Digital adiabatic evolution is universally accurate

Yihua Lu, Yifei Huang, Dong Sung An, Qi Zhao, Dingshun Lv, Xiao YuanarXiv 2025 · DOI · arXiv

TL;DR

Digital adiabatic evolution—a cornerstone of quantum state preparation—is far more efficient than previously believed. This paper proves that algorithmic errors in these simulations do not accumulate over time but instead exhibit a "self-canceling" behavior, allowing for high-accuracy results with significantly shallower quantum circuits.

What Problem Does This Paper Address?

Adiabatic evolution is a method where a quantum system is slowly changed from a simple starting state to a complex target state (like the ground state of a molecule). To do this on a digital quantum computer, we have to "discretize" the continuous process into many small time steps.

The conventional wisdom in quantum computing is that these discretization errors (often called Trotter errors) pile up over time. Because adiabatic processes require very long evolution times ($T$) to remain accurate, researchers assumed that digital simulations would require impossibly deep circuits to keep the total error under control. This paper challenges that "impracticality" assumption by showing that the physics of adiabatic evolution naturally suppresses these errors.

Methods

The authors provide a rigorous mathematical framework to analyze the infidelity of digital adiabatic processes. They focus on two primary Hamiltonian simulation techniques:

  1. Trotterization: The standard method of breaking an evolution into a sequence of simple gates.
  2. Generalized Quantum Signal Processing (GQSP): A more advanced, near-optimal method for simulating quantum dynamics.

The core of their approach is a novel error analysis using oscillatory integrals. They demonstrate that the transition amplitudes (the "leaks" into unwanted states caused by simulation errors) oscillate rapidly. When you sum these up over the entire evolution, most of the error cancels itself out, leaving only small residual terms from the very beginning and the very end of the process.

Datasets

While not a machine learning paper, the authors validated their theory using numerical simulations of several benchmark quantum problems:

  • Transverse Field Ising Model (TFIM): A classic model in condensed matter physics used to test the tightness of their new error bounds.
  • Molecular Nitrogen ($N_2$): Simulating the electronic structure (ground state preparation) using the STO-3G basis set.
  • Linear Systems of Equations: Solving $Ax=b$ via adiabatic quantum computing, a common task in scientific computing.

Key Insights and Ideas

  • The "Self-Healing" Effect: For Trotterization, the authors prove "weak" error cancellation—the error stays constant regardless of how long the simulation runs ($T$), rather than growing linearly with $T$.
  • Strong Cancellation with GQSP: For the more advanced GQSP method, they found "strong" cancellation: the simulation error actually decreases as the total evolution time $T$ increases.
  • Time-Independence is Sufficient: Surprisingly, they show that you can simulate time-dependent adiabatic evolution using simpler, time-independent algorithms. This eliminates the need for complex "Dyson series" methods that usually require extra "ancilla" qubits.
  • Path Optimization: By choosing a "smooth" scheduling function (a $Q$-th order path where derivatives at the start and end are zero), simulation errors can be suppressed exponentially.

Key Results

  • Massive Bound Tightening: For a small Transverse Field Ising Model (less than 6 qubits), the authors' new error estimate for first-order Trotterization was $10^6$ times tighter than previous analytical bounds.
  • Query Complexity Reduction: For first-order Trotterization, the number of quantum operations (query complexity) required to reach a target accuracy $\epsilon$ was reduced from $O(\epsilon^{-3/2})$ to $O(\epsilon^{-1})$.
  • Constant Error: Numerical tests on the $N_2$ molecule confirmed that as the evolution time $T$ increased, the simulation error (infidelity) remained flat or decreased, directly contradicting the traditional "error accumulation" model.

Limitations and Future Work

  • The Gap Closure Problem: The theory relies on the existence of an energy gap between the ground state and excited states. If the gap becomes extremely small (as happens during quantum phase transitions), the adiabatic condition might break down, potentially affecting the error cancellation.
  • Physical Noise vs. Algorithmic Error: This paper focuses on algorithmic errors (the math of the simulation). It does not account for physical noise (decoherence and gate errors) present in today's "noisy" (NISQ) quantum hardware.
  • Future Directions: The authors suggest exploring how these findings could be combined with other techniques like "counter-diabatic driving" (which speeds up adiabatic evolution) to create even more efficient algorithms for near-term quantum devices.

Summary updated

Abstract

Adiabatic evolution is a central paradigm in quantum physics. Digital simulations of adiabatic processes are generally viewed as costly, since algorithmic errors typically accumulate over the long evolution time, requiring exceptionally deep circuits to maintain accuracy. This work demonstrates that digital adiabatic evolution is intrinsically accurate and robust to simulation errors. We analyze two Hamiltonian simulation methods -- Trotterization and generalized quantum signal processing -- and prove that the simulation error does not increase with time. Numerical simulations of molecular systems and linear equations confirm the theory, revealing that digital adiabatic evolution is substantially more efficient than previously assumed. Remarkably, our estimation for the first-order Trotterization error can be 10^6 times tighter than previous analyses for the transverse field Ising model even with less than 6 qubits. The findings establish fundamental robustness of digital adiabatic evolution and provide a basis for accurate, efficient implementations on fault-tolerant -- and potentially near-term -- quantum platforms.
BibTeX
@misc{lu2025digital,
  author = {Yihua Lu and Yifei Huang and Dong Sung An and Qi Zhao and Dingshun Lv and Xiao Yuan},
  title = {{Digital adiabatic evolution is universally accurate}},
  year = {2025},
  eprint = {2510.12237},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2510.12237},
  howpublished = {arXiv},
}

Loading related content...

Figure