ODE: Boundary Value Problems and Eigenvalues

Solving partial differential equations (PDEs) that model heat transfer, vibration, and wave propagation is fundamental to engineering. When you apply the separation of variables technique to these PDEs, you consistently reduce them to a special class of ordinary differential equation (ODE) problems defined not by initial conditions, but by conditions at two different points in space. Mastering these boundary value problems (BVPs) and their associated eigenvalues and eigenfunctions is the key to constructing complete, practical solutions for complex physical systems.

From Separation of Variables to Eigenvalue Problems

Consider the classic one-dimensional heat equation for a rod of length $L$: $u_t=\alpha u_{xx}$. Assuming a product solution $u(x,t)=X(x)T(t)$ leads to two ODEs: $$\frac{T^{\prime }}{\alpha T}=\frac{X^{\prime \prime }}{X}=-\lambda .$$ The constant $-\lambda$ is the separation constant. The spatial ODE becomes $X^{\prime \prime }+\lambda X=0$. However, this ODE has infinitely many possible solutions. The physical constraints—the boundary conditions—determine which solutions are valid. For a rod with both ends held at zero temperature, we have $u(0,t)=0$ and $u(L,t)=0$, which translates to $X(0)=0$ and $X(L)=0$.

This combination, the ODE $X^{\prime \prime }+\lambda X=0$ together with the boundary conditions $X(0)=0$ and $X(L)=0$, is a boundary value problem. Crucially, this BVP only has non-trivial (i.e., non-zero) solutions for specific, discrete values of the parameter $\lambda$. These special values are the eigenvalues, and the corresponding non-zero solutions $X(x)$ are the eigenfunctions. Finding them is the first critical step.

Solving for Eigenvalues and Eigenfunctions

The solution to $X^{\prime \prime }+\lambda X=0$ depends on the sign of $\lambda$. We must test all cases, but the boundary conditions will force the correct one.

• Case $\lambda <0$: The general solution is $X(x)=c_1{e}^{\sqrt{-\lambda }x}+c_2{e}^{-\sqrt{-\lambda }x}$. Applying $X(0)=0$ and $X(L)=0$ leads only to the trivial solution $c_1=c_2=0$. No eigenvalues here.
• Case $\lambda =0$: The solution is $X(x)=c_{1}x+c_2$. The boundary conditions again force $c_1=c_2=0$. $\lambda =0$ is not an eigenvalue.
• Case $\lambda >0$: Let $\lambda ={\omega }^2$. The general solution is $X(x)=c_1\cos (\omega x)+c_2\sin (\omega x)$. Applying $X(0)=0$ gives $c_1=0$. The condition $X(L)=0$ then becomes $c_2\sin (\omega L)=0$. To avoid the trivial solution, we require $c_2\ne 0$, which forces $\sin (\omega L)=0$. Therefore, $\omega L=n\pi$ for $n=1,2,3,...$.

Thus, the eigenvalues are ${\lambda }_n={\omega }_n^2=(\frac{n\pi }{L}{)}^2$, and the corresponding eigenfunctions are $X_n(x)=\sin (\frac{n\pi x}{L})$, for $n=1,2,3,...$. Notice the infinite, discrete set. Each eigenvalue-eigenfunction pair represents a fundamental mode of the system—a specific spatial pattern in which the rod can heat or vibrate.

The Sturm-Liouville Framework

The problem above is a canonical example of a regular Sturm-Liouville problem. This powerful general framework has the form: $$\frac{d}{dx}\left[p(x)\frac{dy}{dx}\right]+q(x)y+\lambda w(x)y=0,a<x<b,$$ with boundary conditions like ${\alpha }_{1}y(a)+{\alpha }_2{y}^{\prime }(a)=0$ and ${\beta }_{1}y(b)+{\beta }_2{y}^{\prime }(b)=0$. Here, $p(x),q(x),$ and the weight function $w(x)>0$ are given, and $\lambda$ is the eigenvalue parameter.

Our example $X^{\prime \prime }+\lambda X=0$ fits this form with $p(x)=1$, $q(x)=0$, and $w(x)=1$. The theory of Sturm-Liouville problems guarantees that for regular problems, there exists an infinite sequence of real eigenvalues ${\lambda }_1<{\lambda }_2<{\lambda }_3<...$ tending to infinity. Each eigenvalue ${\lambda }_n$ has a unique (up to a constant multiple) eigenfunction $y_n(x)$. This structured outcome is what makes the separation of variables method work predictably.

Orthogonality of Eigenfunctions

A property of paramount importance for Sturm-Liouville eigenfunctions is orthogonality with respect to the weight function $w(x)$. Two different eigenfunctions $y_m(x)$ and $y_n(x)$ corresponding to distinct eigenvalues ${\lambda }_m\ne {\lambda }_n$ satisfy: $${\int }_a^{b}w(x)y_m(x)y_n(x)dx=0.$$

In our sine eigenfunction example ($w(x)=1$), this is the familiar Fourier sine series orthogonality: $${\int }_0^L\sin \left(\frac{m\pi x}{L}\right)\sin \left(\frac{n\pi x}{L}\right)dx=0,\text{for }m\ne n.$$ This orthogonality is not just a mathematical curiosity; it is the engine that allows us to break down complex initial shapes into simpler components.

Expanding Initial Conditions in an Eigenfunction Series

The final step in solving the original PDE is to satisfy the initial condition, e.g., $u(x,0)=f(x)$, the initial temperature distribution. Our separated solution gives a family of product solutions for each eigenvalue: $u_n(x,t)=X_n(x)T_n(t)$. By the linearity of the PDE, the general solution is a superposition (infinite series) of these: $$u(x,t)=\sum _{n=1}^{\infty }{c}_n{X}_n(x)T_n(t).$$ At time $t=0$, $T_n(0)=1$, so the condition becomes: $$f(x)=u(x,0)=\sum _{n=1}^{\infty }{c}_n{X}_n(x).$$ We must represent the arbitrary function $f(x)$ as a series of our eigenfunctions $X_n(x)$. This is precisely where orthogonality is used. To find the coefficient $c_m$, multiply both sides by $w(x)X_m(x)$ and integrate over $[a,b]$: $${\int }_a^{b}w(x)f(x)X_m(x)dx=\sum _{n=1}^{\infty }{c}_n{\int }_a^{b}w(x)X_n(x)X_m(x)dx.$$ By orthogonality, every term in the sum on the right is zero except when $n=m$. This collapses the infinite series to a single term: $${\int }_a^{b}w(x)f(x)X_m(x)dx=c_m{\int }_a^{b}w(x)[X_m(x){]}^{2}dx.$$ Solving for $c_m$ gives the formula: $$c_m=\frac{{\int }_a^{b}w(x)f(x)X_m(x)dx}{{\int }_a^{b}w(x)[X_m(x){]}^{2}dx}.$$ For the sine eigenfunctions, this yields the classic Fourier sine coefficients: $c_n=\frac{2}{L}{\int }_0^{L}f(x)\sin (\frac{n\pi x}{L})dx$.

Common Pitfalls

1. Misidentifying the Eigenvalue Parameter: When separating variables, consistently define your separation constant (e.g., $-\lambda$, $\lambda$, $k^2$). The sign convention must be maintained throughout the problem. A common error is to set $X^{\prime \prime }/X=\lambda$ and then later use a solution form for $-\lambda$, leading to sign confusion and missing valid eigenvalues.

• Correction: Stick rigidly to your chosen definition. For second-order spatial problems where oscillatory solutions are physically expected, starting with $X^{\prime \prime }+\lambda X=0$ (implying $\lambda >0$ for oscillations) is often most efficient.

1. Ignoring the Trivial Solution in the Eigenvalue Search: It's tempting to immediately apply boundary conditions to the general solution without considering that they might force all constants to zero. The entire purpose is to find parameters $\lambda$ that allow a non-trivial solution.

• Correction: Explicitly state the condition for a non-trivial solution. For example, from $c_2\sin (\omega L)=0$, deduce that the necessary condition is $\sin (\omega L)=0$, because $c_2=0$ would give the trivial solution.

1. Forgetting the Weight Function in Orthogonality and Series Expansion: In a general Sturm-Liouville problem, orthogonality is defined with respect to $w(x)$. Using the simple integral $\int y_m{y}_{n}dx=0$ when $w(x)\ne 1$ is incorrect and will yield wrong expansion coefficients.

• Correction: Always identify $p(x)$, $q(x)$, and $w(x)$ from the standard Sturm-Liouville form. The coefficient formula must include $w(x)$ in both the numerator and denominator.

1. Incorrectly Handling Different Boundary Condition Types: Dirichlet ($y=0$), Neumann ($y^{\prime }=0$), and Robin ($\alpha y+\beta y^{\prime }=0$) conditions lead to different eigenfunctions (sines, cosines, or mixes). Assuming sines for every fixed-end problem is a mistake.

• Correction: Solve the characteristic equation systematically for each case dictated by the boundary conditions. For a Neumann condition at $x=0$ ($X^{\prime }(0)=0$), the cosine term survives, changing the fundamental set of eigenfunctions.

Summary

• Boundary value problems arise naturally when applying separation of variables to PDEs. They consist of an ODE and spatial boundary conditions.
• Non-trivial solutions to a BVP exist only for specific eigenvalues ${\lambda }_n$. The corresponding solutions are eigenfunctions, which represent the fundamental spatial modes of the physical system.
• Sturm-Liouville theory provides a unified framework for these problems, guaranteeing a discrete, increasing set of real eigenvalues and orthogonal eigenfunctions.
• Orthogonality of eigenfunctions with respect to a weight function $w(x)$ is the critical property that allows an arbitrary function (like an initial condition) to be expressed as an infinite eigenfunction series.
• The complete solution to the original PDE is constructed as a superposition of all modal solutions, with series coefficients determined by projecting the initial condition onto each eigenfunction using the orthogonality relations.