Context-aware gate set tomography: Improving the self-consistent characterization of trapped-ion universal gate sets by leveraging non-Markovianity
P. Viñas, A. Bermudez · 2025
Read Context-aware gate set tomography: Improving the self-consistent characterization of trapped-ion universal gate sets by leveraging non-Markovianity by P. Viñas, A. Bermudez (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-2507-02542.txt.
Introduction
Pablo Viñas and Alejandro Bermudez wrote this paper in two thousand twenty-five. It improves gate set tomography, a method that checks a quantum computer by estimating its gates, state setups, and readouts all at once in a self-consistent way.
The standard method treats noise as fixed and free of past effects, but real trapped-ion devices show drift tied to motion of the ions. The authors model the light-shift gates used in such devices in fine detail. They show the drift can be folded into the gate set itself, which cuts the number of samples the check needs. The same tack may help on other platforms where the noise source can be modeled.
Catalog record
| CREATOR | P. Viñas, A. Bermudez |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 103 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-02542
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Context-aware gate set tomography: Improving the self-consistent characterization of trapped-ion universal gate sets by leveraging non-Markovianity P. Viñas, A. Bermudez Abstract To progress in the characterization of noise for current quantum computers, gate set tomography (GST) has emerged as a self-consistent tomographic protocol that can accurately estimate the complete set of noisy quantum gates, state preparations, and measurements. In its original incarnation, GST improves the estimation precision by applying the gates sequentially, provided that the noise makes them a set of fixed completely-positive and trace preserving (CPTP) maps independent of the history of previous gates in the sequence. This ‘Markovian’ assumption is sometimes in conflict with experimental evidence, as there might be time-correlated noise leading to non-Markovian dynamics or, alternatively, slow drifts and cumulative calibration errors that lead to context dependence, such that the CP-divisible maps composed during a sequence actually change with the circuit depth. In this work, we address this issue for trapped-ion devices with phonon-mediated two-qubit gates. By a detailed microscopic modeling of high-fidelity light-shift gates, we tailor GST to capture the main source of context dependence: motional degrees of freedom. Rather than invalidating GST, we show that context dependence can be incorporated in the parametrization of the gate set, allowing us to reduce the sampling cost of GST. Our results identify a promising research avenue that might be applicable to other platforms where microscopic modeling can be incorporated: the development of a context-aware GST. I Introduction Quantum information processors (QIPs) promise a speedup of certain computations relative to their classical counterparts Montanaro 2016 . To achieve this aim, however, several challenges still have to be overcome. One of these obstacles relates to the development of efficient methods for the characterization of noise. In the near term, detailed understanding and description of undesired interactions is central to advancing error mitigation techniques Temme et al. 2017 ; Nation et al. 2021 ; Yang et al. 2022 ; Cai et al. 2023 , which might allow quantum computers to enter a utility stage prior to fault tolerance (FT) Kim et al. 2023 . In the long term, quantum error correction (QEC) Calderbank and Shor 1996 ; Calderbank and Shor 1996 ; Steane 1996 ; Terhal 2015 offers a clear direction towards FT. When error rates of physical gates are small enough, one can obtain an exponential reduction on logical errors by a polynomial increase of the physical qubit number Nielsen and Chuang 2010 . From this perspective, a precise characterization of noise is crucial, not only to understand how to reduce error rates, but also to adapt and optimise the QEC protocols. With this objective, the community working in quantum characterization, verification, and validation (QCVV) has developed a wide variety of methods Eisert et al. 2020 ; Kliesch and Roth 2021 ; Gebhart et al. 2023 . Some approaches provide unified metrics to estimate average device performance, as exemplified by randomized benchmarking Emerson et al. 2005 ; Knill et al. 2008 ; Dankert et al. 2009 ; Magesan et al. 2011 ; Magesan et al. 2012 . In contrast, tomographic techniques Gill and Guta 2004 ; Banaszek et al. 2013 reconstruct the actual faulty operations in noisy QIPs. While the amount of information provided by the latter makes them much more appealing, there is a considerable associated experimental cost, as epitomized by gate set tomography (GST) Nielsen et al. 2021 ; Greenbaum 2015 ; Blume-Kohout et al. 2013 ; Stark 2014 ; Merkel et al. 2013 . What makes GST fundamentally different from the rest of techniques of the tomographic family Chuang and Nielsen 1997 ; Poyatos et al. 1997 ; Fiurášek and Hradil 2001 ; Sacchi 2001 ; Ježek et al. 2003 is that it dispenses with the assumption of error-free state preparation and measurement (SPAM). Because of this, it is often referred to in the literature as a self-consistent protocol. This is only possible by the simultaneous estimation of a reference initial state of the system \rho_{0} , a positive operator-valued measure (POVM) M_{0} with m possible outcomes Watrous 2018 , and an informationally-complete (IC) set of operations \big\{G_{i}:\,i\in\mathbb{G}\,\big\} . The union of all these components is known as the gate set \mathcal{G}=\left\{\rho_{0},\big\{G_{i}:\,i\in\mathbb{G}\,\big\},M_{0}\right\}, (1) and the goal of GST is to estimate it both accurately and with a high precision using circuit-based measurement data. Succinctly, GST provides an estimate of the gate set under study by fitting experimentally-measured frequencies to a model describing the gates. Optionally, some gate sequences might be chained several times to produce frequencies that enhance the precision of the estimation. This is the ‘long-sequence’ version of GST Nielsen et al. 2021 . Having outlined the main idea behind GST, we introduce now its two main drawbacks, which we address with novel approaches in this work. The more apparent one is the large number of experiments it requires. A generic two-qubit gate set description takes N_{\mathrm{param}}\sim 10^{3} parameters Ostrove et al. 2023 . As a consequence, the number of different circuits required to run a given experimental GST design is N_{\mathrm{GST}}\sim 10^{4} , with some variations depending on the circuit depth employed. This amount can substantially increase when crosstalk is modeled. Even with N_{\mathrm{samples}}=10^{3} samples per circuit, which might already be too conservative given the minute errors of current high-fidelity QIPs one aims at estimating, one obtains a total of N_{\mathrm{shots}}\sim 10^{7} executions. Considering the gate speed and readout times of some of these QIPs, such as trapped-ion devices Bruzewicz et al. 2019 , the total execution time of two-qubit GST is often impractical for a real time characterization. The second issue which restricts GST is more fundamental and platform-agnostic: it is unable to characterize context-dependent noise Rudinger et al. 2019 . GST relies entirely on the assumption that a noisy gate does not change with the number of times it is applied, nor with the specific previous history of operations along the circuit. When deviations from this assumption become significant, the GST protocol fails to provide reliable estimates. These violations of the constant-map behavior are often referred to in the GST literature as evidence of “non-Markovianity” Nielsen et al. 2021 . Because laboratory experiments across different platforms often suffer from Markovian deviations Rebentrost et al. 2009 ; Wallman et al. 2016 ; Rebentrost et al. 2009 ; Jing and Wu 2018 ; Piltz et al. 2014 , the overhead required to run GST is not generally justified. This highlights the importance of developing characterization techniques compatible with non-Markovianity. In this manuscript, we show that one can go beyond this limitation with a specific setup in mind: trapped-ion QIPs, focusing on a geometric entangling gate known as light-shift (LS) gate Leibfried et al. 2003a . As we extensively discussed in our previous work for single-qubit GST Viñas and Bermudez 2025 , a microscopically-motivated parametrization of GST can greatly reduce the number of parameters required to describe a gate set. This directly translates into a substantial reduction of the sampling complexity needed to achieve a desired GST precision. In the present work, we use our microscopic parametrization to include noise in the LS gate, extending GST to the fully universal gate set that has demonstrated the highest fidelities to date Ballance et al. 2016a ;