3 ms·
Following on my comment, back around 2004 or so when I was doing my PhD in physics, I made a Java applet showing how resonant tunnelling works (ie, the phenomen
by jeffwass 10y ago
Following on my comment, back around 2004 or so when I was doing my PhD in physics, I made a Java applet showing how resonant tunnelling works (ie, the phenomenon that drives the negative differential resistance in the RTD above), and also how band structure arises in an ordered crystal.
I put these on GitHub, if anyone wants to play with it, this applet is from the top screenshot :
https://github.com/jeffwass/Physics_Simulations/blob/master/README.md https://github.com/jeffwass/Physics_Simulations/blob/master/...
Basically, in one dimension, we represented an atom as an infinitely-thin 'spike' in potential, a Dirac Delta function.
For a single 'atom', regardless of how much energy an incoming electron wave function has, some energy is always reflected back and a smaller amount goes through. (Here, energy and frequency are basically synonymous, where E = h * nu, h is Planck constant and nu is the frequency.)
For two atoms (a double barrier), certain input frequencies cause destructive interference of the reflected quantum wave function at the first atom (ie - the reflections bouncing off the first and second atom cancel out), and exactly 100% of the incoming energy transfers through! At certain other frequencies, the output waves cancel and the wave is entirely reflected. The plot of transmission vs input frequency looks sinusoidal, between 0% and 100%. The applet shows the impact of adjusting the spacing and relative 'height' of the barrier atoms.
This interference is the quantum mechanical basis of resonant tunnelling.
In the applet, you can control how many atoms are in the crystal. As you add more atoms to the system, the output transfer function starts forming notable regions of high transmission and regions of low transmission.
This is for a perfectly ordered crystal (same potential and spacing). The applet also shows what happens if you randomise the spacing or the potential heights.
Question for the HN community : as Java web applets are seemingly obsolete these days, any recommendations on current front-end tech for making these types of simulations?