Quantum randomness beyond projective measurements
Fionnuala Curran · 2026
Read Quantum randomness beyond projective measurements by Fionnuala Curran (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-2605-18291.txt.
Introduction
Fionnuala Curran asks in a 2026 paper how much intrinsic randomness an extremal measurement can yield beyond the projective kind, a question left open for general cases. She characterizes the randomness from any unbiased extremal rank-one case on any state, solving it fully in dimension two. Four-outcome qubit cases of this type are tomographic, and she proves the maximum amount can be generated wherever a SIC measurement exists.
Catalog record
| CREATOR | Fionnuala Curran |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 135005 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.18291
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quantum randomness beyond projective measurements Fionnuala Curran Abstract The unpredictability of quantum physics gives rise to intrinsic randomness. In an adversarial scenario, any additional degrees of freedom must be attributed to an eavesdropper with correlations to the measurement set-up. The true randomness is then quantified by the probability that she correctly guesses the measurement outcomes, optimised over all possible strategies. Extremal measurements are appealing here, since they do not allow information to leak to such an eavesdropper. Beyond projective measurements, however, a simple question remains open: how much intrinsic randomness can be generated by a given extremal measurement? In a step towards solving it, we characterise the randomness generated by any unbiased extremal rank-one measurement acting on any state, solving the problem explicitly in dimension two. Four-outcome qubit measurements of this type are tomographic, so these results hold for fully source -device-dependent randomness too. The tetrahedral symmetric informationally complete (SIC) measurement, we find, has the least intrinsic randomness within this class. We also present the skewed SIC family of measurements, and use them to partially solve an open problem: we prove that 2 log d 2\log d bits of randomness, the maximal amount, can be generated device-dependently (or source-device-independently) in any dimension in which there exists a SIC measurement. 1 Introduction Quantum measurements generate an intrinsic form of randomness with no counterpart in classical physics. That the outcomes of a well-chosen measurement on a quantum system are unpredictable, even to an eavesdropper with perfect knowledge of the pre-measurement set-up, is exploited in the design of quantum random number generators [ 31 , 27 , 32 ] . Some measurements are more random than others, however. Of particular interest are the extremal measurements, which cannot be decomposed into a probabilistic mixture of other measurements [ 21 ] , and are therefore secure against quantum adversaries [ 18 , 43 ] . A question of practical, as well as theoretical, importance then arises: how much randomness can be generated using an extremal measurement, and which measurements are the most random? When an extremal measurement acts on a generic (mixed) quantum state, we can quantify the resulting randomness by the probability with which a correlated eavesdropper correctly guesses the outcomes [ 30 ] . The eavesdropper may have quantum correlations, such that she holds a purification of the state, or, equivalently, classical correlations, such that she knows which state from a probabilistic mixture is actually prepared in a given measurement round. The guessing probability relates directly to the conditional min-entropy [ 28 ] , which lower bounds the number of (almost) perfectly random bits that can be extracted from the measurement outcomes 1 1 1 We use logarithms of base two throughout. [ 41 , 29 ] . Recent works have studied the maximal randomness that can be produced from any state using projective [ 34 ] or general [ 2 ] measurements, as well as the maximal intrinsic randomness of several noisy (non-extremal) measurements [ 17 ] . At the limits of randomness generation, it has been shown [ 2 ] that no more than 2\log d bits of conditional min-entropy can be generated in dimension d , and that this bound can be approached arbitrarily closely. We begin by introducing the class of skewed SIC measurements, which are derived from symmetric informationally complete (SIC) measurements. These measurements are biased, in that they produce a non-uniform probability distribution from the maximally mixed ‘junk’ state. While this bias makes them poor candidates for quantum state tomography, it allows them to produce a uniform distribution from a pure state instead. We thereby prove that one can generate maximal randomness from any pure or isotropically noisy state in any dimension in which there exists a SIC measurement, partially answering an open question in [ 2 ] . Since the skewed SIC measurements are informationally complete, this result holds in a source-device-independent setting as well. These measurements can be seen as higher-order analogues of a two-dimensional class that was previously shown to generate maximal device-independent randomness [ 47 ] , an area in which skewed SIC measurements may also find application. Turning to the more familiar unbiased extremal measurements, we study the randomness they generate from any given state. This, we find, is proportional to the maximal quantum fidelity of the chosen state with any state that lives in the convex hull of the measurement projectors. Given a measurement with n outcomes, by choosing the worst possible state, we derive a lower bound of \log\frac{n}{d} bits of conditional min-entropy, which is relevant for source-device-independent randomness generation [ 6 ] . Notably, in the extreme case where n=d^{2} , we can always reach at least \log d bits of randomness, which is the maximum that can be achieved using a projective measurement. In pursuit of maximal randomness, we also introduce the one-parameter ‘scissors’ family of unbiased measurements in dimensions two, three and four, which all come arbitrarily close to achieving 2\log d bits of randomness. We then explicitly solve the maximal intrinsic randomness of all unbiased extremal measurements in dimension two, as well as the randomness they generate from any state. On the way, we prove that all four-outcome measurements of this type are group covariant, and present a characterisation in terms of two angles, a result which may be of independent interest. Leveraging the tomographic nature of these measurements, one could adapt our results to randomness generation in a fully source-device-independent setting. A final result shows that the task of randomness generation can itself give an insight into the landscape of quantum measurements: among all four-outcome unbiased extremal qubit measurements, the SIC measurement, which is optimal for quantum state tomography [ 42 ] , is singled out as having the least intrinsic randomness. This paper is laid out as follows. In Section 2 , we set up our problem and give the necessary preliminaries. In Section 3 , we present the family of skewed SIC measurements. In Section 4 , we study unbiased extremal rank-one measurements in any dimension, while in Section 5 , we give our results on unbiased extremal qubit measurements. Finally, in Section 6 , we derive the scissors family of measurements in dimensions two, three and four. We present our conclusions and outlook in Section 7 . 2 Set-up We work throughout in a Hilbert space of finite dimension d . A quantum system is represented by a state \rho , a positive semidefinite matrix of unit trace, while a measurement of n outcomes is represented by a positive operator-valued measure (POVM) \m=\{M_{j}\}_{j} , where the positive semidefinite elements \{M_{j}\} represent the outcomes. The probability of obtaining the outcome j is given by the Born rule, p_{j}=\tr\left(\rho M_{j}\right) . To ensure that the probabilities are non-negative and normalised for any state, the POVM elements must also satisfy \sum_{j=1}^{n}M_{j}=\mathbbm{1} . We call a POVM unbiased if all of its elements have equal trace, and biased otherwise 2 2 2 Note that the ‘unbiased’ quality of a POVM has nothing to do with unbiased bases. . Projective measurements are POVMs whose elements satisfy M_{i}M_{j}=M_{i}\delta_{i,\,j} for all i,\,j\in\{1,\,\ldots,\,n\} ; any POVM can be realised, however, by applying a joint projective measurement to the