MIT8 324F10 Lecture23
MIT8 324F10 Lecture23
MIT8 324F10 Lecture23
Lecture 23
dλi
Λ = 0, (1)
dΛ
meaning λi (Λ) = λ∗i is a constant, independent of scale. At a fixed point, a theory is scale invariant.
As a trivial example of this second case, we will consider a Gaussian fixed point. Let the Lagrangian be given by
1 2
L = (∂ϕ) , (2)
2
describing a free, massless particle. In this case, Si = 0, and all of the couplings satisfy λi = 0, and this is clearly
invariant under the flow. In fact, our earlier discussion of RG flow and the notion of relevant and irrelevant couplings
are defined implicitly with respect to this fixed point.
L = L0 + Lint , (3)
2 ∑
where L0 = 12 (∂ϕ) and Lint = i gi Oi . The dimensions ∆i of Oi are found from the canonical dimension of ϕ,
which is defined using L0 . For a generic fixed point, λi (Λ) = λ∗i constant, consider a small deviation from the fixed
λ6
λ4
Figure 1: Flow diagram for the dimensionless couplings, λ4 and λ6 . λ4 is marginally irrelevant.
point,
δλi = λi − λ∗i . (4)
Then,
dδλi ({ ∗ })
Λ = βi λj + δλj
dΛ
∑ dβi
= δλj + . . .
j
dλj α�
j
∑
= Tij δλj + . . .
j
1
Lecture 23 8.324 Relativistic Quantum Field Theory II Fall 2010
d�i (�)
with Tij = dαj . We now consider the left eigenvectors, ei , of T , with eigenvalues t� , where α indexes the
α�
j
eigenvectors. That is, ∑ (�) (�)
ei Tij = t� ej . (5)
i
Then
dU�
Λ = t� U� + O(U 2 ). (7)
dΛ
(0)
Disregarding the non-linear terms, we suppose at some Λ0 that U� = U� . Then we have that
( )tα
Λ
U� (Λ) = U�(0) . (8)
Λ0
U� is the scaling variable, and −t� = [U� ] is the scaling dimension, as defined with respect to the fixed point {λ∗ } .
We have that
> 0 relevant,
−t� = [U� ] = = 0 marginal, (9)
< 0 irrelevant.
Now,
L = L∗ + Lint , (10)
where
∑
L∗ = λ∗i Oi ,
i
∑ ∑
Lint = δλi Oi = U� Õ� ,
i �
[ ]
˜ � are the scaling operators with respect to L∗ . We have that Õ� = d + t� . Now consider the case of U1 ,
where O
U2 with t1 < 0 (relevant) and t2 > 0 (irrelevant). Then
( ) ( )
U1 δλ1
=M . (11)
U2 δλ2
The critical surface is the submanifold in the space of couplings spanned by irrelevant couplings near a fixed
U2 λ2
U1 λ1
Figure 2: Flow in the U1 , U2 plane and the λ1 , λ2 plane about the fixed point λ∗ .
point. Any point on the critical surface will flow to the fixed point. A point lying off the critical surface will flow
away from it, as shown in figure 3. The critical surface has a finite co-dimension, since the number of relevant and
marginal couplings is always finite.
2
Lecture 23 8.324 Relativistic Quantum Field Theory II Fall 2010
Figure 3: The critical surface submanifold. Couplings on the manifold are irrelevant and flow towards the fixed
point. Couplings off the manifold flow away from it.
2 ∑
Let us consider, at the ultraviolet scale, Λ0 , L = L0 + Lint , with L0 = 1
2 (∂ϕ) , and Lint = i λi Oi giving
deviations from the fixed point.
2
Oi = ϕ2 , (∂ϕ) , ϕ4 , . . . (12)
are scaling operators with respect to L0 . In general near {λ∗i } ,
the spectrum of scaling operators {O� } and their
conformal dimensions will be very different from those at the Gaussian fixed point. Such non-trivial fixed points
are very rare at a finite distance away from the Gaussian fixed point, particularly in d ≥ 3. For this reason, they
provide a very limited theoretical tool. We consider one famous example:
1 2 λ
L = (∂ϕ) + ϕ4 (13)
2 4!
at Λ0 in d = 4. Then we have
βα = cλ2 + . . . , (14)
where c is a positive constant, giving a marginally irrelevant flow back to the Gaussian fixed point. If we now
consider d = 4 − ϵ, then
d−2 ϵ
[ϕ] = = 1 − , [λ] = ϵ, (15)
2 2
for ϵ ≪ 1. Then we have
βα = −ϵλ + cλ2 + . . . , (16)
and so, the fixed point is given by
ϵ
λ∗ = ≪ 1. (17)
c
This fixed point is infinitesimally close to the Gaussian fixed point and can be studied using perturbation theory.
We now conclude our discussion of the Wilsonian approach to renormalization. Our discussion has been more
focused on the conceptual picture, rather than the explicity calculation. The Wilsonian approach is conceptually
very appealing, but technically very awkward, as the couplings depend on the scale. We want to find how to see it
from the conventional approach.
3
Lecture 23 8.324 Relativistic Quantum Field Theory II Fall 2010
where
1 2 1 g ϵ
L = − (∂ϕ) − m2 ϕ2 + µ 2 ϕ3 (19)
2 2 6
and
1 2 1 g ϵ
Lct = − A (∂ϕ) − Bm2 ϕ2 + µ 2 Cϕ3 . (20)
2 2 6
2. We impose the following renormalization conditions to determine the counter terms:
Π(k 2 ) = (21)
k
k2
V (k1 , k2 , k3 ) = (22)
k3
k1
V (0, 0, 0) = g (Physical coupling condition) (23)
where K = K(m2 , k1 , k2 , k3 , µ). The explicit form of this function is not important here. In order to cancel the
divergences, we require
α
A = −+ a,
6ϵ
α
B = − + b,
ϵ
α
C = − + c.
ϵ
where a, b and c are finite constants determined by the renormalization conditions. Therefore, Π(k 2 ) and V (k1 , k2 , k3 )
are independent of the auxillary scale µ and depend only on m2 and g. But we can actually choose a, b and c
arbitrarily, and physical observables will be finite. The previously stated renormalization conditions will generally
not be satisfied, and so m2 , ϕ and g are no longer the physical mass, field and couplings. They can be interpreted
as parameters of the Lagrangian. The conditions given by
Π(k 2 = −m2phys ) = 0,
′
Π (k = 2
−m2phys ) = 0,
V (0, 0, 0) = gphys ,
4
Lecture 23 8.324 Relativistic Quantum Field Theory II Fall 2010
Here, the left-hand side is µ-independent. The splitting on the right-hand side requires introducing a scale µ. ϕ,
g and m2 should be µ−dependent, so that the physical observables are µ−independent, but the finite part of the
split is arbitrary. A different choice of a,b and c is a different choice of scheme: the one we have been using until
now is the on-shell scheme. Different schemes give different g(µ), ϕ(µ) and m2 (µ). Again, the physics should not
depend on such a choice, corresponding to the different ways of doing the splitting.
5
MIT OpenCourseWare
http://ocw.mit.edu
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.