Measured vs predicted (independent check)
T measured (final, from the PDE)
T predicted: packet average of T(k)
T plane wave at mean energy
|measured - predicted|
R + T measured
The PDE is solved numerically; the prediction uses the exact closed-form transmission of a rectangular barrier averaged over the packet's momentum distribution. They share no code.
The science
A quantum particle is described by a wave function ψ(x,t); |ψ|² is the probability density of finding it at x. In units ħ = m = 1 it obeys the time-dependent Schrödinger equation:
i dpsi/dt = -1/2 d2psi/dx2 + V(x) psi
V(x) = V0 for |x| < a/2, else 0
Plane wave, E = k^2/2 < V0:
T(k) = 1 / (1 + V0^2 sinh^2(kappa a) / (4 E (V0 - E))), kappa = sqrt(2 (V0 - E))
E > V0:
T(k) = 1 / (1 + V0^2 sin^2(q a) / (4 E (E - V0))), q = sqrt(2 (E - V0))
Inside the barrier the wave does not stop dead: for E < V0 it becomes a decaying exponential e^(-κx), so a fraction of the wave leaks through. That is tunnelling. Classically the transmission would be exactly 0 below the barrier and exactly 1 above it. Quantum mechanically even E > V0 gives partial reflection, and when q a = nπ the reflections from both edges cancel and transmission reaches 100% (resonant transmission).
A single packet does not split into two particles: |ψ|² splits into a reflected and a transmitted part, and a measurement would find the particle in one of them with probabilities R and T (R + T = 1).
Method: split-operator Fourier method (Strang splitting: half potential step, full kinetic step in k-space via FFT, half potential step), 2048 grid points on x in [-60, 60], time step 0.01. It is unitary, so the norm is conserved to rounding error, and second-order accurate in time. The barrier edges are smoothed over about 0.08 length units (a tanh profile, roughly one grid cell) because a perfectly sharp edge makes the splitting scheme ring; with the smoothing, energy is conserved to about 1e-4. Weak absorbing layers at the far edges stop wrap-around. The final T is measured when the packet has left the barrier region. The prediction averages T(k) over the packet's Gaussian momentum distribution.
Honest simplifications: one space dimension; the picture is |ψ|² (the colours encode the phase); the 'particle' is a 1D wave packet in dimensionless units, not a specific atom or electron; a finite-width packet always has some spread in energy, so measured values match the packet-averaged prediction rather than the single-energy curve; measuring the particle is not modelled.