Almost all old LEDs will be photosensitive and will operate as photodiodes even at near IR and visible wavelenghs.
After all an IR LED and IR photodiode both have to have around the same bandgap and some kind of transparent area over the junction. Not much has to be different in the gross sense.
Direct vs. indirect bandgap.
Silicon has an indirect bandgap, and as a diode, emits very little light indeed. And that light would be at or below the bandgap (>1400nm?), where other Si detectors have little or no sensitivity.
GaAs is a direct bandgap material, so it makes a good LED. The bandgap is somewhat higher than Si, so the emitted light (~1000nm) is well received by Si diodes.
Many old photodiodes will, in fact, emit light as well if you drive a current through them. Sometimes by a LED like mechanism and sometimes as an incandescent thermal source, depending on how you drive them.
You can light up a 2N3055 by reverse-biasing B-E. At 1A (and about 8V), it's just barely visible if you cup your hands over it (ah, and cut off the metal can first...
). It gets hot pretty fast at that power level, so don't leave it too long. (And probably toasts the junction fairly well... but it's a 2N3055, so good riddance anyway.
)
I don't know that the mechanism of avalanche light emission is well known or understood... I haven't heard anything about it.
...Nevermind, it's been known for a long time:
http://cdn.intechopen.com/pdfs-wm/47465.pdf...which explains the otherwise ridiculous sounding proposals for putting phosphors in/on silicon. Because, what's going to excite them? Well, hot carriers, of course, from an avalanche (or near-avalanche) condition.
BTW, hot carriers are why JFET gates leak more at high Vds, and why EPROM and Flash are possible (the floating gate is charged by dumping high Vds through the transistor). Physics!
So if you have something that is in a very generic package that could be either a LED or a photodiode and it acts as a weak near IR LED at least a little and it acts as a photodiode to near IR or visible at least a little, I'm not sure how you'd tell the difference between them.
I would think GaAs photodiodes are very rare indeed, even throughout history. (On a quick search, it looks like they're used for Gbit fiber. But probably more as modules rather than components, except for specialized applications?)
I wonder what the best "LEDs as photodiodes" are today for common-ish devices.
Probably, for most of them, the gain is very poor...
Red or IR should be best, as the sensitivity will include the visible spectrum, assuming that's what you want to look at. (Whereas green LEDs shouldn't respond to red light, and so on. Minding that GaP is a weird semiconductor, and probably does respond to red light anyway...)
An important difference between photodiodes and LEDs: current density. Photodiodes (and solar cells) can afford very thin connections, maximizing absorption. LEDs can't, because they have to operate at high currents.
Tim