Unambiguous Turing machine

A non-deterministic Turing machine is represented formally by a 6-tuple, ${\displaystyle M=(Q,\Sigma ,\iota ,\sqcup ,A,\delta )}$, as explained in the page non-deterministic Turing machine. An unambiguous Turing machine is a non-deterministic Turing machine such that, for all input w = a₁a₂ ... a_n, there exists at most one sequence of configurations c₀,c₁, ..., c_m with the following conditions:

1. c₀ is the initial configuration with input w
2. c_i+1 is a successor of c_i and
3. c_m is an accepting configuration.

In other words, if w is accepted by M, there is exactly one accepting computation.

A (deterministic) Turing machine is an unambiguous Turing machine. Indeed, for each input, there is exactly one computation possible.

On the one hand, unambiguous Turing machine have the same expressivity as a Turing machine. Indeed, they are a subset of non-deterministic Turing machines, which have the same expressivity as Turing machines.

On the other hand, unambiguous non-deterministic polynomial time is suspected to be strictly less expressive than non-deterministic polynomial time.

Lane A. Hemaspaandra and Jorg Rothe, Unambiguous Computation: Boolean Hierarchies and Sparse Turing-Complete Sets, SIAM J. Comput., 26(3), 634–653