In a first quantum mechanics course, tunneling probabilities are usually computed for barriers with a simple shape, a rectangular “top hat” or maybe a parabola. The WKB (Wentzel–Kramers–Brillouin) approximation lets us treat an arbitrary barrier V(x) with one and the same set of steps. The way to organise this is a transfer matrix that relates the wave amplitudes on the two sides of the barrier, and building that matrix is what this post is about.
Take a particle with energy E incident on a barrier V(x). The classical turning points, where E=V(x), sit at x=α and x=β and split the line into three regions.
Region I (x<α). Classically allowed, so the wavefunction oscillates.
Region II (α<x<β). Classically forbidden, so the wavefunction is a combination of a decaying and a growing exponential.
Region III (x>β). Classically allowed again, and this is where the transmitted wave lives.
In the allowed regions we use the local wavenumber k(x)=2m(E−V)/ℏ2, and in the forbidden region the decay constant κ(x)=2m(V−E)/ℏ2. The general WKB solutions in the three regions are then
To use them we need ψ1 in the same cosine/sine form. Multiplying and dividing the two terms by e±iπ/4 does this, and we can then match the coefficients A,B with c,d,
Let us define θ≡exp(∫αβκdx) , which is the exponential of the integral of κ over the full width of the barrier. All of the tunneling physics will end up depending on this single number.
The connection formula at β is written with integrals that start at β, while ψ2 has integrals starting at α. Splitting ∫αx=∫αβ−∫xβ and comparing with the standard form gives two simple relations,
A′=Bθ,B′=A/θ
So the exponential that grows across the barrier picks up a factor θ, and the one that decays picks up 1/θ.
For a particle coming in from the left there is no wave coming in from the right, so we set g=0. The first row of the transfer matrix then gives
a=21(2θ+2θ1)f=(θ+4θ1)f
and the transmission coefficient T=∣f∣2/∣a∣2 follows as
T=θ+4θ121=θ21+4θ2121
In the deep tunneling limitθ≫1 (a thick or tall barrier), the correction 4θ21→0 and we are left with
T≃θ−2=exp(−2∫αβκ(x)dx)
This is the standard WKB tunneling formula. The transmission probability is the exponential of −2 times the integral of κ(x) over the classically forbidden region. I think the nice part is that the shape of the barrier only enters through that one integral. This is why WKB is the standard estimate for tunneling rates in alpha decay, scanning tunneling microscopy and field emission.