Anticoncentration and State Design of Doped Real Clifford Circuits and Tensor Networks

Beatrice Magni, Markus Heinrich, Lorenzo Leone, Xhek Turkeshi · 2025

Read Anticoncentration and State Design of Doped Real Clifford Circuits and Tensor Networks by Beatrice Magni, Markus Heinrich, Lorenzo Leone, Xhek Turkeshi (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-2512-15880.txt.

Introduction

Beatrice Magni and co authors study anticoncentration and state design in doped real Clifford circuits in a 2025 paper. Clifford circuits are a basic, classically tractable family of quantum circuits. Adding small amounts of magic, the resource that makes quantum computation powerful, pushes them toward full quantum behavior, and the cost of that transition matters.

Using Weingarten calculus for the real Clifford group, the team derives the exact overlap distribution of real stabilizer states, a new universality class. Local real circuits recover that global statistic at logarithmic depth. A sharp hierarchy appears: recovering Haar statistics needs a polylogarithmic number of magic states, while full unitary Clifford statistics need only a single phase gate.

Catalog record

CREATORBeatrice Magni, Markus Heinrich, Lorenzo Leone, Xhek Turkeshi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-15880

Browse the full library · About the Tower of Babel

Full text

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

Anticoncentration and State Design of Doped Real Clifford Circuits and Tensor Networks Beatrice Magni, Markus Heinrich, Lorenzo Leone, Xhek Turkeshi Abstract We investigate the statistical properties of orthogonal, or real, Clifford circuits doped with magic and imaginary resources. By developing the Weingarten calculus for the real Clifford group, we derive the exact overlap distribution of real stabilizer states, identifying a new universality class: the orthogonal Clifford Porter-Thomas distribution. We prove that local real architectures recover this global statistic in logarithmic depth. Furthermore, we uncover a sharp hierarchy in resource requirements: while retrieving Haar statistics necessitates a polylogarithmic amount of magic states, recovering the full unitary Clifford statistics requires only a single phase gate. I Introduction Clifford circuits and magic states are central ingredients to fault-tolerant quantum computing Gottesman 1997 ; Gottesman 1998 ; Bravyi and Kitaev 2005 . Indeed, Clifford circuits drive quantum error correction and admit efficient fault-tolerant implementations. However, Clifford dynamics alone are efficiently classically simulatable Aaronson and Gottesman 2004 ; Bravyi and Gosset 2016 ; Bravyi et al. 2019 ; Sierant et al. 2023 . Only the additional supply of magic states can surpass efficient simulation schemes and enable universal quantum computation. Clifford circuits alone, or “doped” with a controlled number of magic resources, have proven an invaluable tool in various fields, from the characterization of out-of-equilibrium systems Leone et al. 2021 ; Magni and Turkeshi 2025 ; Lami and Collura 2024 ; Aditya et al. 2025 ; Tirrito et al. 2025a ; Turkeshi et al. 2025a ; Tirrito et al. 2025b ; Falcão et al. 2025 ; Odavić et al. 2025 ; Bejan et al. 2024 ; Hou et al. 2025 ; Varikuti et al. 2025 ; Szombathy et al. 2025a ; Szombathy et al. 2025b ; Niroula et al. 2024 ; Turkeshi 2024 ; Zhou et al. 2020 , to quantum applications Yoganathan et al. 2019 ; Leone et al. 2025 ; Lóio et al. 2025 ; Liu and Clark 2024 ; Haug et al. 2024 ; Haferkamp et al. 2022a ; Mao et al. 2025 ; Haug et al. 2025 ; Chia et al. 2024 ; Hinsche et al. 2023 ; Fux et al. 2025 ; Harper et al. 2025 ; Nakhl et al. 2025 , even touching questions of quantum gravity and black-hole physics Leone et al. 2022a ; Leone et al. 2024 ; Oliviero et al. 2024 ; Haferkamp et al. 2022b ; Yoshida and Yao 2019 . Crucial to a wide range of applications is the ability of doped Clifford circuits to generate random quantum states. Two fundamental ways of quantifying this are anticoncentration and state designs. The former implies that the circuit spreads probability amplitudes broadly across the Hilbert space rather than localizing on a few basis states Bouland et al. 2018 ; Boixo et al. 2018 ; Hangleiter et al. 2018 ; Dalzell et al. 2022 ; Heinrich et al. 2025 . The latter ensures that the output ensemble reproduces the lower-order moments of the Haar measure Nakata et al. 2025 ; Lami et al. 2025 ; Christopoulos et al. 2025 . Previous results on doped Clifford circuits established that a logarithmic amount of magic states suffices to mimic these statistical properties of Haar-random states Magni et al. 2025 . However, the statistical properties of further restricted ensembles remain largely unexplored. Here, we address this gap by characterizing the anticoncentration and state design properties of real, or orthogonal, Clifford circuits and tensor networks, both in the pure regime and when doped with controlled amounts of magic Liu and Winter 2022 ; Chitambar and Gour 2019 or imaginary resources Wu et al. 2021a ; Deneris et al. ; Shi 2025a ; Shi 2025b injected into them. We address three key questions: (i) What characterizes the output distribution and state design of random real Clifford states? (ii) Under what conditions do local tensor networks or circuits recover these global statistics? (iii) How does doping with different resources affect the anticoncentration and state design? To address these questions, we exploit the Weingarten calculus for the real Clifford group. Combining this with the stabilizer tableau formalism, we derive rigorous results for random tensor networks and local circuits, identifying a distinct universality class: the orthogonal Clifford Porter-Thomas (OCPT) distribution. We show that Random Matrix Product States (RMPS) and shallow local circuits saturate to this distribution with polynomial bond dimension and logarithmic depth, respectively. Furthermore, we characterize the impact of resource doping: while polylogarithmic real or complex magic is required to approach Haar statistics, a single imaginary state suffices to recover unitary Clifford statistics strictly. These findings establish a sharp separation of scales between magic and imaginarity, quantifying the distinct costs required to access orthogonal and unitary state designs. Dynamical Ensemble Complexity Cost Doping Resource Amount State-design Ref. unitary Haar ( \mathcal{U} ) Circuit [ \mathrm{depth}\sim\log N ] RMPS [ \chi\sim\text{poly}(N) ] — — \mathbb{C} Haar Schuster et al. 2025 ; Lami et al. 2025 orthogonal Haar ( \mathcal{O} ) \varnothing — \mathbb{R} Haar Magni et al. 2025 ; Zhang et al. 2025a Imaginary States |\pm i\rangle O(1) \mathbb{C} Haar Haug et al. ; SM unitary Clifford ( \mathcal{C} ) \varnothing — \mathbb{C} Stab Magni et al. 2025 ; Zhang et al. 2025a Real/Complex Magic States |H\rangle O(\log N) \mathbb{C} Haar orthogonal Clifford ( \mathcal{R} ) (This work) Circuit [ \mathrm{depth}\sim\log N ] RMPS [ \chi\sim\text{poly}(N) ] \varnothing — \mathbb{R} Stab Eq. ( 13 ) Real Magic |H\rangle O(\log N) \mathbb{R} Haar Eq. ( 17 ) Complex Magic |T\rangle O(\log N) \mathbb{C} Haar Eq. ( 18 ) Imaginary state |\pm i\rangle O(1) \mathbb{C} Stab Eq. ( 19 ) Table 1: Comparison of state designs. Note the vertical and horizontal separators emphasizing the resource transitions. II Methods We consider an N -qubit system with Hilbert space dimension d\equiv 2^{N} , initialized in the state |\psi_{0}\rangle . The system evolves via a depth- t quantum circuit, yielding |\psi\rangle=\mathbf{C}_{t}|\psi_{0}\rangle . The circuit is decomposed into layers as \mathbf{C}_{t}=\prod_{s=1}^{t}\left(\bigotimes_{\lambda\in\Lambda_{s}}C_{\lambda}\right) , where \Lambda_{s} specifies the set of interaction supports active in layer s . Each local gate C_{\lambda} is a random orthogonal Clifford, i.e. C_{\lambda}\in\mathcal{R}_{n} s.t. C_{\lambda}C_{\lambda}^{T}=\mathbb{1} , and acts on the contiguous qubits in \lambda 1 1 1 Explicitly, C_{\lambda}\equiv C_{\lambda_{1},\ldots,\lambda_{n}}=\mathbb{I}^{\otimes(\lambda_{1}-1)}\otimes C\otimes\mathbb{I}^{\otimes(N-\lambda_{n})} , where n=|\lambda| depends on the geometry under consideration and C\in\mathcal{R}_{n} . For simplicity, we use the shorthand C_{\lambda}\in\mathcal{R}_{n} . . We defer the detailed definition of the architectures \mathfrak{A}=\{\Lambda_{s}\}_{s=1}^{t} to the next section. Here, we introduce the central quantities of interest. To probe anticoncentration, we study the probability distribution of the scaled output probabilities w=d|\langle\mathbf{y}|\psi\rangle|^{2} P(w)=\mathbb{E}_{\mathbf{C}_{t}}\left[\frac{1}{d}\sum_{\mathbf{y}\in\mathcal{B}}\delta\left(w-d|\langle\mathbf{y}|\psi\rangle|^{2}\right)\right], (1) where \mathcal{B}=\{|\mathbf{y}\rangle\}_{\mathbf{y}=0}^{d-1} is the computational basis and the circuit average \mathbb{E}_{\mathbf{C}_{t}}[\bullet] is over each independent gate. The distribution ( 1 ) is equivalently captured by the Inverse Participation Ratios (IPRs) Luitz et al. 2014 ; Macé et al. 2019 ; Liu et al. 2025 ; Sierant et al. 2022 ; Sierant and Turkeshi 2022 , which are proportional to its moments I_{k}\equiv\mathbb{E}_{\mathbf{C}_{t}}\sum_{\mathbf{y}\in\mathcal{B}}|\langle\mathbf{y}|\psi\rangle|^{2k}=d^{1-k}\int w^{k}P(w)dw\,. (2) Complementarily, we assess state k -design properties via the frame potential De Luca et al. 2025 , \mathcal{F}_{k}\equiv\mathbb{E}_{\mathbf{C}_{t},\mathbf{C}_{t}^{\prime}}[|\langle\psi^{\prime}|\psi\rangle|^{2k}]\,, (3) which measures the overlap between states |\psi\rangle and |\psi^{\prime}\rangle generated by two random circuits \mathbf{C}_{t} and \mathbf{C}_{t}^{\prime} . All aforementioned quantities involve statistical averages over the

Read the complete text