# Is the interference stable?

In a Mach Zehnder interferometer if we use one at a time photons and tune it so the first several to arrive in channel D1 after BS2 will all next arrive also in D1? Are there reasons that some will not land on D1 f.e. if BS2 is wide and the probability waves do not overlap at BS2 in some cases. If we turn it to Feynman trajectories maybe trajectories in BS2 will not intersect in some cases and photon can land in D2 for that reason? We assume wavelength is fixed by laser and filters!

• It's all about the path lengths of either arm ... if both are bad then no detector gets light .... if both are good then 50/50 like a bean splitter. Bad path is path length = n(lambda) + 0.5(lambda), good is path length = n(lambda). Commented Feb 8 at 23:53
• What a connection has the length of a path? I think only path difference counts. If l1-l2 = lambda we have detector D1, if 0.5£ than D2.????? Commented Feb 9 at 19:06
• In MZI you can tune both arms to be bad path. No light goes in the MZI … just like in the DSE where no light goes to dark areas/bands because all paths to dark band averages to lambda + 0.5 lambda. Commented Feb 9 at 20:03
• The important length is between the input and output beam splitter faces. And also there can be a phase change for the beam that goes thru the glass part of beam splitter. Commented Feb 9 at 20:07
• Simple experiment: laser diode to glass plate 1 and then thru glass plate 2 and then ccd. If you adjust path length between the glass to be lambda + 0.5 then no light at detector. Commented Feb 9 at 20:10