Differential Equations › Section H · Boundary value problems and Fourier series › Topic H.1

Boundary value problems

A guitar string is pinned at both ends and a metal rod has a temperature set at each end, so the conditions arrive at two different places rather than one starting point. Problems like these behave very differently from initial value problems, and that difference is the doorway to Fourier series.

Conditions at two places

Until now, a second order equation came with initial conditions: the value of \(y\) and of \(y'\) at a single point. A boundary value problem (BVP) instead gives conditions at two different points, usually the ends \(x = a\) and \(x = b\) of the interval where the solution lives. Think of a rope tied to two posts: the physics is known at the edges, not at a starting instant. Solving still means finding the general solution and fixing the constants. What changes is how many solutions survive that last step.

Kinds of boundary conditions

This topic works with linear second order equations \(y'' + p(x)y' + q(x)y = g(x)\). The two conditions can fix values, \(y(a) = \alpha\) and \(y(b) = \beta\), which are called Dirichlet conditions. They can fix slopes, \(y'(a) = \alpha\) and \(y'(b) = \beta\), called Neumann conditions. Or they can mix the two, such as \(y(a) = \alpha\) with \(y'(b) = \beta\). In a heat problem a value condition sets the temperature at an end, while \(y' = 0\) says no heat flows through that end, which is what an insulated end means.

Homogeneous boundary value problems

For a BVP, homogeneous means two things at once: the equation has \(g(x) = 0\), and both boundary values are zero. So \(y'' + 9y = 0\) with \(y(0) = 0\) and \(y(\pi) = 0\) is homogeneous, but the same equation with \(y(\pi) = 2\) is not. A homogeneous BVP always has the trivial solution \(y = 0\), since zero satisfies the equation and every zero condition. The interesting question is whether any nonzero solution exists as well.

Worked example: exactly one solution

Solve \(y'' + 9y = 0\) with \(y(0) = 2\) and \(y(\frac{\pi}{6}) = 5\). 1. The characteristic roots are \(\pm 3i\), so \(y = c_1\cos 3x + c_2\sin 3x\). 2. At \(x = 0\): \(c_1 = 2\). 3. At \(x = \frac{\pi}{6}\), the angle is \(\frac{\pi}{2}\), where cosine is 0 and sine is 1, so \(c_2 = 5\). 4. The solution is \(y = 2\cos 3x + 5\sin 3x\). 5. Check: \(y(0) = 2\) and \(y(\frac{\pi}{6}) = 0 + 5 = 5\). Each condition pinned down one constant, so this solution is the only one.

Same equation, new boundary

Keep \(y'' + 9y = 0\) and \(y(0) = 2\), so \(c_1 = 2\) again, but move the second condition to \(x = \frac{\pi}{3}\). There the angle is \(\pi\), and \(y(\frac{\pi}{3}) = -c_1 + 0 \cdot c_2 = -2\) whatever \(c_2\) is. If the condition is \(y(\frac{\pi}{3}) = -2\), it holds automatically, and every function \(y = 2\cos 3x + c\sin 3x\) is a solution: infinitely many. If the condition is \(y(\frac{\pi}{3}) = 4\), it demands \(-2 = 4\), which is false, so there is no solution at all. A small shift in the data moved the problem from one solution to none or infinitely many.

Why uniqueness can fail

The boundary conditions always produce two linear equations in \(c_1\) and \(c_2\). For \(y(0) = \alpha\) and \(y(b) = \beta\) with the equation above, they are

c_1 = \alpha, \qquad c_1\cos 3b + c_2\sin 3b = \beta

When \(\sin 3b \ne 0\), the system has exactly one solution. When \(\sin 3b = 0\), the second equation loses \(c_2\), and the system has either no solution or infinitely many, exactly like a square system \(A\mathbf{c} = \mathbf{b}\) with a zero determinant. Initial conditions never hit this problem for these equations, because the Wronskian of a fundamental set is never zero.

Worked example: slope conditions

Solve \(y'' + 16y = 0\) with \(y'(0) = 8\) and \(y'(\frac{\pi}{8}) = -12\). 1. General solution: \(y = c_1\cos 4x + c_2\sin 4x\). 2. Differentiate: \(y' = -4c_1\sin 4x + 4c_2\cos 4x\). 3. At \(x = 0\): \(4c_2 = 8\), so \(c_2 = 2\). 4. At \(x = \frac{\pi}{8}\) the angle is \(\frac{\pi}{2}\), so \(y' = -4c_1 = -12\) and \(c_1 = 3\). 5. The solution is \(y = 3\cos 4x + 2\sin 4x\). 6. Check: \(y'(0) = 0 + 8 = 8\), and \(y'(\frac{\pi}{8}) = -12 + 0 = -12\).

Worked example: a forced equation

Solve \(y'' + 4y = 8x\) with \(y(0) = 3\) and \(y(\frac{\pi}{4}) = \frac{\pi}{2} - 1\). 1. Complementary solution: \(c_1\cos 2x + c_2\sin 2x\). 2. Undetermined coefficients with \(Y = Ax + B\) gives \(4Ax + 4B = 8x\), so \(Y = 2x\). 3. Then \(y = c_1\cos 2x + c_2\sin 2x + 2x\). 4. At \(x = 0\): \(c_1 = 3\). 5. At \(x = \frac{\pi}{4}\): \(c_2 + \frac{\pi}{2} = \frac{\pi}{2} - 1\), so \(c_2 = -1\). 6. The solution is \(y = 3\cos 2x - \sin 2x + 2x\). The conditions go on the full general solution, never on the complementary part alone.

Homogeneous problems: trivial or not

Take \(y'' + 9y = 0\) with \(y(0) = 0\) and \(y(\pi) = 0\). The first condition gives \(c_1 = 0\), and the second gives \(c_2\sin 3\pi = 0\), which is true for every \(c_2\) because \(\sin 3\pi = 0\). So \(y = c\sin 3x\) is a nonzero solution for any \(c \ne 0\). Change the equation to \(y'' + 5y = 0\) with the same conditions: now \(c_2\sin(\sqrt{5}\,\pi) = 0\), and since \(\sin(\sqrt{5}\,\pi) \approx 0.675\) is not zero, \(c_2 = 0\). Only the trivial solution remains. Which coefficients allow nonzero solutions is the subject of "Eigenvalues and eigenfunctions".

An applied problem: a rod at steady state

A thin rod 2 m long has its ends held at 20 °C and 80 °C. Assume the temperature has stopped changing, the rod's material is uniform, no heat is produced inside it and none escapes from its sides. Then the temperature \(T(x)\) satisfies \(T'' = 0\) with \(T(0) = 20\) and \(T(2) = 80\). Integrating twice gives \(T = c_1 + c_2x\); the conditions give \(c_1 = 20\) and \(c_2 = 30\), so \(T = 20 + 30x\), rising steadily along the rod. If the far end were insulated instead, \(T'(2) = 0\) would force \(c_2 = 0\) and the whole rod would sit at 20 °C.

Common mistakes to avoid

First, assuming a BVP has exactly one solution because the matching initial value problem would; always check for none or infinitely many. Second, calling a problem homogeneous because the equation has zero on the right, while a boundary value is nonzero. Third, applying the boundary conditions to the complementary solution before adding the particular solution. Fourth, evaluating a slope condition with \(y\) instead of \(y'\). Fifth, concluding "no solution" when a condition reduces to a true statement such as \(-2 = -2\); that means the constant is free.

Key takeaways

A boundary value problem fixes conditions at two points, through values, slopes or a mix. It is homogeneous only when the equation and both boundary values are zero, and then \(y = 0\) is always a solution. Applying the conditions gives a linear system for \(c_1\) and \(c_2\): a nonzero determinant means exactly one solution, while a zero determinant means none or infinitely many. Small changes to the interval, the data or the equation can switch between these cases, which is why each BVP must be checked rather than assumed.

Flashcards

Flashcard

Boundary value problem

A differential equation with conditions given at two different points

Usually the two ends of the interval.

Test your understanding

Question 1 of 5

Solve y″ + y = 0 with y(0) = 3 and y(π/2) = −1.

Question 2 of 5

How many solutions does y″ + 16y = 0 with y(0) = 1 and y(π/4) = −1 have?

Question 3 of 5

Which of these is a homogeneous boundary value problem?

Question 4 of 5

A rod 5 m long at steady state has T″ = 0, T(0) = 10 and T(5) = 60. What is the temperature at x = 2?

Question 5 of 5

Why does y″ + 9y = 0 with y(0) = 0 and y(π) = 0 have nonzero solutions?

← Previous topicNext topic →