The "collapse" postulate says that if part of the wave-function produces a click in a detector, the rest of the wave-function disappears. In the experiment described here, it is shown that no part of the wave-function disappears, namely, given a superposition of two wave-packets, while one wave-packet produces a click in a detector, the other wave-packet produces observable interference effects.
A quantum system is prepared in a state with maximum one particle, a photon, A;
(1) |ψ> = q{ |0>A + p( |1;a>A |0;b>A + eiθ |0;a>A |1;b>A ) },
see figure.
It is shown below that while the wave-packet |1;a>A
The wave-packet |1;b>A illuminates one side of the 50-50%beam-splitter BS, and on the other side lands a coherent beam
(2) |α> = N( |0>B + peiα|1>B + . . . ).
where N is the normalization factor. Thus, we have the total wave-function
(3) Φ = |α>|ψ> = Nq( |0>B + peiα|1>B + . . . ){ |0>A + p( |1;a>A |0;b>A + eiθ|0;a>A |1;b>A ) }.
At the beam-splitter the following transformations take place
(4) |1>B → (1/√2) ( |1;c> |0;d> + i|0;c> |1;d>);
(5) |1;b>A → (1/√2) (i|1;c> |0;d> + |0;c> |1;d>).
Introducing them in (3) one gets the following IMPLICATIONS:
(6) For θ = α - π/2, every click in the detector D is preceded by a the detection of the wave-packet |1>a in the detector U.
(7) For θ = α + π/2, every click in the detector C is preceded by a the detection of the wave-packet |1>a in the detector U.
Thus, one can see that by changing the phase θ, carried by the wave-packet |1;b>A , one can switch between a joint click in D and U, and a joint click in C and U.
CONTRADICTION: one can see in the figure that BS is more distant from the preparation region than the detector U. So, if the collapse hypothesis were correct, the tunning of θ would have no effect, since when the detector U clicks, the wave-packet |1;b>A would disappear instead of reaching the beam-splitter BS.
CONCLUSION: No part of the wave-function disappears - no collapse.