The question of how a Mach–Zehnder interferometer works arose out of reading through Alevel Research and Analysis essays. Becauseour students only study division-of-wavefrontinterferometers and not division of amplitude theyhad simply accepted the standard explanationgiven on websites such as www.qubit.org. On thatsite, the basic interferometer is shown as in figure 1and it is explained that interference between thetwo paths ensures that the photon always strikesdetector A. If one of the paths is lengthened thenthe interference can be altered to ensure that allphotons strike detector B.Whilst this seems eminently plausible (indeed,so plausible that we all accepted it without toomuch worry) it is grossly misleading. First of all,consider the phase of the photon on following eachof the two paths, the lower and upper. Initiallywe shall assume that there is no phase shift onreflection or transmission.The phase on reaching the second beamsplitter is simply the path length divided by thewavelength, multiplied by 2π. On recombinationat the beamsplitter, if the two paths are ofequal length, then the phases are equal. Sowhich path shows constructive interference, thepath towards A or B? The answer is unresolved.In fact, the entire situation is symmetrical withrespect to the two detectors and should onepath allow constructive interference, so will theother. Similarly if one path suffers destructiveinterference, so does the other. This violatesconservation of energy.