1. ODEs & Green's Functions

Introduction

Green's functions are the physicist's Swiss Army knife. Given any linear differential equation with a source term, the Green's function converts the problem into a single integral — one function encodes the entire response of the system to an arbitrary driving force.

The idea is deceptively simple: if you know how the system responds to a point impulse (a Dirac delta function), then by superposition you can construct its response to any source. This principle permeates all of theoretical physics — from electrostatics and heat conduction to quantum field theory and signal processing.

George Green introduced these ideas in his remarkable 1828 essay on mathematical analysis of electricity and magnetism, decades before the rigorous theory of distributions was available. Today, Green's functions form the backbone of perturbation theory, scattering theory, and the path integral formulation of quantum mechanics.

Derivation 1: Second-Order Linear ODEs

The General Framework

Consider the second-order linear ODE in standard form:

with boundary conditions at $x = a$ and $x = b$. The homogeneous equation$\mathcal{L}\,y = 0$ has two linearly independent solutions $y_1(x)$ and $y_2(x)$.

The Wronskian

The Wronskian of $y_1$ and $y_2$ is defined as:

Abel's theorem tells us that $W(x) = W(x_0)\exp\!\bigl[-\int_{x_0}^x p_1(t)/p_0(t)\,dt\bigr]$. For $y_1$ and $y_2$ linearly independent, $W(x) \neq 0$ everywhere in the interval.

Variation of Parameters

We seek a particular solution of the form:

Imposing the gauge condition $u_1' y_1 + u_2' y_2 = 0$ (to eliminate second derivatives of $u_1, u_2$), substitution into the ODE yields:

Solving the two simultaneous equations by Cramer's rule:

Arriving at the Green's Function

Integrating and combining:

where $x_< = \min(x, x')$ and $x_> = \max(x, x')$. Reading off the kernel, the Green's function is:

The full solution is therefore:

Derivation 2: Green's Function Definition and Properties

The Defining Equation

The Green's function $G(x, x')$ for the operator $\mathcal{L}$ is defined as the solution to:

subject to the same homogeneous boundary conditions as the original problem. The subscript on $\mathcal{L}_x$ emphasizes that the operator acts on the first argument.

The Jump Condition

Consider the self-adjoint form $\mathcal{L}\,y = \frac{d}{dx}\bigl[p(x)\,y'\bigr] + q(x)\,y$. Integrating $\mathcal{L}\,G = \delta(x - x')$ over a small interval $[x' - \epsilon,\, x' + \epsilon]$:

As $\epsilon \to 0$, the second integral vanishes (since $G$ is continuous), leaving:

Construction from Homogeneous Solutions

For $x \neq x'$, the Green's function satisfies the homogeneous equation. So we write:

where $y_1$ satisfies the BC at $x = a$ and $y_2$ at $x = b$. Applying the continuity condition and the jump condition determines $A$ and $B$:

Solving yields $A(x') = y_2(x')/[p(x')\,W(x')]$ and $B(x') = y_1(x')/[p(x')\,W(x')]$, recovering the same expression derived via variation of parameters.

Symmetry of the Green's Function

For a self-adjoint operator $\mathcal{L}$, the Green's function satisfies:

Proof: Let $G_1 = G(x, x_1)$ and $G_2 = G(x, x_2)$ satisfy the same homogeneous BCs. Then:

by the self-adjointness of $\mathcal{L}$ and the homogeneous BCs. The left side evaluates to$G(x_2, x_1) - G(x_1, x_2)$, hence $G(x_1, x_2) = G(x_2, x_1)$. This reciprocity is the mathematical expression of the physical principle that the response at $x$due to a source at $x'$ equals the response at $x'$ due to a source at $x$.

Derivation 3: Green's Function for the 1D Helmholtz Equation

The 1D Helmholtz equation arises in wave propagation, quantum mechanics (time-independent Schrödinger equation in 1D with constant potential), and normal modes of vibrating strings:

Case 1: Infinite Line

On $x \in (-\infty, \infty)$ with outgoing wave / radiation boundary conditions. The homogeneous solutions are $e^{\pm ikx}$. Requiring the Green's function to be bounded as $x \to \pm\infty$:

Continuity at $x = x'$: $A\,e^{ikx'} = B\,e^{-ikx'}$. Jump condition on the derivative: $ikA\,e^{ikx'} + ikB\,e^{-ikx'} = 1$. Solving:

Case 2: Half-Line with Dirichlet BC

On $x \in [0, \infty)$ with $G(0, x') = 0$ and outgoing waves as $x \to \infty$. The method of images gives:

The second term is the contribution from an “image source” at $x = -x'$ with opposite sign, ensuring the Dirichlet condition $G(0, x') = 0$ is satisfied.

Case 3: Finite Interval

On $x \in [0, L]$ with $G(0, x') = G(L, x') = 0$. The homogeneous solutions satisfying each BC are $y_1(x) = \sin(kx)$ and $y_2(x) = \sin\bigl(k(L - x)\bigr)$. The Wronskian is $W = -k\sin(kL)$. Therefore:

Method of Images

The method of images, familiar from electrostatics, constructs the Green's function for a bounded domain by superposing free-space Green's functions and their reflections. The key insight is that image sources placed outside the domain enforce boundary conditions:

• Dirichlet BCs ($G = 0$ on boundary): use images with opposite sign
• Neumann BCs ($G' = 0$ on boundary): use images with same sign

For the half-line, a single image suffices. For a finite interval, an infinite series of images is needed (though often the direct construction via homogeneous solutions is more practical).

Derivation 4: Eigenfunction Expansion

Setup

Suppose the self-adjoint operator $\mathcal{L}$ (with homogeneous BCs) has a complete set of orthonormal eigenfunctions $\{\phi_n(x)\}$ with eigenvalues $\{\lambda_n\}$:

Derivation

Expand the Green's function in this basis:

Substituting into $\mathcal{L}\,G = \delta(x - x')$:

where we used the completeness relation for the delta function in the last step. Matching coefficients: $c_n(x')\,\lambda_n = \phi_n(x')$, hence $c_n(x') = \phi_n(x')/\lambda_n$(assuming $\lambda_n \neq 0$).

More generally, if we seek the Green's function for $(\mathcal{L} - \lambda)\,G = \delta$(the “resolvent”), the expansion becomes:

Connection to Spectral Theory

The eigenfunction expansion reveals profound connections:

• Poles of G: The Green's function $G_\lambda$ has simple poles at$\lambda = \lambda_n$. The eigenvalues are precisely the points where the resolvent blows up — this is the starting point of spectral theory.
• Residues: The residue at $\lambda = \lambda_n$ is $-\phi_n(x)\,\phi_n(x')$, giving the projection operator onto the $n$-th eigenstate.
• Contour integral representation: By contour integration,$G_\lambda = \frac{1}{2\pi i}\oint \frac{G_z}{z - \lambda}\,dz$, connecting Green's functions to complex analysis.
• Continuous spectrum: When $\mathcal{L}$ has a continuous spectrum, the sum becomes an integral: $G_\lambda(x, x') = \int \frac{\phi_\mu(x)\,\phi_\mu^*(x')}{\mu - \lambda}\,d\mu$, and the poles become a branch cut.

Relation to the Resolvent Operator

In operator notation, the Green's function for $(\mathcal{L} - \lambda)G = \delta$ is the integral kernel of the resolvent $R(\lambda) = (\mathcal{L} - \lambda)^{-1}$. The resolvent exists for all $\lambda$ not in the spectrum of $\mathcal{L}$. The set of such $\lambda$ is called the resolvent set, and its complement is the spectrum. This viewpoint connects Green's functions directly to functional analysis and motivates the use of contour integrals to extract spectral information.

Derivation 5: Retarded and Advanced Green's Functions

Time-Dependent Problems

For equations with a time variable, such as the 1D wave equation:

there are multiple Green's functions depending on the boundary conditions in time.

Retarded Green's Function

The retarded Green's function $G_R$ satisfies the causality condition:

Physically, the response occurs after the source acts. For the 1D wave equation:

Advanced Green's Function

The advanced Green's function $G_A$ is nonzero only for $t < t'$:

It describes “backward-in-time” propagation and, while non-physical for classical causal systems, plays an essential role in time-symmetric formulations (Wheeler–Feynman absorber theory).

The Feynman Propagator

In quantum field theory, neither purely retarded nor advanced boundary conditions are appropriate. Instead, one uses the Feynman propagator, which propagates positive-frequency modes forward in time and negative-frequency modes backward:

The $i\epsilon$ prescription (with $\epsilon \to 0^+$) determines how the contour in the complex $\omega$-plane avoids the poles at $\omega = \pm ck$:

• Retarded: Both poles pushed below the real axis ($\omega \to \omega + i\epsilon$)
• Advanced: Both poles pushed above the real axis ($\omega \to \omega - i\epsilon$)
• Feynman: Positive-frequency pole below, negative-frequency pole above ($\omega^2 \to \omega^2 + i\epsilon$)

Applications

Worked Example: Charged Rod near a Grounded Plane

A thin rod of charge density $\lambda$ lies along the $z$-axis at distance$d$ from an infinite grounded conducting plane at $x = 0$. By the method of images, we place an image charge $-\lambda$ at $x = -d$. The Green's function for the half-space $x > 0$ with Dirichlet BC on the plane is:

where $\mathbf{r}'_{\text{image}}$ is the reflection of the source through the plane. This is the 3D Poisson Green's function with Dirichlet boundary conditions — a direct generalization of the 1D half-line construction we derived above.

The Born Series

The Lippmann–Schwinger equation can be iterated to produce the Born series:

Each term corresponds to an additional scattering event. The first Born approximation retains only the first correction, and is accurate when the potential is weak. This series is the prototype for all perturbative expansions in quantum field theory.

Historical Context

Python Simulation

We construct Green's functions for several 1D boundary value problems and verify them by solving the ODE directly using finite differences. All computations use only numpy.

Summary of Key Results

Common Pitfalls and Subtleties

Further Reading