Complexity Theory for Quantum Promise Problems

Quantum computing introduces many problems rooted in physics, asking to compute information from input quantum states. Determining the complexity of these problems has implications for both computer science and physics. However, as existing complexity theory primarily addresses problems with classical inputs and outputs, it lacks the framework to fully capture the complexity of quantum-input problems. This gap is relevant when studying the relationship between quantum cryptography and complexity theory, especially within Impagliazzo's five worlds framework, as characterizing the security of quantum cryptographic primitives requires complexity classes for problems involving quantum inputs. To bridge this gap, we examine the complexity theory of quantum promise problems, which determine if input quantum states have certain properties. We focus on complexity classes p/mBQP, p/mQ(C)MA, $\mathrm{p/mQSZK_{hv}}$, p/mQIP, and p/mPSPACE, where "p/mC" denotes classes with pure (p) or mixed (m) states corresponding to any classical class C. We establish structural results, including complete problems, search-to-decision reductions, and relationships between classes. Notably, our findings reveal differences from classical counterparts, such as p/mQIP $\neq$ p/mPSPACE and $\mathrm{mcoQSZK_{hv}} \neq \mathrm{mQSZK_{hv}}$. As an application, we apply this framework to cryptography, showing that breaking one-way state generators, pseudorandom states, and EFI is bounded by mQCMA or $\mathrm{mQSZK_{hv}}$. We also show that the average-case hardness of $\mathrm{pQCZK_{hv}}$ implies the existence of EFI. These results provide new insights into Impagliazzo's worlds, establishing a connection between quantum cryptography and quantum promise complexity theory. We also extend our findings to quantum property testing and unitary synthesis, highlighting further applications of this new framework.