⚛️ quantum physics

Optimal Ground-State Preparation with a Guiding State

This paper presents two optimal algorithms for preparing a ground state with high probability and precision, leveraging a guiding state and known energy estimates to achieve query complexities that are optimal up to constant factors in terms of Hamiltonian evolution and state preparation operations.

Original authors: Stacey Jeffery, Rolando D. Somma, Freek Witteveen, Ronald de Wolf

Published 2026-10-06

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Stacey Jeffery, Rolando D. Somma, Freek Witteveen, Ronald de Wolf

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the quantum world, the most stable and lowest-energy configuration of a system is known as its ground state. Finding this state is a fundamental task for scientists trying to understand how molecules bond, how new materials might behave, or how complex chemical reactions unfold. However, locating this ground state is notoriously difficult. Imagine a vast, foggy landscape of hills and valleys, where the deepest valley represents the ground state. A computer trying to find it must navigate this terrain without getting stuck in a shallow dip that looks like the bottom but isn't. To make matters more challenging, the computer often starts with only a rough guess of where the valley might be, and the tools it uses to explore the landscape are imperfect, introducing small errors that can accumulate and lead it astray.

For years, researchers have struggled to prepare a quantum computer to reliably settle into this ground state. They have had to choose between methods that were fast but prone to error, or methods that were accurate but required so many steps that they were impractical. A team of researchers has now developed a new approach that solves this problem efficiently. By combining two distinct strategies, they have created a method that prepares the ground state with high precision using the fewest possible steps allowed by the laws of physics. Their work proves that it is possible to reach the target state without wasting computational resources on unnecessary corrections, effectively closing the gap between what is theoretically possible and what can be achieved in practice.

The researchers focused on a scenario where they already have a rough estimate of the energy level of the ground state, denoted as ˜E0, which is guaranteed to be within a small distance δ of the true ground state energy E0. They also have a starting point, or "guiding state," that is somewhat close to the target. Think of this guiding state as a map that points generally in the right direction but lacks the fine detail needed to find the exact spot. Crucially, the algorithms require that the energy gap between the ground state and all other energy levels is at least three times the size of this estimation error (a gap of at least 3δ). The goal was to refine this map and guide the system into the precise ground state. The team demonstrated that by using a specific type of filtering process to isolate the correct energy level, followed by a technique to amplify the probability of finding that state, they could achieve the result with optimal efficiency. They showed that the number of steps required depends directly on how close the starting guess is to the target and how distinct the ground state is from other nearby energy levels.

To achieve this, the team employed two different algorithms, both of which arrive at the same optimal result. The first method uses a technique called amplitude amplification, which is a way of boosting the likelihood of the correct outcome while suppressing the wrong ones. In a standard approach, this process would require many extra steps to correct for the small errors introduced by the initial filtering. The researchers avoided this penalty by carefully interleaving the amplification with error reduction, ensuring that the process remained efficient even when the starting information was imperfect. This allowed them to reach the ground state without the extra overhead that had plagued previous methods.

The second method relies on a more modern mathematical framework known as transducers. This approach treats the quantum algorithm as a machine that transforms inputs into outputs in a way that allows different parts of the process to be combined seamlessly. By constructing the algorithm as a series of these transducers, the researchers could combine the filtering and amplification steps without the usual loss of efficiency that occurs when error-prone components are linked together. This composition allowed them to build a single, streamlined process that handles the entire task in one go, avoiding the need for repeated corrections. The result is a method that uses the minimum number of operations required to solve the problem, matching the theoretical lower bound for how fast such a task can be completed.

The significance of this work lies in its optimality. The researchers proved that their method cannot be improved upon in terms of the number of times the quantum computer needs to interact with the system's energy landscape. This is a crucial finding because it sets a definitive limit on the resources needed for ground-state preparation. By showing that the process can be done with a number of steps proportional to the inverse of the starting overlap and the energy gap, they have provided a clear roadmap for future quantum simulations. This means that as quantum computers grow in power, scientists will be able to use these optimal methods to study increasingly complex systems, from new drugs to advanced materials, with a level of efficiency that was previously out of reach.

The paper also addresses the practical details of implementing these algorithms on real hardware. The researchers accounted for the additional memory and control gates required to run the process, showing that the overhead is manageable and scales reasonably with the size of the problem. They demonstrated that the method works even when the initial guess is not perfect, as long as it is within a certain range of the true ground state and the energy levels are sufficiently separated. This robustness is essential for real-world applications, where perfect information is rarely available. By proving that the method works under these realistic conditions, the team has provided a reliable tool for the next generation of quantum experiments.

Ultimately, this research represents a maturation of quantum algorithm design. It moves beyond the era of trial and error, where methods were often chosen based on what seemed to work, to a stage where the best possible approach is known and proven. The ability to prepare the ground state with optimal efficiency removes a major bottleneck in quantum computing. It allows researchers to focus on the science of the systems they are studying, rather than struggling with the limitations of the tools they use to study them. As quantum technology continues to evolve, these foundational improvements will enable more accurate and powerful simulations, bringing us closer to solving some of the most complex problems in chemistry and physics.

Technical Summary: Optimal Ground-State Preparation with a Guiding State

Problem Statement
The paper addresses the problem of preparing the unique ground state $|\psi_0\rangle$ of a Hamiltonian $H$ given a gap $\Delta$ between the ground energy $E_0$ and the first excited energy $E_1$ . The available resources are:

A unitary $U = e^{iH}$ (accessible via Hamiltonian simulation).
A "guiding state" preparation unitary $A$ such that $A|0\rangle$ has an overlap of at least $\gamma$ with $|\psi_0\rangle$ .
An estimate $\tilde{E}_0$ of the ground energy such that $|\tilde{E}_0 - E_0| \le \delta$ , where $\delta \le \Delta/3$ .

The goal is to output a state $\rho$ such that $\||\psi_0\rangle\langle\psi_0| - \rho\|_1 \le \eta$ with success probability at least $1-\varepsilon$ . The authors aim to achieve this with optimal query complexity, specifically minimizing the number of applications of $U$ and $U^\dagger$ , which are typically the dominant cost factors.

Methodology
The paper presents two distinct algorithms to solve this problem, both achieving optimal query complexity. Both approaches rely on the high-level strategy of "marking" the ground state using the energy estimate $\tilde{E}_0$ and then amplifying the amplitude of the marked state.

Algorithm 1: Robust Amplitude Amplification (Section 2)
- Core Idea: This approach adapts the "quantum search on bounded-error inputs" technique from [HMdW03]. It addresses the challenge that marking the ground state via phase estimation introduces errors (false positives).
- Mechanism: The algorithm recursively interleaves amplitude amplification steps with error-reduction steps.
  - Amplification: A standard Grover-like step amplifies the component of the state where the flag qubit indicates the ground state. However, this also amplifies "false positives" (excited states incorrectly marked).
  - Error Reduction: Between amplification rounds, a phase-estimation-based verifier $Q_\xi$ is applied to suppress the amplitude of these false positives.
- Optimization: By carefully choosing the error parameters $\eta_\ell$ at each recursion level $\ell$ , the algorithm ensures that the ratio of false positives to true ground state components decreases rapidly. The recursion depth $L$ is chosen such that the overlap grows from $\gamma$ to a constant. The total complexity avoids the logarithmic overhead ( $O(\log(1/\gamma))$ ) typically incurred by straightforward error reduction in amplitude amplification.
Algorithm 2: Transducer Composition (Section 3)
- Core Idea: This approach utilizes the "transducer" framework introduced in [BJY24], which allows for the composition of bounded-error quantum algorithms without incurring logarithmic overheads from error reduction.
- Mechanism: The algorithm constructs a transducer as a composition of two sub-transducers:
  - Filtering Transducer (LCU): Implements a linear combination of unitaries (LCU) to apply a filter function $f_\delta(H)$ that suppresses excited states while preserving the ground state. This is constructed using Fourier coefficients of a specific filter function, allowing for efficient implementation with a complexity proportional to the first moment of the coefficients rather than the degree of the polynomial.
  - Amplitude Amplification Transducer: A novel construction for amplitude amplification that uses a "leak schedule" to peel off the marked state incrementally. This transducer is made "canonical" with respect to the oracle $A$ , allowing the Las Vegas query complexity (the expected number of calls to $A$ ) to be bounded by $O(1/\gamma)$ rather than the worst-case complexity.
- Composition: These transducers are composed to form a single transducer for ground-state preparation. When mapped back to a standard quantum algorithm via Theorem 3.2, the resulting algorithm achieves the optimal query complexity.

Key Results
The main result is stated in Theorem 1.3: There exists a quantum algorithm for guided ground-state preparation with estimate that uses:

$O\left(\frac{1}{\gamma\Delta}\log\frac{1}{\varepsilon} + \frac{1}{\Delta}\log\frac{1}{\eta}\right)$ controlled calls to $U$ and $U^\dagger$ .
$O\left(\frac{1}{\gamma}\log\frac{1}{\varepsilon}\right)$ controlled calls to $A$ and $A^\dagger$ .

Corollary 1.4 combines this with the optimal ground-state energy estimation algorithm from [JW26] to solve the problem without a priori energy estimates, incurring only a polylogarithmic overhead in the estimation precision.

The paper establishes that these bounds are optimal up to a constant factor regarding the number of applications of $U$ and $U^\dagger$ (when $\varepsilon = \eta$ ), matching the lower bound proven in [SdW26].

Significance and Claims

Optimality: The paper claims to close the gap between the upper and lower bounds for ground-state preparation. Previous best-known algorithms incurred an $O(\log(1/\gamma))$ factor in the query complexity of $U$ , whereas this work removes that factor.
Technique Advancement:
- The first algorithm demonstrates that the [HMdW03] technique for bounded-error search can be effectively adapted to the continuous spectrum of Hamiltonians, working in the unknown eigenbasis of $H$ .
- The second algorithm introduces new transducer constructions for LCU and amplitude amplification. Specifically, the amplitude amplification transducer improves upon previous work [ARZ26] by reducing the space complexity (ancilla qubits) by a multiplicative logarithmic factor.
Context: This work serves as the state-preparation follow-up to the authors' recent ground-state energy estimation papers [JW26, SdW26]. Together, these results provide a complete, optimal solution for the ground-state preparation problem given a guiding state.

The authors note that the majority of high-level ideas are human-derived, though specific components (such as the leak schedule in the transducer and certain proof details) received assistance from Large Language Models. All proofs and text have been rigorously checked and rewritten by the authors.

Technical Summary: Optimal Ground-State Preparation with a Guiding State

More like this