Title:On the Implications of the Gottesman-Knill Theorem for our Understanding of the Resources Involved in Quantum Speedup

Abstract:According to the Gottesman-Knill theorem, any quantum algorithm or protocol which exclusively utilises the elements of a certain restricted set of quantum operations can be efficiently simulated by classical means. Since some of the algorithms and protocols falling into this category involve entangled states, it is usually concluded that entanglement cannot be sufficient to enable quantum speedup.
In this short note I argue that this conclusion is misleading. As I explain, the quantum operations to which the Gottesman-Knill theorem applies are precisely those which will never yield a violation of the Bell inequalities; i.e., they all involve rotations that are multiples of \pi/2. It is well known, however, that the correlations present in entangled quantum systems whose subsystems never take on orientations with respect to one another that are not multiples of \pi/2 are reproducible by a classical hidden variables theory. Thus it should be no surprise that entangled quantum states which only undergo operations in the Gottesman-Knill group are efficiently simulable by a classical computer.
What the Gottesman-Knill theorem shows us, therefore, is that one must \emph{use} an entangled quantum state to its full potential in order to achieve a quantum speedup. If one only utilises the portion of the system's state space efficiently accessible by a classical system, no speedup will be achieved, even when the system is entangled.
Nevertheless, there is a meaningful sense in which an entangled quantum state is sufficient for quantum speedup: an entangled quantum state \emph{provides sufficient physical resources to enable} quantum speedup, whether or not one elects to use these resources to their full potential.