Matthew Schwartz 2012

Appendix A: Conventions and Signs

which turns joules into inverse seconds. This makes all quantities have dimensions of energy (or mass, using E = mc2) to some power. Quantities with positive mass dimension, (e.g. momentum p) can be thought of as energies and quantities with negative mass dimension (e.g. position x) can be thought of as lengths. 1 Sometimes we write the mass dimension of a quantity with brackets, as in [p] = x = M 1 meaning these quantities have mass dimension 1. Other examples are [dx] = [x] = [t] = M 1 [ ] = [p ] = M 1 [velocity] = Thus The action should be a dimensionless quantity [S] = [ d4x L] = M 0 So Lagrangians (really, Lagrangian densities) have dimension 4 [L] = M 4 For example, a free scalar eld has Lagrangian L = 2 so [] = M 1 (9) and so on. In general bosons, whose kinetic terms have 2 derivatives, have mass dimension 1 and 3 fermions, whose kinetic terms have 1 derivative, have mass dimension 2 . You can always put the and c factors back by dimensional analysis. For example, a cross section has units of area, which might measured in picobarns (pb), with 1 picobarn = 1040 meters2 A quantum eld theory calculation might produce something like = m2
P

(3) (4) (5) (6)

[x] = M0 [t]

[d4x] = M 4

(7)

(8)
1

(10)
1 1 , G eV 2

where (11)

1 GeV =1.6021010 joules

So we need a combination of and c with which converts GeV 2 into area. The unique answer is 2c2 = 9.996 1052 joules2 meters2. Thus 1 GeV2
2 2

c = 3.8941032 meters2 = 3.894108 picobarns

(12)

which is a useful conversion factor. 1

Section 3

2 2s
Keeping the factors of 2 straight is important. The origin of all the 2s is the relation

(x) =

dp e2ipx

(13)

This identity holds with either sign; our sign convention for quantum elds is discussed below. To remove the 2 from the exponent, we can rescale either x or p. Since position generally is not an angular coordinate, it makes sense to rescale p. Then

dp eipx = 2(x)

(14)

This relation is so important you should have it completely internalized. Our convention for Fourier transforms is that momentum space integrals have while position space integrals have no 2 factors: f (x) = d4 p f (p)eipx , (2)4
d4 p

1 2

factors

f (p) =

d4xf(x)ei px

(15)

Thus you should get used to writing (2)4 in momentum space integrals. Although physical quantities dont care about our 2 convention, the factors of 2 have important physical eects. If you nd yourself with a lot of free time, try computing something like g 2 = in QED using a dierent 2 convention to check that you get the same answer.

3 Signs
Although the meat of most calculations is independent of the signs, physical results care very much that you get the sign right. Here we tabulate some of the signs in important equations. First of all, we will never use curved space backgrounds, so the metric g and the Minkowski metric are interchangeable. The metric we use has sign convention 1 1 g = = (16) 1 1

This convention makes p2 = p2 Q 2 = m2 > 0. The alternative g = diag(1, 1, 1, 1) makes p2 < 0. p 0 The sign of terms in Lagrangians is set so that they have positive energy density. It is easiest to remember the signs by writing Lagrangian is L = Lkin V where V is the potential energy which should be positive in a stable system. For example, for a scalar eld the mass term 1 2 2 1 1 m should give positive energy, so V = 2 m2 2 and L = 2 m2 2. The kinetic term sign can 2 2 then be recalled from p2 = in Fourier space and p2 = m2 on-shell, so that the equations of motion should be ( + m2) = 0. Therefore we have 1 1 1 L = ( + m2) = ( )( ) m2 2 2 2 2
1

(17)

The factor of 2 makes the kinetic term contribute ( + m2) to the equations of motion (instead of 2( + m2)). For a complex scalar, the Lagrangian is L = ( + m2) = ( )( ) m2 without the 2 , since now variation with respect to will give ( + m2). For gauge bosons, the Lagrangian is 1 2 1 1 1 1 L = F = A A + AA = A A A ( )A 4 2 2 2 2 (19)
1

(18)

Signs

where F = A A . In this equation and many others we employ the modern summation convention where contracted indices can be raised or lowered without ambiguity: x p = x p = x p = x p . All of these contractions are equal to g x p = gx p . 1 The sign and normalization of the 4 factor in Eq.(19) be understood as follows. In Lorenz 1 1Q 1 Q gauge A = 0 the Lagrangian is just L = 2 A A = 2 A0 A0 2 A A . This gives the three Q , which actually contain the propagating transverse degrees of freedom, the spatial components A same kinetic term as for scalars. (Why the scalar component A0 with the wrong sign is not problematic is discussed in Lecture II-1.) Dirac fermions are normalized so that L = (i e A m) (20) where = and A = A . As in the scalar case, the m is xed so that the corresponding energy density is positive. The covariant derivative is D = ieQA (21)

where e is the strength of the force, and Q = 1 is the electrons U (1) quantum number, which leads to the sign in the eA term. For simplicity, in QED we just write D = + ie A with e = 0.303 in dimensionless units. Another common convention is that e is taken to be negative and Q is omitted. Peskin and Schroeder and Srednicki both do this, but have dierent signs for the interaction, Lint = e A and Lint = e A in Eqs. (4.3) and (58.2) respectively. Zee has A in Eq. (11.7.1). Lint = e These Lagrangians lead to propagators 0|T {(x)(y)}|0 = for a real scalar and 0|T {A (x)A (y)}|0 =
d4 p i p (xy) i g (1 ) e p2 + i (2)4 p p p2

d4 p ip (x y) i e (2)4 p2 m2 + i

(22)

(23)

for a massless spin 1 eld in covariant gauges (see Lecture II-1). The i the photon propagator 1 1 versus the +i in the scalar propagator is the same sign dierence as in L = 2 + 2 A A . The Dirac fermion propagator is 0|T {(x) (y)}|0 = d4 p ip (x y) i e = (2)4 p m + i d4 p ip (xy) i(p + m) e p2 m2 + i (2)4 (24)

Here we write (x) (y) not (x)(y) since this is a matrix in spinor space, like the product of ( ). a row vector and a column vector gives a matrix When we expand elds in terms of creation and annihilation operators, we write for a single real scalar eld 1 d3 p px pQ (x) = a p(t)ei Q x + a p(t)ei Q Q (25) (2)3 2p where p Q 2 + m2 . Including the free-eld time dependence and generalizing to the complex p case, this becomes d3 p 1 a peipx + bpeipx (26) (x) = (2)3 2 p (x) = Similarly, we take (x) =
s

d3 p (2)3 d3 p (2)3 d3 p (2)3

1 a peipx + bpei px 2 p 1 (as us ei px + bs v s eipx) p p p p 2p 1 s s (as u p eipx + bs v p ei px) p p 2 p

(27) (28)

(x) =
s

(29)

Section

The sign of the phases follows from a(t) = eita(0) for annihilation operators by Heisenbergs equations of motion in any simple harmonic oscillator.

4 Dirac algebra
The Dirac matrices satsify { , } = 2g . We dene 5 i 0 1 2 3 which leads to { 5, } = 0. We also dene i = [ , ] 2 g g = 4 Tr[ ] = Tr[5] = Tr[ ] = 0 Tr[ ] = 4(g g g g + g g ) The projectors are PL = 1 5 , 2 PR = 1 + 5 2 (35) (31) (32) (33) (34) (30)

Some useful identities are

so that left-handed elds satisfy 5 L = L and right-handed elds satisfy 5 R = R. Spinor sums are, for particles:
2

us(p)us(p) = p + m and for antiparticles:

(36)

vs(p)vs(p) = p m
s=1

(37)

Problems
1. Dimensional analysis. a) A photon coupled to a complex scalar eld in D-dimensions has action S= 1 2 dDx F + gA + 3 + 4 (38) , D 1.

where F = ( A A ) and = as always but now = 0, 1, What is are the mass dimensions of A , , g and (as functions of D)?

b) An interaction is said to be renormalizable if its coupling constant is dimensionless. In what dimension D is the electromagnetic interaction renormalizable? How about the 3 interaction?