Green's Functions
CSIR NET Physical Sciences: Green's functions for ODE boundary and initial value problems, eigenfunction expansion, Poisson, Helmholtz, diffusion, images.
- 26 min
On this page · 15 sections
- Prerequisites
- The Hook: One Point Charge Is Enough
- The Defining Idea
- Green’s Functions for Second-Order Boundary Value Problems
- Eigenfunction Expansion
- Initial Value Problems: the Causal Green’s Function
- Green’s Functions of the Standard PDE Operators
- Bounded Regions: Green’s Theorem and the Method of Images
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Linear Algebra, Matrices and Eigenvalues: a Green’s function is the continuous version of a matrix inverse, and the eigenfunction expansion below is the spectral decomposition of that inverse.
- Dimensional Analysis and Vector Calculus: the divergence theorem, Green’s identities and \( \nabla^2(1/r) = -4\pi\delta^3(\mathbf r) \).
- The proof-level treatment (existence, symmetry, Sturm–Liouville theory) is in Sturm–Liouville Problems and Green’s Function (Mathematical Sciences). This page is the physicist’s working kit.
The Hook: One Point Charge Is Enough
The Defining Idea
Let \( \hat L \) be a linear differential operator acting on functions of \( x \), with homogeneous boundary conditions. We want to solve
$$\hat L\,u(x) = f(x).$$The Green’s function \( G(x, x') \) is the solution for a unit point source at \( x' \), satisfying the same homogeneous boundary conditions in \( x \):
$$\boxed{\hat L\,G(x, x') = \delta(x - x')}$$Then, by linearity,
$$\boxed{u(x) = \int G(x, x')\,f(x')\,dx'}$$Apply \( \hat L \) under the integral and you get \( \int\delta(x - x')f(x')\,dx' = f(x) \). The boundary conditions hold because each \( G(\cdot, x') \) satisfies them.
Sign conventions
Books differ. Physics texts often define \( \nabla^2G = -\delta \) (so that \( G = 1/4\pi r \) is positive, like a potential), while maths texts use \( \hat LG = +\delta \). Both are fine; the solution formula carries the matching sign. In a question, always read off the convention from the defining equation before comparing with options. This page uses \( \hat LG = \delta \) unless it says otherwise.
Green’s Functions for Second-Order Boundary Value Problems
Take the Sturm–Liouville operator on \( a \leq x \leq b \):
$$\hat Ly = \frac{d}{dx}\left[p(x)\frac{dy}{dx}\right] + q(x)\,y, \qquad \text{with homogeneous conditions at } x = a \text{ and } x = b.$$Any second-order linear operator \( y'' + P(x)y' + Q(x)y \) can be brought to this form by multiplying through by \( p = e^{\int P\,dx} \).
What G must do at x = x'
Away from \( x = x' \), \( \hat LG = 0 \). Integrate \( \hat LG = \delta(x - x') \) across a small interval around \( x' \):
- \( G \) is continuous at \( x = x' \) (otherwise \( G'' \) would contain \( \delta' \)).
- \( \partial G/\partial x \) jumps by \( 1/p(x') \):
The construction recipe
- Find \( u_1(x) \), a solution of \( \hat Lu = 0 \) that satisfies the left boundary condition.
- Find \( u_2(x) \), a solution of \( \hat Lu = 0 \) that satisfies the right boundary condition.
- Compute the Wronskian \( W = u_1u_2' - u_1'u_2 \). For a Sturm–Liouville operator, \( p(x)W(x) \) is a constant (Abel’s theorem).
- Write
Continuity is automatic (both pieces equal \( u_1(x')u_2(x')/pW \) at \( x = x' \)), and the jump is \( [u_1u_2' - u_1'u_2]/pW = 1/p \), as required.
graph TD
A["Operator L and two boundary conditions"] --> B["u1: solves Lu = 0, obeys LEFT condition"]
A --> C["u2: solves Lu = 0, obeys RIGHT condition"]
B --> D["Wronskian W = u1 u2' - u1' u2"]
C --> D
D --> E{"Is W = 0?"}
E -->|"yes: u1 and u2 proportional, a zero mode exists"| F["No ordinary Green's function. Fredholm alternative applies"]
E -->|"no"| G["G = u1(x<) u2(x>) / (p W)"]
G --> H["u(x) = integral of G(x,x') f(x') dx'"]When G does not exist: the Fredholm alternative
If \( \hat Lu = 0 \) has a non-trivial solution \( u_0 \) obeying both boundary conditions (a “zero mode”), then \( u_1 \propto u_2 \), \( W = 0 \), and no Green’s function exists. The equation \( \hat Lu = f \) then has a solution only if \( f \) is orthogonal to the zero mode, \( \int u_0f\,dx = 0 \), and that solution is not unique. One can still build a modified Green’s function that solves \( \hat LG = \delta(x - x') - u_0(x)u_0(x') \) (with \( u_0 \) normalised), but Part C questions usually only ask you to recognise the failure.
Classic cases of failure:
| Problem | Zero mode | Why |
|---|---|---|
| \( y'' + n^2y \), \( y(0) = y(\pi) = 0 \), integer \( n \) | \( \sin nx \) | resonance |
| \( y'' \), \( y'(0) = y'(1) = 0 \) (pure Neumann) | constant | adding a constant changes nothing |
| \( y'' \) with periodic conditions on \( [0, 2\pi] \) | constant | same reason |
For the pure Neumann case, integrating \( y'' = f \) gives \( y'(1) - y'(0) = \int_0^1f\,dx \), which must vanish: the net source must be zero. This is the 1D version of “the total charge inside a box with zero normal field must vanish”.
Symmetry and reciprocity
For a self-adjoint operator with self-adjoint boundary conditions (the Sturm–Liouville case with Dirichlet, Neumann or mixed real conditions),
$$\boxed{G(x, x') = G(x', x)}$$This is reciprocity: the response at \( x \) to a source at \( x' \) equals the response at \( x' \) to a source at \( x \). It is visible in the formula \( u_1(x_<)u_2(x_>)/pW \), which treats \( x \) and \( x' \) identically. For a non-self-adjoint operator (for example \( y'' + y' \) without the factor \( p \)), \( G(x, x') \neq G(x', x) \) in general, and for causal (initial-value) problems \( G \) is never symmetric, because cause precedes effect.
Eigenfunction Expansion
Suppose \( \hat L\phi_n = \lambda_n\phi_n \) with orthonormal eigenfunctions satisfying the boundary conditions, \( \int\phi_m^*\phi_n\,dx = \delta_{mn} \), forming a complete set: \( \sum_n\phi_n(x)\phi_n^*(x') = \delta(x - x') \). To solve \( (\hat L - \lambda)u = f \), expand and invert term by term:
$$\boxed{G_\lambda(x, x') = \sum_n\frac{\phi_n(x)\,\phi_n^*(x')}{\lambda_n - \lambda}}$$Three things are visible at once:
- \( G_\lambda \) has poles at the eigenvalues. The Green’s function fails to exist exactly when \( \lambda \) hits an eigenvalue: the series view of the Fredholm alternative.
- Near a pole, \( G_\lambda \approx \phi_n(x)\phi_n^*(x')/(\lambda_n - \lambda) \): the response is dominated by the resonant mode. This is why a driven system near resonance oscillates in the shape of that normal mode.
- \( G \) is Hermitian, \( G(x, x') = G^*(x', x) \), when the eigenvalues are real.
In quantum mechanics this is the resolvent \( (E - \hat H)^{-1} = \sum_n\lvert n\rangle\langle n\rvert/(E - E_n) \) (note the opposite sign convention): its poles are the bound-state energies.
Initial Value Problems: the Causal Green’s Function
For problems in time, the “boundary conditions” are initial conditions: the system is at rest before the source acts. The Green’s function must vanish for \( t < t' \). This is the retarded or causal Green’s function, and it is the impulse response of the system.
The driven oscillator
For \( m\ddot x + kx = F(t) \), with \( \omega_0 = \sqrt{k/m} \) and the system at rest before the force acts, take \( m\ddot G + kG = \delta(t - t') \). For \( t > t' \), \( G \) is a free oscillation that starts from \( G = 0 \) with a velocity jump of \( 1/m \) (the impulse of a unit delta force). So
$$\boxed{G(t, t') = \theta(t - t')\,\frac{\sin\omega_0(t - t')}{m\omega_0}}, \qquad x(t) = \int_{-\infty}^tG(t, t')F(t')\,dt'.$$With damping, \( m(\ddot x + 2\gamma\dot x + \omega_0^2x) = F \), and \( \gamma < \omega_0 \):
$$G(t, t') = \theta(t - t')\,\frac{e^{-\gamma(t - t')}\sin\omega_1(t - t')}{m\omega_1}, \qquad \omega_1 = \sqrt{\omega_0^2 - \gamma^2}.$$The frequency-domain view
With \( x(t) = \int\tilde x(\omega)e^{-i\omega t}\,d\omega/2\pi \), the damped equation gives
$$\tilde G(\omega) = \frac{1}{m(\omega_0^2 - \omega^2 - 2i\gamma\omega)}.$$Its poles sit at \( \omega = -i\gamma \pm\omega_1 \), both in the lower half plane. For \( t < t' \) you close the contour in the upper half plane, enclose no poles, and get zero: causality is the statement that \( \tilde G(\omega) \) is analytic in the upper half plane. (This is the origin of the Kramers–Kronig relations between the real and imaginary parts of a response function.) The contour integral itself is a residue calculation of the kind covered in Series, Singularities and Residues.
Green’s Functions of the Standard PDE Operators
These are the results to know cold. Each satisfies the stated equation in all of space, with the solution decaying (or outgoing) at infinity.
| Operator equation | Green’s function | Physics |
|---|---|---|
| \( \dfrac{d^2G}{dx^2} = \delta(x - x') \) (1D) | \( \tfrac{1}{2}\lvert x - x'\rvert \) (plus any linear function) | sheet of charge |
| \( \nabla^2G = \delta^2(\boldsymbol\rho - \boldsymbol\rho') \) (2D) | \( \dfrac{1}{2\pi}\ln\lvert\boldsymbol\rho - \boldsymbol\rho'\rvert \) | line charge |
| \( \nabla^2G = -\delta^3(\mathbf r - \mathbf r') \) (3D) | \( \dfrac{1}{4\pi\lvert\mathbf r - \mathbf r'\rvert} \) | point charge |
| \( (\nabla^2 + k^2)G = -\delta^3 \) | \( \dfrac{e^{\pm ik\lvert\mathbf r - \mathbf r'\rvert}}{4\pi\lvert\mathbf r - \mathbf r'\rvert} \) | outgoing (+) or incoming (−) spherical waves |
| \( (\nabla^2 - \kappa^2)G = -\delta^3 \) | \( \dfrac{e^{-\kappa\lvert\mathbf r - \mathbf r'\rvert}}{4\pi\lvert\mathbf r - \mathbf r'\rvert} \) | screened (Yukawa, Debye) potential |
| \( \left(\dfrac{d^2}{dx^2} - \kappa^2\right)G = \delta(x - x') \) | \( -\dfrac{e^{-\kappa\lvert x - x'\rvert}}{2\kappa} \) | 1D screened source |
| \( (\partial_t - D\nabla^2)G = \delta^3(\mathbf r)\delta(t) \) | \( \dfrac{\theta(t)}{(4\pi Dt)^{3/2}}\,e^{-r^2/4Dt} \) | heat or particles released at a point |
| \( \left(\nabla^2 - \dfrac{1}{c^2}\partial_t^2\right)G = -\delta^3(\mathbf r)\delta(t) \) | \( \dfrac{\delta(t - r/c)}{4\pi r} \) (retarded) | retarded potentials of electrodynamics |
Note the dimensional pattern: in \( d \) dimensions the Laplacian Green’s function goes as \( r^{2-d} \) for \( d \geq 3 \), becomes logarithmic at \( d = 2 \), and linear at \( d = 1 \). The 2D and 1D Green’s functions grow at infinity, which is why a uniformly charged infinite line has no zero of potential at infinity.
The Helmholtz Green’s function and scattering
Writing the time-independent Schrödinger equation as \( (\nabla^2 + k^2)\psi = \frac{2m}{\hbar^2}V\psi \) and using the outgoing Green’s function gives the integral equation (Lippmann–Schwinger form)
$$\psi(\mathbf r) = e^{i\mathbf k\cdot\mathbf r} - \frac{m}{2\pi\hbar^2}\int\frac{e^{ik\lvert\mathbf r - \mathbf r'\rvert}}{\lvert\mathbf r - \mathbf r'\rvert}V(\mathbf r')\psi(\mathbf r')\,d^3r'.$$Replacing \( \psi \) by the incident plane wave inside the integral is the first Born approximation, which gives the scattering amplitude as a Fourier transform of the potential:
$$f(\mathbf q) = -\frac{m}{2\pi\hbar^2}\int e^{-i\mathbf q\cdot\mathbf r'}V(\mathbf r')\,d^3r', \qquad q = 2k\sin\frac{\theta}{2}.$$For the Yukawa potential \( V = \beta e^{-\mu r}/r \) this gives \( f = -\dfrac{2m\beta}{\hbar^2(\mu^2 + q^2)} \); letting \( \mu \to 0 \) recovers the Rutherford dependence \( \lvert f\rvert^2 \propto 1/\sin^4(\theta/2) \).
Bounded Regions: Green’s Theorem and the Method of Images
The general solution formula
For Poisson’s equation \( \nabla^2u = -f \) in a volume \( V \) with surface \( S \), Green’s second identity gives (with \( \nabla'^2G = -\delta^3(\mathbf r - \mathbf r') \))
$$u(\mathbf r) = \int_VG\,f\,d^3r' + \oint_S\left(G\,\frac{\partial u}{\partial n'} - u\,\frac{\partial G}{\partial n'}\right)dS'.$$- Dirichlet problem (\( u \) given on \( S \)): choose \( G_D = 0 \) on \( S \). The first surface term drops and only the known \( u \) appears.
- Neumann problem (\( \partial u/\partial n \) given): you cannot make \( \partial G/\partial n = 0 \) on all of \( S \) (the flux of \( \nabla G \) must equal \( -1 \)), so one takes \( \partial G_N/\partial n' = -1/A \) (with \( A \) the area of \( S \)), and the solution is fixed only up to an additive constant.
The full-space Green’s function \( 1/4\pi\lvert\mathbf r - \mathbf r'\rvert \) plus any harmonic function inside \( V \) still satisfies the equation; that harmonic piece is chosen to meet the boundary condition. For simple geometries it is the potential of image charges placed outside \( V \).
Method of images: the standard set
| Geometry (grounded conductor) | Real charge | Image |
|---|---|---|
| Infinite plane \( z = 0 \) | \( q \) at height \( d \) | \( -q \) at \( -d \) |
| Sphere of radius \( R \) | \( q \) at distance \( a > R \) from centre | \( -qR/a \) at distance \( R^2/a \) |
| Two perpendicular planes | \( q \) at \( (x_0, y_0) \) | \( -q \), \( -q \), \( +q \) at the three reflections |
| Line charge \( \lambda \) outside a grounded cylinder of radius \( R \) | \( \lambda \) at distance \( a \) | \( -\lambda \) at \( R^2/a \) |
For the plane, the induced surface charge is \( \sigma = -\dfrac{qd}{2\pi(\rho^2 + d^2)^{3/2}} \), integrating to \( -q \), and the force on the charge is \( q^2/(4\pi\epsilon_0(2d)^2) \), attractive. If the sphere is isolated and neutral instead of grounded, add a second image \( +qR/a \) at the centre so the total induced charge is zero.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Boundary value \( G \): symmetric, \( G(x, x') = G(x', x) \) for self-adjoint problems | Causal \( G \): proportional to \( \theta(t - t') \), never symmetric | |
| \( \nabla^2G = -\delta^3 \Rightarrow G = +1/4\pi r \) | \( \nabla^2G = +\delta^3 \Rightarrow G = -1/4\pi r \) | |
| 3D: \( 1/4\pi r \), decays | 2D: \( (1/2\pi)\ln\rho \), grows; 1D: \( \lvert x\rvert/2 \), grows | |
| Helmholtz \( (\nabla^2 + k^2) \): oscillating \( e^{ikr}/4\pi r \) | Screened \( (\nabla^2 - \kappa^2) \): decaying \( e^{-\kappa r}/4\pi r \) | |
| Grounded sphere: one image, induced charge \( -qR/a \) | Neutral isolated sphere: extra \( +qR/a \) at the centre, induced charge zero | |
| Dirichlet \( G \): exists and is unique | Pure Neumann \( G \): needs zero net source; solution unique only up to a constant |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper.
Level 1: Foundation
- For \( \hat L = d^2/dx^2 \), by how much does \( \partial G/\partial x \) jump at \( x = x' \)?
- Which function satisfies \( \nabla^2G = -\delta^3(\mathbf r) \) and vanishes at infinity?
- For the undamped oscillator \( m\ddot x + kx = F(t) \), what is \( G(t, t') \) for \( t < t' \)?
Answers
- By 1 (here \( p = 1 \)). In general the jump is \( 1/p(x') \).
- \( G = 1/(4\pi r) \).
- Zero. The causal Green’s function vanishes before the impulse is applied.
Level 2: Part B Style
Practice Q1. The function \( y(x) \) satisfies \( y'' = x \) on \( 0 \leq x \leq 1 \) with \( y(0) = y(1) = 0 \). Using the Green’s function \( G(x, x') = x_<(x_> - 1) \), the value \( y(1/2) \) is:
(a) \( -1/16 \) (b) \( -1/48 \) (c) \( 1/16 \) (d) \( -1/12 \)
Answer: (a) −1/16
\( y\left(\tfrac{1}{2}\right) = \int_0^{1/2}x'\left(\tfrac{1}{2} - 1\right)x'\,dx' + \int_{1/2}^1\tfrac{1}{2}(x' - 1)x'\,dx' = -\tfrac{1}{2}\cdot\tfrac{1}{24} + \tfrac{1}{2}\left(\tfrac{7}{24} - \tfrac{3}{8}\right) = -\tfrac{1}{48} - \tfrac{1}{24} = -\tfrac{1}{16} \).
Check by direct integration: \( y = (x^3 - x)/6 \) satisfies \( y'' = x \) and both conditions, and \( y(1/2) = (1/8 - 1/2)/6 = -1/16 \). Option (b) keeps only the first integral; option (c) has the sign wrong (a positive source in \( y'' \) with zero ends makes \( y \) sag below zero).
Practice Q2. For \( y'' + y = f(x) \), \( y(0) = 0 \), \( y(\pi/2) = 0 \), the Green’s function (with \( \hat LG = \delta \)) evaluated at \( x = \pi/6 \), \( x' = \pi/3 \) is:
(a) \( -1/4 \) (b) \( 1/4 \) (c) \( -\sqrt3/4 \) (d) \( -3/4 \)
Answer: (a) −1/4
Left solution \( u_1 = \sin x \) (vanishes at 0). Right solution \( u_2 = \cos x \) (vanishes at \( \pi/2 \)). \( W = \sin x(-\sin x) - \cos x\cos x = -1 \), \( p = 1 \).
\( G(x, x') = -\sin x_<\cos x_> \). With \( x_< = \pi/6 \) and \( x_> = \pi/3 \): \( G = -\tfrac{1}{2}\cdot\tfrac{1}{2} = -\tfrac{1}{4} \).
Option (d), \( -3/4 \), comes from attaching the solutions to the wrong arguments (\( -\sin x_>\cos x_< \)); option (c) mixes the sine and cosine of the same angle \( \pi/6 \); option (b) drops the sign of \( W \).
Practice Q3. For which of the following problems does a Green’s function fail to exist?
(a) \( y'' + y = f \), \( y(0) = y(\pi/2) = 0 \) (b) \( y'' + 4y = f \), \( y(0) = y(\pi) = 0 \) (c) \( y'' - y = f \), \( y(0) = y(\pi) = 0 \) (d) \( y'' = f \), \( y(0) = 0 \), \( y'(1) = 0 \)
Answer: (b)
In (b) the homogeneous problem has the solution \( \sin2x \), which vanishes at both 0 and \( \pi \). So \( u_1 \propto u_2 \), \( W = 0 \), and no Green’s function exists.
(a) \( \sin x \) does not vanish at \( \pi/2 \): fine. (c) \( y'' - y = 0 \) has only \( \sinh \) and \( \cosh \) solutions, and \( \sinh x \) vanishes only at 0: fine. (d) \( u_1 = x \), \( u_2 = 1 \) (constant, zero slope at 1), \( W = -1 \neq 0 \): fine, and \( G = -x_< \).
Practice Q4. The Green’s function of \( \left(\dfrac{d^2}{dx^2} - \kappa^2\right)G = \delta(x - x') \) on the whole real line, vanishing as \( \lvert x\rvert \to \infty \), is:
(a) \( -\dfrac{e^{-\kappa\lvert x - x'\rvert}}{2\kappa} \) (b) \( \dfrac{e^{-\kappa\lvert x - x'\rvert}}{2\kappa} \) (c) \( -\dfrac{\sinh\kappa\lvert x - x'\rvert}{\kappa} \) (d) \( -\dfrac{e^{-\kappa\lvert x - x'\rvert}}{\kappa} \)
Answer: (a)
Try \( G = Ae^{-\kappa\lvert x - x'\rvert} \). It solves the homogeneous equation for \( x \neq x' \) and decays. The slope jumps from \( +A\kappa \) to \( -A\kappa \), a jump of \( -2A\kappa \), which must equal \( 1 \). So \( A = -1/(2\kappa) \).
(b) has the wrong sign (it would solve the equation with \( -\delta \)); (c) grows at infinity; (d) gives a jump of 2.
Level 3: Part C Style
Practice Q5. An undamped oscillator of mass \( m = 0.5 \) kg and spring constant \( k = 50 \) N/m is at rest at \( x = 0 \). A constant force \( F_0 = 2 \) N acts from \( t = 0 \) to \( t = T = \pi/10 \) s and then switches off. Consider:
- The causal Green’s function is \( G(t, t') = \theta(t - t')\,\tfrac{1}{5}\sin[10(t - t')] \) in SI units.
- The displacement at \( t = T \) is \( 0.08 \) m.
- The amplitude of the free oscillation after the pulse is \( 0.08 \) m.
- If the pulse instead lasted \( T = \pi/5 \) s, the amplitude after the pulse would be \( 0.16 \) m.
Which are correct?
(a) 1, 2 and 3 only (b) 2 and 3 only (c) 1, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (a) 1, 2 and 3 only
\( \omega_0 = \sqrt{50/0.5} = 10 \) rad/s and \( 1/(m\omega_0) = 1/5 \). Statement 1 true.
During the pulse, \( x(t) = \frac{F_0}{k}(1 - \cos\omega_0t) \). At \( \omega_0T = \pi \): \( x = 0.04\times2 = 0.08 \) m. Statement 2 true.
At that instant the velocity is \( \frac{F_0}{k}\omega_0\sin\pi = 0 \), so the particle is at a turning point of the free motion about \( x = 0 \): amplitude 0.08 m. Equivalently \( A = \frac{2F_0}{k}\lvert\sin(\omega_0T/2)\rvert = 0.08\times1 \). Statement 3 true.
For \( T = \pi/5 \) s, \( \omega_0T = 2\pi \) and \( A = 0.08\lvert\sin\pi\rvert = 0 \). The oscillator returns to rest. Statement 4 false.
Practice Q6. A point charge \( q \) is held at distance \( 2R \) from the centre of a grounded conducting sphere of radius \( R \). Consider:
- The image charge is \( -q/2 \), located \( R/2 \) from the centre.
- The total charge induced on the sphere is \( -q/2 \).
- The force on \( q \) has magnitude \( \dfrac{q^2}{8\pi\epsilon_0R^2} \).
- If the sphere were instead isolated and uncharged, the attractive force on \( q \) would be smaller than for the grounded sphere.
Which are correct?
(a) 1 and 2 only (b) 1, 2 and 4 only (c) 1, 2 and 3 only (d) 2, 3 and 4 only
Answer: (b) 1, 2 and 4 only
- \( q' = -qR/a = -q/2 \) at \( R^2/a = R/2 \). True.
- The induced charge equals the image charge for a grounded sphere. True.
- Separation \( 3R/2 \): \( F = \frac{q(q/2)}{4\pi\epsilon_0(9R^2/4)} = \frac{2}{9}\frac{q^2}{4\pi\epsilon_0R^2} = \frac{q^2}{18\pi\epsilon_0R^2} \), not \( q^2/8\pi\epsilon_0R^2 \). False.
- A neutral isolated sphere needs an extra image \( +q/2 \) at the centre, which repels \( q \). Net force \( \frac{q^2}{4\pi\epsilon_0R^2}\left(-\frac{2}{9} + \frac{1}{8}\right) = -\frac{7}{72}\frac{q^2}{4\pi\epsilon_0R^2} \), still attractive but smaller. True.
Practice Q7. For the problem \( y'' = f \), \( y(0) = y(1) = 0 \), consider the following statements about its Green’s function \( G(x, x') \) (defined by \( \partial_x^2G = \delta(x - x') \)):
- \( G(x, x') = -2\displaystyle\sum_{n = 1}^\infty\frac{\sin n\pi x\,\sin n\pi x'}{n^2\pi^2} \).
- Evaluating the closed form and the series at \( x = x' = \tfrac{1}{2} \) shows that \( \sum_{n\ \text{odd}}1/n^2 = \pi^2/8 \).
- \( G(x, x') = G(x', x) \).
- \( G(x, x') > 0 \) for all \( 0 < x, x' < 1 \).
Which are correct?
(a) 1 and 3 only (b) 1, 2 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
- Eigenfunctions \( \sqrt2\sin n\pi x \) with eigenvalues \( -n^2\pi^2 \); \( G = \sum\phi_n\phi_n/\lambda_n \) gives the series. True.
- Closed form \( G(\tfrac12, \tfrac12) = \tfrac12(\tfrac12 - 1) = -\tfrac14 \); series \( -\frac{2}{\pi^2}\sum_{\text{odd}}1/n^2 \). Equating gives \( \pi^2/8 \). True.
- Self-adjoint operator with Dirichlet conditions: reciprocity holds. True.
- \( G = x_<(x_> - 1) \) is negative inside the interval. (With the opposite convention \( -\partial_x^2G = \delta \), it would be positive: another reason to read the definition.) False.
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| Definition | \( \hat LG(x, x') = \delta(x - x') \), same homogeneous conditions as \( u \) |
| Solution | \( u(x) = \int G(x, x')f(x')\,dx' \) |
| Matching at \( x' \) | \( G \) continuous; \( \partial_xG \) jumps by \( 1/p(x') \) |
| Construction | \( G = u_1(x_<)u_2(x_>)/(pW) \), \( pW \) constant |
| Existence | fails iff a zero mode obeys both conditions (\( W = 0 \)) |
| Symmetry | \( G(x, x') = G(x', x) \) for self-adjoint problems |
| Eigenfunction form | \( G_\lambda = \sum\phi_n(x)\phi_n^*(x')/(\lambda_n - \lambda) \) |
| \( y'' = f \), \( y(0) = y(1) = 0 \) | \( G = x_<(x_> - 1) \) |
| Causal oscillator | \( \theta(t - t')\sin\omega_0(t - t')/(m\omega_0) \) |
| Pulse of length \( T \) | amplitude \( (2F_0/k)\lvert\sin(\omega_0T/2)\rvert \) |
| 3D Laplace | \( \nabla^2(1/4\pi r) = -\delta^3 \) |
| 2D Laplace | \( \nabla^2(\ln\rho/2\pi) = \delta^2 \) |
| Helmholtz, screened | \( e^{ikr}/4\pi r \); \( e^{-\kappa r}/4\pi r \) |
| Diffusion kernel | \( (4\pi Dt)^{-3/2}e^{-r^2/4Dt} \) |
| Sphere image | \( -qR/a \) at \( R^2/a \) |
Key Takeaways
What’s Next
- Partial Differential Equations: Laplace, Wave and Heat: separation of variables in Cartesian, cylindrical and spherical coordinates, the other half of solving the same equations.
- Fredholm and Volterra Equations (Mathematical Sciences): a Green’s function turns a differential equation into an integral equation.
- One-Dimensional Problems: Box, Oscillator and Tunnelling: the delta-function potential uses the same slope-jump condition.