First-hand information for everyone
This paper introduces HNCC, a new algorithm to reduce error in Hamiltonian simulation without using extra qubits. The problem they address is that practical “product formula” or Trotter methods for simulating quantum dynamics have simple circuits and error bounds that exploit nested commutators. But their circuit size grows as a polynomial when you ask for higher precision. HNCC aims to keep the advantages of product formulas while making the circuit size grow only polylogarithmically with the required precision.
The authors build HNCC (high-order nested-commutator compensation) by writing the short-time Trotter error in terms of nested commutators using a truncated Baker–Campbell–Hausdorff (BCH) expansion. They then approximate the corrective operation as a linear combination of quantum channels (LCQC). Each channel in that combination is implemented by sequences of Pauli-rotation channels and sampled at random. Because the correction is applied at the channel, or “superoperator,” level, their scheme avoids Hadamard-test measurements and needs no ancillary qubits. The randomized implementation recovers the ideal evolution in expectation, and the classical bookkeeping only records sampled signs.
What this delivers in complexity terms is a notable trade. For a K-th order product formula applied to a k-local Hamiltonian on N qubits with Γ Pauli terms and local interaction strength g0, HNCC estimates an observable to additive precision ε with the usual O(ε−2) number of circuit repetitions. The maximum gate count per circuit scales as O(N^{2/(2K+1)} (k g0 t log(1/ε))^{1+1/(2K+1)} k(Γ+log(1/ε))). In plain language: the method preserves the nested-commutator (system-size) scaling of product formulas, achieves a time dependence like a product formula of effective order 2K+1, and replaces the previously polynomial dependence on 1/ε in circuit size by a polylogarithmic one.
The paper gives concrete resource estimates for a periodic Heisenberg chain. When applied to the common second-order product formula (K=2), HNCC produced the lowest estimated gate counts among the product-formula-based methods they considered. The authors report up to a 17.5× reduction in CNOT gates per circuit and a 19.8× reduction in T gates per circuit compared with the uncompensated second-order Trotter step.
There are important caveats and limits reported by the authors. The method relies on truncating the BCH series and using bounds based on doubly right-nested commutators; the full BCH series does not generally converge for many-body systems, so an exact series convergence is not assumed. That truncation introduces a factor of log(1/ε) in the gate-count bound, and the authors note it is an open question whether that factor can be reduced with sharper truncation bounds. Also, while the per-circuit gate cost becomes polylogarithmic in 1/ε, the sampling repetition cost remains the standard O(ε−2). The paper also describes an “unpaired” variant with different trade-offs and a small biased error if the number of compensation gates per circuit is artificially limited.
Overall, HNCC offers a practical, ancilla-free path to reduce Trotter errors while keeping the favourable nested-commutator scaling of product formulas. The approach is particularly relevant when one wants low-depth, ancilla-free circuits and better precision scaling in the maximum gate count per run. Further work is needed to tighten truncation bounds and to test the method across more Hamiltonians and hardware connectivity constraints.