Abstract

We introduce the concept of quantum processing focussing on optical implementation. Bits are encoded interferometrically using the relative phase or polarisation of single photons. As we are dealing with quantum objects we sec that the quantum bit (qubit)is extended from the clasical bit as it can be prepared in a superposition state of two orthogonal properties (polarisations for example). Random number generation is also quite a natural for a quantum processor, optics requires a simple 50/50 beamsplitter with outputs labelled as ‘0’ or ‘1’. In quantum cryptography where we would like to securely generate a random number at two remote locations, a simple quantum processing protocol has emerged [1]. Binary information is coded randomly at the single quantum level into four possible states: two orthogonal states such as the zero and 90° polarisation (representing ‘0’ and ‘1’ respectively)and equal superpositions of zero and 90° (which is in fact 45° ‘0’and 135° ‘1’). The receiver uses a polariser which can separate orthogonal polarisations into ‘0’ and ‘1’ channels. The base angle corresponding to ‘0’ is randomly switched between 0° and 45° and thus discriminates in an error free way only when matched to the senders coding basis. A classical channel is used to determine (and erase) the bits lost due to losses and those received using the wrong measurement basis thus allowing identical random numbers to be generated. Security is guaranteed by the fact that single photons encode each bit and by randomly selecting from the two possible non-orthogonal measurement bases. It is impossible for an eavesdropper to predict which base is used or to measure in both bases on the same photon as this would violate Heisenbergs uncertainty principle. Thus any eavesdropping (with reinjection of 'copies’) will necessarily introduce errors in the final key which can easily be checked for using a subset of keybits that are then discarded. We have developed [2] similar schemes where time division interferometry is used in a similar way. This allows use over fibre optic links (up to 30km in experiments) which do not necessarily conserve polarisation. More recently the polarisation scheme has been shown to be possible over 20km in fibre if active control of slow polarisation drift is used [3]. We expect eventual fibre based ranges out to 100km while recent proposals suggest free space systems with ranges extending to satellites in low earth orbit.

© 1996 IEEE