I am reading Field Theories of Condensed Matter Physics by Fradkin and in equation (4.10) it shows that an operator transforms irreducibly under scalings as

$$\phi_n(xb^{-1}) = b^{\Delta_n}\phi_x(x)$$

where $\Delta_n$ in the so called scaling dimension of the operator. I am confused as to how exactly I should be thinking about this quantity. I understand homogeneous functions and scale invariance, but I am just wondering how the scaling dimension fits into it. Is it telling you the degree to which the range of the operator expands or contracts based on a scaling of the domain? How is this quantity thought of in practice?

For example, I could define in one dimension $\phi(x) = x^2$, and scale it as $\phi(xb^{-1}) = (xb^{-1})^2 = b^{-2} \phi(x) = b^{\Delta_n}\phi(x)$ and so the scaling dimension would be $\Delta_n = -2$. I can see its value and that its related to the exponent in $x^2$, but should I be thinking of this as any deeper?



1 Answer 1


The equation in Fradkin's book does not make much sense as written. He is trying to write that under a renormalization group transformation, the field transforms according to a particular rescaling. The field on the left of the equation is therefore not the same as the field on the right; if the equation were taken too literally, then one could differentiate both sides with respect to $b$ and obtain a differential equation giving an explicit form of the field. But clearly this does not make sense, these are supposed to be integration variables in the path integral!

Here's how I would fix the equation. In one iteration of the momentum-shell RG, one first writes $\phi(x) = \phi^<(x) + \phi^>(x)$, integrates out the fast-momenta contained in $\phi^>(x)$, and then rescales $x \to x' = x/b$ to restore the original cutoff. Upon rescaling, we obtain a new field $\phi'(x')$ defined by $$ \phi'(x') = b^{\Delta} \phi^<(x) $$ This new field contains Fourier modes up to the original cutoff. To zeroth order in the perturbative RG, a coupling of such a field will transform under a single RG step as $$ \int d^d x \ \lambda \phi(x) \to \int d^d x\ \lambda \phi^<(x) = \int d^d x' \ b^{d-\Delta} \lambda \phi'(x') = \int d^d x' \ \lambda' \phi'(x') $$ where I've introduced $\lambda' = b^{d-\Delta} \lambda$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.