Random measurements are almost maximally incompatible

Andreas Bluhm, Cécilia Lancien, Ion Nechita · 2025

Read Random measurements are almost maximally incompatible by Andreas Bluhm, Cécilia Lancien, Ion Nechita (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-2507-20600.txt.

Introduction

Andreas Bluhm, Cécilia Lancien, and Ion Nechita study random quantum measurements in a 2025 paper. Some quantum measurements cannot be performed together, a property called incompatibility. Past work mostly asked how much noise a fixed set can bear before becoming compatible. This paper asks a different question: how incompatible are typical, randomly chosen measurements?

For two natural random classes, they show the measurements come close to the maximum possible incompatibility, under the right settings. Their proofs use witness tests for incompatibility together with tools from random matrices and free probability. The result frames incompatibility as the generic case, not an exception.

Catalog record

CREATORAndreas Bluhm, Cécilia Lancien, Ion Nechita
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE106 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-20600

Browse the full library · About the Tower of Babel

Full text

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

Random measurements are almost maximally incompatible Andreas Bluhm, Cécilia Lancien, Ion Nechita Abstract . In this work, we investigate the incompatibility of random quantum measurements. Most previous work has focused on characterizing the maximal amount of white noise that any fixed number of incompatible measurements with a fixed number of outcomes in a fixed dimension can tolerate before becoming compatible. This can be used to quantify the maximal amount of incompatibility available in such systems. The present article investigates the incompatibility of several classes of random measurements, i.e., the generic amount of incompatibility available. In particular, we show that for an appropriate choice of parameters, both random dichotomic projective measurements and random basis measurements are close to being maximally incompatible. We use the technique of incompatibility witnesses to certify incompatibility and combine it with tools from random matrices and free probability. 1. Introduction and main results 1.1. Introduction One of the prime examples of how quantum mechanics differs from classical mechanics is the existence of incompatible measurements. A set of measurements is incompatible if the outcomes of the measurements do not arise as classical post-processings of some joint measurement. Equivalently, measurements are incompatible if they do not arise as marginals from a joint measurement [ HMZ16 ] . In this sense, they cannot be measured simultaneously [ Hei27 , Boh28 ] . The position and momentum measurement are arguably the best-known examples of incompatible measurements, but interesting situations also occur in finite dimensions [ HMZ16 , GHK+23 ] . For example, two projective measurements are compatible if and only if their elements commute. The existence of incompatible measurements is crucial for the violation of Bell inequalities [ Fin82 ] and therefore enables many quantum information processing protocols [ BCP+14 ] , for example device-independent quantum key distribution [ PAB+09 ] . Therefore, incompatibility plays a similarly fundamental role for quantum technologies as entanglement, and can hence be seen as a resource for quantum information tasks [ HKR15 ] . It is well known that entanglement vanishes given enough noise, which is part of what makes building quantum computers very challenging. The same is true for measurement incompatibility [ BHSS13 ] . It is therefore important to quantify how much noise a given set of measurements can tolerate before the measurements become compatible. Most of the work so far has focused either on concrete sets of measurements [ GHK+23 ] , for example qubit measurements [ Bus86 , BH08 , GU24 ] or mutually unbiased bases [ CHT12 , DSFB19 ] , or on the maximal noise robustness of incompatibility for a given set of physical parameters, such as the dimension of the quantum system, the number of measurements, or the number of outcomes. The latter question has been investigated in [ BHSS13 , Gud13 ] and recently also in [ BN18 , BN20 , BJN22 , BN22b ] , where some of the present authors discovered a connection to matrix convex sets and tensor norms. Less explored is the question of quantifying the incompatibility of generic or typical quantum measurements, i.e., understanding how incompatible random quantum measurements are. Different models of random measurements have been defined in [ HJN20 ] , starting from random completely positive maps [ CN16 ] . In that paper, the authors in particular compared different compatibility criteria for random measurements. The article [ ZXLJF19 ] calculated the incompatibility of pairs of random qubit measurements. The present article continues this line of work by considering special classes of random measurements, in particular random dichotomic projective measurements and random basis measurements. Our leading question is to determine whether such measurements are close to being maximally incompatible. We use the technique of incompatibility witnesses to certify incompatibility and combine it with tools from random matrices and free probability. Analogously to entanglement witnesses, incompatibility witnesses are hyperplanes that separate certain incompatible measurements from the convex set of compatible measurements. 1.2. Main results To quantify the amount of incompatibility between measurements, we use the minimal compatibility degree \tau(d,g,(k_{1},\ldots,k_{g})) , i.e., 1-\tau(d,g,(k_{1},\ldots,k_{g})) is the minimal amount of white noise needed to make any set of g measurements in dimension d with k_{x} outcomes for the x -th measurement compatible (see Definition 2.4 ). For the compatibility degree of a fixed set of measurements \{\mathrm{M}_{1},\ldots,\mathrm{M}_{g}\} , we write \tau(\mathrm{M}_{1},\ldots,\mathrm{M}_{g}) (see Definition 2.3 ). Our first results are about dichotomic measurements (i.e., k_{x}=2 for all x ). For those the minimal compatibility degree can be computed exactly, at least in the regime where the dimension is exponentially larger than the number of measurements. Concretely, it has been shown in [ BN18 ] that (1) \tau(d,g,(2,\ldots,2))\geq\frac{1}{\sqrt{g}}, with equality for d=\Omega(2^{g}) . We first look at the case where g=2 . We consider E,F\subset\mathbb{C}^{d} to be independent uniformly distributed random subspaces of respective dimensions \alpha d,\beta d , for some fixed parameters 0<\alpha,\beta<1 . We denote by P_{E} and P_{F} the orthogonal projections onto these subspaces and by \mathrm{P}_{E}=(P_{E},I-P_{E}) and \mathrm{P}_{F}=(P_{F},I-P_{F}) the dichotomic measurements defined by these projections, where I is the identity matrix. We find that, for \alpha and \beta close enough to 1/2 , these measurements are close to being maximally incompatible. The following is an informal restatement of Corollary 3.6 : Theorem . Suppose that (\alpha-\frac{1}{2})^{2}+(\beta-\frac{1}{2})^{2}\leq\frac{1}{4} . Then, \tau(\mathrm{P}_{E},\mathrm{P}_{F})\xrightarrow[d\to\infty]{}\frac{1}{\sqrt{2}}. For \alpha and \beta not close enough to 1/2 , the behavior changes and the measurements become less and less incompatible. For instance, the following is an informal restatement of Corollaries 3.8 and 3.14 : Theorem . Suppose that \beta=\alpha and that either \alpha>\frac{1}{2}(1+\frac{1}{\sqrt{2}}) or \alpha<\frac{1}{2}(1-\frac{1}{\sqrt{2}}) . Then, setting \lambda_{\alpha}=4\alpha(1-\alpha) , \tau(\mathrm{P}_{E},\mathrm{P}_{F})\xrightarrow[d\to\infty]{}\frac{1}{\sqrt{\lambda_{\alpha}}+\sqrt{1-\lambda_{\alpha}}}. Turning to the case where g>2 , we consider g independent uniformly distributed random projections P_{x} of rank d/2 on \mathbb{C}^{d} . We write \mathrm{P}_{x}=(P_{x},I-P_{x}) for the associated dichotomic measurements. Then, we find that for g large (but scaling at most linearly with d ), this set of measurements is close to being maximally incompatible. The following is a consequence of Theorem 4.5 : Theorem . Let g\leq 3d . Then, with probability at least 1-e^{-2d} , \tau(\mathrm{P}_{1},\ldots,\mathrm{P}_{g})\leq\frac{31}{\sqrt{g}}. Combined with the lower bound recalled in Eq. ( 1 ), this result implies that (2) \tau(d,g,(2,\ldots,2))=\Theta\left(\frac{1}{\sqrt{g}}\right)\qquad\forall d\geq\frac{g}{3}, while previously we only knew this to hold for d\geq 2^{\lceil(g-1)/2\rceil} . Our next results concern a set of g independent uniformly distributed random basis measurements \mathrm{B}_{x} in dimension d . As these measurements are not dichotomic, we are no longer comparing the value of \tau(\mathrm{B}_{1},\ldots,\mathrm{B}_{g}) to 1/\sqrt{g} . In fact, little is known about the compatibility degree of measurements which are not dichotomic. In terms of lower bounds, there is the universal lower bound that we can obtain from approximate cloning [ Wer98 ] , which in fact holds for any g basis measurements with d outcomes in dimension d , namely: (3) \tau(\mathrm{B}_{1},\ldots,\mathrm{B}_{g})\geq\frac{g+d}{g(d+1)}. For g=2 , it is in fact easy to show that two independent random basis measurements are asymptotically maximally incompatible. The following is a consequence of Corollary 5.2 : Theorem . It holds that \tau(\mathrm{B}_{1},\mathrm{B}_{2})\xrightarrow[d\to\infty]{}\frac{1}{2} . In fact, we find that random basis measurements are also close to being maximally incompatible for g large enough. The following is a consequence of Theorem 5.5 : Theorem . Let g\geq 2d .

Read the complete text