State preparation with parallel-sequential circuits

Zhi-Yuan Wei, Daniel Malz · 2025

Read State preparation with parallel-sequential circuits by Zhi-Yuan Wei, Daniel Malz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-14645.txt.

Introduction

Zhi-Yuan Wei and Daniel Malz design quantum circuits for preparing many-body states in a 2025 paper. Loading a desired quantum state into a device is a basic step for simulation, and circuit layout shapes how well it survives noisy hardware. Two known layouts sit at extremes: layered walls of gates and one long chain.

Their new family blends the two, with knobs that trade entanglement reach against correlation range. Computer tests show the blend prepares ground states of one-dimensional systems well. On simulated noisy devices, with idle errors and faulty two-qubit gates, the blend beat both parent layouts and a recent low-depth design across a wide range. Random versions of the layout also damped error spread. Used as a flexible trial circuit, it trained better than rivals. The paper offers hardware teams a middle path worth trying.

Catalog record

CREATORZhi-Yuan Wei, Daniel Malz
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-14645

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

State preparation with parallel-sequential circuits Zhi-Yuan Wei, Daniel Malz Abstract We introduce parallel-sequential (PS) circuits, a family of quantum circuit layouts that interpolate between brickwall and sequential circuits, which introduces control parameters governing the ratio of over the amount of entanglement and the maximum correlation distance they can express. We provide numerical evidence that PS circuits can efficiently prepare many-body ground states in one dimension. On noisy devices, characterized through both idling errors and two-qubit gate errors, we show that in a wide parameter regime, PS circuits outperform brickwall, sequential, and log-depth circuits from [Malz, Styliaris, Wei, Cirac, PRL 132, 040404 (2024)]. Additionally, we demonstrate that properly chosen noisy random PS circuits suppress error proliferation and, when employed as a variational ansatz, exhibit superior trainability. Introduction .—Quantum circuits are a unified framework to describe operations on quantum processors Nielsen and Chuang 2000 , but can also serve as a model of many-body dynamics. In this vein, different circuit layouts (arrangement of gates) correspond to distinct processes and can be used to define different variational families of states. For example, brickwall circuits naturally model evolution with spatial locality Vidal 2004 and thus efficiently prepare states obtained in a quench from a product state. In contrast, sequential circuits prepare states with constant entanglement entropy, which in one dimension includes all matrix product states (MPS) Schön et al. 2005 and in higher dimensions a rich family of states including topologically ordered ones Bañuls et al. 2008 ; Zaletel and Pollmann 2020 ; Wei et al. 2022 ; Liu et al. 2022 ; Chen et al. 2024 . The multi-scale entanglement renormalization ansatz (MERA) Shi et al. 2006 can model critical states and can be interpreted as arising from holographic dynamics. The advantage of using a variational family that is equivalent to a quantum circuit applied to a product state is that one can directly prepare states in that family on quantum devices. This approach is thus widely used to prepare quantum states represented as tensor network states by converting (or compiling) them into a circuit Smith et al. 2022 ; Ran 2020 ; Lin et al. 2021 ; Haghshenas et al. 2022 ; Rudolph et al. 2023 , but also in variational quantum algorithms that employ a quantum computer to optimize with a family defined through a circuit layout (usually brickwall circuits) Cerezo et al. 2021a ; Bharti et al. 2022 ; Tilly et al. 2022 . When preparing states on noisy devices, the circuit that achieves the highest fidelity in the noiseless setting (has the lowest approximation error ) may no longer be optimal. Idling errors accumulate with circuit depth and gate errors with the number of gates. The optimal circuit must balance these three sources of error. This depends on the architecture considered, and for concreteness, we consider one dimensional devices with spatially local gates, although the considerations can of course be generalized. In this setting, sequential circuits and MERA have a depth (in terms of local gates) linear in system size N and are therefore very susceptible to idling error. Brickwall circuits are the densest possible arrangement of spatially local gates and thus achieve the lowest depth for any task, but may use too many gates. If the goal is to prepare gapped ground states, which have constant (in system size) entanglement entropy Hastings 2007 ; Eisert et al. 2010 , as well as exponentially decaying correlations Hastings 2004 , one might heuristically expect that the optimal circuit has logarithmic depth, but a constant number of gates crossing any particular bond, which is fulfilled by none of the aforementioned layouts. While the RG circuits in Ref. Malz et al. 2024 can be used to prepare gapped ground states in terms of MPS with such an asymptotical scaling, they incur significant overhead when compiled into two-qubit gates sup . In this work, we introduce parallel-sequential (PS) circuits, designed to manage this trade-off optimally. PS circuits interpolate between brickwall and sequential circuits, with tunable degree of entanglement and correlation range. In contrast to sequential circuits, PS circuits are shallow, which suppresses idling errors, and compared to brickwall circuits they use fewer gates, which mitigates gate errors. Focusing on one dimension (see SI sup for higher-dimensional generalizations), we find that PS circuits are sufficiently expressive, more noise robust than other circuits, and are also easier to find variationally. Specifically, we first show numerically that in the absence of noise, single-layer PS circuits can represent short-range correlated MPS of bond dimension D=2 with a small infidelity \varepsilon using a circuit depth T\sim\log(N/\varepsilon) that is substantially shallower than the state-of-the-art RG circuits Malz et al. 2024 . Then we provide evidence that the PS circuits outperform the brickwall and sequential circuits for finding ground state of the XY model in noisy devices. To further support our findings, we analyse random PS circuits and show that those layouts with a constant number of layers have better trainability and less error proliferation compared to brickwall circuits of the same depth and sequential circuits. Figure 1: (a) Sequential, (b) brickwall, and (c) PS circuits. All three circuits layouts are obtained by repeating the same M -layer stack of gates, differing only in the amount by which neighbouring blocks are shifted relative to each other. Definition .—Before we introduce 1D parallel-sequential circuits, we note that M -layer sequential circuits [cf. fig. 1 (a)] and depth- 2M brickwall circuits [cf. fig. 1 (b)] can both be viewed as placing the same stack of M gates on each pair of neighbouring qubits, but with the difference that in sequential circuits, each subsequent stack is shifted up relative to its predecessor, whereas in brickwall circuits, they are alternately shifted up and down. In both cases, the entanglement entropy across any cut is bounded by S\leq 2M , but the correlations in a brickwall circuit extend only over 4M sites, whereas in a sequential circuit, they extend over the entire system. In PS circuits [cf. fig. 1 (c)], we take a mixed approach: the stack is shifted up l times (generating a sequential chunk) and then once down by l-1 (a break). To smoothen the break, we allow the sequential chunks to extend, such that neighboring chunks overlap over a distance q (red gates, q=2 in the figure). This defines PS circuits with n_{C}=\lceil N/l\rceil chunks and circuit depth T=l+q+2M-3 . In the rest of this article, we understand brickwall and sequential circuits as opposite limits of PS circuits, where M -layer PS circuits with l=2,q=1 , coincide with depth- 2M brickwall circuits, while those with l=N-1,q=1 are M -layer sequential circuits. With gate parameters \bm{\theta} and layout hyperparameters N,M,l,q , PS circuits can be expressed as a unitary U_{\rm PS}(\bm{\theta},N,M,l,q) , which also defines the associated variational family |\Psi_{\rm PS}(\bm{\theta},N,M,l,q)\rangle=U_{\rm PS}(\bm{\theta},N,M,l,q)|0\rangle^{\otimes N}. (1) The entanglement entropy of |\Psi_{\rm PS}\rangle across any non-overlapping cut is bounded by S\leq 2M , and the maximal correlation distance can be tuned via M , l , and q sup . By choosing appropriate scaling of parameters, PS circuits can create states with diverse entanglement and correlation scalings. For instance, by choosing constant M and l,q\sim\log N , PS circuits produce states with the same entanglement and

Read the complete text