Wightman Fields, Domains, and Axioms

A Wightman field is not an operator $\phi(x)$ at each spacetime point. It is a continuous linear map from test functions to generally unbounded operators, all acting on one specified dense domain. The Wightman axioms couple that distributional statement to Poincaré covariance, positive energy, a vacuum, locality, and cyclicity. Keeping the domain in the definition is essential: without it, products of smeared fields and even the covariance law may be meaningless.

Required background. Domains, signatures, supports, and regularity supplies the common-domain discipline used below; positivity, spectrum, covariance, and locality separates the structural hypotheses; test functions, distributions, and support supplies distributional smearing; and unbounded operators, domains, closure, and adjoints supplies the operator-domain language.

Helpful background. Quantum fields as operator-valued distributions gives the physical motivation, while locally convex, nuclear, and rigged Hilbert spaces explains why Schwartz-space continuity is the natural topology.

Fields on a common invariant domain

Let $M=\mathbb R^{1,3}$ with metric $(+---)$, let $\mathcal S(M)$ be Schwartz space, and let $\mathcal H$ be a Hilbert space. For a Hermitian scalar field, the basic object is a linear map

$$\phi:\mathcal S(M)\longrightarrow \operatorname{Op}(\mathcal D), \qquad f\longmapsto\phi(f),$$

where $\mathcal D\subset\mathcal H$ is dense and every $\phi(f)$ maps $\mathcal D$ into itself. For all $\Psi,\Phi\in\mathcal D$, the matrix element

$$f\longmapsto \langle\Psi,\phi(f)\Phi\rangle$$

must be a tempered distribution. This weak continuity, together with a shared invariant $\mathcal D$, is what makes words such as $\phi(f_1)\cdots\phi(f_n)\Omega$ well-defined. The symbolic point field is only the distributional kernel of this map; it is not normally an operator on $\mathcal H$.

For a multiplet $\phi_a$, the index $a$ transforms in a finite-dimensional representation $S$ of the Lorentz cover. Charged fields occur with their adjoints, and the axiom is $\phi_a(f)^*\supseteq \phi_{a^*}(\bar f)$ on $\mathcal D$, not the assertion that every smeared field is bounded or self-adjoint. Essential self-adjointness, closability, and strong commutativity are additional conclusions requiring additional hypotheses.

The Wightman axioms

A standard four-dimensional formulation asks for the following data and properties; equivalent presentations package them differently. The precise classical formulation and its variants are given in Streater and Wightman 2016, § 3-1, pp. 96–101.

1. Hilbert space and Poincaré representation. A strongly continuous unitary representation $U(a,\Lambda)$ of the proper orthochronous Poincaré group, or its spin cover, acts on $\mathcal H$ and leaves $\mathcal D$ invariant.

2. Spectrum condition. If $U(a,1)=e^{iP\cdot a}$, the joint spectrum of $P^\mu$ lies in the closed forward cone $\overline V_+$. This is positive energy in every inertial frame, not merely $P^0\geq0$ in one frame.

3. Vacuum. There is a unit vector $\Omega$, invariant under $U$, usually unique up to phase. The vectors obtained by applying finite polynomials of smeared fields to $\Omega$ are dense. Uniqueness and cyclicity are logically distinct.

4. Covariance. With $f_{(a,\Lambda)}(x)=f(\Lambda^{-1}(x-a))$,

   $$U(a,\Lambda)\phi_a(f)U(a,\Lambda)^{-1} =S(\Lambda^{-1})_a{}^b\phi_b(f_{(a,\Lambda)})$$

   on $\mathcal D$, with the index convention adjusted consistently for the chosen representation.

5. Local commutativity. If the supports of $f$ and $g$ are spacelike separated, bosonic fields commute and fermionic fields anticommute on $\mathcal D$. For neutral scalar fields, $[\phi(f),\phi(g)]\Psi=0$ for every $\Psi\in\mathcal D$.

6. Temperedness and adjoints. The field is an operator-valued tempered distribution and the stated adjoint relation holds on the common domain.

The cluster property is often imposed or derived after assuming a unique vacuum and suitable spectral information; it is not interchangeable with locality. An equation of motion, a canonical commutation relation, a Lagrangian, a mass gap, and an asymptotic particle interpretation are not Wightman axioms.

Free massive scalar as a complete check

The corresponding physical treatment is quantum fields as operator-valued distributions. Here the first application is the exact axiom-level verification.

On symmetric Fock space over the positive-energy mass shell, let $\mathcal D_{\mathrm{fin}}$ be the finite-particle vectors and set

$$\phi(f)=a(K\bar f)+a^\dagger(Kf), \qquad (Kf)(\mathbf p)=\frac{\widetilde f(E_{\mathbf p},\mathbf p)}{\sqrt{(2\pi)^3\,2E_{\mathbf p}}}, \quad E_{\mathbf p}=\sqrt{\mathbf p^2+m^2}.$$

Creation and annihilation operators preserve $\mathcal D_{\mathrm{fin}}$, and their standard number-operator bounds make every matrix element continuous in the Schwartz topology. The second-quantized Poincaré representation preserves this domain, its translation spectrum consists of finite sums of future mass-shell momenta, and the Fock vacuum is invariant. Covariance follows from invariance of the mass-shell measure. Finally,

$$[\phi(f),\phi(g)] =i\int \mathrm d^4x\,\mathrm d^4y\,f(x)\Delta_m(x-y)g(y),$$

and the Pauli–Jordan distribution $\Delta_m$ is supported in the closed light cone, so the commutator vanishes for spacelike-separated supports. Polynomial Fock vectors generated from the vacuum are dense. This checks every item rather than treating the familiar mode expansion as a substitute for the axioms.

The same construction is developed distributionally in Wightman 1956, pp. 860–866 and systematized in Streater and Wightman 2016, §§ 3-1–3-3, pp. 96–116.

Boundaries of the formulation

Writing $\phi(x)\Psi$ at a point is an adversarial test: the delta distribution is not a Schwartz test function, so the axioms do not define that vector. Likewise, even when $\phi(f)$ and $\phi(g)$ separately act on a dense domain, the product is not defined unless the first factor maps the chosen domain into the domain of the second. A calculation that silently changes domains between steps has not established a field theory.

The axioms also do not say that products at coincident points exist. Composite fields require their own construction. Gauge potentials in a positive-metric Hilbert space, theories with indefinite metric, curved backgrounds, and low-dimensional braid statistics require modified frameworks; failure of this particular axiom system is not by itself inconsistency.

An independent check is dimensional and spectral. In four dimensions the free scalar has engineering dimension one, so $\phi(f)$ has the dimension of $f$ integrated against $\mathrm d^4x$; the normalization above agrees with the Lorentz-invariant measure $\mathrm d^3p/(2E_{\mathbf p})$. Every momentum produced from the vacuum is a sum of future-directed on-shell momenta, hence remains in $\overline V_+$ because that cone is convex.

Exercises

Show that locality for compactly supported test functions implies $\langle\Psi,[\phi(f),\phi(g)]\Phi\rangle=0$ whenever $\operatorname{supp}f$ and $\operatorname{supp}g$ are spacelike separated.

Solution

Local commutativity is an operator identity on the common domain: $[\phi(f),\phi(g)]\Phi=0$ for every $\Phi\in\mathcal D$. Pairing this zero vector with any $\Psi\in\mathcal D$ gives the stated matrix element. The converse from matrix elements follows because $\mathcal D$ is dense: if a vector has zero inner product with every $\Psi\in\mathcal D$, it is zero.

References

• Streater, Raymond F., and Arthur S. Wightman. 2016. PCT, Spin and Statistics, and All That. Princeton Landmarks in Physics. Princeton University Press. DOI.
• Wightman, Arthur S. 1956. “Quantum Field Theory in Terms of Vacuum Expectation Values.” Physical Review 101: 860–866. DOI.