Information Theoretic Foundations for Physics

One of the most deeply rooted concepts in science is causality: the idea that events in the present are caused by events in the past and, in turn, act as causes for what happens in the future. If an event A is a cause of an effect B, then B cannot be a cause of A. Recently we proposed a framework that assumes that operations in local laboratories are described by quantum mechanics, but where no reference is made to any global causal relations between these operations. The central notion of the formalism is “process” which is a generalization of the causal notion of “quantum state”. The framework allows for processes in which two operations are neither causally ordered nor in a probabilistic mixture of definite causal orders, i.e. one cannot say that A is before or after B. However a physical interpretation of such correlations was lacking. I will show that the “superposition of quantum circuits” – in which the gate ordering is not fixed but controlled by a quantum system – is an example of a process with indefinite causal order. This process provides a reduction in query complexity of certain computational problems and can be realized by placing the laboratories in the gravitational field of a massive object in a spatial superposition.

In QBism, a quantum state represents an agent's personal degrees of belief regarding the consequences of her actions on any part of her external world. The quantum formalism provides consistency criteria that enable the agent to make better decisions. QBism thus gives a central role to the agent, or user of the theory, and explicitly rejects the ontological model framework introduced by Harrigan and Spekkens. This talk addresses the status of agents and the notion of locality in QBism. Our definition of locality is independent of the assumption of an ontological model. Instead it depends on an appropriate formalization of the idea of a causal structure.

In the classical world of Newton and Laplace, fundamental physics and thermodynamics do not blend well: the former puts forward a picture of nature where states are pure and processes are fundamentally reversible, while the latter deals with scenarios where states are mixed and processes are irreversible. Many attempts have been made at reconciling the two paradigms, but ultimately the source of all troubles remains: if every particle possesses a definite position and a definite velocity, why should experimental data depend on the expectations of agents who have only partial information? Quantum theory offers a way out. Thanks to entanglement, a system and its environment can be jointly in a pure state, while the system alone is in a mixed state. Similarly, the evolution of system and environment can be jointly reversible, but locally irreversible. At the level of axioms, these facts are captured by the Purification Principle [1], from which the whole of quantum theory was derived in Ref. [2]. In this talk I will report on a new research programme, aimed at establishing the Purification Principle as the conceptual foundation for thermodynamics. I will start from the idea that thermodynamical transformations can be mapped into transformations that degrade entanglement. This will lead to a duality between entanglement and thermodynamics, which will be used to define measures of entanglement/entropy and to explore operational tasks like information erasure. At this level, thermodynamical requirements lead directly to requirements on the particular type of Purifications allowed by the theory: for example, forbidding the existence of Entropy Sinks (systems that can absorb entropy from the environment without increasing their internal entropy) is equivalent to imposing the existence of Symmetric Purifications (purifications in which the environment is a mirror image of the system). I will conclude by giving a glimpse on quantitative measures of entanglement/entropy. By adding a requirement about the existence of sharp measurements, I will show that Purification leads directly the an operational version of the spectral theorem, allowing one to define a zoo of well-behaved entropies. [1] G. Chiribella, G. M. D’Ariano, and P. Perinotti, Probabilistic Theories with Purification, Phys. Rev. A 81, 062348 (2010) [2] G. Chiribella, G. M. D’Ariano, and P. Perinotti, Informational Derivation of Quantum Theory, Phys. Rev. A 84, 012311 (2011)

Much progress has recently been made on the fine-grained thermodynamics and statistical mechanics of microscopic physical systems, by conceiving of thermodynamics as a resource theory: one which governs which transitions between states are possible using specified "thermodynamic" (e.g. adiabatic or isothermal) means. In this talk we lay some groundwork for investigating thermodynamics in generalized probabilistic theories. We describe simple, but fairly strong, postulates: unique spectrality, projectivity, and symmetry of transition probabilities, that imply that a system has a well-behaved convex analogue of the spectrum, and show that the spectrum of a state majorizes the outcome probabilities of any fine-grained measurement, allowing the operationally defined measurement entropy (and Schur-concave analogues) to be calculated from the spectrum. These are implied by, but probably weaker than, Axioms 1 (weak spectrality) and 2 (strong symmetry) of a recent characterization of Jordan-algebraic and quantum systems by Barnum, Mueller, and Ududec. It is an open question whether theories beyond the Jordan-algebraic ones satisfy them. We describe how part of von Neumann's argument that spectral entropy is a good candidate for thermodynamic entropy generalizes to systems satisfying our postulates, and discuss whether its assumptions are reasonable there, suggesting that the extendibility of certain processes to reversible ones is crucial. We will discuss further postulates and results that might suffice to obtain, in this more general setting, a thermodynamical resource theory similar to the one that is emerging for quantum theory. (Joint work with Jon Barrett, Marius Krumm, Matt Leifer, Markus Mueller.)

How should we describe the thermodynamics of extreme quantum regimes, where features such as coherence and entanglement dominate? I will discuss possible limitations of a traditional statistical mechanics approach, and then describe work that applies modern techniques from the theory of quantum information to the foundations of thermodynamics. In particular I discuss recent progress in quantum resource theories and argue that to properly encapsulate the thermodynamic structure of quantum coherence and entanglement we must make use of concepts beyond free energies.

I will argue that, apart from their ever growing number of applications to physics, information theoretic concepts also offer a novel perspective on the physical content and architecture of quantum theory and spacetime. As a concrete example, I will discuss how one can derive and understand the formalism of qubit quantum theory by focusing only on what an observer can say about a system and imposing a few simple rules on the observer’s acquisition of information.

It has become conventional wisdom to say that quantum theory and gravitational physics are conceptually so different, if not incompatible, that it is very hard to unify them. However, in the talk I will argue that the operational view of (quantum) information theory adds a very different twist to this picture: quite on the contrary, quantum theory and space-time are highly fine-tuned to fit to each other. After a recap of ideas by von Weizsacker, Wootters, and Popescu and Rohrlich, I will show how uncertainty relations, the number of degrees of freedom of the Bloch ball, and the existence of entangled states and possibly the Tsirelson bound can be understood from space-time geometry alone. Conversely, I will show how the 3+1 Lorentz group of space-time can be derived from a purely informational communication scenario of two observers that describe local quantum physics in different Hilbert space bases (joint work with Philipp Hoehn).

I will discuss my work (in progress) to formulate General Relativity as an operational theory which includes probabilities and also agency (knob settings). The first step is to find a way to discuss operational elements of GR. For this I adapt an approach due to Westman and Sonego. I assert that all directly observable quantities correspond to coincidences in the values of scalar fields. Next we need to include agency. Usually GR is regarded as a theory in which a solution is simply stated for all space and time (the Block Universe view). Here, instead, we find a way to treat agents as making choices. Finally, we need to incorporate probabilities. For this purpose we take a compositional point of view. We associate a generalized state with regions of the space consisting of coincidences in the values of scalars. We then show how to combine these generalized states to make probabilistic predictions using the duotensor machinery developed previously.

To make precise the sense in which nature fails to respect classical physics, one requires a formal notion of "classicality". Ideally, such a notion should be defined operationally, so that it can be subjected to a direct experimental test, and it should be applicable in a wide variety of experimental scenarios, so that it can cover the breadth of phenomena that are thought to defy classical understanding. Bell's notion of local causality fulfills the first criterion but not the second, because it is restricted to scenarios with two or more systems that are space-like separated. The notion of noncontextuality fufills the second criterion, because it is applicable to any experiment (even those on a single system), but it is a long-standing question whether it can be made to fulfill the first. Previous attempts to experimentally test noncontextuality have all presumed certain idealizations that do not hold in real experiments, namely, noiseless measurements and exact operational equivalences. In this talk, I will describe how one can devise experimental tests that are free of these idealizations using an operational notion of noncontextuality that applies to both preparations and measurements. These new theoretical insights raise the bar significantly for any claim of an experimental demonstration of nonclassicality. They also provide the means of determining, for any phenomenon that is typically thought to defy classical explanation, which experimentally-testable features of that phenomenon, if any, conflict with the assumption of a noncontextual model.

Quantum mechanics is derived from the principle that the universe contain as much variety as possible, in the sense of maximizing the distinctiveness of each subsystem. This is an expression of Leibniz's principles of sufficient reason and the identity of the indiscernible. The quantum state of a microscopic system is defined to correspond to an ensemble of subsystems of the universe with identical constituents and similar preparations and environments. A new kind of interaction is posited amongst such similar subsystems which acts to increase their distinctiveness, by extremizing the variety. In the limit of large numbers of similar subsystems this interaction is shown to give rise to Bohm's quantum potential. As a result the probability distribution for the ensemble is governed by the Schroedinger equation. The measurement problem is naturally and simply solved. Microscopic systems appear statistical because they are members of large ensembles of similar systems which interact non-locally. Macroscopic systems are unique, and are not members of any ensembles of similar systems. Consequently their collective coordinates may evolve deterministically. This proposal could be tested by constructing quantum devices from entangled states of a modest number of quits which, by its combinatorial complexity, can be expected to have no natural copies.