Part 9 of 10 · Maxwell's Equations & Electromagnetic Theory Series overview →

This article builds on Part 8: Deriving the Electromagnetic Wave Equation. Next up: Part 10: Applications of Maxwell's Equations in RF Engineering and Antennas.

Fig 1 — An interface between two media: a pillbox (Gauss's laws → normal components) and a thin rectangular loop (Faraday / Ampère–Maxwell → tangential components)
Medium 1: ε₁, μ₁, σ₁ Medium 2: ε₂, μ₂, σ₂ interface: ρ_s and 𝐉_s may exist on it n̂ D₁ₙ, B₁ₙ D₂ₙ, B₂ₙ Pillbox, height Δh → 0 E₁ₜ, H₁ₜ (length Δw) E₂ₜ, H₂ₜ Rectangular loop, height Δh → 0

Introduction to Electromagnetic Boundary Conditions

The differential forms of Maxwell's equations hold at every point inside a uniform material. But real RF hardware is built from many materials joined together: copper traces on a laminate, a laminate in air, a coaxial dielectric inside a metal shield, an antenna behind a plastic radome. At each interface ε, μ or σ changes abruptly, the fields may jump, and derivatives such as ∇·𝐃 or ∇ × 𝐄 are not defined.

Boundary conditions tell us how the fields on one side of an interface are related to the fields on the other. They follow directly from the integral forms of Maxwell's equations (Part 7) and, together with the wave solutions of Part 8, they let us solve real problems: how much power reflects from a substrate, why fields concentrate in a microstrip dielectric, how currents flow on a ground plane, and how a metal shield blocks a wave.

Understanding Material Interfaces

Boundary Between Two Dielectric Materials

Two insulators with different permittivities — air and FR-4, PTFE and a ceramic filler, glass and air. There is normally no free surface charge and no surface current, so the fields change smoothly in a well-defined way: tangential 𝐄 and 𝐇 are continuous, and so are normal 𝐃 and 𝐁. Because ε differs, normal 𝐄 jumps and the field lines bend (refract) at the interface.

Boundary Between Dielectric and Conductive Materials

A dielectric against a good conductor — the substrate–copper interface of every PCB, the inside wall of a waveguide, the dielectric of a coaxial cable. Inside a good conductor the time-varying fields decay within a skin depth, so at RF a metal behaves almost as a perfect electric conductor (PEC): the fields inside are zero, and the field outside terminates on surface charges and surface currents.

Normal and Tangential Field Components

At any point on an interface, define a unit normal n̂ pointing from medium 2 into medium 1. Any field vector can be split into a normal component along n̂ (e.g. D_n = 𝐃·n̂) and a tangential component lying in the plane of the interface (𝐄_t = 𝐄 − E_n n̂). Boundary conditions are always stated separately for these two parts: Gauss's laws control normal components, and the curl equations control tangential components.

Boundary Conditions for Electric Fields

Tangential Electric Field

Tangential 𝐄 is continuous
E1t=E2t 𝐧^×(𝐄1−𝐄2)=0
Holds at every interface, static or time-varying (there are no magnetic surface currents in nature). At a perfect conductor, where 𝐄₂ = 0, it requires E₁t = 0: the electric field meets a conductor at right angles.

Normal Electric Flux Density

Normal 𝐃 jumps by the surface charge density
D1n−D2n=ρs 𝐧^·(𝐃1−𝐃2)=ρs
ρ_s (C/m²) is the free surface charge on the interface. Between two dielectrics ρ_s = 0, so D_n is continuous and E_n jumps in the ratio ε₂/ε₁: ε₁E₁n = ε₂E₂n.

Effect of Surface Charge Density

Surface charge is what lets the electric field start or stop at an interface. On a conductor, every electric field line arriving from outside ends on a surface charge, and Gauss's law gives ρ_s = D_n = εE_n. At RF, these surface charges oscillate with the field, and their motion along the surface is the RF surface current. Between two dielectrics without free charge, the normal field changes by the permittivity ratio — and at a sloping interface this bends the field lines:

Refraction of static field lines at a charge-free dielectric interface
tanθ1tanθ2=ε1ε2
θ is measured from the normal. Field lines bend away from the normal on entering a higher-ε material: the field inside a high-ε dielectric is pulled more nearly parallel to the surface.

Boundary Conditions for Magnetic Fields

Tangential Magnetic Field

Tangential 𝐇 jumps by the surface current density
𝐧^×(𝐇1−𝐇2)=𝐉s |H1t−H2t|=Js
𝐉_s (A/m) is a current flowing in an infinitely thin sheet on the interface. Between two dielectrics 𝐉_s = 0 and H_t is continuous. At a finite-conductivity metal, the current spreads over a skin depth and is not a true sheet, so H_t is still continuous mathematically — but the PEC approximation treats it as a sheet.

Normal Magnetic Flux Density

Normal 𝐁 is continuous
B1n=B2n 𝐧^·(𝐁1−𝐁2)=0
Always true, since there are no magnetic charges. As a consequence μ₁H₁n = μ₂H₂n: normal H jumps by the permeability ratio — the effect that concentrated all the magnetomotive force in the air gap of the ferrite core in Part 7.

Effect of Surface Current Density

Surface current is how a conductor terminates a tangential magnetic field. On a PEC, 𝐇 inside is zero, so the tangential field just outside equals the surface current: J_s = H_t, flowing perpendicular to 𝐇. On a microstrip ground plane, the return current flows exactly where the tangential H of the trace is strongest — directly beneath the trace — which is why a slot or gap in the ground plane under a trace forces the return current to detour, increasing loop area, inductance and radiation.

Derivation of Electromagnetic Boundary Conditions From Maxwell's Equations

Normal components — the pillbox. Apply Gauss's law ∮𝐃·d𝐀 = Q_enc to a short cylinder (pillbox) of face area ΔA straddling the interface, as in Fig 1. Let its height Δh → 0. The curved side wall then has vanishing area and contributes no flux. The top face contributes D₁n ΔA, the bottom face −D₂n ΔA (its outward normal points into medium 2), and the enclosed charge is ρ_s ΔA:

Pillbox result
D1nΔA−D2nΔA=ρsΔA ⟹ D1n−D2n=ρs
The identical argument with ∮𝐁·d𝐀 = 0 gives B₁n = B₂n.

Tangential components — the thin loop. Apply Faraday's law ∮𝐄·d𝐥 = −d/dt ∫𝐁·d𝐀 to a rectangular loop of length Δw along the interface and height Δh → 0. The short sides vanish, and so does the flux through the loop (its area Δw·Δh → 0, and 𝐁 is finite). The long sides give E₁t Δw − E₂t Δw = 0, so E₁t = E₂t.

Applying the Ampère–Maxwell law to the same loop, the displacement-current flux also vanishes as Δh → 0, but a surface current — a finite current squeezed into zero thickness — still threads the loop: H₁t Δw − H₂t Δw = J_s Δw. Hence H₁t − H₂t = J_s (with 𝐉_s perpendicular to the tangential 𝐇).

FieldGeneral interfaceTwo dielectrics (ρ_s = 0, 𝐉_s = 0)Medium 1 against a perfect conductorFrom
Tangential 𝐄E₁t = E₂tE₁t = E₂tE₁t = 0Faraday
Normal 𝐃D₁n − D₂n = ρ_sε₁E₁n = ε₂E₂nD₁n = ρ_sGauss
Tangential 𝐇n̂ × (𝐇₁ − 𝐇₂) = 𝐉_sH₁t = H₂tn̂ × 𝐇₁ = 𝐉_sAmpère–Maxwell
Normal 𝐁B₁n = B₂nμ₁H₁n = μ₂H₂nB₁n = 0Gauss (magnetic)
📌 Rules of thumb at a metal surface

Electric field lines hit metal perpendicularly (no tangential 𝐄); magnetic field lines skim parallel to metal (no normal 𝐁). Remembering these two pictures is enough to sketch the field pattern in a coax, a waveguide or a microstrip line.

Reflection and Transmission at Material Boundaries

Normal Incidence

A uniform plane wave travelling in medium 1 (impedance η₁) hits a flat interface with medium 2 (η₂) head-on. Part of it is reflected back into medium 1 and part is transmitted into medium 2. In medium 1 the total field is the sum of incident and reflected waves; in medium 2 only the transmitted wave exists. Both 𝐄 and 𝐇 are tangential to the interface, so with no surface current both must be continuous:

Continuity of tangential 𝐄 and 𝐇 at z = 0
Ei+Er=Et Eiη1−Erη1=Etη2
The minus sign arises because the reflected wave travels backwards, so its 𝐇 is reversed relative to its 𝐄 (𝐄 × 𝐇 must point in its direction of travel).

Reflection Coefficient

Reflection coefficient at normal incidence
Γ=ErEi=η2−η1η2+η1 non-magnetic: Γ=εr1−εr2εr1+εr2
Exactly the same form as the transmission-line reflection coefficient Γ = (Z_L − Z₀)/(Z_L + Z₀). A perfect conductor (η₂ = 0) gives Γ = −1: total reflection with a phase reversal, forming a standing wave with an 𝐄-field null at the metal.

Transmission Coefficient

Transmission coefficient and power balance
τ=EtEi=2η2η2+η1=1+Γ PrPi=|Γ|2 PtPi=1−|Γ|2
τ can exceed 1 (when η₂ > η₁) without violating energy conservation, because the power density in medium 2 is E_t²/η₂ — the power fractions always add to one (for lossless media).

Impedance Matching

Reflections vanish when η₁ = η₂. When two different media must be joined — air to a dielectric lens, air to a radome, one laminate to another — a quarter-wave matching layer can cancel the reflection at one frequency. The layer has an intrinsic impedance equal to the geometric mean of the two media and a thickness of one quarter of the wavelength inside it, so the reflections from its front and back faces return half a wavelength apart and cancel:

Quarter-wave matching layer
ηm=η1η2 ⟺ εr,m=εr1εr2 d=λ04εr,m
The direct analogue of the quarter-wave transformer Z_T = √(Z₀Z_L) on a transmission line. Anti-reflection coatings on camera lenses use the same principle at optical wavelengths.

Applications in RF and Microwave Engineering

Dielectric Substrates in Microstrip Circuits

A microstrip line has fields partly in the substrate (ε_r) and partly in the air above. The boundary conditions at the substrate–air surface — continuous E_t and continuous D_n — force a field pattern that is neither entirely in air nor entirely in the dielectric. Because the dielectric and air would support different wave speeds, no single TEM wave satisfies the boundary conditions exactly; the result is a "quasi-TEM" mode with an effective permittivity ε_eff between 1 and ε_r (Part 10). At the copper strip and ground plane, the PEC conditions make 𝐄 perpendicular to the metal and place the surface charge and current on the facing surfaces — and the current crowds at the strip's edges, where the field is strongest.

Metal-Conductor Interfaces

For a good conductor the PEC model gives the field pattern, and the finite conductivity gives the loss. The tangential H at the surface sets the surface current J_s = H_t, and the power dissipated per unit area is ½|J_s|²R_s with surface resistance R_s = 1/(σδ) (Part 8). This "perturbation" approach is how conductor loss in waveguides, cavities and microstrip is calculated, and why smooth, highly conductive plating (silver, gold) on the surfaces carrying current improves Q.

Antenna Substrate Interfaces

A patch antenna radiates from the fringing fields at its edges, where the substrate meets air. The boundary conditions there shape how much field escapes into air (radiation) and how much stays bound in the substrate (stored energy and surface waves). A thick, high-ε_r substrate can launch surface waves trapped by total internal reflection at the substrate–air boundary, lowering efficiency and distorting the pattern. Radomes and dielectric lenses are additional interfaces in the radiated beam; their reflections are controlled by thickness — making the wall a half-wavelength thick inside the material, or very thin compared with λ — or by matching layers.

Electromagnetic Shielding

A metal shield attenuates an incident wave in two ways, both direct consequences of the boundary conditions and wave solutions:

  • Reflection loss — the huge impedance mismatch between free space (377 Ω) and metal (|η_c| = √(2πfμ/σ), milliohms) gives |Γ| ≈ 1 at the front surface; most of the incident power bounces off.
  • Absorption loss — the fraction that enters decays by 8.7 dB per skin depth inside the metal.

For copper at 1 GHz, |η_c| ≈ 0.0117 Ω, so |Γ| ≈ 0.99996 and only about 0.009% of the incident power (≈ −41 dB) enters the metal (Example 5) — and a 35 μm sheet is 17 skin depths thick, absorbing a further ≈ 145 dB. In practice, shielding is limited not by the metal but by seams, slots, apertures and cable penetrations, through which fields leak.

Numerical Examples

01

Electric Field Crossing an Air–Dielectric Interface

Problem: A static field of 1000 V/m in air (ε_r1 = 1) meets a charge-free dielectric (ε_r2 = 4) at 60° from the normal. Find the magnitude and direction of the field in the dielectric.

Components in air: E₁t = 1000 sin 60° = 866 V/m; E₁n = 1000 cos 60° = 500 V/m
E2t=866 V/m E2n=ε1ε2E1n=5004=125 V/m E2=8662+1252=875 V/m
Angle in the dielectric
θ2=tan−1866125=81.8° check: tan60°tan81.8°=1.7326.93=0.25=ε1ε2 ✓
Answer: E₂ = 875 V/m at 81.8° from the normal (nearly parallel to the surface)

The normal component is reduced four-fold while the tangential component passes unchanged, so the field bends towards the interface.

02

Surface Charge and Surface Current on a Conductor

Problem: Just above a ground plane in air, a probe measures a normal electric field of 50 kV/m and a tangential magnetic field of 40 A/m. Treating the plane as a perfect conductor, find the surface charge density and surface current density.

ρ_s = D_n = ε₀E_n; J_s = H_t
ρs=8.854×10−12×5×104=0.443 μC/m² Js=40 A/m
Answer: ρ_s = 0.443 μC/m²  |  J_s = 40 A/m, flowing perpendicular to the tangential 𝐇

J_s = 40 A/m means 40 mA flows across every millimetre of width — measuring H_t just above a ground plane maps the return-current distribution directly, which is exactly what a near-field H-probe scanner does.

03

Reflection From an FR-4 Surface

Problem: A plane wave in air strikes a thick slab of FR-4 (ε_r = 4.4, treat as lossless) at normal incidence. Find Γ, τ, the fraction of power reflected and transmitted, and the return loss.

η₂ = 376.7/√4.4 = 376.7/2.098 = 179.6 Ω
Γ=179.6−376.7179.6+376.7=−0.354 τ=1+Γ=0.646 |Γ|2=0.126 RL=−20log100.354=9.0 dB
Answer: Γ = −0.354, τ = 0.646  |  12.6% of power reflected, 87.4% transmitted  |  RL = 9.0 dB

The negative Γ means the reflected 𝐄 is phase-reversed (wave going into a lower-impedance medium). Placing a device behind a plastic or FR-4 cover therefore causes real reflections that detune antennas — unless the cover thickness is chosen to cancel them.

04

Quarter-Wave Matching Layer for an Alumina Window

Problem: A 10 GHz beam passes from air into a thick alumina window (ε_r = 9.8). (a) What fraction of power is reflected without matching? (b) Design a quarter-wave matching layer.

(a) √9.8 = 3.130
Γ=1−3.1301+3.130=−0.516 |Γ|2=0.266
(b) ε_r,m = √(1 × 9.8) = 3.13; λ₀ = 29.98 mm
d=29.9843.13=29.984×1.769=4.24 mm
Answer: (a) 26.6% reflected — return loss 5.7 dB, transmission loss 1.3 dB  |  (b) a 4.24 mm layer with ε_r ≈ 3.1

With the matching layer the reflection is cancelled at 10 GHz and remains small over roughly ±10–20% bandwidth. Materials around ε_r ≈ 3 (e.g. some filled PTFE or polymer composites) are readily available. The same calculation designs anti-reflection layers on dielectric lenses for mmWave radar.

05

Reflection From a Copper Shield at 1 GHz

Problem: A 1 GHz plane wave in air hits a copper sheet (σ = 5.8 × 10⁷ S/m). Find the magnitude of copper's intrinsic impedance, the reflection coefficient, and the fraction of power that enters the metal.

Good conductor: η_c = (1 + j)√(πfμ/σ), |η_c| = √(2πfμ/σ)
|ηc|=2π×109×4π×10−75.8×107=0.0117 Ω Γ≈−1+2ηcη0 1−|Γ|2≈4Rsη0=4×0.00825376.7=8.8×10−5
Answer: |η_c| ≈ 0.0117 Ω  |  |Γ| ≈ 0.99996  |  ≈ 0.009% of incident power (−40.6 dB) enters the copper

R_s = Re{η_c} = 8.25 mΩ (Part 8). The 1 − |Γ|² ≈ 4R_s/η₀ result is also the fraction of power lost each time a wave reflects from a metal wall — the basis of conductor-loss and cavity-Q calculations. Copper is nearly a perfect mirror at RF, which justifies the PEC boundary conditions used throughout RF design.

Summary

  • Boundary conditions follow from applying the integral forms of Maxwell's equations to vanishingly thin pillboxes (normal components) and loops (tangential components).
  • Tangential 𝐄 and normal 𝐁 are always continuous; normal 𝐃 jumps by the surface charge ρ_s; tangential 𝐇 jumps by the surface current 𝐉_s.
  • At a perfect conductor 𝐄 is purely normal and 𝐇 purely tangential, terminated by surface charge and surface current.
  • At normal incidence the reflection coefficient Γ = (η₂ − η₁)/(η₂ + η₁) has the same form as on a transmission line; a quarter-wave layer of impedance √(η₁η₂) cancels the reflection.
  • The same conditions explain microstrip quasi-TEM fields, ground-plane return currents, antenna substrate effects, radome reflections and shielding.

In the final article, Part 10: Applications of Maxwell's Equations in RF Engineering and Antennas, we put all of these tools to work on transmission lines, waveguides, microstrip circuits and antennas, and look at how EM simulators solve Maxwell's equations numerically.

Put It Into Practice

Use the NEXT ANTENNAS calculators to apply the ideas from this article to real designs.