Against probability: A quantum state is more than a list of probability distributions

Ladina Hausmann and Renato Renner · 2026

Read Against probability: A quantum state is more than a list of probability distributions by Ladina Hausmann and Renato Renner (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-2601-18872.txt.

Introduction

Ladina Hausmann and Renato Renner argue that a quantum state holds more than a list of chances. A state can be written as the set of outcome chances for a chosen group of measures. Such lists show up across physics, from field theory to broad chance based frames.

The pair press on what such a list leaves out. If two states give the same chances for the chosen measures, are they the same state? Their title gives the stance: no, the state is more than its chance lists. The paper probes what extra facts a full state carries past any one set of measures.

Catalog record

CREATORLadina Hausmann and Renato Renner
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE117 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-18872

Browse the full library · About the Tower of Babel

Full text

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

Against probability: A quantum state is more than a list of probability distributions Ladina Hausmann and Renato Renner Abstract The state ρ \rho of a quantum system can be represented by a vector 𝐏 ℳ ​ ( ρ ) \mathbf{P}_{\mathcal{M}}(\rho) of outcome probabilities for a set of measurements ℳ \mathcal{M} . Such representations appear throughout physics, for example, in quantum field theory via correlation functions and in quantum foundations within generalized probabilistic frameworks. In this work, we identify an unavoidable tension: to enable operationally meaningful statements, the map ρ ↦ 𝐏 ℳ ​ ( ρ ) {\rho\mapsto\mathbf{P}_{\mathcal{M}}(\rho)} must be topologically robust—preserving the notion of closeness between states. Yet, a probability representation that is topologically robust cannot simultaneously retain other essential structure, such as the subsystem structure. 1 Introduction and a first example Let P_{M}(\rho) denote the probability distribution obtained by measuring a quantum state \rho\in\mathcal{D}(\mathcal{H}) on a separable Hilbert space \mathcal{H} . If one collects these probability distributions for all measurements M from a tomographically complete set \mathcal{M} , the resulting tuple \mathbf{P}_{\mathcal{M}}(\rho)\coloneq\bigl(P_{M}(\rho)\bigr)_{M\in\mathcal{M}} uniquely specifies the quantum state \rho . Such probablity representations arise, for instance, when considering correlation functions in quantum field theories [ 1 ] , and are also widespread in quantum information theory and quantum foundations [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 14 , 15 , 16 , 17 ] . In this work, we ask the following question: does \mathbf{P}_{\mathcal{M}}(\rho) represent \rho faithfully? If faithfulness meant injectivity, the answer would be yes, as this is precisely the notion of tomographic completeness. However, this criterion is too weak. To be faithful in a physically meaningful sense, a representation must also be topologically robust . A main contribution of this paper is to define and explain what this means. We begin with an example where the use of probability representations appears natural: generating random bits R_{1},\ldots,R_{\ell} from a quantum process. For concreteness, imagine a protocol that produces each bit R_{i} as follows: prepare n unstable atoms and count the number of decays within a fixed time interval (e.g., one second). Set R_{i}=0 if this number is even and R_{i}=1 otherwise. An \ell -bit string R=R_{1}\cdots R_{\ell} is said to be random if it cannot be predicted with probability greater than 2^{-\ell} , even by an all-powerful agent with access to any side information E available prior to the process, such as the internal state of the source supplying the atoms. Formally, this means that the joint state \rho=\rho_{RE} of R and E belongs to the set \Sigma^{\mathrm{rand}}\coloneq\left\{\bar{\Pi}^{(\ell)}_{R}\otimes\sigma_{E}\Big|\ell\in\mathbb{N},\sigma_{E}\in\mathcal{D}(E)\right\} (1.1) where \bar{\Pi}^{(\ell)}_{R} is the uniform state on bit strings of length \ell , and \mathcal{D}(E) is the set of density operators on E . In any practical randomness-generation protocol, the output R inevitably has residual correlations with other systems. Moreover, since R is produced by a quantum process, it is not automatically classical: in general, the joint state \rho_{RE} is not a cq-state. Consequently, condition ( 1.1 ) cannot be achieved exactly, but only approached asymptotically by investing additional resources [ 18 , 19 , 20 ] , for instance by increasing the number n of atoms in our example protocol. For this reason, it is standard to adopt an approximate condition of the form \lim_{n\to\infty}\delta(\rho^{(n)},\Sigma^{\mathrm{rand}})=0, (1.2) where \delta(\rho^{(n)},\Sigma^{\mathrm{rand}})\coloneq\inf_{\sigma\in\Sigma^{\mathrm{rand}}}\delta(\rho^{(n)},\sigma) , with \delta(\cdot,\cdot) denoting the trace distance. Operationally, this quantifies the maximal probability with which an agent with access to both R and E can distinguish \rho^{(n)} from the ideal behaviour defined by \Sigma^{\mathrm{rand}} [ 21 , 22 ] . To work at the level of probability representations, one requires that such approximate conditions be expressible via a distance measure defined directly on probability distributions, such as the statistical distance. The non-triviality of this requirement is best illustrated by a concrete example: a sequence of states \rho^{(n)} that violates Eq. 1.2 , while this violation remains invisible in the corresponding probability representation. Example . For any n\in\mathbb{N} , let R and E be 2^{n} -dimensional systems, and the joint state of R and E after the randomness generation protocol be \displaystyle\rho^{(n)}_{RE}\propto\sum_{1\leq u<v\leq 2^{n}}\pi_{u,v} (1.3) where \pi_{u,v} denotes the projector on the subspace of RE spanned by {\ket{u}_{R}\ket{v}_{E}-\ket{v}_{R}\ket{u}_{E}} , for an arbitrary choice of orthonormal bases. This state is entangled and thus distinct from the states in \Sigma^{\mathrm{rand}} , which are separable. Concretely, as shown in [ 23 , Example II.9.] , \forall n\in\mathbb{N}:\delta(\rho^{(n)},\Sigma^{\mathrm{rand}})\geq\frac{1}{4}. (1.4) Therefore, R is not approximately random according to criterion ( 1.2 ). Before resuming this example at the level of the probability representation \mathbf{P}_{\mathcal{M}} , we must clarify the requirements on the underlying set \mathcal{M} of measurements. While one could, in principle, allow arbitrary measurements on the joint system RE , these generally fail to respect the subsystem structure, thus erasing the distinction between R and E . Yet, this distinction is precisely what allows us to phrase definitions like ( 1.1 ). Accordingly, it is common to restrict \mathcal{M} to measurements acting locally on R and E . \rho^{(1)} \rho^{(2)} \rho^{(3)} \dots \Sigma \dots (a) \Sigma \dots \rho^{(1)} \rho^{(2)} \rho^{(3)} \dots (b) Figure 1 : Non-robustness of \mathbf{P}_{\mathcal{M}_{\otimes}} . A region of the quantum state space \mathcal{D}(\mathcal{H}) is shown, containing the sequence of states \rho^{(n)} from the example (black) and the corresponding closest points in \Sigma=\Sigma^{\mathrm{rand}} (blue). Panel (a) uses the trace distance \delta , for which the distance to \Sigma stays constant. Panel (b) uses the metric d_{\mathcal{M}_{\otimes}} induced by the product-measurement representation \mathbf{P}_{\mathcal{M}_{\otimes}} , for which the distance to \mathbf{P}_{\mathcal{M}_{\otimes}}(\Sigma) shrinks with increasing n . Example (continued) . For illustrative purposes, we focus on measurements of the form M\otimes M where M consists of rank- 1 projectors only. 1 1 1 The conclusions of this example can be easily generalized to any measurement acting locally on R and E . For this, it suffices to observe that the state on E conditioned on any outcome of a rank- 1 measurement applied to R is maximally mixed on a subspace of dimension 2^{n}-1 , which is 2^{-n} -close to \bar{\Pi}^{(n)}_{E} . Exploiting the antisymmetry of \rho^{(n)}_{RE} , it is straightforward to verify that the joint probability distribution P_{M\otimes M}=P_{M\otimes M}(\rho^{(n)}_{RE}) of the outcomes X and Y is given by P_{M\otimes M}(x,y)=\begin{cases}\quad 0&\text{if }x=y\\ \frac{1}{2^{n}(2^{n}-1)}&\text{else}.\end{cases} (1.5) From this, one finds that \frac{1}{2}\|P_{M\otimes M}(\rho^{(n)}_{RE})-P^{(n)}\times P^{(n)}\|_{1}\leq 2^{-n} , where P^{(n)} is the uniform distribution on an alphabet of size 2^{n} . Consequently, at the level of such probability representations, \rho^{(n)}_{RE} is approximately indistinguishable from \Sigma^{\mathrm{rand}} ; see also Fig. 1 . The result of the example means that, with respect to the metric defined by \displaystyle d_{\mathcal{M}}(\rho,\sigma)\coloneq\frac{1}{2}\sup_{M\in\mathcal{M}}\bigl\|P_{M}(\rho)-P_{M}(\sigma)\bigr\|_{1}\ , (1.6) when we choose \mathcal{M}=\mathcal{M}_{\otimes} to be the set of local measurements, the sequence (\rho_{RE}^{(n)})_{n\in\mathbb{N}} converges to \Sigma^{\mathrm{rand}} . Yet the same is not true for the metric \delta . This motivates the following definition. (Throughout, we assume that \mathcal{M} is a tomographically complete measurement set, i.e., the map \mathbf{P}_{\mathcal{M}} from the space of density matrices \mathcal{D}(\mathcal{H}) to the representation space is injective.) Definition 1 . We say that \mathbf{P}_{\mathcal{M}} is topologically robust (or simply robust ) if for all subsets \Sigma and sequences (\rho^{(n)})_{n\in\mathbb{N}} of states \lim_{n\to\infty}d_{\mathcal{M}}(\rho^{(n)},\Sigma)=0\ \implies\

Read the complete text