Separation of Variables: Laplace, Heat and Wave Equations
CSIR NET PDE: separation of variables, eigenvalue problems, Fourier series solutions of heat, wave and Laplace equations, d'Alembert, maximum principles.
- 20 min
On this page · 14 sections
Prerequisites
- Classification of Second-Order PDEs: which data suit the heat (parabolic), wave (hyperbolic) and Laplace (elliptic) equations.
- Higher-Order PDEs with Constant Coefficients: the exponential solutions \( e^{hx + ky} \) that separation produces.
- Fourier sine and cosine series and their convergence, and term-by-term operations under uniform convergence, from Sequences and Series of Functions.
The Hook: Fourier’s Rod
The Method in Five Steps
graph TD
A["Linear homogeneous PDE
+ homogeneous boundary conditions"] --> B["Try u = X(x) T(t)"]
B --> C["Divide: X''/X = (something in t) = -lambda"]
C --> D["Eigenvalue problem for X
with the boundary conditions"]
D --> E["Eigenvalues lambda_n, eigenfunctions X_n"]
E --> F["Solve the T (or Y) ODE for each lambda_n"]
F --> G["Superpose: u = sum c_n X_n T_n"]
G --> H["Fix c_n from the remaining data
by Fourier coefficients"]Requirements. The PDE must be linear and homogeneous, and the boundary conditions on the variable you separate first must be homogeneous (zero), so that the products satisfying them form a vector space and can be superposed. Non-zero boundary values are first removed by subtracting a steady state.
The eigenvalue problems you need
\( X'' + \lambda X = 0 \) on \( 0 < x < L \):
| Boundary conditions | Eigenvalues \( \lambda_n \) | Eigenfunctions | \( n \) |
|---|---|---|---|
| \( X(0) = X(L) = 0 \) (Dirichlet) | \( \left(\frac{n\pi}{L}\right)^2 \) | \( \sin\frac{n\pi x}{L} \) | \( 1, 2, \dots \) |
| \( X'(0) = X'(L) = 0 \) (Neumann) | \( \left(\frac{n\pi}{L}\right)^2 \) | \( \cos\frac{n\pi x}{L} \) | \( 0, 1, 2, \dots \) |
| \( X(0) = 0 \), \( X'(L) = 0 \) (mixed) | \( \left(\frac{(2n - 1)\pi}{2L}\right)^2 \) | \( \sin\frac{(2n - 1)\pi x}{2L} \) | \( 1, 2, \dots \) |
| periodic, period \( 2L \) (\( -L < x < L \)) | \( \left(\frac{n\pi}{L}\right)^2 \) | \( \cos\frac{n\pi x}{L} \), \( \sin\frac{n\pi x}{L} \) | \( 0, 1, 2, \dots \) |
Fourier coefficients on \( (0, L) \).
$$b_n = \frac{2}{L}\int_0^L f(x)\sin\frac{n\pi x}{L}\,dx, \qquad a_n = \frac{2}{L}\int_0^L f(x)\cos\frac{n\pi x}{L}\,dx, \qquad \frac{a_0}{2} = \frac{1}{L}\int_0^L f(x)\,dx.$$The Heat Equation
$$u_t = k\,u_{xx}, \quad 0 < x < L,\ t > 0, \qquad k > 0.$$Dirichlet ends (ends held at 0)
With \( u(0, t) = u(L, t) = 0 \), \( u(x, 0) = f(x) \): \( u = X(x)T(t) \) gives \( \frac{X''}{X} = \frac{T'}{kT} = -\lambda \), so \( X_n = \sin\frac{n\pi x}{L} \), \( T_n = e^{-k(n\pi/L)^2 t} \), and
$$\boxed{u(x, t) = \sum_{n=1}^{\infty} b_n\, e^{-k n^2\pi^2 t/L^2}\sin\frac{n\pi x}{L}}, \qquad b_n = \frac{2}{L}\int_0^L f\sin\frac{n\pi x}{L}\,dx.$$Insulated ends
With \( u_x(0, t) = u_x(L, t) = 0 \): cosine eigenfunctions including \( n = 0 \):
$$u(x, t) = \frac{a_0}{2} + \sum_{n=1}^{\infty} a_n\,e^{-kn^2\pi^2 t/L^2}\cos\frac{n\pi x}{L} \ \longrightarrow\ \frac{a_0}{2} = \frac{1}{L}\int_0^L f\,dx \quad (t \to \infty).$$No heat leaves the rod, so the total heat \( \int_0^L u\,dx \) is constant and the temperature tends to the average of the initial temperature.
Non-zero end temperatures
For \( u(0, t) = T_1 \), \( u(L, t) = T_2 \), subtract the steady state \( w(x) = T_1 + (T_2 - T_1)\frac{x}{L} \) (the solution of \( w'' = 0 \) with these ends). Then \( v = u - w \) has zero end values and initial data \( f - w \), and \( u \to w \) as \( t \to \infty \).
Qualitative facts
| Property | Statement |
|---|---|
| Decay | with Dirichlet ends, \( u \to 0 \) like \( e^{-k\pi^2 t/L^2} \) (the slowest mode, \( n = 1 \)) |
| Smoothing | for \( t > 0 \) the factors \( e^{-kn^2\pi^2 t/L^2} \) make the series converge with all derivatives: \( u \) is \( C^\infty \) in \( x \) even if \( f \) is only piecewise continuous |
| Maximum principle | the maximum of \( u \) on \( [0, L] \times [0, T] \) is attained at \( t = 0 \) or on \( x = 0, L \) |
| Uniqueness | from the maximum principle, or from \( \frac{d}{dt}\int_0^L w^2\,dx = -2k\int_0^L w_x^2\,dx \le 0 \) for the difference \( w \) of two solutions |
| Backward in time | ill-posed: modes grow like \( e^{+n^2 \lvert t \rvert} \) |
The Wave Equation
$$u_{tt} = c^2 u_{xx}, \quad 0 < x < L,\ t > 0.$$String with fixed ends
With \( u(0, t) = u(L, t) = 0 \), \( u(x, 0) = f(x) \), \( u_t(x, 0) = g(x) \): \( \frac{X''}{X} = \frac{T''}{c^2T} = -\lambda \) gives sine eigenfunctions and \( T'' + c^2\lambda_n T = 0 \), so
$$\boxed{u(x, t) = \sum_{n=1}^{\infty}\left(A_n\cos\frac{n\pi ct}{L} + B_n\sin\frac{n\pi ct}{L}\right)\sin\frac{n\pi x}{L}}$$$$A_n = \frac{2}{L}\int_0^L f\sin\frac{n\pi x}{L}\,dx, \qquad B_n = \frac{2}{n\pi c}\int_0^L g\sin\frac{n\pi x}{L}\,dx.$$(\( B_n \) carries the extra factor \( \frac{L}{n\pi c} \) because \( u_t(x, 0) = \sum B_n\frac{n\pi c}{L}\sin\frac{n\pi x}{L} \).)
| Quantity | Value |
|---|---|
| angular frequency of mode \( n \) | \( \omega_n = \frac{n\pi c}{L} \) |
| fundamental frequency | \( \frac{c}{2L} \) cycles per unit time |
| period of the motion | \( \frac{2L}{c} \) (every mode returns) |
| normal modes | \( \sin\frac{n\pi x}{L}\cos\omega_n t \), nodes at \( x = \frac{jL}{n} \) |
| energy | \( E = \frac{1}{2}\int_0^L \left(u_t^2 + c^2 u_x^2\right)dx \), constant in time |
d’Alembert’s formula
On the whole line (\( -\infty < x < \infty \)), the factorisation \( (\partial_t - c\partial_x)(\partial_t + c\partial_x)u = 0 \) gives \( u = F(x - ct) + G(x + ct) \), and fitting the data:
$$\boxed{u(x, t) = \frac{f(x - ct) + f(x + ct)}{2} + \frac{1}{2c}\int_{x - ct}^{x + ct} g(s)\,ds}$$- Domain of dependence: \( u(x_0, t_0) \) depends only on data in \( [x_0 - ct_0, x_0 + ct_0] \). Signals travel at speed \( c \).
- No smoothing: a kink in \( f \) travels along the characteristics \( x \pm ct = \text{const} \) for ever.
- For the fixed-end string, extend \( f, g \) as odd, \( 2L \)-periodic functions and apply d’Alembert; this reproduces the Fourier series.
Laplace’s Equation
$$\Delta u = u_{xx} + u_{yy} = 0.$$Solutions are harmonic. There is no time; the data are boundary values.
Rectangle
On \( 0 < x < a \), \( 0 < y < b \), with \( u = 0 \) on \( x = 0 \), \( x = a \), \( y = 0 \), and \( u(x, b) = f(x) \): \( X'' + \lambda X = 0 \) with Dirichlet conditions, then \( Y'' - \lambda_n Y = 0 \) with \( Y(0) = 0 \), so \( Y_n = \sinh\frac{n\pi y}{a} \):
$$\boxed{u(x, y) = \sum_{n=1}^{\infty} c_n\sin\frac{n\pi x}{a}\,\sinh\frac{n\pi y}{a}}, \qquad c_n\sinh\frac{n\pi b}{a} = \frac{2}{a}\int_0^a f\sin\frac{n\pi x}{a}\,dx.$$If several sides carry non-zero data, solve one problem per side (zero on the other three) and add: superposition.
Disc
In polar coordinates \( u_{rr} + \frac{1}{r}u_r + \frac{1}{r^2}u_{\theta\theta} = 0 \). Separating \( u = R(r)\Theta(\theta) \): \( \Theta'' + n^2\Theta = 0 \) with \( 2\pi \)-periodicity forces integer \( n \); then \( r^2R'' + rR' - n^2R = 0 \) (Euler) gives \( R = r^n, r^{-n} \) (\( n \ge 1 \)) and \( 1, \log r \) (\( n = 0 \)).
| Region | Keep | Solution |
|---|---|---|
| inside \( r < \rho \) | bounded at 0: \( r^n \), \( 1 \) | \( \frac{a_0}{2} + \sum \left(\frac{r}{\rho}\right)^n(a_n\cos n\theta + b_n\sin n\theta) \) |
| outside \( r > \rho \) | bounded at \( \infty \): \( r^{-n} \), \( 1 \) | \( \frac{a_0}{2} + \sum \left(\frac{\rho}{r}\right)^n(a_n\cos n\theta + b_n\sin n\theta) \) |
| annulus | all four kinds, including \( \log r \) | fix both boundary series |
Here \( a_n, b_n \) are the Fourier coefficients of the boundary data \( f(\theta) \). Summing the series gives Poisson’s integral for the disc of radius \( \rho \):
$$u(r, \theta) = \frac{1}{2\pi}\int_0^{2\pi}\frac{\rho^2 - r^2}{\rho^2 - 2\rho r\cos(\theta - \phi) + r^2}\,f(\phi)\,d\phi.$$Properties of harmonic functions
| Property | Statement |
|---|---|
| Mean value | \( u(\text{centre}) = \) average of \( u \) over any circle (or disc) inside the domain |
| Maximum principle | a non-constant harmonic function on a bounded connected domain attains its max and min only on the boundary |
| Uniqueness | the Dirichlet problem has at most one solution (difference is harmonic with zero boundary values) |
| Neumann problem | solvable only if \( \oint \frac{\partial u}{\partial n}\,ds = 0 \); solution unique up to a constant |
| Smoothness | harmonic functions are \( C^\infty \), in fact real analytic |
The Three Equations Side by Side
| Heat \( u_t = ku_{xx} \) | Wave \( u_{tt} = c^2u_{xx} \) | Laplace \( u_{xx} + u_{yy} = 0 \) | |
|---|---|---|---|
| Type | parabolic | hyperbolic | elliptic |
| Second ODE | \( T' = -k\lambda T \) | \( T'' = -c^2\lambda T \) | \( Y'' = \lambda Y \) |
| Its solutions | \( e^{-k\lambda t} \) | \( \cos, \sin(c\sqrt{\lambda}\,t) \) | \( \sinh, \cosh(\sqrt{\lambda}\,y) \) |
| Data needed | \( u(x, 0) \) + boundary | \( u(x, 0) \), \( u_t(x, 0) \) + boundary | boundary only |
| Long-time / interior | decays to steady state | oscillates for ever | determined by all boundary data |
| Conserved / monotone | \( \int u^2 \) decreases | energy constant | mean value property |
| Regularity | smooths instantly | keeps kinks | always smooth inside |
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Dirichlet heat: \( u \to 0 \) | Neumann heat: \( u \to \) average of \( f \) | |
| slowest heat mode \( e^{-k\pi^2t/L^2} \) | \( e^{-\pi^2 t} \) (only when \( k = 1 \), \( L = 1 \)) | |
| wave period \( \frac{2L}{c} \) | fundamental angular frequency \( \frac{\pi c}{L} \) | |
| \( A_n \) from \( f \) (no extra factor) | \( B_n \) from \( g \) (divide by \( \frac{n\pi c}{L} \)) | |
| interior disc: \( r^n \) | exterior disc: \( r^{-n} \) | |
| maximum principle (max on boundary) | mean value property (centre value = average) |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. In Level 3, as in Part C of the Mathematical Sciences paper, one or more options may be correct.
Level 1: Foundation
- Find the eigenvalues of \( X'' + \lambda X = 0 \), \( X(0) = X(\pi) = 0 \).
- What is the steady-state temperature of \( u_t = u_{xx} \) on \( (0, 2) \) with \( u(0, t) = 10 \), \( u(2, t) = 30 \)?
- State the mean value property of harmonic functions.
Answers
- \( \lambda_n = n^2 \), \( n = 1, 2, \dots \), eigenfunctions \( \sin nx \).
- \( w(x) = 10 + 10x \).
- If \( u \) is harmonic on a domain containing the closed disc of radius \( r \) about \( z_0 \), then \( u(z_0) = \frac{1}{2\pi}\int_0^{2\pi}u(z_0 + re^{i\theta})\,d\theta \).
Level 2: Part B Style
Practice Q1. Let \( u \) solve \( u_t = u_{xx} \), \( 0 < x < \pi \), \( u(0, t) = u(\pi, t) = 0 \), \( u(x, 0) = 3\sin 2x - \sin 5x \). Then \( \lim_{t\to\infty} e^{4t}\,u\!\left(\frac{\pi}{4}, t\right) \) equals:
(a) 0 (b) 1 (c) 3 (d) \( \infty \)
Answer: (c)
\( u = 3e^{-4t}\sin 2x - e^{-25t}\sin 5x \). Then \( e^{4t}u\!\left(\frac{\pi}{4}, t\right) = 3\sin\frac{\pi}{2} - e^{-21t}\sin\frac{5\pi}{4} \to 3 \).
Practice Q2. Let \( u \) solve \( u_t = u_{xx} \), \( 0 < x < \pi \), \( u_x(0, t) = u_x(\pi, t) = 0 \), \( u(x, 0) = x \). Then \( \lim_{t\to\infty}u(x, t) \) is:
(a) 0 (b) \( \frac{\pi}{2} \) (c) \( x \) (d) \( \pi \)
Answer: (b)
Insulated ends: the limit is the average \( \frac{1}{\pi}\int_0^\pi x\,dx = \frac{\pi}{2} \).
Practice Q3. Let \( u \) solve \( u_{tt} = 4u_{xx} \), \( 0 < x < \pi \), \( u(0, t) = u(\pi, t) = 0 \), \( u(x, 0) = 0 \), \( u_t(x, 0) = \sin x \). Then \( u\!\left(\frac{\pi}{2}, \frac{\pi}{4}\right) \) equals:
(a) \( \frac{1}{2} \) (b) 1 (c) \( \frac{1}{\sqrt{2}} \) (d) 0
Answer: (a)
\( u = B_1\sin 2t\,\sin x \) with \( 2B_1 = 1 \). So \( u = \frac{1}{2}\sin 2t\sin x \), and \( u\!\left(\frac{\pi}{2}, \frac{\pi}{4}\right) = \frac{1}{2}\cdot 1\cdot 1 = \frac{1}{2} \).
Practice Q4. Let \( u \) be harmonic in the unit disc, continuous up to the boundary, with \( u(1, \theta) = \cos^2\theta \). Then \( u \) at the point \( r = \frac{1}{2} \), \( \theta = 0 \) equals:
(a) \( \frac{1}{2} \) (b) \( \frac{5}{8} \) (c) \( \frac{1}{4} \) (d) \( \frac{3}{4} \)
Answer: (b)
\( u = \frac{1}{2} + \frac{r^2}{2}\cos 2\theta \); at \( r = \frac{1}{2} \), \( \theta = 0 \): \( \frac{1}{2} + \frac{1}{8} = \frac{5}{8} \).
Level 3: Part C Style
Practice Q5. Let \( u \) solve \( u_t = u_{xx} \), \( 0 < x < \pi \), \( u(0, t) = u(\pi, t) = 0 \), \( u(x, 0) = 100 \). Which statements are true?
(a) In the sine series of the solution, the coefficients of \( \sin nx \) vanish for even \( n \). (b) \( u(x, t) \to 0 \) as \( t \to \infty \) for every \( x \). (c) \( u\!\left(\frac{\pi}{2}, t\right) \le 100 \) for all \( t > 0 \). (d) \( u\!\left(\frac{\pi}{2}, t\right)e^{\pi^2 t} \) tends to a finite non-zero limit as \( t \to \infty \).
Answer: (a), (b) and (c)
- (a) True. \( b_n = \frac{200}{n\pi}(1 - (-1)^n) \).
- (b) True. Every mode decays and the series is dominated by \( \frac{400}{\pi}\sum_{n \text{ odd}} \frac{e^{-n^2t}}{n} \to 0 \).
- (c) True. Maximum principle: the data never exceed 100.
- (d) False. On \( (0, \pi) \) with \( k = 1 \) the slowest decay is \( e^{-t} \): \( u\!\left(\frac{\pi}{2}, t\right) \approx \frac{400}{\pi}e^{-t} \), so \( e^{\pi^2t}u \to \infty \). The finite non-zero limit belongs to \( e^{t}u \), namely \( \frac{400}{\pi} \).
Practice Q6. Let \( u \) solve \( u_{tt} = u_{xx} \), \( 0 < x < \pi \), fixed ends, \( u(x, 0) = \sin x \), \( u_t(x, 0) = 0 \). Which statements are true?
(a) \( u \) is periodic in \( t \) with period \( 2\pi \). (b) \( u(x, \pi) = 0 \) for all \( x \). (c) The energy \( \frac{1}{2}\int_0^\pi(u_t^2 + u_x^2)\,dx \) equals \( \frac{\pi}{4} \) for all \( t \). (d) \( u(x, t) = \frac{\sin(x - t) + \sin(x + t)}{2} \).
Answer: (a), (c) and (d)
\( u = \sin x\cos t \).
- (a) True. \( \cos t \) has period \( 2\pi \) (and \( \frac{2L}{c} = 2\pi \)).
- (b) False. \( u(x, \pi) = -\sin x \).
- (c) True. \( u_t^2 + u_x^2 = \sin^2 x\sin^2 t + \cos^2 x\cos^2 t \), whose integral over \( (0, \pi) \) is \( \frac{\pi}{2}(\sin^2 t + \cos^2 t) = \frac{\pi}{2} \); half of that is \( \frac{\pi}{4} \).
- (d) True. d’Alembert with \( g = 0 \); the sum-to-product formula gives \( \sin x\cos t \).
Practice Q7. Let \( u \) solve \( u_{xx} + u_{yy} = 0 \) in \( (0, \pi) \times (0, 1) \), with \( u = 0 \) on \( x = 0 \), \( x = \pi \) and \( y = 0 \), and \( u(x, 1) = \sin 2x \). Which statements are true?
(a) \( u\!\left(\frac{\pi}{2}, y\right) = 0 \) for \( 0 \le y \le 1 \). (b) The maximum of \( u \) on the closed rectangle is 1. (c) \( u\!\left(\frac{\pi}{4}, \frac{1}{2}\right) = \frac{1}{2\cosh 1} \). (d) \( u\!\left(\frac{\pi}{4}, \frac{1}{2}\right) = \frac{1}{2} \).
Answer: (a), (b) and (c)
\( u = \frac{\sin 2x\sinh 2y}{\sinh 2} \).
- (a) True. \( \sin \pi = 0 \).
- (b) True. By the maximum principle the maximum is on the boundary; the boundary values are at most \( \sin\frac{\pi}{2} = 1 \), attained at \( \left(\frac{\pi}{4}, 1\right) \).
- (c) True. \( \frac{\sinh 1}{\sinh 2} = \frac{\sinh 1}{2\sinh 1\cosh 1} = \frac{1}{2\cosh 1} \approx 0.324 \).
- (d) False. It is about 0.324, not 0.5. Because \( \sinh \) is convex, \( \frac{\sinh 2y}{\sinh 2} < y \) for \( 0 < y < 1 \): the solution at half height is below half the top value.
Quick Revision Table
| Problem | Solution |
|---|---|
| Heat, Dirichlet | \( \sum b_n e^{-kn^2\pi^2t/L^2}\sin\frac{n\pi x}{L} \) |
| Heat, Neumann | \( \frac{a_0}{2} + \sum a_n e^{-kn^2\pi^2t/L^2}\cos\frac{n\pi x}{L} \to \) average |
| Heat, ends \( T_1, T_2 \) | steady state \( T_1 + (T_2 - T_1)\frac{x}{L} \) + transient |
| Wave, fixed ends | \( \sum (A_n\cos\omega_nt + B_n\sin\omega_nt)\sin\frac{n\pi x}{L} \), \( \omega_n = \frac{n\pi c}{L} \) |
| d’Alembert | \( \frac{f(x - ct) + f(x + ct)}{2} + \frac{1}{2c}\int_{x-ct}^{x+ct}g \) |
| Laplace, rectangle | \( \sum c_n\sin\frac{n\pi x}{a}\sinh\frac{n\pi y}{a} \) |
| Laplace, disc | \( \frac{a_0}{2} + \sum r^n(a_n\cos n\theta + b_n\sin n\theta) \) |
| Harmonic | mean value, maximum principle, uniqueness for Dirichlet |
Key Takeaways
What’s Next
- Partial Differential Equations: Quick Revision: the whole chapter on one page.
- Classification of Second-Order PDEs to revisit why each equation needs the data it does.
- Riemann, Improper and Lebesgue Integration for the integrals behind Fourier coefficients.
- Back to the chapter overview.