Haag-Theorem Variants, Domains, and Proposed Evasions

Haag-type results come in several forms, but proposed “evasions” are usually constructions outside a theorem’s hypotheses: a regulator removes exact locality or covariance, a formal series is not a unitary operator, or an algebraic theory never posits a global free–interacting intertwiner. The right conclusion is to identify the changed assumption and then ask whether the regulated or alternative construction has a controlled physical limit.

Required background. Haag’s theorem and inequivalent representations gives the core no-go result; counterexamples, nonconverses, and hypothesis tests supplies the logical discipline; and domains, signatures, supports, and regularity keeps equal-time and spacetime domains explicit.

Helpful background. Interacting fields, asymptotic observables, and effective descriptions separates exact fields from asymptotic ones; effective field theory as a controlled expansion explains cutoff-dependent predictions; and Haag’s theorem: physical meaning and scope provides orientation.

Which theorem is being tested?

A Haag claim is meaningful only after specifying at least:

• the field class—neutral scalar, general Wightman fields, canonical fields, or local algebras;
• the spacetime dimension and whether a time-zero restriction exists;
• the common invariant domains of fields and conjugate momenta;
• the exact intertwining relation and whether it holds at one time or all times;
• irreducibility, vacuum uniqueness, Euclidean or Poincaré covariance, and spectrum assumptions;
• the conclusion—vacuum matching, equality of low-order correlators, freeness, or disjointness of representations.

The Haag–Hall–Wightman variant and its proof limits are stated in Earman and Fraser 2006, §§ 3–4, author manuscript pp. 8–15, while field-theoretic variants and Borchers classes are treated in Streater and Wightman 2016, §§ 4-5–4-6, pp. 161–174.

Changing a hypothesis may be exactly the right way to formulate interacting physics. It is not, however, a disproof of the theorem that used that hypothesis.

Three constructions compared

Finite-volume, finite-cutoff λφ⁴

With a spatial box, lattice or momentum cutoff, and finitely many modes, the system is a quantum-mechanical collection of oscillators. A Hamiltonian

$$H_{\Lambda,L}=H_{0,\Lambda,L} +\lambda\int_{L^3}\mathrm d^3x\,:\!\phi_{\Lambda,L}(x)^4\!:$$

can be defined on a common oscillator Hilbert space under standard operator-domain conditions. Finite-dimensional canonical uniqueness permits unitary identifications unavailable in the continuum. But the cutoff breaks exact local field behavior or Lorentz covariance, the box breaks translations and boosts, and the finite-mode system is not a Wightman field on Minkowski space. The decisive question is whether correlation functions converge after renormalization as $\Lambda,L\to\infty$; the finite-cutoff unitary need not converge.

Gell-Mann–Low formal series

The normalized expression

$$\frac{\langle\Omega_0,T\{\phi_I(x_1)\cdots\phi_I(x_n) \exp[-i\int \mathrm d^4x\,\mathcal H_I(x)]\}\Omega_0\rangle} {\langle\Omega_0,T\exp[-i\int \mathrm d^4x\,\mathcal H_I(x)]\Omega_0\rangle}$$

is normally interpreted order by order after regularization and renormalization. A formal power series is not a bounded or densely defined unitary operator on free Fock space, and the adiabatic/infinite-volume limits may not exist there. The interaction picture’s clash with Haag’s assumptions is analyzed in Earman and Fraser 2006, § 5, author manuscript pp. 15–20. Perturbation theory can yield controlled coefficients without supplying the forbidden exact global intertwiner.

Algebraic interacting net

An algebraic QFT assigns local observable algebras $\mathcal A(O)$ to spacetime regions, with covariance, isotony, and locality. States produce Hilbert-space representations by the GNS construction. An interacting net need not be represented as $V\mathcal A_0(O)V^{-1}$ for one global unitary $V$ acting on free Fock space. Different physically relevant states or phases may generate inequivalent representations. The framework therefore does not assume the target of the no-go theorem; it must instead establish its own existence, locality, covariance, and spectral properties.

These three cases connect to ultraviolet sensitivity and the renormalization problem: regulators and formal expansions can be useful precisely because their intermediate objects are not exact continuum Wightman intertwiners.

Genuine and spurious evasions

Asymptotic in/out fields are free but are not asserted to equal the interacting Heisenberg field at finite time under one unitary. Local perturbative constructions may compare algebras only in compact regions and order by order. Constructive models may use an interacting vacuum representation from the outset. Each is a coherent change of setup.

By contrast, writing a time-dependent $V_\Lambda(t)$ for a cutoff theory and calling it a counterexample is spurious if the claimed theorem concerns exact covariant continuum fields. At fixed cutoff, hypotheses are missing; after removing the cutoff, existence and unitarity of the limit remain to be proved. Likewise, merely saying “the domains differ” is not an explanation until the affected field products and theorem step are identified.

An independent check is to apply every candidate intertwiner to the vacuum. If $V$ genuinely intertwines the Euclidean symmetry representations and both invariant vacua are unique, then $V\Omega_1$ must be proportional to $\Omega_2$. A cutoff Dyson operator that sends the free vacuum to a vector with a divergent norm or no infinite-volume limit already fails this necessary consequence.

Exercises

Classify the status of a finite-cutoff unitary $V_{\Lambda,L}(t)$ whose matrix elements converge order by order in $\lambda$ but whose strong operator limit is unknown.

Solution

It is a valid regulated or perturbative object. It is neither an exact unitary on the continuum Hilbert space nor a counterexample to a continuum Haag theorem. One must separately prove existence of the regulator limit, preservation of the relevant field domains, covariance and locality, and strong/unitary convergence.

References

• Earman, John, and Doreen Fraser. 2006. “Haag’s Theorem and Its Implications for the Foundations of Quantum Field Theory.” Erkenntnis 64: 305–344. Author manuscript.
• Streater, Raymond F., and Arthur S. Wightman. 2016. PCT, Spin and Statistics, and All That. Princeton Landmarks in Physics. Princeton University Press. DOI.