The human eye detects light via a family of proteins called opsins. Different forms of photopsins are sensitive to different wavebands, which is what gives us color vision. Rhodopsin is sensitive mainly to greenish-blue light, and provides us with monochromatic night vision. Rhodopsin works by changing its conformation when it absorbs a photon; that change of conformation allows ions to flow through the rod cell's membrane and generate a signal. The signal from each rod cell is processed through the retina and passed to the visual cortex, where a representation of the visual field is constructed.
Human rhodopsin has a quantum efficiency (QE) of about 25% (there's a 25% chance a single photon will be absorbed and produce the rod-cell signal). By comparison, cat rhodopsin is more than 90% quantum efficient. 25% QE is sufficiently high to be observable -- a source of single photons can be seen by a dark-adpated person with normal vision.
Step 1: Not Ready for Prime Time
This is a lab we did when I was an undergraduate, more than 20 years ago. I haven't done the setup myself since then, so I'm just going to describe it; if I have the opportunity run it again, then I can take pictures and publish this as an I'ble.
If someone else decides to tackle it, please feel free to contact me and I'll make this a collaboration.
Step 2: Producing Single Photons
NOTE: I have modified the ND notation below to refer explicitly to the optical density, rather than fractional lens area. I was unware of the confusion until Instructables user jefc_uk pointed it out. Thanks!
You'll need a steady source of well-collimated photons. A green laser pointer (~532 nm) will do nicely. But how many photons does it generate? A wavelength of 532 nm corresponds to 3.53×10-19 joules. So a small 1mW laser pointer puts out 2.8×1015 photons per second (watt = joule/s). You can use a red laser pointer, but your eye will be less efficient (see intro). Estimate the number of photons your pointer produces given its wavelength and power rating.
How do you reduce that to one photon at a time? With filters. An ND3.0 neutral density filter (optical density of 3.0) reduces the output light by 10-3 compared to the input, so a stack of just five ND3.0's in front of this laster pointer would result in (on average) just 2.8 photons per second! A stack of four ND4.0's (each reducing the output by 10-4)would give you 0.28 photons/s on average.
If you don't have neutral density filters, you can make a decent approximation, by stacking sheets of black trash-bag plastic. To make this work you have to measure the attenuation yourself, so you'll need a photodetector, something which gives an output (voltage, resistance, current, whatever) proportional to the intensity of light.
Shine the laser on your detector with no filters in place, and record the output. Do the same with one, two, etc. filter layers, and make a plot of output (on a log scale) vs. number of filters. Hopefully, you get down to 0.001 or 0.0001 with just a few layers. with the log plot, you can draw a straight line to extrapolate how many layers you need to get down to a few photons/s.
Ideally, you'd also like a single-photon counter, something like an avalanche photodiode, connected to a piezo-speaker, so that you can hear "clicks" each time a photon comes through the final stack of filters, and confirm that the rate is as low as you expect. Building such a thing is a whole separate project in itself, so I'm just going to assume that you have one.
Step 3: Make a Dark Room -- or a Dark Box
Once you have your single-photon source, you need to set it up in a completely dark room. If you have access to an old-style photographic darkroom, use it.
Otherwise, use thick (3-5 mm) black felt and gaffer's tape to seal any windows and doorframes. When you turn off the lights, you should not see any light coming in through cracks or edges. If you do, fix them and check again.
I'm coming to the conclusion that this is too complicated for the problem at hand. It may turn out to be better to build a light-box with a gasketed hole at one end for the laser pointer, and a viewing window with draped headcover at the other.
Step 4: Set Up Your Photon Gun
Attach the ND filters (or plastic sheets) to the front of your laser, sealing around the edges with gaffer's tape. You don't want any light scattering out the sides.
Put the laser on a table or stand pointed at where you'll be sitting.
Even a 1 mW laser pointer can damage retinal cells from direct impact. Not enough to blind you, but enough to potentially contribute to vision issues later in life. Don't point the laser at your face until you have the ND filters installed.
If you're doing this by yourself, you may want to have a piece of tape set up to hold the pointer's button down. Otherwise, your lab partner will take care of it.
Step 5: Sit in the Dark
The human eye requires 20 to 30 minutes to fully dark adapt. Turn off all the lights in your room and wait. This will seem like forever, so you may want something to help you keep track of the time. A standard CD will be about half finished, or you can get through ten pop sons on your iPod, when your eyes become dark adapted.
Step 6: Fire Away!
Turn on the laser. You'll see intermittent flashes all coming from nearly the same place in your visual field; if you turn your head, the location will move in the opposite direction. If you've used filters to get down to a few flashes per second, POV will make them easier to see. At less than one photon per second, you'll see them individually.
If you have a lab partner, you can make a real experiment out of it. Sit quietly, and have them turn the laser on and off without telling you. If things are working correctly, then you should be able to identify when the laser has been turned on, and when it's off. Keep track of hits and misses for many (at least 20) trials, and measure whether or not what I've described actually works.
Step 7: Other Single-photon Experiments
Now that you have a reliable source of (on average) single photons, you can use it to explore some of the other weirdness in quantum mechanics.
How about "self-interference" and wave-particle duality with a double-slit experiment?
If you are really ambitious, and have access to a proper optical table and research equipment, you could even try measuring entanglement, or trying a delayed-choice experiment.
Of course, if you have access to that kind of equipment, you're probably already a postdoc or facult in quantum optics, and are about to write a long comment about all the stuff I've gotten wrong :-)