Diode selection for AM radio

It's just the basic operating principle of a diode. An ideal diode would allow current to flow in one direction but block current flow in the other direction. This is based on how it's made, with a p-type region, an n-type region, and a depletion zone in between. Like the bottom diode in this picture:

When you apply a some voltage, in your case 0.9V then the p-type holes and the n-type electrons move into the depletion region because they are repelled by their respective battery terminal. With enough voltage (0.9V in your case) the free electrons in the depletion region get moving and current begins to flow like this:

Now in the ideal case if you were to reverse that battery the opposite will happen and you'll get no current flowing:

In the real world though you can only apply so much reverse voltage or push before you hit the breakdown voltage and current begins to flow freely in the reverse direction. Zener diodes take advantage of this fact and are constructed to breakdown at lower voltages such as your 3.3V.

Sources:
You can read more about how zeners are made here Or see the article I got all the pictures from here

There are many plans on the XTAL SET SOCIETY and Ben Tounge's websites, including both biasing of diodes for optimised detection and modeling of crystal radio circuits in SPICE.

Consider also using a Schottky diode (some of the HP types are fantastic for this and 'foxhole' razor blade diodes are essentially the same!) and zero-Vgs MOSFETs as reported about in Bob Cutler's 'High Sensitivity Crystal Radio Set' which is available online at the ARRL.

Of course, if you can get your Q high and tuning not so touchy and your local clear-channel is good enough, maybe that silicon diode is OK.