Boundary conditions describe how electromagnetic fields behave at the interface between two different media. They're derived from Maxwell's equations and enforce physical constraints (like charge conservation and energy conservation) across every boundary. Without them, you can't solve any problem that involves more than one material or region.

Derivation from Maxwell's equations

The boundary conditions come from applying the integral forms of Maxwell's equations to tiny regions that straddle the interface.

For the normal component conditions, you use Gauss's law. Picture a small pillbox-shaped volume with one flat face in medium 1 and the other in medium 2. As you shrink the pillbox height to zero, the volume contribution vanishes and you're left with a relationship between the normal field components on each side.

For the tangential component conditions, you use Faraday's law. Draw a small rectangular loop with its long sides parallel to the interface, one in each medium. Shrinking the loop height to zero eliminates any flux through the loop, giving you a condition on the tangential fields.

The key mathematical tools here are the divergence theorem (for the pillbox argument) and Stokes' theorem (for the loop argument).

Electric field component perpendicular to boundary

The normal component of $\vec{D}$ is generally discontinuous across a boundary. The jump is set by the free surface charge density $\sigma_f$ at the interface:

$\hat{n} \cdot (\vec{D}_2 - \vec{D}_1) = \sigma_f$

Here $\hat{n}$ points from medium 1 into medium 2. Physically, this says the net outward electric flux from the pillbox equals the enclosed free charge. If there's no free charge at the interface ( $\sigma_f = 0$ ), then $D_{1n} = D_{2n}$ .

Note: this involves $\vec{D}$ , not $\vec{E}$ . Since $\vec{D} = \epsilon \vec{E}$ , the normal component of $\vec{E}$ itself will generally be discontinuous even when $\sigma_f = 0$ , because $\epsilon$ changes across the boundary.

Electric field component parallel to boundary

The tangential component of $\vec{E}$ is always continuous across the boundary:

$\hat{n} \times (\vec{E}_2 - \vec{E}_1) = 0$

This follows from Faraday's law applied to the rectangular loop. Because $\nabla \times \vec{E} = 0$ in electrostatics (or $-\partial \vec{B}/\partial t$ in the time-varying case, but the loop area shrinks to zero), the line integral around the loop vanishes, forcing $E_{1t} = E_{2t}$ .

This also guarantees that the electrostatic potential is continuous across the boundary.

Boundary conditions for perfect conductors

A perfect conductor is an idealization with infinite conductivity ( $\sigma \to \infty$ ). Real metals like copper or aluminum approximate this well at low frequencies.

Electric field inside ideal conductors

Under static (or steady-state) conditions, the electric field inside a perfect conductor is exactly zero:

$\vec{E}_{\text{inside}} = 0$

Why? If any field existed inside, the free charges would experience a force and move. With infinite conductivity, they redistribute instantaneously until the internal field is completely canceled. All excess charge ends up on the surface.

Surface charge density on conductors

Since $\vec{E} = 0$ inside and $\vec{D} = \epsilon_0 \vec{E}$ in the vacuum/dielectric region outside, the general boundary condition reduces to:

$\hat{n} \cdot \vec{D}_{\text{outside}} = \sigma_f \quad \Longrightarrow \quad E_n^{\text{outside}} = \frac{\sigma_f}{\epsilon_0}$

where $\hat{n}$ points outward from the conductor surface. The surface charge density $\sigma_f$ adjusts itself to whatever value is needed to terminate the external field lines. Its distribution depends on the conductor geometry and any applied external fields.

Tangential electric field at conductor surface

The tangential electric field at the surface of a perfect conductor is zero:

$\hat{n} \times \vec{E} = 0$

This follows directly from the continuity of $E_t$ across the boundary combined with $\vec{E} = 0$ inside. A consequence: the conductor surface is an equipotential surface. Any path along the surface has zero tangential field, so no work is done moving a charge along it, meaning the potential doesn't change.

Boundary conditions for dielectrics

Dielectrics are insulating materials that become polarized in an external electric field. Unlike conductors, they support nonzero internal fields.

Derivation from Maxwell's equations, Maxwell's equations - Wikipedia

Electric field in dielectrics vs conductors

The electric field inside a dielectric is generally not zero. Instead, the applied field displaces bound charges slightly from equilibrium, creating a polarization $\vec{P}$ . This polarization partially opposes the applied field but doesn't cancel it.

The total electric flux density accounts for both free and bound charge effects:

$\vec{D} = \epsilon_0 \vec{E} + \vec{P} = \epsilon \vec{E}$

where $\epsilon = \epsilon_0 \epsilon_r$ is the permittivity of the dielectric and $\epsilon_r$ is the relative permittivity (dielectric constant).

Normal component of electric displacement field

The normal component of $\vec{D}$ obeys the same condition as at any interface:

$\hat{n} \cdot (\vec{D}_2 - \vec{D}_1) = \sigma_f$

For two dielectrics with no free charge at the interface ( $\sigma_f = 0$ ), this simplifies to:

$\epsilon_1 E_{1n} = \epsilon_2 E_{2n}$

So the normal component of $\vec{E}$ jumps by the ratio of permittivities. A higher-permittivity material will have a smaller normal $\vec{E}$ component.

Tangential component of electric field

The tangential component of $\vec{E}$ remains continuous, just as at any interface:

$\hat{n} \times (\vec{E}_2 - \vec{E}_1) = 0 \quad \Longrightarrow \quad E_{1t} = E_{2t}$

Combined with the normal condition, these two equations fully determine how the field refracts at a dielectric interface. You can derive a "Snell's law" analog for electrostatics:

$\frac{\tan\theta_1}{\tan\theta_2} = \frac{\epsilon_1}{\epsilon_2}$

where $\theta_1$ and $\theta_2$ are the angles the field makes with the surface normal in each medium.

Surface charge density at dielectric boundaries

Even without free charge, a bound surface charge $\sigma_b$ appears at the interface due to the discontinuity in polarization:

$\sigma_b = \hat{n} \cdot (\vec{P}_1 - \vec{P}_2)$

Note the sign convention: $\hat{n}$ points from medium 1 into medium 2, and the bound charge equals the polarization "leaving" medium 1 minus that "entering" medium 2. This bound charge is what causes the normal component of $\vec{E}$ to be discontinuous even when no free charge is present.

Boundary conditions for permeable materials

Although this topic focuses on electric fields, the magnetic boundary conditions are closely analogous and often appear alongside them in boundary-value problems.

Permeable materials are characterized by their magnetic permeability $\mu$ , relating $\vec{B} = \mu \vec{H}$ .

Magnetic field boundary conditions

The magnetic conditions mirror the electric ones, with the roles of $\vec{D}$ / $\vec{E}$ swapped for $\vec{B}$ / $\vec{H}$ :

The normal component of $\vec{B}$ is continuous (no magnetic monopoles).
The tangential component of $\vec{H}$ can be discontinuous if surface currents are present.

Normal component of magnetic flux density

$\hat{n} \cdot (\vec{B}_2 - \vec{B}_1) = 0$

This comes from $\nabla \cdot \vec{B} = 0$ (Gauss's law for magnetism) applied to the pillbox. There's no magnetic charge analog to $\sigma_f$ , so the normal component of $\vec{B}$ is always continuous.

Tangential component of magnetic field intensity

$\hat{n} \times (\vec{H}_2 - \vec{H}_1) = \vec{K}_f$

where $\vec{K}_f$ is the free surface current density (units: A/m). If no surface current exists, $H_{1t} = H_{2t}$ , and the normal $\vec{B}$ condition then gives:

$\frac{\tan\theta_1}{\tan\theta_2} = \frac{\mu_1}{\mu_2}$

This is the magnetic analog of the electrostatic refraction law.

Derivation from Maxwell's equations, The Divergence Theorem · Calculus

Applications of boundary conditions

Solving electrostatic boundary value problems

Boundary conditions provide the constraints needed to find unique solutions to Laplace's equation ( $\nabla^2 V = 0$ ) or Poisson's equation ( $\nabla^2 V = -\rho/\epsilon$ ). Common solution techniques include:

Method of images: Replace a boundary (e.g., a grounded conductor) with fictitious image charges that enforce the correct boundary condition.
Separation of variables: Assume a product-form solution and apply boundary conditions to determine the coefficients.
Numerical methods: Finite difference, finite element, and boundary element methods discretize the domain and enforce boundary conditions at each interface.

In every case, the boundary conditions are what make the solution unique (via the uniqueness theorem).

Capacitors with multiple dielectric layers

For a parallel-plate capacitor with multiple dielectric layers stacked between the plates:

$D_n$ is continuous across each interface (assuming no free charge between layers), so $D$ is the same in every layer.
$E$ differs in each layer: $E_i = D / \epsilon_i$ .
The total voltage is the sum of the voltage drops across each layer: $V = \sum_i E_i \, d_i$ .
The result is equivalent to capacitors in series, with $\frac{1}{C_{\text{total}}} = \sum_i \frac{1}{C_i}$ .

For layers side by side (parallel to the plates), $E_t$ is continuous instead, and the layers act as capacitors in parallel.

Transmission and reflection at dielectric interfaces

When an electromagnetic wave hits a boundary between two dielectrics, boundary conditions on both $\vec{E}$ and $\vec{H}$ determine the reflected and transmitted wave amplitudes. This leads to the Fresnel equations, which give the reflection and transmission coefficients as functions of the angle of incidence and the refractive indices ( $n = \sqrt{\epsilon_r \mu_r}$ ) of the two media.

Special cases include Brewster's angle (zero reflection for one polarization) and total internal reflection (when going from a higher to lower refractive index beyond the critical angle).

Waveguides and cavity resonators

In waveguides and resonators, the conducting walls impose $E_t = 0$ and $B_n = 0$ at every wall surface. These constraints:

Restrict the allowed field patterns to discrete modes (TE, TM, or TEM).
Set cutoff frequencies below which a given mode cannot propagate.
Determine resonant frequencies in closed cavities.

The boundary conditions effectively quantize the transverse wavenumbers, which is why only certain field configurations can exist inside these structures.

Limitations and approximations

Ideal vs real material properties

The boundary conditions above assume ideal material properties: perfect conductors with $\sigma \to \infty$ , lossless dielectrics with real-valued $\epsilon$ , and sharp interfaces. Real materials deviate from these assumptions, and the deviations matter more at higher frequencies and smaller length scales.

Finite conductivity and dielectric loss

Real conductors have finite conductivity, so fields penetrate a small distance into the surface (the skin depth $\delta = \sqrt{2/(\omega \mu \sigma)}$ ). The "perfect conductor" boundary condition $E_t = 0$ is a good approximation only when the skin depth is much smaller than other relevant dimensions.

Dielectrics exhibit loss from mechanisms like dipole relaxation and ionic conduction. This is captured by a complex permittivity $\epsilon = \epsilon' - j\epsilon''$ , which modifies the field distributions and introduces attenuation.

Local vs macroscopic electric fields

Boundary conditions apply to macroscopic fields, which are spatial averages over many atoms. At the nanoscale or near surfaces, the local field can differ significantly from the macroscopic value. For example, the Clausius-Mossotti relation connects the macroscopic permittivity to the microscopic polarizability, but the local field correction becomes important when you're working at scales comparable to interatomic spacing.

Boundary conditions in time-varying fields

For time-varying fields, the boundary conditions on $\vec{E}$ and $\vec{D}$ still hold in the same form. The tangential $\vec{E}$ condition remains valid because the magnetic flux through the shrinking Faraday loop vanishes as the loop area goes to zero, regardless of $\partial \vec{B}/\partial t$ .

What changes in the time-varying case is that $\vec{E}$ and $\vec{B}$ are now coupled through Faraday's law and the Ampère-Maxwell equation. You must solve the boundary conditions for electric and magnetic fields simultaneously, which is exactly what leads to the Fresnel equations and waveguide mode analysis discussed above.

2,589 studying →