Home / Learn / Exam prep / GATE Physics / Lesson 1.7
All lessons

Module 1: Mathematical physics · Lesson 1.7 · about 65 min

Partial differential equations and separation of variables

Classifying second-order PDEs as elliptic, parabolic or hyperbolic, separation of variables for the heat, wave and Laplace equations with Fourier sine and cosine series, steady states and transients, decay times and normal modes, d'Alembert's solution and characteristics, Laplace's equation in polar, cylindrical and spherical coordinates, and uniqueness.

Key points, meanings and formulas on one page: choose "Save as PDF" in the print box.

Module 1 · about 65 minutes. By the end you will be able to classify a second-order PDE, solve the heat, wave and Laplace equations by separation of variables in Cartesian, polar and spherical coordinates, find Fourier coefficients from initial or boundary data, split a problem into a steady state plus transients, read off decay times and mode frequencies, and use d’Alembert’s solution and the domain of dependence. Fourier series themselves are in the JAM lesson on Fourier series, and the wave equation on a string in waves.

Separation of variables turns one PDE into several ODEs, one per coordinate, linked by separation constants. The boundary conditions pick out a discrete set of modes; the initial or remaining boundary data fix how much of each mode is present. Linearity lets you add them up.

1. Three kinds of equation

For Auxx+Buxy+Cuyy+(lower-order terms)=0, the discriminant B2−4AC decides the type:

B2−4AC type model behaviour
negative elliptic Laplace ∇2u=0 equilibrium; data on a closed boundary; smooth inside
zero parabolic heat ut=κuxx diffusion; smooths; infinite signal speed
positive hyperbolic wave utt=c2uxx propagation at finite speed along characteristics

Lower-order terms never change the type: the telegraph equation utt+γut=c2uxx is still hyperbolic. With variable coefficients the type can change from region to region (for uxx+xuyy=0, elliptic for x>0, hyperbolic for x<0).

2. The heat equation on a rod

ut=κuxx on 0<x<L with u(0,t)=u(L,t)=0. Try u=X(x)T(t):

T′κT=X′′X=−k2

Each side depends on a different variable, so both equal a constant. A zero or positive constant gives only X≡0 with two zero ends, so it must be negative. Then X=sin⁡kx with sin⁡kL=0, i.e. kn=nπ/L, and T=e−κkn2t:

u(x,t)=∑n=1∞bnsin⁡nπxLe−κn2π2t/L2,bn=2L∫0Lu(x,0)sin⁡nπxLdx

Mode n decays with time constant τn=L2n2π2κ. The n2 means fine detail disappears first; after a while only the half-sine n=1 is left.

Worked example. A rod at a uniform 100 °C has its ends put into ice. bn=200nπ(1−cos⁡nπ): 400nπ for odd n, zero for even. When κt/L2=0.1, the midpoint is at 400πe−0.987≈47.5 °C; the n=3 term adds only −0.006 °C.

Other end conditions. Insulated ends (ux=0) give cosines, including n=0: a constant that never decays, so the rod ends up at its initial mean temperature. Ends held at different temperatures T1, T2: first subtract the steady state us=T1+(T2−T1)x/L (it solves uxx=0 with the right end values); the remainder has zero ends and decays as above.

3. The wave equation: modes and d’Alembert

On a string with fixed ends the same separation gives sin⁡(nπx/L), but now T′′=−c2kn2T, so each mode oscillates with ωn=nπc/L:

u=∑nsin⁡nπxL(Ancos⁡ωnt+Bnsin⁡ωnt),An=sine coefficients of u(x,0),ωnBn=those of ut(x,0)

A string plucked at its midpoint has no even harmonics (each has a node there). Released from rest means cosines in time; struck from a straight position means sines, with amplitude = initial velocity coefficient divided by ωn.

On an infinite line, d’Alembert’s form is often faster:

u(x,t)=12[f(x−ct)+f(x+ct)]+12c∫x−ctx+ctg(s)ds

for u(x,0)=f, ut(x,0)=g. A bump released from rest splits into two half-height copies. The value at (x0,t0) depends only on data in [x0−ct0,x0+ct0], between the two characteristics x±ct= const: nothing travels faster than c.

The figure runs both equations on the same initial profile, mode by mode.

A profile on [0, L] built from sine modes (thin curves: the first three). Under the heat equation each mode decays as e^(−n²π²κt/L²); under the wave equation each oscillates as cos(nπct/L). Fewer modes cannot reproduce sharp corners.
Separation constant
the common value of two expressions in different variables; boundary conditions decide its sign and allowed values
Mode
one separated solution, a shape that keeps its form in time (decaying or oscillating)
Steady state
the time-independent solution carrying the non-zero boundary values; the rest is transient
Characteristic
a line along which a hyperbolic equation carries information
Physical meaning
Diffusion forgets detail at a rate set by length squared over diffusivity; waves keep every mode forever (without damping) and only rearrange their phases.

4. Laplace’s equation in a rectangle

uxx+uyy=0 in 0<x<a, 0<y<b, with u=0 on three sides and u(x,b)=f(x). The direction with two zero conditions oscillates; the other gets hyperbolic functions:

u=∑ncnsin⁡nπxasinh⁡nπya,cn=fnsinh⁡(nπb/a)

where fn are the sine coefficients of f (sinh⁡, not cosh⁡, because u=0 at y=0). With several non-zero sides, solve one side at a time and add. Symmetry shortcut: in a square with one side at V0 and three at 0, the centre is at exactly V0/4, because the four rotated problems add up to u=V0 everywhere.

A harmonic function equals its average over any circle inside its domain, so it has no interior maximum or minimum. The Dirichlet problem has exactly one solution; the Neumann problem has one up to a constant, and only if the total outward flux is zero.

5. Polar, cylindrical and spherical coordinates

Polar (2D). u=R(r)Θ(θ) gives Θ=cos⁡mθ,sin⁡mθ (integer m for a single-valued answer) and R=rm,r−m (or 1,ln⁡r for m=0). Inside a disc of radius a with u(a,θ)=∑(amcos⁡mθ+bmsin⁡mθ), each harmonic simply picks up (r/a)m. Worked example: u(a,θ)=8cos⁡2θ gives u=8(r/a)2cos⁡2θ, so u(a/2,0)=2.

Cylindrical. With Z=e±kz the radial equation is Bessel’s equation of order m in kr: solutions Jm(kr) (regular on the axis) and Ym(kr). The Helmholtz equation and a drumhead give the same radial functions with the zeros of Jm setting the allowed k (lesson 1.6).

Spherical, axial symmetry. u=∑(Aℓrℓ+Bℓr−(ℓ+1))Pℓ(cos⁡θ). Inside a sphere keep rℓ, outside keep r−(ℓ+1), and match the surface data term by term in Legendre polynomials.

Worked example. A sphere of radius R has surface potential V0cos⁡θ=V0P1. Inside, V=V0rRcos⁡θ (a uniform field); outside, V=V0R2r2cos⁡θ (a dipole). At r=2R on the axis, V0/4.

For the Helmholtz equation ∇2u+k2u=0 in spherical coordinates the radial functions become spherical Bessel functions jℓ(kr), with j0=sin⁡krkr. A hot ball with its surface held at 0 cools through j0 with k=π/a, four times faster than a slab of thickness 2a.

“The solution of the heat equation is the first Fourier term, so its amplitude is the initial temperature.” The first coefficient of a uniform profile U is 4U/π≈1.27U, larger than U itself: the sine series has to overshoot in the middle to fit a flat profile that drops to zero at the ends. Always compute b1 from the integral, and subtract the steady state first when the ends are not at zero.

  • B2−4AC: negative elliptic, zero parabolic, positive hyperbolic; lower-order terms do not matter.
  • Heat on a rod: sin⁡(nπx/L)e−κn2π2t/L2, τ1=L2/(π2κ); insulated ends give cosines and a conserved mean; subtract the linear steady state for non-zero ends.
  • Waves: ωn=nπc/L, coefficients from u(x,0) and ut(x,0)/ωn; d’Alembert splits pulses; the domain of dependence is [x−ct,x+ct].
  • Laplace in a rectangle: sines along the two-zero direction, sinh⁡ across, cn=fn/sinh⁡(nπb/a). Max and min on the boundary.
  • Polar: r±me±imθ. Cylindrical: Jm(kr). Spherical: rℓ inside, r−(ℓ+1) outside, times Pℓ(cos⁡θ).

Practise: exam questions on this lesson

From warm-up to top-rank level, with full solutions. More on the partial differential equations practice page.

fibub’s Heat conduction solver and Standing waves calculator let you try other rods and strings. Before this: lesson 1.6, series solutions and special functions. Next: lesson 1.8, the Dirac delta and Green’s functions, which solve the same equations with sources. Back to the GATE Physics course home.

Going on to CSIR NET? Numerical methods solves the same equations by finite differences when separation of variables fails.

Check yourself

1On a rod with both ends held at 0, how does the decay rate of the mode sin(nπx/L) depend on n?
2Inside a sphere with surface potential V₀P₂(cos θ) and no charge, how does the potential depend on r?
3Which equation is hyperbolic?

Sources

  • M. L. Boas, Mathematical Methods in the Physical Sciences, 3rd ed. (2006), chapter 13
  • G. B. Arfken, H. J. Weber and F. E. Harris, Mathematical Methods for Physicists, 7th ed. (2013), chapter 9
  • W. A. Strauss, Partial Differential Equations: An Introduction, 2nd ed. (2008), chapters 1 to 6
  • D. J. Griffiths, Introduction to Electrodynamics, 4th ed. (2017), section 3.3 (separation of variables)
  • GATE Physics (PH) syllabus, Mathematical Physics section, published by the organising institute each year

Your ticks stay in this browser. Save them as a file on the course home.

Updated 2026-11-22. Not affiliated with the IITs, IISc or the GATE organising institute. The general aptitude part of GATE is not covered here. Check the official GATE website for the current pattern. All questions are original.