Paradox of a nonlinear beam splitter and selection of photon fluctuations

It has been found that a nonlinear beam splitter is an interesting object for investigation for the following reasons. First, the classical and quantum descriptions of this object give opposite results for the behavior of phase fluctuations: according to the classical theory, the phase remains unchanged, whereas the quantum theory predicts that phase fluctuations increase or decrease in correlation with a decrease or increase in amplitude fluctuations. Second, quantum fluctuations of the input mode can be separated into amplitude and phase fluctuations, so that phase fluctuations are directed predominantly into one output mode, say the reflected mode, whereas amplitude fluctuations are directed predominantly into the other (transmitted) mode.