ODE: Power Series Solutions of ODEs

Power series solutions transform intractable differential equations into manageable algebraic problems, enabling you to tackle models with variable coefficients that defy elementary methods. This technique is indispensable in engineering fields like quantum mechanics, aerodynamics, and structural analysis, where equations describing wave propagation, boundary layers, or stress distributions often lack closed-form solutions. Mastering this approach equips you with a flexible tool to approximate solutions with arbitrary precision near key points.

The Foundation: Ordinary Points and Series Assumptions

A point $x=x_0$ is called an ordinary point of a linear second-order ODE if the coefficient functions, when the equation is written in standard form $y^{\prime \prime }+P(x)y^{\prime }+Q(x)y=0$, are analytic there. In simpler terms, $P(x)$ and $Q(x)$ have convergent power series expansions around $x_0$. This analyticity guarantees that a power series solution exists and converges in some interval around that point. The method begins by assuming a power series solution of the form $y={\sum }_{n=0}^{\infty }{a}_n(x-x_0{)}^n$, where $a_n$ are the unknown coefficients to be determined. For simplicity, we often take $x_0=0$, shifting the variable if necessary. This assumption is not a guess but a systematic exploration, similar to using a Fourier series to represent a periodic function; you are expressing the unknown solution as a sum of known basis functions.

Deriving Recurrence Relations for Coefficients

After substituting the infinite series for $y$, $y^{\prime }$, and $y^{\prime \prime }$ into the differential equation, the core algebraic task is to align powers of $x$. You combine series into a single summation, typically by shifting indices so that all terms share the same power, like $x^n$. This process yields a single power series equal to zero. For the series to be identically zero for all $x$ in the convergence interval, the coefficient of each power $x^n$ must itself be zero. This condition generates a recurrence relation, which is an equation that expresses a given coefficient $a_n$ in terms of previous ones (e.g., $a_{n+2}$ in terms of $a_n$). These relations are the engine of the method, allowing you to compute all coefficients sequentially from one or two initial values, $a_0$ and $a_1$, which are determined by initial conditions. For example, in a simple equation like $y^{\prime \prime }-y=0$, substitution leads to ${\sum }_{n=0}^{\infty }[(n+2)(n+1)a_{n+2}-a_n]x^n=0$, giving the recurrence $a_{n+2}=a_n/((n+2)(n+1))$.

Understanding the Radius of Convergence

The radius of convergence $R$ of the resulting power series solution is not merely a theoretical detail; it defines the interval $(x_0-R,x_0+R)$ where your computed series actually represents the solution to the differential equation. For a series solution about an ordinary point $x_0$, $R$ is at least as large as the distance from $x_0$ to the nearest singularity of $P(x)$ or $Q(x)$ in the complex plane. You can often find $R$ directly from the recurrence relation using the ratio test. Consider the limit $L={\lim }_{n\to \infty }|a_{n+1}/a_n|$; the radius of convergence is $R=1/L$. If the recurrence relation yields coefficients that grow or decay in a predictable pattern, this limit can be evaluated. Ensuring you operate within this radius is critical for any engineering application, as using the series outside it leads to divergent, meaningless results.

Solving Airy's and Hermite's Equations

Airy's equation, $y^{\prime \prime }-xy=0$, models phenomena like diffraction patterns and the turning point in quantum mechanics. With $x_0=0$ as an ordinary point, assume $y={\sum }_{n=0}^{\infty }{a}_n{x}^n$. Substituting and aligning terms gives: $$\sum _{n=0}^{\infty }(n+2)(n+1)a_{n+2}{x}^n-\sum _{n=0}^{\infty }{a}_n{x}^{n+1}=0.$$ To combine, shift the index in the second series by letting $m=n+1$, or $n=m-1$. This yields: $$\sum _{n=0}^{\infty }(n+2)(n+1)a_{n+2}{x}^n-\sum _{m=1}^{\infty }{a}_{m-1}{x}^m=0.$$ Rewriting the second sum to start at $n=1$ and combining, we get for $n=0$: $2a_2=0$, so $a_2=0$. For $n\ge 1$: $$(n+2)(n+1)a_{n+2}-a_{n-1}=0\text{or}{a}_{n+2}=\frac{a_{n-1}}{(n+2)(n+1)}.$$ This three-term recurrence produces two independent solutions based on $a_0$ and $a_1$. The radius of convergence is infinite because $P(x)=0$ and $Q(x)=-x$ are entire functions, so the series converges for all $x$.

Hermite's equation, $y^{\prime \prime }-2x{y}^{\prime }+2\nu y=0$ (where $\nu$ is a constant), is central to the quantum harmonic oscillator. Assuming $y={\sum }_{n=0}^{\infty }{a}_n{x}^n$ and substituting: $$\sum _{n=0}^{\infty }(n+2)(n+1)a_{n+2}{x}^n-2\sum _{n=0}^{\infty }n{a}_n{x}^n+2\nu \sum _{n=0}^{\infty }{a}_n{x}^n=0.$$ Combining coefficients for $x^n$ gives: $$(n+2)(n+1)a_{n+2}-2n{a}_n+2\nu a_n=0.$$ This simplifies to the recurrence relation: $$a_{n+2}=\frac{2n-2\nu }{(n+2)(n+1)}{a}_n.$$ For non-negative integer $\nu$, this relation terminates one series, yielding polynomial solutions known as Hermite polynomials, crucial in probability and physics. The radius of convergence is again infinite.

When Power Series Methods Are Necessary Versus Optional

Power series methods are necessary when dealing with linear ODEs with variable coefficients that are not constant or separable, and where no elementary solution exists in terms of standard functions. This includes equations like Airy's, Bessel's, or Legendre's, which frequently arise in engineering models involving cylindrical symmetry, heat conduction, or wave equations. They are also essential when solving initial value problems where the solution must be expressed locally near an ordinary point.

Conversely, these methods are optional when the ODE admits simpler techniques. For instance, equations with constant coefficients are efficiently solved via characteristic equations, and first-order linear equations yield to integrating factors. In engineering practice, you might choose a power series approach even when other methods work if you need a series approximation for computational implementation or perturbation analysis. The decision hinges on the equation's form, the desired solution type, and application constraints.

Common Pitfalls

1. Ignoring the radius of convergence: Using a power series solution beyond its interval of convergence is a critical error. For example, if $P(x)$ has a singularity at $x=1$, a series about $x=0$ may only be valid for $|x|<1$. Always determine $R$ from the recurrence relation or the locations of singularities.

Correction: Explicitly calculate the radius using the ratio test on the coefficients or identify singular points of $P(x)$ and $Q(x)$. State the interval of validity alongside your solution.

1. Incorrect index shifting during substitution: Misaligning powers of $x$ when combining series leads to erroneous recurrence relations. A common mistake is failing to adjust summation indices so that all series sum over the same power.

Correction: Perform index shifts methodically. Write out each series, shift indices so that exponents match, and then combine. For example, if you have a term like $\sum a_n{x}^{n+1}$, set $m=n+1$ to rewrite it as $\sum a_{m-1}{x}^m$.

1. Overlooking two independent solutions: From a second-order ODE, you expect two linearly independent solutions. The recurrence relation often splits into even and odd indexed coefficients based on $a_0$ and $a_1$.

Correction: Express the general solution as $y=a_0{y}_1(x)+a_1{y}_2(x)$, where $y_1$ and $y_2$ are the series solutions obtained by setting $a_0=1,a_1=0$ and vice versa. This captures the full solution space.

1. Forcing termination incorrectly: In equations like Hermite's, polynomial solutions occur only for specific parameter values. Assuming termination for arbitrary parameters is a mistake.

Correction: Analyze the recurrence relation: if $a_{n+2}=0$ for some $n$, it imposes a condition on the parameter (e.g., $\nu$ must be a non-negative integer for Hermite's). Only then does the series reduce to a polynomial.

Summary

• Ordinary points are where coefficient functions are analytic, guaranteeing a convergent power series solution of the form $y=\sum a_n(x-x_0{)}^n$.
• Substituting the series into the ODE leads to recurrence relations for the coefficients, determined by setting the coefficient of each power $x^n$ to zero.
• The radius of convergence is crucial and is at least the distance to the nearest singularity; it ensures the series solution is valid within a specific interval.
• Airy's equation ($y^{\prime \prime }-xy=0$) and Hermite's equation ($y^{\prime \prime }-2x{y}^{\prime }+2\nu y=0$) exemplify the method, yielding series and polynomial solutions with infinite radii due to entire coefficients.
• Power series methods are necessary for variable-coefficient ODEs without elementary solutions but optional when simpler techniques like characteristic equations apply; the choice depends on the problem context and solution needs.