Algebraic QFT and Modular Theory¶

Job: develop Haag–Kastler nets at frontier depth — the Reeh–Schlieder theorem (cyclicity of the vacuum for local algebras as a consequence of locality + spectrum), Tomita–Takesaki modular theory in QFT (Borchers, Bisognano–Wichmann: modular group of the wedge = Lorentz boosts, modular conjugation = CPT), the split property / type III$_1$ local algebras, and the DHR superselection theory.

Core¶

Haag–Kastler nets¶

A Haag–Kastler net on Minkowski $\mathbb{R}^{1,d-1}$ is an assignment $\mathcal{O}\mapsto\mathcal{A}(\mathcal{O})$ of a von Neumann algebra $\mathcal{A}(\mathcal{O})\subset B(\mathcal{H})$ to each bounded open region $\mathcal{O}$, satisfying:

Isotony. $\mathcal{O}_1\subset\mathcal{O}_2\Rightarrow\mathcal{A}(\mathcal{O}_1)\subset\mathcal{A}(\mathcal{O}_2)$.
Locality. If $\mathcal{O}_1,\mathcal{O}_2$ are spacelike-separated ($\mathcal{O}_1\subset\mathcal{O}_2'$, the causal complement), then $[\mathcal{A}(\mathcal{O}_1),\mathcal{A}(\mathcal{O}_2)]=\{0\}$.
Covariance. A strongly continuous unitary representation $U$ of $\mathcal{P}_+^\uparrow$ with $U(a,\Lambda)\mathcal{A}(\mathcal{O})U(a,\Lambda)^{-1}=\mathcal{A}(\Lambda\mathcal{O}+a)$.
Vacuum. A unit vector $\Omega\in\mathcal{H}$, unique Poincaré-invariant, cyclic for the quasi-local $C^*$-algebra $\mathcal{A}=\overline{\bigcup\mathcal{A}(\mathcal{O})}^{\|\cdot\|}$.
Spectrum condition. Joint spectrum of the translation generators $P^\mu$ in $\overline{V}_+$.

The net is the basic object; fields (Wightman operators) are one presentation. The causal complement $\mathcal{O}'$ is the interior of the set of points spacelike to all of $\mathcal{O}$; Haag duality is $\mathcal{A}(\mathcal{O}')=\mathcal{A}(\mathcal{O})'$ (the commutant), a non-trivial property holding for the free field and conjectured generally for "well-behaved" theories.

The Reeh–Schlieder theorem¶

Theorem (Reeh–Schlieder). For any nonempty open region $\mathcal{O}$, the vacuum $\Omega$ is cyclic for $\mathcal{A}(\mathcal{O})$ — i.e. $\mathcal{A}(\mathcal{O})\Omega$ is dense in $\mathcal{H}$. If $\Omega$ is also separating for $\mathcal{A}(\mathcal{O})$ (no kernel; this holds e.g. if $\mathcal{O}'\ne\varnothing$ and Haag duality or additivity holds), $\Omega$ is separating as well.

Proof sketch. Suppose $\psi\perp\mathcal{A}(\mathcal{O})\Omega$. Then for all $a\in\mathcal{A}(\mathcal{O})$, the function $f(x)=\langle\psi,U(x)\Omega\rangle$ (a matrix element of the translation) is a boundary value of a holomorphic function on the forward tube $T^+=\mathbb{R}^{1,d-1}+iV_+$ by the spectrum condition (the spectral theorem for $P^\mu$). Locality forces $f$ to vanish on $\mathcal{O}$ (translating $a$ by small $x\in\mathcal{O}-\mathcal{O}$ keeps $a$ in $\mathcal{A}(\mathcal{O})$ by isotony, and $\langle\psi,U(x)a\Omega\rangle=\langle\psi,a(x)\Omega\rangle=0$). By the edge-of-the-wedge theorem, a function holomorphic on the forward tube vanishing on an open set of the boundary extends to vanish identically — so $f\equiv 0$, hence $\psi\perp U(x)\Omega$ for all $x$, i.e. $\psi\perp\mathcal{H}$ (by cyclicity of $\Omega$ for $\mathcal{A}$). Hence $\psi=0$. $\square$

This is a theorem (locality + spectrum ⇒ cyclicity), not an axiom. The vacuum is a highly entangled state: local operations in any region can approximate any global state.

Structural consequence	Statement
Cyclicity of vacuum	$\overline{\mathcal{A}(\mathcal{O})\Omega}=\mathcal{H}$ for any open $\mathcal{O}$
Separating	$\Omega$ separating for $\mathcal{A}(\mathcal{O})$ if $\mathcal{O}'\ne\varnothing$ (so Tomita–Takesaki applies)
Local algebras type III$_1$	(Buchholz–Fredenhagen) the local von Neumann algebras are hyperfinite III$_1$ factors

Tomita–Takesaki modular theory in QFT¶

For a von Neumann algebra $M$ with a cyclic separating vector $\Omega$, the Tomita operator $S_M:a\Omega\mapsto a^*\Omega$ ($a\in M$) has polar decomposition $$ S_M=J_M\,\Delta_M^{1/2},\qquad \Delta_M=S_M^S_M,\quad J_M \text{ antiunitary}, $$ the *modular operator** $\Delta_M$ and modular conjugation $J_M$. The Tomita–Takesaki theorem: $\Delta_M^{it}M\Delta_M^{-it}=M$ (modular automorphism group) and $J_M M J_M^{-1}=M'$ (modular conjugation exchanges $M$ and its commutant).

In QFT, applied to the local algebra $\mathcal{A}(\mathcal{O})$ with the vacuum $\Omega$:

Theorem (Borchers, 1966). Let $W_R=\{x\in\mathbb{R}^{1,d-1}:x^1>|x^0|\}$ be the right wedge. The modular group $\Delta_{W_R}^{it}$ acts as translations along the lightlike direction tangent to the boundary of $W_R$ (the edge $x^0=x^1$), and $J_{W_R}$ acts as a $PT$-like symmetry. More precisely, the modular group of the wedge recovers a one-parameter subgroup of the Poincaré group.

Theorem (Bisognano–Wichmann). For the free scalar field (and any Wightman theory satisfying the HW axioms with the free-field wedge algebra), the modular operator of the right wedge is $$ \Delta_{W_R}^{it}=U\bigl(\Lambda_{W_R}(2\pi t)\bigr),\qquad J_{W_R}=\Theta\cdot J_{\mathrm{CPT}}, $$ where $\Lambda_{W_R}(s)$ is the Lorentz boost in the $x^0$-$x^1$ plane (rapidity $s$), $U$ the Poincaré representation, and $J_{\mathrm{CPT}}$ the CPT conjugation, $\Theta$ the spatial reflection. The modular group is the Lorentz boosts and the modular conjugation is (CPT times the spatial reflection). The result holds for free fields and for fields satisfying the BW conditions; the general interacting case is the Borchers / modular localization program (not a theorem).

Modular data	Spacetime symmetry
$\Delta_{W_R}^{it}=U(\Lambda(2\pi t))$	Lorentz boosts along the $x^0$-$x^1$ plane
$J_{W_R}=\Theta\cdot J_{\mathrm{CPT}}$	CPT composed with spatial reflection
Bisognano–Wichmann (free)	modular group = boosts; modular conjugation = CPT
Borchers (general)	modular group recovers lightlike translations for the wedge tangent

The modular localization program (Brunetti–Fredenhagen–Verch) reconstructs spacetime geometry from the net + vacuum: the wedge algebras and their modular groups determine the Lorentz group and the spacetime itself.

The split property and type III¶

Theorem (Buchholz–Fredenhagen). In a physically reasonable QFT, the local algebras $\mathcal{A}(\mathcal{O})$ are type III$_1$ factors (the hyperfinite III$_1$ factor, the expected universal local algebra).

The split property: for $\mathcal{O}_1\Subset\mathcal{O}_2$ (closure of $\mathcal{O}_1$ compactly contained in $\mathcal{O}_2$), there is a type I factor $N$ with $$ \mathcal{A}(\mathcal{O}_1)\subset N\subset\mathcal{A}(\mathcal{O}_2). $$ The split property is the algebraic version of independent local subsystems: it implies the existence of product states on $\mathcal{A}(\mathcal{O}_1)\vee\mathcal{A}(\mathcal{O}_2)$ (uncorrelated states, the "absence of entanglement at large separation in bounded energy").

Theorem (Buchholz–Wichmann nuclearity). The split property follows from the nuclearity condition: the map $a\mapsto a\Omega$ from the unit ball of $\mathcal{A}(\mathcal{O})$ to $\mathcal{H}$ is nuclear (Hilbert–Schmidt at bounded energy), i.e. the set $\{a\Omega:a\in\mathcal{A}(\mathcal{O}),\|a\|\le1\}$ is nuclear in $\mathcal{H}$. A sufficient condition is the energy nuclearity bound $\mathrm{Tr}(e^{-\beta L_0})<\infty$ for the local algebra's energy content.

The split property + type III$_1$ is the standard AQFT structure: local algebras are III$_1$ (genuinely infinite, no pure states, no traces), but the split interpolates with a type I factor for separated regions.

Superselection sectors and DHR theory¶

The Doplicher–Haag–Roberts (DHR) analysis of charged sectors: a representation $\pi$ of the quasi-local algebra $\mathcal{A}$ is a DHR sector if it is local — $\pi|_{\mathcal{A}(\mathcal{O}')}\cong\pi_0|_{\mathcal{A}(\mathcal{O}')}$ for some bounded $\mathcal{O}$ (the representation differs from the vacuum only inside a bounded region; the charge is localized).

The DHR theory constructs: - the charge group / the interchangeable charges via localized endomorphisms $\rho\in\mathrm{End}(\mathcal{A})$; - the statistics of a charge (Bose / Fermi / anyonic) via the braiding / permutation of double-cone localized sectors — recovering the spin-statistics theorem algebraically; - the DHR category (a braided / symmetric monoidal category of localized endomorphisms).

Theorem (DHR spin-statistics). In a DHR theory (mass gap assumed), the statistics of a sector matches its spin: the statistics phase $e^{i\pi s}$ equals the spin $s$ — recovering the spin-statistics theorem from the algebraic structure.

Buchholz–Fredenhagen extension. For massive particles (relaxing the bounded localization to cone localization), the DHR framework extends to massive charged sectors and the infraparticle problem (the electron is not a DHR sector because of the soft-photon cloud).

See it¶

Bisognano–Wichmann theorem for the free scalar field. For the free real scalar field in $d=4$ with vacuum $\Omega$ on the right wedge $W_R=\{x^1>|x^0|\}$, the local algebra is $\mathcal{A}(W_R)=\{\varphi(f):\mathrm{supp}\,f\subset W_R\}''$ (the von Neumann closure). The Tomita operator $S_{W_R}:a\Omega\mapsto a^*\Omega$ for $a\in\mathcal{A}(W_R)$ is computed via the explicit two-point (Källén–Lehmann) function $$ \langle\Omega,\varphi(x)\varphi(y)\Omega\rangle=\int\frac{d^3k}{(2\pi)^3 2\omega_k}\,e^{-ik(x-y)}. $$ The action of $S_{W_R}$ on the smeared field $\varphi(f)\Omega$ (for $f$ supported in $W_R$) is identified with the CPT operation composed with the spatial reflection $\Theta: x\mapsto -x^1$ (the wedge-flipping isometry), by the analyticity of the Wightman functions under the boost flow. The polar decomposition gives $$ \Delta_{W_R}^{it}=U(\Lambda_{W_R}(2\pi t)),\qquad J_{W_R}=\Theta\cdot J_{\mathrm{CPT}}, $$ with $\Lambda_{W_R}(s)$ the boost in the $x^0$-$x^1$ plane (rapidity $s$). The factor $2\pi$ is the Unruh periodicity — the modular flow of the wedge is the geometric action of Lorentz boosts, and the KMS condition at temperature $1/2\pi$ in the boost parameter is the Unruh effect (the vacuum restricted to the wedge is a KMS state for the boost flow at the Unruh temperature). This identifies the modular group with the spacetime boosts — the foundational result of "modular localization", reconstructing the Lorentz group from the algebra + vacuum.

Check¶

Show that Reeh–Schlieder cyclicity for the wedge $W_R$ + Haag duality for $W_R$ implies the vacuum is separating for $\mathcal{A}(W_R)$.

Haag duality for the wedge: $\mathcal{A}(W_R')=\mathcal{A}(W_L)=\mathcal{A}(W_R)'$ (the left wedge is the causal complement of the right). Reeh–Schlieder applied to $W_L$ (nonempty open) gives $\overline{\mathcal{A}(W_L)\Omega}=\mathcal{H}$. If $a\in\mathcal{A}(W_R)$ and $a\Omega=0$, then for $b\in\mathcal{A}(W_L)$, $b\Omega$ is dense and $a(b\Omega)=b(a\Omega)=0$ by locality (spacelike commutativity), so $a=0$ on a dense set, hence $a=0$. Thus $\Omega$ is separating for $\mathcal{A}(W_R)$.

Why does the Bisognano–Wichmann theorem fail to generalize immediately to interacting theories, and what replaces it?

The BW theorem uses the explicit form of the free-field two-point function and the Wightman field operator on the wedge; for an interacting theory the wedge algebra and its Tomita operator are not explicitly computable, and the modular group need not coincide with the boosts a priori. The replacement is the Borchers theorem (which derives only lightlike-translation action of the modular group for the wedge tangent, without identifying the boosts) and the modular localization program (Brunetti–Fredenhagen–Verch), which assumes the BW form and reconstructs the spacetime; the interacting generalization remains a program, not a theorem.

Pitfalls¶

Reeh–Schlieder cyclicity is a THEOREM from locality + spectrum, not an axiom — do not list it among the Haag–Kastler axioms.
Local algebras are type III$_1$ (not type I) — there are no pure states, no traces, no finite matrix decomposition. The split property provides a type-I interpolation for separated regions, but the local algebra itself is III$_1$.
The Bisognano–Wichmann theorem is for the FREE field (and fields satisfying the BW wedge-algebra hypothesis); the interacting generalization (modular localization) is a program, not a theorem. Do not state "the modular group of the wedge equals the boosts" for an arbitrary interacting theory without the BW caveat.
The DHR superselection theory assumes a mass gap (localized charges in bounded regions); the massless / QED case fails — the electron is an infraparticle (no discrete one-particle pole, the charge is not localized in a bounded region due to the soft-photon cloud).
Haag duality $\mathcal{A}(\mathcal{O}')=\mathcal{A}(\mathcal{O})'$ is a non-trivial property, conjectured for "well-behaved" theories but not an axiom; failures (Buchholz–Summers) occur in theories with long-range forces or topological charges.
The split property requires the energy nuclearity bound; it can fail for theories with too many local degrees of freedom (e.g. very high-energy content per unit volume).

Refs¶

Haag, Local Quantum Physics (Springer, 2nd ed. 1996); Araki, Mathematical Theory of Quantum Fields
Bratteli–Robinson, Operator Algebras and Quantum Statistical Mechanics Vol. I (Tomita–Takesaki), Vol. II (modular theory)
Bisognano–Wichmann, "On the duality condition for a Hermitian scalar field" (CMP 1975); Borchers, "The CPT theorem in the theory of scalar quantized fields" (1966)
Buchholz, "Product states for local algebras" (CMP 1974); Buchholz–Wichmann nuclearity (CMP 1986); Doplicher–Haag–Roberts, "Fields, observables and gauge fields I–IV"
Brunetti–Fredenhagen–Verch, "Modular localization and Lorentz invariance" (2003)

Next¶

AQFT index · Foundations · Constructive QFT · CFT & TFT Axioms · Scattering Theory · Operator algebras · NCG spectral action · Mathematical physics