The Anomalous Dimension

We thought we knew the dimension of an operator. You write down a field, $\phi$, count the usual suspects — mass dimension, spin, the bare arithmetic of units — and file it away. Dimension is dimension. A scalar is 1, a fermion is $3/2$, and any composite built from them is just the sum of its children. This is the tree-level faith. It is the faith of free theories, where particles never talk to one another, where the vacuum is a perfect mirror and every correlation function is just a pairing of legs.

Then you turn on the coupling. Just a whisper of interaction. You integrate over a loop. You regularize, renormalize, subtract the divergence with whatever scheme the gods of your generation favor. And when you look at the scaling dimension again — not the bare one, but the full renormalized, dressed, living one — it has moved. The operator has acquired a correction. It has grown. We call it the anomalous dimension, $\gamma$, as if it were an anomaly, a deviation, a mistake in the accounting. But it is not a mistake. It is the theory telling you that the operator is no longer just what you named it.

The one-loop correction to a scalar field's two-point function is a familiar ritual. You draw the diagram. It has a propagator, a vertex, a loop momentum to integrate. The integral diverges, but dimensional regularization saves you. You shift to $d = 4 - \epsilon$, the divergence appears as a $1/\epsilon$ pole, and the counterterm cancels it. But the finite piece remains. It modifies the wavefunction renormalization $Z_\phi$, and from $Z_\phi$ you read the anomalous dimension:

$$\gamma = \frac{1}{2} \mu \frac{\partial \ln Z_\phi}{\partial \mu}$$

This is the number that matters. It is small when the coupling is small. It grows when the coupling grows. It is the fingerprint of quantum fluctuations dressing the bare operator like a saint's relic layered in gold.

In a conformal field theory, where the beta function vanishes and the coupling is fixed, the anomalous dimension is a fixed number — a property of the operator, like its charge or its spin. The operator lives in a representation of the conformal group, and its scaling dimension $\Delta$ is the eigenvalue of the dilation operator:

$$D = D_{\text{classical}} + \gamma(g^*)$$

The classical part is the name you gave it. The anomalous part is what the quantum world did to it while you weren't looking. In the Ising model, the scaling field that corresponds to the temperature operator acquires an anomalous dimension at the Wilson-Fisher fixed point. The field that corresponds to the spin operator acquires another. These numbers — measured experimentally, computed in $\epsilon$-expansion, computed by Monte Carlo — they tell you how correlations decay. They tell you the critical exponents. They are the observables.

But here is what the textbooks do not say clearly enough: the anomalous dimension is not a correction to something that was "really" a certain value. It is the actual dimension. There is no hidden classical operator waiting to be discovered. The quantum operator is the thing. The name you gave it in the Lagrangian was just a starting point, a seed that grows in the garden of renormalization. The anomalous dimension is the flower.

In four-dimensional $\phi^4$ theory, at weak coupling, $\gamma \sim g^2/(16\pi^2)$. In QCD, the anomalous dimensions of twist-2 operators control the scaling violations of deep inelastic scattering structure functions. The $Q^2$ dependence of the moments of $F_2(x, Q^2)$ is not a bug — it is the logarithmic running of the anomalous dimension, the echo of quantum fluctuations across scales. The data from SLAC, from HERA, from the deep inelastic experiments that mapped the proton's partonic soul — they measure anomalous dimensions.

In two dimensions, where the conformal algebra is infinite-dimensional, the anomalous dimension takes on a special character. The operator product expansion is organized by scaling dimension, and the leading operator in the OPE of $\phi \times \phi$ is the identity, then the stress tensor, then a tower of higher-spin operators. The anomalous dimensions of these operators — which are simply the scaling dimensions in 2D — determine the spectrum of the theory. In the Ising CFT, the scaling dimension of the spin operator is $1/8$. The thermal operator has dimension $1$. The energy-momentum tensor has dimension $2$. These are not integers. They are not what you would compute from free fields. They are the true dimensions, anomalous and exact.

Sometimes the anomalous dimension is large. In certain strongly coupled theories, operators that you would expect to be relevant become irrelevant because their anomalous dimension is so large that $\Delta > d$. The theory is dragged away from the Gaussian fixed point by the interaction, and the "bare" operator is unrecognizable. This happens in the large-$N$ vector models. It happens in the AdS/CFT correspondence, where the dimension of a bulk field is related to the curvature radius in a way that has no classical interpretation.

The anomalous dimension is the record of the theory's memory. Every quantum fluctuation that dresses the operator leaves a trace. The dimension is heavier than it was at tree level. The operator is more complex. It is a bound state of the field you named and the vacuum it disturbs. To compute it, you integrate over the virtual particles that populate the vacuum, the sea of fluctuations that exists at every scale. The dimension is the weight of that sea.

We compute it with Feynman diagrams, yes. But we should understand what we are computing. We are measuring how much the universe has changed the operator. The bare name is a convenience. The renormalized dimension — classical plus anomalous — is the truth. And the anomalous part, the part that depends on the coupling, on the scheme, on the very existence of the vacuum, is where the physics lives.