Credit: AMBLIN/UNIVERSAL/THE KOBAL COLLECTION

It is a tricky problem that Marty McFly has to solve in the first film of the Back to the Future trilogy. Sent back 30 years, on board a time machine built by the slightly mad scientist Doc Brown (pictured), he accidentally interrupts the first meeting of his future parents, then has to pull all kinds of strings to make their love affair happen — and thereby secure his own existence.

The possibility of time travel inspires physicists too. In the dawning age of quantum computation, one question in particular is keeping them busy: could closed timelike curves (CTCs) — where the path of a particle through spacetime returns to its starting point — help a computer to solve (computationally) hard problems more efficiently? In recent work it has been suggested that, at least in certain circumstances, this should be the case. But Charles Bennett and colleagues now argue that, in practically relevant settings, it isn't (Phys. Rev. Lett. in the press; preprint at http://arxiv.org/abs/0908.3023v1; 2009).

The predicted advantages of CTC-assisted quantum-state evolution include more than computational benefits. It has been conjectured that CTCs also help to perfectly distinguish quantum states, even if these are non-orthogonal — something that is impossible in standard quantum mechanics. But Bennett et al. claim there was a problem with these earlier works: the conclusions drawn were based on considering only fixed pure input states, but the results thus obtained do not hold for the general case in which the input consists of a distribution of several states.

This is so because, unlike in standard quantum mechanics, in models including nonlinearities such as CTCs, the evolution of a mixture of states is not equal to the mixture of the evolutions of the individual states. These findings suggest that CTCs not only fail to help computations (or quantum state discrimination), but also affect conclusions reached about other nonlinear extensions of quantum mechanics. However, quantum mechanics doesn't have to be, by necessity, linear — Bennett et al. expect that, after all, a 'well behaved' nonlinear theory may be possible.

As far as we know, CTCs exist only in fiction. But still, thinking about a world with CTCs has already led to consistent physical models that, for example, avoid the 'grandfather paradox' (the consequences of which Marty McFly tried so hard to escape otherwise — successfully, of course). And for the future, the study of CTC-assisted computation should lead to a fuller understanding of the foundations of quantum information. Or perhaps it already has?