Bessel's equation x²y'' + xy' + (x² - ν²)y = 0 arises in cylindrical symmetry. Solutions are Bessel functions J_ν (first kind) and Y_ν (second kind). These functions are orthogonal with respect to a weighted inner product, enabling Fourier-Bessel expansions. Tables, recursion relations, and asymptotic approximations make Bessel functions practical for engineering and physics applications.
The Frobenius method you mastered handles ODEs with regular singular points by assuming power series solutions of the form x^r Σ aₙxⁿ. Bessel's equation x²y'' + xy' + (x² − ν²)y = 0 is the most important example of this class, arising whenever a physical problem has cylindrical symmetry — the vibrating circular drumhead, heat conduction in a cylindrical rod, electromagnetic modes in a fiber-optic cable. In all these settings, the natural radial coordinate is distance r from the central axis, and separating variables in cylindrical coordinates produces Bessel's equation with x = r.
Applying the Frobenius method at x = 0 (a regular singular point) yields the Bessel function of the first kind J_ν(x), given by the series J_ν(x) = Σ_{k=0}^∞ (−1)^k / (k! Γ(ν+k+1)) · (x/2)^(2k+ν). The key intuition for its behavior: for large x, J_ν(x) ≈ √(2/πx) cos(x − νπ/2 − π/4) — a damped oscillation whose amplitude decays like 1/√x. This is why Bessel functions describe outward-spreading waves in cylindrical geometry: they oscillate like sin and cos but gradually decrease in amplitude as the wave spreads over a larger and larger circumference. Think of ripples on a circular pond.
The second linearly independent solution, Y_ν(x) (the Bessel function of the second kind, or Neumann function), diverges logarithmically as x → 0. This singularity at the origin determines which solutions are physically acceptable. For problems on a full disk including the center — like a drumhead clamped at its edge — the solution must remain finite at r = 0, so Y_ν is discarded and only J_ν appears. For an annular region that excludes the origin, both J_ν and Y_ν contribute to the general solution. The physical boundary condition, not abstract algebra, makes the choice.
The most practically important property is orthogonality with a weight function. If λ_{ν,m} and λ_{ν,n} are distinct zeros of J_ν(x), then ∫₀^a x J_ν(λ_{ν,m} x/a) J_ν(λ_{ν,n} x/a) dx = 0 for m ≠ n. The extra factor of x in the integrand comes from the cylindrical coordinate area element. This weighted orthogonality enables Fourier-Bessel expansions: any reasonable function on [0, a] can be written as a sum of Bessel functions, exactly as Fourier series expand functions in sines and cosines. In practice, recursion relations J_{ν−1}(x) + J_{ν+1}(x) = (2ν/x)J_ν(x) and tabulated zeros allow computation without rederiving the series every time.