Why fields are distributions

In QFT, $\phi(x)$ looks like an operator attached to a point. Taken literally, that notation is often too singular. The useful object is a field averaged against a smooth test function:

$$\phi(f)=\int d^dx\, f(x)\phi(x).$$

An operator-valued distribution assigns an operator to each suitable $f$, linearly and continuously in a specified test-function topology. This formulation retains locality while giving precise meaning to the expressions we use.

The delta function is the simplest model

Prerequisites: Fourier transforms and integration by parts.
Outcomes: define a field by smearing; distinguish ultraviolet and operator-domain questions; show why convergence of distributions need not preserve products.

The delta distribution is defined through its action,

$$\delta(f)=f(0).$$

It is not an ordinary function that is “infinite at zero.” Likewise, its derivative is defined by $\delta'(f)=-f'(0)$. We can manipulate these objects reliably because their action on test functions is specified.

Quantum fields follow a similar logic, except their values after smearing are operators. These operators are generally unbounded and act on a common suitable domain. The word “distribution” handles dependence on spacetime; it does not remove the separate operator-domain problem.

Why smearing matters physically

For a free Euclidean scalar field, the formal covariance is

$$C_m(x-y)=\int\frac{d^dp}{(2\pi)^d} \frac{e^{ip\cdot(x-y)}}{p^2+m^2}.$$

At coincident points the high-momentum integral can diverge. With Schwartz test functions $f,g$, the smeared covariance instead contains rapidly decreasing Fourier transforms:

$$C_m(f,g)=\int\frac{d^dp}{(2\pi)^d} \frac{\overline{\widetilde f(p)}\widetilde g(p)}{p^2+m^2}.$$

For $m>0$ this integral is well defined. The ultraviolet divergence of $C_m(0)$ does not prevent the smeared free field from existing.

Smearing is not merely an instrumental resolution trick. It is part of the mathematical definition. The smearing function also tells us where the observable is supported, so locality can be expressed by the separation of supports.

Products are a new problem

If $\phi$ is a distribution, the expression $\phi(x)^2$ is not automatically defined. Knowing two distributions separately does not guarantee that their pointwise product exists.

At a finite cutoff, a product such as $\phi_a(x)^2$ may be perfectly finite. Removing the cutoff can require subtraction or renormalization. In a Gaussian model, Wick ordering motivates

$$:\!\phi_a(x)^2\!: =\phi_a(x)^2-\mathbb E[\phi_a(x)^2].$$

Whether a smeared limit exists, in which sense, and with what properties are further questions. A subtraction that works in one model is not a universal construction recipe.

This is directly relevant to Yang–Mills theory. Gauge-invariant expressions such as $\operatorname{tr}F_{\mu\nu}F^{\mu\nu}$ are natural physical observables, but constructing their quantum versions requires control of composite operators. Writing the classical expression does not complete that task.

A small calculation

Let $\delta_\epsilon(x)=\epsilon^{-1}\rho(x/\epsilon)$ on the line, where $\rho$ is smooth, compactly supported, nonnegative, and integrates to one. Then

$$\int \delta_\epsilon(x)f(x)\,dx =\int \rho(u)f(\epsilon u)\,du\longrightarrow f(0).$$

However,

$$\int\delta_\epsilon(x)^2f(x)\,dx =\epsilon^{-1}\int\rho(u)^2f(\epsilon u)\,du$$

generically diverges. Convergence of each factor as a distribution did not imply convergence of the product.

Exercise. What information about $\rho$ appears in the leading divergence?

Solution. It is $\epsilon^{-1}f(0)\int\rho(u)^2du$. Its dependence on the regulator illustrates why defining a product requires additional prescriptions.

Continue

The operator-domain bridge explains the independent domain requirement for smeared quantum fields. Weak convergence and tightness explains why a limit must be taken in a declared space and topology.

Read the continuum-limit lesson and Gaussian measures. For the free-field starting point, see David Tong’s QFT lecture notes, chapter 2. Jaffe and Witten’s problem description, sections 3–4, places local quantum fields and gauge-invariant composites in the target construction.