Integral Equations

Integral equations arise naturally when solving boundary value problems, in scattering theory, and wherever Green's functions appear. They provide a powerful alternative to differential equations, often incorporating boundary conditions directly into the formulation. This chapter develops the theory of Fredholm and Volterra equations, solution methods including Neumann series and the resolvent kernel, the Fredholm alternative theorem, and applications to physics including scattering and the Wiener-Hopf technique.

1. Classification of Integral Equations

An integral equation is an equation in which the unknown function appears under an integral sign. The general linear integral equation takes the form:

$$g(x)\,\phi(x) = f(x) + \lambda \int_a^b K(x, t)\,\phi(t)\, dt$$

where $\phi(x)$ is the unknown, $K(x,t)$ is the kernel,$f(x)$ is a known function, and $\lambda$ is a parameter.

Fredholm Equations

Fredholm equations have fixed integration limits. They come in two kinds:

$$\text{First kind:} \quad f(x) = \int_a^b K(x,t)\,\phi(t)\, dt$$

$$\text{Second kind:} \quad \phi(x) = f(x) + \lambda \int_a^b K(x,t)\,\phi(t)\, dt$$

In the first kind, $\phi$ appears only inside the integral. In the second kind, $\phi$ appears both inside and outside. If $f(x) = 0$, the equation is homogeneous.

Volterra Equations

Volterra equations have a variable upper limit:

$$\text{Second kind:} \quad \phi(x) = f(x) + \lambda \int_a^x K(x,t)\,\phi(t)\, dt$$

Volterra equations can be viewed as Fredholm equations with $K(x,t) = 0$ for $t > x$. They arise naturally in initial value problems and causal systems. A key difference from Fredholm equations: Volterra equations of the second kind always have a unique solution for any $\lambda$.

Connection to Differential Equations

Any linear ODE with boundary conditions can be converted to an integral equation using the Green's function. Consider $Ly = f$ with operator $L$ and boundary conditions. If $G(x,t)$ is the Green's function satisfying $LG = \delta(x-t)$, then:

$$y(x) = \int_a^b G(x,t)\, f(t)\, dt$$

This reformulation has advantages: the boundary conditions are built into $G$, and we can apply powerful operator-theoretic tools.

2. Neumann Series Solution

The Neumann series is the integral-equation analog of a geometric series. Consider the Fredholm equation of the second kind:

$$\phi(x) = f(x) + \lambda \int_a^b K(x,t)\,\phi(t)\, dt$$

Iterative Solution

We solve by iteration. Define $\phi_0 = f$ and substitute repeatedly:

$$\phi_1(x) = f(x) + \lambda \int_a^b K(x,t)\, f(t)\, dt$$

$$\phi_2(x) = f(x) + \lambda \int_a^b K(x,t)\left[f(t) + \lambda\int_a^b K(t,s)\,f(s)\,ds\right] dt$$

Continuing to all orders, we obtain the Neumann series:

$$\boxed{\phi(x) = f(x) + \sum_{n=1}^{\infty} \lambda^n \int_a^b K_n(x,t)\, f(t)\, dt}$$

where $K_n$ are the iterated kernels defined recursively:

$$K_1(x,t) = K(x,t), \qquad K_n(x,t) = \int_a^b K(x,s)\, K_{n-1}(s,t)\, ds$$

Convergence

For a bounded kernel with $\|K\|^2 = \int_a^b\int_a^b |K(x,t)|^2\, dx\, dt < \infty$ (Hilbert-Schmidt condition), the Neumann series converges absolutely and uniformly when:

$$|\lambda| < \frac{1}{\|K\|}$$

This parallels the geometric series $(1-x)^{-1} = \sum x^n$ converging for $|x| < 1$. In operator notation, the Neumann series is $\phi = (I - \lambda \hat{K})^{-1} f = \sum_{n=0}^\infty (\lambda \hat{K})^n f$.

Volterra Case: Always Convergent

For Volterra equations, a remarkable simplification occurs. Since $K(x,t) = 0$ for $t > x$, the iterated kernels $K_n(x,t)$ satisfy the bound:

$$|K_n(x,t)| \leq \frac{M^n (x-t)^{n-1}}{(n-1)!}$$

where $M = \max|K|$. The factorial in the denominator ensures the Neumann series converges for all $\lambda$. This is why Volterra equations of the second kind always have a unique solution.

3. The Resolvent Kernel

The Neumann series can be resummed into a compact form using the resolvent kernel(also called the reciprocal kernel). Define:

$$\boxed{R(x,t;\lambda) = \sum_{n=1}^{\infty} \lambda^{n-1} K_n(x,t)}$$

Then the solution to the Fredholm equation of the second kind is:

$$\phi(x) = f(x) + \lambda \int_a^b R(x,t;\lambda)\, f(t)\, dt$$

Resolvent Equation

The resolvent kernel satisfies its own integral equation:

$$R(x,t;\lambda) = K(x,t) + \lambda \int_a^b K(x,s)\, R(s,t;\lambda)\, ds$$

Proof: Using the definition $R = \sum_{n=1}^\infty \lambda^{n-1} K_n$ and the recursion$K_n(x,t) = \int K(x,s)\, K_{n-1}(s,t)\, ds$:

$$R = K_1 + \sum_{n=2}^{\infty}\lambda^{n-1} K_n = K + \sum_{n=2}^{\infty}\lambda^{n-1}\int K(x,s)\,K_{n-1}(s,t)\,ds$$

$$= K + \lambda \int K(x,s) \sum_{m=1}^{\infty}\lambda^{m-1} K_m(s,t)\, ds = K + \lambda \int K(x,s)\, R(s,t;\lambda)\, ds$$

Symmetric Kernels and Eigenexpansion

If $K(x,t) = K(t,x)$ (symmetric kernel), the resolvent has a spectral representation:

$$R(x,t;\lambda) = \sum_{n=1}^{\infty} \frac{\phi_n(x)\,\phi_n(t)}{\lambda_n - \lambda}$$

where $\lambda_n$ and $\phi_n$ are the eigenvalues and normalized eigenfunctions of the kernel. The resolvent has poles at $\lambda = \lambda_n$, analogous to the poles of $(A - \lambda I)^{-1}$ in finite-dimensional linear algebra.

4. The Fredholm Alternative

The Fredholm alternative is the infinite-dimensional generalization of the statement that$Ax = b$ has a unique solution if and only if $\det(A) \neq 0$.

Statement

Consider the Fredholm equation of the second kind and its homogeneous adjoint:

$$\phi(x) = f(x) + \lambda\int_a^b K(x,t)\,\phi(t)\,dt \qquad \text{(inhomogeneous)}$$

$$\psi(x) = \lambda\int_a^b K(t,x)\,\psi(t)\,dt \qquad \text{(homogeneous adjoint)}$$

Theorem (Fredholm Alternative): Exactly one of the following holds:

Case 1: The homogeneous adjoint has only the trivial solution $\psi = 0$. Then the inhomogeneous equation has a unique solution for every $f$.
Case 2: The homogeneous adjoint has $n$ linearly independent solutions$\psi_1, \ldots, \psi_n$. Then the inhomogeneous equation has a solution if and only if$f$ satisfies the $n$ solvability conditions:
$$\int_a^b f(x)\,\psi_k(x)\,dx = 0, \qquad k = 1, \ldots, n$$

Physical Interpretation

The Fredholm alternative appears throughout physics. In quantum mechanics, when solving the Lippmann-Schwinger equation for scattering, the alternative determines whether a bound state exists at a given energy. In electrostatics, it determines whether a charge distribution consistent with given boundary conditions exists. The solvability conditions are orthogonality requirements: the source $f$ must be orthogonal to the null space of the adjoint operator.

Fredholm Determinant

Fredholm showed that the integral equation can be analyzed through a determinant-like object:

$$D(\lambda) = \sum_{n=0}^{\infty} \frac{(-\lambda)^n}{n!} \int_a^b \cdots \int_a^b \det\left[K(x_i, x_j)\right]_{i,j=1}^n \, dx_1 \cdots dx_n$$

The equation has a unique solution when $D(\lambda) \neq 0$. The zeros of $D(\lambda)$ are precisely the eigenvalues of the kernel, analogous to zeros of $\det(A - \lambda I)$.

5. Eigenvalue Problems for Integral Equations

The homogeneous Fredholm equation defines an eigenvalue problem:

$$\int_a^b K(x,t)\,\phi(t)\, dt = \mu\, \phi(x)$$

where $\mu = 1/\lambda$ are the eigenvalues and $\phi$ are the eigenfunctions.

Hilbert-Schmidt Theory

For a symmetric, square-integrable kernel $K(x,t) = K(t,x)$:

The eigenvalues $\mu_n$ are real and form a countable set with no finite accumulation point.
The eigenfunctions $\phi_n(x)$ form a complete orthonormal set in the range of the operator.
The kernel admits the bilinear expansion (Mercer's theorem for positive kernels):
$$K(x,t) = \sum_{n=1}^{\infty} \mu_n\, \phi_n(x)\, \phi_n(t)$$
The Hilbert-Schmidt norm satisfies $\|K\|^2 = \sum_n \mu_n^2$.

Variational Characterization

The largest eigenvalue can be found variationally. For a symmetric kernel:

$$\mu_1 = \max_{\phi} \frac{\int_a^b \int_a^b K(x,t)\,\phi(x)\,\phi(t)\, dx\, dt}{\int_a^b |\phi(x)|^2\, dx}$$

Successive eigenvalues are found by restricting to the orthogonal complement of previously found eigenfunctions. This is the infinite-dimensional Rayleigh quotient.

Connection to Sturm-Liouville Problems

The Sturm-Liouville eigenvalue problem $-(p\,y')' + q\,y = \lambda\, w\, y$ with boundary conditions can be recast as an integral eigenvalue problem using the Green's function $G(x,t)$:

$$y(x) = \lambda \int_a^b G(x,t)\, w(t)\, y(t)\, dt$$

The kernel $K(x,t) = G(x,t)\, w(t)$ is symmetric (with respect to the weight $w$), and Hilbert-Schmidt theory provides the completeness of Sturm-Liouville eigenfunctions.

6. Connection to Green's Functions

Green's functions and integral equations are intimately connected. The Green's function is essentially the resolvent kernel for a differential operator.

From ODE to Integral Equation

Consider the BVP $Ly(x) = f(x)$ with $L = -d^2/dx^2 + V(x)$ on $[0, \ell]$and Dirichlet conditions $y(0) = y(\ell) = 0$. Write $L = L_0 + V$ where$L_0 = -d^2/dx^2$. If $G_0$ is the free Green's function for $L_0$:

$$y(x) = \int_0^\ell G_0(x,t)\,[f(t) - V(t)\,y(t)]\, dt = y_0(x) - \int_0^\ell G_0(x,t)\, V(t)\, y(t)\, dt$$

where $y_0(x) = \int G_0(x,t)\, f(t)\, dt$ is the solution without the potential. This is a Fredholm equation of the second kind with kernel $K(x,t) = -G_0(x,t)\, V(t)$.

Lippmann-Schwinger Equation

In quantum scattering theory, the same construction yields the Lippmann-Schwinger equation. For a particle scattering off a potential $V(r)$:

$$\psi(\mathbf{r}) = e^{i\mathbf{k}\cdot\mathbf{r}} + \int G_0^{(+)}(\mathbf{r}, \mathbf{r}')\, V(\mathbf{r}')\, \psi(\mathbf{r}')\, d^3r'$$

where $G_0^{(+)}$ is the outgoing free-particle Green's function:

$$G_0^{(+)}(\mathbf{r}, \mathbf{r}') = -\frac{m}{2\pi\hbar^2}\frac{e^{ik|\mathbf{r}-\mathbf{r}'|}}{|\mathbf{r}-\mathbf{r}'|}$$

Born Series

The Neumann series applied to the Lippmann-Schwinger equation gives the Born series:

$$\psi = \psi_0 + G_0 V \psi_0 + G_0 V G_0 V \psi_0 + \cdots$$

The first-order term gives the Born approximation, widely used in scattering calculations:

$$f(\theta) \approx -\frac{m}{2\pi\hbar^2}\int V(\mathbf{r}')\, e^{i(\mathbf{k}-\mathbf{k}')\cdot\mathbf{r}'}\, d^3r'$$

This is simply the Fourier transform of the potential evaluated at the momentum transfer$\mathbf{q} = \mathbf{k} - \mathbf{k}'$.

7. The Wiener-Hopf Technique

The Wiener-Hopf technique is a powerful method for solving integral equations on semi-infinite intervals, arising in diffraction, crack problems, and radiative transfer.

The Wiener-Hopf Equation

Consider an integral equation of the form:

$$\phi(x) = f(x) + \lambda \int_0^{\infty} K(x-t)\, \phi(t)\, dt, \qquad x > 0$$

where $K(x-t)$ is a difference kernel (translation-invariant). The key idea is to extend the equation to all of $\mathbb{R}$ and work in Fourier space.

Fourier Transform Approach

Define $\Phi_+(k) = \int_0^\infty \phi(x)\, e^{ikx}\, dx$ (analytic in upper half-plane) and introduce an unknown function $\Phi_-(k)$ for $x < 0$ (analytic in lower half-plane). The Fourier transform of the convolution gives:

$$\Phi_+(k) + \Phi_-(k) = F_+(k) + \lambda\, \hat{K}(k)\, \Phi_+(k)$$

Rearranging:

$$[1 - \lambda\, \hat{K}(k)]\, \Phi_+(k) = F_+(k) + \Phi_-(k)$$

The Factorization Step

The crucial step is to factorize $1 - \lambda\hat{K}(k)$ as a product of functions analytic in the upper and lower half-planes:

$$1 - \lambda\hat{K}(k) = \frac{H_+(k)}{H_-(k)}$$

where $H_+$ is analytic and nonzero in Im$(k) > -\tau$ and $H_-$ in Im$(k) < \tau$for some $\tau > 0$. The factorization is achieved via:

$$\ln H_{\pm}(k) = \pm\frac{1}{2\pi i}\int_{-\infty}^{\infty}\frac{\ln[1-\lambda\hat{K}(k')]}{k'-k}\, dk'$$

with the contour indented appropriately. After factorization, we separate terms analytic in the upper and lower half-planes, apply Liouville's theorem (a bounded entire function is constant), and solve for $\Phi_+$.

Application: Half-Plane Diffraction

The Sommerfeld half-plane diffraction problem reduces to a Wiener-Hopf equation. A plane wave$e^{ikx}$ incident on a semi-infinite perfectly conducting screen ($x < 0, y = 0$) produces a scattered field that satisfies a Wiener-Hopf integral equation. The factorization involves$\sqrt{k \pm k_0}$ factors, and the solution reproduces Sommerfeld's exact result including the diffracted wave that bends around the edge.

8. Applications to Scattering Theory

Scattering theory is one of the richest applications of integral equations in physics. The transition from the differential Schrodinger equation to the Lippmann-Schwinger integral equation automatically incorporates the boundary conditions of the scattering problem.

T-Matrix Formulation

Define the T-matrix (transition matrix) through:

$$V|\psi\rangle = T|\phi_0\rangle$$

where $|\psi\rangle$ is the full scattering state and $|\phi_0\rangle$ is the incident wave. The T-matrix satisfies the operator equation:

$$T = V + V G_0 T = V + V G_0 V + V G_0 V G_0 V + \cdots$$

This is again a Neumann series. The scattering amplitude is directly related to matrix elements of $T$:

$$f(\mathbf{k}' \leftarrow \mathbf{k}) = -\frac{m}{2\pi\hbar^2}\langle\mathbf{k}'|T|\mathbf{k}\rangle$$

Optical Theorem

The unitarity of the S-matrix, combined with the integral equation structure, leads to the optical theorem:

$$\boxed{\sigma_{\text{tot}} = \frac{4\pi}{k}\, \text{Im}\, f(0)}$$

The total cross section equals $4\pi/k$ times the imaginary part of the forward scattering amplitude. This remarkable result connects the total probability of scattering in any direction to a single number — the forward amplitude.

Partial Wave Analysis

For spherically symmetric potentials, the angular momentum decomposition reduces the 3D integral equation to a set of 1D equations for each partial wave $\ell$:

$$\psi_\ell(r) = j_\ell(kr) + \int_0^\infty G_\ell(r, r')\, V(r')\, \psi_\ell(r')\, r'^2\, dr'$$

where $j_\ell$ is a spherical Bessel function and $G_\ell$ is the partial-wave Green's function. The scattering phase shifts $\delta_\ell$ are extracted from the asymptotic behavior of $\psi_\ell$.

9. Python Simulation: Integral Equation Methods

Below we demonstrate the Neumann series solution, eigenvalue computation for symmetric kernels, and Born approximation for scattering — all using numpy only.

Python

script.py175 lines

#!/usr/bin/env python3
"""
integral_equations_sim.py
Neumann series, kernel eigenvalues, and Born scattering.  numpy only.
"""
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

# ─── Part 1: Neumann Series for Fredholm Eq ────────────────────────
# Solve phi(x) = f(x) + lambda * integral K(x,t) phi(t) dt
# on [0, 1] with K(x,t) = exp(-|x-t|), f(x) = x, lambda = 0.3

N = 200
x = np.linspace(0, 1, N)
dx = x[1] - x[0]

# Build kernel matrix
X, T = np.meshgrid(x, x, indexing='ij')
K = np.exp(-np.abs(X - T))
f = x.copy()
lam = 0.3

# Neumann series: phi = sum_{n=0}^inf lambda^n K^n f
phi_neumann = f.copy()
K_power_f = f.copy()  # K^0 f = f
neumann_errors = []

for n in range(1, 30):
    K_power_f = lam * K @ K_power_f * dx  # apply K once more
    phi_neumann = phi_neumann + K_power_f
    # Check convergence by looking at correction size
    correction = np.max(np.abs(K_power_f))
    neumann_errors.append(correction)
    if correction < 1e-14:
        break

# Direct solution: (I - lambda K dx) phi = f
A = np.eye(N) - lam * K * dx
phi_direct = np.linalg.solve(A, f)

fig, axes = plt.subplots(2, 2, figsize=(14, 10))

ax = axes[0, 0]
ax.plot(x, phi_direct, 'b-', lw=2.5, label='Direct (matrix solve)')
ax.plot(x, phi_neumann, 'r--', lw=2, label=f'Neumann ({len(neumann_errors)} terms)')
ax.plot(x, f, 'g:', lw=1.5, label='f(x) = x')
ax.set_xlabel('x', fontsize=12)
ax.set_ylabel(r'$\phi(x)$', fontsize=12)
ax.set_title('Neumann Series vs Direct Solution', fontsize=14, fontweight='bold')
ax.legend(fontsize=10)
ax.grid(True, alpha=0.3)

# Plot convergence
ax2 = axes[0, 1]
ax2.semilogy(range(1, len(neumann_errors)+1), neumann_errors, 'bo-', ms=5)
ax2.set_xlabel('Iteration n', fontsize=12)
ax2.set_ylabel('Max |correction|', fontsize=12)
ax2.set_title('Neumann Series Convergence', fontsize=14, fontweight='bold')
ax2.grid(True, alpha=0.3)

# ─── Part 2: Eigenvalues of Symmetric Kernel ──────────────────────
# K(x,t) = min(x,t)(1 - max(x,t)) = Green's fn for -d^2/dx^2 on [0,1]
K_green = np.minimum(X, T) * (1.0 - np.maximum(X, T))
K_sym = K_green * dx  # discrete approximation to integral operator

# Find eigenvalues
eigenvalues, eigenvecs = np.linalg.eigh(K_sym)
# Sort by magnitude (descending)
idx = np.argsort(-np.abs(eigenvalues))
eigenvalues = eigenvalues[idx]
eigenvecs = eigenvecs[:, idx]

ax3 = axes[1, 0]
# Plot first 4 eigenfunctions
colors = ['#4f46e5', '#7c3aed', '#ec4899', '#f59e0b']
for i in range(4):
    ef = eigenvecs[:, i]
    ef = ef / np.max(np.abs(ef))  # normalize for display
    ax3.plot(x, ef, color=colors[i], lw=2,
             label=f'n={i+1}, mu={eigenvalues[i]:.5f}')

# Theoretical: eigenvalues of Green's fn for -d^2/dx^2 are 1/(n*pi)^2
ax3.set_xlabel('x', fontsize=12)
ax3.set_ylabel('Eigenfunction', fontsize=12)
ax3.set_title("Eigenfunctions of Green's Function Kernel", fontsize=14, fontweight='bold')
ax3.legend(fontsize=9, loc='upper right')
ax3.grid(True, alpha=0.3)

# ─── Part 3: Born Approximation for Scattering ───────────────────
# Gaussian potential V(r) = V0 * exp(-r^2/a^2)
# Born approx: f(theta) ~ -(m/(2*pi*hbar^2)) * FT of V at q = 2k*sin(theta/2)
# For Gaussian: FT is proportional to exp(-q^2 a^2 / 4)

V0 = 1.0
a = 1.0
k = 3.0  # incident wavenumber

theta = np.linspace(0, np.pi, 300)
q = 2 * k * np.sin(theta / 2)  # momentum transfer magnitude

# Born amplitude (3D, including factors)
# f_Born(theta) = -(2m/hbar^2) * V0 * (a^3 * pi^(3/2) / (2*pi)) * exp(-q^2 a^2/4)
# In natural units with 2m/hbar^2 = 1:
f_born = -V0 * (a**3 * np.pi**(1.5)) * np.exp(-q**2 * a**2 / 4) / (2 * np.pi)

# Differential cross section
dsigma = np.abs(f_born)**2

# Second Born correction (approximate)
# sigma_Born = integral of dsigma * sin(theta) dtheta * 2*pi
sigma_born = 2 * np.pi * np.trapz(dsigma * np.sin(theta), theta)

ax4 = axes[1, 1]
ax4.semilogy(np.degrees(theta), dsigma, 'b-', lw=2.5, label='Born approximation')
ax4.set_xlabel('Scattering angle (degrees)', fontsize=12)
ax4.set_ylabel(r'$d\sigma/d\Omega$', fontsize=12)
ax4.set_title('Born Approximation: Gaussian Potential', fontsize=14, fontweight='bold')
ax4.legend(fontsize=10)
ax4.grid(True, alpha=0.3)
ax4.set_xlim(0, 180)

for ax in axes.flat:
    ax.set_facecolor('#0a0a2e')

plt.tight_layout()
plt.savefig('output.png', dpi=130, bbox_inches='tight',
            facecolor='#0a0a2e', edgecolor='none')
plt.close()

# ─── Numerical Results ─────────────────────────────────────────────
print("=" * 65)
print("INTEGRAL EQUATIONS: NUMERICAL RESULTS")
print("=" * 65)

print(f"\n--- Neumann Series (Fredholm, 2nd kind) ---")
print(f"  Kernel: K(x,t) = exp(-|x-t|) on [0,1]")
print(f"  lambda = {lam}")
print(f"  Terms needed:  {len(neumann_errors)}")
print(f"  Max |phi_Neumann - phi_direct|: {np.max(np.abs(phi_neumann - phi_direct)):.2e}")

# HS norm
hs_norm = np.sqrt(np.sum(K**2) * dx**2)
print(f"  Hilbert-Schmidt norm ||K|| = {hs_norm:.4f}")
print(f"  Convergence condition |lambda| < 1/||K||: {lam} < {1/hs_norm:.4f} => {lam < 1/hs_norm}")

print(f"\n--- Eigenvalues of Green's Function Kernel ---")
print(f"  Kernel: K(x,t) = min(x,t)(1-max(x,t))")
print(f"  (Green's function for -d^2/dx^2 on [0,1] with Dirichlet BC)")
print(f"  {'n':>4}  {'Numerical':>14}  {'Exact 1/(n*pi)^2':>16}  {'Rel Error':>12}")
for i in range(6):
    mu_exact = 1.0 / ((i+1) * np.pi)**2
    mu_num = eigenvalues[i]
    rel_err = abs(mu_num - mu_exact) / mu_exact
    print(f"  {i+1:4d}  {mu_num:14.8f}  {mu_exact:16.8f}  {rel_err:12.2e}")

print(f"\n--- Born Approximation Scattering ---")
print(f"  Potential: V(r) = {V0} * exp(-r^2/{a}^2)")
print(f"  Wavenumber: k = {k}")
print(f"  Total cross section (Born): {sigma_born:.6f}")
print(f"  Forward amplitude: f(0) = {f_born[0]:.6f}")
print(f"  Optical theorem check: sigma = 4*pi/k * Im(f(0))")
print(f"    Note: Born is real for real V, so Im(f) = 0 at 1st order")
print(f"    (Optical theorem satisfied at 2nd+ order in Born series)")

# Verify Fredholm alternative: check eigenvalue spectrum
print(f"\n--- Fredholm Alternative Check ---")
print(f"  For lambda not an eigenvalue, unique solution exists.")
print(f"  Eigenvalues (1/mu_n) of our kernel:")
for i in range(4):
    if abs(eigenvalues[i]) > 1e-10:
        print(f"    lambda_{i+1} = 1/mu_{i+1} = {1.0/eigenvalues[i]:.4f}")
print(f"  Our lambda = {lam} is not an eigenvalue => unique solution exists")

Click Run to execute the Python code

Code will be executed with Python 3 on the server

10. Summary and Connections

Integral equations provide a powerful complementary approach to differential equations, with distinct advantages in many physical applications:

Type	Key Property	Application
Fredholm (2nd kind)	Fredholm alternative	Boundary value problems
Volterra (2nd kind)	Always solvable	Initial value problems
Neumann series	Iterative convergence	Perturbation theory
Resolvent kernel	Spectral poles	Resonance theory
Eigenvalue theory	Hilbert-Schmidt completeness	Quantum bound states
Lippmann-Schwinger	Incorporates BCs naturally	Quantum scattering
Wiener-Hopf	Factorization in Fourier space	Diffraction, cracks

The integral equation viewpoint reveals deep structural connections: the Green's function is the kernel of the inverse operator, the resolvent encodes the spectral properties, and the Fredholm alternative is a manifestation of the rank-nullity theorem in infinite dimensions. In quantum field theory, the Dyson equation for propagators and the Bethe-Salpeter equation for bound states are integral equations whose structure mirrors the Fredholm theory developed here.

← Variational Calculus Part IV Overview →