Skip to content
Back to people
QUBITWIRE 100

Matthias Troyer

Technical Fellow and Corporate Vice President of Quantum

Microsoft

Microsoft Research
Identification checked 2026-09-19
WHY INCLUDED

Connects quantum many-body physics, computational complexity and machine learning with the design of quantum-computing architectures and applications that could outperform classical approaches.

arXiv / Physical Review LettersarXiv / Science

Troyer’s research helps define both the difficulty of simulating quantum matter and the tools available to approach it. His work on the fermionic sign problem identifies a fundamental obstacle to a generic classical simulation method, while neural-network quantum states offer a different representation for selected many-body systems. At Microsoft he works on quantum architecture and applications. The thread across these activities is concrete computational cost: understanding what makes a problem hard and what an alternative method would need to improve.

Microsoft ResearcharXiv / Physical Review LettersarXiv / Science

Defining contributions

Work, in context
  1. 2005

    Complexity of the fermionic sign problem

    Troyer and Uwe-Jens Wiese showed that the generic fermionic sign problem in quantum Monte Carlo is NP-hard. This explains why a universally efficient cure is unlikely under standard complexity assumptions, while leaving room for useful solutions in special cases.

    Co-author of the complexity result with Uwe-Jens Wiese.

    Source-supported recordarXiv / Physical Review Letters
  2. 2017

    Neural-network representations of quantum states

    With Giuseppe Carleo, Troyer introduced an artificial-neural-network representation of many-body wavefunctions and demonstrated it on interacting spin systems. The work explored a compact computational approach to ground states and quantum dynamics.

    Co-author of the neural-network quantum-state work with Giuseppe Carleo.

    Source-supported recordarXiv / Science

Keep in perspective

The complexity result concerns the generic sign problem; neural-network demonstrations do not solve every many-body system. A current personal work-base country was not established.

Follow the evidence

3 sources

Primary papers, institutional records and attributed announcements. Each source supports the claims linked above.

QubitWire editorial · Content edition 2026-09-19.1Independent coverage. Inclusion does not imply endorsement.