• Register
PhysicsOverflow is a next-generation academic platform for physicists and astronomers, including a community peer review system and a postgraduate-level discussion forum analogous to MathOverflow.

Welcome to PhysicsOverflow! PhysicsOverflow is an open platform for community peer review and graduate-level Physics discussion.

Please help promote PhysicsOverflow ads elsewhere if you like it.


New printer friendly PO pages!

Migration to Bielefeld University was successful!

Please vote for this year's PhysicsOverflow ads!

Please do help out in categorising submissions. Submit a paper to PhysicsOverflow!

... see more

Tools for paper authors

Submit paper
Claim Paper Authorship

Tools for SE users

Search User
Reclaim SE Account
Request Account Merger
Nativise imported posts
Claim post (deleted users)
Import SE post

Users whose questions have been imported from Physics Stack Exchange, Theoretical Physics Stack Exchange, or any other Stack Exchange site are kindly requested to reclaim their account and not to register as a new user.

Public \(\beta\) tools

Report a bug with a feature
Request a new functionality
404 page design
Send feedback


(propose a free ad)

Site Statistics

145 submissions , 122 unreviewed
3,930 questions , 1,398 unanswered
4,847 answers , 20,601 comments
1,470 users with positive rep
501 active unimported users
More ...

Relation between Wilson approach to renormalization group and 'standard' RG

+ 7 like - 0 dislike

While studying renormalization and the renonormalization group i felt that there wasn't any completely satisfying physical explanation that would justify those methods and the perfect results they get.
Looking for some clarity i began to study Wilson's approach to renormalization; while i got a lot of insight on how a QFT works and what's the role of quantum fluctuations ecc i could not find a direct clear connection between the "standard" approach and the Wilson one. I'll try to be more specific.

To my understanding Wilson approach says (very) bascially this: given a quantum field theory defined to have a natural cutoff $\Lambda$ and quantized via path integrals (in euclidean spacetime)$$W=\int \mathscr D\phi_{\Lambda} \; e^{-S[\phi]}$$ it is possible to study the theory at a certain scale $\Lambda_N<\Lambda$ by integrating, in an iterative fashion, off high momentum modes of the field. Such rounds of integration can be viewed as a flow of the paramters of the lagrangian which is bounded  in it's form only by symmetry principles. For example, given a certain $$\mathscr L_0=a_0\; \partial_{\mu}\phi+b_0\;\phi^2+c_0\;\phi^4$$ we will get something like $$\mathscr L_N=\sum_n a_n \;(\partial_{\mu}\phi)^n+b_n\phi^n+\sum_{n,m}c_{nm}(\partial_{\mu}\phi)^n(\phi)^m$$ where the new parameters $a_n \quad b_n \quad c_{nm}$ have evolved from the original paramters via some relation which depends fromt the cutoff in some way. Now from some dimensional analysis we understeand that the operators corresponing to these parameters organize themselves in three categories which are marginal, relevant and irrelrevant and this categories are the same as renormalizable, super renormalizable and non renormalizable. Then there is the discussion about fixed points and all that stuff needed to have a meaningfull perturbative expansion ecc.

 My question(s) is (are):

 How do i put in a single framework the Wilsonian approach in witch the relations are between the parameters at the scale $\Lambda_N$ with those of the lagrangian $\mathscr L_0$ and their renomrmalization goup flow descibes those changes in scale with the "standard" approach in which we take $\Lambda\rightarrow+\infty$ and relate the bare paramters of the theory $g_0^i$ with a set of parameters $g_i$ via renormalization prescriptions at a scale $\mu$ and then control how the theory behaves at different energy scales using callan-symanzik equation ?

How different are the relations between the paramters in the wilson approach and the one in the "standard" approach? Are these even comparable?

What is the meaning (expecially in the wilsonian approach?) of sending $\Lambda$ to infinity besides of getting completley ridden of non renormalizable terms in the theory?

 Is, in the standard approach, giving a renormalization prescription which experimentally fixes the paramters  $g_i$ at a scale $\mu$ basically the same as integrating from $\Lambda\rightarrow +\infty$ to the scale $\mu$ in the wilsonian approach?

I'm afraid i have some confusion here, any help would be appreciated! 

asked May 20, 2015 in Theoretical Physics by fra (155 points) [ no revision ]
retagged May 20, 2015 by fra

1 Answer

+ 5 like - 0 dislike

Let's get some potential obstacles out of the way first. One theme that's central to all renormalization narratives is to relate the bare theory to physical observables, which is often manifested as the study of the mathematical relations between bare and renormalized couplings. This is what was studied from the beginning of QFT renormalization program, and this central point hasn't been changed ever since, in particular renormalization group perspective doesn't change it. What was troublesome at the time was the insistence on sending cutoff to infinity, which is more than often impossible(e.g. Landau pole), but today people no longer insist on that. Instead, we keep the weaker requirement that cutoff may stay but the low energy physics must be insensitive to the cutoff, for example we can construct our QFT such that S-matrix elements depend on $O(\Lambda^{-1})$, so we might as well just take $\Lambda\to\infty$, but beware this by no means suggest the cutoff can be consistently removed from all parts of our QFT(again, e.g. Landau pole). When the cutoffs can genuinely be removed, we say we have a completion, this is the case for Yang-Mils theory, for example.

When we say "Wilsonian renormalization group (philosophy)", especially when used in the sense of "We only understand what QFT is after Wilson"(somewhat an exaggeration in my opinion), we are typically referring to the realization that renormalization is a sort of coarse graining process (and indeed technically very close to what OP described in the 2nd last paragraph), that a completion of QFT corresponds to the existence of fixed points(modulo limit cycles that are usually not of interest), that nonrenormalizable operators naturally die out at low energy, and that renormalization is intimately connected to critical phenomena etc. These equip us with not only much better intuitions but also very powerful new tools for calculations.

The Wilson(-Polchinski) integrating-momentum-shell approach you referred to in your main post is just one of the many ways to do coarse graining, it's not absolutely necessary to use this approach to talk about Wilsonian philosophy, but it is a convenient one.

All said above involve changing the cutoffs, and are more or less attacking the same problem, with different levels of sophistication: find a family of theories/Lagrangians labeled by cutoffs, such that they give precisely the same prescribed physics at some energy scale lower than the cutoff. Then by changing the cutoff, your coupling constants, as functions of cutoff,  will form a trajectory. If you can make your cutoff arbitrarily large without any inconsistency, congratulations, you have a continuum completion; if not, it's not necessarily a catastrophe, maybe you are dealing with a situation with a physical cutoff, such a solid state lattice system, or that at certain scale a qualitatively different unknown theory has to step in.

As for Callan-Symanzik equations, there are several Callan-Symanzik-like equations, and I suspect all of them have been referred to as "Callan-Symanzik equation" in literatures.

For example, one may ask, as discussed at length above, if we keep the renormalized couplings(at a fixed energy scale) fixed, how would bare couplings change as a function of cutoff. One may also ask if we keep the bare couplings constant, how would renormalized couplings(at a fixed energy scale) change as a function of cutoff. Clearly these two are mathematically equivalent, they both yield  Callan-Symanzik-like equations, and both are in accordance with the philosophy described in the lengthy discussion above.

The last sort keeps bare couplings fixed, cutoff fixed, but ask how the renormalized couplings change as the reference energy scale changes, this is what you are looking for, but unfortunately I myself am not sure if this is equivalent to the former two Callan-Symanzik-like equations, all I know is that  the beta functions of all these three equations have the same power series expansion coefficients for the two leading terms, so at least there should be some weak equivalence in the practical sense. I'll leave it to who knows better.



1. I should add that in the famous review by Wilson and Kogut on RG and epsilon expansion, they call the curve that is produced by taking $\Lambda$ to infinity "canonical curve", while they call the curve produced by changing the coarse-graining/measurementl scale ($\Lambda_N$ or $\mu$ in OP's noatation) "RG trajectory", so they made it quite explicit that the two are in principle two different things. 

2. I now understand exactly why perturbatively the canonical curve  and RG trajectory have the same form of $\beta$ function up to the first two orders, it's very lucidly explained in the brilliantly written book "Quarks, gluons and lattices" by Michael Creutz: 

For a theory with only a dimensionless coupling, any dimensionless observable H must have the functional dependence on cutoff scale $\Lambda$, measurement scale $\mu$ and the bare coupling $g_0(\Lambda)$ in the particular form of $H(\mu,\Lambda, g_0)= H(\frac{\mu}{\Lambda},g_0)$, by simple dimensional analysis. This immediately gives

$$\Lambda \frac{\partial}{\partial\Lambda} H=-\mu\frac{\partial}{\partial\mu}H$$

The renormalized coupling is always defined as the value of some observable at certain scales, so it also satisfies 

$$\Lambda \frac{\partial}{\partial\Lambda} g_R=-\mu\frac{\partial}{\partial\mu} g_R$$

On the other hand, as we push the cuttoff to infinity, the $g_R$ should gradually become a constant with respect to $\Lambda$. (Of course, if keeping $g_R$ fixed at a certain scale is one your renormalization conditions, then this is automatically true for all values of $\Lambda$.) This means 

$$0=\Lambda \frac{d}{d\Lambda} g_R=\Lambda \frac{\partial}{\partial\Lambda} g_R+\frac{\partial g_R}{\partial g_0} \Lambda  \frac{d}{d\Lambda} g_0.$$

Put together them we have 

$$\beta^R(g_R)\equiv \mu\frac{\partial}{\partial\mu} g_R=\frac{\partial g_R}{\partial g_0} \Lambda  \frac{d}{d\Lambda} g_0 \equiv\frac{\partial g_R}{\partial g_0} \beta(g_0).$$

On top of all these, to deserve the name "renormalized coupling", $g_R$ should be defined such that it equals $g_0$ at leading order, so $g_R= g_0+bg_0^3+O(g_0^5)$.

At this stage one can apply the standard calculation to show why the first two coeffiencients in the expansion $\beta^R(g_R)$ (in $g_R$) are the same as those of $\beta(g_0)$ (in $g_0$). 

answered May 21, 2015 by Jia Yiyang (2,465 points) [ revision history ]
edited Oct 23, 2016 by Jia Yiyang

Nice answer,  Jia Yiyang, would  you please give me some references in the literature about these matters?

@fra, I personally don't know any all-encompassing introduction, but http://arxiv.org/abs/cond-mat/0702365 is a very decent one.

Thank you so much!

Your answer

Please use answers only to (at least partly) answer questions. To comment, discuss, or ask for clarification, leave a comment instead.
To mask links under text, please type your text, highlight it, and click the "link" button. You can then enter your link URL.
Please consult the FAQ for as to how to format your post.
This is the answer box; if you want to write a comment instead, please use the 'add comment' button.
Live preview (may slow down editor)   Preview
Your name to display (optional):
Privacy: Your email address will only be used for sending these notifications.
Anti-spam verification:
If you are a human please identify the position of the character covered by the symbol $\varnothing$ in the following word:
Then drag the red bullet below over the corresponding character of our banner. When you drop it there, the bullet changes to green (on slow internet connections after a few seconds).
To avoid this verification in future, please log in or register.

user contributions licensed under cc by-sa 3.0 with attribution required

Your rights