Differential Equations › Section F · Series solutions › Topic F.5
Special functions in applications
Series solutions of a few important equations come up so often that they have names, tables and software, just like sine and cosine. Knowing their key properties lets you use them as confidently as any familiar function.
Named solutions
Sine and cosine are, among other things, the named solutions of \(y'' + y = 0\). When an equation with variable coefficients keeps appearing in physics and engineering, its series solutions earn names too: Legendre polynomials, Hermite and Laguerre polynomials, Bessel functions and Airy functions. Each family comes with recurrences that generate it, orthogonality relations, known zeros and reliable software. Libraries evaluate them with series for small arguments, other approximations for large ones and recurrences in between, so engineers rarely sum a series by hand. What you need is to recognise each family and to know what its properties let you do.
Legendre polynomials
Legendre's equation, met in the topic Series solutions near ordinary points, is
Substituting \(y = \sum c_k x^k\) gives the recurrence
When \(\ell\) is a whole number, the numerator is zero at \(k = \ell\), so the series of the same parity as \(\ell\) stops: it is a polynomial of degree \(\ell\). Scaled so that its value at \(x = 1\) is 1, it is the Legendre polynomial \(P_\ell(x)\). The first two are \(P_0 = 1\) and \(P_1 = x\). They arise in problems with spherical symmetry, where \(x = \cos\theta\).
Worked example: P₂ from a stopping series
Find \(P_2\). 1. With \(\ell = 2\), \(\ell(\ell+1) = 6\), and the even series starts from \(c_0\). 2. \(k = 0\): \(c_2 = \frac{0 - 6}{2}c_0\), so \(c_2 = -3c_0\). 3. \(k = 2\): the numerator is \(6 - 6 = 0\), so \(c_4 = 0\), and every later even coefficient is 0. 4. So \(y = c_0(1 - 3x^2)\). At \(x = 1\) this is \(-2c_0\), so choose \(c_0 = -\frac{1}{2}\):
5. Check in the equation: \(y' = 3x\) and \(y'' = 3\), so the left side is
Worked example: building P₃ by recurrence
Solving a new series for every degree is slow. Legendre polynomials satisfy a three-term recurrence that builds each from the previous two:
Find \(P_3\). 1. Take \(n = 2\): \(3P_3 = 5xP_2 - 2P_1\). 2. Substitute:
3. Divide by 3:
4. Check: \(P_3(1) = \frac{5 - 3}{2} = 1\), as the scaling requires, and \(P_3\) is odd, matching the parity of \(\ell = 3\).
Generating functions and multipoles
A generating function packs a whole family into one expression. For Legendre polynomials,
Expanding the left side with the binomial series gives \(1 + xt + P_2(x)t^2 + \cdots\), matching the polynomials above. Physically, this is the distance formula in disguise. The potential of a point charge at distance \(d\) from the origin, measured at distance \(r > d\), contains the reciprocal of the distance between the two points,
which expands into powers of \(\frac{d}{r}\) with coefficients \(P_n(\cos\theta)\). That is the multipole expansion used for gravity fields of planets and for molecules.
Orthogonality and Gaussian quadrature
Legendre polynomials are orthogonal on \([-1, 1]\):
Hermite polynomials are orthogonal on the whole real line with the weight \(e^{-x^2}\), and Laguerre polynomials on \([0, \infty)\) with weight \(e^{-x}\). Like the sines in a Fourier series, such families can be used to expand other functions. They also power Gaussian quadrature: an \(n\)-point rule with nodes at the zeros of \(P_n\) integrates every polynomial of degree up to \(2n - 1\) exactly. With nodes \(\pm\frac{1}{\sqrt{3}}\), \(f(x) = x^3 + x^2\): the cubed terms cancel and the squared terms give \(\frac{2}{3}\), the exact integral over \([-1, 1]\).
Hermite and Laguerre polynomials
Hermite's equation \(y'' - 2xy' + 2ny = 0\) has a polynomial solution for each whole number \(n\): \(H_0 = 1\), \(H_1 = 2x\), \(H_2 = 4x^2 - 2\) and \(H_3 = 8x^3 - 12x\), generated by \(H_{n+1} = 2xH_n - 2nH_{n-1}\). In suitable units, \(H_n(x)e^{-x^2/2}\) are the states of the quantum harmonic oscillator, the standard model of a vibrating molecule. Laguerre polynomials, in a generalised form, describe the radial part of the hydrogen atom's states. A warning about conventions: probability texts use a rescaled Hermite family with \(\mathrm{He}_2 = x^2 - 1\), so check which version a book or library uses before mixing sources.
Bessel functions in practice
The Bessel function \(J_0\) from the topic The Frobenius method looks like a cosine whose amplitude slowly shrinks, roughly in proportion to \(\frac{1}{\sqrt{x}}\). Its zeros are not evenly spaced: the first three are about 2.405, 5.520 and 8.654. For a circular drumhead clamped at its rim, the symmetric modes vibrate at frequencies proportional to these zeros. The second is about 2.30 times the first, not 2 times, so a drum's overtones are not whole-number multiples of its fundamental the way a string's are. That is why most drums sound less clearly pitched than a guitar. Bessel functions also model heat flow in cylinders and waves in cables.
Truncated series as approximations
Cutting a series off gives a polynomial that is accurate near its centre. An undamped pendulum of length \(L\) obeys
Replacing \(\sin\theta\) by \(\theta\) gives simple harmonic motion with period \(T_0 = 2\pi\sqrt{L/g}\). Keeping the next term, \(\theta - \frac{\theta^3}{6}\), and working carefully gives the correction
for amplitude \(\theta_0\) in radians. At \(\theta_0 = 0.5\) this predicts 1.56% longer, against a true 1.59%. At \(\theta_0 = 1\), it gives 6.25% against a true 6.63%: the further from the centre, the worse a truncated series does.
Common mistakes to avoid
First, forgetting the scaling: the stopping series for \(\ell = 2\) is \(1 - 3x^2\), but \(P_2\) is that times \(-\frac{1}{2}\), chosen so \(P_2(1) = 1\). Second, mixing conventions, such as the two Hermite families. Third, dropping the weight from an orthogonality integral: Hermite polynomials are orthogonal only with \(e^{-x^2}\) included. Fourth, misusing the recurrence index: to get \(P_3\), use \(n = 2\), not \(n = 3\). Fifth, trusting a truncated series far from its centre, as the large-swing pendulum shows.
Key takeaways
Special functions are named solutions of recurring equations. Legendre's equation has polynomial solutions \(P_\ell\) when \(\ell\) is a whole number, scaled so \(P_\ell(1) = 1\) and built quickly by a three-term recurrence; their generating function gives the multipole expansion. Legendre, Hermite and Laguerre polynomials are orthogonal with their own weights and intervals, which underlies series expansions and Gaussian quadrature. Hermite polynomials describe the quantum oscillator, and the uneven zeros of \(J_0\) explain a drum's overtones. Truncated series are good local approximations only.
Flashcards
Flashcard
A named solution of an important differential equation
Test your understanding
Question 1 of 5
Using (n + 1)Pₙ₊₁ = (2n + 1)xPₙ - nPₙ₋₁ with P₁ = x and P₂ = (3x² - 1)/2, what is P₃?
Question 2 of 5
Using Hₙ₊₁ = 2xHₙ - 2nHₙ₋₁ with H₀ = 1 and H₁ = 2x, what is H₂?
Question 3 of 5
With which weight and interval are the Hermite polynomials orthogonal?
Question 4 of 5
Using T ≈ T₀(1 + θ₀²/16), about how much longer than T₀ is the period of a pendulum swinging with amplitude 0.4 rad?
Question 5 of 5
Why does a circular drum's overtone series not match a string's?
