Non-Hermitian Gates: The Price of Imaginary Speedup ✛
Context
Non-Hermitian Hamiltonians describe the conditional evolution of a quantum system, for example, when we monitor the absence of quantum jumps. They arise in lossy optics, superconducting circuits, and other platforms. Alongside this work on the arXiv are studies of quantum algorithms and error correction, where non-Hmiticity is also discussed, but mostly from the perspective of specific protocols. This article stands apart: it asks a fundamental question about the computational complexity of non-Hermitian models in general, not about particular implementations.
Methods
The authors construct an explicit 'postselection gadget' from any irreversible gate surrounded by suitable unitary operations. The key step is the singular value decomposition of the non-Hermitian operator, which isolates the largest and smallest stretch factors. By repeating such a gate a polynomial number of times, one can exponentially suppress the unwanted component of the state, mimicking a projector. For the upper bound, they use purification of the non-Hermitian evolution to a unitary on an extended system, followed by postselection on the measurement qubit.
Results
It is proved that the class NHBQP(U), the problems solvable by a quantum circuit with a fixed irreversible gate and universal unitary gates, is exactly PostBQP = PP. This holds for any gate not proportional to a unitary, and does not depend on its specific form. For families of gates that depend on the input size, the result holds provided the singular gap decays no faster than an inverse polynomial. At the same time, if the purified unitary circuit is strongly simulable (e.g., Clifford or matchgate), then the non-Hermitian dynamics remains classically simulable for postselection probability at least $2^{-poly(n)}$.
Implications
The work shows that non-Hermiticity in quantum computing is disguised postselection. Any claim of non-Hermitian advantage must be accompanied by an accounting of the physical costs of implementing conditional evolution. Unlike neighboring works that study specific quantum algorithms or error correction, here a general complexity-theoretic boundary is given. This is the first work that links an arbitrary irreversible gate to the full power of postselection, not just isolated examples.
Future development
A natural next step is to build postselection gadgets directly from local non-Hermitian Hamiltonians, bypassing singular value decomposition, and to refine resource estimates: the number of gates, success probability, and error accumulation. It is also important to understand which classes of non-Hermitian models admit weak simulation without strong simulation, and how quantum trajectories with a limited number of jumps behave.
Impact
The results directly affect the assessment of the realism of non-Hermitian quantum protocols in metrology, entanglement generation, and quantum sensing. They also provide guidance for developers of quantum simulators with losses.
Next steps
Investigate constructions of postselection gadgets for specific physical systems, such as chains with dissipation, and check whether the required resources can be reduced by exploiting the Hamiltonian structure. Additionally, study the complexity of models with feedback based on measurement outcomes.
Key open problems
In quantum complexity theory, the exact power of intermediate classes between BQP and PostBQP remains open. This work shows that any non-Hermitian extension of BQP immediately jumps to PostBQP, sharpening the problem of classifying 'physically realizable' computational models. It also remains unclear whether the requirement of strong simulation can be relaxed to weak simulation for some non-Hermitian circuits.
The knowledge machine's view
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors construct an explicit postselection gadget from an arbitrary non-Hermitian gate using singular value decomposition, and prove that the class NHBQP(U) coincides with PostBQP = PP. For restricted classes (Clifford, matchgate), non-Hermiticity adds no computational power.
Where its strength is
The work provides a general complexity-theoretic criterion: it links any irreversible gate to the full power of postselection, rather than isolated examples, and clearly delineates when non-Hermiticity is potent and when it is harmless.
Where it could go next
- It might be interesting to compare the obtained bound with the behavior of non-Hermitian systems in physical implementations where conditional evolution arises naturally. Nearby is work on a quantum motor in atomic collisions, where non-Hermiticity is used to control heat; one can see how constraints on success probability manifest in a specific platform.based on: A Quantum Motor Powered by Atomic Collisions: Steering Heat Through Microscopic Coupling arXiv:2608.12055v1
- A possible continuation is the analysis of non-Hermitian circuits with noise or limited gate fidelity. Nearby is work on quantum droplets where precision proves unattainable; its approach to accounting for realistic limitations may suggest how to modify the NHBQP model in the presence of errors.based on: Quantum Droplets: Precision Unattainable arXiv:2605.19706
- Another move is to check whether the PostBQP equivalence holds for non-Hermitian Hamiltonians with continuous time, not just discrete gates. In the archive there is work on quantum echo and the hierarchy of time correlations that considers non-Hermitian evolutions; perhaps their methods can help transfer the result to Hamiltonian dynamics.based on: Death and Rebirth of Quantum Echo: A Hierarchy of Temporal Correlations in a Qubit arXiv:2505.01379v1
Worth trying
- Test the postselection gadget construction on a specific non-Hermitian gate from an optical or superconducting implementation to estimate the required resources.
- Investigate how the singular gap behaves for typical dissipative gates and how realistic the condition of its polynomial decay is.
- Consider intermediate classes of circuits between universal and strongly simulable ones to find a finer boundary of computational power.
Why this may matter
This work sets a framework for discussing non-Hermitian quantum computing: it shows that any advantage must be weighed against the cost of postselection, thereby helping to separate genuine physical effects from formal speedups.
Nearby works in our archive
- Quantum Droplets: Precision Unattainablesimilarity 0.55
- Universe as a Precise Mechanism: No 'Boltzmann Brains'similarity 0.50
- Quantum Chip Outperforms Supercomputers: Two Hours vs. Yearssimilarity 0.50
- Universe Without a Beginning: Quantum Cosmology Confirms Hawking's Ideasimilarity 0.50
- Quantum Computer Simulates Turbulence for the First Timesimilarity 0.50
- Bacteria Teach Quantum Batteries to Store Lightsimilarity 0.50
- Death and Rebirth of Quantum Echo: A Hierarchy of Temporal Correlations in a Qubitsimilarity 0.50
- A Quantum Motor Powered by Atomic Collisions: Steering Heat Through Microscopic Couplingsimilarity 0.50
The archive is sparse around this work: the closest review is only 0.55, against a usual 0.64. It looks like few have drilled nearby.
A single-qubit non-Hermitian Hamiltonian H = iσ_z suppresses the |1⟩ component as $e^{-2t}$ over time t. If t grows linearly with the number of qubits, the suppression becomes exponential, which is the simplest postselection gadget.
The idea of postselection as a computational resource is explored in Ted Chiang's story 'Story of Your Life': the protagonist sees only those branches of reality where the required condition is met. Non-Hermitian dynamics formalizes exactly such branch selection.
Key numbers
- complexity class: PostBQP = PP
- suppression of unwanted component: e^{-2t}
- postselection probability threshold: 2^{-poly(n)}
- gadget error: 2^{-poly(n)}