0
$\begingroup$

Goal: to compute accurate tunneling probabilities with an arbitrary barrier using numerical solutions to Schrodinger's equation.

Main question: what method is best for generating a robust numerical solution to Schrodinger equation for tunneling through an arbitrary barrier? A resource link would be great.

Current solution: I have a time-dependent numerical solution already made using a moving Gaussian wave packet toward a potential barrier.

The tunneling probability is computed as $$T_{prob}=\frac{\Sigma\psi_3^2}{\Sigma\psi_1^2}$$ where $\psi_1, \psi_2, \psi_3$ are the wave function in regions 1, 2, and 3. Region 1 is before the barrier where $V=0$, region 2 is inside the barrier where $E<V$, and region 3 is after the barrier where again $V=0$.

I've discovered, however, that the tunneling probability is dependent on the wavepacket starting position and of course the width $\sigma$ of the Gaussian wave packet. Is there some way to solve this? Wavepacket broadening is expected in a time-dependent solution like this.

I am just looking for a robust method. Perhaps I am missing a concept with this simulation, or there is a better method. The resources I have found almost exclusively focus on bound systems, not tunneling.

Some resources used: https://arxiv.org/html/2403.13857v1#S3 https://www.reed.edu/physics/faculty/wheeler/documents/Quantum%20Mechanics/Miscellaneous%20Essays/Gaussian%20Wavepackets.pdf

$\endgroup$
3
  • $\begingroup$ You might want to look at Numerov's method. I remember from my computational physics days this is especially useful for a Schrödinger-type differential equation. I once used it for scattering off an inverted Morse potential, works very well. $\endgroup$
    – Guliano
    Commented Jun 12 at 5:50
  • $\begingroup$ I can't recommend the link, but instead of considering TDSE, I would rather solve stationary Schroedinger equation with BC representing a plane wave. Tunneling probability can be then extracted from the transmission coefficient. This will at least solve the problem with the dependence on the wavepacket form. $\endgroup$
    – E. Anikin
    Commented Jun 12 at 11:33
  • $\begingroup$ @E.Anikin How would I set up boundary conditions for TISE numerically? I have a solution for this, but get inconsistent results because I am just trying to set $\psi = 0$ at the edges, but this makes it inherently a bound eigenvalue problem, which I am not looking for. The Eigenvalue shoudl be the kinetic energy of the incoming particle. $\endgroup$ Commented Jun 12 at 23:56

1 Answer 1

1
$\begingroup$

I would consider stationary Schroedinger equation instead of solving TDSE: $$ -\frac{1}{2}\frac{\partial^2 \psi}{\partial x^2} = (E - V(x))\psi $$ For large $|x|$, the solution can be written as $$ \psi = \begin{cases} e^{-ikx}, & x < 0\\ ae^{-ikx} + b e^{ikx}, & x > 0,\\ \end{cases} $$ where reflection and transmission amplitudess are $t = 1/a$, $r = b/a$. (It is assumed that the particle is falling on the potential from the right.)

Such a solution can be found by considering the Cauchy problem for the stationary Schroedinger equation with the following initial conditions: $$ \psi(x_0) = 1, $$ $$ \frac{\partial\psi}{\partial x}(x_0) = -ik, $$ where $x_0$ is some large negative coordinate away from the scattering region. It is easy to verify that these conditions indeed correspond to a plane wave $e^{-ikx}$. The coefficients $a$ and $b$ can be extracted from the solution at large positive $x$.

Some drawbacks of this method:

  1. The potential should quickly decay away from $x=0$ (ideally, become zero at $|x|$ larger than some $L$)
  2. If the potential is high enough, $\psi(x)$ can grow very rapidly in the classically forbidden region, so numerical difficulties may arise.
$\endgroup$
7
  • $\begingroup$ And I would assume this tunneling probability should be independent of $x_0$? $\endgroup$ Commented Jun 14 at 0:54
  • $\begingroup$ It seems like the distance from $x_0$ to the barrier (assuming a square for now) has a large impact on the relative amplitudes of the before and after waves. Based on my first simulation results. $\endgroup$ Commented Jun 14 at 1:47
  • $\begingroup$ For a rectangular barrier, it certainly should be independent. There may be dependence, say, for a Gaussian potential, but it vanishes for large $x_0$. But I am also afraid that for the rectangular potential there may be numerical difficulties due to discontinuity of the potential. $\endgroup$
    – E. Anikin
    Commented Jun 14 at 17:00
  • $\begingroup$ I can numerically solve with a rectangular potential, and a coulomb potential and it seems to work fine. My problem is I'm seeing dependence on the value of $x_0$ as it seems to determine the phase of the incoming wave when it reaches the barrier, which in turn propagates differently through the barrier. Maybe I have set the problem up incorrectly? $\endgroup$ Commented Jun 19 at 21:27
  • $\begingroup$ Here's my setup (in Mathematica) *** V[x_] := Piecewise[{{0, x < 0}, {21, 0 < x <= 4}, {0, x > 4}}]. *** schr = -1/2 D[[Psi][x], x, x] == (Ee - V[x]) [Psi][x]; *** bdrs = {[Psi][xmin] == 1, [Psi]'[xmin] == -I k}; *** psol = NDSolve[{schr, bdrs}, [Psi][x], {x, xmin, xmax}, MaxStepSize -> 0.0001]; $\endgroup$ Commented Jun 19 at 21:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.