Presets
Playback
Photons
Delay between the photons, their bandwidth and their polarisation.
Splitter and counting
Splitting ratio, automatic delay scan and the pair rate.
Layers
Model checks
The science
Two photons, one beam splitter
Each photon alone leaves a beam splitter by port 1 or port 2 with probabilities set by the reflectivity R and transmission T = 1 - R. For two photons arriving one at each input there are two ways to get one photon in each output (both transmitted or both reflected), and for a symmetric splitter their amplitudes carry opposite signs. If the photons are completely indistinguishable the two amplitudes cancel and the pair always leaves together: bunching.
The formula
With Gaussian wave packets of spectral width sigma (standard deviation of the intensity spectrum) the coincidence probability is P(tau) = R^2 + T^2 - 2 R T cos^2(phi) exp(-sigma^2 tau^2), where tau is the relative delay and phi the angle between the polarisations. At R = T = 1/2 with equal polarisations the dip goes to zero. The exchange term is the overlap integral psi*(w1, w2) psi(w2, w1), which the page evaluates as a double integral over frequency as an independent check of the closed form.
Why the dip has this width
The photons are distinguishable by arrival time when tau exceeds their coherence time. The dip has full width at half minimum 2 sqrt(ln 2) / sigma; a broader spectrum gives a shorter wave packet and a narrower dip, which is why the HOM interferometer can measure photon timing at femtosecond precision. The packet length drawn on the bench is this coherence length.
Counting
Each pair is a random event with probability P(tau) of a coincidence; the yellow dots are the measured fractions per delay bin, and the sampling check compares counts with probabilities in standard errors.
Simplifications
Perfect single-photon sources, no losses or dark counts, a lossless symmetric-phase beam splitter, Gaussian spectra, a single polarisation angle and no frequency entanglement. The bench is schematic: one scene unit of path stands for 100 fs.
What to try
- Press the perfect-overlap preset: the predicted coincidence probability drops to zero, and the two photons go to the same detector.
- Slide the delay away from zero: the packets separate on the bench and the coincidence probability climbs back to the classical 50 %.
- Choose the broad-bandwidth preset: the packets are shorter and the dip narrower.
- Rotate the polarisation to 60 degrees: the photons become partly distinguishable and the dip no longer reaches zero.
- Use the unbalanced splitter: even perfect photons cannot cancel the coincidences completely because the two amplitudes differ in size.