the angular memory ${\nu }^2$.

PICTURE. Same equation, new axis: the physical spoke of length $a$ becomes a clean $x$-axis. The three terms are colour-tagged to their physical origins from Step 2.

Step 6 — Why the solution must fade: the shape of $J_{\nu }$

WHAT. The bounded solution is the Bessel function of the first kind $J_{\nu }(x)$ (built by the Frobenius method; the $x=0$ point is a regular singular point). Its large-$x$ shape is $J_{\nu }(x)\approx \underset{\text{fading amplitude}}{\underbrace{\sqrt{\frac{2}{\pi x}}}}\underset{\text{the wiggle}}{\underbrace{\cos \left(x-\frac{\nu \pi }{2}-\frac{\pi }{4}\right)}}.$

WHY it fades — and what the picture does and does not explain. Remember Step 1, now made exact: energy spreads over a ring of length $2\pi x$. Energy $\propto (\text{amplitude}{)}^2\times (\text{ring length})$; hold energy fixed and the amplitude must drop like $\frac{1}{\sqrt{x}}$. The energy-spreading picture explains only this $1/\sqrt{x}$ scaling — the shape of the envelope. The exact constant multiplying it, $\sqrt{2/\pi }$, and the phase shift $-\frac{\nu \pi }{2}-\frac{\pi }{4}$ inside the cosine, do not come from the simple ring argument; they fall out of a careful large-$x$ (asymptotic) analysis of the series. So: trust the picture for the fade rate, trust the analysis for the precise numbers. The $\cos (\dots )$ part is the ordinary wiggle, just phase-shifted.

PICTURE. $J_0$ (blue) and $J_1$ (orange) plotted: real oscillations wrapped inside the dashed $\pm \sqrt{2/\pi x}$ envelope (gray). Note $J_0(0)=1$, $J_1(0)=0$, and that the zero-crossings drift apart — not evenly spaced like a cosine.

Step 7 — The degenerate direction: what happens at the centre $x=0$

WHAT. The radial equation has a second solution $Y_{\nu }(x)$, the Bessel function of the second kind. Near $x=0$ it behaves like $Y_0(x)\sim \frac{2}{\pi }\ln x\to -\infty ,Y_{\nu }(x)\sim -x^{-\nu }\to -\infty (\nu >0).$

WHY it must be dropped on a solid disc. At the very centre $r=0$ (i.e. $x=0$) the drum has one physical height — it cannot be infinitely deep. $J_{\nu }$ stays finite there; $Y_{\nu }$ plunges to $-\infty$. So for any region that includes the centre we are forced to set the coefficient of $Y_{\nu }$ to zero. (Only an annulus — a drum with a hole punched out — keeps $Y_{\nu }$, because then $r=0$ is not in the region.)

PICTURE. $J_0$ (blue, finite, starts at 1) versus $Y_0$ (red, diving to $-\infty$ as $x\to 0$). A gray "solid disc includes $x=0$" band shows why only the blue curve is admissible.

Step 8 — The rim condition: how $k$ gets chosen and where the drum's notes come from

WHAT. After Step 7 the bounded radial shape is $R(r)=J_{\nu }(kr)$ (we dropped $Y_{\nu }$). But $k$ is still free. The drum's edge at $r=a$ is clamped — a real drumhead is nailed to the rim, so it cannot move there: $R(a)=0⟹J_{\nu }(ka)=0.$

WHY this quantizes $k$. $J_{\nu }$ is zero only at special isolated points, its zeros ${\alpha }_{\nu ,1}<{\alpha }_{\nu ,2}<\dots$ (for $J_0$ these are $2.405,5.520,\dots$). The clamp forces $ka$ to be one of these zeros: $ka={\alpha }_{\nu ,m}⟹k=\frac{{\alpha }_{\nu ,m}}{a}.$ So $k$ is not free after all — the rim picks out a discrete ladder of allowed values, one per zero. Each allowed $k$ is one vibration mode; its frequency is $f=\frac{ck}{2\pi }=\frac{c{\alpha }_{\nu ,m}}{2\pi a}$ (with $c$ the wave speed). Because the ratios $2.405:5.520:\dots$ are not whole numbers, a drum's overtones are inharmonic — the very reason a drum sounds like a drum and not a flute. This turns the whole problem into an eigenvalue problem, the subject of Sturm-Liouville theory.

PICTURE. $J_0(x)$ with its zeros marked; the vertical dashed lines at $x=ka=2.405,5.520,\dots$ show the only stretches of the ruler for which the clamped edge can sit at zero.

The one-picture summary

Everything collapses into a single frame. This figure stacks the whole derivation on one radial ruler $x=kr$: the fading $1/\sqrt{x}$ envelope (Step 1 & 6), the actual $J_0$ wiggle it houses (Step 5), the finite value at the centre that vetoed $Y_{\nu }$ (Step 7), and the marked zeros that become the drum's notes (Step 8). Reading it left-to-right is replaying the derivation.

Active-recall

Connections

• Separation of variables — the split in Step 3.
• Laplacian in polar and cylindrical coordinates — source of the circle-tax term.
• Frobenius method — builds $J_{\nu }$ once we reach the singular point.
• Regular singular points — why $x=0$ is special.
• Gamma function — supplies $\Gamma (m+\nu +1)$ in the $J_{\nu }$ series.
• Sturm-Liouville theory — frames the rim-clamp zeros as an eigenvalue problem.
• Parent: Bessel's equation — the result this page derives.