Cyclic and Separating Vectors and the Standard Form

Intuition Beginner

Take an algebra of operators acting on a space of arrows, and pick a single arrow inside that space. The arrow is called cyclic if the operators can push it around enough to reach everywhere: sweep the arrow with every operator in the algebra and the results fill up the whole space, leaving no corner unvisited. Such an arrow is rich — one starting point from which the entire space grows.

The same arrow is called separating if it never gets fooled. Apply any operator in the algebra to it; if the result is the zero arrow, then the operator itself had to be zero to begin with. Nothing nonzero in the algebra can quietly annihilate this arrow. So the arrow is honest — it remembers every operator distinctly and refuses to collapse any of them to nothing.

A single arrow can be both at once, and that pairing is the whole point. The neat fact tying the two together is a swap: an arrow that reaches everywhere for one algebra is exactly an arrow that fools no operator in the partner algebra of everything-that-commutes. Richness for one side is honesty for the other.

Visual Beginner

One marked arrow seen two ways: rich enough to generate the whole space by sweeping it with the algebra, and honest enough that no nonzero algebra operator sends it to zero.

The dictionary reads: cyclic means the algebra times the arrow is dense — sweeping reaches everywhere; separating means the only algebra operator killing the arrow is the zero operator — no honest arrow is fooled; and the two notions trade places when you pass to the partner algebra of commuting operators.

Worked example Beginner

Work with two-by-two complex matrices acting on flat four-dimensional space, viewed as the matrices themselves carrying an inner product: the overlap of two matrices is the sum, over all four entries, of one entry times the matching entry of the other (with a complex flip on the second). The algebra is left multiplication by a matrix; the partner algebra is right multiplication.

Pick the identity matrix as the marked arrow. Is it cyclic for left multiplication? Sweep it: left-multiplying the identity by any matrix returns that matrix, and as the matrix ranges over all of them you reach every arrow in the four-dimensional space. So the identity is cyclic — sweeping reaches everywhere.

Is it separating? Left-multiply the identity by a matrix and ask when you get the zero matrix. Left-multiplying the identity by a matrix gives back that very matrix, so the result is zero only when the matrix was already zero. No nonzero matrix fools the identity. So the identity is separating too.

What this tells us: the identity matrix is a single arrow that is at once rich and honest. The same arrow that reaches everywhere by left sweeping is the one no nonzero left operator can collapse — and the swap shows up as the pairing of left and right multiplication.

Check your understanding Beginner

Formal definition Intermediate+

Let $M\subseteq B(H)$ be a von Neumann algebra 39.03.01 on a Hilbert space $H$ 02.11.08, with commutant $M^{\prime }$. A vector $\xi \in H$ is:

• cyclic for $M$ if the orbit $M\xi =\{a\xi :a\in M\}$ is dense in $H$, i.e. $\overline{M\xi }=H$;
• separating for $M$ if $a\xi =0$ with $a\in M$ forces $a=0$; equivalently the map $M\to H$, $a\mapsto a\xi$, is injective.

The two notions are dual under the commutant.

Lemma (cyclic-separating duality). A vector $\xi$ is cyclic for $M$ if and only if it is separating for $M^{\prime }$. ^{[Bratteli-Robinson §2.5]}

Proof. Suppose $\xi$ is cyclic for $M$ and let $a^{\prime }\in M^{\prime }$ with $a^{\prime }\xi =0$. For every $a\in M$, $a^{\prime }(a\xi )=a(a^{\prime }\xi )=0$, so $a^{\prime }$ vanishes on the dense set $M\xi$; by boundedness $a^{\prime }=0$. Hence $\xi$ separates $M^{\prime }$. Conversely suppose $\xi$ is not cyclic for $M$, so $K=\overline{M\xi }\ne H$. Let $p$ be the orthogonal projection onto $K$. The subspace $K$ is $M$-invariant, so $p\in M^{\prime }$ (the projection onto the closure of an $M$-invariant subspace lies in the commutant, by the argument of 39.03.01). Then $1-p\in M^{\prime }$ is nonzero (since $K\ne H$) yet $(1-p)\xi =0$ because $\xi =1\cdot \xi \in M\xi \subseteq K$. So $\xi$ fails to separate $M^{\prime }$. The contrapositive gives the claim. $\square$

Interchanging $M$ and $M^{\prime }$ (using $M^{\prime \prime }=M$) yields the symmetric statement: $\xi$ is cyclic for $M^{\prime }$ iff it is separating for $M$. A vector that is simultaneously cyclic and separating for $M$ is therefore cyclic and separating for $M^{\prime }$ as well; this is the symmetric datum on which Tomita-Takesaki theory rests.

A normal state $\omega$ on $M$ (a $\sigma$-weakly continuous state 39.03.02) is faithful if $\omega (a^{\ast }a)=0\Rightarrow a=0$. A representation $(\pi ,H,\xi )$ of $M$ with cyclic vector $\xi$ implementing $\omega$, $\omega (a)=⟨\pi (a)\xi ,\xi ⟩$, has $\xi$ separating for $\pi (M)$ precisely when $\omega$ is faithful: $\pi (a)\xi =0\Rightarrow ⟨\pi (a^{\ast }a)\xi ,\xi ⟩=\omega (a^{\ast }a)=0\Rightarrow a=0$.

A standard form of a von Neumann algebra is a quadruple $(M,H,J,\mathfrak{P})$ where $M$ acts on $H$, $J$ is an antiunitary involution (the modular conjugation, $J=J^{\ast }=J^{-1}$, conjugate-linear, $J^2=1$), and $\mathfrak{P}\subseteq H$ is a self-dual cone ($\mathfrak{P}={\mathfrak{P}}^{\circ }:=\{\eta :⟨\eta ,\zeta ⟩\ge 0\forall \zeta \in \mathfrak{P}\}$), subject to: $JMJ=M^{\prime }$; $JcJ=c^{\ast }$ for $c\in Z(M)=M\cap M^{\prime }$ (the centre acts by complex conjugation); $J\eta =\eta$ for all $\eta \in \mathfrak{P}$; and $aJaJ\mathfrak{P}\subseteq \mathfrak{P}$ for all $a\in M$.

Counterexamples to common slips

• Cyclic does not imply separating. On $H={\mathbb{C}}^2$ let $M=B(H)$; the vector $\xi =e_1$ is cyclic ($B(H)e_1=H$) but not separating, since the rank-one operator $e_2\otimes e_1^{\ast }$ kills it while being nonzero. Separation fails because $M^{\prime }=\mathbb{C}1$ is too small to be separated by a single nonzero vector only when $M$ is large.
• A faithful state is needed for a separating vector, not merely a faithful representation. The GNS vector of a non-faithful state is cyclic but not separating: if $\omega (a^{\ast }a)=0$ for some $a\ne 0$ then ${\pi }_{\omega }(a){\xi }_{\omega }=0$ with ${\pi }_{\omega }(a)$ possibly nonzero, so ${\xi }_{\omega }$ separates ${\pi }_{\omega }(M)$ only after dividing out the kernel.
• The cone $\mathfrak{P}$ is not the cone of positive operators sitting in $H$. It is the self-dual cone, the closure of $\{aJaJ{\xi }_0:a\in M\}$ for a cyclic-separating ${\xi }_0$; the naive image of $M_{+}$ under $a\mapsto a{\xi }_0$ is generally neither self-dual nor $J$-fixed.

Key theorem with proof Intermediate+

Theorem (existence of a cyclic-separating vector from a faithful normal state). Let $M$ be a von Neumann algebra admitting a faithful normal state $\omega$. Then the GNS representation $({\pi }_{\omega },H_{\omega },{\xi }_{\omega })$ is normal and faithful, and ${\xi }_{\omega }$ is a cyclic and separating vector for ${\pi }_{\omega }(M)$. ^{[Takesaki Ch. VI §1; Bratteli-Robinson §2.5]}

Proof. Apply the GNS construction 39.01.03 to the state $\omega$: it produces a Hilbert space $H_{\omega }$, a representation ${\pi }_{\omega }:M\to B(H_{\omega })$, and a cyclic vector ${\xi }_{\omega }$ with $\omega (a)=⟨{\pi }_{\omega }(a){\xi }_{\omega },{\xi }_{\omega }⟩$ and $\overline{{\pi }_{\omega }(M){\xi }_{\omega }}=H_{\omega }$. Cyclicity of ${\xi }_{\omega }$ for ${\pi }_{\omega }(M)$ is immediate from the construction.

Normality of ${\pi }_{\omega }$ follows from normality of $\omega$. The vector functionals $a\mapsto ⟨{\pi }_{\omega }(a)\eta ,\zeta ⟩$ are $\sigma$-weakly continuous, and $\omega$ being $\sigma$-weakly continuous means the form $⟨{\pi }_{\omega }(\cdot ){\xi }_{\omega },{\xi }_{\omega }⟩=\omega$ is too; cyclicity propagates $\sigma$-weak continuity to all matrix coefficients $⟨{\pi }_{\omega }(\cdot ){\pi }_{\omega }(b){\xi }_{\omega },{\pi }_{\omega }(c){\xi }_{\omega }⟩=\omega (c^{\ast }(\cdot )b)$, which are $\sigma$-weakly continuous since $\omega$ is, and these span a dense set of functionals. Hence ${\pi }_{\omega }$ is $\sigma$-weakly continuous, i.e. normal, and ${\pi }_{\omega }(M)$ is a von Neumann algebra (the normal image of a von Neumann algebra is $\sigma$-weakly closed).

Faithfulness of ${\pi }_{\omega }$ follows from faithfulness of $\omega$: if ${\pi }_{\omega }(a)=0$ then $\omega (a^{\ast }a)=\Vert {\pi }_{\omega }(a){\xi }_{\omega }{\Vert }^2=0$, so $a=0$ as $\omega$ is faithful. Thus ${\pi }_{\omega }$ is an isomorphism onto ${\pi }_{\omega }(M)$, and we may identify $M$ with ${\pi }_{\omega }(M)$.

It remains to show ${\xi }_{\omega }$ is separating for ${\pi }_{\omega }(M)$. Suppose ${\pi }_{\omega }(a){\xi }_{\omega }=0$. Then $$ \omega(a^*a) = \langle \pi_\omega(a^*a)\xi_\omega, \xi_\omega\rangle = \langle \pi_\omega(a)\xi_\omega, \pi_\omega(a)\xi_\omega\rangle = |\pi_\omega(a)\xi_\omega|^2 = 0, $$ and faithfulness of $\omega$ gives $a=0$, hence ${\pi }_{\omega }(a)=0$. So the only element of ${\pi }_{\omega }(M)$ annihilating ${\xi }_{\omega }$ is $0$, and ${\xi }_{\omega }$ is separating. By the duality lemma ${\xi }_{\omega }$ is then cyclic for ${\pi }_{\omega }(M{)}^{\prime }$ as well, so it is cyclic and separating on both sides. $\square$

A faithful normal state exists whenever $M$ has a separable predual (equivalently $H_{\omega }$ is separable): take a faithful normal state as a $\sigma$-convergent convex combination ${\sum }_n{2}^{-n}{\omega }_n$ of normal states ${\omega }_n$ whose supports increase to $1$. So every von Neumann algebra with separable predual sits in standard position.

Bridge. This existence result builds toward the entire Tomita-Takesaki apparatus, and it appears again in 39.04.02 where the densely defined antilinear operator $S_0:a{\xi }_{\omega }\mapsto a^{\ast }{\xi }_{\omega }$ — well defined precisely because ${\xi }_{\omega }$ is cyclic and separating — has polar decomposition $S=J{\Delta }^{1/2}$ producing the modular conjugation $J$ and the modular operator $\Delta$. The foundational reason the cyclic-separating pair is the right datum is exactly the duality lemma: cyclicity for $M$ is separation for $M^{\prime }$, so a single vector lets the algebra and its commutant be analysed symmetrically, and this is exactly the symmetry the modular involution $J$ will turn into the spatial isomorphism $JMJ=M^{\prime }$. The construction generalises the GNS picture of 39.01.03, where a state already became a cyclic vector; faithfulness is the extra honesty that upgrades cyclic to cyclic-and-separating. Putting these together, the standard form is the canonical home in which $M$ and $M^{\prime }$ are mirror images under $J$, and the bridge is that the modular flow ${\Delta }^{it}$ acts inside this home as a one-parameter automorphism group intrinsic to $M$.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none — Mathlib has Hilbert spaces, adjoints, projections, and antilinear maps via the conjugate scalar action, but no VonNeumannAlgebra predicate (a gap inherited from 39.03.01), no cyclic/separating predicates for a von Neumann algebra, no cyclic-separating duality theorem, and nothing of the standard form $(M,H,J,\mathfrak{P})$ with its self-dual cone and the relation $JMJ=M^{\prime }$.

The intended statements read schematically:

import Mathlib.Analysis.InnerProductSpace.Adjoint
import Mathlib.Topology.Algebra.Module.WeakDual

variable {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H]

/-- A vector is cyclic for a set of operators when its orbit is dense. -/
def IsCyclic (M : Set (H →L[ℂ] H)) (ξ : H) : Prop :=
  Dense (Set.range fun a : M => (a : H →L[ℂ] H) ξ)

/-- A vector is separating when only the zero operator annihilates it. -/
def IsSeparating (M : Set (H →L[ℂ] H)) (ξ : H) : Prop :=
  ∀ a ∈ M, a ξ = 0 → a = 0

/-- Cyclic for M ↔ separating for the commutant M'. -/
theorem cyclic_iff_separating_commutant
    (M : Set (H →L[ℂ] H)) (hM : IsVonNeumannAlgebra M) (ξ : H) :
    IsCyclic M ξ ↔ IsSeparating (commutant M) ξ :=
  sorry  -- projection onto closure of Mξ lies in M'

Advanced results Master

The cyclic-separating datum is the entrance to modular theory, and the standard form is its canonical packaging. The structural results below organise what the datum buys.

The standard form and its uniqueness (Haagerup). A standard form of $M$ is a quadruple $(M,H,J,\mathfrak{P})$ with $J$ an antiunitary involution, $\mathfrak{P}$ a self-dual cone, $JMJ=M^{\prime }$, $JcJ=c^{\ast }$ on the centre, $J\eta =\eta$ for $\eta \in \mathfrak{P}$, and $aJaJ\mathfrak{P}\subseteq \mathfrak{P}$ for $a\in M$ ^{[Haagerup 1975]}. Haagerup proved existence for every von Neumann algebra and uniqueness up to a unique spatial isomorphism: if $(M,H_1,J_1,{\mathfrak{P}}_1)$ and $(M,H_2,J_2,{\mathfrak{P}}_2)$ are two standard forms, there is a unique unitary $u:H_1\to H_2$ with $ua{u}^{\ast }=a$ (intertwining the $M$-actions through the fixed isomorphism), $u{J}_1{u}^{\ast }=J_2$, and $u{\mathfrak{P}}_1={\mathfrak{P}}_2$. The cone rigidifies the form: the antiunitary $J$ alone is not canonical, but $J$ together with the self-dual cone it fixes is. For a von Neumann algebra with a cyclic-separating vector $\xi$, the cone is $\mathfrak{P}=\overline{\{aJaJ\xi :a\in M\}}=\overline{\{aj(a)\xi \}}$ where $j(a)=JaJ\in M^{\prime }$, and $J,\Delta$ come from 39.04.02.

Normal states are vector states in $\mathfrak{P}$, uniquely. In a standard form, every normal positive functional $\omega \in M_{\ast }^{+}$ is implemented by a unique vector ${\xi }_{\omega }\in \mathfrak{P}$: $\omega (a)=⟨a{\xi }_{\omega },{\xi }_{\omega }⟩$, and the map $\omega \mapsto {\xi }_{\omega }$ is a homeomorphism from $M_{\ast }^{+}$ (norm topology) onto $\mathfrak{P}$ with $\Vert {\xi }_{\omega }-{\xi }_{{\omega }^{\prime }}{\Vert }^2\le \Vert \omega -{\omega }^{\prime }\Vert \le \Vert {\xi }_{\omega }-{\xi }_{{\omega }^{\prime }}\Vert \Vert {\xi }_{\omega }+{\xi }_{{\omega }^{\prime }}\Vert$. The self-duality of $\mathfrak{P}$ is exactly what removes the phase ambiguity of the GNS vector — without the cone, $\omega$ determines ${\xi }_{\omega }$ only up to a unitary in $M^{\prime }$. This is the precise sense in which the standard form encodes the algebra together with its state space: $M$, $M_{\ast }$, and the order structure are all read off $(H,J,\mathfrak{P})$.

Spatial implementation of automorphisms. Every normal $\ast$-automorphism $\alpha \in \mathrm{Aut}(M)$ is spatially implemented by a unique unitary $u_{\alpha }$ preserving the cone: $u_{\alpha }a{u}_{\alpha }^{\ast }=\alpha (a)$, $u_{\alpha }J=J{u}_{\alpha }$, and $u_{\alpha }\mathfrak{P}=\mathfrak{P}$. The map $\alpha \mapsto u_{\alpha }$ is a group isomorphism from $\mathrm{Aut}(M)$ onto the unitaries preserving $\mathfrak{P}$, splitting the projective-representation ambiguity that plagues general spatial implementations. Thus the symmetry group of $M$ acts honestly (not merely projectively) on the standard Hilbert space, which is what makes the standard form the natural arena for crossed products, the flow of weights, and the equivariant analysis of dynamical systems on $M$.

The modular operator and Tomita's theorem. From a cyclic-separating $\xi$ the closed antilinear $S=J{\Delta }^{1/2}$ (Exercise 7) has positive self-adjoint $\Delta =S^{\ast }S$ (the modular operator) and antiunitary $J$ (the modular conjugation). Tomita's theorem 39.04.02 asserts $JMJ=M^{\prime }$ and ${\Delta }^{it}M{\Delta }^{-it}=M$ for all $t\in \mathbb{R}$, so ${\sigma }_t^{\omega }=\mathrm{Ad}({\Delta }^{it})$ is a one-parameter automorphism group of $M$ — the modular automorphism group intrinsic to the pair $(M,\omega )$. The cyclic-separating vector is the indispensable input: $S_0$ is only well defined because $\xi$ separates $M$ and only densely defined because $\xi$ is cyclic.

Spatial derivative and independence of the state (Connes cocycle). Two faithful normal states $\omega ,\psi$ give modular groups ${\sigma }^{\omega },{\sigma }^{\psi }$ related by the Connes cocycle $(D\psi :D\omega {)}_t\in M$, a ${\sigma }^{\omega }$-cocycle of unitaries with ${\sigma }_t^{\psi }=\mathrm{Ad}((D\psi :D\omega {)}_t)\circ {\sigma }_t^{\omega }$. The class of ${\sigma }^{\omega }$ modulo inner automorphisms is therefore independent of the state, an invariant $\delta :\mathbb{R}\to \mathrm{Out}(M)$ of $M$ alone. This canonical outer flow is the engine of the type III classification: the period/spectrum of $\delta$ distinguishes the subtypes III${}_{\lambda }$, $0\le \lambda \le 1$, that lie beyond the trace-bearing types.

Synthesis. The cyclic-separating vector is the foundational reason a von Neumann algebra carries an intrinsic dynamics: it is exactly the datum making $S_0:a\xi \mapsto a^{\ast }\xi$ well defined and densely defined, and this is exactly what the duality lemma guarantees by trading cyclicity for $M$ against separation for $M^{\prime }$. The standard form putting these together is the canonical home where $JMJ=M^{\prime }$ realises the commutant as a mirror image, every normal state appears once as a vector in the self-dual cone, and every automorphism is spatially implemented — the central insight that the algebra, its predual, and its symmetry group all live inside one rigid geometric object $(H,J,\mathfrak{P})$. This is dual to the GNS picture of 39.01.03, where a single state already became a cyclic vector; here faithfulness upgrades that to the symmetric cyclic-and-separating datum, and Haagerup's uniqueness theorem generalises the GNS uniqueness from a single state to the whole state space at once. The modular flow built on this datum appears again in 39.04.02 as Tomita's theorem and in the type III classification as the state-independent outer flow, so the whole of modular theory is the systematic exploitation of the one cyclic-separating vector whose existence the key theorem secured.

Full proof set Master

Proposition (cyclic-separating duality, both directions). Let $M$ be a von Neumann algebra with commutant $M^{\prime }$ ($M^{\prime \prime }=M$). A vector $\xi$ is cyclic for $M$ iff separating for $M^{\prime }$, and separating for $M$ iff cyclic for $M^{\prime }$. Proof. The first equivalence is the lemma of the Formal-definition section: cyclicity of $\xi$ for $M$ gives separation for $M^{\prime }$ by pushing an annihilating $a^{\prime }\in M^{\prime }$ onto the dense orbit $M\xi$; the converse uses that non-cyclicity yields a proper invariant subspace $K=\overline{M\xi }$ whose complementary projection $1-p\in M^{\prime }$ is nonzero yet annihilates $\xi$. Replacing $M$ by $M^{\prime }$ and using $M^{\prime \prime }=M$ gives the second equivalence verbatim. $\square$

Proposition (separating $⟺$ faithful normal state). Let $\xi$ be a unit vector cyclic for $M$ and $\omega ={\omega }_{\xi }$. Then $\xi$ is separating for $M$ iff $\omega$ is faithful. Proof. If $\xi$ separates $M$ and $\omega (a^{\ast }a)=\Vert a\xi {\Vert }^2=0$, then $a\xi =0$, so $a=0$: $\omega$ is faithful. Conversely if $\omega$ is faithful and $a\xi =0$ with $a\in M$, then $\omega (a^{\ast }a)=\Vert a\xi {\Vert }^2=0$ forces $a=0$: $\xi$ separates $M$. $\square$

Proposition (existence of a faithful normal state, separable predual). If $M_{\ast }$ is separable, $M$ has a faithful normal state. Proof. Choose a countable family $({\omega }_n)$ of normal states whose support projections $s({\omega }_n)$ have supremum ${\bigvee }_{n}s({\omega }_n)=1$ (possible because the normal states separate $M$, so no nonzero projection annihilates all of them; separability lets a countable subfamily already achieve the supremum). Set $\omega ={\sum }_n{2}^{-n}{\omega }_n$, a norm-convergent sum of normal states, hence a normal state. Its support is ${\bigvee }_{n}s({\omega }_n)=1$, so $\omega (a^{\ast }a)=0\Rightarrow s(\omega )\ge$ the support of $a^{\ast }a$ fails unless $a^{\ast }a=0$; concretely $\omega (a^{\ast }a)={\sum }_n{2}^{-n}{\omega }_n(a^{\ast }a)=0$ forces every ${\omega }_n(a^{\ast }a)=0$, so $a^{\ast }a\le 1-s({\omega }_n)$ for all $n$, whence $a^{\ast }a\le 1-\bigvee s({\omega }_n)=0$ and $a=0$. Thus $\omega$ is faithful. $\square$

Proposition (the standard form recovers normal states uniquely in the cone). In a standard form $(M,H,J,\mathfrak{P})$ the map $\mathfrak{P}\to M_{\ast }^{+}$, $\xi \mapsto {\omega }_{\xi }$, is a bijection. Proof sketch. Surjectivity: a normal positive $\omega$ has a GNS vector, and the cone is built so that the modular flow rotates any implementing vector to a unique representative in $\mathfrak{P}$ (the element ${\Delta }^{1/4}$-image of the positive part). Injectivity: if $\xi ,\eta \in \mathfrak{P}$ implement the same $\omega$, they differ by a unitary $u^{\prime }\in M^{\prime }$ with $u^{\prime }\xi =\eta$; self-duality and $J$-invariance of $\mathfrak{P}$ force $u^{\prime }=JuJ$ for $u\in M$ unitary and then $u=1$ by a positivity argument on $⟨u^{\prime }\xi ,\xi ⟩$, giving $\eta =\xi$. The metric inequalities follow from $\Vert \xi -\eta {\Vert }^2\le \Vert {\omega }_{\xi }-{\omega }_{\eta }\Vert$ via the Powers-Størmer inequality. $\square$ (Full proof: Haagerup 1975 ^{[Haagerup 1975]}.)

Proposition (automorphisms are implemented in the cone). For $\alpha \in \mathrm{Aut}(M)$ there is a unique unitary $u_{\alpha }$ on $H$ with $u_{\alpha }a{u}_{\alpha }^{\ast }=\alpha (a)$, $u_{\alpha }J=J{u}_{\alpha }$, $u_{\alpha }\mathfrak{P}=\mathfrak{P}$, and $\alpha \mapsto u_{\alpha }$ is a homomorphism. Proof sketch. Existence: $\alpha$ transports the cyclic-separating vector $\xi$ to a vector implementing $\omega \circ {\alpha }^{-1}$; the cone-representative of that state defines $u_{\alpha }\xi$, and $u_{\alpha }(a\xi )=\alpha (a)u_{\alpha }\xi$ extends to a unitary by density. Cone-preservation and commutation with $J$ are forced by the rigidity of the previous proposition. Uniqueness: two such unitaries differ by a unitary in $M^{\prime }\cap M=Z(M)$ fixing the cone, hence by $1$. Homomorphism property is immediate from uniqueness. $\square$ (Haagerup 1975; Takesaki Ch. VI ^{[Takesaki Ch. VI §1]}.)

Connections Master

• The modular operator and Tomita's theorem 39.04.02 — the cyclic-separating vector makes $S_0:a\xi \mapsto a^{\ast }\xi$ well defined and densely defined; its closure's polar decomposition $S=J{\Delta }^{1/2}$ produces the modular conjugation and modular operator, and Tomita's theorem $JMJ=M^{\prime }$, ${\Delta }^{it}M{\Delta }^{-it}=M$ is the structural payoff this unit sets up.

• Von Neumann algebras and the bicommutant theorem 39.03.01 — the duality lemma "cyclic for $M$ iff separating for $M^{\prime }$" runs on the very projection-in-the-commutant argument that drives the bicommutant theorem; the standard form is the canonical representation in which the bicommutant relation $M=M^{\prime \prime }$ is realised with $M$ and $M^{\prime }$ as $J$-mirror images.

• States, the GNS construction, and Gelfand-Naimark 39.01.03 — the existence theorem is the GNS construction applied to a faithful normal state, upgrading the cyclic GNS vector to a cyclic-and-separating vector; the standard form is the canonical, phase-rigidified GNS representation that works for the whole normal state space simultaneously.

• The predual, normal states, and the $\sigma$-weak topology 39.03.02 — normality of the implementing state is what makes the GNS representation normal and the image a von Neumann algebra; in the standard form the predual $M_{\ast }^{+}$ is realised isometrically (up to the Powers-Størmer comparison) as the self-dual cone $\mathfrak{P}$.

• The Kaplansky density theorem 39.03.03 — Kaplansky density is the technical tool extending normal maps from a strongly dense C*-subalgebra to the von Neumann completion, the step used to verify that the modular and cone structures built on a generating subalgebra propagate to all of $M$ in the standard-form construction.

Historical & philosophical context Master

Cyclic and separating vectors were isolated in the 1930s rings-of-operators work of Murray and von Neumann, where the existence of a cyclic and separating "trace vector" for a finite factor underlies the canonical trace; the general framework matured with Tomita's 1967 manuscripts and Takesaki's 1970 Lecture Notes Tomita's Theory of Modular Hilbert Algebras and its Applications ^{[Takesaki Ch. VI §1]}, which gave the closed antilinear operator $S=J{\Delta }^{1/2}$ its definitive treatment and proved $JMJ=M^{\prime }$. The standard form as an intrinsic, representation-independent object — the quadruple $(M,H,J,\mathfrak{P})$ with the self-dual cone — was constructed by Uffe Haagerup in 1975 ^{[Haagerup 1975]}, who proved both existence for arbitrary von Neumann algebras and uniqueness up to a unique cone-preserving unitary, and established that normal states are exactly the vectors of the cone and that automorphisms are spatially implemented within it. Parallel cone constructions were given by Araki, Connes, and Woronowicz in the same period.

The standard form supplied the geometric substrate for Connes' classification of type III factors, for which he received the Fields Medal in 1982, and the Bisognano-Wichmann identification of the modular operator of a wedge algebra with a Lorentz boost made the cyclic-separating vacuum of algebraic quantum field theory the source of the Unruh and Hawking temperatures. Reeh and Schlieder had shown in 1961 that the QFT vacuum is cyclic and separating for every local algebra, which is precisely the hypothesis modular theory consumes.

Bibliography Master

• Takesaki, M., Theory of Operator Algebras II, Encyclopaedia of Mathematical Sciences 125, Springer, 2003. Ch. VI.
• Haagerup, U., "The standard form of von Neumann algebras", Mathematica Scandinavica 37 (1975), 271-283.
• Bratteli, O. and Robinson, D. W., Operator Algebras and Quantum Statistical Mechanics I, 2nd ed., Springer, 1987. §2.5.
• Takesaki, M., Tomita's Theory of Modular Hilbert Algebras and its Applications, Lecture Notes in Mathematics 128, Springer, 1970.
• Connes, A., "Une classification des facteurs de type III", Annales scientifiques de l'École Normale Supérieure 6 (1973), 133-252.
• Reeh, H. and Schlieder, S., "Bemerkungen zur Unitäräquivalenz von Lorentzinvarianten Feldern", Il Nuovo Cimento 22 (1961), 1051-1068.
• Araki, H., "Some properties of modular conjugation operator of von Neumann algebras and a non-commutative Radon-Nikodym theorem with a chain rule", Pacific Journal of Mathematics 50 (1974), 309-354.

---

Operator-algebras spine, foundational Tomita-Takesaki unit. The cyclic-separating vector as the symmetric datum (cyclic for $M$ = separating for $M^{\prime }$), its existence from a faithful normal state via GNS, and the standard form $(M,H,J,\mathfrak{P})$ with its self-dual cone — the canonical home where $JMJ=M^{\prime }$, normal states are cone vectors, and automorphisms are spatially implemented; sets up the modular operator and Tomita's theorem (39.04.02).