Double Slit Experiment: How do scientists ensure that there's only one photon? Many documentaries regarding the double slit experiment state that they only send a single photon through the slit. How is that achieved and can it really be ensured that it is a single photon?
 A: I'd like to add to gregsan's answer about using quantum dots by also raising diamond nanowire one photon sources or similar devices made on CVD-grown diamond waveguides in contemporary or future experiments - of course this this method was not used in the historical one-photon experiments!. See:
Mark P. Hiscocks, Kumaravelu Ganesan, Brant C. Gibson, Shane T. Huntington,
François Ladouceur, and Steven Prawer,"Diamond waveguides fabricated by reactive ion etching", Optics Express, 16, Issue 24, pp. 19512-19519 (2008) http://dx.doi.org/10.1364/OE.16.019512
This one can be freely downloaded, and I should declare my close contact with several of the authors, although I was not part of this work.
Thomas M. Babinec, Birgit J. M. Hausmann, Mughees Khan, Yinan Zhang, Jeronimo R. Maze, Philip R. Hemmer & Marko Lončar, "A diamond nanowire single-photon source", Nature Nanotechnology 5, 195 - 199 (2010)
This one is paywalled, but independent from me.
These devices might find experimental uses in more sophisticated one photon experiments both now and in the future because one can trigger them to emit lone photons almost on demand, whereas dimmed light sources simply emit photons randomly following a Poisson process and cannot be triggered. I have heard several times the opinion of several very bright (much brighter than I am) experimentalists that "no-one would bother", but I just can't shake the feeling that triggering might be useful in yet-to-be thought of experiments - I'd call myself a lousy experimentalist but once upon a time I could drive an analogue oscilloscope pretty well and triggering is surely what makes a useful piece of kit ten times more useful!
Here's how the diamond waveguide one-photon source works. One lays down a diamond waveguide by chemical vapour deposition and shapes the surrounding environment by reactive ion etching. Then a highly controlled amount of nitrogen is included into the diamond lattice. A nitrogen atom normally only makes three covalent bonds with its neighbours, whereas carbon normally makes four, so one is left with carbon atoms with one covalent bond forming electron "dangling" in the lattice wherever there is an included nitrogen. This "dangling electron" is then a fluorophore. One controls the concentration of nitrogen carefully and dices the waveguides so that there is exactly one fluorophore centre in each waveguide device: you can do this by building many devices at once then testing each and throwing out any with none or more than one fluorophores. The waveguide device can then be coupled readily to a single mode optical fibre which links the device to the outside world. The extremely high diamond to air refractive index difference (diamond $n=2.4$) and the ability to build waveguides down to below wavelength dimensions means that we can make "nanowires", which are single moded but also their strong waveguiding properties means that there is near unity probability for the photon fluoresced from the dangling electron into the single mode waveguide. So now we have something which is very readily connected to other experimental kit and doesn't need alignment.
When the time comes to use the device, a trigger signal causes an optical pump laser in the device to pulse the device with an intense beam of light, so that the lone fluorophore in the device is almost certainly raised to its metastable state. At the same time, a very high performance optical switch gates the device's optical output so no light at all escapes to the outside world. After the pumping pulse has ebbed, the output optical gate opens and the fluorophore relaxes a short time later - the fluorescence lifetime is a few nanoseconds. Although this fluorescence is governed by an exponential time-till-emission probability distribution so that, strictly speaking, it is stochastic just like the traditional dimmed light sources, the gating process means that one can control the time of photon emission to within a few nanoseconds. 
Moreover, the near unity photon coupling probability into the single mode waveguide system means that there is an extremely low coupling probability for outside photons incoherent with the waveguide mode to couple in. This system is therefore extremely immune to stray light even though one photon experiments are being done: the exposure of the waveguide system to normal room light levels means that almost no photons get in - and that's before one blackens the system or puts it in a box. There is zero contamination in this device in the hands of even the most ham fisted experimentalist. 
The original and main motivation for these devices is not one-photon double slit experiments, but rather as sources for quantum cryptography protocols such as the Bennett and Brassard (BB84) quantum key distribution protocol.
A: Quantum dots. nanoscale semiconductor materials that can confine photons in 3 dimensions and release them a measurable time after. Based on material used the decay time is known empirically. frequency is also known. the latter is sufficient to calculate the energy of one photon. The former is then sufficient to calculate the rate of photon re emission from the QD. If the peaks at the detector are further apart than the decay time and each peak is measurable to one photon's worth of energy then you know you have a beam of single photons.
A: In the double slit experiment, if you decrease the amplitude of the output light gradually, you will see a transition from continuous bright and dark fringe on the screen to a single dots at a time. If you can measure the dots very accurately, you always see there is one and only one dots there. It is the proof of the existence of the smallest unit of each measurement which is called single photon: You either get a single bright dot, or not.
So, probably you may ask why it is not a single photon composite of two "sub-photon", each of them passing through the slit separately and then interference with "itself" at the screen so that we only get one dot. However, the same thing occurs for three slits, four slits, etc... but the final results is still a single dot. It means that the photon must be able to split into infinitely many "sub-photon". If you get to this point, then congratulation, you basically discover the path-integral formalism of quantum mechanics.
A: The practical answer (which I also wrote in a comment on the linked question) is that you turn the intensity of the light source down until the expectation value for the number of photons on the optical path is low enough to suit you.
If $\bar{n} = 0.1$ then very few of the events that are recorded on the screen will come from events where more than one photon was present on the optical path and the data will be dominated by single photon event.
Not good enough for you? Turn it down until $\bar{n} = 0.01$. Or $0.001$ or whatever suits you.
At some point the exercise becomes silly.
