Bessel Function Calculator

Two solutions, and which one is physical

A second-order equation has two independent solutions, and Bessel’s is no exception. The choice between them is made by the geometry rather than by the mathematics:

• Jₙ — first kind. Finite at x = 0 (J₀(0) = 1, and Jₙ(0) = 0 for n ≥ 1), oscillating with a slowly decaying amplitude ~√(2/πx). This is the solution that survives on a solid disc.
• Yₙ — second kind. Diverges to −∞ as x → 0, oscillating like J but a quarter-period out of phase. Only appears when the origin is outside the domain — an annulus, a coaxial cable.

J is the one that appears on a solid disc or a full cylinder, because it is the only solution that stays finite at the centre — and a real drumhead is not infinitely displaced at its middle. Its zeros set the vibration modes.

Y is needed exactly when the origin is NOT in the domain: an annular membrane, a coaxial line, heat in a pipe wall. On a solid disc its coefficient is forced to zero by the requirement that the solution be finite at the centre.

The zeros are the answer to most physical questions

A circular drumhead clamped at radius a vibrates in modes determined by Jₙ(ka) = 0. So the frequencies come from the ZEROS of the Bessel function, not from its values.

The first zeros of J₀ are 2.404825558, 5.52007811, 8.653727913, 11.79153444, 14.93091771. For J₁ they are 3.83170597, 7.01558667, 10.17346814, 13.32369194, and for J₂ 5.135622302, 8.41724414, 11.61984117.

The same numbers set the cutoff frequencies of a circular waveguide, the resonances of a cylindrical cavity, and the diffraction pattern from a circular aperture — the first zero of J₁ is what gives the Airy disc its size, and therefore the resolution limit of every telescope and microscope with a round lens.

Why a circular drum has no pitch

The first zeros of J0 are listed above. They are NOT evenly spaced: the gaps here run from 3.11525 to 3.14083, approaching π only in the limit. That unevenness is why a circular drum has no pitch — its overtones are not whole-number multiples of its fundamental, unlike a string, whose modes are spaced exactly evenly.

Compare a string: its modes are at 1, 2, 3, 4 times the fundamental — exact whole-number ratios, which is what makes a plucked string sound like a note. A drumhead’s modes are in the ratios 1 : 1.593 : 2.136 : 2.296 and so on, which are not ratios of small whole numbers at all.

That is why a timpani has to be built with a large air cavity and struck off-centre to suppress the worst of the inharmonicity, and why an undamped circular membrane sounds like a thud rather than a pitch. The unevenness of the Bessel zeros is audible.

The gaps do approach π eventually — the asymptotic form of Jₙ is a cosine with a decaying amplitude — but slowly, and the early zeros are the ones that matter for the audible modes.

The modified pair, where the sign flips

Change one sign in the equation, from (x² − n²) to −(x² + n²), and the oscillation becomes growth and decay:

• Iₙ. Monotonic and growing, like eˣ/√(2πx) for large x. No zeros for n ≥ 0 except I₀ never and Iₙ only at the origin, so it never oscillates. I₀(3) = 4.88079258587.
• Kₙ. Monotonic and decaying, like √(π/2x)·e⁻ˣ. Diverges at the origin. This is the one that describes something dying away with radius. K₀(3) = 0.0347395043863.

I appears where the equation has the opposite sign — diffusion and static fields rather than waves. It grows rather than oscillates, so it describes something building up towards a boundary rather than ringing.

K is the decaying modified solution, describing a field dying away with radius — the evanescent part of a waveguide mode, or the screened potential around a charge in a plasma.

The modified functions have no real zeros at all — I grows monotonically and K decays monotonically, so neither ever crosses the axis. That is the clearest difference from J and Y.

Spherical Bessel functions, which are elementary

Separating in spherical rather than cylindrical coordinates gives jₙ and yₙ, which are J and Y at half-integer order rescaled. Unusually for this subject, they are elementary: every one is a finite combination of sin x, cos x and powers of 1/x.

j₀(x) is simply sin x / x — 0.841470984808 at x = 1 — and y₀(x) is −cos x / x. Higher orders get longer but stay elementary.

The spherical Bessel functions come from the same separation in spherical coordinates, and are elementary: every one is a finite combination of sin x, cos x and powers of 1/x. They are the radial parts of the free-particle wavefunction.

Their zeros are a different sequence from Jₙ’s, and the page computes them separately: j₀ vanishes at the multiples of π, while J₀ vanishes at 2.4048, 5.5201 and so on. Confusing the two is easy and gives wrong mode frequencies.

How accurate, and where the weak point is

Each family is computed by whichever method is stable in its own range — ascending series near the origin, asymptotic expansions far from it, Miller’s downward recurrence where upward recurrence would amplify the wrong solution. The resulting accuracy differs by family, and the page states each rather than implying they are all the same:

• J and Y — about 6e-12 worst case, from the J/Y Wronskian over 0.2 ≤ x ≤ 40.
• I — about 1e-14 worst case, from the order recurrence over n ≤ 25, x ≤ 90.
• K — about 4e-9 worst case near the x = 8 crossover between its two expansions; better than 1e-12 away from it.
• Spherical — about 7e-10 against the closed forms, which themselves cancel badly at large x.

The K family is the weakest, and the reason is structural: its ascending series is a difference of two large nearly-equal terms and loses digits as x grows, while its asymptotic expansion is divergent and only reaches machine precision past about x = 15. Neither is exact in between, and the crossover was placed where the worst case is smallest.

Those figures come from comparison against integral-representation quadrature and the Wronskian identities, not from a claim. The tool checks the relevant Wronskian at whatever argument you enter: J1(x)Y0(x) − J0(x)Y1(x) = 2/(πx) gives 0.254647908947 against 0.254647908947.

Sources and methodology

Bessel functions, their zeros and the Wronskian relations are standard; the references below carry the canonical statements.

• NIST Digital Library of Mathematical Functions — §10.2, Bessel Functions of the First and Second Kinds
• NIST DLMF — §10.21, Zeros
• How we calculate — methodology

Method. Every figure on this page comes from src/lib/bessel-function.ts over src/lib/algebra/bessel.ts. J and Y use the ascending series below x = 12 and an optimally truncated Hankel asymptotic expansion above it; Jₙ at higher orders uses Miller's downward recurrence, and Iₙ an all-positive series that cannot cancel. The crossovers were placed by measuring both methods around them rather than by rule of thumb. That engine is verified on every change against 76 hand-written assertions, including that the J/Y and I/K Wronskians hold across the tested range and that the order recurrences are satisfied at every order up to 25. The count and the per-case breakdown are published on the formula verification page.

Related calculators

Where this goes next:

More in Math, or browse all calculators.

Read the guide

Bessel functions are what separation of variables produces in circular geometry, the way sines and cosines do in rectangular. The Gamma Function Calculator covers the function their series coefficients are built from, and the Hyperbolic Functions Calculator the growth-and-decay pair the modified equation resembles.