Quantum Field Theory I
Physics 523 & 524: Quantum Field Theory I
Kevin Cahill Department of Physics and Astronomy University of New Mexico, Albuquerque, NM 87131
Contents
1 Quantum fields and special relativity page 1 1.1 States 1 1.2 Creation operators 1 1.3 How fields transform 3 1.4 Translations 7 1.5 Boosts 8 1.6 Rotations 9 1.7 Spin-zero fields 9 1.8 Conserved charges 12 1.9 Parity, charge conjugation, and time reversal 13 1.10 Vector fields 15 1.11 Vector field for spin-zero particles 18 1.12 Vector field for spin-one particles 18 1.13 Lorentz group 23 1.14 Gamma matrices and Clifford algebras 25 1.15 Dirac’s gamma matrices 27 1.16 Dirac fields 28 1.17 Expansion of massive and massless Dirac fields 41 1.18 Spinors and angular momentum 44 1.19 Charge conjugation 49 1.20 Parity 51 2 Feynman diagrams 53 2.1 Time-dependent perturbation theory 53 2.2 Dyson’s expansion of the S matrix 54 2.3 The Feynman propagator for scalar fields 56 2.4 Application to a cubic scalar field theory 60 2.5 Feynman’s propagator for fields with spin 62 ii Contents 2.6 Feynman’s propagator for spin-one-half fields 63 2.7 Application to a theory of fermions and bosons 66 2.8 Feynman propagator for spin-one fields 71 2.9 The Feynman rules 73 2.10 Fermion-antifermion scattering 74 3 Action 76 3.1 Lagrangians and hamiltonians 76 3.2 Canonical variables 76 3.3 Principle of stationary action in field theory 78 3.4 Symmetries and conserved quantities in field theory 79 3.5 Poisson Brackets 84 3.6 Dirac Brackets 85 3.7 Dirac Brackets and QED 86 4 Quantum Electrodynamics 93 4.1 Global U( 1) Symmetry 93 4.2 Abelian gauge invariance 94 4.3 Coulomb-gauge quantization 96 4.4 QED in the interaction picture 97 4.5 Photon propagator 98 4.6 Feynman’s rules for QED 100 4.7 Electron-positron scattering 101 4.8 Trace identities 104 4.9 Electron-positron to muon-antimuon 105 4.10 Electron-muon scattering 111 4.11 One-loop QED 112 4.12 Magnetic moment of the electron 124 4.13 Charge form factor of electron 132 5 Nonabelian gauge theory 134 5.1 Yang and Mills invent local nonabelian symmetry 134 5.2 SU(3) 135 6 Standard model 138 6.1 Quantum chromodynamics 138 6.2 Abelian Higgs mechanism 138 6.3 Higgs’s mechanism 140 6.4 Quark and lepton interactions 146 6.5 Quark and Lepton Masses 148 6.6 CKM matrix 153 6.7 Lepton Masses 156 6.8 Before and after symmetry breaking 157 Contents iii 6.9 The Seesaw Mechanism 161 6.10 Neutrino Oscillations 163 7 Path integrals 165 7.1 Path integrals and Richard Feynman 165 7.2 Gaussian integrals and Trotter’s formula 165 7.3 Path integrals in quantum mechanics 166 7.4 Path integrals for quadratic actions 170 7.5 Path integrals in statistical mechanics 175 7.6 Boltzmann path integrals for quadratic actions 180 7.7 Mean values of time-ordered products 183 7.8 Quantum field theory 185 7.9 Finite-temperature field theory 189 7.10 Perturbation theory 192 7.11 Application to quantum electrodynamics 196 7.12 Fermionic path integrals 200 7.13 Application to nonabelian gauge theories 208 7.14 Faddeev-Popov trick 208 7.15 Ghosts 211 7.16 Integrating over the momenta 212 8 Landau theory of phase transitions 214 8.1 First- and second-order phase transitions 214 9 Effective field theories and gravity 217 9.1 Effective field theories 217 9.2 Is gravity an effective field theory? 218 9.3 Quantization of Fields in Curved Space 221 9.4 Accelerated coordinate systems 228 9.5 Scalar field in an accelerating frame 229 9.6 Maximally symmetric spaces 233 9.7 Conformal algebra 236 9.8 Conformal algebra in flat space 237 9.8.1 Angles and analytic functions 241 9.9 Maxwell’s action is conformally invariant for d = 4 247 9.10 Massless scalar field theory is conformally invariant if d = 2 248 9.11 Christoffel symbols as nonabelian gauge fields 249 9.12 Spin connection 254 10 Field Theory on a Lattice 256 10.1 Scalar Fields 256 10.2 Finite-temperature field theory 257 iv Contents 10.3 The Propagator 261 10.4 Pure Gauge Theory 261 10.5 Pure Gauge Theory on a Lattice 264 References 266 Index 267 1 Quantum fields and special relativity
1.1 States A Lorentz transformation Λ is implemented by a unitary operator U(Λ) which replaces the state |p, σi of a massive particle of momentum p and spin σ along the z-axis by the state s 0 (Λp) X (j) U(Λ)|p, σi = D (W (Λ, p)) |Λp, s0i (1.1) p0 s0σ s0 where W (Λ, p) is a Wigner rotation
W (Λ, p) = L−1(Λp)ΛL(p) (1.2) and L(p) is a standard Lorentz transformation that takes (m,~0) to p.
1.2 Creation operators The vacuum is invariant under Lorentz transformations and translations
U(Λ, a)|0i = |0i. (1.3)
A creation operator a†(p, σ) makes the state |p, σi from the vacuum state |0i |pσi = a†(p, σ)|0i. (1.4)
The creation and annihilation operators obey either the commutation rela- tion
† 0 0 † 0 0 † 0 0 (3) 0 [a(p, s), a (p , s )]− = a(p, s) a (p , s ) − a (p , s ) a(p, s) = δss0 δ (p − p ) (1.5) 2 Quantum fields and special relativity or the anticommutation relation
† 0 0 † 0 0 † 0 0 (3) 0 [a(p, s), a (p , s )]+ = a(p, s) a (p , s ) + a (p , s ) a(p, s) = δss0 δ (p − p ). (1.6) The two kinds of relations are written together as
† 0 0 † 0 0 † 0 0 (3) 0 [a(p, s), a (p , s )]∓ = a(p, s) a (p , s ) ∓ a (p , s ) a(p, s) = δss0 δ (p − p ). (1.7) A bracket [A, B] with no signed subscript is interpreted as a commutator. Equations (1.1 & 1.4) give
s 0 (Λp) X (j) U(Λ)a†(p, σ)|0i = D (W (Λ, p)) a†(Λp, s0)|0i. (1.8) p0 s0σ s0
And (1.3) gives
s 0 (Λp) X (j) U(Λ)a†(p, σ)U −1(Λ)|0i = D (W (Λ, p)) a†(Λp, s0)|0i. (1.9) p0 s0σ s0
SW in chapter 4 concludes that
s 0 (Λp) X (j) U(Λ)a†(p, σ)U −1(Λ) = D (W (Λ, p)) a†(Λp, s0). (1.10) p0 s0σ s0
If U(Λ, b) follows Λ by a translation by b, then
s 0 (Λp) X (j) U(Λ, b)a†(p, σ)U −1(Λ, b) = e−i(Λp)·a D (W (Λ, p)) a†(Λp, s0) p0 s0σ s0 s 0 (Λp) X †(j) = e−i(Λp)·a D (W −1(Λ, p)) a†(Λp, s0) p0 s0σ s0 s 0 (Λp) X ∗(j) = e−i(Λp)·a D (W −1(Λ, p)) a†(Λp, s0) p0 σs0 s0 (1.11) 1.3 How fields transform 3 The adjoint of this equation is s 0 (Λp) X ∗(j) U(Λ, b)a(p, σ)U −1(Λ, b) = ei(Λp)·a D (W (Λ, p)) a(Λp, s0) p0 s0σ s0 s 0 (Λp) X †(j) = ei(Λp)·a D (W (Λ, p)) a(Λp, s0) p0 σs0 s0 s 0 (Λp) X (j) = ei(Λp)·a D (W −1(Λ, p)) a(Λp, s0). p0 σs0 s0 (1.12) These equations (1.11 & 1.12) are (5.1.11 & 5.1.12) of SW.
1.3 How fields transform The “positive frequency” part of a field is a linear combination of annihila- tion operators Z + X 3 ψ` (x) = d p u`(x; p, σ) a(p, σ). (1.13) σ The “negative frequency” part of a field is a linear combination of creation operators of the antiparticles Z − X 3 † ψ` (x) = d p v`(x; p, σ) b (p, σ). (1.14) σ To have the fields (1.13 & 1.14) transform properly under Poincar´etrans- formations X U(Λ, a)ψ+(x)U −1(Λ, a) = D (Λ−1)ψ+(Λx + a) ` ``¯ `¯ `¯ Z X −1 X 3 = D``¯(Λ ) d p u`¯(Λx + a; p, σ) a(p, σ) `¯ σ X U(Λ, a)ψ−(x)U −1(Λ, a) = D (Λ−1)ψ−(Λx + a) ` ``¯ `¯ `¯ Z X −1 X 3 † = D``¯(Λ ) d p v`¯(Λx + a; p, σ) b (p, σ) `¯ σ (1.15) the spinors u`(x; p, σ) and v`(x; p, σ) must obey certain rules which we’ll now determine. 4 Quantum fields and special relativity First (1.12 & 1.13) give
Z + −1 X 3 −1 U(Λ, a)ψ` (x)U (Λ, a) = U(Λ, a) d p u`(x; p, σ) a(p, σ)U (Λ, a) σ Z X 3 −1 = d p u`(x; p, σ) U(Λ, a)a(p, σ)U (Λ, a) σ (1.16) s Z 0 X (Λp) X (j) = d3p u (x; p, σ) ei(Λp)·a D (W −1(Λ, p)) a(Λp, s0). ` p0 σs0 σ s0
Now we use the identity
d3p d3(Λp) = (1.17) p0 (Λp)0 to turn (1.16) into
Z + −1 X 3 i(Λp)·a U(Λ, a)ψ` (x)U (Λ, a) = d (Λp) u`(x; p, σ) e σ s 0 p X (j) × D (W −1(Λ, p)) a(Λp, s0). (1.18) (Λp)0 σs0 s0
Similarly (1.11, 1.14, & 1.17) give
Z − −1 X 3 † −1 U(Λ, a)ψ` (x)U (Λ, a) = U(Λ, a) d p v`(x; p, σ) b (p, σ)U (Λ, a) σ Z X 3 † −1 = d p v`(x; p, σ) U(Λ, a)b (p, σ)U (Λ, a) (1.19) σ s Z 0 X (Λp) X ∗(j) = d3p v (x; p, σ) e−i(Λp)·a D (W −1(Λ, p)) b†(Λp, s0) ` p0 σs0 σ s0 s Z 0 X p X ∗(j) = d3(Λp) v (x; p, σ) e−i(Λp)·a D (W −1(Λ, p)) b†(Λp, s0). ` (Λp)0 σs0 σ s0
So to get the fields to transform as in (1.15), equations (1.18 & 1.19) say 1.3 How fields transform 5 that we need
X X X Z D (Λ−1)ψ+(Λx + a) = D (Λ−1) d3p u (Λx + a; p, σ) a(p, σ) ``¯ `¯ ``¯ `¯ `¯ `¯ σ Z X −1 X 3 = D``¯(Λ ) d (Λp) u`¯(Λx + a;Λp, σ) a(Λp, σ) `¯ σ Z X 3 i(Λp)·a = d (Λp) u`(x; p, σ) e (1.20) σ s 0 p X (j) × D (W −1(Λ, p)) a(Λp, s0) (Λp)0 σs0 s0 Z X 3 0 i(Λp)·a = d (Λp) u`(x; p, s ) e s0 s 0 p X (j) × D (W −1(Λ, p)) a(Λp, σ) (Λp)0 s0σ σ and
X X X Z D (Λ−1)ψ−(Λx + a) = D (Λ−1) d3p v (Λx + a; p, σ) b†(p, σ) ``¯ `¯ ``¯ `¯ `¯ `¯ σ Z X −1 X 3 † = D``¯(Λ ) d (Λp) v`¯(Λx + a;Λp, σ) b (Λp, σ) `¯ σ Z X 3 −i(Λp)·a = d (Λp) v`(x; p, σ) e (1.21) σ s 0 p X ∗(j) × D (W −1(Λ, p)) b†(Λp, s0) (Λp)0 σs0 s0 Z X 3 0 −i(Λp)·a = d (Λp) v`(x; p, s ) e s0 s 0 p X ∗(j) × D (W −1(Λ, p)) b†(Λp, σ). (Λp)0 s0σ σ
Equating coefficients of the red annihilation and blue creation operators, we find that the fields will transform properly if the spinors u and v satisfy the 6 Quantum fields and special relativity rules
s 0 X −1 p X (j) −1 0 i(Λp)·a D ¯(Λ ) u¯(Λx + a;Λp, σ) = D (W (Λ, p))u (x; p, s ) e `` ` (Λp)0 s0σ ` `¯ s0 (1.22) s 0 X −1 p X ∗(j) −1 0 −i(Λp)·a D ¯(Λ )v¯(Λx + a;Λp, σ) = D (W (Λ, p))v (x; p, s )e `` ` (Λp)0 s0σ ` `¯ s0 (1.23)
which differ from SW’s by an interchange of the subscripts σ, s0 on the rotation matrices D(j). (I think SW has a typo there.) If we multiply both sides of these equations (1.22 & 1.23) by the two kinds of D matrices, then we get first
X −1 D`0`(Λ)D``¯(Λ ) u`¯(Λx + a;Λp, σ) = u`0 (Λx + a;Λp, σ) `,`¯ s 0 p X (j) −1 0 i(Λp)·a = D (W (Λ, p))D 0 (Λ)u (x; p, s ) e (1.24) (Λp)0 s0σ ` ` ` s0,` X −1 D`0`(Λ)D``¯(Λ )v`¯(Λx + a;Λp, σ) = v`0 (Λx + a;Λp, σ) `,`¯ s 0 p X ∗(j) −1 0 −i(Λp)·a = D (W (Λ, p))D 0 (Λ)v (x; p, s )e (1.25) (Λp)0 s0σ ` ` ` s0,` 1.4 Translations 7 and then with W ≡ W (Λ, p)
X (j) Dσs¯ (W )u`0 (Λx + a;Λp, σ) σ s 0 p X (j) −1 (j) 0 i(Λp)·a = D (W )D (W )D 0 (Λ)u (x; p, s ) e (Λp)0 s0σ σs¯ ` ` ` s0,σ,` s 0 p X i(Λp)·a = D 0 (Λ)u (x; p, s¯) e (1.26) (Λp)0 ` ` ` ` X ∗(j) Dσs¯ (W )v`0 (Λx + a;Λp, σ) σ s 0 p X ∗(j) −1 ∗(j) 0 −i(Λp)·a = D (W )D (W )D 0 (Λ)v (x; p, s )e (Λp)0 s0σ σs¯ ` ` ` σ,s0,` s 0 p X −i(Λp)·a = D 0 (Λ)v (x; p, s¯)e (1.27) (Λp)0 ` ` ` ` which are equations (5.1.13 & 5.1.14) of SW: s 0 X (j) p X i(Λp)·a u¯(Λx + a;Λp, s¯)D (W (Λ, p)) = D¯ (Λ)u (x; p, σ) e ` sσ¯ (Λp)0 `` ` s¯ ` s 0 X ∗(j) p X −i(Λp)·a v¯(Λx + a;Λp, s¯)D (W (Λ, p)) = D¯ (Λ)v (x; p, σ)e . ` sσ¯ (Λp)0 `` ` s¯ ` (1.28) These are the equations that determine the spinors u and v up to a few arbitrary phases.
1.4 Translations When Λ = I, the D matrices are equal to unity, and these last equations (1.28) say that for x = 0
ip·a u`(a; p, σ) = u`(0; p, σ) e −ip·a (1.29) v`(a; p, σ) = v`(0; p, σ)e . Thus the spinors u and v depend upon spacetime by the usual phase e±ip·x
−3/2 ip·x u`(x; p, σ) = (2π) u`(p, σ) e (1.30) −3/2 −ip·x v`(x; p, σ) = (2π) v`(p, σ)e 8 Quantum fields and special relativity in which the 2π’s are conventional. The fields therefore are Fourier trans- forms: Z + −3/2 X 3 ip·x ψ` (x) = (2π) d p e u`(p, σ) a(p, σ) σ Z (1.31) − −3/2 X 3 −ip·x † ψ` (x) = (2π) d p e v`(p, σ) b (p, σ) σ and every field of mass m obeys the Klein-Gordon equation
2 2 2 2 (∇ − ∂0 − m ) ψ`(x) = (2 − m ) ψ`(x) = 0. (1.32)
Since exp[i(Λp · (Λx + a))] = exp(ip · x + iΛp · a), the conditions (1.28) simplify to s 0 X (j) p X u¯(Λp, s¯)D (W (Λ, p)) = D¯ (Λ)u (p, σ) ` sσ¯ (Λp)0 `` ` s¯ ` s (1.33) 0 X ∗(j) p X v¯(Λp, s¯)D (W (Λ, p)) = D¯ (Λ)v (p, σ) ` sσ¯ (Λp)0 `` ` s¯ ` for all Lorentz transformations Λ.
1.5 Boosts Set p = k = (m,~0) and Λ = L(q) where L(q)k = q. So L(p) = 1 and
W (Λ, p) ≡ L−1(Λp)ΛL(p) = L−1(q)L(q) = 1. (1.34)
Then the equations (1.33) are
rm X u¯(q, σ) = D¯ (L(q))u (~0, σ) ` q0 `` ` ` (1.35) rm X v¯(q, σ) = D¯ (L(q))v (~0, σ). ` q0 `` ` `
Thus a spinor at finite momentum is given by a representation D(Λ) of the Lorentz group (see the online notes of chapter 10 of my book for its finite-dimensional nonunitary representations) acting on the spinor at zero 3-momentum p = k = (m,~0). We need to find what these spinors are. 1.6 Rotations 9 1.6 Rotations Now set p = k = (m,~0) and Λ = R a rotation so that W = R. For rotations, the spinor conditions (1.33) are X ~ (j) X ~ u`¯(0, s¯)Dsσ¯ (R) = D``¯ (R)u`(0, σ) s¯ ` (1.36) X ~ ∗(j) X ~ v`¯(0, s¯)Dsσ¯ (R) = D``¯ (R)v`(0, σ). s¯ `
(j) The representations Dsσ¯ (R) of the rotation group are (2j + 1) × (2j + 1)- dimensional unitary matrices. For a rotation of angle θ about the θ~ = θ axis, they are the ones taught in courses on quantum mechanics (and discussed in the notes of chapter 10)
(j) h −iθ·J (j) i Dsσ¯ (θ) = e (1.37) sσ¯ where [Ja,Jb] = iabcJc. The representations D``¯ (R) of the rotation group are finite-dimensional unitary matrices. For a rotation of angle θ about the θ~ = θ axis, they are h −iθ·J i D``¯ (θ) = e (1.38) ``¯ in which [Ja, Jb] = iabcJ . For tiny rotations, the conditions (1.36) require (because of the complex conjugation of the antiparticle condition) that the spinors obey the rules X ~ (j) X ~ u`¯(0, s¯)(Ja )sσ¯ = (Ja)``¯ u`(0, σ) s¯ ` (1.39) X ~ ∗(j) X ~ v`¯(0, s¯)(−Ja)sσ¯ ) = (Ja)``¯ v`(0, σ) s¯ ` for a = 1, 2, 3.
1.7 Spin-zero fields Spin-zero fields have no spin or Lorentz indexes. So the boost conditions p 0 p 0 (1.210) merely require that u(q) = m/q √u(0) and v(q) = √m/q v(0). The conventional normalization is u(0) = 1/ 2m and v(0) = 1/ 2m. The spin-zero spinors then are
u(p) = (2p0)−1/2 and v(p) = (2p0)−1/2. (1.40)
For simpicity, let’s first consider a neutral scalar field so that b(p, s) = 10 Quantum fields and special relativity a(p, s). The definitions (1.13) and (1.14) of the positive-frequency and negative- frequency fields and their behavior (1.30) under translations then give us Z d3p φ+(x) = a(p) eip·x p(2π)32p0 (1.41) Z d3p φ−(x) = a†(p) e−ip·x. p(2π)32p0 Note that φ±(x)† = φ∓(x). (1.42) 0 Since [a(p), a(p )]± = 0, it follows that + + − − [φ (x), φ (y)]∓ = 0 and [φ (x), φ (y)]∓ = 0 (1.43) whatever the values of x and y as long as we use commutators for bosons and anticommutators for fermions. But the commutation relation † (3) [a(p, s), a (q, t)]∓ = δst δ (p − q) (1.44) makes the commutator Z 3 3 0 + − d pd p ip·x −ip0·y 3 0 [φ (x), φ (y)]∓ = e e δ (p − p ) (2π)3p2p02p00 (1.45) Z d3p = eip·(x−y) = ∆ (x − y) (2π)32p0 + nonzero even for (x − y)2 > 0 as we’ll now verify. For space-like x, the Lorentz-invariant function ∆+(x) can only depend 2 0 upon x > 0 since the time x and its sign are√ not Lorentz invariant. So we choose a Lorentz frame with x0 = 0 and |x| = x2. In this frame, Z 3 d p ip·x ∆+(x) = e (2π)32pp2 + m2 (1.46) Z p2dp d cos θ = eipx cos θ (2π)22pp2 + m2 where p = |p| and x = |x|. Now Z d cos θ eipx cos θ = eipx − e−ipx /(ipx) = 2 sin(px)/(px), (1.47) so the integral (1.46) is 1 Z ∞ sin(px) pdp ∆+(x) = 2 p (1.48) 4π x 0 p2 + m2 1.7 Spin-zero fields 11 with u ≡ p/m Z ∞ m sin(mxu) udu m 2 ∆+(x) = √ = K1(mx ) (1.49) 2 2 2 4π x 0 u + 1 4π x a Hankel function. To get a Lorentz-invariant, causal theory, we use the arbitrary parameters κ and λ setting φ(x) = κφ+(x) + λφ−(x) (1.50) Now the adjoint rule (1.42) and the commutation relations (1.45 and 1.45) give † + − ∗ − + [φ(x), φ (y)]∓ = [κφ (x) + λφ (x), κ φ (y) + λφ (y)]∓ 2 + − 2 − + = |κ| [φ (x), φ (y)]∓ + |λ| [φ (x), φ (y)]∓ 2 2 = |κ| ∆+(x − y) ∓ |λ| ∆+(y − x) + − + − (1.51) [φ(x), φ(y)]∓ = [κφ (x) + λφ (x), κφ (y) + λφ (y)]∓ + − − + = κλ [φ (x), φ (y)]∓ + [φ (x), φ (y)]∓
= κλ (∆+(x − y) ∓ ∆+(y − x)) .
2 But when (x − y) > 0, ∆+(x − y) = ∆+(y − x). Thus these conditions are † 2 2 [φ(x), φ (y)]∓ = |κ| ∓ |λ| ∆+(x − y) (1.52) [φ(x), φ(y)]∓ = κλ∆+(x − y)(1 ∓ 1). The first of these equations implies that we choose the minus sign and so that we use commutation relations and not anticommutation relations for spin-zero fields. This is the spin-statistics theorem for spin-zero fields. SW proves the theorem for arbitrary massive fields in section 5.7. We also must set |κ| = |λ|. (1.53) The second equation then is automatically satisfied. The common magnitude and the phases of κ and λ are arbitrary, so we choose κ = λ = 1. We then have φ(x) = φ+(x) + φ−(x) = φ+(x) + φ+†(x) = φ†(x). (1.54) Now the interaction density H(x) will commute with H(y) for (x − y)2 > 0, and we have a chance of having a Lorentz-invariant, causal theory. The field (1.54) Z d3p h i φ(x) = a(p) eip·x + a†(p) e−ip·x (1.55) p(2π)32p0 12 Quantum fields and special relativity obeys the Klein-Gordon equation
2 2 2 2 (∇ − ∂0 − m ) φ(x) ≡ (2 − m ) φ(x) = 0. (1.56)
1.8 Conserved charges If the field φ adds and deletes charged particles, an interaction H(x) that is a polynomial in φ will not commute with the charge operator Q because φ+ will lower the charge and φ− will raise it. The standard way to solve this problem is to start with two hermitian fields φ1 and φ2 of the same mass. One defines a complex scalar field as a complex linear combination of the two fields 1 φ(x) = √ (φ1(x) + iφ2(x)) 2 Z 3 d p 1 ip·x 1 † † −ip·x = √ (a1(p) + ia2(p)) e + √ a (p) + ia (p) e . p(2π)32p0 2 2 1 2 (1.57) Setting
1 † 1 † † a(p) = √ (a1(p) + ia2(p)) and b (p) = √ a (p) + ia (p) (1.58) 2 2 1 2 so that
1 † 1 † † b(p) = √ (a1(p) − ia2(p)) and a (p) = √ a (p) − ia (p) (1.59) 2 2 1 2 we have Z d3p h i φ(x) = a(p) eip·x + b†(p) e−ip·x (1.60) p(2π)32p0 and Z d3p h i φ†(x) = b(p) eip·x + a†(p) e−ip·x . (1.61) p(2π)32p0 Since the commutation relations of the real creation and annihilation oper- ators are for i, j = 1, 2
† 0 3 0 0 † † 0 [ai(p), aj(p )] = δij δ (p − p ) and [ai(p), aj(p )] = 0 = [ai (p), aj(p )] (1.62) the commutation relations of the complex creation and annihilation opera- tors are [a(p), a†(p0)] = δ3(p − p0) and [b(p), b†(p0)] = δ3(p − p0) (1.63) 1.9 Parity, charge conjugation, and time reversal 13 with all other commutators vanishing. Now φ(x) lowers the charge of a state by q if a† adds a particle of charge q and if b† adds a particle of charge −q. Similarly, φ†(x) raises the charge of a state by q [Q, φ(x)] = − qφ(x) and [Q, φ†(x)] = qφ†(x). (1.64) So an interaction with as many φ(x)’s as φ†(x)’s conserves charge.
1.9 Parity, charge conjugation, and time reversal If the unitary operator P represents parity on the creation operators
† −1 † † −1 † Pa1(p)P = η a1(−p) and Pa2(p)P = η a2(−p) (1.65) with the same phase η. Then
−1 ∗ −1 ∗ Pa1(p)P = η a1(−p) and Pa2(p)P = η a2(−p) (1.66) and so both Pa†(p)P−1 = ηa†(−p) and Pa(p)P−1 = η∗a(−p) (1.67) and Pb†(p)P−1 = ηb†(−p) and Pb(p)P−1 = η∗b(−p). (1.68) Thus if the field Z 3 d p h ip·x † −ip·xi φ1(x) = a1(p) e + a (p) e (1.69) p(2π)32p0 1 or φ2(x), or the complex field (1.60) is to go into a multiple of itself under parity, then we need η = η∗ so that η is real. Then the fields transform under parity as −1 ∗ 0 0 Pφ1(x)P = η φ1(x , −x) = ηφ1(x , −x) −1 ∗ 0 0 Pφ2(x)P = η φ(x , −x) = ηφ2(x , −x) (1.70) Pφ(x)P−1 = η∗φ(x0, −x) = ηφ(x0, −x). Since P2 = I, we must have η = ±1. SW allows for a more general phase by having parity act with the same phase on a and b†. Both schemes imply that the parity of a hermitian field is ±1 and that the state Z |abi = d3p f(p2) a†(p) b†(−p) |0i (1.71) has even or positive parity, P|abi = |abi. 14 Quantum fields and special relativity Charge conjugation works similarly. If the unitary operator C represents charge conjugation on the creation operators
† −1 † † −1 † Ca1(p)C = ξa1(p) and Ca2(p)C = − ξa2(p) (1.72) with the same phase ξ. Then
−1 ∗ −1 ∗ Ca1(p)C = ξ a1(p) and Ca2(p)C = − ξ a2(p) (1.73) √ √ and so since a = (a1 + ia2)/ 2 and b = (a1 − ia2)/ 2 Ca(p)C−1 = ξ∗ b(p) and Cb(p)C−1 = ξ∗a(p) (1.74) √ √ † † † † † † and since a = (a1 − ia2)/ 2 and b = (a1 + ia2)/ 2 Ca†(p)C−1 = ξ b†(p) and Cb†(p)C−1 = ξa†(p). (1.75)
Thus under charge conjugation, the field (1.60) becomes Z d3p h i Cφ(x)C−1 = ξ∗ b(p) eip·x + ξ a†(p) e−ip·x (1.76) p(2π)32p0 and so if it is to go into a multiple of itself or of its adjoint under charge conjugation then we need ξ = ξ∗ so that ξ is real. We then get
Cφ(x)C−1 = ξ∗ φ†(x) = ξ φ†(x). (1.77)
Since C2 = I, we must have ξ = ±1. SW allows for a more general phase by having charge conjugation act with the same phase on a and b†. Both schemes imply that the charge-conjugation parity of a hermitian field is ±1 and that the state Z |abi = d3p f(p2) a†(p) b†(p) |0i (1.78) has even or positive charge-conjugation parity, C|abi = |abi. The time-reversal operator T is antilinear and antiunitary. So if −1 ∗ −1 ∗ Ta1(p)T = ζ a1(−p) and Ta2(p)T = − ζ a2(−p) (1.79) † −1 † † −1 † Ta1(p)T = ζa1(−p) and Ta2(p)T = − ζa2(−p) then
−1 1 −1 1 −1 −1 Ta(p)T = T√ a1(p) + ia2(p) T = √ Ta1(p)T − iTa2(p)T 2 2 ∗ 1 ∗ = ζ √ a1(−p) + ia2(−p) = ζ a(−p) 2 (1.80) 1.10 Vector fields 15 and 1 1 Tb†(p)T−1 = T√ a†(p) + ia†(p)T−1 = √ Ta†(p)T−1 − iTa†(p)T−1 2 1 2 2 1 2 1 = ζ √ a†(−p) + ia†(−p) = ζ b†(−p) 2 1 2 (1.81) then one has
Z d3p h i Tφ(x)T−1 = T a(p) eip·x + b†(p) e−ip·x T−1 p(2π)32p0 Z d3p h i = Ta(p)T−1 e−ip·x + Tb†(p)T−1 eip·x (1.82) p(2π)32p0 Z d3p h i = ζ∗ a(−p) e−ip·x + ζ b†(−p) eip·x . p(2π)32p0
So if ζ is real, then after replacing −p by p, we get
Z 3 −1 ∗ d p h ip0x0+ip·x † −ip0x0+ip·xi Tφ(x)T = ζ p a(p) e + b (p) e (2π)32p0 (1.83) = ζ∗φ(−x0, x) = ζφ(−x0, x).
Since T2 = I, the phase ζ = ±1. SW lets ζ be complex but defined only for complex scalar fields and not for their real and imaginary parts.
1.10 Vector fields Vector fields transform like the 4-vector xi of spacetime. So
`¯ D``¯ (Λ) = Λ ` (1.84) for `,¯ ` = 0, 1, 2, 3. Again we start with a hermitian field labelled by i = 0, 1, 2, 3
X Z φ+i(x) = (2π)−3/2 d3p eip·xui(p, s) a(p, s) s (1.85) X Z φ−i(x) = (2π)−3/2 d3p e−ip·xvi(p, s) a†(p, s). s 16 Quantum fields and special relativity The boost conditions (1.210) say that
r m X ui(p, s) = L(p)i uk(~0, s) p0 k k (1.86) r m X vi(p, s) = L(p)i vk(~0, s). p0 k k
The rotation conditions (1.39) give
X i ~ (j) X i k ~ u (0, s¯)(Ja )ss¯ = (Ja) ku (0, s) s¯ k (1.87) X i ~ ∗(j) X i k ~ − v (0, s¯)(Ja )ss¯ = (Ja) kv (0, s). s¯ k
(j) The (2j + 1) × (2j + 1) matrices (Ja )ss¯ are the generators of the (2j + 1) × (2j + 1) representation of the rotation group. (See my online notes on group theory.) You learned that
3 3 j X h (j) 2i X X (j) (j) (Ja ) = (Ja )ss¯ (Ja )ss0 = j(j + 1)δss¯ 0 (1.88) ss¯ 0 a=1 a=1 s=−j and that
0 1 0 0 −i 0 1 0 0 1 1 J1 = √ 1 0 1 ,J2 = √ i 0 −i ,J3 = 0 0 0 2 0 1 0 2 0 i 0 0 0 −1 (1.89) in courses on quantum mechanics. i For k = 1, 2, 3, the three 4 × 4 matrices (Jk) j are the generators of rotations in the vector representation of the Lorentz group. Their nonzero components are
i (Jk) j = − iijk (1.90)
0 0 i for i, j, k = 1, 2, 3, while (Jk) 0 = 0, (Jk) j = 0, and (Jk) 0 = 0 for i, j, k = 1, 2, 3. So
2 i i (J ) j = 2δ j (1.91)
2 0 2 0 2 i with (J ) 0 = 0, (J ) j = 0, and (J ) 0 = 0 for i, j = 1, 2, 3. Apart from a factor of i, the Jk’s are the 4 × 4 matrices Ja = iRa of my online notes on 1.10 Vector fields 17 the Lorentz group
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 i 0 0 −i 0 J1 = J2 = J3 = . 0 0 0 −i 0 0 0 0 0 i 0 0 0 0 i 0 0 −i 0 0 0 0 0 0 (1.92) 0 Since (Ja) k = 0 for a, k = 1, 2, 3, the spin conditions (1.87) give for i = 0 X 0 ~ (j) X 0 ~ ∗(j) u (0, s¯)(Ja )ss¯ = 0 and − v (0, s¯)(Ja )ss¯ = 0. (1.93) s¯ s¯
(j) Multiplying these equations from the right by (Ja )ss0 while summing over (j) 2 a = 1, 2, 3 and using the formula (1.88) [(J ) ]ss0 = j(j + 1) δss0 , we find
j(j + 1) u0(~0, s) = 0 and j(j + 1) v0(~0, s) = 0. (1.94)
Thus u0(~0, σ) and v0(~0, σ) can be anything if the field represents particles of spin j = 0, but u0(~0, σ) and v0(~0, σ) must both vanish if the field represents particles of spin j > 0. Now we set i = 1, 2, 3 in the spin conditions (1.87) and again multiply (j) from the right by (Ja )ss0 while summing over a = 1, 2, 3 and using the formula (1.88) (J (j))2 = j(j + 1). The Lorentz rotation matrices generate a j = 1 representation of the group of rotations.
3 X i k i i (Ja) k (Ja) ` = j(j + 1) δ` = 2δ`. (1.95) ka=1 So the remaining conditions on the fields are
i ~ 0 X i ~ (j) (j) X i k ~ (j) j(j + 1) u (0, s ) = u (0, s¯)(Ja )ss¯ (Ja )ss0 = (Ja) ku (0, s)(Ja )ss0 ssa¯ ksa X i k ` ~ 0 X i ` ~ 0 i ~ 0 = (Ja) k (Ja) ` u (0, s ) = 2 δ` u (0, s ) = 2 u (0, s ) k`a k i ~ 0 X i ~ ∗(j) ∗(j) X i k ~ (j) j(j + 1) v (0, s ) = v (0, s¯)(Ja )ss¯ (Ja )ss0 = (Ja) kv (0, s)(Ja )ss0 ssa¯ ksa X i k ` ~ 0 X i ` ~ 0 i ~ 0 = (Ja) k (Ja) ` v (0, s ) = 2 δ` v (0, s ) = 2 v (0, s ). k`a k (1.96)
Thus if j = 0, then for i = 1, 2, 3 both ui(~0, s) and vi(~0, s) must vanish, while if j > 0, then since j(j + 1) = 2, the spin j must be unity, j = 1. 18 Quantum fields and special relativity 1.11 Vector field for spin-zero particles The only nonvanishing components are constants taken conventionally as
u0(~0) = ipm/2 and v0(~0) = − ipm/2. (1.97)
At finite momentum the boost conditions (1.210) give them as p p uµ(~p) = ipµ/ 2p0 and vµ(~p) = − ipµ/ 2p0. (1.98)
The vector field φµ(x) of a spin-zero particle is then the derivative of a scalar field φ(x) Z d3p h i φµ(x) = ∂µφ(x) = ipµ a(p) eip·x − ipµ b†(p) e−ip·x (1.99) p(2π)32p0
1.12 Vector field for spin-one particles We start with the s = 0 spinors ui(~0, 0) and vi(~0, 0) and note that since (j) (J3 )s¯0 = 0, the a = 3 rotation conditions (1.87) imply that i k ~ i ~ i k ~ i ~ (J3) k u (0, 0) = iR3 u (0, 0) = 0 and (J3) k v (0, 0) = iR3 v (0, 0) = 0. (1.100) Referring back to the explicit formulas for the generators of rotations and setting u, v = (0, x, y, z) we see that
0 0 0 0 0 0 0 0 0 −i 0 x −iy 0 J3 u = = = (1.101) 0 i 0 0 y x 0 0 0 0 0 z 0 0 and 0 0 0 0 0 0 0 0 0 −i 0 x −iy 0 J3 v = = = . (1.102) 0 i 0 0 y x 0 0 0 0 0 z 0 0 Thus only the 3-component z can be nonzero. The conventional choice is 0 µ µ 1 0 u (~0, 0) = v (~0, 0) = √ . (1.103) 2m 0 1
We now form the linear combinations of the rotation conditions (1.87) 1.12 Vector field for spin-one particles 19
(1) (1) (1) that correspond to the raising and lowering matrices J± = J1 ± iJ2
0 1 0 0 0 0 (1) √ (1) √ J+ = 2 0 0 1 and J− = 2 1 0 0 . (1.104) 0 0 0 0 1 0
Their Lorentz counterparts are
0 0 0 0 (1) (1) (1) 0 0 0 ∓1 J = J ± iJ = . (1.105) ± 1 2 0 0 0 −i 0 ±1 i 0
In these terms, the rotation conditions (1.87) for the j = 1 spinors ui(~0, s) are
X i ~ (1) X i k ~ u (0, s¯)(J± )ss¯ = (J±) ku (0, s). (1.106) s¯ k
But
∗(1) ∗(1) (1) (1) J1 ± iJ2 = J1 ∓ iJ2 = J∓. (1.107)
So the rotation conditions (1.87) for the j = 1 spinors vi(~0, s) are
X i ~ (1) X i k ~ − v (0, s¯)(J∓ )ss¯ = (J±) kv (0, s). (1.108) s¯ k
So for the plus sign and the choice s = 0, the condition (1.106) gives ui(~0, 1) as
0 0 0 0 0 X (1) √ 0 0 0 −1 1 0 ui(~0, s¯) J = 2 ui(~0, 1) = (J )i uk(~0, 0) = √ +¯s0 + k 0 0 0 −i 0 s¯ 2m 0 1 i 0 1 (1.109) or 0 i 1 −1 u (~0, 1) = √ . (1.110) 2 m −i 0 20 Quantum fields and special relativity Similarly, the minus sign and the choice s = 0 give for ui(~0, −1) 0 0 0 0 0 X (1) √ 0 0 0 1 1 0 ui(~0, s¯) J = 2 ui(~0, −1) = (J )i uk(~0, 0) = √ −s¯0 − k 0 0 0 −i 0 s¯ 2m 0 −1 i 0 1 (1.111) or 0 i 1 1 u (~0, −1) = √ . (1.112) 2 m −i 0
The rotation condition (1.108) for the j = 1 spinors vi(~0, s) with the minus sign and the choice s = 0 gives 0 0 0 0 0 X (1) √ 0 0 0 −1 1 0 − vi(~0, s¯)J = − 2 vi(~0, −1) = (J )i vk(~0, 0) = √ −s¯0 + k 0 0 0 −i 0 s¯ 2m 0 1 i 0 1 (1.113) or 0 i 1 1 v (~0, −1) = √ . (1.114) 2 m i 0 Similarly, the plus sign and the choice s = 0 give 0 0 0 0 0 X (1) √ 0 0 0 1 1 0 − vi(~0, s¯)J = − 2 vi(~0, 1) = (J )i vk(~0, 0) = √ +¯s0 − k 0 0 0 −i 0 s¯ 2m 0 −1 i 0 1 (1.115) or 0 i 1 −1 v (~0, 1) = √ . (1.116) 2 m i 0 The boost conditions (1.210) now give for i, k = 0, 1, 2, 3
i ∗i p 0 i k ~ i p 0 u (~p,s) = v (~p,s) = m/p L k(~p) u (0, s) = e (~p,s)/ 2p (1.117) 1.12 Vector field for spin-one particles 21 where i i k ~ e (~p,s) = L k(~p) e (0, s) (1.118) and 0 0 0 0 1 1 1 1 e(~0, 0) = , e(~0, 1) = − √ , and e(~0, −1) = √ . 0 2 i 2 −i 1 0 0 (1.119) A single massive vector field is then
1 X Z d3p φi(x) = φ+i(x)+φ−i(x) = ei(~p,s) a(~p,s)eip·x+e∗i(~p,s) a†(~p,s)e−ip·x. p 3 0 s=−1 (2π) 2p (1.120) The commutator/anticommutator of the positive and negative frequency parts of the field is
Z 3 +i −k d p ip·(x−y) ik [φ (x), φ (y)]∓ = e Π (~p) (1.121) p(2π)32p0 where Π is a sum of outer products of 4-vectors
1 X Πik(~p) = ei(~p,s)e∗k (~p,s). (1.122) s=−1 At ~p = 0, the matrix Π is the unit matrix on the spatial coordinates
0 0 0 0 1 X 0 1 0 0 Π(~0) = ei(~0, s)e∗k (~0, s) = . (1.123) 0 0 1 0 s=−1 0 0 0 1
So Π(~p) is
1 0 0 0 0 0 0 0 Π(~p) = LΠ(0)LT = LηLT + L LT (1.124) 0 0 0 0 0 0 0 0 or Π(~p)ik = ηik + pipk/m2. (1.125) 22 Quantum fields and special relativity This equation lets us write the commutator (1.126) in terms of the Lorentz- invariant function ∆+(x − y) (1.45) as Z 3 +i −k ik i k 2 d p ip·(x−y) [φ (x), φ (y)]∓ = (η − ∂ ∂ /m ) p e (2π)32p0 (1.126) ik i k 2 = (η − ∂ ∂ /m ) ∆+(x − y). As for a scalar field, we set vi(x) = κ φ+i(x) + λ φ−i(x) (1.127)
2 and find for (x − y) > 0 since ∆+(x − y) = ∆+(y − x) for x, y spacelike † 2 2 ik i k 2 [v(x), v (y)]∓ = |κ| ∓ |λ| (η − ∂ ∂ /m ) ∆+(x − y) (1.128) ik i k 2 [v(x), v(y)]∓ = (1 ∓ 1) κλ(η − ∂ ∂ /m ) ∆+(x − y). So we must choose the minus sign and set |κ| = |λ|. So then vi(x) = v+i(x) + v−i(x) = v+i(x) + v+i†(x) (1.129) is real. This is a second example of the spin-statistics theorem. If two such fields have the same mass, then we can combine them as we combined scalar fields i +i −i v (x) = v1 (x) + iv2 (x). (1.130) These fields obey the Klein-Gordon equation (2 − m2)vi(x) = 0. (1.131) And since both i i j k k ` p = L jk and e (~p) = L `e (0) (1.132) it follows that p · e(~p) = k · e(0) = 0. (1.133) So the field vi also obeys the rule i ∂i v (x) = 0. (1.134) These equations (1.133) and 1.134 are like those of the electromagnetic field in Lorentz gauge. But one can’t get quantum electrodynamics as the i m → 0 limit of just any such theory. For the interaction H = Ji v would lead to a rate for v-boson production like ik JiJkΠ (~p) (1.135) which diverges as m → 0 because of the pipk/m2 term in Πik(~p). One can 1.13 Lorentz group 23
i avoid this divergence by requiring that ∂iJ = 0 which is current conserva- tion. Under parity, charge conjugation, and time reversal, a vector field trans- forms as a −1 ∗ a b Pv (x)P = − η P bv (Px) Cva(x)C−1 = ξ∗va†(x) (1.136) a −1 ∗ a b Tv (x)T = ζ P bv (−Px).
1.13 Lorentz group The Lorentz group O(3, 1) is the set of all linear transformations L that leave invariant the Minkowski inner product
xy ≡ x · y − x0y0 = xTηy (1.137) in which η is the diagonal matrix
−1 0 0 0 0 1 0 0 η = . (1.138) 0 0 1 0 0 0 0 1
So L is in O(3, 1) if for all 4-vectors x and y
(Lx) T ηL y = xTLT η Ly = xTη y. (1.139)
Since x and y are arbitrary, this condition amounts to
LTη L = η. (1.140)
Taking the determinant of both sides and recalling that det AT = det A and that det(AB) = det A det B, we have
(det L)2 = 1. (1.141)
So det L = ±1, and every Lorentz transformation L has an inverse. Multi- plying (1.140) by η, we get
ηLTηL = η2 = I (1.142) which identifies L−1 as L−1 = ηLTη. (1.143) 24 Quantum fields and special relativity The subgroup of O(3, 1) with det L = 1 is the proper Lorentz group SO(3, 1). The subgroup of SO(3, 1) that leaves invariant the sign of the time compo- nent of timelike vectors is the proper orthochronous Lorentz group SO+(3, 1). To find the Lie algebra of SO+(3, 1), we take a Lorentz matrix L = I + ω that differs from the identity matrix I by a tiny matrix ω and require L to obey the condition (1.140) for membership in the Lorentz group