⚛️ quantum physics

Quantum Imaginary Time Evolution on an Infinite 1D Chain

This paper introduces a quantum-circuit algorithm for performing imaginary-time evolution on infinite one-dimensional lattice systems using a parameterized uniform matrix product state ansatz, which is benchmarked on the transverse-field Ising model via classical and IBM Quantum simulations to analyze the impact of finite-sampling noise on convergence.

Original authors: Hao-Ti Hung, Tung Tsao, Ying-Jer Kao

Published 2026-09-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hao-Ti Hung, Tung Tsao, Ying-Jer Kao

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 vast landscape of modern physics, there is a persistent challenge: understanding how countless tiny particles interact to create the complex behaviors we see in the world around us. When these particles are governed by the strange rules of quantum mechanics, the mathematics becomes so tangled that even the most powerful supercomputers struggle to keep up. Scientists have long relied on a clever shortcut called a "matrix product state," which simplifies these infinite chains of particles by focusing on how each one connects to its immediate neighbors. This approach has been a workhorse for classical computers, but a new frontier has opened with the arrival of quantum computers. These machines, still in their noisy and imperfect early stages, promise to simulate quantum systems directly. The goal is to use these devices to find the "ground state" of a system, which is its most stable, lowest-energy configuration. Knowing this state is like knowing the resting position of a complex machine; it reveals the fundamental nature of the material and how it will behave under different conditions.

A team of researchers at National Taiwan University has taken a significant step toward this goal by teaching a quantum computer how to simulate an infinite line of particles. Their work focuses on a specific mathematical technique called imaginary-time evolution. In the real world, time moves forward, and quantum systems evolve in ways that preserve their total probability, much like a spinning top that never falls over. However, to find the lowest energy state, scientists often use a mathematical trick that treats time as if it were imaginary. This process acts like a filter, gradually washing away the high-energy, unstable parts of a system until only the most stable, lowest-energy state remains. The problem is that this "imaginary" filter is not a natural fit for quantum computers, which are built to perform operations that preserve probability. The researchers had to devise a way to force this non-standard process to work within the rigid rules of quantum hardware.

To solve this, the team developed a method that represents the infinite chain of particles using a parameterized quantum circuit. Think of this circuit as a flexible, adjustable machine made of quantum gates. The researchers set this machine to mimic the structure of the particle chain, then asked it to evolve toward the lowest energy state. They faced a choice in how to handle the non-standard "imaginary" filter. One approach was to approximate the filter using standard quantum gates, essentially faking the process. The other was to add an extra helper particle, known as an ancilla qubit, to the system. This helper allowed them to embed the difficult filter into a larger, perfectly valid quantum operation. By testing both methods on a classical simulator and on real quantum hardware from IBM, they found that the method using the extra helper particle was far more reliable. It kept the simulation stable and accurate, whereas the approximation method began to drift and produce incorrect results after a certain amount of time.

The researchers then pushed their findings further by applying a refinement technique called the quantum Lanczos algorithm. This step is akin to taking a rough sketch of the lowest energy state and polishing it into a precise portrait. By collecting data from different stages of the simulation and combining them intelligently, they were able to correct for the errors introduced by the imperfect nature of current quantum machines. When they ran these simulations on actual IBM quantum devices, they encountered a new hurdle: the noise inherent in real hardware. Because the computers must measure the outcome of the simulation many times to get a reliable answer, the results fluctuate due to statistical randomness. The team discovered that these fluctuations cause the simulation to stall before it reaches the perfect lowest energy, stopping just short of the target. However, by analyzing the distribution of these stalled results, they could still extract a highly accurate estimate of the true ground state.

The study confirms that it is possible to simulate an infinite quantum system on a quantum computer using a surprisingly small number of qubits—just seven for the specific model they tested. This is a crucial demonstration because it shows that the complexity of the system does not require the computer to grow indefinitely in size. The researchers found that while the approximation method is simpler, it introduces errors that grow over time, making it unsuitable for long simulations. In contrast, the method using the extra helper qubit, though it requires slightly more hardware resources, maintains its accuracy and converges toward the correct answer. The work also highlights the critical role of statistical noise in these experiments. The team showed that the errors are not random chaos but follow a predictable pattern, allowing them to be accounted for and corrected.

Ultimately, this research provides a practical roadmap for using today's imperfect quantum computers to solve problems that were previously out of reach. By successfully simulating an infinite chain of particles and refining the results to account for hardware noise, the team has shown that these devices can do more than just mimic small, finite systems. They have demonstrated a path toward understanding the fundamental properties of materials that stretch on forever. The findings suggest that with the right combination of algorithms and error-correction techniques, quantum computers can begin to tackle the most complex questions in condensed matter physics, offering a glimpse into a future where we can design new materials by simulating their quantum behavior directly.

Technical Summary: Quantum Imaginary Time Evolution on an Infinite 1D Chain

Problem Statement
The advent of Noisy Intermediate-Scale Quantum (NISQ) devices has spurred interest in simulating quantum many-body systems. While algorithms for real-time evolution (e.g., TEBD, TDVP) have been successfully adapted to quantum circuits by leveraging the unitary nature of quantum gates, extending these methods to Imaginary Time Evolution (ITE) remains challenging. The ITE operator, $e^{-H\tau}$ , is non-unitary, rendering it incompatible with standard unitary quantum gates. Existing approaches to this problem—Quantum ITE (QITE), Variational ITE (VITE), and Probabilistic ITE (PITE)—have primarily been restricted to finite-size systems. Furthermore, on real quantum hardware, the statistical noise inherent in estimating cost functions via repeated measurements significantly impacts the convergence of variational optimization. This work addresses the gap in performing ITE on infinite one-dimensional lattice systems using quantum circuits and analyzes the statistical effects of finite-sampling noise on convergence.

Methodology
The authors propose a quantum-circuit algorithm based on the Time-Dependent Variational Principle (TDVP) applied to a uniform Matrix Product State (uMPS) ansatz. The core methodology involves:

Mapping uMPS to Quantum Circuits:
- A translationally invariant uMPS is mapped to a parameterized quantum circuit (PQC). For a bond dimension $D=2^N$ , a single-site unit cell maps to a unitary gate acting on $N+1$ qubits.
- To handle the non-unitary environment tensors ( $l$ and $r$ ) required for the TDVP cost function, the authors construct unitary gates by enlarging the Hilbert space (embedding non-unitary tensors into larger unitary matrices).
- The cost function, representing the overlap between the evolved state and the target state, is formulated as a quantum circuit where the output probability of the all-zero state serves as the metric for optimization.
Handling Non-Unitarity (Two Strategies):
- Unitary Approximation (QITE-like): The non-unitary ITE operator is approximated by a unitary operator $e^{-iA(\tau)d\tau}$ , where $A(\tau)$ is a Hermitian operator determined by minimizing the distance between the normalized ITE state and its unitary approximation. This approach does not require additional qubits but introduces approximation errors.
- Ancilla-Assisted Embedding (PITE-like): The non-unitary operator is embedded into a larger unitary operator using an additional ancilla qubit via block encoding. This method avoids approximation errors but increases the qubit count (e.g., from 6 to 7 qubits for $D=2$ ).
Quantum Lanczos (QLanczos) Refinement:
- To mitigate errors from unitary approximations and statistical noise, the authors employ the Quantum Lanczos algorithm.
- Instead of generating a Krylov subspace from $H^k|\phi\rangle$ , QLanczos constructs a subspace from the sequence of ITE states $|\phi_l\rangle = e^{-l d\tau H}|\phi_0\rangle$ .
- By solving a generalized eigenvalue problem using the overlap matrix $S$ and Hamiltonian matrix $H$ constructed from these states, the algorithm refines the ground-state energy estimate.

Key Results
The authors benchmarked their algorithm using the Transverse-Field Ising Model (TFIM) at its critical point ( $H = \sum_i (-\sigma^z_i \sigma^z_{i+1} + \sigma^x_i)$ ) with a bond dimension of $D=2$ .

Statevector Simulations:
- The ancilla-assisted method closely matched classical Tensor Network (TN) TDVP results, converging to the expected ground-state energy density ( $\approx -1.2732$ ).
- The unitary approximation method exhibited an unexpected increase in energy at long evolution times ( $\tau \approx 1$ ), attributed to the limited correlation-domain size inherent in the infinite-system setup and the unitary approximation error.
Real Quantum Device Simulations (IBM Quantum):
- Simulations on devices (e.g., ibm_nairobi, ibm_auckland) demonstrated that while raw ITE trajectories are noisy due to finite-sampling statistics, the QLanczos post-processing significantly reduced errors, bringing the energy estimates closer to the ground-state value.
Statistical Analysis of Noise:
- The study analyzed the distribution of cost values and energy densities derived from measurement-based estimates.
- While individual cost estimates follow a normal distribution (Bernoulli process), the aggregate distribution of energy densities across multiple optimization runs forms a right-skewed, Maxwell-Boltzmann-like envelope.
- This distribution arises from a combination of run-to-run variations in "stall points" (where optimization halts due to noise) and Gaussian fluctuations within individual runs. The analysis indicates that optimization trajectories often stall before reaching the variational minimum when the optimizer tolerance becomes comparable to the finite-sampling noise.

Significance and Claims
The paper claims to demonstrate the first implementation of quantum-circuit ITE for infinite 1D systems. Its primary contributions include:

Algorithmic Extension: Successfully adapting the TDVP framework to infinite lattices using parameterized quantum circuits, maintaining a constant number of gates regardless of evolution time.
Noise Characterization: Providing a detailed statistical study of how finite-sampling noise affects variational convergence in ITE, identifying the specific distributional characteristics of the resulting energy errors.
Error Mitigation: Validating the utility of the QLanczos algorithm in refining ground-state energy estimates on both simulators and real hardware, effectively compensating for both unitary approximation errors and measurement noise.
Scalability: While the current implementation focuses on $D=2$ (requiring only 6 or 7 qubits), the authors note that the mapping between uMPS and quantum gates can be generalized to larger bond dimensions, offering a pathway to simulate more complex, strongly entangled systems in the future.

The work concludes that while current NISQ devices introduce significant statistical challenges, the combination of variational ansätze, ancilla-assisted embedding, and QLanczos refinement provides a viable route for simulating infinite quantum systems and estimating ground-state properties.

Technical Summary: Quantum Imaginary Time Evolution on an Infinite 1D Chain

More like this