Witnesses of non-Gaussian features as lower bounds of stellar rank

Jan Provazník, Šimon Bräuer, Vojtěch Kala, Jaromír Fiurášek, Petr Marek · 2026

Read Witnesses of non-Gaussian features as lower bounds of stellar rank by Jan Provazník, Šimon Bräuer, Vojtěch Kala, Jaromír Fiurášek, Petr Marek (2026) 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-2603-03185.txt.

Introduction

Jan Provazník, Šimon Bräuer, Vojtěch Kala, and colleagues turn easy lab measurements into hard guarantees about quantum light. Truly non-Gaussian states of light power quantum computing and precision sensing beyond classical limits. The stellar rank grades that non-Gaussian character, but measuring it directly is impractical.

Witnesses, by contrast, use simple averages and spreads of measurable quantities to certify non-Gaussian behavior. The team proves those witnesses set rigorous lower bounds on stellar rank, and arranges them into a hierarchy of thresholds matching the rank ladder. Normalized versions of the expectation and variance quantifiers make the tests comparable across setups. The bridge lets experimentalists certify how deep a state's non-Gaussianity runs with scalable, routine measurements.

Catalog record

CREATORJan Provazník, Šimon Bräuer, Vojtěch Kala, Jaromír Fiurášek, Petr Marek
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31097
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.03185

Browse the full library · About the Tower of Babel

Full text

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

Witnesses of non-Gaussian features as lower bounds of stellar rank Jan Provazník, Šimon Bräuer, Vojtěch Kala, Jaromír Fiurášek, Petr Marek Abstract Quantum non-Gaussian states and operations serve as fundamental resources for universal quantum computation, error correction, and high-precision metrology, extending beyond the Gaussian limits. While the stellar rank provides a rigorous hierarchical measure of non-Gaussianity, it remains challenging to determine experimentally. Conversely, witnesses of non-Gaussian features, based on the expectation values and variances of measurable observables, offer an accessible method for certifying non-Gaussian behavior but lack a direct connection to stellar rank. In this work, we establish a quantitative connection between these witnesses and stellar rank, demonstrating that the former can provide certifiable lower bounds on stellar rank. We introduce normalized expectation value and variance-based quantifiers and show that these witnesses form a consistent hierarchy of thresholds corresponding to stellar rank. Our results bridge the gap between abstract hierarchical measures and experimentally accessible quantifiers, enabling scalable certification of non-Gaussian states. I Introduction Bosonic modes of traveling light, described by continuous-variable (CV) quadrature operators, provide a robust, naturally scalable platform suitable for universal fault-tolerant quantum computation 23 ; 1 ; 39 and quantum metrology 50 ; 22 ; 49 ; 47 ; 26 . Achieving a quantum advantage in these applications requires quantum non-Gaussian states 45 , but alas, their reliable preparation remains one of the main experimental challenges of the platform 59 ; 67 ; 48 ; 46 ; 1 ; 39 . While the experimental techniques for preparation of such states are limited to various configurations of linear optics, squeezed states, and photon number resolving detectors 12 ; 19 ; 17 ; 43 ; 41 ; 66 ; 65 ; 46 ; 42 ; 54 ; 30 ; 56 ; 14 ; 36 ; 57 ; 15 ; 64 ; 60 ; 16 , theoretical approaches to observing their non-Gaussianity are more varied, utilizing state decompositions 25 , negative areas in Wigner functions 21 ; 29 ; 55 ; 58 , specific properties of photon number distributions 18 ; 2 ; 38 ; 10 ; 11 ; 20 , and other numerous features discernible from reconstructed density matrices. Among the many existing approaches to certifying non-Gaussian features, stellar rank 10 ; 11 , equivalent to genuine n -photonic quantum non-Gaussianity 38 , has emerged as a hierarchical and operationally motivated measure of non-Gaussianity, capturing the discrete structure of continuous-variable quantum states. Stellar rank of pure states is defined as the number of zeros of the Husimi Q-function. It can be interpreted operationally as the minimal number of photon additions required to generate the state from Gaussian resources 10 ; 11 ; 62 ; 20 . Its extension to mixed states is defined through a convex-roof construction, giving stellar rank a rigorous theoretical foundation but rendering it experimentally difficult to determine in practice. Quantum non-Gaussian features can also be detected using witnesses based on statistical moments of quadrature operators, relying on expectation values or variances of experimentally accessible operators. Typical examples include Gottesman-Kitaev-Preskill (GKP) state squeezing 44 , catability 6 , generalized quadrature nullifiers 33 , and non-linear squeezing 36 ; 46 ; 35 ; 4 ; 5 ; 31 ; 33 , with practical appeal of frequently requiring only a limited number of measurements, for example, of statistical moments of lower orders, often directly accessible on optical platforms without the need for full quantum state tomography. These criteria, introduced independently of one another, are motivated by state- or protocol-specific considerations and share the common goal of detecting statistical signatures that Gaussian states cannot reproduce, while only loosely connecting to the discrete hierarchy of stellar rank. In this work, we demonstrate that witnesses of non-Gaussian features, based on statistical moments, yield rigorous lower bounds on stellar rank, and conversely, that the stellar rank imposes constraints on their values. We further show that both expectation-value and variance-based criteria form a monotonic hierarchy consistent with stellar-rank ordering. The rest of this paper is organized into a pair of primary sections. The first section introduces the theoretical methods, whereas the numerical results and findings are discussed in the latter part of the paper. II Stellar rank and continuous witnesses A pure single-mode quantum state \ket{\psi_{m}} with stellar rank m can be expressed 20 as \ket{\psi_{m}}=\hat{G}\sum\limits_{n=0}^{m}c_{n}\ket{n}\,\text{,} (1) where {\lvert c_{m}\rvert>0} and \hat{G} denotes an arbitrary single-mode Gaussian unitary operation, which can be decomposed 9 into a sequence {\hat{G}\coloneqq\hat{F}(\vartheta)\hat{D}(z)\hat{S}(r)\hat{F}(\theta)} of elementary single-mode Gaussian operations \begin{gathered}\hat{F}(\vartheta)\coloneqq\exp(\imath\vartheta\hat{a}^{\dagger}\hat{a})\,\text{,}\quad\hat{D}(z)\coloneqq\exp(z\hat{a}^{\dagger}-\bar{z}\hat{a})\,\text{,}\\ \text{and}\quad\hat{S}(r)\coloneqq\exp\left(\frac{r}{2}(\hat{a}^{2}-\hat{a}^{\dagger 2})\right)\,\text{,}\end{gathered} (2) where \hat{a} and \hat{a}^{\dagger} represent the bosonic annihilation and creation operators. The general Gaussian operation \hat{G} is specified by five real parameters in total. A pair of real parameters forms the complex displacement amplitude z , the squeezing rate is given by r , and finally, \theta and \vartheta define the two phase shifts. A mixed state has stellar rank m if it can be expressed as a statistical mixture of pure states with stellar rank at most m , but not as a mixture of states with lower stellar rank 10 . Stellar rank naturally forms a hierarchy of quantum states. Let \mathscr{C}_{m} denote the set of all states with stellar rank at most m . These sets clearly satisfy {\mathscr{C}_{m}\subset\mathscr{C}_{n}} for every {m<n} . The set of Gaussian states \mathscr{C}_{0} lies at the bottom of this hierarchy. While stellar rank captures the discrete structure of a quantum state, witnesses based on the moments of directly measurable operators provide a complementary perspective. Let us first consider a witness based on the expectation value of an observable non-negative lower-bounded Hermitian operator \hat{W} and examine its minimal value achievable by quantum states with stellar rank m W_{m}=\inf\limits_{\hat{\varrho}_{m}\in\mathscr{C}_{m}}\trx(\hat{\varrho}_{m}\hat{W})\,\text{.} (3) The practical evaluation of ( 3 ) deserves further discussion. It depends linearly on \hat{\varrho}_{m} , making it sufficient to consider only pure states ( 1 ) in the minimization W_{m}=\inf\limits_{\hat{G},\ket{\psi_{m}}}\braket{\psi_{m}|\hat{G}^{\dagger}\hat{W}\hat{G}|\psi_{m}}\,\text{.} (4) This approach can be further simplified 20 . Instead of minimizing over all the pure states {\ket{\psi_{m}}\in\mathscr{C}_{m}} , it is enough to find the least eigenvalue W_{m}=\inf\limits_{\hat{G}}\left[\min\eig\left(\hat{\Pi}_{m}\hat{G}^{\dagger}\hat{W}\hat{G}\hat{\Pi}_{m}\right)\right] (5) of the transformed operator \hat{W} projected onto the finite-dimensional Hilbert space spanned by the first {(m+1)} Fock states using the projection operator \hat{\Pi}_{m}=\sum_{n=0}^{m}\ket{n}\!\bra{n}\,\text{.} (6) The optimization in ( 5 ) is performed over all the possible single-mode Gaussian operations \hat{G} , that is, at most over four real parameters instead of all five required to characterize the general Gaussian transformation. The right-most phase-shift \hat{F}(\theta) can be naturally absorbed into the eigen-decomposition as it maps the truncated Fock space onto itself, thus reducing the number of parameters by one. The required number of parameters could be further reduced due to symmetries within specific witness operators. In general, linear witnesses can be improved by introducing additional non-linear terms 24 . Let us now consider a class of witnesses based on variance, specifically on non-linear squeezing, which effectively quantifies the suppression of fluctuations in a general Hermitian operator \hat{Q} beyond the limits attainable by some well-defined class \mathscr{C} of reference states 5 . Non-linear squeezing offers experimental advantage as it can be observed directly in many relevant cases 46 ; 36 ; 4 ; 32 ; 35 ; 44 ; 5 ; 31 . It can be formally defined as \xi(\hat{\varrho})=\frac{\langle\{\hat{Q}-\langle\hat{Q}\rangle_{\hat{\varrho}}\}^{2}\rangle_{\hat{\varrho}}}{\inf\limits_{\hat{\mu}\in\mathscr{C}}\langle\{\hat{Q}-\langle\hat{Q}\rangle_{\hat{\mu}}\}^{2}\rangle_{\hat{\mu}}}\,\text{.} (7) When

Read the complete text