nLab state on a star-algebra

Contents

This entry describes a concrete formalization of the general notion of state in the context of quantum probability theory and algebraic quantum field theory and operator algebra. For other conceptualizations of states see there.

Context

Measure and probability theory

measure theory

probability theory

Contents

Idea

The concept of state on a star-algebra is the formalization of the general idea of states from the point of view of quantum probability theory and algebraic quantum theory.

In order to motivate the definition from more traditional formulations in physics, recall that there a state $\langle - \rangle$ is the information that allows to assign to each observable $A$ the expectation value $\langle A\rangle$ that this observable has when the physical system is assumed to be in that state.

Often this is formalized in the Schrödinger picture where a Hilbert space of states $\mathcal{H}$ is taken as primary, and the observables are represented as suitable linear operators $A$ on $\mathcal{H}$. Then for $\psi \in \mathcal{H}$ a state (pure state) the expectation value of $A$ in this state is the inner product $\langle \psi \vert A \vert \psi \rangle \coloneqq (\psi, A \psi)$. This defines a linear function

$\langle \psi \vert (-) \vert \psi \rangle \;\colon\; \mathcal{A} \longrightarrow \mathbb{C}$

on the algebra of observables $\mathcal{A}$, satisfying some extra properties.

Conversely, in the Heisenberg picture one may take the “abstract” algebra of observables as primary (i.e. not necessarily manifested as an operator algebra), and declare that a state is any linear functional

$\langle - \rangle \;\colon\; \mathcal{A} \longrightarrow \mathbb{C}$

which is positive in that $\langle A^\ast A\rangle \geq 0$ and normalized in that $\langle 1\rangle = 1$. Under suitable conditions a Hilbert space of states may be (re-)constructed from this a posteriori via the GNS construction.

Traditionally this definition is considered for algebras of observables which are C*-algebras (as usually required for non-perturbative quantum field theory, see e.g. Fredenhagen 03, section 2), but the definition makes sense generally for plain star-algebras (Meyer 95, I.1.1), such as for instance for the formal power series algebras that appear in perturbative quantum field theory (e.g. Bordemann-Waldmann 96, def. 1, Fredenhagen-Rejzner 12, def. 2.4, Khavkine-Moretti 15, def. 6, Dütsch 18, def. 2.11).

The perspective that states are normalized positive linear functionals on the algebra of observables is implicit in traditional perturbative quantum field theory, where it is encoded in the 2-point function corresponding to a vacuum state or more generally a quasi-free quantum state (the Hadamard propagator). The perspective is made explicit in algebraic quantum field theory (see e.g. Fredenhagen 03, section 2) and for star-algebras of observables that are not necessarily C*-algebras in perturbative algebraic quantum field theory (e.g. Bordemann-Waldmann 96, Fredenhagen-Rejzner 12, def. 2.4, Khavkine-Moretti 15, def. 6, Dütsch 18, def. 2.11).

Definition

For unital star algebras

Definition

(state on a unital star algebra)

Let $\mathcal{A}$ be a unital star-algebra over the complex numbers $\mathbb{C}$. A state on $\mathcal{A}$ is a linear function

$\rho \;\colon\; \mathcal{A} \longrightarrow \mathbb{C}$

such that

1. (positivity) for all $A \in \mathcal{A}$ the value of $\rho$ on the product $A^\ast A$ is

1. real$\,\rho(A^\ast A) \in \mathbb{R} \hookrightarrow \mathbb{C}$

2. as such non-negative:

(1)$\rho(A^\ast A) \geq 0$
2. (normalization)

(2)$\rho(\mathbf{1}) = 1$

for $\mathbf{1} \in \mathcal{A}$ the unit in the algebra.

Remark

(probability theoretic interpretation of state on a star-algebra)

A star algebra $\mathcal{A}$ equipped with a state is also called a quantum probability space, at least when $\mathcal{A}$ is in fact a von Neumann algebra.

Remark

(states form a convex set)

For $\mathcal{A}$ a unital star-algebra, the set of states on $\mathcal{A}$ according to def. is naturally a convex set: For $\rho_1, \rho_2 \colon \mathcal{A} \to \mathbb{C}$ two states then for every $p \in [0,1] \subset \mathbb{R}$ also the linear combination

$\array{ \mathcal{A} &\overset{p \rho_1 + (1-p) \rho_2}{\longrightarrow}& \mathbb{C} \\ A &\mapsto& p \rho_1(A) + (1-p) \rho_2(A) }$

is a state.

Definition

(pure state)

A state $\rho \colon \mathcal{A} \to \mathbb{C}$ on a unital star-algebra (def. ) is called a pure state if it is extremal in the convex set of all states (remark ) in that an identification

$\rho = p \rho_1 + (1-p) \rho_2$

for $p \in (0,1)$ implies that $\rho_1 = \rho_2$ (hence $= \rho$).

For $C^\ast$-Algebras

The following discusses states specifically on C*-algebras.

Definition

An element $A$ of an (abstract) $C^*$-algebra is called positive if it is self-adjoint and its spectrum is contained in $[0, \infinity)$. We write $A \ge 0$ and say that the set of all positive operators is the positive cone (of a given $C^*$-algebra).

Remark

This definition is motivated by the Hilbert space situation, where an operator $A \in \mathcal{B} (\mathcal{H})$ is called positive if for every vector $x \in \mathcal{H}$ the inequality $\langle x, A x \rangle \ge 0$ holds. If the abstract $C^*$-algebra of the definition above is represented on a Hilbert space, then we see that by functional calculus we can define a self adjoint operator $B$ by $B \coloneqq f(A)$ with $f(t) := t^{1/2}$ and get $\langle x, A x \rangle = \langle B x, B x \rangle \ge 0$. This shows that the positive elements of the abstract algebra, if represented on a Hilbert space, become positive operators as defined here in the Hilbert space setting.

Definition

A linear functional $\rho$ on an $C^*$-algebra is positive if $A \ge 0$ implies that $\rho(A) \ge 0$.

A state of a unital $C^*$-algebra is linear functional $\rho$ such that $\rho$ is positive and $\rho(1) = 1$.

Though the mathematical notion of state is already close to what physicists have in mind, they usually restrict the set of states further and consider normal states only. We let $\mathcal{R}$ be an $C^*$-algebra and $\pi$ an representation of $\mathcal{R}$ on a Hilbert space $\mathcal{H}$.

Definition/Theorem

A normal state $\rho$ is a state that satisfies one of the following equivalent conditions:

• $\rho$ is weak-operator continuous on the unit ball of $\pi(\mathcal{R})$.

• $\rho$ is strong-operator continuous on the unit ball $\pi(\mathcal{R})$.

• $\rho$ is ultra-weak continuous.

• There is an operator $A$ of trace class of $\mathcal{H}$ with $tr(A) = 1$ such that $\rho(\pi(R)) = tr(A \pi(R))$ for all $R \in \mathcal{R}$.

This appears as KadisonRingrose, def. 7.1.11, theorem 7.1.12

Remark

This list is not complete, there are more commonly used equivalent characterizations of normal states.

The last one is most frequently used by physicists, in that context the operator $A$ is also called a density matrix or density operator.

Sometimes the observables of a system are described by an abstract $C^*$-algebra, in this case an important notion is the folium:

Definition

The folium of a representation $\pi$ of an $C^*$-algebra $\mathcal{R}$ on a Hilbert space is the set of normal states of $\pi(\mathcal{R})$.

Definition

A state $\rho$ of a representation is called a vector state if there is a $x \in \mathcal{H}$ such that $\rho(\pi(R)) = \langle \pi(R)x, x \rangle$ for all $R \in \mathcal{R}$.

Theorem

Normal states are vector states if $\mathcal{R}$ is a von Neumann algebra with a separating vector. More precisely: Let $\mathcal{R}$ be a von Neumman algebra acting on a Hilbert space $\mathcal{H}$, let $\rho$ be a normal state of $\mathcal{R}$ and $x \in \mathcal{H}$ be a separating vector for $\mathcal{R}$, then there is a $y \in \mathcal{H}$ such that $\rho(R) = \langle Ry, y \rangle$ for all $R \in \mathcal{R}$.

This appears as KadisonRingrose, theorem 7.2.3.

The set of states of an $C^*$-algebra is sometimes called the state space.

The state space is non-empty (define a state on the subalgebra $\mathbb{C} 1$ and extend it to the whole $C^*$-algebra via the Hahn-Banach theorem), convex and weak$^*$-compact, so it has extreme points. By the Krein-Milman theorem? (see Wikipedia: Krein-Milman theorem) it is the weak$^*$-closure of its extreme points.

Definition

A pure state is a state that is an extreme point of the state space.

The term “pure” originates from the notion of entanglement, a pure state is not a mixture of two distinct other states.

Examples

Proposition

(classical probability measure as state on measurable functions)

For $\Omega$ classical probability space, hence a measure space which normalized total measure $\int_\Omega d\mu = 1$, let $\mathcal{A} \coloneqq L^1(\Omega)$ be the algebra of Lebesgue measurable functions with values in the complex numbers, regarded as a star algebra by pointwise complex conjugation. Then forming the expectation value with respect to $\mu$ defines a state (def. ):

$\array{ L^1(\Omega) &\overset{\langle (-)\rangle_\mu}{\longrightarrow}& \mathbb{C} \\ A &\mapsto& \int_\Omega A d\mu }$
Example

(elements of a Hilbert space as pure states on bounded operators)

Let $\mathcal{H}$ be a complex separable Hilbert space with inner product $\langle -,-\rangle$ and let $\mathcal{A} \coloneqq \mathcal{B}(\mathcal{H})$ be the algebra of bounded operators, regarded as a star algebra under forming adjoint operators. Then for every element $\psi \in \mathcal{H}$ of unit norm $\langle \psi,\psi\rangle = 1$ there is the state (def. ) given by

$\array{ \mathcal{B}(\mathcal{H}) &\overset{\langle (-)\rangle_\psi}{\longrightarrow}& \mathbb{C} \\ A &\mapsto& \langle \psi \vert\, A \, \vert \psi \rangle &\coloneqq& \langle \psi, A \psi \rangle }$

These are pure states (def. ).

More general states in this case are given by density matrices.

Properties

Closure properties

Example

(mixtures, convex combinations)

For $k \in \mathbb{N}_+$, let

• $\big( \rho_i \colon \mathcal{A} \to \mathbb{C} \big)_{i = 1}^k$

be a $k$-tuple of states;

• $\big( p_i \in \mathbb{R}_{\geq 0} \big)_{i = 1}^k$, $\underoverset{i = 1}{k}{\sum} p_i \;=\; 1$

be a probability distribution on the finite set $\{1, \cdots, k\}$

then the convex combination

$\underoverset{i = 1}{k}{\sum} p_i \cdot \rho_i \;\;\; \in \; \mathcal{A}^\ast$

is another state on the star-algebra $\mathcal{A}$.

Proposition

(operator-state correspondence)

For $\rho \;\colon\; \mathcal{A} \to \mathbb{C}$ a state, with a non-null observable $O \in \mathcal{A}$, $\rho(O^\ast O) \neq 0$, then also

(3)$\rho_O \;\colon\; A \;\mapsto\; \tfrac{1}{ \rho(O^\ast O) } \cdot \rho\big( O^\ast \cdot A \cdot O \big)$

is a state.

Proof

To check positivity (1), we compute for any $A \in \mathcal{A}$ as follows:

\begin{aligned} \rho_O \big( A^\ast \cdot A \big) & = \tfrac{1}{\rho(O^\ast O)} \cdot \rho \big( O^\ast \cdot ( A^\ast \cdot A ) \cdot O \big) \\ & = \tfrac{1}{\rho(O^\ast O)} \cdot \rho \big( ( A \cdot O )^\ast \cdot ( A \cdot O ) \big) \\ & \geq 0 \,, \end{aligned}

where the first step is the definition (3) the second step uses the anti-homomorphisms-property of the star-involution, and the last step follows by the assumed positivivity of $\rho$.

To check normalization (2), we observe that:

\begin{aligned} \rho_O(\mathbf{1}) & = \tfrac{1}{\rho(O^\ast O)} \cdot \rho \big( O^\ast \cdot \mathbf{1} \cdot O \big) \\ & = \tfrac{1}{\rho(O^\ast O)} \cdot \rho \big( O^\ast \cdot O \big) \\ & = 1 \,. \end{aligned}

Fell’s theorem

See at Fell's theorem.

Gleason’s theorem

See at Gleason's theorem.

References

Exposition

• Paul-André Meyer, Section I.1.1 in: Quantum Probability for Probabilists, Lecture Notes in Mathematics 1538, Springer 1995 (doi:10.1007/BFb0084701)

• Jonathan Gleason, The $C^\ast$-algebraic formalism of quantum mechanics, 2009 (pdf, pdf)

With an eye towards density matrices and their entropy:

With an eye towards Gelfand-Tsetlin algebras and in the generality of conditional expectation values:

Original articles

For more references see at operator algebra.

Last revised on May 19, 2021 at 08:07:38. See the history of this page for a list of all contributions to it.