Setting up the microwave bench…
-

📡 Microwave bench controls

-

Time t-
Wavelength λ (air / medium)-
Cells per λ   dt-
Fresnel R (one face / slab)-
Energy UE / UH (fJ per m)-
Ez at the probe (last 1.2 ns)
mean intensity along the white line
The science

Model. Maxwell's curl equations for the transverse-magnetic (TMz) polarisation, ε0εr ∂Ez/∂t = ∂Hy/∂x − ∂Hx/∂y, μ0 ∂Hx/∂t = −∂Ez/∂y, μ0 ∂Hy/∂t = ∂Ez/∂x, solved by Yee's leapfrog scheme: E and H live on interleaved grids advanced half a step apart, so both curls are centred differences (second-order accurate).

Stability (CFL). In 2D the explicit scheme needs c·dt ≤ dx/√2; the slider shows ν = c dt √2 / dx and the page resets with a message if you push it above 1.

Energy and Poynting. u = ½(εE2 + μH2), S = E × H, ∂u/∂t + ∇·S = 0. In the closed metal cavity the discrete energy ½εE2 + ½μHoldHnew is constant to round-off.

Fresnel. One vacuum–dielectric face reflects R = ((n−1)/(n+1))2 of the power; a slab of thickness d gives the Airy result R = F sin2δ / (1 + F sin2δ), F = 4R1/(1−R1)2, δ = 2π n d/λ; a quarter-wave layer (d = λ/4n) is an anti-reflection coating when n = √nsubstrate.

Double slit. Bright directions satisfy d sinθ = mλ (dashed lines) under an envelope set by the slit width; with only 12 cells per slit the pattern is a near-field Fresnel pattern and the lines are asymptotic guides.

Absorbing edges. A graded lossy layer with σ/ε = σ*/μ keeps the impedance matched and soaks up outgoing waves; it reflects a little at oblique incidence.

Honest simplifications. 2D only, one polarisation, lossless non-dispersive dielectrics, a 1.5 mm grid (a microwave-scale model; the physics is scale-free in wavelengths). Grid dispersion makes short waves slightly slow (see the cavity check). The colours are Ez (blue negative, orange positive) or its smoothed square.

Independent checks (run on load, same core as the 3D page):

running…
Made with 🟠 Blender