P. 1
Lecture Notes for Physics 253a (Sidney Coleman)

Lecture Notes for Physics 253a (Sidney Coleman)

|Views: 28|Likes:
Published by Chris Walsh

More info:

Published by: Chris Walsh on Apr 20, 2012
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less





Notes from Sidney Coleman’s Physics 253a

Sidney Coleman∗ Harvard, Fall 1985-1986 semester†

Edit and typeset by Bryan Gin-ge Chen and Yuan-Sen Ting from scans of the handwritten notes of Brian Hill. Please contact Bryan Gin-ge Chen (bryangingechen@gmail.com) and Yuan-Sen Ting (ting.yuansen@gmail.com) if you have any questions or comments. † A L TEX version of October 21, 2011


0. Preface

Notes from Sidney Coleman’s Physics 253a




It’s unexpected and heart-warming to be asked by Bryan Chen to write something about these notes, 25 years after taking them. I was the teaching assistant for Sidney’s quantum field theory course for three years. In the first year, I sat in, because frankly, I hadn’t learned quantum field theory well enough the first time that I took it. When I have the good fortune to hear a really good lecturer, I often re-copy my notes, preferably the evening on the day that I took them. Once in a while, students would miss a class, and then ask me if they could look at my notes. At some point, the requests started happening enough that it was suggested that a copy be put on reserve in the Harvard physics library. From there, copies of the notes just kept spreading. Sidney once expressed disappointment about the spread of the notes. For one thing, I even wrote down some of his anecdotes and jokes, and that made it less fun for him to re-tell them. For another, he wrote Aspects of Symmetry which shared a lot of material with what he taught in Physics 253b. He may have had in mind that he would write a field theory book as a companion volume. Of course, he never did write a field theory book, or you’d be reading that, and he never tried to rein the copies in. Now that he is gone, we are lucky that his clarity lives on. A Thanks to Bryan Chen and Yuan-Sen Ting for creating this lovely L TEX version. Sometimes transcription can seem tedious, but I hope it was as valuable for them as the first re-copying was for me, and that for you – fellow student of quantum field theory – the existence of these notes is similarly valuable. –Brian Hill, www.lingerhere.org, March 10, 2011 Typesetters’ notes The great field theorist Sidney Coleman for many years taught the course Physics 253 at Harvard on Quantum Field Theory. The notes you are reading were typeset from a scanned version of handwritten lecture notes by Brian Hill from the Fall 1985-1986 semester of the first half of the course: Physics 253a. The Harvard Physics Department has made films of the lectures from the 1975-1976 version of the course available on their website as well. The typesetting for lectures 1-11 was done by Bryan Gin-ge Chen and for lectures 12-28 by Yuan-Sen Ting, who also recreated most of the figures. We have attempted to stay as faithful as possible to the scanned notes, correcting only obvious typos and changing some A of the in-text references. We have noted some of these changes in comments in the L TEX source, though this was not done systematically.

0. Preface

Notes from Sidney Coleman’s Physics 253a


0 Preface 1 September 23 Units; Translational invariance, Rotational invariance and Lorentz invariance for a single free particle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 September 25 Position operator; Violation of Causality; Pair Production; Fock Space; Occupation number representation; SHO; Oscillator-like formalism for Fock space. . . . . 3 September 30 Causality, observables and quantum fields; Constructing the quantum field from Fock space operators axiomatically; Translational invariance; Lorentz invariance and Relativistic normalization; Field constructed satisfies Klein-Gordon equation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 October 2 Constructing Fock space from the quantum field axiomatically; The Method of the Missing Box, Classical Particle Mechanics, Quantum Particle Mechanics, Classical Field Theory, Quantum Field Theory; Quantum field from free scalar theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 October 7 Hamiltonian recovered in free scalar theory up to infinite constant; Normal ordering; Symmetries and conservation laws, Noether’s theorem; Noether’s theorem in field theory, conserved currents; ambiguity in currents; Energy-momentum tensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 October 9 Lorentz transformations; Angular momentum conservation; Internal symmetries; SO(2) internal symmetry; Charged field; SO(n) internal symmetry. . . . . . 2 7 7 14 14 21

21 28

28 38

38 49 49

7 October 14 58 Lorentz transformation, properties of conserved quantities; Discrete symmetries; φ → −φ; Charge conjugation; Parity; Ambiguity of choice of parity; Time reversal; Unitary and anti-unitary operators, angular momentum; Dilatations. 58 8 October 16 70

0. Preface

Notes from Sidney Coleman’s Physics 253a


Scattering theory overview; Low budget scattering theory; Turning on and off function; Schr¨dinger picture, Heisenberg picture and interaction picture; Evoluo tion operator; Time ordered product; Three models; Wick’s theorem. . . . . 9 October 21 Diagrammatic perturbation theory in Model 3; Vertex in model 1; Connected diagrams; Thm: all Wick diagrams = : e connected Wick diagrams :; Model 1 solved; Model 2 begun. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 October 23 Model 2 finished; Vacuum energy c.t.; S matrix is 1; Ground state energy; Yukawa potential; Ground state wave function; Model 3 and Mass renormalization; Renormalized perturbation theory. . . . . . . . . . . . . . . . . . . . . . . .

70 83

83 93


11 October 28 103 Feynman diagrams in Model 3; Feynman rules in model 3; A catalog of all Feynman diagrams in model 3 to O(g 2 ); Scattering amplitude at O(g 2 ); Direct and exchange Yukawa potentials. . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 12 October 30 112 2 “Nucleon”-anti“nucleon” scattering at O(g ); Energy eigenstate probe; Meson“nucleon” scattering; “Nucleon”-anti“nucleon” annihilation; Assembling the amplitudes for various processes into one amplitude; Mandelstam variables; Mandelstam-Kibble plot; crossing symmetry; CPT; Phase space and the S tran. matrix; Differential time prob. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 unit 13 November 4 131 Differential tran. prob. ; Decay; Cross sections, flux; Final state phase Applications of unit time dσ space simplified for two bodies; dΩ ; Optical theorem; Final state phase space for three bodies; Feynman diagrams with external lines off the mass shell; they could be an internal part of a larger diagram. . . . . . . . . . . . . . . . 131 14 November 6 144 Fourier transform of the new blob; A second interpretation of Feynman diagrams with lines off the mass shell; they are the coefficients of ρn in 0|S|0 ρ ; A third interpretation of the blob; they are the Fourier transform of the VEV of a string of Heisenberg fields; Reformulation of scattering theory; S matrix elements without the turning on and off function; LSZ formula stated. . . . . 144 15 November 13 163

0. Preface

Notes from Sidney Coleman’s Physics 253a


LSZ formula proved; A second look at Model 3 and its renormalization; Renormalization conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 16 November 18 171 Perturbative determination of a c.t.; Problems with derivative couplings; Rephrasing renormalization conditions in terms of Green’s functions; Lehmann-Kallen spectral representation for the propagator; Rephrasing renormalization conditions in terms of 1PI functions. . . . . . . . . . . . . . . . . . . . . . . . . 171 17 November 20 182 Perturbative determination of c.t.; Corrections to external lines in the computation of S matrix elements; One loop correction to meson self energy; Feynman’s trick for combining 2 denominators; Shift to make denominator SO(3, 1) invariant; Wick notation to make denominator O(4) invariant; Integral tables for convergent combinations; Self-energy at one loop studied; Combining lots of denominators; The shift in the general case to reduce any multi-loop integral to an integral over Feynman parameters. . . . . . . . . . . . . . . . . . . 182 18 November 25 196 Rephrasing coupling constant renormalization in terms of a 1PI function; Experimental significance of the definition; Renormalization versus the infinities; Renormalizable Lagrangians; Unstable particles, Decay products. . . . . . . 196 19 December 2 207 Unstable particles, lifetime, method of stationary phase; Where it begins again; Lorentz transformation laws of fields; Equivalent representations; Reducible reps; The finite dimensional inequivalent irreducible representations of SO(3); Unitarity; Complex conjugation; Direct product; Projection operators and reducibility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 20 December 4 226 Parametrizing the Lorentz group; Commutation relations for the generators; decomposition into two sets obeying SO(3) commutation relations; The catalog; Complex conjugation properties; Tensor product properties; Restriction to SO(3); The vector; Rank 2 tensors; Spinors. . . . . . . . . . . . . . . . . . 226 21 December 9 240 Lagrangian made of two component spinors; Solution of Weyl equations of motion; Weyl particles; Dirac Lagrangian; Four-component spinors; Weyl basis, Dirac basis; Plane wave solutions of the Dirac equation. . . . . . . . . . . . . . . . 240

0. Preface

Notes from Sidney Coleman’s Physics 253a


22 December 11 249 Plane wave solutions of the Dirac equation; Pauli’s theorem; Dirac adjoint; PauliFeynman notation; Parity; Bilinears; Orthogonality; Completeness; Summary. 249 23 December 16 267 Canonical quantization of the Dirac Lagrangian. . . . . . . . . . . . . . . . . . . . 267 24 December 18 273 Perturbation theory for spinors; Time ordered product; Wick’s theorem; Calculation of the contraction (propagator); Wick diagrams; Feynman diagrams; Matrix multiplication; Spin averages and spin sums. . . . . . . . . . . . . . . 273 25 January 6 289 Parity for spinors; Fermion and antifermion have opposite intrinsic parity; Charge conjugation; Majorana basis; Charge conjugation properties of fermion bilinears; Decay of ortho and para positronium; UP UC = UC UP (−1)NF ; PT. . . . 289 26 January 8 301 Effect of PT on states; Proof of CPT theorem in perturbation theory; Renormalization of spinor theories; Propagator. . . . . . . . . . . . . . . . . . . . . . . 301 27 January 13 313 1PI part of propagator; Spectral representation for propagator; (p ) to O(g 2 ) in meson-nucleon theory; Coupling constant renormalization; Is renormalization sufficient to eliminate ∞’s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 28 January 15 323 Regularization; Regulator fields; Dim’l regularization; Minimal subtraction; BPHZ renormalizability; Renormalization and symmetry; Renormalization of composite operators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323

Class: Why not use Feynman’s lecture notes? Gell-Mann: Because Feynman uses a different method than we do. Class: What is Feynman’s method? Gell-Mann: You write down the problem. Then you look at it and you think. Then you write down the answer.

2 For calculations of high accuracy effects at low energy. September 23 Notes from Sidney Coleman’s Physics 253a 7 1 September 23 In NRQM. but you already know from NRQM that that isn’t true. At these energies p + p → p + p + π0 is possible. without considering multi-particle intermediate states. complicate quantum mechanics? The addition of relativity is necessary at energies E mc2 . This is a fluke caused by some c unusually low electrodynamic matrix elements. If the Dirac equation is used.. only a finite number.1. You might think that for a given E. even a small number of processes actually contribute. suppressed by energy denominators. Units =c=1 Because we’re doing relativistic (c) quantum mechanics ( ) [m] = [E] = [T −1 ] = [L−1 ] Sometimes 1 = −1 = 2π 1 and 2π = 1=“one-bar” (1 fermi)−1 ≈ 197 GeV . Intermediate states with extra particles will contribute corrections of order E ← mc2 ← Typical energies in problem Typical energy denominator ∼ mv 2 mc2 = v 2 . The exact solution of a high energy scattering problem necessarily involves many particle processes. (It is a phenomenal fluke that relativistic kinematic corrections for the Hydrogen atom work. the addition of L.) We will see that you cannot have a consistent relativistic picture without pair production. relativistic effects of order v c can be included. We can’t even solve the zero-body problem. At slightly higher energies p+p→p+p+p+p ¯ can occur. rotational invariance simplifies scattering problems. Why does the addition of relativity.I. the addition of relativity forces you to consider many-body problems. H → H + δV δE0 = 0|δV |0 + n | 0|δV |n |2 + ··· E0 − En Intermediate states of all energies contribute. c As a general conclusion: the corrections of relativistic kinematics and the corrections from multiparticle intermediate states are comparable. corrections of O v can be obtained.

The theory is not manifestly L.. Let’s be more precise about this. Metric convention: + − −−.I. Normalization k|k = δ (3) (k − k ) The statement that this is a complete set of states. The theory is manifestly rotationally and translationally invariant.1) (1.? Just because it contains one relativistic formula.2) (1. U (a).3) . 1). and 1. is 1= |ψ = d3 k|k k| d3 kψ(k)|k ψ(k) ≡ k|ψ (If we were doing NRQM. satisfying: U (a)U (a)† = 1. How do we know this theory is L. a. Momentum operator → P| k = k|k State of a spinless particle is completely specified by its momentum.I. that there are no others. to preserve probability amplitudes U (0) = 1 U (a)U (b) = U (a + b) (1.) We take H|k = |k|2 + µ2 |k ≡ ωk |k That’s it. By Lorentz invariance we mean SO(3. there should be a linear operator. September 23 Notes from Sidney Coleman’s Physics 253a 8 Lorentz Invariance Every Lorentz transformation is the product of an element of the connected Lorentz group. Theory of a single free spinless particle of mass µ The components of momentum form a complete set of commuting variables. k|2 and thus the time evolution: H|k = |2µ |k . we’d finish describing the theory by giving the Hamiltonian. P (reflects all three space components). T (time reversal). specifying a translation (active). 1). Translational Invariance Given a four-vector. SO(3. made relativistic. it is not necessarily relativistic. the theory of a single free spinless particle. or P T .1.

∂U 2. Pj ] = 0.3) [Pi .1) P = P † ). but correct: ˆ qe−iP ·a |q = (q + a)|q + a ˆ eiP ·a qe−iP ·a |q = (q + a)|q ˆ ˆ ⇒ eiP ·a qe−iP ·a = q + a Looks like the opposite of e−iP ·a O(x)eiP ·a = O(x + a) ˆ The q operator is not an operator localized at q. .1. September 23 Notes from Sidney Coleman’s Physics 253a 9 The U satisfying these is U (a) = eiP ·a where P = (H. 4. No reason for these (last two equations) to look alike. by (1. O(x + a) = U (a)O(x)U (a)† a|O(x + a)|a = 0|O(x)|0 Non-relativistic reduction U (a) = e−iP ·a e−iP ·a |q = |q + a Something hard to digest.) More translational invariance U (a)|0 = |a U (a) = eiP ·a where |0 here means state centered at zero and |a means state centered at a. 3. P ). The theory would contain unitary translation operators U (a). not logical order. Define P i = −i ∂qi ||a=0 . Define H = p2 + µ2 . Declare P i to be a complete set and classify the states by momentum. That we want to set up a translationally invariant theory of a spinless particle. The logical order would be to state: 1. thus giving the time evolution. (by (1. (This lecture is in pedagogical.

4) is satisfied U (R)U (R)† = U (R) = = d3 k|k k|U (R)† k = Rk.6) d3 k|Rk Rk| d3 k |k k|=1 Proof that (1.6) = U (R−1 )P U (R−1 )† = U (R−1 )P = U (R−1 ) = = d3 k|k k|U (R−1 )† d3 k k|k k|U (R−1 )† k = Rk . U (R)† P U (R) = RP and U (R)† HU (R) = H A U (R) satisfying all these properties is given by U (R)|k = |Rk That (1.4) U (R)−1 P (U (R)−1 )† ↓ by (1.6) are satisfied is trivial. d3 k = d3 k (1.e. d3 k = d3 k d3 k k|R−1 k R−1 k| d3 k Rk |k k| = RP .5) and (1. there should be a U (R) satisfying U (R)U (R)† = 1 U (1) = 1 U (R1 )U (R2 ) = U (R1 R2 ) Furthermore denote |ψ = U (R)|ψ ψ |P |ψ = R ψ|P |ψ for any |ψ i.5) and (1.1.5) (1.7) is satisfied U (R)† P U (R) = by (1. September 23 Notes from Sidney Coleman’s Physics 253a 10 Rotational Invariance Given an R ∈ SO(3).7) (1. Proof that (1.4) (1.8) (1.

I.8).e. Suppose a silly physicist took |k s s s = 2 1 + kz |k k|k 2 = (1 + ks )δ (3) (k − k ) 1 1 = d3 k |k s s k| 2 1 + kz If he took Us (R)|k s = |Rr s his proofs of (1. d3 k is not a Lorentz invariant measure. k 0 > 0. k = (ωk . This is the template for studying L. We want a Lorentz invariant measure on the hyperboloid k 2 = µ2 . So factors of 2π come out right in the Feynman rules a few months from now k o k 1 Think of the δ function as a function of k 0 and use the general formula δ(f (k 0 )) = zeroes of f . = (2π)3 2ωk |k . 2ωk ωk = k 2 + µ2 . This yields the measure on the hyperboloid1 So we take |k relativistically normalized d3 k . (1.1. Ki δ(k0 −Ki ) |f (Ki )| . d4 k is a Lorentz invariant measure. September 23 Notes from Sidney Coleman’s Physics 253a 11 You supply proof of (1. k). d3 k is not a rotationally invariant measure! 2 1 + kz d3 k d3 k = 2 1 + kz 1 + kz2 Lorentz Invariance k|k = δ (3) (k − k ) is a silly normalization for Lorentz invariance. .4). Restrict it to the hyperboloid by multiplying it by a Lorentz invariant d4 kδ(k 2 −µ2 )Θ(k 0 ).8) would break down because i.7). (1.

This is an illusion. Xj ] = δij .9) (1.13) (1.12) We have a fairly complete theory except we still don’t know where anything is. Determination of X in position space ψ(k) ≡ k|ψ k|X|ψ = i ∂ ∂k ψ (k) + kF (|k|2 ) ψ(k) an arbitrary vector commuting with the P ’s. September 23 Notes from Sidney Coleman’s Physics 253a 12 (Using d3 xδ(x2 − R2 ) can get R sin θdθdφ) 2 3 Looks like factor multiplying d k ought to get larger as |k| gets large.15) of (1.14) (1. satisfying X = X† RX = U (R) XU (R) eiP ·a Xe−iP ·a = X + a Take ∂ ∂ai † (1.10) (1. The moving twin’s path looks longer. but in fact. using 1= d3 k |k k| (2π)3 2ωk d3 k d3 k = (2π)3 2ωk (2π)3 2ωk (1. a complete set. Now the demonstration of Lorentz invariance: Given any Lorentz transformation Λ define U (Λ)|k = |Λk U (Λ) satisfies U (Λ)U (Λ)† = 1 U (1) = 1 U (Λ1 )U (Λ2 ) = U (Λ1 Λ2 ) U (Λ)P U (Λ) = ΛP The proofs of these are exactly like the proofs of rotational invariance.11) (1. We need a position operator.1. caused by graphing on Euclidean paper. can be written this way k|P |ψ = kψ(k) to satisfy inhomogeneous part of commutation relation . This is the same illusion as in the twin paradox. its proper time is shorter.15) to get i[Pi .

it is N k|X|ψ n = d3 kN k|k eiG(|k| ) δ (3) (k−k ) 2 2 k |X|ψ [i ∂∂ k N |2 ) ψ(k +k F (|k |2 )]eiG(|k ) = =i d3 k e−iG(|k| ) δ (3) (k − k ) ∂ ∂k ψ(k) ∂ ∂ ¨¨ $ ¨ k F $$ − ¨G(|k |2 ) + $$(|k |2 ) eiG ψ + eiG ψ ¨k ∂ ∂k ¨ ¨ Up to an unimportant phase definition. and we have done it without using L.) G(|k|2 ) = kF (|k|2 ) The only formula this affects in all that we have done so far is the expression for k|X|ψ . September 23 Notes from Sidney Coleman’s Physics 253a 13 The extra arbitrary vector can be eliminated by redefining the phases of the states.1. In effect.I. With the redefined states. . call it U . U † XU is our new position operator. we have shown that the obvious definition for X is the unique definition. or the explicit form of H. In the momentum state basis let |k → |k Here N = eiG(|k| ) |k 2 (This is a unitary transformation.

x is an observable. k = |k| d(cos θ) 1 1 (2π)2 ir −i = (2π)2 r = ∞ kdk(eikr − e−ikr )e−iωk t 0 ∞ ωk = k2 + µ 2 kdkeikr e−iωk t −∞ These steps can be applied to the F. and this leads to a conflict with causality. Introduce position eigenstates k|x = 1 e−ik·x (2π)3/2 We will evaluate x|e−iHt |x = 0 and see that our particle has a nonzero amplitude to be found outside its forward light cone 2 . 2 .T. x|e−iHt |x = 0 = = = 0 d3 k x|k k|e−iHt |x = 0 d3 k ∞ H|k = ωk |k ωk = k 2 + µ2 1 ik·x −iω t e e k (2π)3 k2 dk π sin θdθ (2π)3 0 1 −1 2π dφ eikr cos θ e−iωk t 0 2π r = |x|. k branch cut for Im <0 iu Im >0 -iu By translational invariance and superposition we could easily get the evolution of any initial configuration from this calculation. September 25 Notes from Sidney Coleman’s Physics 253a 14 2 September 25 It is possible to measure a particle’s position in our theory. of any function of |k|.2.

how does nature get out of it? Particle trapped in container with reflecting walls . Given r > 0 and r > t deform the contour to: half circle at infinity does not contribute The integral becomes −i (2π)2 r ∞ µ k=iz (−z) d(iz)e−zr [e √ z 2 −µ2 t √ e−µr 2 2 − e− z −µ t ] = − 2 2π r ∞ zdze−(z−µ)r sinh( z 2 − µ2 t) µ The integrand is positive definite. In principle. but if it is at all possible. how does this affect atomic physics? Not at all. Given r > 0 and r > t the product eikr e−iωk t decreases exponentially as you go up the branch cut. and that makes it extremely unlikely that I could go back and convince my mother to have an abortion. and we can release it and detect it outside of its forward light cone. the integral is nonzero. because we never tried to localize particles to spaces of order their Compton wavelength when doing atomic physics.2. September 25 Notes from Sidney Coleman’s Physics 253a 15 √ 2 2 ei k +µ is a growing exponential as you go up the right side of the upper branch cut. Admittedly. In practice. the chance that the particle is found outside the forward light cone falls off exponentially as you get further from the light cone. it is still an unacceptable contradiction. We can measure a particle’s position in this theory. The particle can travel faster than light and thus it can move backwards in time. with all the associated paradoxes. We can trap it in a box of arbitrarily small size. the ability to localize a single particle is a disaster. We say that the electron is in the TV picture tube and there is not much chance that it is actually out in the room with you. and a decreasing exponential on the left side.

k2 etc. k2 |k1 . k1 indistinguishability. 27 particles or 0 particles. k2 = |k2 . k2 = (ωk1 + ωk2 )|k1 . If the box contained a photon and the walls were mirrors the photon would pick up energy as it reflected off the descending mirror. September 25 Notes from Sidney Coleman’s Physics 253a 16 If the particle is localized to a space with dimensions on the order of L the uncertainty in 1 the particle’s momentum is ∼ L . This formalism is also used in thermodynamics with the grand canonical ensemble. Any number of one type of free spinless mesons The space we construct is called Fock space. fluctuates around a value determined by the chemical potential. k2 = (k1 + k2 )|k1 . Bose statistics k1 . Basis for single particle states |k k |k = δ (3) (k − k ) H|k = ωk |k P |k = k|k This is the same as last time. Two particle states |k1 . Particle number instead of being fixed. 3 particles. As L gets less than µ states with more than one particle are energetically accessible. This strongly suggests that the next thing we should do is develop a multi-particle theory. k2 = δ (3) (k1 − k1 )δ (3) (k2 − k2 ) + δ (3) (k1 − k2 )δ (3) (k2 − k1 ) H|k1 . Relativistic causality is inconsistent with a single particle theory.2. k2 P |k1 . If we try to localize a particle in a box with dimensions smaller or on the order of a Compton wavelength it is unknown whether what we have in the box is 1 particle. . In the relativistic regime this tells us that the uncertainty 1 1 in the particle’s energy is ∼ L . The real world evades the conflict through pair production. it could turn into two photons as it reflected. except now this is just part of the basis.

Completeness relation 1 = |0 0| + d3 k|k k|+ 1 2! d3 k1 d3 k2 |k1 k2 k1 . k2 = δk1 k δk2 kh2 + δk1 k δk2 k 1 2 1 Occupation number representation Each basis state corresponds to a single function |k1 ↔ n(k) = δkk1 |k1 . 0. a wave function contains a number. the momenta must be restricted to allowed values k= 2πnx 2πny 2πnz . k2 ↔ n(k) = δkk1 + δkk2 |0 ↔ n(k) = 0 . This would be a mess. As a pedagogical device. To specify a state. k2 | to avoid double counting. 0. Alternatively. we will work in a periodic cubical box of side L for a while. k2 |k1 .k Now we could proceed by setting up equations for wave functions.2. Since a translation by L does nothing. 0) = 2ny π k · (L. All observers agree that the state with no particles is the state with no particles. a function of six variables. L. Interactions involving a change in particle number will connect a function of six variables to a function of nine variables. it must satisfy U (Λ)|0 = |0 . . 0) = 2nx π Dirac deltas become Kronecker deltas and integrals become sums δ (3) (k − k ) → δkk k|k = δkk d3 k → k k1 . just check that this works on |k. a function of three variables. L) = 2nz π satisfying k · (0. etc. September 25 Notes from Sidney Coleman’s Physics 253a 17 Also need a no particle state |0 0 |0 =1 H|0 = 0 P |0 = 0 not part of a continuum The vacuum is unique. We need a better description. L L L k · (0.

We will review the oscillator using the operator formalism. Define raising and lowering operators q + ip a≡ √ 2 † † [H.2. a] = −ωa H = ωa† a [a. September 25 Notes from Sidney Coleman’s Physics 253a 18 Given a function n(k) in the occupation number description. to emphasize that the state depends on the whole function n. A] = 0 then A = λI. N (k) gives the excitation level of the oscillator labelled by k. In a system of oscillators. 1 H = ω[p2 + q 2 − 1] 2 If [p. We will then exploit the formal similarity to Fock space to get an operator formulation of our multi-particle theory. a ] = ωa q − ip a† = √ 2 [H. not just its value for one specific k. q] = −i Ha† |E = a† H|E + ωa† |E = (E + ω)a† |E Ha|E = (E − ω)a|E a† is the raising operator . a† ] = 1 [p. A] = [q. The two systems are completely different. Note that H for our system has the form it would have if the system we were dealing with was actually a bunch of harmonic oscillators. In ours the N (k) tells how many particles are present with a given momentum. Review of the simple harmonic oscillator No physics course is complete without a lecture on the simple harmonic oscillator. Inner product n(·)|n (·) = k n(k)!δn(k)n (k) Define an occupation number operator N (k)|n(·) = n(k)|n(·) H= k ωk N (k) P = k kN (k) This is a better formalism. we write the state | n(·) no argument. but it could still use improvement. It would be nice to have an operator formalism that did not have any wave functions at all. This is all we need to get the spectrum.

P] = 0 ⇒ P = λI The projector onto the ladder is proportional to the identity. having E0 = 0 is denoted |0 . k2 . A] = 0 ⇒ A = λI to show that this ladder built from a state with E0 = 0 is in fact the whole space. ψ|H|ψ = ω ψ|a† a|ψ = ω||a|ψ ||2 ≥ 0 the ladder of states must stop going down or else E becomes negative. September 25 Notes from Sidney Coleman’s Physics 253a 19 Ladder of states at at a E+2 E+ E E- Because. Define creation and annihilation operators for each momentum. There is nothing besides the states we have found. P] = [q. k3 k k k 1 2 3 ak |0 = 0 ka† ak k k H= k ω k a† ak k P = . P] = 0 ⇒ [p. ak . Now we will apply this to Fock space. P] = [a† . The lowest state of the ladder. |k = a† |0 k a† a† a† |0 = |k1 . Since a and a† keep you within the ladder [a. A] = [q. a† . The only way this can happen is if a|E0 = 0. Then H|E0 = 0. We do this by considering the projector. ak ] = 0 [a† . The higher states are made by (a† )n |0 ∝ |n Get normalizations right a |n = cn |n + 1 |cn |2 = n|aa† |n = n + 1 ⇒ a|n = dn |n − 1 cn = † H|n = nω|n √ n+1 dn = |dn |2 = n √ n Now we use [p.2. a† ] = 0 k k Hilbert space built by acting on |0 with strings of creation operators. P. onto the states in the ladder. satisfying k [ak . a† ] = δkk k [ak .

a† ] k k P = d3 k ka† ak k Check energy and normalization of one particle states. A] =⇒ k 20 A = λI This tells us there are no other degrees of freedom. k2 = a† a† |0 . An extra meson in a k plane wave state added to any state is enough to make it nonnormalizable. [ak . ak ] = ωk ak ⇒ k H|k = Ha† |0 = [H. d3 kf (k)ak is acceptable with a “sufficiently smooth” function. A] = [a† . An operator takes any normalizable vector in k Hilbert space to another normalizable vector: A|ψ is normalizable whenever |ψ is normal¨ izable. k2 |k1 . a† ]|0 = ωk |k k k Mathematical Footnote: We’ve been calling the ak and a† operators. Using commutation relak1 k2 tions check k1 . Let’s drop the box normalization and see if it is working. . a† ] = δ (3) (k − k ) k H= d3 kωk a† ak k [ak . We’ve laid out a compact formalism for Fock space. ak ] = 0 = [a† . An unbounded operator has A|ψ normalizable for a dense set3 of |ψ . 3 Any |ψ is the limit of a sequence in the dense set. You only get something a mathematician would be happy with after integration. Even x in 1-d QM is not an operator. x is ⇒ an unbounded operator.2. dx|f (x)|2 < ∞¨ dx|f (x)|2 x < ∞. ak is an operator valued distribution. September 25 Notes from Sidney Coleman’s Physics 253a If ∀k 0 = [ak . a† ]|0 k k = δ (3) (k − k ) 0|0 = δ (3) (k − k ) [H. a† ] = ka† ⇒ P |k = k|k k k Check normalization of two-particle states |k1 . k2 = δ (3) (k1 − k1 )δ (3) (k2 − k2 ) + δ (3) (k2 − k1 )δ (3) (k2 − k1 ) d3 k ωk [a† ak . k |k = 0|ak a† |0 = 0|[ak . ak and a† are even more awful than unbounded operators. a† ] = k [P.

3. observables are given by operators. but it does have an analog in classical field theory. O2 ] = 0 Observables are attached to space time points. Ez (x). In electromagnetism. By (x). nor in the classical theory of a single particle. any hermitian operator is observable. Bx (x). We can’t design an apparatus here that measures the Ex field now in the Andromeda galaxy. even in principle for everyone to measure everything. The attachment of observables to space-time points has no analog in NRQM. We don’t have any way of localizing particles. relativistic or nonrelativistic. It is a little hard to mathematically state the obvious experimental fact that I can’t measure the spin of an electron in the Andromeda galaxy. yet. If x1 ∈ R1 and x2 ∈ R2 then (x1 − x2 )2 < 0. there are six observables at each point: Ex (x) = Ex (x. Bz (x). September 30 Notes from Sidney Coleman’s Physics 253a 21 3 September 30 In ordinary QM. Imagine two experiments that are at space-like separation. an operator for each spacetime point. t). This is faster than light communication. A given observer cannot measure all observables. This can’t be true in relativistic quantum mechanics. only a restricted set can be measured at a space-time region. Ey (x). any position operator. It is not possible. R1 space-time region accessible to observer 1 R2 space-time region accessible to observer 1 Suppose observer 2 has an electron in his lab and she measures σk . The fields will become quantum fields. these observables are numbers. an impossibility. If observer 1 can measure σy of that electron it will foul up observer 2’s experiment. Half the time when she remeasures σx it will have been flipped. In quantum mechanics. Out of the hordes of observables. only the ones associated with his or her region of space-time. In classical field theory. This tells her that observer 1 has made a measurement. Localization of measurements is going to substitute for localization of particles. pith . We can see in another way that the electric field is going to have to become a quantum field: How would you measure an electric field? You mount a charged ball. We can make a mathematical statement in terms of observables: If O1 is an observable that can be measured in R1 and O2 is an observable that can be measured in R2 and R1 and R2 are space-like separated then they commute: [O2 .

like charge density is f (x) = ψ|ρ(x)|ψ . NOT BOTH! What’s embodied in assumption (3): Given U (a) the unitary operator of space translation by a (U (a) = e−iP ·a ) the translation of a state |ψ is a state |ψ = U (a)|ψ . In fact. These five conditions will determine them: 2. observable and that they have proper translation and Lorentz transformation properties 3. and so Ex must be an operator.. We don’t have a proof. e−iP ·a φa (x)eiP ·a = φa (x − a) 4. Suppose the value of some observable.N operator valued functions of space-time. to some springs and see how much the springs stretch. September 30 Notes from Sidney Coleman’s Physics 253a 22 ball. The location of the ball is given by an equation like: q¨ = Ex (x) ←− x might be d4 xf (x)Ex (x) where f (x) gives some suitable average over the pith ball. We will try to build our observables from a complete commuting set of quantum fields. or as transformations on the operators A → U † AU . Observables in a region R will be built out of φa (x) with x ∈ R. and if the world is quantum mechanical. φa (x) = a d3 k[Fk (x)ak + Ga (x)a† ] k k We can think of our unitary transformations in two ways. [φa (x). Observables in space-like separated regions will be guaranteed to commute if 1.3. The amount the ball moves is related to the E field.. φa (x) a=1. We are going to construct our fields out of the creation and annihilation ops.. φb (y)] = 0 whenever (x − y)2 < 0. but what is strongly suggested is that QM and relativistic causality force us to introduce quantum fields. as well as a change of argument 5. as transformations on the states |ψ → U |ψ .. x must be an operator. U (Λ)† φa (x)U (Λ) = φa (Λ−1 x) If this were not a scalar field. there would be extra factors here reflecting a change of basis. φa (x) = φa† (x) hermitian. relativistic QM is practically synonymous with quantum field theory.

. The equation with a = (t. f(x) f(x. Recall the relativistically normalized one particle states |k = (2π)3/2 2ωk |k Introduce α† (k) = (2π)3/2 2ωk ak . .a) x a Rewrite the first equation with x → x − a x f (x − a) = ψ|ρ(x − a)|ψ Equate ψ|ρ(x − a)|ψ = ψ |ρ(x)|ψ = ψ|eiP ·a ρ(x)e−iP ·a |ψ A hermitian operator is determined by its expectation values eiP ·a ρ(x)e−iP ·a = ρ(x − a) (3) is just the full relativistic form of this equation. α† (k)|0 = |k . Λ−1 x appears in the RHS of (4). (4) gives the Lorentz transformation properties of a :::::: field. In order to apply condition (4). September 30 Notes from Sidney Coleman’s Physics 253a 23 then it should be that ψ |ρ(x)|ψ = f (x − a). Out of vector or tensor fields. We can get fields transforming as vectors or tensors by taking derivatives of the φa . . This is not much of scalar an assumption. kn . Multiparticle states are made by α† (k1 ) · · · α† (kn )|0 = |k1 .3. For the exact same reason as x − a appears in the RHS of (3). . 0) is just the time evolution for Heisenberg fields. we could make scalars. it is nice to have the discussion phrased in terms of relativistically normalized creation and annihilation operators.

(1.1)-(1. . . . . .3) of the Sept. . .I. k1 . . . kn = (k1 + · · · + kn )µ |k1 . Λ−1 k1 . . . . . We can determine Lorentz transformation and translation properties of the α† (k). kn is an arbitrary state in our complete basis so we have determined its action completely U (Λ)α† (k)U (Λ)† = α† (Λk) Similarly. . September 30 Notes from Sidney Coleman’s Physics 253a 24 The Lorentz transformation properties of the states are U (Λ)|0 = 0 U (Λ)|k1 . kn = |Λk1 . . U (Λ)α† (k)U (Λ)† |k1 .3. .9)-(1.12) and Eqs. . . . . At the time our Hilbert space consisted only of the one particle part of the whole Fock space we have now. . or by taking the adjoint of this equation U (Λ)α(k)U (Λ)† = α(Λk) An analogous derivation shows that eiP ·x α† (k)e−iP ·x = eik·x α† (k) eiP ·x α(k)e−iP ·x = eik·x α(k) Now to construct the field φ (if there is more than one we’ll label them when we’ve found them) satisfying all 5 conditions. First we’ll satisfy condition (5) except we’ll write the linear combination of ak and a† in terms of our new α(k) and α† (k) k It would be stupid not to use the L. . . Λ−1 kn = |Λk. . . . kn That is U (Λ)α† (k)U (Λ)† |k1 . . . . kn In Eqs. Λkn and U (a) = eiP ·a is found from P µ |0 = 0 P µ |k1 . . kn |k1 . kn = α† (Λk)|k1 . . . . . . . . (1. . Consider. kn = U (Λ)α† (k)|Λ−1 k1 . . . 23 lecture we set up the criteria that U (Λ) and U (a) must satisfy. measure φ(x) = d3 k [fk (x)α(k) + gk (x)α† (k)] 3 2ω (2π) k . . You should check that the criteria are satisfied in Fock space. Λ−1 kn = U (Λ)|k. . .

but it was est’d by Heisenberg and Pauli 50 years ago. September 30 Notes from Sidney Coleman’s Physics 253a 25 By (3). So fk (0) and gk (0) are constants. the coefficients of the complex constants f and g. but a Lorentz transformation can turn any of these k’s into any other. a† eik·x k Now we’ll apply hermiticity. (Switching back to our old creation and annihilation ops. A special case of (4) is φ(0) = U (Λ)φ(0)U (Λ)† d3 k [fk (0)α(k) + gk (0)α† (k)] = (2π)3 2ωk d3 k [fk (0) U (Λ)α(k)U (Λ)† +gk (0) U (Λ)α† (k)U (Λ)† ] 3 2ω (2π) k α(Λk) α† (Λk) change variables measure is unchanged 3 k → Λ−1 k = dk [fΛ−1 k (0)α(k) + gΛ−1 k (0)α† (k)] (2π)3 2ωk The coefficients of α(k) and α† (k) must be unchanged ⇒ fk (0) = fΛ−1 k (0) and gk (0) = gΛ−1 k (0). φ(x) = d3 k [f e−ik·x α(k) + geik·x α† (k)] (2π)3 2ωk We have two linearly independent solutions of conditions (3). φ(x) = eiP ·x φ(0)e−iP ·x . independent of k. k ranges all over the mass hyperboloid (k 0 > 0 sheet). that is.) φ+ (x) = − + d3 k (2π)3/2 † 2ωk ak e−ik·x φ− (x) = d3 k (2π)3/2 Note φ (x) = φ (x) 2ωk ± convention is bananas.3. Two independent combinations satisfying (2) are 1 φ(x) = φ+ (x) + φ− (x) and φ(x) = [φ+ (x) − φ− (x)] i . φ(x) = = d3 k [fk (0)eiP ·x α(k)e−iP ·x + gk (0)eiP ·x α† (k)e−iP ·x ] (2π)3 2ωk d3 k [fk (0)e−ik·x α(k) + gk (0)eik·x α† (k)] (2π)3 2ωk We have found the x dependence of fk (x) and gk (x) now we will use (4) to get their k dependence. We’ll name them. (4) and (5).

Possibility B: Only one combination is acceptable. and we could weaken (5) or think harder. In fact. φ (y)] = −∆+ (x − y. φ(y)] = 0 when x − y is space-like. Possibility A: Both of the above combinations are OK. We’ll choose θ = 0. and we can see this without any calculations. This and the fact that ∆+ is Lorentz invariant. φ− (y)] = (2π)3 2ωk or just ∆+ (x−y) also [φ (x). Possibility C: The program crashes. but we can’t choose more than one θ. While θ may be arbitrary. We have two fields φ1 and φ2 commuting with themselves and each other at spacelike separation. k . µ ) ∆+ is manifestly Lorentz invariant ∆+ (Λx) = ∆+ (x). We can choose θ arbitrarily. tells us that [φ(x). In this possibility φ+ (x) and φ− (x) commute with each other at spacelike separation. ::: Possibility A is DEAD. φ+ (y)] d3 k e−ik·(x−y) ≡ ∆+ (x − y. September 30 Notes from Sidney Coleman’s Physics 253a 26 These are two independent cases of the most general choice satisfying (2): φ(x) = eiθ φ+ (x) + e−iθ φ− (x) Now to satisfy (1).3. Using their expansions in terms of the ak and a† k and the commutation relations for ak and a† we find k [φ+ (x). It is of the form φ = eiθ φ+ +e−iθ φ− . and we found that it did not vanish when (x − y)2 < 0. Let’s first calculate some commutators. i∂0 ∆+ (x − y) = d3 k −ik·(x−y) e (2π)3 − + 2 is the very integral we studied. Possibility A runs only if ∆+ (x − y) = 0 for (x − y)2 < 0. Any phase could be absorbed into the ak and a† . We have encountered a similar integral when we were looking at the evolution of one particle position eigenstates. only one θ is acceptable. µ2 ) and [φ+ (x). On to possibility B. φ(y)] = ∆+ (x − y) − ∆+ (y − x) θ dependence drops out Does this vanish when (x − y)2 < 0? Yes. A space-like vector has the property that it can be turned into minus itself by a (connected) Lorentz transformation. There are three possible outcomes of trying to satisfy (1). φ− (y)] = 0 = [φ− (x). Take φ(x) = eiθ φ+ (x) + e−iθ φ− (x) and calculate [φ(x).

It is called the Klein-Gordon equation. If we had quantized the electromagnetic field it would have satisfied Maxwell’s equations. he knew about relativity. He immediately saw that something was wrong with the equation though. Schr¨dinger first wrote down the Klein-Gordon equation. so he also obtained ( +µ2 )φ(x) = 0 o from p2 = µ2 . but with 50 years hindsight we see that this is no problem for a field that can create and destroy particles. The equation has both positive and negative energy solutions. . i Schr¨dinger was no dummy. and we don’t have to go on to C.3. We have our free scalar field of mass µ φ(x) = φ+ (x) + φ− (x) = d3 k (2π)3/2 2ωk [ak e−ik·x + a† eik·x ] k Our field satisfies an equation (show using k 2 = µ2 ) ( + µ2 )φ(x) = 0 = ∂ µ ∂µ This is the Heisenberg equation of motion for the field. He got it at the same o time as he got the Schr¨dinger equation: o i∂0 ψ = − 1 2µ 2 ψ 2 p This equation is obtained by starting with E = 2m noting that E = ω (when = 1) and pi = k i and for plane waves ω = i∂0 and k i = 1 ∂i . Actually. September 30 Notes from Sidney Coleman’s Physics 253a 27 Possibility B is ALIVE. For a free particle the energies of its possible states are unbounded below! This is a disgusting relativistic generalization of a single particle wave equation.

let’s define These funny factors will make the commutation relations for the ak and a† k come out nice a† † k βk = αk = 3/2 (2π) 2ωk αk = ak (2π)3/2 2ωk . φ(x)] = 0 ⇒ A = λI. i. Conversely. Because of † property (1). K. we can reconstruct the entire theory from the quantum field. and the plane wave solutions are a complete set. Suppose we started with a quantum field satisfying (as our φ does). ( + µ2 )φ(x) = 0.-G. hermiticity 2. Just to please my little heart. we could reconstruct the creation and annihilation operators and the whole theory. φ(x) = φ† (x). that φ is a solution of the Klein-Gordon equation. [φ(x). The structure we built in the last three lectures is rigid. Defining ak and a† and recovering their properties k Property (2) of φ. all these properties follow from the expression for φ(x) in terms of the creation and annihilation operators. October 2 Notes from Sidney Coleman’s Physics 253a 28 4 October 2 We have constructed the quantum field. 1. φ(x) is a complete set of operators.. It is the object observables are built from.-G. U (Λ)† φ(x)U (Λ) = φ(Λ−1 x) U (a)† φ(x)U (a) = φ(x − a) d3 k [e−ik·(x−y) (2π)3 2ωk − eik·(x−y) ] φ is a scalar field 5. Because φ is an operator. the coefficients in the expansion. then from these properties. tells us that φ(x) = d3 k[αk e−ik·x + βk eik·x ] (k 0 = ωk = k 2 + µ2 ) This is because any solution of the Klein-Gordon equation can be expanded in a complete set of solutions of the K. We can make the top story the foundation.e. More than that though. equation 3. φ(y)] = ∆+ (x − y) − ∆+ (y − x) = 4. eqn. The two beginning points are logically equivalent. if ∀x [A. αk = βk . αk and βk .4. are operators.

0)e−ik·x (2π)3/2 1 eik·x (2π)3/2 The coefficient of ωk (−iak 2 ˙ in the expansion for φ(x. Write out k the double integral for [ak . 0)e−ik·x 3/2 − That’s just the inverse Fourier transformation applied. 0) + ωk (−iak 2 1 2 i ˙ φ(x. a† ] and use property (3’) below (which is weaker than property k (3)). You should get δ (3) (k − k ). 4 It is easier to find the action of Lorentz transformations on a(k) ≡ (2π)3/2 2ωk ak . 0) is + ia† k ) therefore − d3 ˙ + ia† k ) = (2π)x φ(x. 0) is √1 (ak + a† k ) therefore − 2ω ωk (−iak 2 d3 x φ(x. October 2 Notes from Sidney Coleman’s Physics 253a 29 Then the expression is φ(x) = d3 k (2π)3/2 2ωk [ak e−ik·x + a† eik·x ] k This defines implicitly the ak and a† . k Having solved for the ak and a† . 0) = While ˙ φ(x) = And ˙ φ(x. . From property (4). 0) e−ik·x ωk Take the hermitian conjugate to get an expression for a† . Note that φ(x. you can derive. To find their commutation relations we’ll first have k to solve for them. Take k 2ωk times the expression for + ia† k ) and divide by 2 to − √1 (ak + a† ) and add −k 2ω get ak = 2 ωk times the expression for d3 x (2π)3/2 φ(x. what U (a)† φ(x)U (a)† ak U (a) and U (Λ)† ak U (Λ)4 are.4. 0) = The coefficient of √1 (ak + a† ) = −k 2ω k d3 k (2π)3/2 d3 k 2ωk [ak e−ik·x + a† eik·x ] k (2π)3/2 2ωk [ak (−iωk )e−ik·x + a† (iωk )eik·x ] k ωk [ak e−ik·x + a† eik·x ] k 2 k d3 k (2π)3/2 1 eik·x (2π)3/2 in the expansion for φ(x. You can also derive and state the analog of property (5). we can now find their commutation relations. Now I can use these two expressions to solve for ak .

(4) and (5). [φ(x. t). φ(y. The equations in (3’) are called equal time commutation relations. t). So we will weaken (3) to 3’. which are easy with x0 = y 0 . and the K. For a given y.4.-G. We will obtain a field satisfying properties (1). qb ] = −iδb In the remainder of this lecture we are going to develop a completely different approach to quantum field theory. t).) [φ(x. ( x + µ2 )[φ(x). φ(y)] = 0 We see that the commutator obeys the K.-G. t) we have chosen our initial surface to be x0 = t. t)] = = d3 k (−iωk )[e−ik·(x−y) + eik·(x−y) ] (2π)3 2ωk Again k → −k in second term d3 k (−iω )2eik·(x−y) & ¡ (2π)3&& &k 2ωk d3 k [e−ik·(x−y) − eik·(x−y) ] (2π)3 2ωk Change variables k → −k in second term = −iδ (3) (x − y) Reminds us a little of a bad dream in a which [pa . t)] = =0 [∂0 φ(x.-G. (2). without losing anything. t)] = −iδ (3) (x − y) We can easily check that this is the right specialization of property (3) by doing the integrals. (3). equation determines its evolution off the surface. The new approach is . (For a given y = (y. t). φ(y. (3’). φ(y. We can weaken one of them. we only need to give the commutator and the time derivative on one initial time surface. (2). and then we’ll stop. t)] = 0 [∂0 φ(x. We have just shown that given a field satisfying these properties we can recover everything we did in the first three lectures. equation. October 2 Notes from Sidney Coleman’s Physics 253a 30 The properties of the field φ(x) are actually a little overcomplete. φ(y. Since φ obeys the K. equation for any given y.

Canonical quantization is a turn the crank way of getting quantum mechanics. q N .g. October 2 Notes from Sidney Coleman’s Physics 253a 31 The method of the missing box continuum limit classical particle mechanics classical field theory canonical quantization quantum particle mechanics quantum field theory Anyone looking at this diagram can see there is a missing box Taking the continuum limit essentially is just letting the number of coordinates. . . It doesn’t matter a whole lot if you have a discrete infinity or a continuous infinity. q 1 . but first. That is. Classical Particle Mechanics We start with a Lagrangian. a ∈ 1. With a continuous infinity you can Fourier transform the coordinates to obtain a discrete set of coordinates. . . we vary S by arbitrarily varying δq a (t) except at the endpoints δq a (t1 ) = δq a (t2 ) = 0 and demand 0 = δS . a lightning review of the essential principles of the three boxes we have. . . We’ll combine these two standard operators to get the missing box. N 1 ˙ e. . q N . t) ˙ ˙ q a (t). a function of the generalized coordinates. . . This is usually done in a cavalier way. and their time derivatives L(q 1 . . L = 2 mq 2 − V (q) t We define the action S = t12 dtL and apply Hamilton’s principle to get the equations of motion. .4. degrees of freedom go to infinity. beginning with a classical Hamiltonian. .

(This is not always possible. p = mq. p = − ˙ ˙ dV dq These are the Euler-Lagrange equations. q1 . . In this system. t) = pa q a − L ˙ a H must be written in terms of the p’s and q’s only. . not the q’s. October 2 Notes from Sidney Coleman’s Physics 253a 32 t2 δS = t1 dt a ∂L a ∂L a δq + a δ q ˙ ∂q a ∂q ˙ parts integration t2 = t1 dt a ∂L d − pa δq a + ∂q a dt ∂L pa ≡ ∂ q a ˙ t2 pa δq a $ $$$|t1 vanishes because of endpoint restriction Since the variation δq a (t) is arbitrary ∂L = pa . ˙ The new variables must also be independent. pN .4. . . polar coordinates for the system. and then trying again. In the passage to the Hamiltonian formulation pλ = 0. one way to et to the Hamiltonian description is to first eliminate λ and one more coordinate.5 pλ is not an independent function. so it is possible to vary them independently.) Vary the coordinates and momenta dH = a dpa q a + $$q a − ˙ pa d $ ˙ Fortunately we didn’t have to expand out dq a ˙ ∂L a ∂L a dq − a dq ˙ ∂q a ∂q ˙ pa by the E-L equations ˙ Read off Hamilton’s equations ∂H = qa ˙ ∂pa 5 ∂H = −pa ˙ ∂q a Notice: ∂L = 0 ⇒ dH = 0 ∂t dt in that case H is called the energy (which is the name reserved for the conserved quantity resulting from time translation invariance) How can I vary that?!? . . Define H(p1 . Examples where the p’s are not complete and independent are electromagnetism and a particle on a sphere in R3 . rewriting the Lagrangian. . . for all a ˙ ∂q a e. . The Lagrangian for the latter system may be taken as 1 ˙ L = mr2 + λ(r2 − a2 ) − V (r) 2 The equation of motion for the variable λ enforces the constraint. qN . φ. taking perhaps θ. Around 1920(!) the Hamiltonian formulation of classical particle mechanics was discovered.g.

Heisenberg equations of motion dq a = i[H. This step depends only on the commutation Fairly easy to see for a polynomial in the p’s and q’s. at least up to the ordering ambiguities. to be able to fix those ambiguities. which determines the dynamics. and any variation due to ordering ambiguities is down by a factor of . We wouldn’t expect any general procedure for turning classical theories into quantum theories. it is not clear. This is the motivation for the commutation relations. They are assumed to be complete. qp2 or pqp. q a by operator valued functions of time satisfying [q a (t). The quantum Hamiltonian. and it may matter. pb (t)] b and [pa (t). q a ] dt ∂H = i −i ∂pa ∂H = ∂pa ∂H dpa =− a Similarly. in central force problems. motivated only by the correspondence principle. It is a way of putting the correspondence principle into the theory. H suffers from ordering ambiguities. October 2 Notes from Sidney Coleman’s Physics 253a 33 Quantum Particle Mechanics We replace the classical variables pa . we quantize in Cartesian coordinates and then transform to central coordinates. q b (t)] = −iδa =1 The p’s and q’s are hermitian observables. For any A if there is explicit dependence on ∂A dA = i[H. p or rather eipx tells you how to relate the phases of the various eigenstates and indeed even that the range of q is R (not [-1. We can be a little more precise about how the classical equations are actually . A basis set can be the eigenstates of q. The quantum equations resemble the classical equations.1] or anything else). the space of square integrable functions. For example. They determine the Hilbert space. is just the classical Hamiltonian except it is now a function of the operator p’s and q’s. Sometimes this ambiguity can be cleared up using other criteria. q b (t)] = 0 = [pa (t). For example 1-d particle mechanics. dt ∂q ⇓ relations for the p’s and q’s.4. whether you write p2 q. Because p and q don’t commute. A] + ←− time other than that implicit in the time dependence of the p’s and q’s dt ∂t H is the generator of infinitesimal time translations just as in classical mechanics.

and this turns our equation among polynomials of quantum operators into an equation among classical variables. that is obviously incomplete. we’ll just replace it with a a a sum on a and an integral over x. For example we are used to giving initial value data at fixed t in CPM.c But note that a CPM Lagrangian does not contain products at different times. as generalization of t. expectations of products can be replaced by products of expectations. In CFT we don’t give initial value data at fixed t and x. In the classical limit. But again with an eye to Lorentz invariance. because 2 only first derivatives with respect to time appear in the Lagrangian. Also. October 2 Notes from Sidney Coleman’s Physics 253a 34 recovered. we will specialize to Lagrangians that are local in space. when fluctuations are small. real number functions of time. y. x It is sometimes a handy mnemonic to think of t.4. t)φb (y. x. t)] a2 ∂ would become ∂xρ in the continuum. t) − 2ρ(na. t) ↔ fabc q a (t)q b (t)q c (t) a. Classical Field Theory We have an infinite set of generalized coordinates. t) labelled by a discrete index. differences of neighboring variables become spatial derivatives and a combination like 1 [ρ((n − 1)a. you know that when taking continuum limits. q. everywhere in CPM we see a sum on a. but that is not the right way to think about it. We give initial value data for fixed t and all x. x.b. and noticing that Lagrangians are local in time. t)φc (z. because if you are just a little cavalier. With cowboy boldness. z)φa (x. the Heisenberg equations of motion for an arbitrary operator A relate one polynomial in p. In general. The other two boxes are actually going to go quite quickly. pq → p q . p and q to one another. t) with different x d3 xd3 yd3 zfabc (x. Because a CPM Lagrangian can contain products of q a with different a. we will only consider 2 . the continuum limit is little more than some new notation. and everywhere we see δb we’ll replace it with δb δ (3) (x − y). t) ↔ q a (t) t↔t a ↔ a. With an eye to Lorentz invariance. The next thing to do is write down Lagrangians. The Dirac delta function has the exact same properties in integrals that the Kronecker delta has in sums. and a continuous index. a. φa (x. t) + ρ((n + 1)a. We can take the expectation value of this ˙ ˙ equation to obtain (a quantum mechanical average of) an equation between observables. pn → p n . our Lagrangian could contain products of φ(x. φa (x.

As usual. October 2 Notes from Sidney Coleman’s Physics 253a 35 first derivatives with respect to the xi . we’ll just be slothful.” Alternatively we could work in a box with periodic boundary conditions. t2 ) = 0 d4 x ∂L ∂φa (x. Now we’ll apply Hamilton’s principle 0 = δS = a under arbitrary variations δφ satisfying δφa (x. not generalizations. Anyway. These have been specializations. t1 ) = δφa (x. x) t2 dtL(t) = d4 xL The Euler-Lagrange equations which come from varying S will be Lorentz covariant if L is a Lorentz scalar. A simple example of a possible L Most general L satisfying (3) conditions 1. ∂µ φa (x). We should say something about the surface terms at spatial infinity which we ignored when we did our parts integration. At three points we have used Lorentz invariance to cut down on the possible forms for L. t) + ∂L δ∂µ φa ∂∂µ φa µ ≡πa ∂µ δφa ⇓ restrictions on δφa make the surface terms at t1 and t2 drop out Do parts integration. One can say “we are only considering field configurations which fall off sufficiently rapidly at spatial infinity that we can ignore surface terms. Build L out of one reals calar field. The correspondence is 0 πa (x. the = a ∂L µ dx − ∂µ πa δφa ∂φa 4 Euler-Lagrange equations ⇒ ∂L µ = ∂µ πa (for all x and a) ∂φa µ πa should not be thought of as a four-vector generalization of pa . So L has the form L(t) = And the action S = t1 d3 kL(φa (x). L is a Lorentz scalar .4. we’re not doing QFT yet) 2. t) δφa (x. φ φ = φ∗ (not φ = φ† . t) ↔ pa (t) 0 In fact πa is often jut written πa .

October 2 Notes from Sidney Coleman’s Physics 253a 36 3. if the energy is to be bounded below. we are always free to rescale φ. and . We are ready to canonically quantize classical field theory. φ → √ . L |a| becomes 1 L = ± [∂µ φ∂ µ φ + bφ2 ] 2 The Euler-Lagrange equations for our example are πµ ≡ ∂µ π µ − ∂L = ±∂ µ φ and ∂∂µ φ ∂L = 0 or ± (∂µ∂µ φ − bφ) = 0 ∂φ a In general the Hamiltonian which was H= a pa q a − L in CPM is ˙ d3 xH 0 d3 x(πa ∂0 φa − L) = H= a 0 πa ∂0 φa − L is the Hamiltonian density In our example 1 1 d3 x[ (π 0 )2 + ( φ)2 − bφ2 ] 2 2 Since each of these terms separately can be made arbitrarily large. the easiest ones to solve. Most general L is 1 L = a[∂µ φ∂ µ φ + bφ2 ] 2 φ One of these constants is superfluous. Since we are not worrying about the passage to an infinite number of variables. A motivation for (3) is that a quadratic action yields linear equations of motion.4. . this is looking familiar now. It would not even have required a notational change if Newton hadn’t chosen two ways of writing S for sum. ± better be + and b better be H=± b = −µ2 definition of µ ≥ 0 The E-L equation is now ∂µ ∂ µ φ + µ2 φ = 0 Gosh. this will be little more than a notational change in the canonical quantization of CPM. they each better have a positive coefficient. L is quadratic in φ and ∂µ φ (1) and (3) are for simplicity. We are ready to fill in the last box.

the quantum field. φ(y. (4) and (5). φb (y. but boy are there going to be booby traps. Looks easy. We have reproduced our quantum field satisfying (1). t) = 0 = [πa (x. πa by quantum operators satisfying 0 0 [φa (x.4. but let’s check that things are working in our example anyway. and about the only theory we know the exact spectrum for is the free theory. H= 1 d3 x [(π 0 )2 + ( φ)2 + µ2 φ2 ] 2 d3 xπ 0 (x. t)δ (3) (x − y) ∂0 φ(y. we can’t put interactions in in this method though. t). Since this is just a change of notation there is no need to redo any proofs. (3’). t)] 0 b and [πa (x. t)] = −iδa δ (3) (x − y) The Dirac delta for the continuous index x is just the continuum generalization of the 0 Kronecker delta. . The physical interpretation comes at the end instead of at the beginning. and we finally obtained a local observable. if there weren’t any booby traps. t) Similarly. t)] = i(−i) = π 0 (y. October 2 Notes from Sidney Coleman’s Physics 253a 37 Quantum Field Theory 0 We replace the classical variables φa . This was accomplished in one lecture rather than three because this is a mechanical method without physical insight. we shook each object we introduced to make sure it made sense. t). t). x) determines the dynamics as usual. φa . because in the very first steps we had to know the whole spectrum of the theory. (2). In our first method. At the first order in perturbation theory the part of λφ4 that has two creation and two annihilation operators produces two-into-two scattering! At second order we’ll get two-into-four and two-into-six 6 scattering. you need the commutator [π 0 (y. . . φ(x. In our magical canonical quantization method it’s easy to put in interactions: just let L → L − λφ4 . t) = i[H. φb (y. ∂0 π 0 (y. the constructive method. t) = 2 φ − µ2 φ. t). πb (y. t)] which is obtained by taking the gradient of the equal time commutation relation (ETCR). . 6 pair production! . The commutation relations are set up to reproduce the Heisenberg equations of motion. H = d3 xH(πa .

) Just as the spectrum of the Hamiltonian H = 1 (p2 + ω 2 q 2 ) starts at 1 ω. These all conserve momentum. H should not be time-dependent. H= @ 1 d3 k @@ 2 2 (−ωk + |k|2 + ak a−k e−2iωk t@@@@@@@ µ2 ) +ak a† (ωk + |k|2 + µ2 ) k 2 2ωk k =−k see footnote 1 That what multiplies e±2iωk t is 0 is good. 1 Using [ak . and ones creating a particle with momentum −k. then destroying one with momentum −k. but the x integration is easily done yielding a δ function.5. H becomes 1 2 k ωk (ak a† + ak a† ) = k k k a† ak + k 1 2 In a box we see that this is just a zero point energy. ones creating a particle with momentum k. Evaluate H= 1 2 d3 x[π 2 + ( φ)2 + µ2 φ2 ] ˙ π=φ Write φ in terms of its Fourier expansion φ(x) = d3 k (2π)3/2 2ωk [ak e−ik·x + a† eik·x ] k Substitution would lead to a triple integral. not zero. ones destroying particles with momentum k and −k. which does one of the k integrals7 . 2 2 (−ωk + |k|2 + + a† ak (ωk + |k|2 + µ2 ) + a† a† k e2iωk t @@@@@@@ µ2 ) k k − @ @@ H had four types of terms: ones creating particles with momentum k and −k. δ (3) (0)! Can get some k k idea of the meaning of this by putting the system in a box. H= 1 2 d3 kωk (ak a† + ak a† ) k k Almost. (It is still infinite though. the spectrum 2 2 7 d3 xeik·x eik ·x = (2π)3 δ (3) (k + k ) ⇒ k = −k . but not quite what we had before. We’ll double check that we get the same Hamiltonian we had in that theory. then destroying one with momentum k. October 7 Notes from Sidney Coleman’s Physics 253a 38 5 October 7 The simplest example obtained from applying canonical quantization to a free scalar field led to the same theory we got by constructing a local observable painstakingly in the theory we had before. a† ] = δ (3) (k − k ) gives H = d3 kωk a† ak + 2 δ (3) (0) .

maybe even to some grand unified theory.” If the energy density of an infinite crystal is changed by an infinite amount.5. Strong interactions have energies typically of 1 GeV and a characteristic length of a fermi. Just as the quantum Hamiltonian for the harmonic oscillator could be chosen as 1 H = (p + iωq)(p − iωq) 2 we could reorder our Hamiltonian (see below). No astronomer has ever observed a nonzero cosmological constant. In general relativity the absolute value of the energy density does matter. Indeed. You couldn’t even get to MIT without coming back where you started. (A) You can’t measure absolute energies. 10−13 cm.8 Our theory is eventually going to be applied to the strong interactions. That this constant is infinite even in a 2 box is because even in a box the field has an infinite number of degrees of freedom. eliminating long wavelengths. They don’t explain it in any course given at Harvard because nobody knows why it is zero. The zero point energy is ultraviolet divergent.] . would still have an infinity (δ (3) (0)) even if the momentum integral were cut off. so it’s stupid to ask what the zero point energy is. Einstein’s equations 1 Rµν − gµν R = −8πGTµν 2 couple directly to the energy density T00 . say by restricting it to |k| < K. Tµν → Tµν − λgµν is just a way of changing the cosmological constant. introducing a change in the vacuum energy density. the total energy is changed by an infinite amount. infinite systems. This is no big deal for two reasons. There can be excitations of modes with arbitrarily short wavelengths. but for some potentials you can’t do that. only energy differences. not zero. a term introduced by Einstein and repudiated by him 10 years later. Our system in infinite space has a zero-point energy that is also infrared divergent. and you have to choose your zero another place. We won’t talk about why the cosmological constant is zero in this course. like an infinite crystal. (B) This is just an ordering ambiguity. measurements on supernovae have indicated that Λ > 0. in a covariant way. Our Hamiltonian for the system in infinite space. In general. also have “infrared divergences. This even occurs in introductory physics. That part of the infinity is eliminated by putting the system in a box. With a cosmic energy density of 1039 GeV/cm3 . We usually put interaction energies to be zero when particles are infinitely separated. the universe would be about 1 km long according to Einstein’s equations. 8 [BGC note: As of 1998. October 7 Notes from Sidney Coleman’s Physics 253a 39 of our Hamiltonian starts at 1 k ωk .

5. Redefine H to be : H :. since the commutation relations are set up to reproduce the classical equations of motion. equation. In any given case you can check to see if ordering ambiguities or extra terms in commutators screw up the calculations. Later on we’ll have to think harder about what we’ve done wrong to get rid of troublesome infinities. 0) = q a (t) What’s wrong? a† a = aa† − 1 : a a : =: a a : −1 0 = −1 † † Normal order both sides Answer. This operation is only defined for free fields satisfying the K. No further specification is needed since all annihilation operators commute with one another as do all annihilation operators. If we wanted to get as quickly as possible to applications of QFT. Normal ordering is not derived from the ordinary product any more than the cross product is derived from the scalar product. But first we are going to get some more exact results from field theory. . a physically uninteresting question. Symmetries and Conservation Laws We will study the relationship between symmetries (or invariances) and conservation laws in the classical Lagrangian framework. . Hopefully everything will go through all right when Poisson brackets become commutators. We’ll encounter more ferocious ones. . about an unobservable quantity. we’d develop scattering theory and perturbation theory next.9 This is the first infinity encountered in this course. . λ) 9 q a (t. q a . which tells us we can write the field in terms of time independent creation and annihilation operators. October 7 Notes from Sidney Coleman’s Physics 253a 40 Reordering the Hamiltonian: Given a set of free fields (possibly with different masses) φ1 (x1 ). Our derivations will only use classical equations of motion. φn (xn ). t) and a transformation of the generalized coor˙ dinates q a (t) → q a (t. Normal ordering cannot accurately be used as a verb. We don’t “normal order” equations. We ran into it because we asked a dumb question. This gets rid of the normal ordering constant. .-G. define the normal ordered product : φ1 (x1 ) · · · φn (xn ) : as the usual product except rearranged so that all the annihilation operators on the right and a fortiori all creation operators on the left. Given some general Lagrangian L(q a .

. the transformed system is a time λ ahead q a (t) → q a (t + λ) Dq a = 3. Time translations for a general system. . a = 1. October 7 Notes from Sidney Coleman’s Physics 253a 41 we can define Dq a ≡ ∂q a ∂λ λ=0 This is useful because the little transformations will be important.) ˆ Most transformations are not symmetries (invariances). all particles in the system moved by eλ Dra = e 2. Dxa = y a . The Lagrangian we’ll use is n ma a 2 |r | − Vab (|ra − rb |) 2 a=1 ab The transformation is ra → ra + eλ. not necessarily satisfying the equations fo motion. n. Dz a = 0. Given some evolution q a (t) of the system. ˙ Definition: A transformation is a symmetry iff DL = dF for some F (q a . Space translation of point particles described by n vectors ra . Why is this a good definition? Hamilton’s principle is t2 t2 0 = δS = t1 dtDL = t1 dF = F (t2 ) − F (t1 ) dt Thus a symmetry transformation won’t affect the equations of motion (we only consider transformations that vanish at the endpoints). . t). . axis (If e = z . Dy a = −xa . q a .5. Some examples: 1. This equality dt a must hold for arbitrary q (t). Rotations in the system of example (1) Rotation matrix ∂q a ∂t ra → R ( λ e )ra Dra = e × ra angle. Back to our examples: .

. dP =0 . all of its time dependence comes from its dependence on q a and q a . we’ll call the conserved quantities the momentum. F = L ˙ ∂t dt 3. dt Whenever we get conserved quantities from spatial translation invariance. The proof comes from considering two expressions for DL.) We can write down the conserved quantities in our examples ˙ 1. dt by equality of mixed partials d Dq a = dt Dq a ˙ By assumption DL = quantity Subtracting these two expressions for DL we see that the pa Dq a − F a Q= satisfies dQ =0 dt (There is no guarantee that Q = 0. ∂L = 0. F = 0 Theorem (E. October 7 Notes from Sidney Coleman’s Physics 253a 42 1. F = 0 2. Noether say ‘Nota’): For every symmetry there is a conserved quantity. in fact the construction fails to produce a Q for gauge symmetries. whether or not the system looks anything like a collection of point particles. DL = a ∂L Dq a + pa Dq a ˙ ∂q a (pa Dq a + pa Dq a ) ˙ ˙ ∂L used pa ≡ ∂ qa ˙ = a used E-L equations = d dt pa Dq a a dF . If L has no explicit time dependence. then DL = dL . The momentum P = a ˙ ma ra is conserved. Q= a ˙ ma e · r a = e · a ˙ ma r a For each of three independent e’s we get a conservation law. F = 0. Dra = e. (pa = ma ra ). DL = 0.5. or that for each independent symmetry we’ll get another independent Q.

we’ll call them the angular momentum.10 Assume Dq b and F depend only on the q a . Dq a = ∂q a . we can usually reconstruct the symmetry. like the energy. for example. Q= a pa q a − L ˙ is conserved when ∂L =0 ∂t Whenever we get a conserved quantity from time translation invariance. had to be noticed from the equations of motion in each new theory. we’ll call the conserved quantity the energy. q a ] Dq b a −iδb = −iDq a This assumption is not at all necessary. Q= a pa · (e × ra ) = a e · (ra × pa ) =e· a ra × pa three laws ra × pa is conserved dJ =0 dt J= a Whenever we get a conserved quantities from rotational invariance. Show: [Qi . the quantum Q’s recreate the algebra of the generators. October 7 2. 3. not the pa . 10 . ∂t Notes from Sidney Coleman’s Physics 253a 43 F = L. There is nothing here that was not already in the Euler-Lagrange equations. From the conserved quantity. q a = b [pb .5. We’ll do it in quantum mechanics using commutators. What this theorem provides us with is a turn the crank method for obtaining conservation laws from a variety of theories. Before this theorem. For a general internal symmetry group. Qj ] = i ijk Qk ] when the Qi ’s are generators of an SO(3) internal symmetry. Dra = e × ra . why a variety of theories. Note: Q is identical to H. not the q a so that their expression in the ˙ a Hamiltonian formulation only depends on the q . It explains. Usually the result holds. or energy (example (2)). For example the energy generates time translations even though the assumption doesn’t hold. This can be done in classical mechanics using Poisson brackets. q a ] = b pb Dq b − F. the existence of conserved quantities. This theorem organizes conservation laws. including ones with velocity dependent potentials all have a conserved Hamiltonian. Then [Q. F = 0.

There is also a current  and there is a much stronger statement of charge conservation than dQ = 0. the electric charge. ∂µ φa . we could invoke special relativity to show this picture is inconsistent. What is this more? Electromagnetism possesses a conserved quantity Q. This picture satisfies global charge conservation. 0) = φa (x) . Given some Lagrangian density L(φa . x) and a transformation of the fields φa (x) → φa (x. and there would have to be a flow of current in between the two charges. Q = d3 xρ(x. The charge is the integral of the charge density p. t). λ) we define Dφa = ∂φa . Field theory which embodies the idea of local measurements. Even if there were not a current and a local conservation law. should have local conservation laws. and for a moment global charge conservation is violated. ∂λ λ=0 φa (x. You have to be able to account for the change in charge in any volume. October 7 Notes from Sidney Coleman’s Physics 253a 44 Symmetries and Conservation Laws in (Classical) Field Theory Field theory is a specialization of particle mechanics.5. but violates local charge conservation.t)) This equation says you can see the charge change in any volume by watching the current flowing out of the volume. There will be more that is true in field theory. You can’t have: t x simultaneously wink out of existence with nothing happening anywhere else. In another frame the charges don’t disappear simultaneously. Local charge conservation dt says ∂ρ + ·=0 ∂t Integrate this equation over any volume V with boundary S to get dQV =− dt dAˆ ·  n S using Gauss’s theorem and (Q= V d3 xρ(x.

In the particle mechanics case. There is an ambiguity in the definition of F µ and J µ since F µ can be changed by any χµ satisfying ∂µ χµ = 0. We are not using special relativity. Why is this a good definition? Hamilton’s principle is 0= = δS = d4 xDL = d4 x∂µ F µ blithe as usual about contributions from spatial infinity d3 x[F 0 (x. This equality must hold for arbitrary φa (x) not necessarily satisfying the equations of motion. ∂µ φa . but only by a . The total amount of stuff. t1 )] Thus a symmetry transformation does not affect the equations of motion (we only consider transformations that vanish at the endpoints when deriving the equations of motion. To wit. October 7 Notes from Sidney Coleman’s Physics 253a 45 Definition: A transformation is a symmetry iff DL = ∂µ F µ for some F µ (φa . Note that the previous definition can be obtained. t2 ) − F 0 (x. DL = D d3 xL = d3 x∂µ F µ = d3 x∂0 F 0 = d F dT where F = d3 xF 0 . Subtracting these two expressions for DL we see that the four quantities µ Jµ = πa Dφa satisfy ∂µ J µ = 0 a which implies dQV = − S dAˆ · J . F µ is not necessarily a 4-vector (just some set of 4 objects).5. n dt This equation justifies calling J 0 the density of stuff and J i the current of stuff. The proof comes from considering two expressions for DL. F was ambiguous. DL = a ∂L µ Dφa + πa D∂µ φa ∂φa µ µ (πa Dφa + πa D∂µ φa ) µ ∂L used πa ≡ ∂∂ φa µ = a used E-L equations = ∂µ a µ πa Dφa by equality of mixed partials D∂µ φa =∂µ Dφa By assumption DL = ∂µ F µ . x).). Theorem (maybe this is Noether’s theorem?): For every symmetry there is a conserved current. S =boundary of V . Q is independent of time.

e. As a result of this change J µ → J µ − ∂ν Aµν .e. for an arbitrary function of space time α(x) parametrizing a transformation. Instead of arguing that the most convenient J µ is the “right” one. Q= d3 xJ 0 → Q − d3 x∂i Ai = Q i. 490 of them are idiotic. By assumption of a symmetry t dF 2 δS = dtDL = dt dt t1 Subtracting we see a pa Dq a − F is time independent. [Aside: Noether’s Theorem derived at the level11 of the action If without using the equations of motion. This is a lot of freedom in the definition of J µ . October 7 Notes from Sidney Coleman’s Physics 253a 46 time independent quantity. you should just be happy that you had some freedom of choice. δS = δ dtL = a pa Dq a |t2 t1 for an arbitrary variation about a solution of the equations of motion. For example in a theory with three fields we can let J µ → J µ − (φ3 ) 4 ∂ν (∂ µ φ1 ∂ ν φ2 − ∂ ν ∂φ1 ∂ ν φ2 ) This J µ is as good as any other. Q is unchanged. for some reason. There are 500 papers arguing about which energymomentum tensor is the right one to use inside a dielectric medium. ignoring contributions from spatial infinity as usual.5. For arbitrary antisymmetric A we can let F µ → F µ + ∂ν Aµν (∂µ ∂ν Aµν = 0). From cranking Hamilton’s principle. They’re all edible! Sometimes one J µ is more useful in a given calculation than another. It’s like if someone passes you a plate of cookies and you start arguing about which copy is #1 and which is #2. to first order in α(x) δL = α(x)∂µ k µ + ∂µ α(x)j µ (Equivalent to DL = ∂µ k µ when α is a constant). 11 i. we can give another derivation of the relation between symmetries and conserved quantities. using Hamilton’s principle rather than mucking around with the equations of motion .

for fields satisfying the equations of motion. j µν = −π µ ∂ ν φ so T µν = π µ ∂ ν φ − g µν L is conserved.] Noether’s Theorem: when the variation is a hermitian traceless matrix. (useful in theories with fields most easily written as matrices. except there is no conservation law associated with the trace. write this statement as ∂µ F µ − 1 Id Tr F µ N =0 dimension of F µ . d4 x(α(x)∂µ k µ − α(x)∂µ j µ ) So it must be that J µ = k µ − j µ is conserved. Example: Space-time translation invariance δφ = − ν ∂ ν φ ∂L ∂L δL = δφ + δ∂τ φ ∂φ ∂∂τ φ ∂L ∂L ∂L ν ν =− ∂τ ν ∂ ν φ ν∂ φ − ν ∂ ∂τ φ − ∂φ ∂∂τ φ ∂∂τ φ = − ν L ν − π µ ∂ ν φ∂µ ν δ∂τ φ = −∂τ ( ν ∂ ν φ) ν is a parameter for the type of transformation. hermitian. 0= = d4 xδL = d4 x(α(x)∂µ k µ + ∂µ α(x)j µ ) Let α(x) vanish at ∞ to do parts integration. If you like. traceless but otherwise arbitrary d4 xTr δω∂µ F µ Then F µ is a matrix of conserved currents.5.) matrix Suppose (without using the equations of motion) δS = Then Hamilton’s principle says 0= d4 x Tr δω ∂µ F µ . October 7 Notes from Sidney Coleman’s Physics 253a 47 Then by Hamilton’s principle. µ plays the same role I read off k µν = −g µν L.

5. ∂L = 0. an arbitrary field. October 7 Notes from Sidney Coleman’s Physics 253a 48 Space-time translations in a general field theory We won’t have to restrict ourselves to scalar fields because under space-time translations. for example gravity and other theories of higher spin. one for each of the four independent directions we can point e. so you don’t have to remember which index is which. transforms the same way. tensor. ∂x φa (x) → φa (x + λe) Dφa = eν ∂ν φa F µ = eµ L T µν e some fixed four vector DL = eµ ∂µ L = ∂µ (eµ L) Jµ = a µ πa eν ∂ν φa − eµ L = eν T µν = a µ πa ∂ ν φa −g L µν There are four conserved currents. T iµ are the three currents from spatial translation invariance. vector. Pi = d3 xT 0i T 0i = the density of the ith component of momentum T ji = the jth component of the current of the ith component of momentum µ For a scalar field theory with no derivative interactions. etc. Just let the index a also denote vector or tensor indices. ∂µ T µν = 0 since ∂µ J µ = 0 for arbitrary e. T µν can be nonsymmetric. It depends on x only through its dependence on φa and ∂µ φa .. i. πa = ∂ µ φ so T µν = a ∂ µ φa ∂ ν φa − g µν L Note that T µν is symmetric. four local conservation laws. . T µ0 is the current that is conserved as a result of time translation invariance. Translations are symmetries as long as L does not have any explicit dependence on x. which can lead to problems. H= d3 xH T 00 is the energy density. T i0 is the current of energy. Can try to symmetrize it. indeed.e. H = T 00 = a ˙ πa φa − L.

fields. µ ν µ ν ν relabel dummy indices ν a bµ + a µ b ⇓ in the second term. ν µ ν µ µ→ν. For example a vector field Aµ (x) → Λµ Aν (Λ−1 x). because there are extra factors in the transformation law when the fields are tensorial. we only consider scalars. so there are 4·(4−1) = 6 2 independent . that is if bµ → Λµ bν then aµ bµ → aµ bµ ν This must be true for arbitrary a and b. µν a b + νµ a b ( µν + νµ )aν bµ ⇒ µν = − νµ since this has to hold for arbitrary a and b. or vector. We’ll define ∂λ DΛµ ≡ ν µ ν defines some matrix µ ν From the invariance of aµ bν . In any case. With Lorentz transformations.6. Λµ must preserve the Minkowski space inner ν µ product. (The equation this condition gives is gµν Λµ Λν = α β gαβ . which is good since we have to generate 3 rotations (about each axis) and 3 boosts (in each direction). λ) = φa (Λ(λ)−1 x) We are restricting ourselves to scalar fields. October 9 Notes from Sidney Coleman’s Physics 253a 49 6 October 9 Lorentz transformations Under a Lorentz transformation all vectors transform as aµ → Λ µ aµ ν where Λµ specifies the Lorentz transformation.ν→µ. µ and ν range from 0 to 3. φa (x) → φa (x. . ν We need to get Dφ = ∂φ λ=0 . the derivation of the conservation of T µν from space-time translation invariance applies to tensor. Even though we only used scalar fields in our examples of T µν . This could be rotations about some specified axis by an angle λ or a boost in some specified direction by a rapidity λ. the Lorentz transformation is given by a family aµ → aµ (λ) = Λ(λ)µ aν ν Under this (active) transformation (we are not thinking of this as a passive change of coordinates) the fields φa transform as (φa is a scalar). we’ll derive a condition on 0 = D(aµ bν ) = (Daµ )bν + aµ (Dbν ) = = = µ ν.) We’ll be interested in one parameter subgroups of the group of Lorentz transformations parametrized by λ.

October 9 Notes from Sidney Coleman’s Physics 253a 50 As a second confidence-boosting check we’ll do two examples. Da1 = a0 is just the infinitesimal version of a0 → cosh λa0 + sinh λa1 a1 → sinh λa0 + cosh λa1 Without even thinking. Now assuming L is a scalar. Take 12 = − 21 = 1. all other components zero. the time. Da0 = Da1 = 0 1 1a 1 0 0a =− =− = a1 0 0 10 a = a 01 a 1 This says x1 . we are all set to get the 6 conserved currents. all other components zero.6. Take 01 = − 10 = +1. which could be the first component of the position of a particle. which is definitely what a boost in the x1 direction does. This is a rotation. . Da0 = a1 . gets a little contribution proportional to x0 . Da1 = Da2 = 1 2 2a 2 2 1a =− =− = −a2 1 1 21 a = +a 12 a 2 x2 a infinitesimally shifted a x1 This says a1 gets a little negative component proportional to a2 and a2 gets a little component proportional to a1 . the great index raising and lowering machine has given us all the right signs. In fact. in the standard sense about the positive z axis.

J 3 is (J i = 1 ijk J jk ). often called. October 9 Notes from Sidney Coleman’s Physics 253a µ ν 51 From Λ−1 (λ)Λ(λ) = 1. we would have additional terms. so the part of the quantity in parentheses that is antisymmetric in λ and σ must be conserved i. 0 = D(Λ−1 Λ) and D Λ−1µ |Λ=1 = − ν ˙ Dφ(x) = There are extra ∂ a −1 µ ν terms in Dφa if φ (Λ (λ)ν x ) φa are not scalars ∂λ λ=0 = ∂σ φa (x)D(Λ−1 (λ)σ xτ ) τ a σ τ = ∂σ φ (x)(− τ )x = − στ xτ ∂ σ φa (x) Using the assumption that L is a scalar depending only on x through its dependence on φ and ∂µ φa we have a DL = The conserved current J µ is Jµ = a µ πa λσ x λ σ ∂ L λ µσ = ∂µ [ λσ x g L] λσ x λ σ a ∂ φ − λσ x λ µσ g L = λσ a µ πa xλ ∂ σ φa − xλ g µσ L This current must be conserved for all six independent antisymmetric matrices λσ . feeding in from the extra terms in Dφa .6. The 6 conserved charges are J λσ = d3 xM 0λσ = d3 x(xλ T 0σ − xσ T 0λ ) For example. λµ M µλσ = 0 where M µλσ = a µ πa xλ ∂ σ φa − xλ g µσ L − (λ ↔ σ) − (λ ↔ σ) = xλ = xλ T a µσ µ πa ∂ σ φa − g µσ L − xσ T µλ If the φa were not scalars. J 12 . 2 J 3 = J 12 = d3 x(x1 T 02 − x2 T 01 ) . the conserved quantity coming from invariance under rotations about the 3 axis.e.

What are they? Consider J 0i = d3 x[x0 T 0i − xi T 00 ] This has an explicit reference to x0 .6. T 00 is the relativistic generalization of mass. and that will be reflected in extra terms in the J ij . You aren’t used to calling this a conservation law. The conservation law is dJ = 0 so we have dt 0= d d 0i J = t d3 xT 0i − d3 xxi T 00 dt dt ¨ d d 3 d¨¨ 0i d3 xT 0i − d3 xxi T 00 =t ¨ xT + dt ¨ dt pi pi Dividing through by p0 gives pi 1 = constant = o = p d dt d3 xxi T 00 = p0 d (“center dt of energy”i ) total energy This says that the “center of energy” moves steadily. will be described by fields of tensorial character. but it is. October 9 Notes from Sidney Coleman’s Physics 253a 52 If we had point particles with T 0i (x. So far we have just found the continuum field theory generalization of three conserved quantities we learn about in freshman physics. But we have three other conserved quantities. the J 0i . t) = a pi δ (3) (x − ra (t)) a J 3 would be (xa1 p2 − xa2 p1 ) = a a a a (ra × pa )3 We have found the field theory analog of the angular momentum. the time. . This is the relativistic generalization of the statement that the center of mass moves steadily. Particles that have intrinsic angular momentum. but there is nothing a priori wrong with that. spin. We can pull the x0 out of the 0i integral over space. We only have the orbital contribution. The particles themselves could have some intrinsic angular momentum. Those contributions to the angular momentum are not in the J i . and in fact it is the Lorentz partner of the angular momentum conservation law. something we’ve not seen in a conservation law before.

because the existence of a conservation law is a qualitative fact that greatly restricts the form of the Lagrange density. Clockwise rotation makes the signs come out conventionally if b and c (which we see a little bit later) are defined like a and b in Itzykson and Zuber p. These additional conservation laws will only occur in specific theories whose Lagrange densities have special properties. October 9 Notes from Sidney Coleman’s Physics 253a 53 Internal Symmetries There are other conservation laws. The Lagrangian is invariant because it only depends on (φ1 )2 + (φ2 )2 and that combination is unchanged by rotations. Dφ2 = −φ1 µ µ J µ = π1 Dφ1 + π2 Dφ2 = (∂ µ φ1 )φ2 − (∂ µ φ2 )φ1 Q= 12 13 DL = 0 Fµ = 0 d3 xj 0 = d3 x(∂0 φ1 φ2 − ∂0 φ2 φ1 ) We have left particle mechanics behind and I’ll often use Lagrangian to mean Lagrange density. The first example is 1 L= 2 2 2 ∂µ φ ∂ φ − µ φ φ a=1 a µ a 2 a a −g a (φ ) a 2 This is a special case of a theory of two scalar fields. They will not relate fields at different spacetime points. the name comes from the idea that what an internal symmetry did was transform internal characteristics of a particle. Dφ1 = φ2 . . and the potential only depends on the combination (φ1 )2 + (φ2 )2 . 121. conservation of baryon number. This is a rotation13 in the φ1 . Internal symmetries are non-geometrical symmetries. φ2 plane. We have already found all the conservation laws that are in a general Lorentz invariant theory. We’ll study internal symmetries with two examples. This Lagrangian12 is invariant (DL = 0) under the transformation φ1 → φ1 cos λ + φ2 sin λ SO(2) symmetry φ2 → φ1 sin λ + φ2 cos λ The same transformation at every space time point. and we expect the symmetries they come from to commute with Lorentz transformations.6. Historically. The transformations will turn fields at the same spacetime point into one another. Both fields have the same mass. All these additional charges are scalars. and conservation of lepton number that we have not found yet. like conservation of electric charge. Conservation laws are the best guide for looking for theories that actually describe the world. For us internal just means non-geometrical.

6. If they didn’t Q wouldn’t be time independent since they are multiplied by e±2iωk t . as is easily checked. If one state is degenerate with another. we need the other combination to reconstruct a1 . Define a1 + ia2 bk = k √ k 2 b† k a1† − ia2† = k √ k 2 That’s not the end of the definitions. October 9 Notes from Sidney Coleman’s Physics 253a 54 We can get some insight into this quantity by going to the case g = 0 in which case φ1 and φ2 are both free fields and can be expanded in terms of creation and annihilation ops. a1 − ia2 ck = k √ k 2 c† k 1† 2† ak + iak √ = 2 These linear combinations of operators are allowable 14 . |k. 14 . They are operators that create (or destroy) a superposition of states with particle 1 and particle 2. Let’s have faith in our formalism and assume that the terms with two creation ops or two annihilation ops go away. ak ] = 0 k Q=i 1† 2† d3 k[ak a2 − ak a1 k k We are within reach of something intuitive. φa (x) = d3 k (2π)3/2 2ωk (aa e−ik·x + aa† eik·x ) k k Now we’ll compute Q. 2 . Q= d3 k 1 2† [ak ak (−iωk − iωk ) + 2iωk a2 a1† ] k k 2ωk 2† We don’t have to worry about the order because a1 . k 1† 2† 2 ak . ak . i(a1† a2 k k − a2† a1 ) k k b† + c † bk − c k b† − c † bk + c k k k =i √ · √ + i k√ k · √ 2 2 2 2 1 † = [2bk bk + 0b† ck + 0c† bk − 2c† ck ] = b† bk − c† ck k k k k k 2 Exchanging the role of bk and ck is the way of making the signs come out conventionally with counterclockwise rotation. 1 is degenerate with |k. Because b† and c† create orthogonal k k states. bk and c† commute with each other. ak . and for a given k. The useful thing about these k linear combinations is that Q has a simple expression. it is often convenient to work with linear combinations of these states as basis states. So define.

b and c type mesons are eigenstates of Q. φa ] = −iDφa 2 [Q. It’s like particles and antiparticles. ψ] = −ψ a Q “eigenfield” There is also the hermitian conjugate of ψ. the c’s carry Q charge -1. ψ† = φ1 − iφ2 √ 2 d3 k [ck e−ik·x + b† eik·x ] = 3/2 k (2π) 2ωk (ψ † will be denoted ψ ∗ in the classical limit. [Q. A quantum field ψ ∗ will be understood to be ψ † . φ1 ] = √ [Q. ψ= φ1 + iφ2 √ 2 d3 k = [b e−ik·x + c† eik·x ] k (2π)3/2 2ωk k ψ always diminishes Q by 1 either by annihilating a b type particle or by creating a c type particle. By the way H and P have familiar forms in terms of the b’s and c’s. φ1 and φ2 have messy commutation relations with Q. ψ † ] = +ψ † . 1 1 [Q. October 9 Notes from Sidney Coleman’s Physics 253a 55 That is. φ ] = iφ1 . An eigenstate of Na and Nb is an eigenstate of Q. Q= d3 k(b† bk − c† ) = Nb − Nc k k The b’s carry Q charge +1. 1 and 2 type mesons are not.) ψ † always increases Q by 1. Q = Nb − Nc is true as an operator equation.6. ψ and ψ † have neat commutation relations with Q. ψ † . H= d3 kωk (b† bk + c† ck ) k k P = d3 k k(b† bk + c† ck ) k k Taking all these linear combinations suggests that we could have changed bases earlier in the calculation and simplified things. an eigenstate of N1 and N2 is not. [Q. ψ + ψ † ] = √ (−ψ + ψ † ) 2 2 2 = −iφ In agreement with a formula which is usually true [Q. the b’s and c’s have the same mass and opposite charge.

is going to canonically quantize this theory. t)] = iδ (3) (x − y) = [ψ † (x. ∂0 ψ ∗ (y. ψ and ψ are as nice as φ1 and φ2 . ψ(y. we’ll get back to symmetries. t). t). How can you work with complex fields. ψ † (y. t)] By God. October 9 Notes from Sidney Coleman’s Physics 253a 56 Under our transformation ψ → e−iλ ψ. t). Here he goes µ πψ = ∂L = ∂ µψ∗ ∂∂µ ψ µ πψ ∗ = ∂ µ ψ The Euler-Lagrange equations he gets are µ ∂µ πψ = ∂L ∂ψ i. t). which can’t be varied independently. treat them as if they can be and nevertheless get the right answers? We’ll demonstrate that this works for the equations of motion. Back to the classical Lagrangian which is L = ∂ µ ψ ∗ ∂ µ ψ − µ2 ψ ∗ ψ 1 (no 2 ) Imagine a person who once knew a lot of quantum field theory. ψ(y. t)] = [ψ(x. Suppose I have some action S(ψ. ψ(y. t)] = 0 = [ψ † (x. t). but has suffered brain damage. t). This is called a U (1) or phase transformation. You can demonstrate that it works for the equal time commutation relations. ψ ∗ = −µ2 ψ ∗ The idiot got it right. It is equivalent to SO(2). The equations of motion come from 0 = δS = d4 x(Aδψ + A∗ δψ ∗ ) . he gets that right. too. Digression. t)] These equations can be obtained by doing something completely idiotic. They obey ( + µ2 )ψ = 0 ( + µ2 )ψ † = 0 They obey simple equal time commutation relations: [ψ(x. t)] ˙ ˙ [ψ(x. t)] = 0 = [ψ † (x.6. t). t)] ˙ ˙ [ψ(x. He has forgotten that ∗ stands for complex conjugate. t)] = [ψ(x. t). ψ † (y. It works very generally. ψ † (y. ψ † (y. t).e. How about that!? Now he says “I’m going to deduce the canonical commutation relations”: 0 iδ (3) (x − y) = [ψ(x. ψ ∗ ). and he is going to treat ψ and ψ ∗ as if they were independent fields. † As quantum fields. πψ (y. He also gets ψ = −µ2 ψ. ψ ∗ → eiλ ψ ∗ .

b] Jµ = ∂µ φa φb − ∂µ φb φa You can’t find combinations of the fields that have simple commutation relations with all the Q[a. so we can read off the current. The notations are. take 1 L= 2 n n 2 (∂µ φ ∂ φ − µ φ φ ) − g a=1 a=1 a µ a 2 a a (φ ) a 2 This is the same as our first example except we have n fields instead of 2.b] ’s. . 2 so there are n(n−1) conserved currents and associated charges. which is a symmetry generated by I1 . If they did they would generate symmetries that commute. the symmetry is just like the one we had in the first example. Anyway. I3 . φa → a Rb φb b n×n rotation matrix There are n(n−1) independent planes in n dimensions and we can rotate in each of them.) We don’t expect the various charges to commute. Just as in the first example the Lagrangian was invariant under rotations mixing up φ1 and φ2 . . You can only choose the particle states π + . I2 and I3 . This is just like isospin. This example is quite differ2 ent from the first one because the various rotations don’t in general commute. October 9 Notes from Sidney Coleman’s Physics 253a 57 Naive Approach: Treating the variations in δψ and δψ ∗ as independent we get A=0 and A∗ = 0 Sharp Approach: However δψ is not independent of ψ ∗ .6. . We can make purely real variations in ψ so that δψ = δψ ∗ and we can make purely imaginary variations in ψ so that δψ = −δψ ∗ . Back to internal symmetries. . We can use the fact that ψ and ψ ∗ are allowed to be complex. Nothing tricky or slick. ψ ∗ = ψr√2 i . +iψ −iψ For the ETCR write ψ = ψr√2 i . this Lagrangian is invariant under rotations mixing up φ1 . because it only depends on (φ1 )2 + · · · + (φn )2 . From the real variation we deduce A + A∗ = 0 and from the imaginary variation we deduce A − A∗ = 0 which implies A = A∗ = 0. For our second example. [a. . (They all commuted in the first one by virtue of the fact that there was only one. say Q[1. π 0 and π − to be eigenstates of one of them. φn . For n = 3 you can choose the fields to have a simple commutation relation with one charge. for any single rotation axis.2] .

7. and M µνλ . but do they have the right Lorentz transformation properties? We’ll prove that P ν = d3 xT 0ν is indeed a Lorentz vector given that T µν is a two index tensor and that T µν is conserved. October 14 Notes from Sidney Coleman’s Physics 253a 58 7 October 14 Lorentz transformation properties of conserved quantities We’ve worked with three currents. Then we can write P ν in a way that makes its Lorentz transformation properties clearer. We integrated the zeroth component of each of these currents to obtain Q. P ν and J νλ respectively. 0. 0. the current of meson number. the current of the [νλ] component of angular momentum. We’ll choose the latter because we need the practice. 0) is a Lorentz vector. These look like tensors with one less index. a unit vector pointing in the time direction. The generalization to currents with more indices or 1 index will be clear. J µ . Introduce n = (1. 0). ∂µ T µν = 0. T µν . Pν = d4 xnµ T µν δ(n · x) Perform a Lorentz transformation on P ν U (Λ)† P ν U (Λ) = = d4 xnµ U (Λ)† T µν (x)U (Λ)δ(n · x) d4 xnµ Λµ Λν T στ (Λ−1 x)δ(n · x) σ τ Change integration variables x = Λ−1 x and define n = Λ−1 n U (Λ)† P ν U (Λ) = = = Λν τ d4 x Λµρ n ρ Λµ Λν T στ (x )δ(n · x ) σ τ d4 x n ρ Λν T ρτ (x )δ(n · x ) τ d4 xnρ T ρτ (x)δ(n · x) Using Λµρ Λµ = gρσ σ . the current of the νth component of momentum. The assumption that T µν is an operator in quantum field theory that transforms as a two-index tensor is phrased mathematically as follows: Given U (Λ) the unitary operator that effects Lorentz transformations in the theory U (Λ)† T µν (x)U (Λ) = Λµ Λν T στ (Λ−1 x) σ τ Now we’ll try to show that Pν = d3 xT 0ν (x. We could do this proof in the classical theory or the quantum theory.

Our active transformation has had the exact same effect as if we had made a passive transformation changing coordinates to x = Λ−1 x. P ν is not even independent of what time you compute at. It’s the same old story: alias (another name) versus alibi (another place). so that if I call the Lorentz transform of P ν . let alone changing the tilt of the surface. passive active t n' n t' x' x You can think of this as Lorentz transforming the field or Lorentz transforming the surface. that the current that flows through the surface t = 0 is the same as the current that flows through the surface t = 0. The only difference between what we have and what we would like to get U (Λ)† P ν U (Λ) = Λν P τ = Λν τ τ d4 xnρ T ρτ δ(n · x) is that n has been redefined. In the far future or the far past θ(n · x) = θ(n · x) so there are no surface terms there. October 14 Notes from Sidney Coleman’s Physics 253a 59 In the last expression we’ve dropped the prime on the variable of integration.7. . and as usual we won’t worry about surface terms at spatial infinity. For an arbitrary tensor. More or less. This doesn’t produce a vector P ν for any tensor T µν . and we take the component nµ T µν in the t direction. what we are trying to show is −1ν 0 = P ν − Λσ P ν = = d4 x[∂µ θ(n · x) − ∂µ θ(n · x)]T µν (x) d4 x∂µ [θ(n · x) − θ(n · x)]T µν (x) This integral over all spacetime is a total divergence. The surface of integration is now t = 0. P ν . We need to use that T µν is conserved. Note that nµ δ(n · x) = ∂µ θ(n · x).

This is a symmetry for any Lagrangian with only even powers of φ. Examples of internal symmetries (there is not a general theory. There is no conserved quantity associated with a discrete symmetry however. satisfying U (θ)† HU (θ) = H You can differentiate this with respect to the parameter to find that [I. 103ff . the inner product. we expect that there is a unitary operator effecting the transformation U † q a (t)U = q a (t) U † HU = H The discrete symmetry could be an element of a continuous symmetry group. (the canonical commutation relations. 15 See Dirac Principles of QM. October 14 Notes from Sidney Coleman’s Physics 253a 60 Discrete Symmetries A discrete symmetry is a transformation q a (t) → q a (t) that leaves the Lagrangian. (It does not include time reversal. but we’ll do prototypical examples) Example (1). wait15 . in particular. this could be rotation by π about the z axis. θ. we will actually be able to construct U . and that you have a unitary operator for each value of the parameter. It could be rotation by 20◦ . that is give its action on the creation and annihilation operators.7. 1 1 L = (∂µ φ)2 − µ2 φ2 − λφ4 → L 2 2 There should be a unitary operator effecting the transformation φ → U † φU = −φ For λ = 0. H] = 0 I ≡ −i dU dθ ≡ −iDU θ=0 There is nothing analogous for discrete symmetries. the Hamiltonian all come from L) and since the Lagrangian is unchanged. The transformation is φ(x) → −φ(x) at every space-time point. pp. What is special about the continuous symmetry is that there is a parameter. and thus its action on the basis states. unchanged L → L.) Since all the properties of the theory are derived from the Lagrangian. This could be parity. L.

Mesons are always produced in pairs. In fact it has an O(2) symmetry.c. whenever n + m is odd. It is invariant under proper and improper rotations. Recall that L= a=1 2 1 1 (∂µ φa )2 − µ2 (φa )2 − λ[(φ1 )2 + (φ2 )2 ]2 2 2 had an SO(2) symmetry. . We’ll take one standard improper rotation φ1 → φ1 φ2 → −φ2 . For if n + m is odd n|S|m = n|U † U SU † U |m = n|U † (U SU † ) U |m S because U HU † =H = (−1) n+m n|S|m = − n|S|m Our second example of a discrete internal symmetry will turn out to be Charge Conjugation.7. |m is a state with m mesons and S is the scattering matrix made from the Hamiltonian. The existence of this unitary operator tells you that you’ll never see 2 mesons scatter into 43 mesons or any odd number of mesons. . it must be that 0 = n|S|m where n is a state with n mesons. equation U † a† U = −a† k k We’ll also make a choice in the phase of U by specifying U |0 = |0 We can determine the action of U on the basis states U |k1 . rotations and rotations with reflections in the φ1 . . More formally. kn As an operator statement we see that U = (−1)N (= eiπN if you prefer ) (N =meson number = d3 ka† ak ) k Suppose we could construct this operator when λ = 0. . φ2 plane. . . kn = U a† a† · · · a† |0 k k k 1 2 n = U a† U † U a† U † · · · U a† U † U |0 k k k 1 2 n = (−1)n a† a† k1 k2 · · · a† |0 kn = (−1)n |k1 . October 14 Notes from Sidney Coleman’s Physics 253a 61 Since φ is a linear function of ak and a† it must be that k U † ak U = −ak and the h. . .

It is easy to write down the unitary operator in the case λ = 0. this is sometimes called a conjugation symmetry. Parity Transformations Any transformation that takes φa (x.7. 16 the charge from the SO(2) symmetry .) A parity transformation transforms each fundamental observable at the point x into some linear combination of fundamental observables at the point −x. t) → b Mba φb (−x. You could write down a Lagrangian with four fields that is invariant under rotation in fourdimensional space and under permutations of any of the four fields. (We have used the fact that we live in an odd number of dimensions (3) and thus that x → −x is an improper rotation. Because UC QUC = −Q it is also called charge16 conjugation or particle-anti-particle conjugation. From the action on ψ and ψ † we see † † UC bk UC = ck and UC ck UC = bk and the equations obtained from these by hermitian conjugation. You could write down a theory with the icosahedral group. we have reduced it to the previous case. Then we have two independent free scalar field theories UC = (−1)N2 As mathematicians are fond of saying. t) we’ll call a parity transformation. but it is boring. in our definition. The action of U is especially nice if we put it in terms of the fields ψ and ψ † already introduced ψ= φ1 + iφ2 φ1 − iφ2 † √ → UC ψUC = √ = ψ† 2 2 † ψ →ψ † For this reason. If we lived in two space dimensions we could do a similar thing with only x2 → −x2 . UC is unitary and hermitian: † † 2 UC = 1 = UC UC ⇒ UC = UC We could continue the discussion of discrete internal symmetries. October 14 Notes from Sidney Coleman’s Physics 253a 62 Any other one can be obtained by composing this one with an element of the internal symmetry group SO(2).

t) L→L (M = −1) Whenever there is an internal symmetry in a theory I can multiply one definition of parity by an element of that symmetry group (discrete or continuous) and get another definition of parity. t). t) we can see that the action on the creation and annihilation operators must be † UP ak Up = a† k a−k a† k − and on the basis states UP |k1 . t) and a theory with invariance under P : φ(x. . −kn Sometimes people distinguish between a theory with invariance under P : φ(x. Our theory is invariant under both. when it is a symmetry. . . . −kn But there is a second possibility for parity P : φ(x. t) → φ(−x. . t) → φ(−x. but all we really demand is that parity takes L(t) → L(t) as usual. t)UP = φ(−x. t) → L(−x. . t) → −φ(−x. October 14 Notes from Sidney Coleman’s Physics 253a The usual confusing notation 63 Usually parity takes L(x. . by calling the first the theory of a scalar meson and the second a theory of a pseudo scalar meson. kn = (−1)n | − k1 . . t) → −φ(−x. . it’s the old plate of cookies . Example (1) Parity is a symmetry of µ2 1 L = (∂µ φ)2 − φ2 2 2 From φ(x. . . kn = | − k1 . In the case at hand the unitary operator UP is given by UP = (−1)N UP or UP |k1 . t) (M = 1) L→L † UP φ(x. . t).7. t) = and d3 k (2π)3/2 2ωk [aa e−ik·x e−iωk t + aa† eik·x eiωk t ] k k P : φ(x. . . . .

three of the derivatives are on space coordinates. Example (4) Sometimes people say that because the product of two reflections is 1. Example (3) Take the perverse Lagrangian of the last example by and make it worse by adding 4 (φa )3 a=1 Now there is no definition of parity that gives a symmetry. The new interaction involving the totally antisymmetric tensor in four indices is invariant under L. We get three minus signs under parity. for the same reason that a·(b×c) is invariant under proper rotations. (a·(b×c) is multiplied by det R under a rotation). This is why the conventions on the parity of some particles is arbitrary (relative parity). Because any φa → −φa is an internal symmetry of the free Lagrangian. All the physicists in the world would agree that this is a scalar meson. P isn’t. An odd number of mesons are going to have to be pseudoscalar to get a net even number of minus signs. After this lecture you won’t see anything this bad again.7. This theory violates parity. As long as you are in agreement about the total set of invariances of the theory. you shouldn’t waste time arguing about what you’ll call each one. Indeed UP cannot be chosen to satisfy UP = 1. October 14 Notes from Sidney Coleman’s Physics 253a 64 problem again. One of the cookies has been poisoned. 4 L= a=1 µ2 1 (∂µ φa )2 − a (φa )2 − g 2 2 µνλσ ∂µ φ1 ∂ν φ2 ∂λ φ3 ∂σ φ4 All four meson masses are different. it doesn’t matter which one you choose or which three you choose to be pseudoscalar. If −λφ4 is added to L both P and P are still invariances of L. Example (2) This awful Lagrangian has been cooked up to illustrate a point. 4 L= a=1 µ2 1 (∂µ φa )2 − a (φa )2 + ∂µ ψ ∗ ∂ µ ψ − m2 ψ ∗ ψ 2 2 4 −h a=1 (φa )3 − g µνλσ ∂ µ 1 ν 2 λ 3 σ 4 φ ∂ φ ∂ φ ∂ φ [ψ 2 + ψ ∗2 ] . the other three spacelike. The theory has a set of invariances. This example is cooked up to show that a theory with 2 2 parity can have indeed UP = 1. Because µνλσ is nonzero only if one of its indices is timelike. the square of parity is 1.T. With a φ3 interaction P is a symmetry.

t) → +φa (x. q(t)]UT = −[p(−t). October 14 Notes from Sidney Coleman’s Physics 253a 65 The transformation of the φa ’s must be φa (x. and if parity were a symmetry (or an approximate symmetry) of the world we would have a name for fields transforming like ψ. q(−t)] = −i Particularly poignant at t = 0 Looks like we would have to give up the canonical commutation relations to implement time reversal. Because p(t) ∝ q(t) ˙ † UT p(t)UT = −p(t) † Consider UT [p. † UT q(t)UT = q(−t) There are two paradoxes I’ll give to show this can’t happen. t).7. From the two relations we have juts obtained we also have † UT [p(t). You might still hope there is a unitary operator that does the job in the quantum theory. Time Reversal First a famous example from classical particle mechanics. q]UT = −i. If it did. The only way to make the last term parity invariant is for ψ ot transform as ψ → ±iψ ψ∗ → iψ ∗ † † 2 In either case UP = 1. UP UP ψUP UP = −ψ. a particle moving in a potential 1 L = mq 2 − V (q) ˙ 2 T : q(t) → q(−t) T is not a discrete symmetry the way we have defined it T : L(t) → L(−t) in the usual confusing notation where L(t) refers to the time dependence through the coordinates Nevertheless T does take one solution of the equations of motion into another. a “semi-pseudo-scalar”. † 1st Paradox Differentiate UT q(t)UT = q(−t) with respect to t. If that isn’t enough to make you abandon the idea of a unitary time reversal operator I’ll continue to the . Fortunately. nothing like this occurs in nature (as far as we know).

A† b) = (Aa. U . Dirac notation is designed to automate the handling of linear operators. For any operator O(t) O(t) = eiHt Oe−iHt † Apply UT the unitary transformation to both sides to obtain † † O(−t) = UT e+iHt UT OUT e−iHt UT O = O(0) let V † ≡UT e−iHt UT but O(−t) = e −iHt UT Oe iHt I’d like to show V = eiHt . b unitarily This is enough of an assumption to show U is linear (see related proof below). b denote states. B denote operators. so for a while we’ll use some more cumbersome notation that does not automate the handling of linear operators. β denote complex numbers and A. b) is the inner product of two states. i. The answer is that time reversal is implemented by an antiunitary operator. † & UT (& −iH)UT = ¡H. b) for all a. because (the spectrum of unitarily related operators is the same and) the spectrum of H is not bounded above. Antiunitary operators are antilinear. Let a. b . October 14 Notes from Sidney Coleman’s Physics 253a 66 † 2nd Paradox Roughly. V e−iHt = 1. b) for all a. (a. The spectrum of H cannot be bounded below. U (αa + βb) = αU a + βU b linearity The adjoint of a linear operator A is denoted A† and is the operator defined by (this definition is not consistent if A is not linear) (a. Now that I’ve proved † UT e−iHt UT = eiHt Take d dt t=0 of this relation. we see that H is unitarily related to −H.e. A unitary operator is an invertible operator. α. i canceling the i’s.7. U b) = (a. AUGGH! A unitary time reversal operator is an object that makes no sense whatsoever. satisfying (U a. UT should reverse time evolution. UT e−iHt UT = eiHt . V e−iHt commutes with any operator O. The simplest unitary operator is 1. I can prove this. What we have is e−iHt OeiHt = V −1 OV which implies V e−iHt O = OV e−iHt .

applying the antiunitarity condition to each term. a → U a. an upside down U ) satisfying (Ωa. Ω(αa + βb) − α∗ Ωa − β ∗ Ωb). For the elements of some basis. U U −1 b) = (U a. (Ωa.) The simplest antiunitary operator is complex conjugation. Ω. Ab) → (U a. The result we want! Indeed. and then expand the 5 terms containing αa + βb some more to show this is zero.7. An antiunitary operator is an invertible operator. A useful fact (especially conceptually) is that any antiunitary operator. (You ask why?!) This is the inner product of Ω(αa + βb) − α∗ Ωa − β ∗ Ωb with itself. Proof by construction: take U = ΩK. (this is a notational gem. a) ⇒ Ω(αa + βb) = a∗ Ωa + β ∗ Ωb antilinearity Consider (Ω(αa + βb) − α∗ Ωa − β ∗ Ωb. can also be thought of as a transformation of the operators in the theory (a. (The product of a unitary operator with an antiunitary operator U Ω is an antiunitary operator. October 14 Notes from Sidney Coleman’s Physics 253a 67 I’ll show that U † = U −1 (which is sometimes given as the definition of unitarity).) U U Ω Ω Ω U We can prove (this is the related proof referred to above) that any operator (not necessarily invertible) satisfying the antiunitarity condition is antilinear. is equal to U K for some unitary U . (The analogous proof for operators satisfying the unitarity condition also only uses properties of the inner product and is even easier. bj ) must be real. AU b) = (a. Ωb) = (b. a) for all a. U −1 b) = (U a. The basis is a complete set of real wave functions. b antiunitarity We can immediately make a little table showing the result of taking products of unitary and antiunitary operators. bi . U † AU b) can alternatively be thought of as A → U † AU . Kbi = bi and on any linear combination K( i αi bi ) = i ∗ αi bi For consistency (bi . . the fact that the inner product is positive definite implies that Ω(αa + βb) − α∗ Ωa − β ∗ Ωb = 0. Ω. unitarity of U (a. K. This is the familiar complex conjugation of nonrelativistic quantum mechanics of position space wave functions. If this is zero. Ωb) = (b. etc. b) A transformation of the states. it is simply a matter of expanding this inner product out into its 9 terms.

It is the probabilities that are measurable. Aa) → (Ωa. Time reversal is antiunitary. He showed that (up to phases) they are the only operators that preserve probabilities. That is ΩP T does nothing. Ω−1 (−iH)ΩT = iH ⇒ Ω−1 HΩT = H T T We can explicitly construct the time reversal operator in free field theory. Ω−1 iΩT = −i T You can’t cancel the i’s as we did in paradox 2. Eugene Wigner proved a beautiful theorem telling us why unitary and antiunitary operators are important in QM. absolutely nothing to ak and a† k Ω−1 ak ΩP T = ak PT and Ω−1 a† ΩP T = a† PT k k . 50 years ago [1931]. Ω−1 AΩa) This transformation can alternatively be thought of as A → Ω−1 AΩ We don’t write Ω† AΩ because adjoint is not even defined for antilinear ops. a → Ωa. Consider the expectation value of a Hermitian operator (observable) in the state a. the transformation of the states by an antiunitary operator Ω. Ωa) = (ΩΩ−1 AΩa. b)|2 then F (a) = eiφ(a) x a unitary op or an antiunitary op (hermiticity) invertibility antiunitarity (φ : H → R) Time reversal is not a unitary operator. Let’s find ΩP T such that Ω−1 φ(x)ΩP T = φ(−x) PT The simplest candidate is just complex conjugation. Look in the appendix of his book on group theory. It transforms as (a. in the momentum state basis. F (b))|2 = |(a. It is easy to see now how the two paradoxes are avoided. AΩa) = (AΩa. It is not necessary to preserve inner products.7. October 14 Notes from Sidney Coleman’s Physics 253a 68 In a more limited sense. they aren’t measurable. Given that F (a) (F : H → H) is continuous and for any a and b |(F (a). can also be thought of as a transformation of the operators in the theory. Ωa) = (a. The simpler thing to look at in a relativistic theory is actually P T .

they just lie there. .7. . . It turns that into −i! Ω−1 φ(x)ΩP T = PT d3 k (2π)3/2 2ωk [aa eik·x + aa† e−ik·x ] k k = φ(−x) ! P T does nothing to momentum states. . October 14 Notes from Sidney Coleman’s Physics 253a 69 Furthermore we’ll take ΩP T |0 = |0 and it follows that ΩP T |k1 . kn = |k1 . (2π)3/2 2ωk But what about that i up in the exponential. . It does nothing to a† . . k 1√ . What does this operator do to φ(x)? φ(x) = d3 k (2π)3/2 2ωk [aa e−ik·x + aa† eik·x ] k k it does nothing to ak . Parity turns k → −k and time reversal changes it back again. kn again nothing. . . and nothing to Apply ΩP T to φ(x). That is expected. .

8. October 16

Notes from Sidney Coleman’s Physics 253a



October 16

Scattering Theory
For a wide class of quantum mechanical systems, the description of any state is simple if you go far enough into the past or far enough into the future. We’ll illustrate this by a sequence of three increasingly complicated systems. 1. NRQM, two particles interacting through a repulsive force, which dies off at large separation. p2 p2 A H= + B + V (|rA − rB |) 2mA 2mB V ≥0 V (∞) = 0

Any state of the system, a real normalizable state, not a plane wave, looks in the far past / future like two particles far apart (actually a superposition of such states). The potential always pushes the particles far apart for times in the far future, and because the potential dies off they then look like noninteracting particles





There is no trace of the interaction in the far future or the far past. The states act like states with dynamics governed by a free Hamiltonian, H0 , which is simple, H0 = p2 p2 A + B 2mA 2mB

The aim of scattering theory is to tell what (superposition of) simple state(s) in the far future a simple state in the far past evolves into.

8. October 16

Notes from Sidney Coleman’s Physics 253a


How far in the “far past” you have to go depends on the initial conditions and the interaction. If the interaction is the nuclear force and we collide two neutrons at low energy which elastically scatter near t = 0, then t = −7 years is far enough in the far past so that the system looks like two non interacting nucleons. If however the initial conditions are set up so that the elastic scattering occurs around t = −1 billion years then you might have to go to t = −(1 billion and seven) years to make the system look simple. It need not be that the simple Hamiltonian in the far future is the same as the simple Hamiltonian in the far past. 2. NRQM, Three particles A, B, C which interact through interactions which are strong enough to make an AB bound state. Start in the far past with C and the AB bound state. Everything still looks like free particles governed by a free Hamiltonian but the free Hamiltonian in the far past is H0 = p2 p2 AB + C 2mAB 2mC

Now let scattering occur. In the far future, we can get states that look like three particles, A, B, C, non-interacting, governed by the free Hamiltonian H0 =

p2 p2 p2 A + B + C 2mA 2mB 2mC






If you had a sufficiently advanced QM course, you may have studied such a system: e+ + H → p + e+ + e− There is no way to truncate this system’s full Hamiltonian into a free part and an interacting part, for which the free part describes the evolution of the system in the

8. October 16

Notes from Sidney Coleman’s Physics 253a


far future and the far past. If you use the far future H0 , you don’t have an AB bound state. 3. (A plausible picture of) the real world. In the real world we have loads of (stable) bound states. If the real world has a laboratory bench as a stable bound state, then I can do chalk-bench scattering, and I’ll need a description of a freely flying piece of chalk and a freely flying laboratory bench. The description of states in the far past requires states with free electrons, hydrogen atoms, protons, Iron atoms, Iron nuclei, laboratory benches, chalk, and the associated free Hamiltonians. Let’s get some formalism up. (The first part of this lecture, with lots of words and few equations, is the part of a lecture that makes some people nervous and some people bored.) Let H be the actual Hamiltonian of the world and H be the actual Hilbert space of the world. If you go sufficiently far in the past, every state in the actual Hilbert space looks simple. Let H0 be the Hilbert space of simple states and let |ψ ∈ H0 . Somewhere in the real world Hilbert space there is a state that looks like |ψ in the far past. We’ll label that state |ψ in , |ψ ∈ ∈ H. Given another state |φ ∈ H0 , there is another state in the real world Hilbert space that looks like |φ in the far future. We’ll label that state |φ out , |φ out ∈ H. States in the complicated space are labelled by what they look like in the far past or the far future. What we are after in scattering theory is the probability, and hence the amplitude, that a given state looking like |ψ in the far past, looks like |φ in the far future. We are after out φ|ψ in The correspondence between |ψ in and |ψ (for every state |ψ ∈ H there is a state that looks like |ψ in ∈ H in the far past) and between |φ in and |φ allows us to define an operator in the simple Hilbert space H0 : S, the scattering matrix, which is defined by φ|S|ψ ≡ out φ|ψ in An ideal scattering theory would have two parts 1. A turn the crank method of obtaining the “descriptor” states |ψ , |φ , that is, generating H0 , from the real world Hamiltonian. We also need H0 which gives the evolution of the descriptor states. H0 evolves the descriptor states without scattering. 2. A turn the crank method of obtaining S.

8. October 16

Notes from Sidney Coleman’s Physics 253a


90% of the rest of this course will be devoted to calculating the matrix elements of S perturbatively. That’s an ideal scattering theory. We want to get calculating so we’ll start with a bargain basement, K-Mart scattering theory.

Low Budget Scattering Theory
Imagine that H can be written as H = H0 + f (t)H , f (t) = 0 for large |t|, and H0 a free Hamiltonian that evolves states simply, without scattering. Unless the interaction is with some externally specified apparatus, interesting Hamiltonians and the real world Hamiltonian are not of this form. We want a simple description of states in the far past / future. Because the interaction is off in the far past / future, the simple descriptor states are simply the states in the full theory far enough in the past / future that f (t) = 0, H = H0 . Furthermore, the Hamiltonian that gives the evolution of the simple states, H0 , is just the full Hamiltonian H, far enough in the past / future that f (t) = 0. Most Hamiltonians don’t have an f (t) in them that goes to zero as |t| → ∞. However, many Hamiltonians are of the form H = H0 + H where H0 is a Hamiltonian we know the solution of. Maybe we could put an f (t) into the Hamiltonian without changing scattering processes much. We know we can’t do this in system (2). No matter how far you go into the far past / future, it is the interaction that holds the stable AB bound state together, and you can’t shut the interaction off without the bound state falling apart, totally changing scattering processes, no matter how long you wait to shut it off. The real world is like system (2) and we can only get a little ways studying the real world if we hack it up like this. We might get a ways studying system (1) like this. In the far past / future the particles in system (1) are far apart and noninteracting. If f (t) → 0 in the far past / future, it should not affect their evolution since they are not interacting then anyway. Suppose we wanted to insert an f (t) to study a theory of electrons interacting through a repulsive Coulomb force. We can see a flow developing. As a single electron goes off to ∞ away from all others, it still has a Coulomb field (cloud of photons) around it. If you weigh an 2 electron, you get a contribution E d3 x to the mass-energy in addition to the contribution 8π to the mass-energy at the heart of the electron. This is one and the same electric field that causes the scattering to take place, and you can’t turn off scattering without turning off this cloud. Maybe, if we turn the interaction off sufficiently slowly the simple states in the real theory will turn into the states in the free theory with probability 1. We want f (t) to look like:

8. October 16

Notes from Sidney Coleman’s Physics 253a


f(t) t T
f (t) turns on and off adiabatically. A more precise way of stating the condition under which we hope inserting f (t) into the theory won’t change scattering processes much is there must be a 1-1 correspondence between the asymptotic (simple) states of the full Hamiltonian and the states of the free Hamiltonian. That means no bound states, no confinement. We hope scattering processes won’t be changed at all under this assumption in the limit ∆ → ∞, T → ∞, ∆ → 0. The last limit is needed so that edge effects are negligible. We want T adiabatic turn on and off, but we also want the interaction to be on much longer than the amount of time we spend turning it on and off. Similar requirements must be imposed if you put a system in a spatial box, depending on what kind of quantities you want to know about. In slightly racy language, the electron without its cloud of photons is called a “bare” electron, and with its cloud of photons a “dressed” electron. The scattering process goes like this: In the far far past a bare electron moves freely along. A billion years before it is to interact it leisurely dresses itself. Then it moves along for a long time as a dressed electron, briefly interacts with another (dressed) electron and moves for a long time again, dressed. Then it leisurely undresses. We need to develop Time dependent Perturbation Theory for Hamiltonians of the form H = H0 + H (t) H (t) may depend on time because of externally varying interactions or because of the insertion of f (t). We’ll do the formalism in the interaction picture developed by Dirac. This is the best formalism for doing time-dependent perturbation theory. If you have laser light shining on an atom, and you know this formalism, it is the most efficient way of calculating what happens to the atom, although it is higher powered than the minimum formalism you need for that problem. Schr¨dinger picture o We have states evolving in time according to i d |ψ(t) dt

= H(pS , qS , t)|ψ(t)


for Schr¨dinger, since we’ll be o working in several pictures

8. October 16

Notes from Sidney Coleman’s Physics 253a


The fundamental operators, pS and qS are time independent qS = qS (t) = qS (0) pS = pS (t) = pS (0)

The only operators that are not time independent are operators that explicitly depend on time. The time evolution operator U (t, t ) is given by |ψ(t)

= U (t, t )|ψ(t )


U is completely determined by the 1st order differential equation in t i d U (t, t ) = H(pS , qS , t)U (t, t ) dt

and the initial condition U (t, t )|t=t = 1. Think of t as a parameter. U is unitary (from the Hermiticity of H), i.e. U (t, t )† = U (t, t )−1 , which expresses the conservation of probability. U also obeys the composition law U (t, t )U (t , t ) = U (t, t ) which implies U (t, t ) = U (t , t)−1 Heisenberg Picture The states do not change with time |ψ
possible explicit time dependence H

= |ψ(0)


= |ψ(0)


If AS (t) is an operator in the Schr¨dinger picture and AH (t) its counterpart in the o Heisenberg picture, demand that

φ(t)|AS (t)|ψ(t)


=H φ(t)|AH (t)|ψ(t) H =S φ(0)|AH (t)|ψ(0) S =S φ(t)|U (0, t)† AH (t)U (0, t)|ψ(t)
† † S

∴ AH (t) = U (0, t)AS (t)U (0, t) = U (t, 0) AS (t)U (t, 0) Suppose we have some function of operators and the time in the Schr¨dinger picture, o itself an operator. For example H itself is a function of pS , qS and t. To get the operator in the Heisenberg picture, all you have to do is replace p’s and q’s by pH and qH . HH (t) = H(pH (t), qH (t), t)
Expand H as a power series and insert U (t,0)U (t,0)† all over.

8. October 16

Notes from Sidney Coleman’s Physics 253a


Interaction Picture Assume H can be written as H(p, q, t) = H0 − (p, q) + H (p, q, t). This defines the relation between the functions H, H0 and H , the arguments of these functions will change. The interaction picture is intermediate between the Schr¨dinger picture and the Heiseno berg picture. You make the transformation you would make to get the Schr¨dinger picture o to the Heisenberg picture, but you do it using just the free part of the Hamiltonian only. |ψ(t)

= eiH0 (pS ,qS ,t)t |ψ(t)


If there were no interaction H , there would be no time evolution of the states. That is what makes the interaction picture so useful.

If H0 depended explicitly on time we would have to define a U0 (t, 0) which would take the place of e−iH0 t , but we will have no occasion to be that general. If AS (t) is an operator in the Schr¨dinger picture, and AI (t) its counterpart in the interaction picture, we get o AI (t) = eiH0 (pS ,qS )t AS (t)e−iH0 (pS ,qS )t by demanding


φ(t)|AS (t)|ψ(t)


=I φ(t)|AI (t)|ψ(t)


We can find a differential equation for |ψ(t) i d |ψ(t) dt

d iJ0 (pS ,qS )t e |ψ(t) S dt = eiH0 (pS ,qS )t [−H0 (pS , qS ) + H(pS , qS , t)]|ψ(t) =i =e
iH0 (pS ,qS )t


[H (pS , qS , t)]e

−iH0 (pS ,qS )t



= H (pI , qI , t)|ψ(t) ≡ HI (t)|ψ(t)

Expand H (pS , qS , t) in a power series and insert e−iH0 t eiH0 t all over.

As promised, if H is zero, no time evolution.

In field theory, HI (t) will contain the free fields φ(x, t) = eiH0 t φS (x)e−iH0 t That is why all our results about free fields are still going to be useful. We can define UI (t, t ) by |ψ(t) I = UI (t, t )|ψ(t ) I
† UI (t, t ) = UI (t, t )−1

UI (t, t )UI (t, t ) = UI (t, t ) UI (t, 0) = eiH0 (pS ,qS )t U (t, 0)

UI (t, t ) = UI (t , t)


In a field theory, φI (x) obey free eq of motion + commutation relations. UI (t, t ) can be determined from the first order differential equation in t it satisfies i d UI (t, t ) = HI (t)U (t, t ) and the initial conditions UI (t, t )|t=t = 1 dt

8. October 16

Notes from Sidney Coleman’s Physics 253a


Now we’ll apply interaction picture perturbation theory to scattering theory. In the interaction picture the scattering process looks like this: In the far past the interaction is not felt, both because of f (t) and the fact that these particles are far apart. The states just like there, although the p’s and q’s are changing. The time of scattering approaches and the state starts changing. After scattering, they stop scattering, like a game of musical chairs, everything freezes again. You want to connect the simple description in the far past to the simple description in the far future. Because of f (t) the simple description in the far past / future is the actual description in the far past / future. The states in the far past and future are their own descriptors (used in arrowed step). φ|S|ψ ≡







= I φ(∞)UI (∞, −∞)|ψ(−∞) = φ|UI (∞, −∞)|ψ ∴ S = UI (∞, −∞)

Our number one priority then is to evaluate UI (∞, −∞). That will cause us to develop Dyson’s formula and Wick’s theorem. Then we’ll apply this formalism to three models. Proof that S = UI (∞, −∞) in the Schr¨dinger picture. o d |ψ in = H|ψ in dt d i |ψ = H0 |ψ dt out φ|ψ in ≡ φ|S|ψ i

|ψ(−∞) = |ψ(−∞)








any time will do

= φ(∞)|U (∞, 0)U (0, −∞)|ψ(−∞) = φ(∞)|U (∞, −∞)|ψ(−∞)



= φ(0)|U0 (∞, 0)†−1 U (∞, −∞)U0 (0, −∞)−1 |ψ(0) = φ|UI (∞, −∞)|ψ [S, H0 ] = 0 because S turns free states of a given energy into other free states of the same energy.

8. October 16

Notes from Sidney Coleman’s Physics 253a


Dyson’s Formula
We would like to find the solution of the equation i d UI (t, t ) = HI (t)UI (t, t ) dt
@ @ t @@@@ ) t dt HI (t

UI (t, t ) = 1

Imagine that [HI (t), HI (t )] = 0, which is not true. Then the solution would be U @ t ) @@@ (t, = e−i @I @@

Let’s define a new exponential so this equation is right. Given a string of operators define the time-ordered product T [A1 (t1 ) · · · An (tn )] to be the string rearranged so that later operators are to the left of earlier operators, with the operator with the latest time one the leftest. The ambiguity of what to do at equal times does not bother us when the operators commute at equal times. This is certainly the case when all the operators are the same, HI (t), evaluated at various t. T , the symbol for the time ordering operation, is not an operator in Hilbert space. Time ordering is a notation. Now we’ll show that the differential equation for UI (t, t ) is satisfied by T e−i
t t

dt HI (t )


Under the time ordering symbol everything commutes, so we can naively take a time derivative to get t t d i T e−i t dt HI (t ) = T HI (t)e−i t dt HI (t ) dt Now t is a special time. It is the latest time, so the time ordering puts HI (t) on the leftest, and we can pull it out on the left to get i d −i Te dt
t t

dt HI (t )

= HI (t)T e−i

t t

dt HI (t )

This solution of the differential equation also obeys the boundary condition (any old ordering does that). The solution of a first order differential equation with given initial value is unique. Therefore UI (t, t ) = T e−i
t t

dt HI (t )


Dyson’s Formula

To illustrate what Dyson’s formula means, we’ll look at the second order term in the power series expansion for the exponential. (−i)2 2!
t t

t t

dt2 T (HI (t1 )HI (t2 ))

” µ2 µ2 .8. which is not a big surprise. We will have to use (and thus gain experience with) our adiabatic turning on and off function. October 16 Notes from Sidney Coleman’s Physics 253a 79 t2 t t' t 1 <t2 t 1 =t2 t 1 >t2 t' t t1 Think of this as an integration over the square. You can easily get the formula for t < t by taking the adjoint. f (t). and is a quantum field. We’ll call model 1 quantum meso-dynamics. The equation of motion is ( + µ2 )φ(x) = −gρ(x) Electromagnetism with an external source which generates the EM field looks like Aµ = −ej µ . We are going to apply Dyson’s formula to three model theories in order of increasing complexity. Model 2 L = 1 ∂µ φ∂ µ φ − 2 φ2 − gρ(x)φ(x) 2 ρ(x) → 0 as |x| → ∞. a source V which will create mesons. This is the quantum scalar analog of electrostatics. ρ(x) is a prescribed c-number function of space-time. If the time ordering symbol were not there. Because of the time ordering symbol. Our formula is only valid for t > t . but it isn’t because the source does not turn off as |t| → ∞. We’ll be able to solve it exactly. Except that our field is massive. in momentum space it is just a bunch of independent forced harmonic oscillators. has no vector index. Model 1 L = 1 ∂µ φ∂ µ φ − 2 φ2 − gρ(x)φ(x) 2 ρ(x) → 0 as x → ∞ in space or time. these two theories look similar. you would be integrating H(t1 )H(t2 ) over the whole square. you get instead twice the integral of H(t1 )H(t2 ) over the lower half of the square. so we’ll call it “mesostatics. the shaded triangle. You might think this time independent problem would be easier than model 1. This is the same as model 1 except the source is static.

Wick’s Theorem When doing perturbative calculations in g in any of these three models. the current j µ is not prescribed.8. from ak1 ak2 ak3 ak4 and ak1 ak2 ak3 a† to a† a† a† a† . ( + m2 )ψ = −gψφ This theory also looks a lot like Yukawa’s theory of the interaction between mesons and nucleons. If we could rearrange these into k4 k1 k2 k3 k4 normally ordered products. only). we would again have great simplifications when we sandwiched this between states of mesons and nucleons. We had better not push this theory too far (we’ll be doing low orders in P. Wick’s theorem turns time ordered products of free fields into normal ordered products of free fields. The classical Hamiltonian contains gψ ∗ ψ and that is not bounded below for either sign of g. The φ field in turn appears in the equation of motion for the ψ field. it is the current of charged particles.T. In model 1. Here we have the charged field ψ as a source for φ (ψ ∗ ψ is like a current). At fourth order in g for a meson scattered by the source we would have to evaluate these are free fields k |T (φ(x1 )φ(x2 )φ(x3 )φ(x4 ))|k The time ordered product contains 16 arrangements of creation and annihilation operators. except our charged particles are spinless and we only have one meson. the only normally ordered product that could contribute would be the one with one creation operator on the left and one annihilation operator on the right. A(x)B(y) ≡ T (A(x)B(y))− : A(x)B(y) : . a great simplification. HI = gρ(x)φ(x). In model 3 we wil have to evaluate time ordered products like T (φ(x1 )ψ ∗ (x1 )ψ(x1 )φ(x2 )ψ ∗ (x2 )ψ(x2 )) If we had an algorithm for normal ordering the time ordered product. we are going to have to evaluate time ordered products of strings of Hamiltonians between states. To state Wick’s theorem we’ll define the contraction. October 16 Notes from Sidney Coleman’s Physics 253a µ2 80 Model 3 L = 1 ∂µ φ∂ µ φ − 2 φ2 + ∂µ ψ ∗ ∂ µ φ − m2 ψ ∗ ψ − gψ ∗ ψφ 2 The equation of motion for the φ field is ( + µ2 )φ = −gψ ∗ ψ This is beginning to look like the real thing. In real electrodynamics.

. Theorem (Gian-Carlo Wick) term with no contractions all the other possible+ terms with one contraction T (φ1 . Convince yourself the ± doesn’t matter. d4 k ±ik·(x−y) i φ(x)φ(y) = 0|T (φ(x)φ(y))|0 = e 4 2 − µ2 + i (2π) k lim →0+ is understood. A(x)B(y) = 0| A(x)B(y) |0 = 0|T (A(x)B(y))|0 − @@:@@@@@ : |0 0| A(x)B(y) @@@ That’s why the calculation of 0|T (φ(x)φ(y))|0 in the first problem set is going to be useful. really. that x0 > y 0 .8. October 16 Notes from Sidney Coleman’s Physics 253a 81 Suppose. That you get all that is no surprise. Choose without loss of generality χ10 ≥ χ20 ≥ · · · ≥ χn0 . Trivial for n = 1. W ] (+) . Proof (By induction) Define the RHS of the expression to be W (φ1 · · · φn ). A(x)B(y) is a c# and is thus equal to its vacuum expectation value. . . etc. + 1 n(n−1) (n−2)(n−3) 2 2 2 possible terms with two contractions + . So whether x0 < y 0 or y 0 < x0 . This is also a c# when x0 < y 0 . The remarkable and graceful thing about this theorem is that each term occurs with coefficient +1. You can also see that ψ(x)ψ ∗ (y) = ψ ∗ (x)ψ(y) = d4 k ±ik·(x−y) i e 4 2 − m2 + i (2π) k A little more obvious notation: : A(x) B(y)C(z)D(w) :≡: A(x)C(z) : B(y)D(w) and let φ1 ≡ φa1 (x1 ). B (−) ] and A(x)B(y) = [A(+) . . just for this proof. . B (−) ] which is a c-number. Then induction step T (φ1 · · · φn ) = φ1 T (φ2 · · · φn ) = φ1 W (φ2 · · · φn ) = φ1 W + W φ1 (−) (+) + [φ1 . . Using its definition Using the defn. .  : φ φ φ φ · · · φ φ : 1 2 3 4 n−1 n + : φ φ φ φ · · · φ φ φ : 1 2 3 4 n−2 n−1 n if n is even if n is odd You draw all possible terms with all possible contractions. . φ2 ≡ φa2 (x2 ).. Then T (A(x)B(y)) = A(x)B(y) = (A(+) + A(−) )(B (+) + B (−) ) =: AB : +[A(+) . without loss in generality. φn ) = : φ1 · · · φn : + : φ1 φ2 · · · φn : + all the other all possible terms with < n−1 2 contractions n(n−1) 2 : φ1 φ2 φ3 φ4 · · · φn : . 2. We want to show W = T .

8. . The third term contains all possible contractions that do include φ1 . Together they contain all possible contractions. October 16 Notes from Sidney Coleman’s Physics 253a 82 This expression is normal ordered. Either a contraction includes φ1 or it doesn’t. The right hand side is thus W (φ1 · · · φn ). The first two terms contain all possible contractions that do not include φ1 .

which for those of you who really love combinatorics can be written n 1 φφ ∂ ∂ T (φ1 · · · φn ) =: e 2 i. −∞) is UI (∞. The ψ ∗ field contains operators that annihilate an anti-nucleon and create a nucleon. HI = d3 xHI . In model 3. October 21 Notes from Sidney Coleman’s Physics 253a 83 9 October 21 Diagrammatic Perturbation Theory Dyson’s formula applied to S = UI (∞.j=1 i j ∂φi ∂φj φ1 · · · φn : We have enough work done to write down diagrammatic perturbation theory for S = UI (∞. −∞). The operator : ψ ∗ ψ φ(x1 )ψ ∗ ψφ(x2 ) :=: ψ ∗ ψ(x1 )ψ ∗ ψ(x2 ) : φ(x1 )φ(x2 ) can contribute to N + N → N + N . −∞) = T e−i dtHI (t) (Without the use of the time ordering notation. UI (∞. The easiest way to see this is to look at a specific model and a contribution to UI (∞. HI = gf (t)ψ ∗ ψφ. and Dyson says he should not have credit for little more than a change in notation. The ψ field contains operators that annihilate a “nucleon” and operators that create an anti-“nucleon”. this formula for UI (t. t ) was written down by Dirac 15 years before Dyson wrote it this way.9. −∞) at a specific order in g.) From this and Wick’s theorem. That is to say final 2 nucleon state| initial 2 : ψ ∗ ψ(x1 )ψ ∗ ψ(x2 ) : |nucleon state . −∞) = T e−i the contribution at second order in g is (−ig)2 2! d4 x1 d4 x2 f (t1 )f (t2 )T (ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 )) d4 xHI = T e−i d4 xgf (t)ψ ∗ ψφ One of the terms in the expansion of the time ordered product into normal ordered products by Wick’s theorem is (−ig)2 2! d4 x1 d4 x2 f (t1 )f (t2 ) : ψ ∗ ψ φ(x1 )ψ ∗ ψφ(x2 ) : This term can contribute to a variety of physical processes.

join the lines of the contracted fields. because the contractions for which they don’t are zero. At N th order in perturbation theory. A single term is capable of contributing to a variety processes because a single field is capable of creating or destroying a particle. It can be thought of as annihilating a nucleon or creating an antinucleon The vertex represents the factor of f ψ ∗ ψφ. The terms in the Wick expansion can be written down in a diagrammatic shorthand according to the following rules. An unarrowed line will never be connected to an arrowed line because that contraction is also zero. you start by writing down N interaction vertices and numbering them 1 to N . Our first term in the expansion of the time ordered product corresponds to the diagram ¤ffi i ¤ 1 2 . 2φ → N + N . Another term in the expansion of the time ordered product into normal ordered products is (−ig)2 d4 x1 d4 x2 f (t1 )f (t2 ) : ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 ) : 2! This term can contribute to the following 2→2 scattering processes: N + φ → N + φ. The arrows will always line up. You can see that there is no combination of creation and annihilation operators in this operator ¯ ¯ that can contribute to N + N → N + N . N + N → 2φ. which would be a disaster. ¯ ¯ ¯ ¯ N + φ → N + φ. It creates or destroys a meson fi fi ¤ ¤ 1 2 This outgoing line is for the ψ ∗ in ψ ∗ ψφ(x2 ). For model 3 at second order in perturbation theory you write down This line is for the φ in ψ ∗ ψφ(x1 ). to give a ¯ ¯ ¯ ¯ ¯ ¯ nonzero matrix element. The ψ fields would have to annihilate the nucleons ∗ and the ψ fields cannot create antinucleons. The coefficient of that term after integrating over x1 and x2 had better tun out to be zero. This is good because this process does not conserve the U (1) symmetry charge. Any time there is a contraction. It can be thought of as creating a nucleon or annihilating an antinucleon This incoming line is for the ψ in ψ ∗ ψφ(x2 ).9. It can also contribute to N + N → N + N and N + N → N + N . From the fact that there are two in this 2 diagram you know to include (−ig) d4 x1 d4 x2 2! Contractions are represented by connecting the lines. October 21 Notes from Sidney Coleman’s Physics 253a 84 is nonzero because there are terms in the two ψ fields that can annihilate the two nucleons in the initial state and terms in the 2 ψ ∗ fields that can then create two nucleons. However it looks like our operator can contribute to ¯ ¯ vacuum→ N + N + N + N .

The vertices are numbered in Wick diagrams and ¦¦e d ¦ 2 1 is distinct from ¦¦e d ¦ 1 2 (there are two distinct terms in the Wick expansion) even though after integrating over x1 and x2 these are identical operators. we can shorten ¦¦¦e ¦¦e d ¦ d ¦ 1 2 to 1 2 These diagrams are in one-to-one correspondence with terms in the Wick expansion of Dyson’s formula. but these aren’t them yet. we are most of the way to Feynman diagrams which stand for matrix elements. They stand for operators and the vertices are numbered. October 21 Notes from Sidney Coleman’s Physics 253a 85 The second term corresponds to ¦¦¦e d ¦ 1 2 The term in the Wick expansion (−ig)2 2! d4 x1 d4 x2 f (t1 )f (t2 ) : ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 ) : is zero because ψ ∗ ψ ∗ = 0. On the other hand 1 fxf 2 is identical to 2 fxf 1 . We’ll call them Wick diagrams. In Feynman diagrams the lines will be labelled by momenta. so we never write down pp¦e d ¦ 1 2 Because the arrows always line up.9.

− − − 1 • f 1 and • •. We have been putting the normal ordering inside the integrand. Rotate the right one by 180◦ and you see they are the same. 2 . but is connected. . At O(g) in UI we have (−ig) d4 x1 ρφ(x1 ) which is represented by • . • • • • is a fully contracted diagram that is not connected. The vertex in this model is That represents ρφ(x).• 1 •• 2 3 • 2 •. Although these two have been written down to look different they aren’t. • is not contracted. 2 . ρ(x) is a prescribed c-number source. Of course we could put it around the whole integral in which case we see that this O(g 4 ) diagram corresponds to (−ig)4 : 4! d4 x1 d4 x2 d4 x3 d4 x4 ρ(x1 )ρ(x2 )ρ(x3 )ρ(x4 ) φ(x1 )φ(x2 ) φ(x3 )φ(x4 ) := (−ig)4 : 4! d4 x1 d4 x2 φ(x1 )φ(x2 ) ρ(x1 )ρ(x2 ) d4 x3 φ(x3 )ρ(x3 ) d4 x4 φ(x4 )ρ(x4 ) : ff f fff f f ff fff • 2 • 2 • 3 2 3 That is. October 21 Notes from Sidney Coleman’s Physics 253a 86 (there is only one way of contracting all three fields at one vertex with all three at the other). It doesn’t mean fully 12 3 4 1 contracted. and • 3 1 4 ff •• 3 2 . 1 At O(g 2 ) in UI we have • At O(g 3 ) in UI we have • 4 1 1 1 f A diagram at O(g ) is • •• • . the integrands factor into products of terms corresponding to each connected17 part of the diagram. This suggests we can sum the series and then normal order in this 17 “Connected” means (in any theory) that the diagram is in one connected piece.9.• 1 fxf The contraction this diagram corresponds to is (−ig)2 2! d4 x1 d4 x2 f (t1 )f (t2 ) : ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 ) : In model 1 HI = gρ(x)φ(x) (we don’t have to insert a turning on and off function because the interaction goes to zero as x → ∞ in any direction and in particular in the time direction in the far past / future). so strongly made we don’t have to worry about the back reaction of the field φ on the source.

. xn(D) . n(D). we won’t solve it. . . Associated with this diagram is an operator : O(D) : n(D)! We have explicitly displayed the n(D)! and we have pulled the normal ordering outside. for example dpe f¦¢d e 1 2 3 4 is not distinct from dpe f¦¢d e 4 1 3 2 (and there are two more cyclic permutations) but it is distinct from the diagrams with noncyclic permutations. . This is in exact correspondence with the question of whether or not there is a new term in the Wick expansion from permuting x1 . . you might expect the sum over all diagrams of a given pattern to exactly cancel the n(D)!. The free energy is just the sum of the connected diagrams. Let D be a general diagram with n(D) vertices. For example for D= f f ¤fki i ¤ 1 2 O(D) = (−ig)2 d4 x1 d4 x2 f (t1 )f (t2 ) φ(x1 )φ(x2 ) ψ ∗ ψ(x1 )ψ ∗ ψ(x2 ) I will define two diagrams to be of the same “pattern” if they differ just by permuting the labels at the vertices. October 21 Notes from Sidney Coleman’s Physics 253a 87 simple theory. . all Wick diagrams =: e connected Wick diagrams : In a theory with only two connected diagrams this theorem is powerful enough to solve the theory exactly in a couple of lines. This is not quite right however. .9. Since after integration over x1 . but since there are still an infinite number of connected diagrams in model 3. It will help a lot in model 3. . . We could do the sum in this theory. n(D). This formula is also useful in condensed matter physics where you develop a perturbation theory for Tr e−βH . . xn(D) two different diagrams of the same pattern give identical contributions to UI and since there are n(D)! permutations of the numbers 1. . . . The free energy which is the logarithm of the partition function is what is actually of interest. . This theorem’s analogue tells you that you don’t have to calculate a huge series for Tr e−βH and then try to take its logarithm. . 1. . but instead we will prove a general theorem. 2. because at all orders in g the diagrams only contain • • and • various numbers of times. For some diagrams there are permutations of the vertices that have no effect. .

. D2 . there are n! whole exchanges.9. there will be some symmetry number. So S(D) contains r nr !. . . Dr . A general diagram. . with one diagram of each pattern. . . October 21 Notes from Sidney Coleman’s Physics 253a 88 For any pattern. Thus summing over all patterns is the same as summing over all sets {nr }. we have to sum over all patterns. . xn(D) that do not give additional contributions in the Wick expansion). The sum over all diagrams with the same pattern as D gives : ∞ [O(Dr )]nr : : O(D) : = ∞r=1 nr S(D) r=1 [S(Dr ) nr !] Now that we have done the sum over all diagrams of a given pattern. If I have n identical factors. What is S(D)? S(D) S(D) certainly contains r [S(Dr )]nr . D. S(D). Notice that there is a 1-1 correspondence between patterns and sets {nr }. So. has nr components of pattern Dr . Summing over all distinct diagrams of the same pattern as D yields : O(D) : S(D) Let D1 . ∞ ∞ all Wick diagrams = n1 =0 n2 =0 ∞ ∞ ··· ··· n1 =0 n2 =0 ∞ : ∞ nr : r=1 [O(Dr )] ∞ nr n !] r r=1 [S(Dr ) ∞ =:   r=1 ∞ ∞ [O(Dr )]nr : S(Dr )nr nr ! r=1 nr  O(D ) r =: S(Dr ) nr =0 nr ! : =: r=1 e S(Dr ) : ∞ O(Dr ) r=1 S(Dr ) O(Dr ) =: e =: e : : connected Wick diagrams . that there are S(D) permutations of x1 . . I can take all the indices on one of them and exchange them with the other. If I have 2 identical factors. . . . Because of the factorization of the integrands and because we have explicitly pulled out the n(D)! ∞ : O(D) :=: r=1 [O(Dr )]nr : Summing over all diagrams with the same pattern as D gives :O(D): . which is the number of permutations that have no effect on the diagram D (and of course there is the analogous statement. be a complete set of connected diagrams.

demanding that UI be unitary. using the expansion for φ(x). −∞) =: eO1 + O1 = −ig O2 2 f 1 f + 1 • f 2 •.9. October 21 Notes from Sidney Coleman’s Physics 253a 89 This is a neat theorem because it expresses a fact about diagrams. Let’s rewrite O1 . Model 1 Solved D1 = • 1 f . algebraically. We’ll get it here by a consistency argument. Now we’ll apply this to model 1. define f (k) ≡ −ig (2π)3/2 2ωk ρ(k. see Nov. 6) ˜ f (k) = d4 xeik·x f (x) f (x) = d4 k −ik·x ˜ e f (k) (2π)4 also in three space dimensions ˜ f (k) = d3 xeik·x f (x) f (x) = d3 k −ik·x ˜ f (k) e (2π)3 So as not to carry around so many factors. D2 = • :=: e• 1 UI (∞. 2 S(D2 ) = 2 2 • d4 x1 ρ(x1 )φ(x1 ) some d4 x1 d4 x2 φ(x1 )φ(x2 ) ρ(x1 )ρ(x2 ) = number = α + iβ O2 = (−ig)2 You will compute α in the homework. pictures. O1 = −ig = −ig d3 k (2π)3/2 dk (2π)3/2 2ωk 3 2ωk d4 xρ(x) e−ik·x ak + eik·x a† k ρ(−k) ak + ρ(k)a† ˜ ˜ k ρ(k)∗ ˜ Using the Fourier transform convention (this convention will not be adhered to. ωk ) ˜ .

. |0 . October 21 Notes from Sidney Coleman’s Physics 253a 90 then UI (∞. −∞)|0 = e 2 (α+iβ) e = e 2 (α+iβ) n=0 1 1 d3 kf (k)a† ∞ k 1 n! d3 k1 · · · d3 kn f (k1 ) · · · f (kn )|k1 . . . −∞) = e 2 (α+iβ) e 1 d3 kf (k)a† − k e d3 kf (k)∗ ak Now that we have solved the model we can answer the usual questions you ask about when a field is driven by an external source.9. what is the probability of finding n mesons in the far future? The state in the far future is UI (∞. 1= n ! P n = eα n 1 n! n d3 k1 |f (k1 )|2 = eα e d3 k1 |f (k1 )|2 α=− So Pn = e−|α| d3 k|f (k)|2 |α|n Poisson distribution. . Given that you start with nothing in the far past. n! This state. . eλa ]|0 = λeλa |0 = λ|λ λ|x(t)|λ and λ|p(t)|λ oscillate sinusoidally like the classical variables. kn (9. is called a coherent state. . Coherent states of the harmonic oscillator are † |λ ≡ eλa |0 They are special because they diagonalize a a|λ = aeλa |0 = [a. of finding n mesons is thus Pn = e 1 (α+iβ) 2 1 n! 2 d k1 · · · d kn f (k1 ) · · · f (kn )|k1 . created by a classical source.1) The probability. † † † . . . Pn . kn 3 3 = eα 1 (n!)2 1 = eα n! d3 k1 · · · d3 kn |f (k1 )|2 · · · |f (kn )|2 n! n d3 k1 |f (k1 )|2 Now is where we demand unitarity of UI to get α.

you get the same state back. i. the vacuum-to-vacuum scattering matrix element ought to be easy to calculate. The interaction doesn’t actually turn off in the far past / future. You start out with nothing you end up with nothing. f(t) t HI = gφ(x)ρ(x)f (t) T Assuming the theory has a ground state. ex nihil nihil. If you remove a particle with φ(+) (x).2) What is the average energy of the final state.9. October 21 Notes from Sidney Coleman’s Physics 253a 91 The coherent states we have constructed are eigenvectors of φ(+) (x) with eigenvalue d3 k (2π)3/2 2ωk e−ik·x f (k) 1 Except for the n! this state’s n particle part is just the product of n 1 particle states. model 2 is identical to model 1. If you calculate 0|S|0 however. Expectations of normal ordered products factorize. We put that in by hand. What is the average number of mesons created? ∞ ∞ N = n=0 nPn = n=1 e−|α| |α|n = |α| (n − 1)! (pull out an |α| and reindex the sum) (9. but physically the content is different. you will not get one. the total energy of all the mesons created? ∞ H = n=0 ∞ e−|α| (n!)2 −|α| d3 k1 · · · d3 kn |f (k1 )|2 · · · |f (kn )|2 (ωk1 + · · · + ωkn ) n! nωk 1 = n=1 e |α|n−1 (n − 1)! d3 k|f (k)|2 ωk = d3 k|f (k)|2 ωk Average momentum? P = d3 k|f (k)|2 k Model 2 solved (beginning) Combinatorically. It is about as uncorrelated as a state of mesons can be.e. withthe whole 0 energy E |0 = ground state of H0 as usual . Let |0 P ground state of = Hamiltonian.

How will we get rid of it? We’ll change the theory. We don’t expect this to be true in model 3. the interaction turns on adiabatically.9. −∞)|0 = e−i(γ− +γ+ +E0 T ) This is disgusting. A divergent phase. e−iγ− . the one particle state energies should be lined up. It counters a problem we ran into in scattering theory. There shouldn’t be though. 2 We can transfer this to the interaction picture. but it fixes up the problem. The number. October 21 Notes from Sidney Coleman’s Physics 253a 92 Let’s look at the scattering process in the Schr¨dinger picture. . to get 0|UI (∞. the state evolves with the full Hamiltonian. HI → g d3 xφ(x. Because the physical states get far away from the potential at large times. If the vacuum energies are lined up. The adiabatic hypothesis says that |0 turns into |0 P with probability 1. in time ∆. The particles interact with themselves and they can never get away from that as a particle can get away from an external potential. It doesn’t change the physics. At t ≈ T . getting one more phase e−iγ+ . The state we have for t > T is e−i(γ− +γ+ +E0 T |0 . Subtract the mismatch and we’ll eliminate the problem. o T T For t < − 2 we have |0 . The problem is that there is a mismatch between the ground state energy of the full Hamiltonian and the ground state energy of the free Hamiltonian. You might worry that there will be energy mismatches in the one or many particle energy states even after we get the energy mismatch in the ground states fixed up. just exponentiates while the interaction is on. a. It can pick up a phase. it 2 2 rotates as e−iE0 t and picks up a total phase e−iE0 T . From t = − T to t = + T . At t ≈ − 2 . as the interaction turns off 2 adiabatically |0 P turns back into |0 . t)ρ(x) − a f (t) a = E0 It’s obvious what will happen. to be the same as the energy difference between the bare particle and the bare vacuum. we expect the energy difference between a state with one physical particle and the physical vacuum. 0|S|0 = e−i[(γ+ +γ− +E0 T )−aT (1+O( T )] = 1 ∆ take a = E0 + O T ! ∆ This is the first example of what is called a counterterm.

it had better be purely real. (3) = −(2) Since a is an addition to the Hamiltonian. a → E0 . S = UI (∞. October 23 Notes from Sidney Coleman’s Physics 253a 93 10 October 23 Model 2 solved (conclusion) HI = f (t) g d3 xφ(x. and diagram (2) in order to be cancellable had better come out pure imaginary. Photons don’t scatter off nailed down charges. Mesons don’t scatter off nailed down nucleons. but there the source was time dependent. It didn’t come out pure imaginary in Model (1). t)ρ(x) − a a is the vacuum energy counterterm chosen so that 0|S|0 = 1 We have already argued that as T → ∞. Diagram (3) is then pure imaginary. −∞) =: e(1)+(2)+(3) := e(2)+(3) : e(1) : (Since (2) and (3) are just numbers. where E0 is the vacuum energy of the interacting theory (without the counterterm a). (1) = −ig and φ(x) = f (t)ρ(x)φ(x)d3 xdt d3 k (2π)3/2 2ωk d3 k (2π)3/2 e−ik·x ak + eik·x a† k  (1) = −ig so ˜ f (ωk )∗    ˜ ˜ f (ω )a† ρ(k) + f (−ω )a ρ(−k) k k ˜ k k˜  2ωk  ←−−→ hermitian conjugates ρ(k)∗ ˜ f (t). Now it is clear what the addition of a constant to the Hamiltonian does to UI (∞.) To set 0|S|0 = 1 is to set e(2)+(3) = 1 i.10. There are now three connected diagrams • 1 f (1) • 1 f (2) • 2 (3) x Diagram (3). Let’s see this come out of our diagrammatic perturbation theory. −∞). which has no lines coming out. Finding a is going to give us E0 . They only scatter off real nucleons (or off nailed down nucleons if there is some dynamical charged field in the theory). our turning on and off function. has a Fourier transform that looks like .e. The constant just exponentiates. is for the counterterm.

This result holds in the massless theory too. Ground State Energy. This is the same thing we do in QM. I’ll show you how to get these quantities out of the time dependent perturbation theory we have already developed. In meson-“nucleon” theory. usually you say that the protons are so much heavier than the electrons and . Since it can only create mesons one at a time and since ω = 0 is not on the mass shell. Not until miraculous cancellations of all the terms at 4th order in the Born series occurred did people realize that they should try to prove S = 1 to all orders. f (ωk ) goes to zero for every ωk = 0 and since ωk ≥ µ f (ωk ) → 0 for all k. That is (1) → 0 as T → ∞. Take ρ = “δ (3) (x − y1 ) + “δ (3) (x − y2 ) The quotes are around the δ functions because we might want to smear them out a little bit. Since there is clearly no scattering for all k = 0. (as long as we can set (3) = −(2)) we have found S=1 This theory is a complete washout as far as scattering is concerned. We calculate the interaction between the two protons + due to their interaction with the electron in H2 by considering how the ground state energy of the electron varies with the separation of the protons. Why is S = 1? A time independent source can impart no energy. does not screw up the wave packet. This is the charge density of two nucleons at y1 and y2 . October 23 Notes from Sidney Coleman’s Physics 253a 94 1 T ˜ ˜ As T → ∞. you only have to prove that for wave packets centered about k = 0. it cannot create mesons. the failure at k = 0. the source will be ψ ∗ ψ. The protons are nailed down in that calculation. You usually use time independent perturbation theory.10. Our classical source theory is a lot like a meson-“nucleon” theory with the “nucleons” nailed down. it was not easy when they were evaluating the theory in the Born approximation. By computing the ground state energy and then by varying the positions we can find the potential between two “nucleons”. Why is the ground state energy interesting? We have been studying the response of the meson field to a classical source. Ground State Wavefunction In most QM courses these are discussed in a model long before scattering. a set of measure 0. While this was easy to see in the formalism we have built up.

Also ∞ −∞ by 0 (and dω ˜ |f (ω)|2 = 2π ∞ dt|f (t)|2 = T −∞ T +O(∆) 1+O famous theorem. Of course the Coulomb potential in QM really comes from the interaction with the photons. Parseval’s theorem ∆ T We could sum up these properties by saying something sloppy like T →∞ ˜ lim |f (ω)|2 = 2πT δ(ω) but what I have just shown is all (no more. the T ’s and i’s cancel. To convert it to position space. Here we are trying to get at the whole internucleon potential by saying it all comes from the interaction with the meson field. Of course in that calculation we also have a Coulomb potential between the protons.10. Define V (x) = −g 2 Then E0 = 1 2 d3 k eik·x (2π)3 k 2 + µ2 d3 xd3 yρ(x)ρ(y)V (x − y) . Now to calculate a = E0 by setting (3) = −(2): (3) = −i (2) = dtf (t)(−a) = iaT 1+O ∆ T = iE0 T 1+O ∆ T (−ig)2 2! −ig 2 2 d4 x1 d4 xf (t1 )f (t2 )ρ(x1 )ρ(x2 ) φ(x1 )φ(x2 ) d4 k i (2π)4 k2 −µ2 +i eik·(x1 −x2 ) = d3 k |˜(k)|2 ρ (2π)3 dk 0 ˜ 1 |f (ω)|2 2π ω2 − k2 + i 1 ω 2 −k2 +i ˜ Now |f (k 0 )|2 is sharply concentrated at k 0 = 0. . no less) than that sloppy statement means. . October 23 Notes from Sidney Coleman’s Physics 253a 95 move so slowly that we can treat the response of the electron field to changes in positions of the protons as if the changes take place adiabatically. E0 = T →∞ −g 2 2 d3 k 1 |˜(k)|2 ρ 3 (2π) k 2 + µ2 The potential has come out in momentum space. (2) = ig 2 T 2 1+O ∆ T d3 k 1 |˜(k)|2 ρ (2π)3 |k|2 + µ2 The moment of truth: Set (3) = −(2). we can replace ω in then the i is not needed any longer).

we’ll use the results of model 1. . . . screw the phase factor . is of course not a position space wavefunction (the expansion of |ψ into |x ’s). −∞)|0 = k1 . then we turn the interaction off abruptly. 0). y2 +V (y1 − y2 ) If “δ → δ. and falls off rapidly for r µ−1 . . The usual procedure for the integration of spherically symmetric Fourier transforms. The ground state wavefunction. Of course it doesn’t include the effect of the whole family of mesons in the real world of multi-meson processes. this part → ∞ Same problem as the self-energy of a charged sphere in E. 2 ρ(x) = “δ (3) (x − y1 ) + “δ (3) (x − y2 ) E0 = something independent of y1 . . followed by a contour integration gives V (r) = −g 2 −µr e 4πr r = |y1 − y2 | Yukawa potential Looks like the Coulomb potential for r µ−1 the Compton wavelength of the meson. it is an expansion of |0 P into the basis states |k1 . . Two will be the interaction of the nucleons with each other (cancelled by the 1 ). there will be four contributions. This potential has some of the essential features of the real nuclear force. kn |UI (∞. Consider the S matrix in this theory k1 . To get the ground state wave function of model 2 using time dependent perturbation theory. which is 1 in the 18 19 MW is much larger and the weak force is thus much shorter ranged. kn |UI (∞. UI (∞. . → 0+ ρ(x. For the two nucleon charge density.10. kn of the noninteracting theory.D. . . −∞)|0 Since the interaction is turned on arbitrarily slowly. . . −∞) should turn19 the bare vacuum into |0 P . The force is attractive (between like charges) (because the particle mediating it has even integer spin) and short18 ranged because the mediating particle is massive. but with µ = mπ it is a start. . UI (0. t) = 0 t>0 That is we turn it on very slowly. October 23 Notes from Sidney Coleman’s Physics 253a 96 1 Then 2 is the usual factor found even in electrostatics from overcounting the interaction when integrating over all space. Consider ρ(x)e t t < 0. 0)UI (0. the evolution by the free Hamiltonian alone. arbitrarily slowly so that at t = 0 we finally have the full interaction of model 2.

Now we can apply the results of model 1 (Oct. kn |0 which is what we are after. . −∞)|0 = k1 . Physically observable quantities like the S matrix and the internucleon potential are hearteningly sensible. Even if we don’t go to the limit “δ → δ. Since N = |α| (Oct. . 21. ρ(k) → 1 ˜ and this is bad news. What about that? . These infinities are scary but not harmful. N = |α| → ∞. . . kn |UI (∞. ωk ) ρ 3/2 (2π) 2ωk 0 |α| = ρ(x) −∞ d3 k|f (k)|2 ρ(k) = ˜ d xe 4 ik·x ρ(x) = d xe 3 −ik·x dteik t e t 0 = ρ(k) ˜ >0 ik 0 1 i − → − 0 ρ(k) − ˜ →0 + k The probability for having n mesons is Pn = e−|α| |α|n n! What is Pn for a point charge at the origin? That is ρ(x) → δ(x) Well. . but instead take the limit of massless mesons µ → 0. . . (9. at high k we have a UV divergence in the integral for |α|. Eq. . the limit of a real point charge) high k high k The integral is log divergent. . . does nothing on the left so k1 . kn |UI (∞.10. but the ground state flees Fock space. (9. −∞)|0 = e− |α| 2 iβ e 2 f (k1 ) · · · f (kn ) 1 1 f (k) = (−ig)˜(k. .1)) k1 . . . kn |0 P P = k1 . . |α| = d3 k|f (k)|2 ∼ d3 k 1 3 ωk ∼ k 2 dk k3 (not “δ smeared.2)) we see that not only the energy of the field becomes infinite in the limit of a point nucleon. . October 23 Notes from Sidney Coleman’s Physics 253a 97 interaction picture. this time because of a small k divergence. Eq. . 21.

Same with P = d3 k|f (k)|2 k So it seems we have been lucky. The change in energy of a particle due to its interaction with another field is called “mass renormalization.10. In fact there is a surprisingly wide class of theories in which the divergences don’t break the quarantine. the divergences have restricted themselves to unobservable quantities. the coupling. but there isn’t. so what. the energy of a “static nucleon” depended on its interaction with the meson field. of this system is not necessarily zero. Indeed. In this simple theory. Maybe the divergences will break this quarantine in more complicated theories. It might be a problem if there were an infinite amount of energy at small k. October 23 Notes from Sidney Coleman’s Physics 253a 98 Answer: So what if there are an infinite number of mesons in the ground state. And the extra factor of ωk moderates the IR divergence. with mass equal to 20 of the water it displaces. Not only did the vacuum energy of a state with one static nucleon depend on g.” Mass renormalization goes all the way back to hydrodynamics. Seems to contradict Yukawa pot. The integral is finite even as µ → 0. Next: Mass renormalization. in perturbation theory. not arbitrarily low.” The cure for this disease is also called “mass renormalization. If there are 1.000 mesons with a wavelength between 1 light 2 year and 1 light year. In this theory. Model 3 and Mass Renormalization The ground state energy. that is not enough to make the 1-particle states of the Hamiltonian equal in energy to the 1-particle states of the full Hamiltonian. it depended on how smeared the nucleon was. An experimentalist will tell you he can only measure the existence of a meson down to some low energy. m0 = 1 ρV 20 20mog 1mog . They are called “renormalizable” theories. H = d3 k|f (k)|2 ωk This is manifestly positive. result.000. 1 Suppose I have a ping pong ball.

leading up to the problem on p. This is nonsense. as anyone who has ever held a ping pong ball underwater knows. The ping pong ball may accelerate up fast. Putting this in Newton’s equation we have m0 a = 19m0 g a = 19g The ball accelerates upwards at 19g. Fluid Mechanics. is thus 1 ρV v + m0 v = 11m0 v 2 which we set equal to the force after taking d dt dp = 11m0 a = 19m0 g = F dt 19 a= g 11 See the derivation in Landau and Lifshitz. The answer is not friction. The total momentum of 2 the system. ball and fluid. the fluid flow has a momentum. you have to accelerate some fluid. you have to move some fluid. moving with velocity v through fluid If a ball that displaces volume V is moving with velocity v through the fluid. of 1 ρV v. flow lines in rest frame of sphere. October 23 Notes from Sidney Coleman’s Physics 253a 99 Elementary hydrostatics tells you that there is an upward force on the ping pong ball equal to g times the mass of the water it displaces. The net force on the ball is thus 19m0 g upward. .10. Stokes solved the fluid motion around a sphere. The answer is that in order to move the ping pong ball. There is also the downward force of gravity on the ping pong ball itself.36. for a more detailed understanding of the problem. You can see that the ping pong ball is not accelerating with 19g even when its velocity is low and friction or viscosity is negligible. in the same direction as the ball. but not at 19g. In order to accelerate the ping pong ball.

However. In fact if we don’t smear it out at all ρ(x) = δ (3) (x) − y) the energy of the meson ground state → −∞. the radius. There is a contribution to the measured mass of the electron other than m0 . as well as other properties of the system in terms of the microphysical parameters. The bare mass is m0 . Another example of renormalized perturbation theory: You could try. the charge.10. e2 m = m0 + 2rc2 measured. The energy of a single meson state or a single nucleon state is going to depend on its interaction. The classical theory of the electron also suffers mass renormalization. in the statistical-mechanical theory of critical phenomena to calculate the critical temperature. r. . Classical field limit: Can read dispersion relation for low amplitude plane waves off of the quadratic part of the Lagrangian. e. and although the features of this effect mass change when the coupling to the φ field goes from ρ(x)φ(x) (nailed down nucleons) to ψ ∗ ψφ(x) (dynamical nucleons) there is no reason to expect it to go away. we better be prepared for mass renormalization. Since the quantum field theory will probably exhibit all the behavior of the worst classical limit. There is the electrostatic energy (divided by c2 ). Recall that even for a single nucleon weighted down at y ρ(x) = “δ (3) (x − y) the energy of the system depends in detail on how we smear out the δ function. Imagine the electron as a charged shell. you may be able to do computations much more easily if you trade in one of the microphysical parameters for the critical temperature. October 23 Notes from Sidney Coleman’s Physics 253a 100 More motivation for mass renormalization: Two limits of classical field theory: Point particle limit: Mass renormalization occurs Example: GR Point particle of mass m creates a gravitational field which itself has energy density and creates further gravitational field. Now the energy of a one nucleon state includes the change in energy of the meson field its presence causes. physical mass Model 3 is going to suffer mass renormalization. We looked at static “nucleons” interacting with the meson field in model 2.

|p P . for sufficiently weak coupling. it can even cause two wave packets that were arranged to collide. can have a different mass. we expect that there will be one nucleon states. which because of this mass renormalization mess. Just as the failure to match up the ground state energy for the noninteracting and full Hamiltonians in model 2 produced T and ∆ dependent phases in 0|S|0 . by adjusting b and c. October 23 Notes from Sidney Coleman’s Physics 253a 101 This is going to be bad news for scattering theory. This eliminates the phases in the one particle matrix elements by matching the energy of the one particle state (the vacuum energy. is adjusted to zero with the help of a). . that the one meson state has mass µ and the one nucleon state has mass m. which is an infinite volume system may be infinite. not to collide.10. and if I turn on the interaction too early or too late. Worse than that. I’ll show how: In our relativistic interacting theory. which are eigenstates of the full Hamiltonian. we are going to introduce new counterterms in our theory µ2 1 L = (∂µ φ)2 − φ2 + ∂µ ψ ∗ ∂ µ ψ − m2 ψ ∗ ψ + f (t) −gψ ∗ ψφ + 2 2 + b 2 φ 2 meson mass counterterm a vacuum energy density counterterm + cψ ∗ ψ “nucleon” mass counterterm (10. the failure to match up one particle state energies in model 3 will yield T and ∆ dependent phases in k|S|k . If I prepare a packet of these free Hamiltonian eigenstates they propagate along with group velocity v= ∂E ∂p = p E when E 2 = p2 + m2 When I turn the interaction on (slowly. being proportional to the volume. When the interaction is off (f (t) = 0) this theory is a free theory with mesons of mass µ and nucleons of mass m. so that the free Hamiltonian eigenstate (bare nucleon) turns into the dressed nucleon (full Hamiltonian eigenstate)) the group velocity changes because the mass in E = p2 + m2 changes. we arrange.1) µ is the measured mass of the meson. m is the measured mass of the nucleon. When the interaction is turned on (f (t) = 1). I could set up a nucleon and meson to scatter. H: H|p P = p2 + m2 |p P There are also eigenstates of the Hamiltonian. they might not even come close! To fix up this problem.

10. The energy of two widely separated mesons. although they have very 2 little physical significance. m and µ. without any additional twiddling. to present your results. . they have got nothing to scatter (only vacuum) p |S| p This procedure should match up the energies of states of widely separated nucleons and mesons. and c are 0|S|0 = 1 ⇒ a one meson states q |S| q one nucleon states = δ (3) (q − q ) ⇒ b = δ (3) (p − p ) ⇒ c The one meson and one nucleon states shouldn’t do anything. in terms of m0 and µ0 too. Although the particles in model 3 never become separated from their own fields. respectively. b. If you treated m0 and µ0 as fundamental. bound state energy levels. you would calculate all your cross sections. even in model 3 when they are affected by self-interaction. the conditions determining a. our perturbation theory 0 0 is expressed in terms of the actual physical masses. You are always free to break up the free and interacting parts of the Hamiltonian any way you like. Instead of a perturbation theory for m2 . you would have to calculate µ and m. the physical masses. should be the sum of their respective energies. and them to make contact with reality. µ2 and all other physical quantities in terms of m2 and µ2 . they always become widely separated from each other. We have bypassed that mess by turning perturbation theory on its head. although you won’t get anywhere unless you can solve the free Hamiltonian. µ0 and m0 . Since no one is interested in your m0 and µ0 . Matching the energies of the vacuum and the one particle state matches the energy of states of widely separated mesons too. Sometimes it is useful to think about µ2 ≡ µ2 − b 0 and m2 ≡ m2 − c 0 the coefficients of 1 φ2 and ψ ∗ ψ in the full Lagrangian. This procedure gives us a BONUS. not the dumb quantities. By making b and c (hence µ2 and m2 ) quantities you compute. 0 0 we have a perturbation theory (for m2 and µ2 ) and all physical quantities in terms of the 0 0 observed masses. b We have put 2 φ2 and cψ ∗ ψ in with the interaction gψ ∗ ψφ because that way the mass of the meson (nucleon) is µ(m) when the interaction is of (manifestly) and the mass of the meson (nucleon) is µ(m) when the interaction is on (by our careful checks of b and c). What our procedure amounts to is breaking up the free and interacting parts of the Hamiltonian in a less naive way. all quantities of interest. you would have to reexpress all your cross sections in terms of µ and m. in terms of them. October 23 Notes from Sidney Coleman’s Physics 253a 102 To summarize. in the far past / future.

p2 = b† (p1 )b† (p2 )|0 b† (p) = (2π)3/2 2ωp bp So we don’t have to worry about Bose statistics demand p1 = p2 and p1 = p2 . it is not misleading to pull the g out in front of the whole interaction and talk about the O(g 0 ) contribution to S. Let’s look at p1 p2 | : ψ ∗ (x1 )ψ(x1 )ψ ∗ (x2 )ψ(x2 ) : |p1 p2 (11. The term in S with two factors of the interaction is (−ig)2 2! d4 x1 d4 x2 T [ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 )] After all the hoopla about the turning on and off function. That is the first order at which there is a contribution to p1 p2 |(S − 1)| p1 p2 two-nucleon states The −1 in S − 1 is there because we aren’t really interested in the no scattering process. Λp2 |p1 . October 28 Notes from Sidney Coleman’s Physics 253a 103 11 October 28 Feynman Diagrams in Model 3 0|(S − 1)|0 = p|(S − 1)|p = k|(S − 1)|k one-nucleon one-meson Let’s look at nucleon-nucleon scattering at O(g 2 ).1) . p2 = |Λp1 . we are abandoning it. which comes from 1.) There are no arrows over p1 . We can recover what we’ve lost by building wave packets concentrated around p1 = p2 . the O(g 0 ) term in S = T e−ig b c d4 x(ψ ∗ ψφ− 2g φ2 − g ψ ∗ ψ ) (Since the power series for b and c begin at order g 1 at the earliest (they are zero if g = 0).11. being sloppy: The only term in the Wick expansion of this term in S that can contribute to two nucleons goes to two nucleons is (−ig)2 2! d4 x1 d4 x2 : ψ ∗ ψ φ(x1 )ψ ∗ ψφ(x2 ) : φ(x1 )φ(x2 ) is some number. p1 = p1 and p2 = p2 or p1 = p2 and p2 = p1 . p2 . p1 and p2 because we are going to use the states that transform nicely under Lorentz transformations U (Λ)|p1 .

11. October 28

Notes from Sidney Coleman’s Physics 253a


The nucleon annihilation terms in ψ(x1 ) and ψ(x2 ) have to be used to annihilate two incoming nucleons. The nucleon creation terms in ψ ∗ (x1 ) and ψ ∗ (x2 ) have to be used to create two nucleons, so as not to get zero inner product (alternatively I could say they have to be used to “annihilate two nucleons on the left”). In equations p1 p2 | : ψ ∗ ψ(x1 )ψ ∗ ψ(x2 ) : |p1 p2 = p1 p2 |ψ ∗ (x1 )ψ ∗ (x2 )|0 0|ψ(x1 )ψ(x2 ) : |p1 p2 You can easily show that the two contributions to the second matrix element are (the c.c. equation is also used) 0|ψ(x1 )ψ(x2 )|p1 p2 = e−ip1 ·x1 −ip2 ·x2 + e−ip1 ·x2 −ip2 ·x1
↑ p1 absorbed at x1


So there are four contributions to our matrix element p1 p2 | : ψ ∗ ψ(x1 )ψ ∗ ψ(x2 ) : |p1 p2 = eip1 ·x1 +ip2 ·x2 + eip1 ·x2 +ip2 ·x1 eip1 ·x1 +ip2 ·x2 + eip1 ·x2 +ip2 ·x1 = eip1 ·x1 +ip2 ·x2 −ip1 ·x1 −ip2 ·x2 + eip1 ·x2 +ip2 ·x1 −ip1 ·x2 −ip2 ·x1 +e
F ip1 ·x2 +ip2 ·x1 −ip1 ·x1 −ip2 ·x2 I L ip1 ·x1 +ip2 ·x2 −ip1 ·x2 −ip2 ·x1 O



Notice that the pair of terms on the first line differ only by x1 ↔ x2 and that the pair of terms on the second line only differ by x1 ↔ x2 . Since x1 and x2 are to be integrated over and since φ(x1 )φ(x2 ) is symmetric under x1 ↔ x2 these pairs give identical contributions to 1 the matrix element. We’ll just write one of the pairs, which cancels the 2! . We have, (−ig)2 d4 x1 d4 x2 φ(x1 )φ(x2 ) eip1 ·x1 +ip2 ·x2 −ip1 ·x1 −ip2 ·x2 + eip1 ·x2 +ip2 ·x1 −ip1 ·x1 −ip2 ·x2 = (−ig)2 d4 x1 d4 x2 d4 k ix1 ·(p1 −p1 +k) ix2 ·(p2 −p2 −k) e e (2π)4 + eix1 ·(p2 −p1 +k) eix2 ·(p1 −p2 −k) k2 i − µ2 + i (11.4)

I have used the expression for φ(x1 )φ(x2 ) and grouped all the exponential factors by spacetime point. With these two integrals (there are two terms in the integrand) go two pictures

¦¦ ¦˜¦
p1 p1 p2 k↓ p2 (a)

N +N →N +N Feynman diagrams @ O(g 2 )

¦¦ ¦˜¦
p2 p1 p2 k↓ p1 (b)

Notice external lines each have an associated momentum. The vertices are not numbered.


11. October 28

Notes from Sidney Coleman’s Physics 253a


and two stories20 . The story that goes with picture (a) is this. A nucleon with momentum p1 comes in and interacts. Out of the interaction point comes a nucleon with momentum p1 and a “virtual meson.” This “virtual meson” then interacts with a nucleon with momentum p2 and out of the interaction point comes a nucleon with momentum p2 . The interaction points x1 and x2 can occur anywhere, and so they are integrated over. Furthermore, this “virtual meson” can have any momentum k and this is integrated over, although you can i see from the factor k2 −µ2 +i , “Feynman’s propagator” that it likes to be on the meson mass shell, although with k 0 = ± k 2 + µ2 . Fairy tales like this helped Feynman discover and think about quantum electrodynamics. In our formalism, they are little more than fairy tales, but in a formulation of quantum particle mechanics called the path integral formulation they gain some justification. The words not only match the pictures, they parallel the mathematics. The x1 and x2 integrations are easy to do. We get (ig)2 i d4 k (2π)4 δ (4) (p1 − p1 + k)(2π)4 δ (4) (p2 − p2 + k) 4 k 2 − µ2 + i (2π) + (2π)4 δ (4) (p2 − p1 + k)(2π)4 δ (4) (p1 − p2 + k)


Because the interaction is spacetime translationally invariant, after integrating the interaction point over all space-time we get delta functions which enforce energy-momentum conservation at every vertex. All the features of this computation generalize to more complicated S matrix element contributions. I’ll give a set of rules for writing down these integral expressions for contributions to S matrix elements. First, I’ll explain in general why there are no combinatoric factors in model 3 to worry about, no symmetry numbers in the integral expressions. Take a given operator in the Wick expansion, which has an associated Wick diagram D, a diagram ←→ : O(D) : n(D)!

Designate which of the lines leading out of the diagram annihilates each incoming particle, and which of the lines creates each outgoing particle. This is one contribution to the S matrix element. Now consider summing over the permutation of the numbered points in the Wick diagram. While only n(D)! of these permutations actually correspond to different terms in the Wick S(D) expansion, in model 3, all n(D)! of these permutations correspond to contributions to the S 1 matrix element. This cancels the n(D)! exactly.


in on right out on left

11. October 28

Notes from Sidney Coleman’s Physics 253a


There are other possible designations (in general) for the way the external lines connect to the vertices of the Wick diagram. If they differ just by a permutation of the vertices 1 then we have already counted them (by cancelling the n(D)! ). If they don’t differ by just a permutation of the vertices then they correspond to a different Feynman diagram (the difference between (a) and (b) in Eq. (11.5) Only in certain theories, like Model 3, do the n(D)! permutations of the vertices al make different contributions to the S matrix, and in fact this is true in model 3 only when a diagram21 has at least one external line. In that case there is an unambiguous way of identifying each vertex in the diagram. Contributions to 0|S|0 , which have no external lines, can have symmetry factors. The unambiguous labelling statement for diagram (a) is

¦¦ ¦˜¦
p1 p1 k↓ p2 p2

The upper vertex is uniquely labelled as the one where p1 is absorbed. The lower vertex is the one connected to the upper vertex by a muon line.

In this theory, as soon as one vertex is labelled (by an external line) they are all uniquely labelled.

Feynman Rules for Model 3
You should convince yourself by taking some other low order terms in the Wick expansion of S and looking at some simple matrix elements they contribute to that the following set of rules applied to the diagram always gives you the correct contribution to the S matrix element. incoming in For external lines momenta are directed . outgoing out

in fact, each connected part has to have at least one external line

11. October 28

Notes from Sidney Coleman’s Physics 253a


Assign a directed momentum to every internal line. For every

internal meson line

internal nucleon line

meson vacuum counterterm •
(2π)4 δ (4) (0) would turn into the volume of all spacetime if the system were in a box. This c.t. diagram is designed to cancel diagrams without external lines which you will see also have a factor of δ (4) (0) although the meson mass counterterm had a 1 in 2 the Lagrangian, there is no 1 here because there are 2 two ways to do the interaction

f ¦ ¦¦ d
←p ←p k

Write d4 k i (2π)4 k 2 − µ2 + i d4 p i 4 p2 − m 2 + i (2π) (−ig)(2π)4 δ (4) (p − p − k)

ia(2π)4 δ (4) (0)

meson mass counterterm

nucleon mass counterterm
←p ←p

fxf ¦x¦
←k ←k

ib(2π)4 δ (4) (k − k )

ic(2π)4 δ (4) (p − p )

A catalog of all Feynman diagrams in model 3 up to O(g 2 ) (except those related by C or T will not be written down twice) Order g (1)


¤f i –ywdf ¤f i ¤f i

= 0 if µ < 2m by energy momentum conservation.


Order g 2


= 02 if µ < 2m

11. October 28

Notes from Sidney Coleman’s Physics 253a





fxf n–f–n

↑ vacuum energy c.t. to O(g 2 )


             Because we demand there be no corrections to  0|S|0 these sum to zero. This fixes the vacuum   energy c.t. to O(g 2 )           

In 4(c) think of the c.t. as O(g 2 ). Its value is determined by the fact that it has to cancel some O(g 2 ) contributions to the vacuum-to-vacuum S matrix element. (a) (5) (b)

↑ nucleon mass c.t. to O(g 2 )

¦f—lnf¦ ¦x¦

      Sum to zero because we demand that  there are no corrections to p|S|p    one nucleon each   


(a) (6) (b)

–n ¦˜¦ f–v–f fxf

this diagram comes out zero for the same reason as (2) does.

   Sum to zero because we demand that   there are no corrections to k|S|k    one meson each 

The remaining order g 2 diagrams are more interesting. They contribute to the following processes ¯ ¯ ¯ ¯ (7) N + N −→ N + N (connected by C to N + N −→ N + N ) ¯ ¯ (8) N + N −→ N + N ¯ ¯ (9) N + φ −→ N + φ (connected by C to N + φ −→ N + φ) ¯ ¯ (10) N + N −→ φ + φ (connected by T to φ + φ −→ N + N )

11. October 28

Notes from Sidney Coleman’s Physics 253a


Although φ + φ −→ φ + φ appears in this theory, indeed it must appear, it does not do so until O(g 4 ). The diagram is

We have already written down the contributions to process (7). The diagrams are
p1 p1 p2 p1 k↓ p2 p2

dpe f¦¢d e ¦¦ ¦¦ ˜ ¦¦ ¦˜¦
k↓ p1 p2 (a) (b)


They give (see Eq. (11.6)) (−ig)2 i (p1 − p1 )2 − µ2 + i (2π)4 δ (4) (p1 + p2 − p1 − p2 ) + (−ig)2 i (p1 − p2 )2 − µ2 + i (2π)4 δ (4) (p1 + p2 − p1 − p2 )

1 Note that the (2π)4 associated with d4 k exactly cancels with the (2π)4 associated with the δ functions used to do the integral. All our formulas have been arranged so that 2π’s 1 always go with δ’s and 2π ’s always go with dk’s. Note that you can shortcut these trivial integrations over δ functions by just assigning internal momenta so as to conserve momentum whenever an internal momentum is determined by the other momenta at a vertex. Finally, note that performing the trivial integrals over δ functions always gives you a factor sum of all initial momenta

(2π)4 δ (4) ( pf − pi )
sum of all final momenta

at least when the diagram is of one connected piece

We define af i , the invariant Feynman amplitude by f |(S − 1)|i = iaf i δ (4) (pf − pi ) The factor of i is inserted to match the phase convention of NRQM. For N + N −→ N + N ia = (−ig)2 i (p1 − p1 )2 − µ2 +i + i (p1 − p2 )2 − µ2 + i

∆. by (p1 − p1 )2 = −∆2 (p1 − p2 )2 = −∆2 c In our COM variables ∆2 = 2p2 (1 − cos θ) The invariant Feynman amplitude is a = g2 1 1 + 2 ∆2 + µ2 ∆c + µ2 ∆2 = 2p2 (1 + cos θ) c We have dropped the i because it is unnecessary. Let’s try to understand our amplitude in NRQM. The probability had better have come out symmetrical. First we’ll find the NR analog of the first term. unit vector p1 p2 p3 p4 =( =( =( =( p2 + m2 . and at low energies they could describe scattering processes adequately with NRQM. θ = 0 is indistinguishable from a scattering angle of θ = π. p is more useful. −pe ) e · e = cos θ. which we will be taking. c The first term is peaked (peaked sharper at higher p) in the forward (θ ≈ 0) direction. Of course when identical particles collide who is to say what is forward and what is backward. p e ) p2 + m2 . ∆c . pe ) p2 + m2 . October 28 Notes from Sidney Coleman’s Physics 253a 110 Let’s look at this in the COM frame. ET = 2 p2 + m2 is often used to characterize collisions. and the crossed momentum transfer. The second term produces an identical peak in the backward direction. θ is the scattering angle in the COM frame. −pe) p2 + m2 . In the nonrelativistic limit.11. Define the momentum transfer. People were scattering nucleons off nucleons long before quantum field theory was around. For physically accessible values of ∆2 and ∆2 the denominators are never less than µ2 . .

11. r2 ∝ g e r Yukawa potential 2 −µr VE exchange Yukawa potential E|r1 . r2 The exchange Yukawa potential is the source of a simple pole in the amplitude at ∆2 = c −µ2 . In a partial wave expansion of the amplitude. The NRQM amplitude is proportional to ˜ ˜ V (∆)+ V (∆exch ) (∆exch = k + k). P. also in a physically unobservable region. r2 = |r2 . two body scattering is simplified to the problem of scattering of a potential (classically and quantum-mechanically). r1 E is the exchange operator r = |r1 − r2 | |r1 . corresponds to the Born approx. October 28 Notes from Sidney Coleman’s Physics 253a 111 In the COM frame. With two identical particles. the exchange identical even potential gives a contribution to the direct potential if l is . the Hamiltonian should contain an exchange potential H = H0 + V + V |r1 . which is characterized by having a simple pole in a physically unobservable region.T. at lowest order gives k |S − 1|k ∝ k |V |k = d3 rV (r)e−i∆·r “Born” approximation ∆=k −k ˜ = V (∆) To explain the first term in our scattering amplitude using NRQM we must have ˜ V (∆) ∝ 1 g 2 e−µr ˜ =⇒ V (r) ∝ ∆2 + µ2 r Our amplitude. at ∆2 = −µ2 . to the Yukawa interaction! The second term also has an analog in NRQM. opposite odd This is because in the COM an eigenstate of angular momentum is an eigenstate of the exchange operator E with eigenvalue (−1)2 . .

We’ve already found what the first term is by going to the COM frame and taking the NR limit. labelling the momenta arbitrarily and 2. From the diagram I write down ia = (−ig)2 i (p1 − p1 )2 − µ2 + i (p1 + p2 )2 − µ2 This has a less symmetric structure than than the amplitude for N + N → N + N . Bose statistics ¯ does not apply to the the incoming and outgoing states of N and N . It is a Yukawa potential. but that is not unexpected. October 30 Notes from Sidney Coleman’s Physics 253a 112 12 October 30 ¯ ¯ N + N → N + N “nucleon antinucleon” scattering p1 ← p1 − p 2 ↓ p2 p2 ← ← p2 ← p1 + p2 ← ← p1 p1 ← p1 ← ← p2 Notice that my labeling of internal lines has been done so as to be consistent with energymomentum conservation.12. performing the trivial integration over the arbitrarily labelled momenta that are actually fixed by δ functions. The symmetry of the amplitude for N + N → N + N was forced upon us because of identical particles in the incoming and outgoing states. This does away with the two steps of: 1. Now 1 1 1 1 = = 2 2 − µ2 2 (p1 + p2 ) ET − µ ET − µ ET + µ 1 1 2m + µ ET − µ . What about the second term? In the COM frame 2 (p1 + p2 )2 = 4 p2 + m2 = [2(p2 + m2 )]2 = ET 2 where ET the total energy in the COM frame.

What is the explanation of this in terms of NRQM? Let us suppose there is an energy eigenstate just below threshold. (Continuum states produce a branch cut in the amplitude). We’ll see this is because the intermediate state is spinless. Notice that this amplitude is independent of the scattering angle. i. We have not replaced ET by 2m in the second term because 2m could be very near µ. A partial wave decomposition would show a contribution only to the S wave. it may cause a significant contribution to the scattering amplitude in the second term of the Born expansion a ∝ f |V |i + n f |V |n n|V |i ET − EN ± i where En could be a possible source of an energy eigenstate pole. Two nucleons 2m Two nucleons at rest u One meson 0 Vacuum In the COM frame the energy spectrum of possible intermediate states looks like a state with En = µ and non vanishing matrix elements could produce a pole in the amplitude.12. A pole occurs in the partial wave that has the same angular momentum as the intermediate state. N + φ ⇒ N + φ meson-nucleon scattering ← q p q p ← q p ← ← p−q ← q p ← p+q . Then even in perturbation theory. This can cause a rapid variation in this factor. October 30 Notes from Sidney Coleman’s Physics 253a 113 in the NR limit. θ.e an energy eigenstate with an energy slightly less than 2m.

which is attractive in odd partial waves. Because the particles have different masses this potential is different from the ones we’ve had before. October 30 Notes from Sidney Coleman’s Physics 253a 114 The language that goes along with the second graph. Usually the s state is the most tightly bound (This is a classical expectation one has because the angular momentum barrier is the lowest). What kind of force could create a p wave state but no s wave state? A repulsive exchange force. then the incoming meson is absorbed. ia = (−ig)2 i i + (p + q)2 − m2 + i (p − q )2 − m2 + i This is an energy eigenstate pole and an exchange Yukawa potential. Usually one thinks of a resonance as caused by an attractive force that nearly creates a bound state. There is a resonance in the p wave of pion nucleon scattering called the N ∗ (12−− ). Of course this is no more kinematically possible for the intermediate nucleon than the first graph. pe) p = ( p2 + m2 .12. The second term in the amplitude −g 2 (p − q )2 1 − m2 + i unnecessary = g2 1 −(p − q )2 + m2 has a denominator in the COM frame of −(p − q )2 + m2 = −( p2 + m2 − p2 + µ2 )2 + p2 (1 + 2 cos θ) +m2 ∆2 c since p = ( p2 + m2 . pe ) . the crossed graph. There is nothing in the s wave. is that the outgoing meson is emitted first. Perhaps it is clear to see that this is an exchange Yukawa potential if we redraw the second graph as p ← p−q ↑ q ← ← q ←p Notice there is no direct Yukawa term.

In NRQM. It can have a long range at low energies if the mass µ is small. We’ll develop some formation which will be useful later to describe the connection. If one is bold. However we still need to develop spin and isotopic spin to really get things right (the sign of the potential for one) and the pionnucleon coupling is strong which means lowest order calculation can’t be trusted (except at long range or high partial waves). Our next topic is a discussion of the connection between Yukawa potentials. The second is an exchange Yukawa potential. i. q← p−q ↓ ←p q ← ← p ↓p−q q ← ←p q ← ←p The first graph is a Yukawa potential. In 1930. October 30 Notes from Sidney Coleman’s Physics 253a 115 q = ( p2 + µ2 . but what we would call the range is p dependent. there is absolutely no connection between these things. exchange Yukawa potentials and energy eigenstate poles in relativistic scattering theory. −pe ) This has the usual p2 (1 + 2 cos θ) exchange Yukawa forward peak. You can have any one (or two) without having all three. Note that as p → ∞ this → m2 and as p → 0 this → −(m − µ)2 + m2 = 2mµ − µ2 = µ(2m − µ). −pe) q = ( p2 + µ2 . we can start applying these ideas to real pion nucleon interactions.12. ¯ N + N → φ + φ ”nucleon-anti-nucleon” annihilation. .e energy dependent 1 = −( (range of pot)2 p2 + m2 − p2 + µ2 )2 + m2 The energy dependent part vanishes when µ2 = m2 . this was sensational.

12. Then who is to say this is not the amplitude for 3 + 4 → ¯ + ¯ reading left to right. i. with momentum p2 . we read graphs as if time flowed from right to left. which is useful (but not necessary) for discussing crossing. with momentum p3 . in analogy with the way we write down matrix elements. the particle of type ¯ is outgoing. its momentum is p1 . we have another method for fixing a convention. or 1 2. Now we are going to abandon that convention. Well. This you can tell because of our conthe particle of type 4 vention of putting incoming lines on the right and outgoing lines on the left. (or some contribution to the amplitude) by 3 1 4 2 where the numbers on the lines tell you what type of particles propagate along these lines (with the arrows). The particle of type 1 is incoming. 3 + 1 → ¯ + ¯ reading top to bottom (or even 1 → ¯ + ¯ + ¯ 2 4. and 3 ¯ is outgoing with momentum p4 . The honest to goodness physical momenta in the theory always are on the upper sheet of their mass hyperboloids. Imagine a general 2 → 2 scattering process 1 + 2 → ¯ + ¯ We’ll denote the amplitude 3 4. 2 3 4??). October 30 Notes from Sidney Coleman’s Physics 253a 116 Crossing [Symmetry] [Symmetry]: Brackets because this has nothing to do with symmetries and particles in the sense we have discussed them.e p2 = m2 3 3 or po 2 = m2 + |p3 |2 3 3 and p0 > 0 3 . and you aren’t supposed to worry about what type of interactions are hidden from view when the lines go behind the shield. the particle of type 2 is also incoming.

p0 and p0 are less that zero 3 1 and p0 and p0 are greater than zero. If you like you could think of the function as a function of four momenta. with p0 < 0. instead of graphically. in all cases.) Just to keep an air of symmetry we’ll display all four momenta in the function. the energy momentum conserving delta function is δ (4) (p1 + p2 + p3 + p4 ).] af i should be thought of as a function of parameters that parametrize the surface p2 = 1 2 p2 = p2 = p2 =. what this actually stands for is (what we used to mean 2 4 by) → −p1 1 2 ← p2 which is the amplitude for 2 + 4 ⇒ ¯ + ¯ Note that the notation has been set up. ia(p1 . and scatters into a particle of type ¯ with momentum −p1 . [But don’t. 1 Ignore everything written about QFT when it starts talking about negative energy particles. 3 independent momenta ⇒ 6 Lorentz invariants. an extra bit of information. p3 . And if. p4 ) This function is the amplitude for a particle of type 2 with momentum p2 and a particle of type 4 with momentum p4 .12. If someone gives you a momentum p. so that 1 3. a flag. what we have defined is a new function of these momenta. (p1 + p2 + p3 + p4 is restricted to be zero.2 . 4 constraints ⇒ 2 3 4 parameters. which is zero whenever p1 + p2 + p3 + p4 is not equal to zero23 . you know what they are really giving you is a physical momentum.22 Thus there will be no confusion if we flag a momentum by sending it to minus itself. p1 + p2 + p3 + p4 = 0. 23 22 → −p33 4 ← p4 . p2 . We will orient all momenta inward on our general 2 → 2 graph: → p3 3 1 ← p1 → p4 4 2 ← p2 The orientation on the page no longer matters. −p and a wink. say. Mathematically. October 30 Notes from Sidney Coleman’s Physics 253a 117 There are no negative energy states in our theories. We’ll use that extra bit of information to specify whether a particle is incoming or outgoing.

p0 .12. Landau rules 3. October 30 Notes from Sidney Coleman’s Physics 253a 118 ¯ and a particle of type 3 with momentum −p3 . Another way of writing the amplitude for 2 + 4 → ¯ + ¯ using this function is to take all the momenta. p2 . when p0 and p0 are > 0 and p0 and p0 are 3 1 4 2 < 0. p2 . macro? causality? 2. p4 ) There is no reason we can’t assemble the amplitude for all these different processes into a single process like this. 90 o 30o 1929 Food for thought (3 → 3 scattering) virtual nucleon [Tools for analysis of this limit: 1. p3 1 3 and p4 to their honest to goodness physical values (p0 . −p3 . Graphs with single poles] . but there is also no obvious reason it is any more useful that graphing the Dow Jones on the positive axis and the temperature in Miami on the same graph on the negative real axis. p0 and p0 > 0) and write 1 2 3 4 ia(−p1 . p1 . It is also the amplitude for a bunch of other processes when the time components of the three independent momenta take on their various possible signs.

there are only two relativistic invariants describing a 2 → 2 scattering process of spinless particles. Let’s define three relativistic invariants to describe 2 → 2 scattering processes. 1 = 2 and ¯ = ¯ and if 1 = ¯ or 2 = ¯ the choice is bassackwards. s ≡ (p1 + p2 )2 = (p3 + p4 )2 t ≡ (p1 + p3 )2 = (p2 + p4 )2 u ≡ (p1 + p4 )2 = (p2 + p3 )2 If particle 3 is outgoing. −t is the momentum transfer 3 4. 1 billion years A scattering in another galaxy One scattering There must be some appropriate limit where 2 → 2 followed by 2 → 2 is a limit of 3 → 3. They are often taken as the COM total energy and scattering angle. October 30 Notes from Sidney Coleman’s Physics 253a 119 Also think of this as 2 → 2 scattering followed 1 billion light years later by another 2 → 2 scattering. s. Now. squared. momentum transfer2 and −t the crossed momentum transfer2 . In all other cases. −p3 is its actual 4-momentum. They must be redundant. 2(s + t + u) = (p1 + p2 )2 + (p3 + p4 )2 + (p1 + p3 )2 + (p2 + p4 )2 + (p1 + p4 )2 + (p2 + p3 )2 4 =3 a=1 m2 + 2 a a>b pa · p b . That’s why you can get a vague description of virtual particles by thinking about them as real particles. t. I have made these last two choices arbitrarily. √ For the process 1 + 2 → ¯ + ¯ s is the total COM energy. and u are three relativistic invariants. Here is a (nice symmetric) derivation of their interdependence. If 3 4 3 4. If 1 = 2 and ¯ = ¯ and if 1 = ¯ or 2 = ¯ the choice is standard. I ought to call −u the 3 4 4 3.12. The virtual internal line must somehow become almost real. and −u is the crossed momentum transfer squared. anybody’s designation is arbitrary.

e s+t+u= a ma 2 There is a symmetrical way of graphing three variables in the plane. when they are restricted like this. Look at the plane in s − t − u space. s + t + u = a m2 a m2 a m2 a m2 a Looking down perpendicular at this plane. s. October 30 Notes from Sidney Coleman’s Physics 253a 120 Now use 0=( a pa )2 = a ma 2 + 2 a>b 4 p a · pb to see 2(s + t + u) = 2 m2 a a=1 i.12. t and u in the plane by 1 m2 s = r · es + ˆ a 3 a t = r · et + ˆ u = r · eu + ˆ 1 3 1 3 m2 a a m2 a a . you get the idea of representing.

12. Johnson. the process 1 + 2 → ¯ + ¯ is called the 3 4 s-channel. and the crossed process 1√ 3 → ¯ + ¯ and 1 + 4 → ¯ + ¯ are called the t-channel + 2 4 2 3 √ and u channel respectively because t and u are the total COM energy in these process. [MORE FOOD: The article “Uniqueness property of the Twofold Vacuum Expectation” by Paul G.] In an abuse of the scattering term “channel”. t. m2 = m2 . We have a ”Mandelstam-Kibble plot”. p0 < 0). p0 . 2. t < 0 is the physically accesible region for the process 1 + 2 → ¯ + ¯ and the process 3 4 ¯+¯ 3 + 4 → 1 2. the lowest order scattering amplitude for the process N + φ → N + φ was (using our new wacky conventions. Federbush and Kenneth A. the shaded area. 120. October 30 Notes from Sidney Coleman’s Physics 253a 121 ˆ eu unit vectors at 120◦ angles ˆ es et ˆ Each vector r in the plane gives you a triple s. Phys. 3 4 . u. 4. a = 1. Rev. In model 3. s > 4m2 . and since es + et + eu is obviously 0 ˆ ˆ ˆ 2 (rotational invariance) the set satisfies s + t + u = a ma . a u < 0. threshold the line s= m2 the line s=0 1 the line s= 3 m2 the line t=0 the line u=0 When all four masses are equal. 3. 1926 (1960) was attached at this point.

p2 → ↓ ← p1 p2 ← ← p1 ↓ ← p4 → ← p3 p4 ← p3 .12. ¯ What about the t channel process N + N → 2φ. 3 = N . October 30 Notes from Sidney Coleman’s Physics 253a 122 p4 → ← p1 ↓ p 1 + p4 + ← p4 → p3 ← p2 ← p1 + p2 p1 ← p3 → = (−ig)2 = (−ig)2 ← p2 )2 i i + 2+i 2 − m2 + i (p1 + p4 − m (p1 + p2 ) i i + 2+i u−m s − m2 + i exchange Yukawa interaction energy eigenstate pole The lowest order scattering amplitude for the corresponding u channel process (I’m thinking of 1 = N . 4 = φ) is N + φ → N + φ ← p2 p3 → p4 ← p1 + p2 p1 ← = (−ig)2 ← p2 → p3 ← ← p4 ← p1 + p2 p1 ← i i + 2+i u−m s − m2 + i energy eigenstate pole exchange Yukawa How about that: the amplitude are the same although the interpretation of the two are different. 2 = φ.

it may look to you like the analytic continuation can’t be performed even for the simple O(g 2 ) amplitudes we have discussed because the poles in s and u show up as lines which cut off one physical region from another. . 8?). it doesn’t matter how you go around them. you get the same analytic continuation. From our picture of the s − t − u plane. The second amplitude is only defined when u > 4m2 and s and t are less than zero. October 30 Notes from Sidney Coleman’s Physics 253a 123 = (−ig)2 i i + 2+i s−m u − m2 + i Again the amplitude is the same although the interpretation is different. but at higher orders they will have branch cuts. t. That is: the first amplitude is only defined in the shaded region u=0 2 pole at u=m t=0 pole at s=m2 s=0 (for simplicity take µ = m) where s.) Every one of these amplitudes is the exact same function of s. do you get the correct amplitude? The answer is yes and you do get the amplitude for one physical process by analytically continuing the amplitude for another. the COM energy squared. (More on this March 19. This is wrong because you can go around these poles by letting the variables become complex. they are avoidable singularities. we get the amplitude for that other process. but you must follow specific prescriptions when going around the essential singularities. so can the analytic continuation from one physical region to another still be performed and if so. and u. (I’d rather not ¯ assign a NR interpretation to the two graphs because non-relativistically N + N → 2φ can’t occur. Furthermore.12. What we have observed is that if we analytically continue an amplitude for some process outside of its physical region to the physical region of some other process. is greater than 4m2 and t and u are less than zero. that is. This brings up a tougher question: at this order in perturbation theory our amplitude just have poles.

The problem is that even if you have an excellent approximation to an analytic function if you analytically continue the approximation a long ways you may get something that doesn’t remotely resemble the analytic function. In the u channel process that same pole looks like an exchange Yukawa potential. c → ∞. Consider ex on the real axis for x < −1 million. The next thing to ask is how do we lose the relationship when we take the nonrelativistic limit. October 30 Notes from Sidney Coleman’s Physics 253a 124 Given this relation between amplitudes for different processes. But now analytically continue this wonderful approximation to x = +1 million. . As c → ∞. They are two aspects of the same analytic function restricted to two disconnected region of the plane.12. These effects are one and the same. Yukawa interactions and exchange Yukawa interaction. An example will suffice. Although c this has been illustrated only for 2 → 2 scattering it applies to any process. It is in the way that the connection between different amplitudes is lost when you take nonrelativistic limit and chuck those teensie but important terms of O( v )2 and higher. In that region 0 is a wonderful approximation to the function. and you discover you have completely missed the boat. the three physical regions on the Mandelstam plot which are separated by a distance of of O(mc2 ) get very far apart. we have related energy eigenstate poles. a pole in s in an s channel process looks like an energy eigenstate pole. mc2 The other thing we do in the NR limit is chuck terms of order (v/c)2 which of course is an arbitrarily good approximation in this limit. three things which had no connection in nonrelativistic quantum mechanics. They are only near each other for finite c.

. . n + m. There is no need to do any analytic continuation to show that the amplitude for these two processes are the same. . a = 1. so I don’t . a = 1. a = 1. 4 and that all the Feynman rules are quadratic in the momenta so there is no way this operation can change the amplitude. . n. October 30 Notes from Sidney Coleman’s Physics 253a 125 CPT symmetry There are three scattering processes related to 1 + 2 → ¯ + ¯ by various crossings.12. The argument applies to any process n particles → m particles by the same argument. . . This is the CP T theorem. . . Although we haven’t discussed theories without parity invariance. This is an argument to all orders in perturbation theory that even in a parity violating theory the amplitude for these two processes are equal. . if there is any grace in the world. This equality has nothing to do with analytic continuation. All you have to do is note that they are related by pa → −pa . . The Feynman rules are invariant under pa → −pa . n + m. . parity violating interactions will involve an tensor contracted with four momenta and that too is an even power of momenta. . it is invariant under pa → −pa . if n+m p n+m 1 p1 2 p2 p n+1 n+1 n pn is calculated in a Lorentz invariant theory. . Graphically. Why is it called the CP T theorem? I’ll only explain why in the 2 → 2 case. . They are 3 4 3+4→¯+¯ 1 2 2+4→¯+¯ 1 3 2+3→¯+¯ 1 4 Note that the physical region for 3+4 → ¯ ¯ is in the same region as that of 1+2 → ¯ ¯ 1+ 2 3+ 4. . n + 1. .

1 2 In a charge conjugation invariant theory this process would have the same amplitude but in general it doesn’t. They extend throughout all of space.12. Our S matrix elements are proportional to δ (4) (pf − pi ). Now let’s consider the time reversed process that would be 3+4→¯+¯ 1 2. p2 . because if you run a movie backward the products of a reaction become the reagents and the reagents become the products. CP T . −p2 . Phase space and the S matrix Our job is to make contact with the numbers experimenters measure. What went wrong? What went wrong is that the states we are using are not normalizable. The scattering process occurs at every point in space. Lagrangian quantum field theory is cooked. p4 ) Whether or not these three operations individually affect the amplitude. October 30 Notes from Sidney Coleman’s Physics 253a 126 have to invent some notation. Furthermore. we have shown above that the combined effect of all three operations. The most general 2 → 2 process is 1+2→ ¯+4 3 ¯ p1 p2 p3 p 4 physical 4 momenta amplitude = a(p1 . Squaring them is senseless. p3 . the velocities are reversed. and since two . what once went north now goes south and what once went up now goes down. and we get a probability that a counter will advance. −p4 ) If I charge conjugate the incoming and outgoing states. To do this we square our S matrix elements and integrate over the possible final states a detector might register. Contrast: If C violation is observed. we just write down C non-invariant interactions. can’t change the amplitude on general grounds. If we also apply parity we undo the reversal of velocities and the final process is 3+4→ ¯+¯ 1 2 p3 p4 p1 p 2 physical 4 momenta amplitude = a(−p1 . that is. −p3 . I get a related process ¯ + ¯ → 3 + 4. If CP T violated is ever observed.

.z 3 ki. a† ] = δk. A more satisfying way to salvage the situation is to build wave normalizable. . L ky dots represent allowed k values 2 L kx The states |k1 .z = . We are in a the world in a box of volume V . what is the probabilities of making a transition to some infinitesimal volume of phase space specified by d3 k1 · · · d3 kn (n final particles). October 30 Notes from Sidney Coleman’s Physics 253a 127 plane wave states never get far apart no matter how long you wait. . kn are built up from the vacuum by creation operators |k1 . so we’ll put The states in a box of volume V with periodic boundary condition are |k1 . It is the probability of . .x.x. . so that we a finite amount of packets.y. . kn = a† · · · a† |0 k k 1 n ak |0 = 0 [a† . . kn where 2πni. a box. a† ] = 0 k k [ak .y. L =V. the scattering process goes on for all time. A half-assed way to salvage the situation is to put the systems in can normalize the plane-wave states and to turn the interaction on for time T . . which are hurry. ak ] = 0 [ak . . and do get far apart in the far past / future.12. . . .k k Kronecker delta = k|k The free field in the box has the expansion φ(x) = k ak e−ik·x √ + 2Ep V a† eik·x k √ 2Ep V We want to know.

so √ 1 the probability that they will scatter from anywhere is ∝ V . ak . Why the factor V ? Without it. we expect f |S − 1|i = iaViT (2π)4 δV T (pi − pf ) f final particles (4) 1 √ 2Ef V √ initial particles 1 √ 2Ei V where (2π)4 δV T (p) ≡ V (4) d3 x dtf (t)eip·x The extra factors you have never seen before come from the expansion of the fields. k2 V Decay Scattering √ |k and |k1 . and the other as having probability one of being somewhere in the box. With these conventions. We will normalize the two particle initial state unconventionally. k2 are box normalized. With the factor V you can think of one particle as having probability 1 of being in any unit volume. d3 k1 d3 kn V ··· V (2π)3 (2π)3 Let’s look at the transition probability to go to one of the final states in that region | f |(S − 1)|i |2 . V 2ωk You get one of these factors for every particle that is annihilated on the left or right by the fields. We will restrict our attention to the two simplest and most important initial states: one particle and two particle initial states. October 30 Notes from Sidney Coleman’s Physics 253a 128 going to one of the final states in that infinitesimal region times the number of states in that infinitesimal region. Of course. each particle has probability 1 of being somewhere in the box. The coefficient of the creation operator a† . The probability that they are both near a given point and can scatter is ∝ V12 . they could both be near any point in the box. which annihilate a particle on the right is 1 √ .12. We will consider |i = |k √ |k1 . which annihilate a particle on the left and the k coefficient of the annihilation operator. We did not get these before because the coefficient of a(k) and a(k)† in the 1 1 field is which exactly cancels the factor we get when a(k) hits the relativistically (2π)3 2ωk normalized state |k a(k)|k = (2π)3 2ωk |0 .

You should get Differential Transition Probability = Unit time 1 VT 2 (4) |af i | (2π)4 δV T (pi − pf ) VT Now we take the limits V . To get something that does not depend on the time we turn the interaction on for T . October 30 Notes from Sidney Coleman’s Physics 253a 129 The product over initial states in our formula has had the single factor of √1V pulled out √ in the decay case.12. aViT −→ af i f |aViT |2 −→ |af i |2 f δV T −→ δ (4) ((2π)4 δV T )2 → Here’s where we have to be careful Recall: δV T is a function concentrated near the origin. It becomes more and more so as V . Also it is normalized to 1 for all V . it is concentrated about p = 0 just as surely as δV T (p) is. T → ∞. T . We can find its normalization 1 (4) dt dt f (t)f (t ) d3 x d3 x d4 peip·x eip·x d4 p[δV T (p)]2 = 8 (2π) V V (2π)4 δ (4) (x+x ) (4) (4) (4) = 1 (2π)4 dt|f (t)|2 V d3 x = 1 VT (2π)4 . T → ∞. To get the transition probability. and the two factors of √1V cancelled to just one by the V we put in by hand in the scattering case. ∞ 1 (4) d4 p δV T (p) = d3 x d4 peip·x = 1 dtf (t) (2π)4 −∞ V (2π)4 δ (4) (0) (4) (4) (4) 2 d3 kf (2V )3 2Ef final particles 1 2Ei initial particles (1 or 2) For these two reasons we say limV T →∞ δV T (p) = δ (4) (p). we divide by T . We multiply this by the numbers of states in the infinitesimal region of final state phase space to get the differential transition probability. leaving again a single factor of √1V which we have explicitly put in as the last factor of the formula. What about (δV T (p))2 ? In the limit V. T → ∞. we square the transition amplitude.

Every 2π goes with a δ 1 function and every 2π goes with a k integration. T → ∞ limit to get differential transition probability = unit time |af i |2 (2π)4 δ (4) (pf − pi ) d3 kf (2π)3 2Ef final particles 1 2Ei initial particles. Note that you have no excuse for not getting the (2π)’s right. 1 or 2 This factor which is manifestly Lorentz invariant is called the “invariant density of states”.12. .T →∞ V T lim Thank God a factor of V1T appears in our formula for the differential transition probability per unit time so we can take the V . D. or the ”relativistic density of final states”. October 30 Notes from Sidney Coleman’s Physics 253a 130 For these two reasons we say 1 (4) (2π)4 (δV T (p))2 = δ (4) (p) V.

The shelf life of a moving particle is longer.1) Decay Diff decay prob 1 = |af i |2 D Unit time 2E The total decay probability per unit time is dΓ summed and integrated over all possible final states 1 Decay Probability = |a|2 D Unit time 2E dΓ = final states It is obvious that this part is L.13. November 4 Notes from Sidney Coleman’s Physics 253a 131 13 November 4 Differential Transition Probability = |af i |2 D Unit time 1 2Ei Applications of initial particles (1 or 2) D = (2π)4 δ 4 (pf − pi ) final particles d3 kf (2π)3 2Ef (13. This is the “decay width”.I. We’ll evaluate the decay probability rate in the rest frame of the decaying particle. Γ= Since the |a|2 D is L.2) Cross sections dσ = Diff Trans Prob 1 1 = |af i |2 D Unit time × Unit flux 4E1 E2 |v1 − v2 | These v’s are initial state particle velocities and energies . Decay probability m dτ = Γ= Γ Unit time E dt where τ is the particle’s proper time. 1 Rest decay probability = Unit time 2m |a|2 D (13.I. Γ. its decay rate is slower exactly by a factor of elapsed proper time / elapsed observer time.

so as to catch the most flux. The cylinder is the volume swept out in the beam in a time t. k2 . The orientation of A is so as to be ⊥ to v2 − v1 . v1 t is the motion of the beam in time t. k2 Our convention The factor The transition probability per unit time is some mess t. Its volume is |v2 − v1 |tA The flux is thus |v2 − v1 | and the (some mess) t.v1 t v2 t is the vector displacement of the particle with momentum k2 in a time t.p = u.p 1 = ( some mess ) u.t. that is. Then the transition probability per unit time would have come out as t. v 1 v 2 A v t 2 A vt 1 A (v2.t. Let’s understand this factor with our |v1 − v2 | conventions. November 4 Notes from Sidney Coleman’s Physics 253a 132 1 takes care of the per unit flux. × unit flux |v2 − v1 | Another Convention is to take |i = |k1 . Think of the particle with momentum k1 as having probability 1 of being in any volume and the particle with momentum k2 as being located somewhere in the whole box with probability one.13. A to the particle beam with momentum k1 . V The flux for this normalization is however 1 |v1 − v2 | V This is the same “(some mess)” as in the previous equation. = (some mess) u. Let’s suppose the particle with momentum k2 presents some area. whatever it is ( .t. √ |i = V |k1 .p.

The underbraced term is invariant under any Lorentz transformation. November 4 Notes from Sidney Coleman’s Physics 253a 133 So is the same √ 1 ¡ (some mess) t. P1 = (E1 . This justifies the interpretation of the cross section as an area. It will be proportional to the flux in that frame. It should be unaffected by these boosts. However the idea of a perpendicular area should be Lorentz invariant for boosts along v1 − v2 direction. p1x . V ¡ = 1 Unit time × Unit flux ¡ |v1 − v2 | V ¡ .p.3) If v1 is parallel or antiparallel to v2 and the total cross section really has the interpretation of an area. . 0) Then E1 E2 |v1 − v2 | = E1 E2 p1x p2x − = |p1x E2 − p2x E1 | = | E1 E2 23µν p1µ p2ν | which is obviously invariant under rotations on the 0 − 1 plane (boosts along v1 ). The total cross section is σ= 1 1 |v1 − v2 | 4E1 E2 |af i |2 D final states (13. 0. I want to emphasize that this is the nonrelativistic velocity and NR velocity addition formula that appears here.13. p2x . A v 1 v / v2 1/ v 2 The total number of particles that smash into the area A will depend on the velocity of the observer. If two beams approach with speed c head on the flux is 2c. What about the factor in front? Take v1 and v2 to be along the x direction. Since σ= 1 1 |v1 − v2 | 4E1 E2 |af i |2 D is supposed to have the interpretation of an area. 0) P2 = (E2 . 0. then it should be unaffected by boosts along v1 and v2 .

E1 = p1 2 + m2 . 4Ei1 Ei2 |v1 − v2 | = 4|Ei2 pi1 − Ei1 pi2 | = 4|Ei2 pi1 + Ei1 pi1 | = 4ET pi1 = 4ET pi (13.4) 2 → 2 scattering in the COM frame In the COM frame. We can do them once and for all.13. as p2 dp1 dΩ1 and use the energy delta function to do the p1 integration.5) . E2 dE2 = p2 dp2 . we have 1 E1 dE1 = p1 dp1 . November 4 Notes from Sidney Coleman’s Physics 253a 134 D for a two body final state in the COM frame D contains both δ functions and integrals that can be trivially performed by using those δ functions. You must now remember that p2 depends on p1 whenever it appear (in E2 or in |af i |2 ). Let’s rewrite d3 p1 . and E2 = p2 2 + m2 = p1 2 + m2 . ∂(E1 + E2 ) p1 p1 p1 E T = + = ∂p1 E1 E2 E1 E2 So D= 1 dΩ1 2E E 16π 1 2 p2 1 ∂(E1 +E2 ) ∂p1 = 1 p1 dΩ1 16π 2 ET (13. Of course this turns independent variables (in |af i |2 and in where they arise kinematically) into dependent variables. In the center of mass frame pi = 0 and E i = ET while D= d3 p1 d3 p2 (2π)4 δ (3)(p1 + p2 ) δ(E1 + E2 − ET ) 6 4E E (2π) 1 2 final particle energies and momenta = d3 p 1 2πδ(E1 + E2 − ET ) (2π)3 4E1 E2 p2 = −p1 The δ (3) (p1 + p2 ) is used to do the p2 integration. 1 2 2 Thought of as functions of p1 .

Compare this with from NRQM and see 1 |af i | 8πET We’ll get the phase when we do the optical theorem. the amount of time the two particles spend 1 in the danger zone near each other. It is simple to understand this.13. This why they slow down the reaction in atomic piles. which goes as pi becomes infinite. dΩ and hence σ can be infinite even when the amplitude af i is finite. As pi → 0. Contact with elastic 2 → 2 scattering in NRQM Our formula for dσ dΩ is 1 pf dσ 1 = |af i |2 |af i |2 = 2 2 dΩ 64π 2 ET pi 64π 2 ET dσ = |f |2 dΩ for elastic scattering. November 4 Notes from Sidney Coleman’s Physics 253a 135 (The i subscript reminds you that these are initial particle momenta and energies) dσ = 1 pf dΩ1 1 1 pf |af i |2 = dΩ1 |af i |2 2 E 2E 2 16π 64π T p1 T 4ET pi dσ 1 pf = |af i |2 2E 2 p dΩ 64π T i dσ Note that for an exothermic reaction we can have pi = 0 while pf = 0. We maximize the chance of neutron capture in the pile by making the neutron cruise out of the pile as slowly as possible. Model 3 p ¤f i q p ¤f i L = −gψ ∗ φφ = −ig = (−ig)(2π)4 δ (4) (p + p − q) . |f | = Example.

the total cross section. so we’ll rephrase the conservation of probability as (S − 1)(S − 1)† = SS † −S − S † + 1 = −(S − 1) − (S − 1)† 1 Now f |(S − 1)|i = iaf i (2π)4 δ 4 (pf − pi ) . so we should be able to get a analog of the optical theorem in our relativistic scattering theory. 136 pretty simple. The outgoing wave is the superposition of the incoming wave that passes right through the target and goes off in the forward direction.13. which mathematically is σ. af i . (13.4) Γ= 1 |a|2 D 2µ p1 dΩ1 1 2 = g 2µ 16π 2 ET µ 2 2 = g p1 1 g ( µ)2 − m2 = 2 2 8πµ 8πµ 2 2 g µ2 − 4m2 = 16πµ2 Optical Theorem The optical theorem in NRQM is based on a simple idea. From Eqs. which are proportional to matrix elements of S − 1. There is nothing in this argument that is nonrelativistic. The total probability for scattering in all directions but exactly forward. couldn’t be simpler. November 4 Notes from Sidney Coleman’s Physics 253a ia = −ig + O(g 3 ). there must be a decrease in the intensity in the beam in the forward direction. There is an incoming wave incident on a target. and since probability is conserved. and the scattered wave which goes off in all directions. Since there is some probability that a particle in the beam is scattered off. which mathematically is an interference term between the wave that passes right through and the scattered wave in the forward direction. must be equal to the decrease in probability of going exactly in the forward direction. The mathematical statement of conservation of probability in the scattering process is SS † = 1 We want to make a statement about our amplitudes.2) and (13. and an outgoing wave.

1)). The RHS of the rephrased statement of the conservation of probability is RHS = −iaf i (2π)4 δ 4 (pf − pi ) + ia∗ (2π)4 δ 4 (pf − pi ) if Both the LHS and RHS are proportional to (2π)4 δ 4 (pf − pi ). we get an overcounting factor if the nm particles are identical intermediate states with nm particles 1 nm ! Dm |aim |2 = 2Im aii . Dm (see Eq. so that we explicitly have that LHS is proportional to δ 4 (pf − pi ). we can replace the pm in δ 4 (pf −pm ) by pi . November 4 Notes from Sidney Coleman’s Physics 253a 137 and f |(S − 1)† |i = −ia∗ (2π)4 δ 4 (pf − pi ) if Our rephrased statement of the conservation of probability has matrix elements of f |(S − 1)(S − 1)† |i = − f |(S − 1)|i − f |(S − 1)† |i By inserting a complete set of intermediate states. (13. we have 1 nm ! intermediate states with nm particles d3 k1 d3 knm (2π)4 δ 4 (pm − pi ) af m a∗ = −iaf i + ia∗ ··· im if (2π)3 2E1 (2π)3 2Enm = 2Im af i The underbraced factor is what we would call the invariant density of states for the process i → m.13. If we choose f = i. the left hand side (LHS) becomes LHS = f |(S − 1)|m m|(S − 1)† |i intermediate states |m an overcounting factor if the nm particles are identical = intermediate states with nm particles 1 nm ! d3 k1 d3 knm af m a∗ (2π)4 δ 4 (pf − pm )(2π)4 δ 4 (pm − pi ) ··· im (2π)3 2E1 (2π)3 2Enm Because of the δ 4 (pm −pi ). Comparing the LHS with the RHS.

From now on p3 and E3 are not independent variables. p 2 p 1 12 . Instead of writing d3 p2 as p2 dp2 dΩ2 . d3 p2 = p2 dp2 dφ12 d cos θ12 . In NRQM the optical theorem for elastic scattering (pi = pf = p) is p σ = Im f |θ=0 4π Barring a different θ dependence in the phase conventions. This proves that the forward scattering amplitude for 2N → 2N is real at O(g 2 )). In the COM frame this says (see Eqs.13. p3 = −(p1 + p2 ) E3 = (p1 + p2 )2 + m2 3 Now we’ll rewrite d3 p1 as p2 dp1 dΩ1 . we can finally state f= 1 a 8πET 3 body final state phase space in the COM frame D= d3 p1 d3 p2 d3 p3 (2π)4 δ (3) (p1 + p2 + p3 ) · δ(E1 + E2 + E3 − ET ) (2π)3 2E1 (2π)3 2E2 (2π)3 2E3 1 3 3 1 = d p1 d p2 δ(E1 + E2 + E3 − ET ) 5 (2π) 8E1 E2 E3 The momentum conserving δ function has been used to eliminate p3 .5)) 4ET pi σ = 2Im aii ¡ ¡ 2 (since the LHS is zero till O(g 4 ) for 2N → 2N scattering.3) and (13. If the process has a two particle initial state. November 4 Notes from Sidney Coleman’s Physics 253a 138 2 This says the total transition probability (statement might be off by a factor of ET ) per unit time is equal to twice the imaginary part of the forward scattering amplitude. we see that the RHS must be zero till O(g 4 ). we can rewrite this as a statement about cross sections. 1 2 2 where φ12 is an azimuthal angle about p1 and θ12 is a polar angle measured from p1 and θ12 is a polar angle measure from p1 . (13.

E3 . because dE1 dE2 is the Euclidean measure on the plane E2 E1 There will be a kinematically imposed boundary to the sprinkling of data. if you average over initial spin states. or for a particle with spin decaying. Our next topic is the beginning of a discussion of Green’s functions. November 4 Notes from Sidney Coleman’s Physics 253a 139 We are going to use the energy conserving δ function to do the θ12 integration.13. If I make a plot of experimental data points as a function of E1 and E2 . E3 depends on θ12 . then we can do the angular integrations (which give 8π 2 ) to get 1 |a|2 dE1 dE2 32π 3 as the differential transition probability per unit time into an energy range E1 for particle 1 and E2 for particle 2. but within those boundaries. they will be distributed according to |a|2 . scattering with wave packets. 2 E3 = p2 + 2p1 · p2 + p2 + m2 = p2 + p2 + m2 + 2p1 p2 cos θ12 1 2 3 1 2 3 ∂E 1 Therefore ∂ cos 3 12 = pEp2 and thus d cos θ12 δ(E1 + E2 + E3 + E4 ) = θ 3 θ12 is now a dependent variable. p1 p2 D= E3 1 2 1 p1 dp1 dΩ1 p2 dp2 dφ12 2 (2π)5 p1 p2 8E1 E2 E3 1 p1 dp1 p2 dp2 = dΩ1 dφ12 (Valid in COM frame) 256π 5 E1 E2 dE1 dE2 (Amazing simple result if you use the right variable) Suppose the amplitude a is independent of Ω1 and φ12 . and the LSZ reduction formula. . the sprinkling of data is directly proportional to |a|2 . as in the decay of a spinless meson (at rest).

If I really wanted to specify parity effect. or whatever. Something neat is that they all agree. . . I just mean uncharged. November 4 Notes from Sidney Coleman’s Physics 253a 140 Feynman Diagrams With External Lines Off The Mass Shell We’ll restrict ourselves (for notational simplicity only) to considering diagrams in which only one type of scalar meson appears on the external lines. I would say “scalar under parity transformations”. with no Lorentz indices on its field. no spin. .)) We’ll still let particles of all types run around on the internal lines. Let a blob like this ˜ ≡ G(4) (k1 . maybe not even satisfying k1 + k2 + k3 + k4 = 0? We are going to come up with three affirmative answers to this question. A parity need not even exist for the formalism we are about to develop to be applicable (“charged scalar” means charged. but no spin. (By scalar. or in our imagination the sum of all diagrams to all orders in perturbation theory. k4 ) represent a sum of Feynman diagrams. not a specification of its parity transformation properties. It could be all Feynman diagrams to some order in perturbation theory. Answer 1 The blob could be an internal part of a more complicated graph . off the mass shell. that is.13. Can we assign any meaning to this blob if the momenta on the external lines are unrestricted. .

for example all of which have the form . in our work calculating those graphs to some order. Then it would be nice not to repeat that work when calculating it would be nice to just plug the result in from a table of blobs.13. that we have already summed the blob. . . November 4 Notes from Sidney Coleman’s Physics 253a 141 k2 k3 k4 (The Feynman rules instruct you to label all internal momenta arbitrarily and integrate over them.) Suppose in our study of other graphs in the theory.

. A big graph that contains G(2) (k1 . even useful. k2 ) to 2 O(g ). which enforces k1 = ˜ −k2 . ˜ Here’s a simple example. ←k2 k1 fd‡dkd‡df k1 → These correspond to contributions to . We still have a couple of conventional choices to make in defining a blob. We’ll include it all. then tell us exactly how to define the blob. k1 → ←k2 and k1 +l→ To order g 2 we have k1 → ←k2 ˜ ≡ G(2) (k1 . We will define it to be what it would be if it were an internal part of a bigger graph (or a sum of internal parts in bigger graphs). Our Feynman rules for the bigger graph. . kn ) and we could include or not include the overall energy momentum conserving δ function. .13. k2 ) is k2 k1 More explicitly various contributions of this type are k2 k2 k2 k1 . fpf fkf fd‡df and k1 ←l k1 → ←k2 of k1 → ←k2 . We could ˜ include or not include the n propagators that hang off G(n) (k1 . there is some ambiguity in the way to write down the contributions to G(2) (k1 . November 4 Notes from Sidney Coleman’s Physics 253a 142 So we have one sensible. which has its external lines on the mass shell. It could just as well have been written i i + 2 (2π) δ (k1 +k2 ) 2 2+i k2 − µ k2 − µ2 + i 4 (4) 2 d4 l i i 4 (−k + l)2 − m2 + i l 2 − m2 + i (2π) 2 . definition of the blob. k2 ) = (2π)4 δ (4) (k1 + k2 ) + 2 k1 fpf i − µ2 + i d4 l i i i i 2 2 (2π)4 (k1 + l)2 − m2 + i l2 − m2 + i k1 − µ2 + i k2 − µ2 + i Because of the overall energy momentum conserving delta function. .

k2 = 2 kr − µ2 ˜ (4) G (−k1 . One thing we can do with these blobs is to recover S matrix elements. the graphs that we wrote out above do not contribute.2. k4 ) = ˜ (4) = fkf fkf fkf fkf fkf fkf k1 → k2 → ←k4 ←k3 k1 → k2 → ←k3 ←k4 k1 → ←k2 + + + O(g 2 ) k3 → ←k4 = (2π)4 δ (4) (k1 +k4 ) 2 k1 i i (2π)4 δ (4) (k2 +k3 ) 2 +2 permutations+O(g 2 ) 2+i −µ k2 − µ2 + i Since the second δ function enforces k2 = −k3 we can rewrite the first δ function as δ (4) (k1 + k2 + k3 + k4 ) if you like.13. k2 |(S − 1)|k1 . . to display over all energy momentum conservation explicitly. k2 ) (*) LSZ reduction formula i r=1. they should not contribute to S − 1. . . We cancel off the external propagator and put the momenta back on their mass shells to cancel the external ˜ propagator we had included in G k1 .1 . k1 . Indeed. −k2 . November 4 Notes from Sidney Coleman’s Physics 253a 143 We can also write down a few contributions to G (k1 .2 Because of the four factors of zero out front when the momenta are on mass shell. . .

21) (The power theorem is We have G(n) (x1 .3) if you don’t know or remember how to get this Feynman rule) Answer 1 (cont’d) We have found one meaning for our blob. . .14. . its Fourier transform. . . kn ) (2π)4 (2π)4 d4 xf (x)g(x) = d4 k ˜ f (k)˜(−k)) g (2π)4 . .1)-(14. xn ) = d4 k1 d4 kn ik1 ·x1 +···+ikn ·xn ˜ (n) ··· e G (k1 . . . Using the Fourier transform convention f (x) = ˜ f (k) = d4 k ˜ f (k)eik·x (2π)4 d4 xf (x)e−ik·x (Which you’ll notice has a different sign in the exponent from what we used on Oct. A source with positive frequencies creates particles while a source with negative frequencies absorbs particles. We can use it to obtain another function. This is summed up in the Feynman rule • f →k ⇐⇒ i˜(−k) ρ (See Eqs. November 6 Notes from Sidney Coleman’s Physics 253a 144 14 November 6 d4 k ˜ f (k)eik·x (2π)4 Fourier transform (convention of Nov. 6): f (x) = This is a little unfortunate because e−iEt+ik·x (E > 0) ∂ ∂ is generally called a positive frequency plane wave (because i ∂t acting on it give E and 1 ∂x i ˜ acting on it gives k) and thus if f (k) has support for positive k 0 . f (x) negative frequency. (14.

kn ) = This is the sum of all (let’s make this definite) Feynman diagrams to all orders (if the blob does not exist then the blob represents a formal power series in g) and the blob includes factors for the external propagators and the factor for overall energy momentum conservation (2π)4 δ (4) (k1 + · · · + kn ) Answer 2 Consider modifying H. November 6 Notes from Sidney Coleman’s Physics 253a 145 where ˜ G(n) (k1 . There are now contributions to 0|S|0 proportional to ρn at low orders in g. . .14. say in model 3. H → H − ρ(x)φ(x) (L → L + ρ(x)φ(x)) where ρ(x) is a specified c number source. . . This adds a new vertex. not an operator. . The modification of the Hamiltonian (density) H → H − ρ(x)φ(x) spoils this. At order f ρ and order g. these are not interesting simple examples. At order ρ2 and order g 0 we have • f •. because unless ρ(0) is nonzero ˜ they vanish because of energy-momentum conservation. Let’s suppose we have got the original Hamiltonian’s vacuum counterterm all calculated out to some high order in perturbation theory so that there are no corrections to 0|S|0 to this high order. we have • fd‡d At order ρ and O(g 3 ) we have Unfortunately. a model 1 type vertex ρ • →k ⇐⇒ i˜(−k) The new Feynman rule was just quoted in class. let’s see it arise in a simple example.

November 6 Notes from Sidney Coleman’s Physics 253a 146 At order ρ and order g we have and 2 2 • y#dfdy# f • fdy#df • . . They are ψ ∗ ψ φ(x1 )ψ ∗ ψφ(x2 ) φ(x3 )φ(x4 )     ψ ∗ ψ φ(x1 )ψ ∗ ψφ(x2 )φ(x3 )φ(x4 )    ∗ ∗ ψ ψ φ(x1 )ψ ψφ(x2 )φ(x3 )φ(x4 ) ←→ y#dfdy# • • • f • ←→ fdy#d fdy#d f f • ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 ) φ(x3 )φ(x4 )      ∗ ∗ ψ ψφ(x1 )ψ ψφ(x2 )φ(x3 )φ(x4 )      ∗ ∗ ψ ψφ(x1 )ψ ψφ(x2 )φ(x3 )φ(x4 ) ←→ • flx¦ • ←→ • • fdy#df • The last two are the ones I want to look at in detail. so we are looking for the completely contracted terms in the Wick expansion of the time ordered product.e. This is a nice simple example. −∞) = T e−i i.14. It comes from the term second order in ρ and second order in g in S = UI (∞. and since these are dummy variables of integration they together make a . They differ by an exchange of x1 ↔ x2 only. as well as • • • fdy#d f fdy#d flx¦ • • . let’s look at it. d4 x(gψ ∗ ψφ−ρφ) (−ig)2 (i)2 d4 x1 d4 x2 d4 x3 d4 x4 ρ(x3 )ρ(x4 )T (ψ ∗ ψφ(x1 )ψ ∗ ψφ(x2 )φ(x3 )φ(x4 )) 2! 2! The process is vacuum → vacuum.

(11. corresponding to space-time point x1 (but I hate to label it as such because it is just a dummy integration variable which is going to be integrated over. and creating a meson with momentum k.2) in the lecture of Oct.2) to Eq. and in our combinatoric arguments for Feynman diagrams. (11. we have i i i i d4 p d4 q d4 k d4 l (2π)4 (2π)4 (2π)4 (2π)4 p2 − m2 + i q 2 − m2 + i k 2 − m2 + i l2 − m2 + i × (−ig)2 (i)2 ρ(k)˜(l)(2π)4 δ (4) (p + q + k)(2π)4 δ (4) (−p − q + l) ˜ ρ . 28.14. creating an antinucleon with momentum q. The x integrals are easy to perform. you’ll see that when we have a factor eix1 ·(p+q+k) that corresponds to a picture with p. (11. q and k flowing out of x1 . November 6 Notes from Sidney Coleman’s Physics 253a 147 contribution to 0|S|0 of (−ig)2 (i)2 2! d4 x1 d4 x2 d4 x3 d4 x4 ρ(x3 )ρ(x4 ) ψ ∗ (x1 )ψ(x2 ) ψ ∗ (x1 )ψ(x2 ) φ(x1 )φ(x3 ) φ(x2 )φ(x4 ) (1) (2) (3) (4) (1) : (2) : (3) : (4) : We have d4 p ip·(x1 −x2 ) i e 4 2 − m2 + i (2π) p d4 q iq·(x1 −x2 ) i e (2π)4 q 2 − m2 + i d4 k ik·(x1 −x3 ) i e (2π)4 k 2 − m2 + i d4 l il·(x2 −x4 ) i e 4 2 − m2 + i (2π) l i i i d4 p d4 q d4 k d4 l i (2π)4 (2π)4 (2π)4 (2π)4 p2 − m2 + i q 2 − m2 + i k 2 − m2 + i l2 − m2 + i × (−ig)2 (i)2 d4 x1 d4 x2 d4 x3 d4 x4 ρ(x3 )ρ(x4 )eix1 ·(p+q+k) eix2 ·(−p−q+l) eix3 ·k eix4 ·l Now if you go back to Eq. we have kept the vertices unlabelled) is creating a nucleon with momentum p.5) and especially Eq. Our picture for the integral at hand is • fd‡df →q ←k →l →p • The left vertex.

−∞)|k = 0| I + i =0+• ←k f d4 xρ(x)φ(x) |k + · · · + ··· (14. If you know how the Feynman rule comes out. no external lines. and it still only applies to diagrams where each connected part has at least one external line. What we have here is a vacuum to vacuum diagram. (14. (11. zeroth order in g) contribution to ˜ 0|S|k = 0|UI (∞.6) of lecture on Oct. which I’ll explain in a moment.14. and we now have to worry about symmetry factors. 28).) But.2) f ←k =i d4 xρ(x)e−ik·x = i˜(k) ρ (14. Suppose we have a graph in the Wick expansion with n powers of ρφ and m powers of 1 1 gψ ∗ ψφ. November 6 Notes from Sidney Coleman’s Physics 253a 148 This is just what you would have directly written down using our old Feynman rules supplanted by ρ • ←k ⇐⇒ i˜(k) 1 except for that out front. which comes with a n! m! from the exponential. we sung and danced about how there were no symmetry factors in model 3.1) (|k is relativistically normalized and I’ll use the relativistically normalized creation and annihilation operators in the expansion for φ(x).4)). 2! f This was a moderately interesting graph to show how the Feynman rule comes out. 0|φ(x)|k = So • d3 k e−ik·x 0|a(k )|k = e−ik·x (2π)3 2ωk (2π)3 2ωk δ (3) (k−k ) (14. but you want to check whether it’s ρ(k) or ˜ ρ(−k) just look at the lowest order (in ρ. That argument still goes through. Earlier (after Eq.3) 1 Now for that 2! out front (see also the short argument after Eq. 8 7 6 5 4 3 3' 1' 2' 2 4' 5' 9 1 .

To see the accounting work. but also 2 ↔ 3. because that has already been counted as 1 ↔ 2 . November 6 Notes from Sidney Coleman’s Physics 253a 149 (In this example n = 9 and m = 5) Each of the m! permutations of the model 3 vertices. Some of them make no new contributions to the Wick expansion. For example 5 ↔ 6 and 8 ↔ 9. is a new term in the Wick expansion. Any of the n! permutations that do make a new contribution to the Wick expansion. that have not already been counted in the permutations of the m model 3 vertices. which is a contribution to 0|S|0 with the source off. k7 k6 k5 k8 k4 k3 k9 k1 k2 1 9! out front consider it as Of course when the momenta are integrated over.14. There are no model 3 vertices that are not somehow attached to a source vertex. that is. are accounted for in another way. this is identical to the integral above. That would be a disconnected bubble. look at the messy example in momentum space. 7 ↔ 8 or 6 ↔ 8 gives a new term in the Wick expansion. the vacuum energy counterterm has been adjusted so that there are no corrections to 0|S|0 with the source off. Now what about the n! permutations of source vertices. ˜ Now both of these would be counted in G(k1 . For example. and by assumption. kq ) and in fact every permutation of the 9 source vertices that leads to a new term in the Wick expansion corresponds to a diagram . They are all uniquely identified by the way they are attached to the source vertices. It is (i)9 9! d4 k9 d4 k1 ··· ρ(−k1 ) · · · ρ(−k9 )× ˜ ˜ k6 (2π)4 (2π)4 k5 k8 k7 k4 k3 Feynman diagram with external lines off the mass shell k9 k1 k2 Instead of using the permutation 7 ↔ 8 to partially cancel off the the same mess m with a new diagram in the integrand. keeping the source vertices fixed. · · · .

kq ) for this too k1 k2 k5 k3 k7 k4 k6 k8 k9 Let denote the sum of all diagrams to all orders in g (i. Notes from Sidney Coleman’s Physics 253a 150 Of course there are diagrams with n = 9 and m = 5 that are not of the same pattern as the one drawn (differ by more than a permutation of vertices). . all m) and at nth order in ρ that contribute to 0|S|0 .14. For example 1 2 5 3 7 1 2 3 4 6 8 5 4 9 ˜ There is a Feynman diagram in G(k1 . November 6 ˜ in G(k1 . . k2 . kq ). . · · · . . .e.

kn . . . Now we integrate over all momenta k1 . . . That contribution to 0|S|0 would be (i)n d4 kn d4 k1 ˜ ··· ρ(−k1 ) · · · ρ(−kn )G(k1 . I’ll try to make a shorter argument. kn ) ˜ ˜ 4 4 (2π) (2π) d4 x1 d4 x2 d4 x3 d4 x4 ρ(x1 ) · · · ρ(xn )G(n) (x1 . and in ˜ ˜ doing so we make an overcounting by n! . while ρ(x1 ) · · · ρ(xn ) comes with coefficient n! . . . xn ) =1+ n=1 . . kn ) ˜ ˜ (2π)4 (2π)4 (14. . . . . . To all orders ∞ 0|S|0 = 1 + n=1 ∞ (i)n n! (i)n n! d4 k1 d4 kn ˜ ··· ρ(−k1 ) · · · ρ(−kn )G(n) (k1 . temporarily think of it as a first order calculation in each of n different sources ρ1 (x). which are distinguishable by virtue of the fact that they all carry different momenta k1 . . November 6 Notes from Sidney Coleman’s Physics 253a 151 What these combinatoric arguments say is (i)n = n! d4 k1 d4 kn ··· ρ(−k1 ) · · · ρ(−kn )× ˜ ˜ (2π)4 (2π)4 = (i)n n! d4 k1 d4 kn ˜ ··· ρ(−k1 ) · · · ρ(−kn )G(k1 . There is no overcounting when you integrate over all momenta. in the exponential ρ1 (x1 ) · · · ρ1 (xn ) ˜ ˜ 1 comes with coefficient 1. . They interact in all possible ways.14. . That gives us ˜ (i)n ρ(−k1 ) · · · ρ(−kn )G(k1 . . . Then the diagram where source 1 creates a particle with momentum k1 and source 2 creates a particle with momentum k2 really is distinguishable from a diagram where source 1 creates k2 and source 2 creates k1 . kn ). . kn ) ˜ ˜ (2π)4 (2π)4 How does this imagined calculation differ from ours? Well. . Instead of considering this as an nth order calculation in ρ. . . . . . . . . BEST ARGUMENT One last way of arguing this.4) Having gone all the way back to Wick expansion arguments to show the combinatorics are right for this. . ρn (x). . . . . kn . The source creates n mesons.

we could just say “function”. You give me a function on spacetime. and hence all S matrix elements) about the system is coded in the vacuum persistence amplitude in the presence of an external source ρ. These are called functional derivatives. . Ex nihil omnes: All physical information (all Green’s functions. 0|S|0 ρ . and I give you back a number. . Green introduced the first Green’s function in the early 19th century. φ(x). we have δ n Z[ρ] δρ(x1 ) · · · δρ(xn ) = (i)n G(n) (x1 . 0|S|0 ρ comes up often enough. 0|S|0 ρ is a functional of ρ.14. Actually. it gets a name. From conservation of probability alone. x )ρ(x ) 1 |x − x | (E = − φ) G(x. and the nomenclature “functional” is redundant. Z[ρ] Z[ρ] ≡ 0|S|0 The square brackets remind you that this is a function of a function ρ ρ Z[ρ] is called a generating functional for the Green’s functions because in the infinite dimensional generalization of a Taylor series. φ(x) = d3 x G(x. From a prescribed charge distribution. Mathematicians don’t call a vector in an infinite dimensional space “vectoral”. holding a (4 − d) continuum of other variables fixed. the value of the source at each spacetime point. In a theory with linear response only G(1) = 0. . x ) = satisfies 2 φ=− · E = −4πρ Let’s explicitly note that the vacuum to vacuum transition amplitude depends on ρ by writing 0|S|0 ρ (don’t confuse the subscript ρ with a p). ρ(x). . xn ) ρ=0 The δ instead of a ∂ reminds you that you are taking a partial derivative of z with respect to ρ(x). The Fourier transform of the sum of Feynman diagrams with n external lines off the mass shell is a Green’s function (that’s what G stands for). . his Green’s function gave you the electrostatic potential. you can see that the response of a quantum mechanical vertex can’t be linear. November 6 Notes from Sidney Coleman’s Physics 253a 152 This is the second answer to our question. it is just a function of an infinite number of variables. ρ(x).

. which are a statement about Green’s functions resulting from a symmetry. Take the vacuum expectation value of both sides. We have Z[ρ] = 0| : e connected diagrams : |0 = e 0| connected diagrams |0 This is true because the terms in the sum in the exponential are themselves normal ordered. . . xn ) is a VEV (“Vacuum Expectation Value”) of a time ordered string of Heisenberg fields. By a cunning trick we will show that G(n) (x1 . kn possibly off shell. ln Z[ρ] = 0| ∞ connected diagrams |0 (i)n n! d4 kn d4 k1 ˜ ··· ρ(−k1 ) · · · ρ(−kn )Gc (k1 . An example will be the Ward identities. . Taking the natural logarithm. xn ). kn ) ˜ ˜ (2π)4 (2π)4 = n=1 ˜ Gc is the sum of all connected Feynman diagrams. which blow up on mass shell. including the overall energy momentum conserving δ function. Apply this to a model which has had a source added. . you can break the Hamiltonian up into a “free” and interacting part in any way you please. As we did in obtaining answer 2. with k1 . . Z[ρ]. November 6 Notes from Sidney Coleman’s Physics 253a 153 Z[ρ] is called a generating functional in analogy with the functions of two variables which when you Taylor expand in one variable. let H → H − ρφ(x) H0 + H → H0 + H − ρφ(x) As far as Dyson’s formula is concerned. .e. . Let’s take the “free” part to be H0 + H and . . .14. . Answer 3 One more way of interpreting G(n) (x1 . Because of our great theorem all Wick diagrams = : e connected diagrams : The sums are sums of normal ordered terms. Sometimes it is very useful to put a whole set of functions in one neat package like that. convince yourself. . . and the external propagator. the coefficients are a set of functions like the Legendre polynomials in the other. . . The LHS is just 0|S|0 ρ . . i. It can be put very compactly in term of Z.

14. I put quotes around “free”. They do not obey the free field equations of motion. Its Fourier transform is the VEV of a time ordered string of Heisenberg fiels. This can all be taken as motivation. . the fields evolve according to φ(x. and thus you can’t do Wick’s theorem. 3. Its Fourier transform (times (i) ) is the coefficient of the nth order term in ρ in the n! expansion of the vacuum to vacuum persistence amplitude in the presence of a source. H = d3 xH (the turning on and off function is gone for good). n . because we are going to start from scratch and do a REFORMULATION OF SCATTERING THEORY No more turning on and off function Imagine you have a well-defined theory. 1. You can’t define a contraction for these fields. e−iHt H= d3 xH H = H0 + H These fields are not free. For this reason we’ll subscript them with an H. . . 0). 2. . ρ. Let’s see what this tells us about Z[ρ]. with a time independent Hamiltonian. whose spectrum is bounded below. November 6 Notes from Sidney Coleman’s Physics 253a 154 the interaction to be −ρφ(x). we have found three meanings for the (sum of all) Feynman diagrams with (n) external lines off the mass shell. xn ) = 0|T (φH (x1 ) · · · φH (xn ))|0 (14.5) To summarize. because in this new interaction picture. t) = eiHt φ(x. These fields are what we would have called Heisenberg fields if there was no source. Z[ρ] = 0|S|0 ∞ ρ = 0|T e+i (i)n n! d4 xρ(x)φH (x) |0 just expand =1+ n=1 d4 x1 d4 x2 d4 x3 d4 x4 ρ(x1 ) · · · ρ(xn ) 0|T (φH (x1 ) · · · φH (xn ))|0 and we read off G(n) (x1 . It is a handy blob we can plaster into the interior of a larger diagram.

(14.T. .2 2 ka − µ2 ˜ G(−k1 . November 6 Notes from Sidney Coleman’s Physics 253a 155 whose lowest lying state. k2 |(S − 1)|k1 . The question is: Is G(n) = GF ? or equivalently. .1 . gives G(n) (x1 . is Z = ZF ? Answer will be “yes. satisfies don’t confuse p with ρ H|0 p =0 The vacuum is translationally invariant and normalized to one p|0 p = 0 and p 0|0 p =1 Now let H → H − ρ(x)φ(x) and define Z[ρ] ≡ = p p 0|S|0 p in the presence of the source ρ p 0| U (−∞. . in the exact same way as we did in Eq. ∞) |0 Schr¨dinger picture evolution operator o for the Hamiltonian d3 x(H−ρφ) and define G(n) (x1 .6) . |0 p .” Answer to question 1: Is G(n) = GF (n)? Using Dyson’s formula. is k1 . xn ) = p 0|T (φH (x1 ) · · · φH (xn ))|0 p (14. Is G(n) defined this way (the F. and the Hamiltonian has actually been adjusted so that this state. the physical vacuum. k2 . Are S − 1 matrix elements obtained from Green’s functions in the same way as before? For example. .5). k2 = a=1. k2 ) ? i Answer will be “almost. .) of the sum of all Feynman graphs? Let’s call the (n) G(n) defined as the sum of all Feynman graphs GF and the Z which generated those (n) ZF . k1 . .” 2. .2. is not part of a continuum. .14. xn ) = Two Questions: 1 δ n Z[ρ] in δρ(x1 ) · · · δρ(xn ) 1.

xn ) we do n functional derivatives w. t1 )φI (x1 )UI (t1 . . . so it suffices to prove they are equal for one ordering which for convenience we choose so that or for short t1 > t2 > · · · > tn x0 > x0 > · · · > x0 n 1 2 The time ordering in the expression for G(n) is just lexicographic ordering G(n) (x1 . xn ) = lim 0|T e −i t+ t− d4 x[φI (x1 )···φI (xn )HI ] −i t+ t− |0 t± →±∞ 0|T e d4 x H I |0 We have got a little work to do to show this is the same as G(n) in Eq. November 6 Notes from Sidney Coleman’s Physics 253a 156 Does ZF [ρ]. . (14. t− )|0 0|U (t+ . . .6). xn ) = = Using the standard shorthand for e−i (n) for GF is GF (x1 .14.r. . ρ and then set ρ = 0 (and divide by in ) (n) GF (x1 . we adjusted the vacuum energy counterterm. both these expressions are manifestly symmetric under the n! permutations of the x1 . you divide by the same thing with ρ = 0. . . Let HI = H (φI ). ta ). xn . . The thing which after Wick’s theorem and a combinatoric argument or two had a graphical expansion is ZF [ρ] = lim t± →±∞ Base eigenstate of H0 H0 |0 =0 0| T e −i t+ t− d4 x[HI −ρφI ] |0 We used to adjust the constant part of HI to eliminate vacuum bubbles in our old scattering theory when ρ = 0. . Fortunately. . . .e. .. . the generating functional you get by blindly summing graphs generate the same Green’s functions? H splits up into H0 + H . i. so that the vacuum to vacuum graphs (with no source vertices) summed to zero. . . . . You divide out of ZF [ρ] the sum of all vacuum to vacuum graphs with no source vertices explicitly.t. and then you don’t have to worry about a vacuum energy c. . t2 )φI (x2 ) · · · φI (xn )UI (tn . ZF [ρ] = lim (n) 0|T e −i t+ t− d4 x[HI −ρφI ] t+ t− |0 t± →±∞ 0|T e −i d4 xHI |0 To get GF (x1 . xn ) = lim (n) t± →±∞ tb ta p p 0|T (φH (x1 ) · · · φH (xn ))|0 0|φH (x1 ) · · · φH (xn )|0 p p d4 xHI = UI (tb . t− )|0 . That is. There is an equivalent way of throwing away the vacuum bubbles.t. the time ordering in the expression 0|UI (t+ .

The integral is a continuous function that oscillates more and more wildly as t− → −∞. 0)U (0. .1) = insert a complete set t− →−∞ lim φ| U (0. t− ) |0 Schr¨dinger picture o = lim t− →−∞ φ|U (0. xn ) = lim (n) t± →±∞ 0|UI (t+ . tb ) and then use φH (xi ) = UI (ti . A similar argument shows lim 0|UI (t+ . but if I wrote more it would take just as long to figure out what I was saying as it would take to convince yourself. t− ) |0 pp 0| + |n n| |0 all other eigenstates of the full Hamiltonian. t− )|ψ = φ|0 p p 0|ψ . Convince yourself.14. t− )|0 Considering the whole mess sandwiched with UI (0. t− )|0 = lim ∗see below t− →−∞ φ|UI (0. ti )φI (xi )UI (ti . 24 . In the limit it integrates to zero. t0 ) is acting on. t− ) eiH0 t− |0 fancy way of inserting 1 Using an easily derivable relationship between the evolution operator in the various pictures. let’s work on t− →−∞ lim φ|UI (0. Now every state but the vacuum is part of a continuum. As long as φ|n n|0 is a continuous function. November 6 Notes from Sidney Coleman’s Physics 253a 157 at least in the limt± →±∞ when t+ > t1 > · · · > tn > t− . . What we are showing is that for two arbitrary fixed states limt− →−∞ φ|U (0. rewrite it as UI (ta . t− )|0 0|U (t+ . 0) = UI (0. H|n =En |n = φ|0 p p 0|0 + lim t− →−∞ eiEn t− φ|n n|0 all other eigenstates *: You see it doesn’t really matter from this point on that it is a bare vacuum that U (0. 0)φH (x1 )φH (x2 ) · · · φH (xn )UI (0.24 Everywhere UI (ta . . tb ) appears. t− )|0 in the numerator or denominator as some fixed state φ|. 0)UI (0. 16. . Oct. (8. 0)|ψ = 0|0 p p 0|ψ t+ →∞ Physically what this theorem about oscillation integrands (the Riemann-Lebesgue lemma) says is that if you look at a state in some fixed region (take its inner product with some I usually put “convince yourself” when I haven’t written enough to make something clear. Eq. the limit vanishes. H H|0 p =0. 0)† φI (xi )UI (ti . 0) to get GF (x1 .

P µ |0 = 0 0|0 = 0 = 1 We will assume there are physical one meson states. kr = 2 la − µ2 i a=1. Question 2: Are S − 1 matrix elements obtained from Green’s functions in the same way as before? By introducing a turning on and off function. . November 6 Notes from Sidney Coleman’s Physics 253a 158 fixed state φ|) and you wait long enough.14. xn ) and there is no longer any reason to distinguish between them.. −ls . . Is this formula right? The answer is “almost. . .. . . ls |(S − 1)|k1 ..” We will show how to obtain S−1 matrix elements from Green’s function without resorting to perturbation theory.r The real world does not have a turning on and off function. We will make no reference to free Hamiltonia. xn ) = ¨ ¨ $ ¨ p 0|0¨ p 0|0 p$$$ p 0|0 ¨ = G(n) (x1 . relativistically normalized. Accordingly. i. .s 2 kb − µ2 ˜ (r+s) G (−l1 .. . .. . All the one and multi-particle states will have run away... . interaction picture fields. . normalizable. . . take φ(x) ≡ φH (x) |0 ≡ |0 p |0 is the ground state of the full Hamiltonian which as usual we assume to be translationally invariant and not part of a continuum. . . . ¨ $ ¨ p 0|0¨ p 0|φH (x1 ) · · · φH (xn )|0 p$$$ p 0|0 (n) GF (x1 . etc. k1 ..e. bare vacua. . we were able to show that l1 . . H|k = k 2 + µ2 |k P |k = k|k k |k = (2π)3 2ωk δ (3) (k − k ) . . kr ) i b=1. . . the only trace of it that will remain is its (true) vacuum component. . |k . .

φ in terms of φ. 0|φ(x)|0 may not be zero. kr = 2 la − µ2 i a=1. then S − 1 matrix elements are given by l1 .. G (n) (x1 . it may have a VEV. . −ls . We define a new field φ which has zero VEV and is normalized to have a standard amplitude to create one meson φ (x) = Z3 2 (φ(x) − 0|φ(0)|0 ) 0|φ (x)|0 = 0 k|φ (x)|0 = eik·x −1 LSZ formula stated Define the renormalized Green’s functions.. However this VEV is independent of x by translational invariance. φ. We correct for these things by defining a renormalized field. 0|φ(x)|0 = 0|eiP ·x φ(0)e−iP ·x |0 = 0|φ(0)|0 We also have by translational invariance k|φ(x)|0 = k|eiP ·x φ(0)e−iP ·x |0 = eik·x k|φ(0)|0 By Lorentz invariance.14. . kr ) i b=1. . you can easily see that k|φ(0)|0 is independent of k. . .r . their Fourier transforms. . . . . ls |(S − 1)|k1 . More precisely. November 6 Notes from Sidney Coleman’s Physics 253a 159 The reason that the answer to question 2 is “almost” is because the field. . . . xn ) ≡ 0|T (φ (x1 ) · · · φ(xn ))|0 ˜ and G (n) . . Z32 (traditionally called the “wave function renormalization”). . k1 . . in general it is not normalized so as to create a one particle state from the vacuum with a standard amplitude. It is normalized to obey the canonical commutation relations. ... . in general = 1.. . . . . . which enters the formula for S−1 matrix elements through G(n) (G(n) (x1 . 1 Z32 ≡ k|φ(0)|0 which we hope is nonzero. ... and second. . does not have quite the right properties.s 2 kb − µ2 ˜ (r+s) G (−l1 . xn ) = 0|T (φ(x1 ) · · · φ(xn ))|0 ). It is some 1 number.. G (n) . First.

The proof of the LSZ reduction formula is as follows. its relationship to φ(x).e. the field that appears in the Lagrangian with a standard kinetic term. The method will be inspired by the way a limiting process gave us the physical vacuum when we started with the bare vacuum. Any field that satisfies these properties. Once we know how to make one meson states we’ll wave our arms some and describe how to get many meson in and out states. whose Green’s functions you have. November 6 Notes from Sidney Coleman’s Physics 253a 160 This is the Lehmann-Symanzik-Zimmermann reduction formula. First we’ll find a way of making one meson wave packets. d3 k F (k)e−ik·x f (x) ≡ (2π)3 2ωk (2 + µ2 )f (x) = 0 Note that as |f → |k (i. The only assumptions needed about the total scalar field φ is that it satisfy 0|φ (x)|0 and k|φ (x)|0 = eik·x In particular.-G. equation.) associated with each of these wave packets. Then we’ll be set to get S matrix elements in terms of the Green’s function of the renormalized fields that were used to create the in and out states. LSZ reduction formula proof A notation for normalizable wave packet states |f ≡ d3 k F (k)|k (2π)3 2ωk (You can recover F (k) from |f : F (k) = k|f . is not used. F (k ) → (2π)3 2ωk δ (3) (k − k )) f (x) → e−ik·x Define an operator which is a function of time only out of any operator which is a function of x and t (here taken to be φ ): φ f (t) = i d3 x(φ ∂0 f − f ∂0 φ ) .14. gives you the S matrix. we have a negative frequency solution of the K.

14. t) |0 d3 x ∂0 f − f ∂0 n|φ ( x. t) |0 (2π)3 2ωk d3 k F (k ) k| − iωk e−ik ·x − e−ik ·x ∂0 k|φ (x. We will now take the limit t → ±∞. n|φ f (t)|0 = n|i =i d3 x φ (x. and in this limit we’ll see that many more matrix elements of φ f (t) look like the matrix elements of a creation operator. in the limit t → ±∞. φ f (t) is acting like a creation operator for a physical meson wave packet. From the properties of φ (x) and f (x) 0|φ f (t)|0 = 0 k|φ f (t)|0 = i =i d3 x d3 x d3 x and d3 k F (k ) k| φ ∂0 e−ik ·x − e−ik ·x ∂0 φ (x.-G. It will turn out to create a particle in state |f . µ Consider any state with two or more particles satisfying P µ |n = Pn |n . t)|0 eiPn ·x n|φ ( 0)|0 =i =i =i 0 d3 x ∂0 f − f iPn eiPn ·x n|φ ( 0)|0 0 d3 x ∂0 − iPn d3 k F (k)e−ik·x eiPn ·x n|φ ( 0)|0 (2π)3 2ωk d3 xe−ik·x eiPn ·x n|φ ( 0)|0 0 −iω t+iPn t k d3 k 0 F (k)(−iωk − iPn ) (2π)3 2ωk (2π)3 δ (3) (k−Pn )·e 0 ωPn + Pn 0 F (Pn ) n|φ (0)|0 e−i(ωPn −Pn )t = 2ωPn . equation will not find this a strange combination. t)∂0 f − f ∂0 φ (x. t)|0 3 2ω (2π) k eik·x =i dk F (k )(−iωk − iωk )e−ik ·x+ik·x (2π)3 2ωk independent of time 3 = F (k) A similar derivation except for one crucial minus sign shows 0|φ f (t)|k = 0 In these few matrix elements. November 6 Notes from Sidney Coleman’s Physics 253a 161 Those of you who are familiar with the initial value theory of the K.

and send the time of reaction to −∞). November 6 Notes from Sidney Coleman’s Physics 253a 0 162 The important thing to notice is that this matrix element contains e−i(ωPn −Pn )t and that 0 ωPn < Pn for any multiparticle state. A two particle state with P = 0 can have any energy from 2µ to ∞. Now consider ψ|φ f (t)|0 in the limit t → ±∞ where |ψ is a definite (not varying with t) normalizable state. The one particle state with P = 0 has energy ωP = µ. Physically what we have shown is this: We have created a state which in part looks like a one meson wave packet. If we look at this state in some definite region in space time (take its inner product with some definite state |ψ . . No multiparticle state has the right dispersion relation to keep the same relationship to the observer under this modification.    lim ψ|φ f (t)|0 = lim ψ| |0 0| +  t→±∞ t→±∞  d3 k |k k| + |n n| φ f (t)|0  (2π)3 2ωk multiparticle states|n =0+ d3 k ψ|k F (k) + lim t→±∞ (2π)3 2ωk ψ|n |n 0 F (Pn )(ωPn + Pn ) 2ωPn = ψ|f + 0 The integrals over the various continua of multiparticle states have integrands which oscillate more and more wildly as t → ±∞. A multiparticle state with momentum Pn always has more energy than a single particle state with momentum Pn . all the multiparticle states will run away. Insert a complete set. We prevent this by modifying the state we create in such a way that the one meson packet always has the same relationship to the meson that’s the funny combination φ f (t). The analogous derivation showing limt→±∞ 0|φ f (t)|ψ = 0 goes through because the phases add (and thus obviously never cancel) for every momentum eigenstate. Of course the one particle state may run away too. The phases were arranged to cancel in the one particle state matrix elements only. They integrate to zero in the limit by the Riemann-Lebesgue lemma. plus a little multiparticle garbage.14. one or multiparticle in the inserted complete set.

For the creation of multiparticle in and out states we are going to have to do some vigorous handwaving. this is time ordered and thus we have succeeded in writing S matrix elements in terms of the renormalized Green’s functions. with only a little effort. This excludes scattering at threshold. e wait another thousand years. However there is still some concept of localization up to a few Compton wave1 lengths. This physical argument says lim t→+(−)∞ ψ|φ f2 (t)|f1 = ψ|f1 . we will get an expression that you can put plane wave states into without getting nonsense. f4 |S|f1 . there is no x operator. k2 .15. f2 out (in) By definition. In doing so. Thus there is no way to measure of define S matrix elements of plane wave states. like an application of φ f2 to the vacuum. is what tells you the probability amplitude that looks like a given state in the far past will look like another given state in the far future. f3 . but we are going to massage this expression. f4 |S|f1 . k4 |(S − 1)|k1 . You may be bothered that a position space picture doesn’t really exist. at least it can be made rigorous. f2 ≡ out f3 . Consider two normalizable wave packets described by F1 (k) and F2 (k) which have no common support in momentum space. k3 . Then the application of φ f2 (t) to |f1 is for all purposes to an observer near the f2 packet. However. we’ll extend the idea of an S matrix element. Physically there is no way to create plane wave states. We’ll make this the definition of S matrix elements of plane wave states. If a teensy exponential tail with 1 distance m is too big for you for some purpose. the S matrix. The utility of this object is that you . after we get done massaging the RHS of the above equation. Because F1 and F2 describe packets headed in different directions if you wait long enough (go back far enough) the two wave packets will be widely separated in position. So we could quit now. f2 lim in = lim lim t4 →∞ t3 →∞ t2 →−∞ t1 →−∞ lim 0|φ f3 † (t3 )φ f4 † (t4 )φ f2 (t2 )φ f1 (t1 )|0 Note that in the limits. November 13 Notes from Sidney Coleman’s Physics 253a 163 15 November 13 What we have done so far has been rigorous. What is lim t→+(−)∞ ψ|φ f2 (t)|f1 ? The handwaving requires we picture this in position space.

k2 . and f → 0 as |x| → ∞ and a . . x4 ) d4 l1 d4 l4 il1 ·x1 +···+il4 ·x4 ˜ = ··· e G (l1 . but it actually isn’t. This is almost exactly the result that came out of our low budget scattering theory. −k4 ) i This says that an S − 1 matrix element is equal to a Green’s function with the external propagators removed. k4 |(S − 1)|k1 . e−ik2 ·x2 by f2 (x2 ). smear it a little. The result we will first obtain won’t be an expression for k3 . satisfying (2 + µ2 )f (x) = 0. what we will show is f3 . k2 = d4 x1 · · · d4 x4 eik3 ·x3 +ik4 ·x4 −ik1 ·x1 −ik2 ·x2 × (i)4 r (2r + µ2 ) 0|T (φ (x1 ) · · · φ (x4 ))|0 That looks unfamiliar and messy. Recall that 0|T (φ (x1 ) · · · φ (x4 ))|0 ≡ G (x1 . Given any function. f (x). Now I’ll tell you the answer. beginning with a lemma. . k4 |(S − 1)|k1 . . f2 which looks just like the expression for k3 . e−ik3 ·x3 by f3 (x3 ) and e−ik4 ·x4 by f4 (x4 ). that is. to recover physically measurable S matrix elements. k4 |(S − 1)|k1 . k2 stated above except e−k1 ·x1 is replaced by f1 (x1 ). k2 . f4 |(S − 1)|f1 . k2 . . . with the only difference being that the Green’s function is of renormalized fields. what we will soon show is a sensible definition for k3 . We get that by abstracting the expression for f3 . l4 ) (2π)4 (2π)4 If you substitute this in the expression for k3 . . f4 |(S − 1)|f1 . k2 = r 2 kr − µ2 ˜ G (k1 .15. −k3 . November 13 Notes from Sidney Coleman’s Physics 253a 164 can integrate it. k4 |(S − 1)|k1 . That is. . . f2 = ∗ ∗ d4 x1 · · · d4 x4 f3 (x3 )f4 (x4 )f1 (x1 )f2 (x2 )× (i)4 r (2r + µ2 ) 0|T (φ (x1 ) · · · φ (x4 ))|0 Let’s get on with the proof. k4 |(S − 1)|k1 . it collapses to k3 .

3. Then the x2 integration t1 →−∞ lim − lim t1 →∞ t2 →−∞ lim − lim t2 →∞ ∗ ∗ d4 x2 d4 x3 d4 x4 f3 (x3 )f4 (x4 )× (i)2 (23 + µ2 )(24 + µ2 ) 0|T (φ f1 (t1 )φ f2 (t2 )φ (x3 )φ (x4 ))|0 Etcetera. you get t1 →−∞ lim − lim t1 →∞ ∗ ∗ d4 x2 d4 x3 d4 x4 f3 (x3 )f4 (x4 )f2 (x2 )× (i)3 r=2.4 (2r + µ2 ) 0|T (φ f1 (t1 )φ (x2 )φ (x3 )φ (x4 ))|0 We push a time derivative through a time-ordered product in these steps. November 13 Notes from Sidney Coleman’s Physics 253a 165 general field A. call it −Af (t) =− = Also. t1 →−∞ lim − lim t1 →∞ t2 →−∞ lim − lim t2 →∞ t3 →∞ lim − lim t3 →−∞ t4 →−∞ lim − lim t4 →∞ note the difference in sign 0|T (φ f1 (t1 )φ f2 (t2 )φ f3 † (t3 )φ f4 † (t4 ))|0 . First do the x1 integration.15. if A is hermitian. then i d4 xf (2 + µ2 )A = i =i = 2 d4 x f ∂0 A + A(− 2 2 d4 x(f ∂0 A − A∂0 f ) 2 + µ2 )f dt ∂0 d3 x i(f ∂0 A − A∂0 f ) this is something that appears often enough that it is worth giving it a name. i dt ∂0 Af (t) lim − lim Af (t) t→−∞ t→∞ dx f ∗ (x)(2 + µ2 )A = t→∞ lim − lim t→−∞ Af † (t) note difference in sign from conjugating the i in the def’n of Af (t) Now we’ll apply this equation to the RHS of the equation we want to prove. It is OK in the limit we have. It is a function of time only.

All 4! orderings lead to the same result however. We have lim − lim lim − lim lim − lim f4 |T (φ f1 (t1 )φ f2 (t2 )φ f3 † (t3 ))|0 t1 →−∞ t1 →∞ t2 →−∞ t2 →∞ t3 →∞ t3 →−∞ note the difference in sign The exact same considerations apply to the t3 limits except we get the matrix elements of out f3 . We get lim − lim out t1 →−∞ t1 →∞ f3 . When t4 → −∞. f4 |f1 . f4 |(S − 1)|f1 . f2 out $ $ − $$$ + $$$ ψ|f1 ψ|f1 The last two terms cancel. f4 |φ f1 (t1 )|f2 − lim out t2 →∞ f3 . Of course you only get something physically measurable when you integrate.15. However φ f4 † (t4 ) with the vacuum on the right and any other state on the left vanishes in the limit. f4 |φ f2 (t2 ) On to the evaluation of the t1 limit. The mathematical expression we started with makes sense even for f1 . The two terms remaining are exactly what we wanted to get. We have obtained f3 . smear. f2 in − out f3 . The second term looks real bad. because there is no difference between a matrix element of limt1 →+∞ φ f2 (t2 )|0 and limt1 →−∞ φ f2 (t2 )|0 ¨ . When t4 → +∞. f4 plane waves. f4 |φ f2 (t2 )φ f1 (t1 )|0 The first term was expected. f4 | with the remaining mess ‘out’ because both fields are applied to the vacuum in the far future. f2 REMARKS: 1. No one can build a point charge. it is the latest time and the time ordering puts it on the left. We’ll make that expression the definition of an S − 1 matrix element of plane waves. f3 . f4 |f1 . We get out f3 . it is the earliest time and thus the time ordering puts it on the right. The situation is very analogous to V (x − y) in the expression U = d3 xd3 yV (x − y)ρ(x)ρ(y). the expression. f2 . Let’s compartmentalize our ignorance by just giving a name to this state we have created ψ| ≡ lim out t2 →∞ f3 . Doing the t2 limits does not result in such a simplification. and thus no one . November 13 Notes from Sidney Coleman’s Physics 253a 166 If we had reduced the integrals in some other order we would have a different order of limits here. and we get the matrix element of f4 | with the rest of the mess. and we’ll just do the order we have arrived at.

There is thus a many to one correspondence between fields 1 ˜ and particle. x2 . and multiply it by another constant to get the renormalized field. From the point of view of the reduction formula. You do not have to begin with one of the fields that seemed to be fundamental in the Lagrangian. you would expect q q(x) to have a nonvanishing vacuum to one meson matrix elements. Using methods of the same type as those used in the derivation of the LSZ formula. 3. “All” you need ¯ to calculate 2 → 2 meson scattering then would be G (4) (x1 . k2 in . ¯ 4. x3 . it may be possible to use these properties in the LSZ formula to get some exact knowledge about S − 1 matrix elements.15. Then you shift that field by some constant. the expression for S − 1 matrix elements of plane waves. November 13 Notes from Sidney Coleman’s Physics 253a 167 can make a charge distribution that directly measures V (x − y). f2 = d3 k1 d3 k4 ∗ ··· F ∗ (k3 )F4 (k4 )F1 (k1 )F2 (k2 )× (2π)3 2ωk1 (2π)3 2ωk4 3 k3 . All you can do is measure U for various charge distributions. 5. For example. that is. other formulas can be derived. Of course no one has gotten G (4) . If we have some exact knowledge of the position space properties of a field. The formula analogous to U = d3 xd3 y V (x − y)ρ(x)ρ(y) is f3 . If q(x) is a quark field. one for which the interaction energy is V (x − y). k2 2. whose Green’s functions are what actually entered the proof.” You only recover something physically measurable when you integrate. “the potential energy of between two point charges. Then you can abstract to the notion of V (x−y). the mesons are bound states of a quark and an antiquark. In the QCD theory of the strong interaction. one can “stop half way” in the reduction formula and obtain k3 . There is no problem in principle of obtaining scattering matrix elements of composite particles and bound states. k4 |(S − 1)|k1 k2 = d4 x3 d4 x4 eik3 ·x3 eik4 ·x4 × (i)2 (23 + µ2 )(24 + µ2 ) 0|T (φ (x3 )φ (x4 ))|k1 . x4 ) ≡ 0|T (¯q(x1 )¯q(x2 )¯q(x3 )¯q(x4 ))|0 q q q q where q q(x) is the renormalized field. k4 |(S − 1)|k1 . smear. φ = φ + 2 gφ2 is just as good as field (at least except for one exceptional value of g that makes the vacuum ˜ to one particle matrix element of φ vanish). The proof only required that the field you begin with have a nonzero vacuum to one particle matrix element. f4 |(S − 1)|f1 .

and then invert the equation to eliminate g0 . The coefficient of − 1 φ2 in L may not be the meson mass squared. when calculating our scattering matrix elements. fn |A(x)|0 = d4 x1 · · · d4 xn f ∗ (x1 ) · · · f ∗ (xn ) (i)n r (2r + µ2 ) 0|T (φ (x1 ) · · · φ (xn )A(x)|0 Applying the methods used above. . . . . We can also use LSZ methods to derive expressions for the matrix elements of fields between in and out states. . . For example. . if that were the coefficient of some term in the electrodynamics Lagrangian. from our expressions for all other quantities of interest. compute the conventionally defined coupling constant from it. these limits can be evaluated to get out f1 . m2 may 0 2 not be the charged muon (nucleon) mass squared. Of course this is really an abstraction of out f1 . We’ll subscript the coupling constant. after the proof of the lemma. What we have a perturbative expansion for if we treat −g0 ψ ∗ ψφ as our . extremely lucky. . which is not directly measured. kn |A(x)|0 = d4 x1 · · · d4 xn eik1 ·x1 +···+ikn ·xn × (i)n r (2r + µ2 ) 0|T (φ (x1 ) · · · φ (xn )A(x)|0 where φ (x) is a correctly normalized field that can create the outgoing mesons and A(x) is an arbitrary field. I can show out k1 . In general it isn’t. we get lim − lim ··· lim − lim 0|T (φ f1 † (t1 ) · · · φ fn † (tn )A(x)|0 t1 →∞ t1 →−∞ tn →∞ tn →−∞ Just as easily as we evaluated the t3 and t4 limits. Furthermore g0 may not be what we want to call the coupling constant. . . Also. fn |A(x)|0 A second look at model 3 and its renormalization: 1 µ2 L = (∂µ φ)2 − 0 φ2 + ∂µ ψ ∗ ∂ µ ψ − m2 ψ ∗ ψ − g0 ψ ∗ ψφ 0 2 2 The upshot of what we have done so far is that some 0 subscripts have been added to the Lagrangian. November 13 Notes from Sidney Coleman’s Physics 253a 168 This is used to derive theorems about the production of “soft” (low energy) photons. It would be lucky.15. . we need Green’s functions. In real electrodynamics there is a parameter e defined by some experiment. .

This determination of φ . . µ0 . the field satisfying 0|φ |0 = 0 p|φ (0)|0 = 1 φ = Z3 2 (φ − φ 0 ) So along the way in calculating quantities of interest. Six unknowns. 3. A. 0|φ |0 = 0 2. is the Green’s functions of φ. m. Those Green’s functions aren’t exactly what we are interested in. g agrees with the conventionally defined g. q| φ (0)|0 = 1 one meson −1 p| ψ (0)|0 = 1 one anti-nucleon 4. the field whose kinetic term has coefficient 1 in the Lagrangian (and thus obeys the canonical commutation . F . Of course. g0 and the above conditions. . six conditions. . It can be avoided. and then the plugging ins of the inverse of these equations into other quantities of interest sounds like a mess.15. if you actually wanted to know the relationship of φ to φ. µ. . LCT = Aφ + 1. We want the Green’s function of φ . The nucleon mass is m 6. We rewrite L with six new parameters. November 13 Notes from Sidney Coleman’s Physics 253a 169 interaction Lagrangian. The meson mass is µ 5. and g from m0 . we’ll have to calculate the Green’s function of φ from the Green’s functions of φ. µ2 1 L = (∂µ φ )2 − φ 2 + ∂µ ψ ∗ ∂ µ ψ − m2 ψ ∗ ψ − gψ ∗ ψ φ + LCT 2 2 B c (∂µ φ )2 − φ 2 + D∂µ ψ ∗ ∂ µ ψ − Eψ ∗ ψ − F ψ ∗ ψ φ + const 2 2 The six new parameters are going to be determined order by order in perturbation theory by six renormalization conditions.

g0 . the things we usually compute. we compute everything in terms of the right quantities. and then inverting to get S matrix elements in terms of µ2 . You can also read off the bare meson mass. you can read it off. 0 Z3 (See the discussion from Eq. There are derivative interactions in LCT . ˙ This parenthetical remark should be emphasized more. 25 . m2 and g. etc. We have to make contact with the committee definition of g. you avoid the infinities which plague quantum field theory. and the relationship of ψ to ψ. µ2 . not φ . (10. 23 for these same ideas expressed before we knew / worried about Z3 and 0|φ|0 ) What you have now is a perturbation theory for the quantities you are really interested in in terms of the conditions on φ and ψ . and experimentally input parameters. the wrong parameters to hold fixed. Then we may still have to worry about defining it in terms of Green’s functions [(2)]. 1. It is φ that satisfies [φ. the bare nucleon mass. (6) are not expressed in terms of Green’s functions. This procedure has a bonus. There are three technical obstacles we will have to overcome to implement this program. m0 and g0 . Renormalization conditions (4). as long as you stick to observable quantities expressed in terms of physical parameters.1) to the end of the lecture of Oct. The differences in the two kinds of perturbation theory is what you call the interaction Lagrangian. m2 and g from µ0 . 2.15. We’ll be taking −gψ ∗ ψ φ + LCT as the interaction. Instead of computing scattering matrix elements µ2 . the experimentally input parameters. 1 1 1 µ2 (1 + B)(∂µ φ )2 + (µ2 + C)φ 2 + Aφ + const = (∂µ φ)2 − 0 φ2 2 2 2 2 Z3 = 1 + B 1 2 µ2 = (µ + C). and has no linear term. 3. m2 and g. November 13 Notes from Sidney Coleman’s Physics 253a 170 relation)25 . (5). This is called renormalized perturbation theory. φ] = iδ (3) with coefficient 1.

Graphically. by hypothesis. A is going to be some power series in g. so the coefficient of g n+1 in its power series must vanish). November 18 Notes from Sidney Coleman’s Physics 253a 171 16 November 18 Let’s go into more detail on how 0|φ |0 = 0 determines A. k← (n+1) xf . A= r Ar Ar ∝ g r Diagrammatically. that is to get An+1 we apply the renormalization condition 0|φ |0 = 0. Thus these graphs only contain known stuff. = 0 at order g n+1 .16. To determine A to order g n+1 . (We demand this for all values of g. If they are going to be of order g n+1 every vertex has to be of order g n or less. There is only one of them. (r) xf xf xf k← k← corresponds to iA(2π)4 δ (4) (k) corresponds to iAr (2π)4 δ (4) (k) = r (r) xf I’ll now explain how to determine A order by order in perturbation theory. The graph with only one vertex at order g n+1 with one external line also have a special property. Suppose that we know all Feynman graphs and have determined all counterterms to order g . We can break n pf pf pf = 0 at O(g n+1 ) into two parts = pf + pf at order g n+1 graphs of order g n+1 graphs of order g n+1 with more than one vertex with only one vertex behind the shield The graphs with more than one vertex behind the shield have a special property.

Consider the same anything but summed over all possible tadpoles that can be attached to that same line. 13) before we can do anything.16. We can cancel completely a potentially momentum dependent sum of ←k graphs by adjusting a single number because it is always ∝ δ (4) (k). . . but first we have to surmount three obstacles (see page 170 in the lecture of Nov. . . Now there is a nice simplification in all graphs because of this cancellation. Physical Review’s editors rejected spermion. The program for determining B. F successively will be similar. (Unless you cared about A). Tadpole all tadpoles including counterterms = =0 The total result is that you can just ignore all tadpoles. November 18 Notes from Sidney Coleman’s Physics 253a 172 Setting k← (n+1) xf =− pf ←k graphs of order g n+1 with more than one vertex behind the shield determines An+1 . = 0 for all k. Suppose we have Any Tadpole pf this can be anything A part of graph that is connected to the rest of the graph by one and only one line is called a tadpole. .

we’ll just note that these two problems frequently cancel out. In terms of φ L = Z3 µ2 1 (∂µ φ )2 − φ 2 2 2 2 1 µ 1 1 = (∂µ φ )2 − φ 2 + (Z3 − 1) (∂µ φ )2 − µ2 φ 2 2 2 2 2 −1 . A simple example Take the simplest field theory 1 µ2 L = (∂µ φ)2 − φ2 2 2 and introduce a new field φ = Z3 2 φ. Later we will develop a new method to deal with these problems. and that in a few simple examples. and you act as if HI = −LI and as if T (∂µ φ · · · ) = ∂µ T (φ · · · ) you get the right answer. we can explicitly show this. (Act as if HI = −LI and as if ∂µ → ikµ ) What do I mean by “cancel out”? If you are naive. take Z3 to be arbitrary. the interaction Lagrangian. in general. The Wick expansion does not apply to derivatives of fields. For now. This means that the interaction Hamiltonian is not just −LI . we had to employ the Wick expansion which turns time ordered products of free fields into normal ordered products.16. and we can’t pull the derivatives out of the time ordered product T (∂µ φ(x) · · · ) = ∂µ T (φ(x) · · · ) and then apply the Wick expansions. November 18 Notes from Sidney Coleman’s Physics 253a 173 (1) Problems with derivative couplings and why they don’t arise here In the presence of a derivative interaction. π µ = ∂ µ φ. A second problem is that to get from Dyson’s formula to Feynman diagrams.

kn ). to be the sum of all connected ˜ graphs with n external line that contribute to G(n) (k1 .16. k2 ). k2 ) are k1 → ←k2 ff fxf fxfxf fxfxfxf + k1 → ←k2 + k1 → k1 → ←k2 + k1 → k1 → k1 → ←k2 + · · · birds on a rail = (2π)4 δ (4) (k1 + k2 ) i 2 k1 − µ2 + i 1+ + 2 −i(Z3 − 1)(−k1 + µ2 )i 2 k1 − µ2 + i 2 −i(Z3 − 1)(−k1 + µ2 )i 2 k1 − µ2 + i 2 + ··· i 1 − (Z3 − 1) + (Z3 − 1)2 + · · · 2+i −µ i 1 = (2π)4 δ (4) (k1 + k2 ) 2 k1 − µ2 + i 1 + (Z3 − 1) i −1 = Z3 (2π)4 δ (4) (k1 + k2 ) 2 k1 − µ2 + i = (2π)4 δ (4) (k1 + k2 ) 2 k1 . k2 ). . What does naive perturbation ˜ (2) ˜ theory give for Gc (k1 . n/2 0|T (φ(x1 ) · · · φ(xn ))|0 = Z3 0|T (φ (x1 ) · · · φ (xn ))|0 We’ll show that this holds perturbatively using the naive method above. . . k2 ) = (2π)4 δ 4 (k1 + k2 ) 2 c k1 − µ2 + i only contribution is ff k2 → ←k1 That’s the right answer in this exactly soluble theory. the sum of all connected graphs that contribute to G(2) (k1 . . . ˜ (n) Define a connected Green’s function Gc (k1 . first we will only show it for one Green’s function but we’ll be more general in a moment. The only nonzero connected Green’s function for one scalar field with no interactions is i G(2) (k1 . Actually. The interaction is (Z3 − 1) 1 1 (∂µ φ )2 − µ2 φ 2 2 2 fxf k2 → ←k1 It has as Feynman rule 1 µ µ ←→ i(Z3 − 1) 2!((−ik1 )(−ik2 ) − µ2 )(2π)4 δ (4) (k1 + k2 ) 2 2 = −i(Z3 − 1)(−k1 + µ2 )(2π)4 δ (4) (k1 + k2 ) (2) The connected con graphs contributing to Gc (k1 . November 18 Notes from Sidney Coleman’s Physics 253a 174 The Green’s functions of φ are simply related to the Green’s functions of φ. because −1/2 φ = Z3 φ. . . . kn ).

but this agrees with the right result. Consider a theory of one scalar meson with arbitrary nonderivative self interactions 1 1 L = (∂µ φ)2 − µ2 φ2 + 2 2 Again let φ = Z3 −1/2 N gr φr r=3 φ N 1 1 1 1 L = (∂µ φ )2 − µ2 φ 2 + (Z3 − 1) (∂µ φ )2 − µ2 φ 2 + 2 2 2 2 gr φ r Z3 r=3 r/2 We are going to compute a general Green’s function in this theory in terms of the Green’s function of the theory without Z3 by making a graphical relation.16. Consider any graph in the Green’s function of φ. Corresponding to this graph. the progenitor. k2 ) c c using a naive method. A slightly less simple example. there are a whole bunch of graphs in the Green’s function of φ which look just like this graph except there are an arbitrary number . November 18 Notes from Sidney Coleman’s Physics 253a 175 We have shown −1 G(2) (k1 . k2 ) = Z3 G(2) (k1 .

. and etc. It looks like the contributions to G(n) depend on Z3 in a graph dependent way. November 18 Notes from Sidney Coleman’s Physics 253a 176 of birds on each line. The graphs we have summed give −n −I Z3 Z3 Z3 rVr /2 = Z3 −n−I+ r rVr /2 r product of n+I independent geometric series times progenitors contributions to G(n) . Every internal line has both ends on a vertex. Since all the contributions to G (n) have this factor −n/2 G (n) = Z3 G(n) as expected. Therefore n + 2I = or −n − I + r −n/2 rVr rVr n =− 2 2 The graphs we have summed give Z3 . Every external line ends on a vertex. The only effect of −1 all these birds is to replace each internal and external propagator by Z3 times the free r/2 propagator.16. We can sum up this bunch of graphs with our naive Feynman rule. Suppose there are n external lines. There is also an effect on the value of the graph coming from all those Z3 factors at the vertices. Every r legged vertex connects to r of these ends. There is a conservation law. but we aren’t done yet. I internal lines and Vr vertices with r legs. conservation of ends.

that renormalization conditions (2). one meson and all other intermediate states and applied renormalization conditions states and applied renormalization conditions (1) and (2). (3).1) + e−iPn ·(x−y) | 0|φ (0)|n |2 all other momentum eigenstates |n besides vacuum and one meson We have broken up the sum into vacuum. µ2 ). November 18 Notes from Sidney Coleman’s Physics 253a 177 This is the right result. It justifies the naive treatment we will apply to LCT on Model 3. We have an name for d3 p e−ip·(x−y) (2π)3 2ωp it is ∆+ (x − y. . This will require a study of G(2) . 0|φ (x)φ (y)|0 = |n e−iPn ·(x−y) | 0|φ (0)|n |2 d3 p e−ip·(x−y) | 0|φ (0) |p |2 (2π)3 2ωp =1 one meson = $ $ | $$$$ |2 0|φ (0)|0 =0 + (16. G(2) (k1 . (4) and (5) aren’t expressed in terms of Green’s function (we will worry about (6) later). Overcoming the second obstacle. 0|φ (x)φ (y)|0 is called a Wightman function. 0|φ (x)φ (y)|0 = 0|φ (x)|n n|φ (y)|0 fpf k1 → ←k2 d4 x d4 ye−ik1 ·x−ik2 ·y 0|T (φ (x)φ (y))|0 complete set of intermediate momentum eigenstates |n Pµ |n =Pnµ |n So.16. k2 ) = = Now T (φ (x)φ (y)) = θ(x0 − y 0 )φ (x)φ (y) + θ(y 0 − x0 )φ (y)φ (x) so it is sufficient to study 0|φ (x)φ (y)|0 and then take this combination at the end.

we are abandoning our ‘every 1 p integration gets a 2π . a e . It is conventionally called 1 2 0 3 σ(p )θ(p ) (2π) δ (p − Pn )| 0|φ (0)|n | = 4 2 d4 p −ip·(x−y) e σ(p2 )θ(p0 ) 3 (2π) (To agree with unfortunate but longstanding conventions. the physical meson mass is what comes out here.T.) The density σ(p2 ) has some definite properties. call it µ2 + . It is always ≥ 0. then that is what we would be calling the meson. every δ function gets a 2π’ rule. 4µ2 ). but now we can do something tricky with it. because there are no bound states in P. a2 ) (Sometimes ρ(a2 ) ≡ δ(a2 − µ2 ) + σ(a2 ) is used) σ(a2 ) ≥ 0 σ(a2 ) = 0 for a2 < µ2 + This is the Lehmann-K¨ll´n (“Chalain”) spectral decomposition.) So what we have found so far is 0|φ (x)φ (y)|0 = ∆+ (x − y. that vanishes when p0 <0. In perturbation theory.16. Let’s massage the sum over all other momentum eigenstates e−iPn ·(x−y) | 0|φ (0)|n |2 = all other |n p2 + µ2 ) and e−iPn ·(x−y) all other |n d4 p δ 4 (p − Pn )| 0|φ (0)|n |2 The integral over p is just a fancy way of writing 1. it equals zero if p2 < min(4m2 .. > 0. It is in ωp ( ωp = it comes from inserting physical one meson momentum eigenstates.T. Take e−iPn ·(x−y) inside the p integration and rewrite it as e−ip·(x−y) . We have d pe 4 −ip·(x−y) all other |n This is a manifestly Lorentz invariant function of p. it still is zero for p2 < mass2 of the lightest neutral bound state. (If the lightest neutral bound state has a mass less than the meson mass. November 18 Notes from Sidney Coleman’s Physics 253a 178 µ2 . µ2 ) + = ∆+ (x − y. µ ) + 0 2 d4 p −ip·(x−y) e σ(p2 )θ(p0 ) 3 (2π) d4 p −ip·(x−y) ∞ 2 e da σ(a2 )θ(p0 ) (2π)3 0 ∞ da2 σ(a2 )∆+ (x − y. Outside of P. µ2 ) + = ∆+ (x − y.

k ). later in the course. t)]|0 = Z3 iδ (3) (x − y) by c.) We set out to study G(2) (k. φin ] = iδ (3) (x − y) because φ and π can be changed to φin . The value on the positive real axis is given by the i prescription. πin by a canonical transformation. November 18 Notes from Sidney Coleman’s Physics 253a 1/2 179 We can use this to make a statement about Z3 using φ = Z3 φ and the fact that φ obeys the canonical commutation relation. t)]|0 = iδ (3) (x − y) + 0 da2 σ(a2 )iδ (3) (x − y) by using ∂ ∆+ (x − y) = iδ (3) (x − y) ∂y ∞ −1 hence Z3 = 1 + 0 da2 σ(a2 ) ≥ 1 ⇒ In general Z3 < 1.c. which says you take the value just above the cut. t).r ∞ ˙ 0|[φ (x. ˙ (*: [φin . t). φ (y. φ (y. −1 ˙ 0|[φ (x. (16. We will show that if Z3 = 1 you have free field theory. The function is analytic except at k 2 = µ2 where it has a pole with residue i and along the positive real axis beginning at k 2 = µ2 + . where it has a branch cut.16.) This is a highly nontrivial expression. k ) = (2π) δ (k + k ) 4 (4) 4 (4) i + 2 − µ2 + i k 2 ∞ da2 σ(a2 ) 0 k2 i − a2 + i = (2π) δ (k + k ) + D (k ) D (k 2 ) = “renormalized propagator” ∞ i i = 2 + da2 σ(a2 ) 2 2+i k −µ k − a2 + i 0 (Note that [−iD (p2 )]∗ = −iD (p2∗ ) Schwarz reflection property.2) . It defines a function everywhere in the complex k 2 plane (even though the propagator was not originally defined there). What we have shown implies that G (2) (k.

or we can have an interaction. k1 k2 k3 kn k4 1PI the sum of all connected graphs that cannot be disconnected by cutting a single internal line ≡ Our convention will be that this does not include the overall energy momentum conserving δ function or the external propagators. Again it will be defined graphically. November 18 Notes from Sidney Coleman’s Physics 253a 180 k 2 physical value is obtained by approaching cut from above 2 2 + This has nothing to do with the i prescription Our renormalization conditions. nothing can happen. The cute thing about this Green’s function when n = 2 is the following expression for 1PI 1PI 1PI G(2) k1 → ←k2 = + + + ··· fpf ff fpf fpfpf That is. the one particle irreducible Green’s function. (2) and (4) are encoded in the function. Define another new kind of Green’s function. (look back and see where (2) was used in the derivation of the expression for D ) We are going to keep massaging G(2) to find a slicker statement of our renormalization conditions.16. (4) The meson mass is µ ⇔ D has a pole at µ2 (2) 0|φ (0)|q = 1 ⇔ The residue at this pole is +i. but before we get to the other external line there is never a point where we get just one line. or there is only one line .

It is i 2 − µ2 + i k 1+ π (k 2 ) k 2 − µ2 + i π (k 2 ) k 2 − µ2 + i 2 + · · · (2π)4 δ (4) (k + k ) k2 1 i (2π)4 δ (4) (k + k ) 2) 2+i −µ 1 − k2π (k2 +i −µ = Now you can see why π (k 2 ) is called the “self-energy”.. November 18 Notes from Sidney Coleman’s Physics 253a 181 like that or ..16. π (k 2 ) = π (µ2 ) + . the LHS is (2π)4 δ (4) (k + k )D (k 2 ). dπ (k 2 − µ2 ) + · · · 2 µ2 dk These two terms must vanish or it screws up the location and residue of the pole.” we can sum the series on the RHS. It is like a momentum dependent mass. D (k 2 ) = k2 − µ2 i − π (k 2 ) + i Now for the slick rephrasing of the renormalization conditions: D has a pole at µ2 ⇔ π (µ2 ) = 0 dπ The residue of this pole is i ⇔ dk2 k2 =µ2 =0 Perhaps this is easier to see if you think of expanding π (k 2 ) around k 2 = µ2 in a power series. Identifying the coefficient of (2π)4 δ (4) (k + k ) on the LHS and RHS. If we define f pf k→ 1PI k→ = iπ (k 2 ) π (k 2 ) = “self-energy. By definition.

17. November 20 Notes from Sidney Coleman’s Physics 253a 182 17 November 20 Having succeeded in expressing renormalization conditions (2) and (4) as statements about the 1PI two-point function. including all the counterterms. I’ll now explain how to determine B and C order by order in perturbation theory (those of you that have taken quantum field theory once or twice before probably recognize that this is going to be a rerun of the argument for determining A). and we’ll show that Bn+1 and Cn+1 can be determined. 1 1 LCT = · · · + B(∂µ φ )2 − Cφ 2 + · · · 2 2 fpf 1P I = iπ (k 2 ) π (µ2 ) = 0 dπ dk 2 =0 µ2 fxf k→ ←k We can express the Feynman rule for the B and C counterterms together as corresponds to i(2π)4 δ (4) (k + k )(−Bk · k − C) = i(2π)4 δ (4) (k + k )(−Bk 2 − C) Writing B and C as power series expansions B= r Br Cr r Br ∝ g r Cr ∝ g r C= We can also write k→ (r) ←k Assume everything is known to O(g n ). ff x k→ fxf →k = r fxf (r) corresponds to i(2π)4 δ (4) (k + k )(Br k 2 − Cr ) fpf 1P I at order g n+1 = known stuff sum of all 1PI graphs with more than one vertex at order g n+1 + k→ (n+1) →k the only O(g n+1 ) 1PI graph with only one vertex fxf iBn+1 µ2 − iCn+1 = − (known stuff)|µ2 iBn+1 = − d(known stuff) dk 2 k2 =µ2 .

We can do some examples before worrying about obstacle (3). However. that is renormalization condition (6). which is not eliminated by just subtracting a constant and a term linear in k 2 . November 20 Notes from Sidney Coleman’s Physics 253a 183 Similar arguments apply to the nucleon self energy. the exact same as that of the free propagator. in renormalized perturbation theory. the only thing that matters about an external line is the location and residue of the pole. Calculation of π (k 2 ) to order g 2 1P I −iπ (k ) = 2 = fpf fd‡df fxf (2) + = iπf (k 2 ) + iB2 k 2 − iC2 . p← 1P I ←p ¦p¦ dΣ dp2 = iΣ (p2 ) which can be used to express renormalization conditions (3) and (5) as Σ (m2 ) = 0 =0 p2 =m2 Of course these subtractions are not going to allow you to ignore corrections to the 1PI two point function. has a complicated momentum dependence in general. k k fpf k→ 1P I →k k →µ lim 2 2 k 2 − µ2 × i k = 2 2 lim k →µ k 2 − µ2 × i The location and residue of the pole in the full propagator is. That is because in the computation of an S matrix element. this does allow you to ignore corrections to lines on the mass shell. that is external lines in the computations of S matrix elements.17.

The integral is divergent. I suppose we could parametrize the integral with a polar angle measured from k but. (This is called log divergent. and it isn’t even spherical symmetry we have.) we use Feynman’s trick for combining two denominators. 1.I. which was implemented to make expansions in the right parameters has saved us from this unexpected infinity.17. −iπf (k 2 ) = dπf dk 2 (17. November 20 Notes from Sidney Coleman’s Physics 253a 184 where −iπf (k 2 ) ≡ The renormalization conditions are fd‡df − B2 = 0 πf (k 2 ) + B2 k 2 − C2 = 0 dπf dk 2 µ2 If you don’t care what B2 and C2 are. To make this thing manifestly spherically symmetric (actually L.) This last problem is the easiest to take care of: πf (k 2 ) − πf (µ2 ) is not divergent. and if it was cut off at some large radius in q4 momentum space Λ. 2. these can be rephrased as π (k 2 ) = πf (k 2 ) − πf (µ2 ) − (k 2 − µ2 ) we should check that B2 and C2 are real however.1) µ2 fd‡df ←k+q ←k ←k →q = (−ig)2 i i d4 q 4 q 2 − m2 + i (q + k)2 − m2 + i (2π) There are three problems in doing this integral. Not spherically symmetric. would be ∼ ln Λ. We are in Minkowski space. 3. Renormalized perturbation theory. 1 0 1 1 dx = [ax + b(1 − x)]2 b−a 1 ax + b(1 − x) 1 = 0 1 b−a 1 1 − a b = 1 ab (17.2) . at high q it looks like q 3 dq d4 q 1 (2π)4 q 4 which if the integral was spherically symmetric would be ∼ .

3) We could do this integral in a moment if we were living in Euclidean space. It is not spherically symmetric though. with a = (q + k)2 − m2 + i b = q 2 − m2 + i −iπf (k ) = g 2 2 = g2 [q = q + kx] = g2 0 d4 q (2π)4 d4 q (2π)4 1 1 dx 0 1 [((q + [q 2 k)2 − m2 1 + i )x + (q 2 − m2 + i )(1 − x)]2 dx 1 + + 2k · qx − m2 + i ]2 0 d4 q 1 (2π)4 [q 2 + k 2 x − k 2 x2 − m2 + i ]2 dx xk 2 (17.17. Now that q 0 runs from −i∞ to +i∞. so that it runs up the imaginary q 0 axis. q4 = −iq0 . So now we’ll study integrals of the form In (a) = d4 q 1 = 4 (q 2 + a)n (2π) d3 q dq 0 1 4 (q 0 2 − q 2 + a)n (2π) where a has a positive imaginary part.) The location of the poles in the q 0 integration splits into two cases Case: Re(q 2 − a) > 0 q0 x x x x Case: Re(q 2 − a) < 0 q0 In either case. the contour can be rotated as shown (called the “Wick rotation”). a = k 2 x(1 − x) − m2 + i . November 20 Notes from Sidney Coleman’s Physics 253a 185 Apply this to the two denominators in πf . Because this rotation does not cross any poles the value of the integral is unchanged. define a new variable q4 that runs from −∞ to ∞. (The case of interest has n = 2.

If we cut the integral off at some large value Λ2 (in a bit we’ll send Λ → ∞) we have Λ2 I1 (a) = 0 z−a+a dz = −z + a Λ2 0 a dz −1 + −z + a 2 Λ2 = 0 dz −1 + a −z + a For large z. November 20 Notes from Sidney Coleman’s Physics 253a dq 0 = idq4 d4 q = id4 qE = idq4 d3 q d4 q E 1 =i 4 [−q 2 − q 2 + a]n (2π) 4 d4 q E 1 4 (−q 2 + a)n (2π) E 186 In (a) = i This is now a spherically symmetric integral in 4-d Euclidean space. Λ a −i lim dz 1 + I1 (a) = 2 Λ→∞ 16π z−a 0 2 −i = lim [z + aln (z − a)]|Λ 0 2 Λ→∞ 16π −i $ & = lim Λ2 aln $2 1 + O & + $$ Λ 2 Λ→∞ 16π vanishes in convergent combinations 2 have a  Λ 2 − aln (−a) =0.17. we have In (a) = i ∞ π2 1 zdz 4 (2π) 0 (−z + a)n ∞ i (−1)n−1 dn−1 1 = zdz 16π 2 (n − 1)! dan−1 0 −z + a n−1 n−1 (−1) d I1 (a) = (n − 1)! dan−1 ∞ 1 This is only a formal expression because I1 (a) = 0 zdz −z+a has a divergent part. Using V (S 3 ) = 2π 2 1 2 3 and setting z = qE . the integrand is −1 − a + O a2 .in Λ→∞ limit (for our purposes) = i aln (−a) 16π 2 . z z I’ll evaluate I1 (a) in a way which is only valid when the integral is part of a convergent combination ∞ Cn zdz −z + an 0 n where n Cn = 0 and n an c n = 0 1 z This will guarantee that those first two terms in the integrand of order 1 and coefficient zero. qE dqE = 2 zdz.

The limit Λ → ∞ will be taken at the end. November 20 Notes from Sidney Coleman’s Physics 253a 187 What about I2 (a) ? −16π iI2 (a) = 2 1 (z − a)2 0 ∞ z−a+a dz = (z − a)2 0 ∞ 1 a dz = + z − a (z − a)2 0 zdz ∞ 1 For large z the integrand is z + O( z12 ). It also gets rid of the −1. II. So I2 (a) = −i ln 16π 2 (−a) for our purposes. etc. What follows is only valid in either of two cases. . The integral is part of a convergent combination ∞ zdz 0 n Cn (z − an )2 where n Cn = 0 II. I. i I2 (a) = 16π 2 Λ2 dz 0 1 a + z − a (z − a)2 Λ2 i 1 = ln (z − a) − a 2 16π z−a 0 i a& a a$ $ $$$ & $ = (1 $ O( 2 )) −ln (−a) − 1 ln (Λ2 ¡ $$ ) 1 + O( 2 ) − $$+ $ & $2 16π 2 Λ Λ & Λ vanishes =0 in Λ→∞ limit vanishes Let’s see why the terms I claim vanish. Taking a derivative w. You plan to differentiate I2 with respect to a to get I3 . vanish in either case. 1 I. a eliminates these terms.17. As before we make sense of I2 (a) by itself by cutting the integral off at some large value Λ2 . I4 . since that is independent of a.t.r. The condition n Cn = 0 which was put in to make the coefficient of z vanish makes the infinite terms as Λ → ∞ vanish. ln (Λ2 ) and 1 are both constants independent of a.

17. For n ≥ 3. 16π 2 where the triple dots indicate terms that cancel in a sum of such terms such that the total integrand vanishes for high q more rapidly than q −4 . and the 1 that vanishes in convergent combinations or when differentiated to get I3 . can be chucked. In (a) = = = = = = (−1)n−1 dn−1 I1 (a) (n − 1)! dan−1 i (−1)n−1 dn−1 aln (−a) n−1 (n − 1)! da 16π 2 i (−1)n−1 dn−2 (ln (−a) + 1) 16π 2 (n − 1)! dan−2 i (−1)n−1 dn−3 1 16π 2 (n − 1)! dan−3 a i (n − 3)! 1 (−1)n−1 (−1)n−3 2 16π (n − 1)! an−2 i 1 2 (n − 1)(n − 2)an−2 16π These facts are summarized on the following table of integrals. πf (k 2 ) = g2 16π 2 1 dx ln (−k 2 x(1−x)+m2 −i )+terms that vanish in convergent combinations 0 . with a = k 2 x − k 2 x2 − m2 + i . etc. 4 (q 2 + a)n (2π) −i ln (−a) + · · · . In (a) = with n integer and Im a > 0. Let’s get I3 . (17. for n ≥ 3. November 20 Notes from Sidney Coleman’s Physics 253a 188 −i 1 (a) Note that if you take − dIda to get I2 (a) you get 16π2 (ln (−a) + 1).3) is an expression for πf to which we can apply our expression for I2 . I4 . Eq. is given by In (a) = i[16π 2 (n − 1)(n − 2)an−2 ]−1 . For n = 1. 2. etc.. 16π 2 1 d4 q . I1 = and I2 = i aln (−a) + · · · . I4 . The Minkowski-space integral.

− 2 Cφ 2 . (16. C is infinite. However.17. π (k 2 ) = πf (k 2 ) − πf (µ2 ) − (k 2 − µ2 ) = g 16π 2 2 1 2 dπf dk 2 dx ln 0 −k x(1 − x) + m − i (k 2 − µ2 )(+x(1 − x)) + −µ2 x(1 − x) + m2 −µ2 x(1 − x) + m2 µ2 2 This thing. after Eq. and i D (k 2 ) = 2 2 − π (k 2 ) + i k −µ (Note that [iD (k 2 )]∗ = iD (k 2∗ ) ⇒ π (k 2 )∗ = π (k 2∗ )) Let’s look at the analytic structure of π (k 2 ) to second order in perturbation theory. dπf ) gets an imaginary part is when the dk 2 µ2 argument of the logarithm in the integral becomes negative. Recall that (Nov. corresponds to the mass coun1 terterm in L. πf (µ2 ). for real k 2 . 18. November 20 Notes from Sidney Coleman’s Physics 253a 189 Eq.2)) we have already found the analytic structure of D (k 2 ). that is unimportant. the function we have just obtained an expression for. This can be seen by graphing x(1 − x) The only way the expression for πf (µ2 ) (and x(1-x) 1 4 0 1 2 1 x Of course in this case we have no business treating the meson as a stable particle anyway. which was subtracted off of πf (k 2 ). which can happen for ranges of x within [0. It is infinite. The bare mass of the meson does not enter into any expression relating physical quantities. (17.1) is an expression for π in terms of πf . but which can be defined by . 1] if µ2 > 4m2 . We should worry whether this infinite term we have stuck into L is real. the bare mass of the meson is infinites.

) This and the fact that our counterterm was real when µ2 < 4m2 (a requirement necessary for the existence of a physical meson). π (k 2 ) = g2 16π 2 1 dx ln 0 −k 2 x(1 − x) + m2 − i (k 2 − µ2 )x(1 − x) + −µ2 x(1 − x) + m2 −µ2 x(1 − x) + m2 This expression is not only well defined. −∞ < k 2 < 4m2 . all internally consistent theories are not right. Compare this with the analytic structure of D and you’ll see that perturbation theory is satisfying formulas obtained outside of perturbation theory. it is analytic for Im k 2 = 0. k 'is real for k2real and < 4m2 2 4m 2 physical value is obtained by approaching cut from above The i prescription when k 2 is real and greater than 4m2 says to define the logarithm and hence π by approaching the cut from above. 1]. because the branch cut in the logarithm needs to be defined for ranges of x ∈ [0. but starting at 4m2 . there is a branch cut in π . When you invert D to get π you get a zero. are satisfying consistency checks. (D had a pole at µ2 . All right theories are internally consistent. It is also analytic for Im k 2 = 0. This is in agreement.17. November 20 Notes from Sidney Coleman’s Physics 253a 190 this expression for complex k 2 . (However.) Loop Lore How do you generalize the wonderful tricks done here to graphs with more propagators and more loops? .

Then we’ll be able to do a shift and a Wick rotation and then a spherically symmetric integral. Combining denominators 1 = ar + i r=1 n ∞ −i r 0 ∞ dβr eiβr (ar +i ) ∞ = (−i)n 0 dβ1 · · · dβn ei r βr (ar +i ) 0 dλ δ λ − s Fancy way of inserting 1 into the integrand βs ∞ ∞ = (−i) n 0 dλ 0 dβ1 · · · dβn ei r βr (ar i ) δ λ− s βs Now for some rescalings. αi [−i (−i)n (n − 1)! n r αr (ar + i )] 1 = (n − 1)! a +i r=1 r n ∞ dα1 · · · dαn δ 1 − 0 i αi [ 1 n r αr (ar + i )] . that combine all the propagators into one denominator. November 20 Notes from Sidney Coleman’s Physics 253a 191 We’ll introduce Feynman parameters. First rewrite δ λ− s βs as 1 δ 1− λ βi λ βs λ Then introduce new integration variables αi = 1 = (−i)n ar + i r=1 ∞ n ∞ ∞ dλ 0 0 dα1 · · · dαn λn−1 eiλ αi i αr (ar +i ) δ 1− i αi αr (ar +i ) ∞ = 0 ∞ dα1 · · · dαn δ 1 − dα1 · · · dαn δ 1 − 0 i (−i)n 0 λn−1 dλeiλ = So. What remains and is very difficult is the integration over the Feynman parameters. n − 1 of them if there are n propagators.17. In difficult but important applications those integrations are done accurately by computer.

leaving n − 1 of them.2). Γ(n + 1) = n! Feynman parameters Aj real. (17. Take n = 2. and used that to stop the αi integration at 1.17. αj > 0 ∞ I= 0 dt tα−1 e−At = ∞ 0 1 Γ(α) Aα ∞ 1 αj = j (Aj ) j dtj tj j e−Aj tj Γ(αj ) ∞ 0 j α −1 ds δ(s − 0 Fancy way of writing 1 tj ) ∞ (tj = sxj ) = 0 ds 1 0 j sαj dxj xj j e−sAj xj δ(1 − Γ(αj ) s ∞ α −1 xj ) = dxj xj j Γ(αj ) Γ(αj ) α −1 0 ds s( Γ( ( j j j αj )−1 −s e j xj A j = j 1 0 α −1 dxj xj j aj ) j xj Aj ) αj Γ( j αj ) 1 αj = j Γ(αj ) j (Aj ) dx1 · · · dxj δ 1 − j xj j xαj −1 αj ( j xj A j ) . as advertised. and you have 1 1 = A1 + i A2 + i 1 dx 0 1 [A1 x + A2 (1 − x) + i ]2 A shorter derivation using the Γ function ∞ Uses: Γ(x) ≡ 0 dt tx−1 e−t . However. 1 = (n − 1)! ar + i r=1 n 1 1−α1 1−α1 −α2 1−α1 −α2 −···−αn−2 dα1 0 0 dα2 0 dα3 · · · 0 n−1 r=1 dαn−1 1 αr ar + (1 − n−1 r=1 [ αr )an + i ]n This generalizes the result of Eq. easy to remember. November 20 Notes from Sidney Coleman’s Physics 253a 192 where I have noticed that the δ function vanishes whenever any of the αi are greater than one. we can use the δ function to do the integral over one of the Feynman parameters. α1 = x. This is a nice symmetric form.

that is L momentum integrals still left to be done after using the energy-momentum conserving δ functions. The integral to be done looks like d4 k d4 l (k 2 − m2 )(l2 − m2 )((k − l)2 − µ2 )((k + p)2 − m2 )((l + p)2 − m2 ) To this we would apply our denominator combining identity. qj . I = 5. pansion. Take Aj = 1. All momenta on the I internal lines are linear combinations of the ki and qj . 1 = AB 1 dx 0 1 (xA + (1 − x)B)2 Γ(αj ) = Γ( j αj ) j 2. how do we make the loop integration trivial? Suppose we have a graph with I internal lines. . . generalized binomial ex- Γ(α1 )Γ(α2 ) = Γ(α1 + α2 ) 1 dx xα1 −1 (1 − x)α2 −1 0 Now that we have introduced the Feynman parameters into the integral. After introducing the Feynman parameters. November 20 Notes from Sidney Coleman’s Physics 253a 193 Examples 1. . Call the independent loop momenta ki . L. and L loops. Beta function δ 1− j xj j dxj xj j α −1 . the integral to be done is of the form 1 dα1 · · · dαI δ 1 − 0 α d4 k1 · · · d4 kL DI . . 3. i = 1. Let’s write down the general case. k l p l-k p l+p k+p has L = 2. and the external momenta.17.

C is a number. depending linearly on the Feynman parameters.j=1 Aij ki · kj + i=1 Bi · ki + C A is an L×L matrix that is linearly dependent on the Feynman parameters.j=1 L Aij Oik kk · Ojl kl + C (OT AO)kl kk · kl + C i. It is positive definite except at the endpoints of the Feynman parameter integrations. It has a small positive imaginary part. B is a vector with L four vector components. Now shift the k integration to eliminate the terms linear in k. It is linear in the Feynman parameters and linear in the external momenta. November 20 L Notes from Sidney Coleman’s Physics 253a L 194 where D = i.j=1 L = = i=1 ai ki · ki + C . and the external momenta squared and the masses squared that appear in the propagators. but now it has some awful dependence on the Feynman parameter because of A−1 . Now diagonalizing Aij with an orthogonal transformation on the set of four vectors ki ki = Oij k2 L det O = 1 L d4 ki = i=1 L i=1 d4 ki D= i. ki = ki + 1 2 (A−1 )ij Bj j d4 ki = d4 ki L D= ij=1 Aij ki · kj + C where C = C − 1 4 Bi A−1 Bj ij ij C is still linear in external momenta squared and the masses squared. It still has a small positive ij imaginary part.17.

Finally we’ll make a transformation to eliminate the ai 1 ki = √ ki ai L L 4 d ki = i=1 i=1 4 1 √ ai −2 d4 ki L = (det A) The integral to be done has been reduced to 1 d4 ki i=1 dα1 · · · dαI δ 1 − 0 L α (det A)−2 d4 k1 · · · d4 kL DI where D = i=1 ki · ki + C 0 Now we can perform Wick rotations on each of the ki 1 variables independently to get d4 k1 E · · · d4 kL E DI dα1 · · · dαI δ 1 − 0 α (det A)−2 in d4 ki = id4 ki E ki 4 = iki L 0 D=− i=1 2 kE + C This is one big spherically symmetric integral in 4L dimensions! Easily done with only a slight generalization of our integral table. November 20 Notes from Sidney Coleman’s Physics 253a 195 where (OT AO)kl = δkl al . We have reduced a general graph to an awful integral over Feynman parameters. this is progress.17. All you need in the end is det A. . Note that you don’t actually have to diagonalize A when applying this formula.

B. p 2 . p 2 . C. it has some experimental significance.18. November 25 Notes from Sidney Coleman’s Physics 253a 196 18 November 25 The definition of g in Model 3 Renormalization condition (6). p 2 and q 2 ? Γ is a Lorentz invariant. q 2 are equally good as far as determining F is concerned. there is one that is more equal than others. q 2 ) The −i is a sensible convention. It might well be the one the committee picks. so no committee has actually gotten together to define g. The proof of the iterative determination of F is identical. q = −p − p . There are only two independent momenta. Define ¤xf i ←q p p = −iF (2π)4 δ (4) (p + p + q) ¤ pf i 1P I ←q p = −i Γ(p2 . The point is p2 = p 2 = m2 q 2 = µ2 . q 2 ) The bars mean some specific point in momentum2 space. the committee definition of g. p 2 and p · p . So here is our committee definition of g: g ≡ Γ (p2 . We’ll play committee. However p · p can be traded in for q 2 . because as we will show. The statement is needed to fix the counterterm in L L = · · · − F ψ∗ ψ φ + · · · which has Feynman rule p Model 3 does not exist in the real world. so it must be a function of Lorentz invariants only. While all points p2 . put there so that at lowest order Γ=g Why can we consider Γ to be a function of p2 . so the only Lorentz invariants are p2 . This is a reasonable if not obvious generalization of the types of conditions we used to determine A. D and E. has not been stated or turned into an equation among Green’s functions. p 2 .

there is some line in the graph which can be cut and the graph falls into two pieces. you have to make some of their components complex. The = 3 possibilities. If 2 2 cutting the internal line separates the incoming meson and nucleon from the outgoing ones the graph must be a contribution to (I hope you can convince yourself) the other two possibilities are . If I ignore interactions on the external legs. the cutting of the internal line separates the graph into two pieces each having 1 4 two external lines. as well as p + p + q = 0. = connected Green's function 1PI stuff that is 1PI + not 1PI stuff that isn't 1PI We can say more about the stuff that isn’t 1PI. This point is not kinematically accessible. considered as a function of three complex variables. What is this point’s experimental significance? Look at the process φ + N → φ + N . By definition. that the domain of analyticity of Γ . t and u channel graphs.18. look like s. One can show in general however. November 25 Notes from Sidney Coleman’s Physics 253a 197 To actually find a trio of four-vectors satisfying these conditions. is sufficiently large to define the analytic continuation of Γ from any of its physically accessible regions to this point. Diagrammatically.

To summarize Experimentally. we can split up dpe i ¤ = −ig 2 + plus analytic or at least no pole. we have −ig i (−ig) + analytic stuff at s = m2 2 s−m Because of our definitions.5. s = m2 . and extrapolating down to s = m2 . You just measure the s wave scattering. When this was done. Consider the process N + N → N + N .18. November 25 Notes from Sidney Coleman’s Physics 253a 198 Using our definitions. so they measured the pole in γ + p → p + π. they found pole-like behavior with g ≈ 13. but we don’t expect them to have a pole at s = m2 . the graph dpe i ¤ 1P I probably has all sorts of cuts. near s = m2 s − m2 . Actually. it put the last nail in the coffin for the attempts to consider the strong interactions perturbatively with the pion and nucleons as fundamental particles. the residue of the pole of these graphs is −ig 2 . s. which measures eg. µ2 )) This has a pole at s = m2 because the full nucleon propagator has a pole there. the first graph (on mass shell) is −iΓ(s. µ2 )D (s)(−iΓ(m2 . they weren’t very good at making pion beams back when they did this. Now what about the other two graphs. we can write down what the residue of that pole is.5 was determined this way. They look like they have poles at u = m2 and t = µ2 . When g = 13. the residue of this pole can be measured by looking at φ + N scattering in the physical region. By similar arguments. but it is unlikely that it has a pole at s = m2 because there is no propagator on the inside of the graph that carries the whole incoming momentum. Furthermore. m2 . and because of our renormalization conditions.

they got agreement (to within 10%). to measure the effect of this pole. where they felt justified π calculating with P. The counterterm is The O(g 3 ) correction to Γ in model 3 needed only to make the theory agree with the committee definition of g.) Renormalization vs. the fit showed that the location of the pole was at t = m2 (They only fit the high partial waves. This extreme convergence is peculiar to model 3 (and other models where all the couplings have positive mass dimension as we will later see). L = · · · + gφ4 + · · · . To see that the graph is finite. Our simple model states that the residue of this pole is the same.18. after doing some work to eliminate electromagnetic effects. Without even combining denominators 4 you can see that at high q the integral looks like dq6q . November 25 Notes from Sidney Coleman’s Physics 253a 199 1PI into four things connected 1PI and two more 1PI The thing to note is that this decomposition leads us to expect ¤pi i ¤ = (−ig)2 i + stuff with no pole at t = µ2 2 t−µ Since t ≈ µ2 is unphysical. Consider a model with a four scalar field interaction. look at its high momentum behavior. you again have to extrapolate. Futhermore. −ig 2 . infinities (3) is finite.T. When they did this experiment with p + p → p + p.

18. and a φ4 counterterm which can cancel off the log divergence What about . ∼ log divergent There is a committee definition of g. November 25 Notes from Sidney Coleman’s Physics 253a 200 φ4 could be ABCD or (ψ ∗ ψ)2 . but it is constant and is really easy to cancel After combining denominators you see that the integral looks like at high q 2 loops d8 q q6 three propagators quadratic divergence Fortunately there are renormalization counterterms What about other graphs in this theory? There is d4 q q4 fxf to cancel this infinity. there is . Look at the lowest order correction to a propagator: Whoops.

you cannot make any predictions. Definition (This is a more stringent definition than is often used. One possibility is to hunt for some µ2 1 relationship among the terms in the infinite series. We have not shown that polynomials of degree 4 or less are renormalizable. It seems fairly clear that to correct this defect. You can see that adding jφ7 and kφ8 to the Lagrangian is not going to help. which is good. we would need a φ6 counterterm to cancel this graph’s divergence if it weren’t convergent. Then is log divergent. . we say φ5 theory is not renormalizable. Suffice it to say that no one has ever been able to construct a renormalizable non-polynomial interaction that is not equivalent to free field theory. This shows that any polynomial interaction of degree higher then 4 is not renormalizable. So gφ5 +hφ6 is not a renormalizable interaction either.) A Lagrangian is renormalizable only if all the counterterms required to remove infinites from Green’s functions are terms of the same type as those present in the original Lagrangian. Since the theory did not originally contain a φ6 interaction. That would require another committee definition. then your φ6 counterterm will be of the same type as the interaction term in the original Lagrangian. say for 3 → 3 scattering at some momentum. and you would need a φ6 counterterm to cancel it. Now these theories with an infinite series of interactions are disgusting because they contain an infinite number of independently adjustable parameters. but what we have found so far suggests it. These are also log divergent and they require φ7 and φ8 counterterms to cancel them. √ Perhaps S = d4 x −gR. Perhaps L = ∂µ φ2 − φ2 − λ cos αφ is 2 2 renormalizable. the Einstein-Hilbert action is renormalizable. Unless you make some additional statement. Suppose a theory has a φ5 (ABCDE) interaction.18. November 25 Notes from Sidney Coleman’s Physics 253a 201 This is finite. there is and to worry about. But then. you just add a φ6 term to your Lagrangian.

In that case. Our subtraction. the elimination of infinities.) D'(k2) 2 k . or anything else involving an external meson. demand Re π (µ2 ) = 0 dπ =0 Re 2 dk k2 =µ2 We will see that with these renormalization conditions for µ > 2m. and is continuously related to the subtractions made for µ < 2m. for µ > 2m. One road is to just say. but they can’t hide. so just drop all renormalization conditions and subtractions related to the meson. I have no business calculating meson-meson scattering or nucleon-meson scattering. For µ > 2m. which says π (µ2 ) = 0 and dπ dk 2 =0 k2 =µ2 would be causing us to subtract imaginary terms from the Lagrangian. that as you increase µ. and then it moves onto the second sheet (The poles can run. the meson is unstable. the pole in D (k 2 ) moves up the real axis until it touches the branch cut. but so that we do not make imaginary subtractions. and we lose the bonus of renormalization. πf (k 2 ) is not real at the subtraction point. Our modified procedure is to quite a degree ad hoc. This is unacceptable because a non-Hermitian Hamiltonian is unacceptable. We will modify our subtraction procedure for µ > 2m so that it still removes ∞’s. The definition of the theory does not change in any smooth way as µ increases beyond 2m. This road is not ideal for two reasons. and the usual one for µ < 2m. but we will see that it is useful.18. November 25 Notes from Sidney Coleman’s Physics 253a 202 Unstable particles Let’s look at model 3 in the regime µ > 2m. You can see this by looking at πf or you can look at the nonperturbative formula Im π (k 2 ) = −π σ(k 2 ) |D (k 2 )|2 σ(k 2 ) = 0 when k 2 = µ2 > 4m2 so Im π = 0.

we can move the cut leading from the branch point to ∞ around any way we like. [−iD (k 2 )]−1 = k 2 − µ2 − π (k 2 ) A formula we will use is Im π (k 2 ) = |D (k 2 )|−2 (− π )σ(k 2 ) 3. What we get when we analytically continue D (k 2 ) to Im k 2 < 0 from its value for Im k 2 > 0. because it is not separated by a discontinuity. is called D (k 2 ) on the second sheet. Now let’s work on [−iD (k 2 )]−1 | n|φ (0)|0 |2 (2π)4 δ (4) (k − Pn ) |n =|0 . D (k 2 ) has a discontinuity across the cut.18. The value of D (k 2 ) for Im k 2 < 0 is defined by that integral expression for D (k 2 ) in terms of σ(a2 ). In some sense this is much closer to the physical region. the value along the real axis is physical and can’t be changed. 1 = − |D (k 2 )|−2 2 using the definition of σ. A couple other ways of saying this: The branch point is fixed.. to get a function that is continuous along the old cut.. November 25 Notes from Sidney Coleman’s Physics 253a 203 What do I mean by second sheet? The value of D (k 2 ) for k 2 > 4m2 (and real) is obtained by taking the limit from positive imaginary k 2 down onto the real axis. The second sheet is what you get from peering down from above the cut. according to the i prescription.|p |p : one meson [−iD (k 2 )]−1 = k 2 − µ2 − π (µ2 ) − (k 2 − µ2 ) this is O(g 2 ) dπ dk 2 this is O(g 2 ) k2 =µ2 + O(g 4 ) = k 2 − µ2 − Re π (µ2 ) −i Im π (µ2 ) + O(g 4 ) Obey our convenient renormalization condition . for µ > 2m.14. we will now compute (−iD )−1 to O(g 2 ) when k 2 − µ2 is order g2. but within those restrictions. In model 3.

[−iD (k 2 )]−1 = k 2 − µ2 + 2 2 i 2 n=|0 . Γ → 0. That is the pole on the second sheet dominated the behavior of D (k 2 ) near k 2 = µ2 . What we will do next does not depend on perturbation theory in the coupling constant. D (k 2 ) is still analytic on the cut complex plane. 13. November 25 Notes from Sidney Coleman’s Physics 253a 204 Now using that formula for Im π (k 2 ). and the actual value of D (k 2 ) along the real axis should be more and more dominated by the presence of this pole when k 2 ≈ µ2 . this pole is close to the physical region. although the way Γ appeared. Our perturbative analysis shows that as g → 0. This does not prove that Γ is a lifetime−1 . with µ > 2m. it is clear how a perturbative calculation would go in other models.2 from Nov. 4. We will now do two thought experiments and show that this behavior is what experimentalists are talking about when they say they have discovered an unstable particle. The |D |−2 serves to exactly eliminate the external propagators you would get in relating n|φ (0)|0 to n|(S − 1)|k ∝ ia. Compare with Eq.18. Our only assumption now will be that D (k 2 ) = i k2 − µ2 + µiΓ + small terms for some range of the real axis near k 2 = µ2 . What we have done so far has depended on perturbation theory in model 3. we have found that in model 3. To summarize what we have found so far.|k | n|φ (0)|0 |2 (2π)4 δ (4) (k − Pn ) [D (k 2 )]2 k2 =µ2 + O(g 4 ) = k − µ + iµΓ + O(g ) = k2 − µ − iΓ 2 2 4 + O(g 4 ) i The nice thing I have noticed in the next to last step is that what multiplies 2 in the first expression is µΓ when k 2 = µ2 . or on any model. That Γ was an inverse lifetime in the theory with a turning on and off function does not suffice to show that it is a lifetime in our fullblown scattering theory. In 2 a sense. in the small g limit. there is a pole in D (k 2 ) at k 2 = (µ − iΓ )2 on the second sheet. . and some ad hoc renormalization conditions.

November 25 Notes from Sidney Coleman’s Physics 253a 205 Our first thought experiment is to blast the vacuum at x = t = 0. A theorist blasts the vacuum by turning on a source L → L + ρ(x. σ(k 2 ) is in turn proportional to −Im π σ(k 2 ) = −(Im π (k 2 ))|D (k 2 )|2 So the probability of finding momentum k not equal to zero. t)φ (x. k 0 > 0 is proportional to −λ2 Im π (k 2 )|D (k 2 )|2 + O(λ3 ) Finally using the form of D which is assumed to dominate near k 2 = µ2 (π (k 2 ) = −µiΓ) we have λ2 µΓ + O(λ3 ) (k 2 − µ2 )2 + µ2 Γ2 Probability of finding an invariant mass of k2 u2 k2 .18. t) ρ(x. The amplitude that you’ll get any momentum eigenstate |n is proportional to λ n|φ (0)|0 + O(λ2 ) The probability of having momentum k in the final state is proportional to λ2 |n | n|φ (0)|0 |2 (2π)4 δ (4) (Pn − k) + O(λ3 ) = 2πλ2 σ(k 2 )θ(k 0 ) + O(λ3 ) In the sum. I don’t have to specify |n = |0 . |k (one meson) (as long as I stay away from k = 0 so δ (4) (Pn − k) = 0 when |n = |0 ) since there are no physical one meson states to emerge from blasting the vacuum when “the meson is unstable”. t) = λδ (4) (x) An experimentalist blasts the vacuum at x = t = 0 by crashing two protons together at the origin of coordinates.

the decay products’ COM energy. and the average lifetime is Γ−1 . the peak gets narrower and higher. The probability of finding a COM energy E in the decay products is proportional to (drop the O(λ3 )) µΓ µΓ = (E 2 − µ2 )2 + µ2 Γ2 (E − µ)2 (E + µ)2 + µ2 Γ2 µΓ ≈ 2 (2µ)2 + µ2 Γ2 (E − µ) µΓ = 2 (4µ2 ) (E − µ)2 + Γ 4 (Approximation preserves the character of the function if Γ µ) Probability of finding energy E in the decay product COM frame P max P max 2 u E This is called a Breit-Wigner or Lorentzian line shape. which locate the pole on the second sheet of D (k 2 ) are the mass and decay width respectively that an experimenter reports when she says she has found an unstable particle. So we have shown that µ and Γ. or decay width. We will now do a second thought experiment to show that this second way of determining Γ is equivalent. we can think of this probability distribution as a function of E. .18. That is. instead of as a function of k 2 . They use a clock. As Γ gets smaller. and it is familiar from QM. Experimenters have another way of measuring Γ. November 25 Notes from Sidney Coleman’s Physics 253a 206 We can look at the center of mass energy of the decay products. which is purported to be equivalent. Γ is the full width at half maximum.

The region of production cannot be too sharply localized as we are going to make states only with k 2 ≈ µ2 . December 2 Notes from Sidney Coleman’s Physics 253a 207 19 December 2 We have explained “width” in the phrase “decay width”. we’ll produce an unstable particle near the origin and detect it a long ways away at some region near y.19. f (x) is fairly ˜ well localized in position space and its Fourier transform (f (k) = d4 x eik·x f (x)) is fairly well localized in momentum space about a momentum k. In thought experiment 2. t x We’ll make the initial state by hitting the vacuum with d4 x f (x)φ (x). Initial state is d4 x f (x)φ (x)|0 f(x) f(k) x k k I’ll detect the particle by finding the amplitude that this state becomes the state d4 x g(x − y)φ (x)|0 . With a second thought experiment we’ll explain “decay”.

˜ 2 Assume that D (k ) is dominated by a stable or unstable particle pole at k 2 ≈ µ2 and that ˜ f (k) is sufficiently tightly concentrated around k that we can make the approximation A(y) = i d4 k −ik·y ∗ ˜ e g (k)f (k) 2 ˜ (2π)4 k − µ2 + iµΓ (The stable case is handled by taking the limit Γ → 0+ ). which is difficult because the phase of the exponential is varying rapidly as k changes. the phase of the .T. December 2 Notes from Sidney Coleman’s Physics 253a 208 g(x-y) g(k) y x l k g(x − y) is concentrated around y. A(y) = = = d4 xd4 x g ∗ (x − y)f (x) 0|T (φ (x )φ (x)|0 d4 xd4 x d4 k d4 k ik·(x −y) ∗ ˜ e g (k)e−ik ·x f (k ) 0|T (φ(x )φ (x))|0 ˜ (2π)4 (2π)4 d4 xd4 x eik·x e−ik ·x G (x . Furthermore as k 2 increase through µ2 . unless k ≈ l ˜ Recall that g (k) is concentrated around l and f (k) is concentrated around k. with k 2 ≈ µ2 .19. g (k) is concentrated around l.k )≡(2π)4 δ (4) (k−k )D (k2 ) Compare this F. param˜ eterized by y is A(y) = 0| d4 x g ∗ (x − y)φ (x ) d4 xf (x)φ (x)|0 If the point y is far later in time than the origin. we can make this a time ordered vacuum expectation value with negligible error. convention with the one for f on the previous page = d4 k −ik·y ∗ ˜ e g (k)f (k)D (k 2 ) ˜ (2π)4 This integral gives zero. The amplitude. x) d4 k d4 k −ik·y ∗ ˜ e g (k)f (k ) ˜ (2π)4 (2π)4 ˜ G (−k. We want to analyze this for large y.

December 2 Notes from Sidney Coleman’s Physics 253a 209 propagator changes rapidly. At those points ti . except at those points where the phase stops wildly varying for a moment. (See Whittaker and Watson Modern Analysis). ∞ i A(y) = 0 d s 2µ s d4 k −ik·y+i 2µ (k2 −µ2 +iµΓ) ∗ ˜ g (k)f (k) ˜ e 4 (2π) . or take g = f . but it can also be done by the method of stationary phase. Unfortuk 2 − µ2 + iµΓ nately we introduce another integral. phi(t) 1000 t1 t2 t -1000 Then the integral gives nothing almost everywhere. where φ is a rapidly varying function of t except for a few points = 0. There is a method for handling these kinds of integrals. call them ti . Method of stationary phase Given I = where dφ dt dteiφ(t) f (t). ∞ s i ds i 2µ (k2 −µ2 +iµΓ) = e 2 − µ2 + iµΓ k 2µ 0 We will rewrite ˜ Assume the variation in phase of g ∗ (k)f (k) is slow compared to the variation in the ˜ exponential. the contribution can be approximated by I= i 2π eiφ(ti ) f (ti ) dt e 2 φ i (ti )(t−ti )2 eiπ/4 ti |d2 φ/dt2 | as an exponential so we can use this method.19.

Call y 2 “s0 ”. There is a factor s0 .19. s dk µ 2 ∞ s µ A(y) ≈ − 0 ds 1 µ 2 s 2µ (2π) 2 g∗ ˜ µ 2 s µ ˜ µ y f y ei 2 (−µ+iΓ) e−i 2s y s s To do the s integration by stationary phase. In a proper time s. set 0= s µ ∂ − µ − y2 ∂s 2 2s s= y2 Note that there is no stationary phase point if y 2 is spacelike. where µ 2 vµ = 1 ⇒ s = y 2 . December 2 Notes from Sidney Coleman’s Physics 253a 210 To do the k integration by stationary phase. (proper time)3/2 . then d2 φ ds2 = s0 µ . So the amplitude falls like 3/2 . As y → ∞. That is there because if you wait long enough. Suppose you classically propagate a stable particle kµ with velocity vµ = . The Lorentz invariant gent t 1 eralization of this is that the amplitude at the center of the packet falls like . every packet because of an initial uncertainty in velocity is spreading out in all directions linearly with time. set 0= ∂ s 2 s −k·y+ (k − µ2 ) = −y µ + k µ ∂kµ 2µ µ = k0 d2 φ µ Thus the only stationary phase point in the k integrations is at k0 = y. there is no probability that a particle will be detected. This is just classical kinematics. In 3-D this means that the probability density at the center 1 1 of the packet goes down like 3 . it will arrive at a point yµ = vµ s. We have recovered causality. and s0 µ s0 2 A(y) = −eiπ/4 2πs0 1 1 µ 2µ (2π)2 g∗ ˜ Γs µ ˜ µ y e−iµs0 e− 20 y f s0 s0 These factors can be understood. The factor e−iµs0 is just e−iEt of quantum mechanics that has come out in a Lorentz −3/2 invariant generalization. y 2 < 0. but you see we have recovered it in the limit of large y from quantum field theory. The conditions of stationary phase are the equations of classical kinematics.

It should be possible to do the computation in the frame where y = (y. and we would start adding interactions. The actual order we have been doing this course. Under Lorentz transformations a set of scalar fields transform like φa (x) a = 1. 1) Λ : φa (x) → φa (Λ−1 x) The only Lorentz scalars you can construct have derivatives. . 0). ∂µ φ2 ∂ µ φ3 µνλσ ∂µ φ1 ∂ν φ2 ∂λ φ3 ∂σ φ4 The list of possible quadratic Lagrangians is pretty short. Γ is the decay rate per unit proper time. to the Lagrangian. we would have begun the course by listing all possible field transformation laws. In fact we have even gone a long ways toward studying the total list of interacting scalar fields since the renormalization vs infinities arguments pretty well rule out Lagrangian that are more than quartic in the fields. scalar fields. ¿¿ This talk of “correct generalization” must be made more precise. At this point we would have all possible free particle theories. which act on the scalars and are completely contracted with g µν or µνλσ . but we are now going to go on to DISCOVERING ALL POSSIBLE LORENTZ TRANSFORMATION LAWS OF FIELDS . We have e−Γs0 /2 in the amplitude. which means that the probability has a factor e−Γs0 . higher order polynomials. December 2 Notes from Sidney Coleman’s Physics 253a 211 Finally there is the unstable case. . then we would have constructed all possible quadratic Lagrangians. . However our stationary phase computations are not justified in that case ?? “WHERE IT BEGINS AGAIN” If we had been proceeding logically. like the Hamiltonian not being bounded below.19. and we have gone a long ways toward exploring them. only after exhausting those. . . that is all possible combinations that are at most quadratic in the fields and transform as scalars under the Lorentz group. making the most general statements about relativistic quantum field theory we could and. Then we would do canonical quantization and in the process discard many of the Lagrangians because of one or another inconsistency. made simplifying assumptions and approximations. 0. . They are indeed decaying and again we have gotten the Lorentz invariant generalization of e−Γt . starting from first principles. 2. n Λ ∈ SO(3. We haven’t exhausted the study of scalars. . 2φ. is to spend a lot of time studying relativistic invariants made up of the simplest kind of fields. of them. 4.

i. Many elements of the group can map into a single D matrix. . U (Λ1 )U (Λ2 ) = U (Λ1 Λ2 )eiφ(Λ1 . It only need be true up to a phase. More concretely. that is it is possible that D(Λ) = D(Λ ) while Λ = Λ . If we think of the D’s as matrices and the φ’s as column vectors. Λ can also be thought of as a 4 × 4 matrix. for all x. that preserves the metric. det(Λ) = 1. is proper. If D(Λ) = D(Λ ) =⇒ Λ = Λ . and you might think from any set of D’s you could reconstruct the group. the representation is “faithful”. Λµ Λν g αβ = g µν . It turns out that for SO(3) and SO(3. “The dimensions of D is N ”) A property of the U ’s reflects itself in the D’s (It only takes a couple of lines to prove this) U (Λ1 )U (Λ2 ) = U (Λ1 Λ2 ) =⇒ D(Λ1 Λ2 ) = D(Λ1 )D(Λ2 ) Also U (1) = 1 =⇒ D(1) = 1 and from these two properties of the D’s. the phase can be removed except in representations . . For each α β 0 Λ there is a unitary transformation U (Λ)† φa (x)U (Λ) = Da b (Λ)φb (Λ−1 x) (Σb implied) For each Λ there is some N × N matrix Da b (Λ) that gives a linear relationship between the complete set of commuting observables at Λ−1 x and those at x. the possibility which is impossible to rule out in quantum mechanics that U (Λ1 )U (Λ2 ) = U (Λ1 Λ2 ). 1). and is orthochronous.e. 1). . Λ0 > 0.19. A person who is tired of group theory is tired of life. December 2 Notes from Sidney Coleman’s Physics 253a 212 We’ll phrase the analysis in a quantum language. An additional complication that we are only going to consider in a very cavalier way. You can’t though. but the all that we are about to do can be carried through classically.Λ2 ) The product law is not exactly true in general. It seems that the D matrices obey all the properties of the group. the part of the Lorentz group connected to the identity. D(Λ−1 ) = D(Λ)−1 . N Let Λ denote an abstract element of SO(3. . A set of D’s that obey the group laws is called a representation. Assume we have a finite number of fields. φa (x) a = 1. we can write U (Λ)† φ(x)U (Λ) = D(Λ)φ(Λ−1 x) (D’s are N × N matrices. The trivial prototypical example is (all fields are scalar) D(Λ) = 1 for all Λ.

If there is no such S. and it is in the basement of Cabot. Vol. The Lorentz transformation of parity is not in SO(3. V.19. If the fields transform as Λ : φ(x) → φ(Λ−1 x). because it is not connected to the identity. 59.. 1). 1). so you would expect to have gotten 1. it is some definite invertible matrix. the new basis φ (x) = Sφ(x) transforms as Λ : φ (x) → D (Λ)φ (Λ−1 x). (You can think of parity. Equivalent to this though is just a redefinition of the basis fields. (If you want to study this the only good reference I know of is Bargmann. Try as you may. I can make a new representation of SO(3.) For every Λ ∈ SO(3. P . I am not recommending this however. Our catalog will only include inequivalent representations of SO(3. 1) Suppose I have a representation D(Λ).) The possibility that the product of the unitary operators is only the unitary operator of the product up to a phase reflects itself identically in the composition law for the D’s. 1). that obeys all three conditions. Define two representations D and D to be equivalent if there exists an invertible S such that D (Λ) = SD(Λ)S −1 for all Λ and write D ∼ D. 1).” Annals of Mathematics. 1).(1954) p. D and D are inequivalent D ∼ D. you cannot use the unitary of the U ’s to prove unitary of the D’s). We will get all the right results with much less effort by being cavalier and lucky. December 2 Notes from Sidney Coleman’s Physics 253a 213 called spinor representations where a rotation by π about any axis e followed by another rotation by π about e gives U ( eπ )U (eπ) = −1 notation for the unitary operator that rotates around e by π The rotations by π are physically equivalent to no rotation at all. This is not worth listing as a new kind of field theory. Shortening the catalog of finite dimensional representations of SO(3. by defining D (Λ) = SD(Λ)S −1 where S does not vary with Λ. Our task of finding all possible Lorentz transformation laws of fields has been reduced to the task of making a catalog of all finite dimensional representations of SO(3. one representation from each “equivalence class. (Another thing to note about the D’s is that they are not in general unitary. “On Unitary Ray Representations of Continuous Groups.” I’ll give a useful example. it has determinant −1. I can obtain another element of . abstractly or as a 4 × 4 matrix.1.

define Dp (Λ) = D(Λp ) This new rep obeys all three conditions. Λp Λp = (ΛΛ )p . say the two index tensor. I did the two index tensor because it is a little less trivial) Shortening the catalog of finite dimensional inequivalent representations of SO(3. If it is interpretable as a change of basis. Starting with a representation D. we must be able to find the similarity transformation relating the two representations. I can construct a new representation of the group from any given representation. 1). .e. let’s look at what the parity automorphism does to one of the representations of SO(3. . . T 0j → −T 0j . December 2 Notes from Sidney Coleman’s Physics 253a 214 SO(3. The association is one to one. Λ : T µν → Λµ α Λν β T αβ (Transformation law of a tensor) We read off Dµν αβ (Λ) = Λµ α Λν β . 1) Suppose someone has two theories. and it preserves the group multiplication law. It may or may not be true that Dp ∼ D. . These properties make it an “automorphism”. To make the example more concrete. 1). Let’s massage the expression for Dp until we find it. 1) we all know and love. N1 . After all what parity is really doing is just turning T 00 → T 00 . one with a set of fields φ1a transforming as Λ : φ1 (x) → D(1) (Λ)φ1 (Λ−1 x) a = 1. Dp (Λ) = SD(Λ)S −1 (you can do the vector case. 1) this way.19. With this automorphism of SO(3. Dp µν αβ = (P ΛP )µ α (P ΛP )ν β = P µ σ Λσ τ P τ α P ν φ Λφ ψ P ψ β = P µ σ P ν φ Λσ τ Λφ ψ P τ α P ψ β = S µν σφ D(Λ)σφ τ ψ S −1τ ψ αβ S µν σφ ≡ P µ σ P ν φ i. The D’s in the representation induced by the parity automorphisms are Dp µν αβ (Λ) = Dµν αβ (Λp ) = (P ΛP )µ α (P ΛP )ν β Now we expect that we have not constructed an inequivalent representation of SO(3. Λp ≡ P ΛP (P = P −1 ). T ij → T ij and T i0 → −T i0 .

The new theory would have to be five dimensional. .19. and they transform like Λ : φ(x) → D(Λ)φ(Λ−1 x) D(Λ) = D(1) (Λ) 0 0 D(2) (Λ) a = 1. The D’s are in block diagonal form. the catalog of finite dimensional inequivalent irreducible representations of SO(3. peculiar to living in 3+1 dimensions. It is specified by giving an axis of rotation. which he has assembled into a vector φ. otherwise. 1) to the problem of finding the representations of SO(9). D = D(1) ⊕ D(2) . December 2 Notes from Sidney Coleman’s Physics 253a 215 and the other with a set of fields φ2a transforming as Λ : φ2 (x) → D(2) (Λ)φ2 (Λ−1 x) Now this person comes to you and says. Actually what is constructed there are the representations of the Lie algebra of SO(3) rather than the representations of the Lie Group but you’ll see that is what we want. . Let’s standardize . e and θ. which 9 + 1 dimensional students solve as undergraduates. the representations of SO(3. there would be no quick reduction of the problem of finding the representations of SO(9. You know all about the representations of SO(3) from undergraduate QM. THE FINITE DIMENSIONAL INEQUIVALENT IRREDUCIBLE REPRESENTATIONS OF SO(3) An element R of SO(3) can be thought of abstractly or as a 3×3 matrix. . It is such a simple extension of what was previously known that it is not worth including in our catalog. They are carefully constructed in a few pages in a way that generalizes to other groups beginning on page 16 of Howard Georgi’s Lie Algebras in Particle Physics. Our task is to build the remarkably shorter catalog. so we will be able to wrap the catalog up by the end of next lecture. By a wonderful fluke. Call a representation “reducible” if it is equivalent to a direct sum. It has dimension N1 + N2 . theory 1 could contain a vector and theory 2 could contain a scalar. 1). This is hardly a big breakthrough. call it “irreducible”. 1) can be rapidly obtained from the representations of SO(3). I have a new theory with N1 + N2 fields. N2 For example. . We will review the representations of SO(3) just enough to refresh your memory. and an angle of rotation about the axis. Define this representation D to be the “direct sum” of D(1) and D(2) . If we lived in 9 + 1 dimensions.

Each point on the equator is still identified with one other point on the equator.] 26 . The projective 2 sphere is S 2 with each pair of antipodal points identified. the ball B 3 ) in three space of radius π. the ball just described. except antipodal points on the surface of the ball are identified. This still include twice the rotations by π because R(eπ) = R(−eπ) The group SO(3) is topologically like the ball (not the sphere S 2 .19. with antipodal points on the ball identified. Its length gives the amount of rotation in the counterclockwise direction when looking down toward the tail of the vector e. π].26 I can explain that in one lower dimension where I can visualize it. But it is clear that (x’s represents a pair of identified points) Just flatten the sphere out into the disk and this disk is the “ball” in two space. The product eθ defines the rotation completely. A method of infinitesimal analysis d R(eθ)|θ=0 ≡ −ie · J dθ [BGC note: The standard name for these spaces are real projective 3-space RP3 and real projective 2-space or the real projective plane RP2 . So we’ll restrict θ ∈ [0. The reason being is that a rotation about e by an angle θ is exactly the same as a rotation about −e by an angle 2π − θ. If we let θ take on any value from 0 to 2π we have included twice every element of SO(3). Chuck the whole bottom half of the sphere leaving the top half and the equator. Each point on the whole bottom half of the sphere below the equator has a point it is identified with the half of the sphere above the equator. December 2 Notes from Sidney Coleman’s Physics 253a 216 the vector that defines the axis by taking it to be a unit vector. is topologically like the projective 3 sphere. with the identification. Just so you have some impressive jargon at your disposed. R(eθ).

It is also unchanged by the LHS because R turns it into e. Thus the LHS is a rotation by θ about R −1 e. Applying the infinitesimal analysis to a representation.19. This loss of information decreases the restrictions on the form of the representations. Assume all the differentiability you desire. even though the formalism hasn’t explicitly included them. that is. D(R(eθ ))D(R(eθ)) = D(R(e[θ + θ ])) Take d d d . Rather than making mathematical statements about 3 × 3 matrices. a set of three 3 × 3 matrices if you think of R as a 3 × 3 matrix. Property (1) is obvious. we’ll take derivative with respect to θ and θ . There are three because SO(3) is a three parameter group. What this derivative is is a directional derivative in the direction e at the center of the ball and −iJ is the gradient. We won’t actually use all the information contained in (1) and (2). which is unchanged by R(eθ) and R −1 turns it back into R −1 e. (1). R −1 R(eθ)R = R(R −1 eθ) These are physically motivated statements about properties of rotations. R(eθ )R(eθ) = R(e[θ + θ ]) (2). we’ll put in two properties physically that are enough to specify the group. and is the reason we pick up representation up to a phase. The D’s are some representation of dimension N and the three J’s are some N × N matrices. and set θ = 0 to get (on the RHS = ) dθ dθ dθ −ie · J D(R(eθ)) = d D(R(eθ)) dθ . (1) can be made into a statement about representations. How do you see that this derivative is linear in e? Recall the 2 picture of SO(3) as a ball. (SO(n) is an n(n−1) parameter group). R −1 e is unchanged by the RHS of (2). December 2 Notes from Sidney Coleman’s Physics 253a 217 This expression defines J. The three matrices are called the Lie algebra of SO(3). and something more abstract if you think of R more abstractly. How to see property (2)? A way of characterizing the axis of rotation is to say it is the axis such that any vector parallel to this axis is unchanged by the rotation. We will only use them in infinitesimal form. Someplace in this analysis we have to put in the properties of SO(3). Take d D(R(eθ))|θ=0 ≡ −ie · J dθ (The J’s are the “generators” of this representation) This is a very concrete equation.

If D(R) = D(1) (R) 0 0 D(2) (R) then J = J (1) 0 0 J (2) and if D ∼ D(1) ⊕ D(2) then J ∼ J (1) ⊕ J (2) The task of finding inequivalent irreducible finite dimensional representations D of SO(3) has been reduced to the task of finding inequivalent irreducible sets of 3 matrices J. What about irreducibility? If a representation is reducible. (2) is D(R −1 )D(R(eθ))D(R ) = D(R(R −1 eθ)) d Take i dθ at θ = 0 to get D(R −1 ) e · J D(R ) = (R −1 e) · J = e · R J (the last equality using the fact rotation matrices preserve scalar products) Dropping the primes.e.J ∼ J Equivalence of two reps is the same as equivalence of their generators. I can even phrase this a little more strongly. whose properties we are considering. As a statement about representations. Reducibility of a rep is the same as reducibility of its generators. So e · J = i dθ D(R(eθ))|θ=0 has block diagonal form. then it is equivalent to a representation that d has block diagonal form for all rotations. and using the fact that e is an arbitrary unit vector. the 3 J’s (N × N matrices) transform like a vector. The solution with the boundary condition D(R(e0)) = 1 is D(R(eθ)) = e−ie·Jθ (The J’s “generate the representation”) Now we can transfer our definitions about inequivalence and irreducibility to the generators. If the rep D is generated by J and the representation D by J and D ∼ D that is D(R) = SD (R)S −1 for some S and all R then Se−ie·J θ S −1 = e−ie·Jθ ⇐⇒ S J S −1 = J i. . this says D(R−1 )JD(R) = RJ which is the statement that the generator of the rotations. December 2 Notes from Sidney Coleman’s Physics 253a 218 This is a simple differential equation.19.

. . that for small θ. −s + 1. . December 2 Notes from Sidney Coleman’s Physics 253a 219 You can go further by writing R. . Every representation is equivalent to a hermitian representation. . s in a usual basis for the 2s + 1 dimensional vector space the J (s) act on. Jy ] = iJz also can get [Ji . · · · 2 2 1 “spin” s (1) J (0) = 0 J ( 2 ) = (s) Jz |m = m| m no sum σ 2 (Ji )j k = −i ijk m = −s. Facts about finite dimensional inequivalent irreducible representation of the Lie algebra of the rotation group A complete set of them is the J (s) 3 1 s = 0. . They satisfy the same commutation relations.19. 1. Jj ] = i ijk Jk The generators J form a representation of the Lie algebra of the group. s − 1. −s + 2. . parametrizing R by e and θ (it shouldn’t cause confusion to use these variables again). What we have found is eiJ·eθ Je−ie·Jθ = J + θe × J + O(θ2 ) where on the RHS I have used a physical property of the rotation group. a rotation matrix acting on a vector changes it by θe × V + O(θ2 ) (use the RH rule to make sure this agrees with the convention and picture before). J] = −ie × J Take e = ex and look at y component to get ˆ [Jx . s − 2. Take d −i dθ of this equation to get [J · e. The J (s) are hermitian.

with an orthonormal basis |i . dim D(s) = 2s + 1 (3). like i. D(g). they are double valued D(s) (R(2πe)) = (−1)2s 1 (2). It is intuitively clear that a matrix of numbers acting on a column vector. Direct product of representations. and they are not unitary. they generate the reps up to a phase. the notation of bras and linear op(s) erators. A new notation. no sum implied) appeared out of the blue. The reps of the Lie algebra just listed not only generate the reps of the rotation group. Now let me define a linear operator. but I would like to make the connection precise. . The hermiticity of the J (s) implies the D(s) are unitary (D(s) (R−1 ) = [D(s) (R)]−1 × [D(s) (R)]† ). that is. The − sign is present because of the i in the exponential. g ∈ G. can be reinterpreted as a linear operator acting on a vector space. . D(g) is defined by D(g)|i = |j j|D(g)|i = |j Dj i (g) j j . the D’s preserve x2 + y 2 + 49 z 2 . Furthermore D(s) ∼ D(s)∗ and D(s) (R(eθ)) = e−ie·Jθ (19. D∗a b (g) = (Da b (g))∗ (no matrix transpose). It may or may not be equivalent to the representation you obtained it from. Suppose I have an n dimensional representation. In SO(3) since there is only one inequivalent irreducible representation of a given dimension. I can define a new representation. More specifically. (I am sorry this notation also been used for the matrix although we shall see why the ambiguity is small). i = 1.1) implies J (s) ∼ −J (s)∗ . 7k for the 1 space D(1) acts on. The half integers s are reps up to a phase. in dumb bases. Every representation of the rotation group is equivalent to a unitary representation of the rotation group. . which will act on an n dimensional vector space.19. n. If I have any representation of any group G. (4). an n×n matrix for every element of a group. . Di j (g). j. December 2 Notes from Sidney Coleman’s Physics 253a 220 Facts about finite dimensional inequivalent irreducible representation up to a (s) phase of the rotation group. D(s) (R(eθ)) = e−ie·J θ (1). (Jz |m = m|m . a rep must be equivalent to its associated complex conjugate rep. The integer s are representations. Of course. the new rep D∗ is irreducible. Da b (g). g ∈ G. (5). If D is irreducible. This new representation has the same dimension as the original representation.

because we also write the composition law of the matrices as D(g)D(g ) = D(gg ) Now we see that is true for the matrices and the abstract linear operators. We might have gotten D(g)D(g ) = D(g g) for the operator composition law. . First to define tensor product space. . December 2 Notes from Sidney Coleman’s Physics 253a 221 The intermediate step was motivational. you can recover the matrix Dj i (g) from the abstract linear operator.19.k |k Dk j (g)Dj i (g ) |k Dk i (gg ) k = = D(gg )|i Therefore D(g)D(g ) = D(gg ). i = 1. I define a ⊗ |j 2 . . implies about our new abstract linear operator D(g)D(g )|i = D(g) j |j Dj i (g ) = j. j|D(g)|i = j|( k |k Dk i (g)) = k δjk Dk i (g) = Dj i (g) Let’s see what the composition law Di j (g)Dj k (g ) = Di k (gg ) j which must be satisfied by a representation. . j ≡ |i 1 j = 1. If I have two vector spaces V1 with a basis |i 1 . There was some danger this was not going to work out. . d2 . the definition is D(g)|i ≡ j |j Dj i (g) From the definition. . In this spiffy notation. . I’ll define the tensor product of two representations. . and the fact that the basis is orthonormal. It is nice to see this work out. d1 and V2 with a basis |j 2 . vector space V1 × V2 with a basis |i.

I can define a linear operator on V1 × V2 called A ⊗ B by (A ⊗ B)|i. A. V = V1 = V2 and D = D = D(2) . B. P+ P− = 0. and a linear operator on V2 . also P+ + P− = 1. Given a linear operator on V1 . 2 |i. that is d = d1 = d2 .19. What I need to show is that (D ⊗ D)(g)P± = P± (D ⊗ D)(g) . If the two representations are the same. j + |j. P− |sym = 1 2 2 2 is put in so that P+ = P+ and P− = P− . we can define a new representation denoted D(1) ⊗ D(2) acting on V1 × V2 : D(1) ⊗ D(2) (g) = D(1) (g) ⊗ D(2) (g) The new representation may or may not be reducible. The trick is that in this case it makes sense to talk about (1) i √ 2 |i ⊗ |j + |j ⊗ |i 1 and √ |i ⊗ |j − |j ⊗ |i 2 and 1 √ 2 1 (or in another notation √ |i. The elements of the symmetric part of the basis satisfy P+ |sym = |sym . i ). i 2 d(d+1) . j |A ⊗ B|i. You can get fancy by defining projection operators 2 1 P+ (|i ⊗ |j ) ≡ (|i ⊗ |j + |j ⊗ |i ) 2 and 1 P− (|i ⊗ |j ) ≡ (|i ⊗ |j − |j ⊗ |i ) 2 d(d+1) 2 (The 0). j = (A|i 1 ) ⊗ (B|j 2 ) or i . j = 1 i |A|i 12 j |B|j 2 Now if we have a representation D(1) acting on V1 and D(2) acting on V2 . You can think of this as a new basis. we can show that D ⊗ D is reducible (except in the case d = 1). I’ll show that D ⊗ D is reducible. The number of basis elements of the first type is the number of the second is d(d−1) . With these projection operators. j − |j. December 2 Notes from Sidney Coleman’s Physics 253a 222 The dimension of V1 × V2 is d1 · d2 .

di . where di is the dimension of the ith subspace). . If you have a set of projection operators Pi . so I’ll make the connection precise and I’ll try to phrase the connection so that you can see a representation is reducible whenever you have a set of projection operator like P+ and P− . Let’s call the rep D. α = 0 j=i The big assumption about these projection operators is that they commute with the representation operators. (i has nothing to do with direct products. Pi D(g) = D(g)Pi for all i and g Let’s look at the matrix associated with D(g) in this basis and see what we can show . (In the example of importance. . m Pi2 = Pi Pi = 1 i Pi Pj = 0. while the other n(n−1) basis vector unaffected by P− and annihilated by P+ ) 2 I’ll write the basis vector as |i. It may not be obvious to you that this shows that the representation is reducible. α = |i. . . just a way of labelling the basis. . since our definition of reducibility was in terms of matrices. i = j then I can choose the basis of the vector space they act on so that it breaks up into bases for various subspace that are either annihilated or unaffected by the Pi . . December 2 Notes from Sidney Coleman’s Physics 253a 223 It’s just a matter of using definitions (somehow math proofs are always just a matter of using the definitions although it is usually beyond me to do it) 1 (D ⊗ D)(g)P± |i ⊗ |j = (D ⊗ D)(g) |i ⊗ |j ± |j ⊗ |i 2 1 = D(g)|i ⊗ D(g)|j ± D(g)|j ⊗ D(g)|i 2 = P± D(g)|i ⊗ D(g)|j = P± (D ⊗ D)(g)|i ⊗ |j That does it. . α Pi |j. i = 1. there are n(n+1) bases vectors unaffected by P+ and anni2 hilated by P− . and α = 1. commuting with it.19. α . where Pi |i. .

. α|D(g)Pj |j. If the n(n+1) dim block is reducible 2 into m irreducible components. β = i.19. D (m) . If the n(n−1) dimension block is reducible into 2 m irreducible components. j. the matrix D(g) is block diagonal. December 2 Notes from Sidney Coleman’s Physics 253a 224 about it. .) This is the statement that in this basis. . α|D(g)|j. β = i. by finding P+ and P− that commute with D ⊗ D. α|Pj D(g)|j. I’ll finally state the fifth fact about the rotation group.   d1 × d1   d2 × d2   . α|D(g) k Pk |j. So after a multipage rambling explanation of tensor product. each of these representations is said to be in the symmetric part of the tensor product.. each of these is said to be in the antisymmetric part of the tensor product.       . Tensoring two irreducible reps together D(s1 ) ⊗ D(s2 ) ∼ D(s1 +s2 ) ⊕ Ds1 +s2 −1 ⊕ Ds1 +s2 −2 ⊕ · · · ⊕ D|s1 −s2 | s1 +s2 =⊕ s=|s1 −s2 | D(s) Tensoring two identical reps togethers D(s) ⊗ D(s) = D(2s) ⊕ D(2s−1) ⊕ D(2s−2) ⊕ · · · ⊕ D(0) . . . α and β but that doesn’t matter.. . dm × dm In the example of importance. p ∝ δji (The proportionality constant depends g.   . Diα jβ = i. D (1) . . . D(1) . we have shown that D ⊗ D is equivalent to [(D ⊗ D)iα jβ (g)] = Some n(n+1) 2 × n(n+1) matrix 2 0 n(n−1) 2 0 × n(n−1) 2 These two blocks may or may not be further reducible. i.. D(n) . β = i.

December 2 Notes from Sidney Coleman’s Physics 253a 225 where D(2s) . . · · · are in the symmetric part. D(0) is symmetric if s is an integer. D(2s−2) . · · · are in the antisymmetric part. D(2s−1) . 2 They just alternate.19. antisymmetric if s is a 1 integer.

φ. We’ll show that SO(3. it just adds. 1) and when R is a 3 × 3 orthogonal matrix. A Lorentz transformation is called a rotation if it takes t→t and x → Rx We’ll denote such a Lorentz transformation by R and you’ll have to understand from context when R ∈ SO(3. A(ex φ). December 4 Notes from Sidney Coleman’s Physics 253a 226 20 December 4 Parametrizing the connected homogeneous Lorentz group The rotation group was parametrized by a direction e and an angle θ. 1) can be parametrized by e. velocity parameter.1) We also have the formula (which can be verified with a little algebra) (the algebra involves using formulas for sinh(φ1 + φ2 ) and cosh(φ1 + φ2 )). Now to prove that any Lorentz transformation can be uniquely decomposed into a rotation followed by a boost. .3).20. by showing that any Lorentz transformation can be decomposed into a rotation and a boost.2) This is why the velocity parameter is such a useful parameter for boosts. The proof is by construction and the construction is unambiguous. In general A(eφ) : t → t cosh φ + e · x sinh φ x → e t sinh φ + x + e (cosh φ − 1) e · x (This generalization is forced upon you by Eq. A boost in the x direction by an “angle”. and another direction and “angle”. θ. (20.T.) To go along with the formula R(e θ )R(e θ) = R(e [θ + θ ]) (20. restricting 0 ≤ φ < ∞ to avoid parametrizing each boost more than one way. has taken t → t cosh φ + x sinh φ x → x sinh φ + x cosh φ y→y z→z We’ll denote such a L. A(eφ )A(eφ) = A(e[φ + φ ]) (20.

Starting with a general Lorentz transformation Λ consider its action on the vector e0 ≡ (1.1) and (20. which implies uniqueness. The key thing about Lorentz transformations is that they leave the length of a vector unchanged. 0). φ > 0. connected to the identity preserve this when the vector is timelike.20. it is A−1 (eφ) A−1 (eφ) Λ : e0 → e0 This means this product is some rotation. αe) where α. e sinh φ) This determination of an “angle” φ and a direction e allows me to (uniquely) read off the boost that will bring e0 back to rest. then γ = cosh φ and the most general thing that e0 can transform into under a Lorentz transformation is Λ : e0 → (cosh φ. dθ d D(A(eφ))|φ=0 ≡ −iM · e. L is playing . formulas like Eqs. 1). In a rep D it implies d D(R(e θ))|θ=0 ≡ −iL · e.T. (20. December 4 Notes from Sidney Coleman’s Physics 253a 227 it has no freedom. we must have Λ : e0 → (γ. you know the inequivalent irreducible reps up to a phase of SO(3. dφ D(R(e θ)) = e−iL·e θ D(A(eφ)) = e−iM ·eφ Just as for SO(3). A−1 (eφ) Λ = R Λ = R A(eφ) Just like when we were working with SO(3). if you know the inequivalent irreducible reps of L and M . Since this vector has time component > 0 and since all L. γ real and greater then zero but otherwise unknown and e is some unit vector. 1). so we know there is a restriction on γ and α γ 2 − α2 = 1 and γ > 0 √ =⇒ γ = 1 + α2 Let’s rename α = sinh φ. the generators of SO(3.2) show that rotations and boosts can be written as exponentials. call it R.

D(R) = RM (20.3) (These are just like the SO(3) arguments. So from the properties of rotations. the commutator of two boost generators must be a linear combination of a boost generator and a rotation generator. D(R−1 ) e · M D(R) = R−1 e · M which implies D(R−1 )M. If the Lie algebra of SO(3. Mj ] is a two index antisymmetric tensor under rotation. Lj ] = i ijk Lk ( implied) k (that came from R −1 R(eθ)R = R(R −1 eθ)) A property of the Lorentz group is that R−1 A(eφ)R = A(R−1 eφ) (can be used to get the general boost from a boost in the x direction) (you can convince yourself that both sides are boosts by φ along R−1 e) Out of this (take d i dφ ) comes the statement (applied to rep D). which transform like vectors. 1) is going to close. in fact we would have reused J if it weren’t that it is conventionally used for something else. Therefore [Mi . we have [Li . The most general thing I can make that is a two index antisymmetric tensor that is linear in the boost and rotation generators.20. December 4 Notes from Sidney Coleman’s Physics 253a 228 the exact same role as J did in our discussion of SO(3). is α ijk Mk + β ijk Lk . Mj ] = i Using D(R)−1 Mi D(R) = Rij Mj 3×3 orthogonal rotation matrix ijk Mk ( j implied) It is easy to see that D(R)−1 Mi Mj D(R) = Rik Rjl Mi Mj that is Mi Mj transforms like a two index tensor under rotations. so I haven’t written them in detail) which implies after a little more work [Li .

December 4 Notes from Sidney Coleman’s Physics 253a 229 Therefore it must be that [Mi . but you can think about it either way) that have these commutation relations – we want a complete set of inequivalent irreducible ones. P RP = R. Define 1 J (±) = (L ± i M ) 2 no i would be here if it were SO(4) we were studying L = J (+) + J (−) −iM = − J (+) − J (−) This − sign is obvious You can verify from the commutators of the L’s and M ’s that [Ji .20. you can check from the definitions. Now for a rotation. Jj ] = i and [Ji . we get a new representation from the one we were working with by defining DP (Λ) = D(ΛP ). Mj ] = i ijk [αMk + βLk ] Still more evasive reasoning shows that α = 0. From the commutation relations of the Lie algebra of any group. Jj ] = 0 The J (+) ’s and J (−) ’s form two independent. Lj ] = i ijk Lk is OK. [Li . In fact with a fair amount of work. Li → Li . Mj ] = −i ijk Lk . and we will not have to go through that method. so DP (R) = D(R). a miracle occurs. [Li . The commutation relations have to still work under this transformation. Fortunately. Using the Parity automorphism introduced on December 2. What we want to find is matrices (my brain is better equipped to think about matrices than abstract linear operator. (The − sign would not be present if this were SO(4) instead of SO(3. The generator of rotations in Dp are the same as those in D. −Mj ] = i ijk (−Mk ) is OK but [−Mi . −Mj ] = i ijk [α(−Mk ) + βLk ] is OK only if α = 0. 1)). [Mi . Mi → −Mi . commuting SO(3) algebra. there is a general method called the method of highest weight. This can be summarized. (±) ( ) (±) (±) (±) ijk Jk . However for P A(eφ)P = A(−eφ) so the generators of boosts are minus the generators of boosts in D.

Jj 1 ] ⊗ Id2s2 +1 (s ) i ijk Jk 1 ⊗ Id2s2 +1 (+) i ijk Jk Now you can solve for the L’s and M ’s in term of the J (±) . I’ll define Λ(eθ. Let me show that this satisfies one of the commutation relations.5) The Ji are our friends from the last lecture.20. Ji = Ji (s1 ) ⊗ Id2s2 +1 (20. 1. . M= 1 (+) J − J (−) i is antihermitian (because of the 1 ) 2 The commutation relations among the J (+) and J (−) (which were derived by using the definitions in term of L and M ). 1. L = J (+) + J (−) = J (s1 ) ⊗ Id2s2 +1 + Id2s1 +1 ⊗ J (s2 ) Because the J (s) are hermitian27 . Exponentiating L and M for any choice of s1 and s2 gives a representation called D(s1 . . 2 . L is hermitian. Jj (s1 ) ⊗ Id2s2 +1 ] ⊗ Id2s2 +1 i↔j −i↔j (s1 ) ⊗ Id2s2 +1 Jj (s1 ) (s ) (s ) Ji 1 Jj 1 ⊗ Id2s2 +1 − (s ) (s ) [Ji 1 .4) (2s2 +1)×(2s2 +1) dimensional identity and Ji (−) = Id2s1 +1 ⊗ Ji (2s1 +1)×(2s1 +1) dimensional identity (s2 ) (20. Jj ] = [Ji = Ji = = = = (+) (+) (s1 ) (s) ⊗ Id2s2 +1 . (You can check the others). take 230 Here are some: For any s1 .s2 ) (Λ(e θ. f φ) D(s1 . . 2 . s2 . December 4 Notes from Sidney Coleman’s Physics 253a 1 s1 = 0. f φ)) = e−iL·eθ e−iM ·f φ = e−i(J (only for e 27 (+) +J (−) )·eθ e−(J (+) −J (−) )·f φ f) = e−iJ (+) (eθ−if φ) e−i(J (−) (eθ+if φ) If you can’t prove this from (A ⊗ B)† = A† ⊗ B † please feel free to come ask me to elaborate. . (+) 1 s2 = 0. give us back the correct commutation relations for L and M .s2 ) . [Ji . · · · . .

D(s1 . . it has three generators.s2 ) is the basis |m+ . . I would like to explain the distinction between D(s1 . the rotations. and they are given by taking i d (D(s1 ) ⊗ D(s2 ) ) R(eθ dθ = (J (s1 ) ⊗ Id2s2 +1 + Id2s1 +1 ⊗ J (s2 ) ) · e θ=0 These three generators (one for each direction e can point). The claim is that these are a complete set of inequivalent irreducible reps up to a phase of the Lorentz group. bear some resemblance to the six generators of SO(3. s2 − 1. −s2 + 1. (s1 ) D(s1 ) ⊗ D(s2 ) (R(eθ)) = e−iJ (s1 ) ·eθ ⊗ e−iJ (s2 ) ·eθ This is a representation of SO(3). and D ⊗D(s2 ) . 1) such that the coefficient of Ji is the (−) same as the coefficient of Ji . m− = m± |m+ . December 4 Notes from Sidney Coleman’s Physics 253a 231 This has a simpler form for a pure boost or a pure rotation. . s1 − 1. D(s1 ) ⊗ D(s2 ) is a reducible rep of SO(3).20.s2 ) is an irreducible rep of SO(3. the coefficient of J (+) is independent of that of J (−) . But. you really have got the same thing. 1) if s1 + s2 is a half integer. m− (+) and It would have been possible to carry out the analysis in a more Lorentz invariant fashion by defining (according to Ramond. m− . D(s1 . 1). the direct product of two representations of SO(3). a representation of SO(3. which is in general reducible.s2 ) . s2 which simultaneously diagonalizing the two commuting hermitian operators Jz (−) Jz : (±) Jz |m+ . s1 m− = −s2 . −s1 + 1. 1).s2 ) (A(eφ)) = e−(J D(s1 . p. m+ = −s1 .10) J ij = ijk Lk . 1): J (+) = J (s1 ) ⊗ Id2s2 +1 and J (−) = Id2s1 +1 ⊗ J (s2 ) (+) and if you restrict yourself to elements of SO(3. . . . that is. . A standard basis for the representation D(s1 . .s2 ) (R(eθ)) = e−(J (+) −J (−) )·eφ (+) +J (−) )·eθ It turns out that these are only representations up to a phase of SO(3.

If people come ask me about them it will force me to do so. however for A ∈ SO(3. 1). This agrees with some general theorem of group theory. D(s1 . a rotation D(s1 . I would enjoy making some of these statement more concrete. Carefully spelling out and proving any of the statements before would be instructive.s2 ) (A) is not.s2 ) summarized (1). a boost. It is kind of nice to look through then calculations and see where i’s are changed to −i’s and where g µν becomes δ µν . worth noting. In fact it is hermitian. at the expense of clarity.s2 ) is (2s1 + 1)(2s2 + 1). Facts about D(s1 . (2). The dimension of D(s1 . For R ∈ SO(3. I feel I have put enough of the outlines here that I could pursue the proofs to my satisfaction. 1).20. the generators can always be chosen to be hermitian in some basis. J αβ ] = ig να J µβ − ig µα J νβ − ig νβ J µα + ig µβ J να A Lorentz transformation would be parametrized by an antisymmetric matrix D(Λ( µν )) µν = e−i µν J µν This approach buys you nothing but elegance. Another thing worth noting is that a parallel analysis can be made of SO(4).s2 ) (R) is unitary. December 4 Notes from Sidney Coleman’s Physics 253a J 0i = −J i0 = −Mi J 00 = J ii = 0 no sum 232 The commutation relation would have been unified into [J µν . . we have representation up to a phase D R(eπ) D R(eπ) = −1 (3). When s1 + s2 is a half integer. The finite dimensional representations of a compact group are always equivalent to unitary representations. This can be seen from the hermiticity of J (+) and J (−) .

D(s1 . 1) is not a compact group..s2 ) (unless s1 = s2 ).s2 )∗ and D(s1 . their coefficients. (19. (4).1) )). this shows that D(s1 .s2 )∗ R(eθ) = e+i(J ∼ e−i(J (+) +J (−) )·eθ (+) +J (−) )∗ ·eθ = D(s1 .s2 ) .s1 ) (Λ) If you believe that D(s1 . December 4 Notes from Sidney Coleman’s Physics 253a 233 SO(3. Although the J ± we have found are hermitian.s2 ) is not equivalent to D(s2 . The unitary representations of a non-compact group are always infinite dimensional. These would presumably describe an ∞ number of particle types. That is D(s1 .20. To agree with reality.The representation of a rotation is equivalent28 D(s1 . What rep do we get by taking the complex conjugate of D(s1 . We can identify the new rep we have made by noting that the exchange of J (+) ↔ J (−) is just like the exchange of s1 and s2 . J (+) ↔ J (−) does nothing to them. This is true for both the rotations. when Λ is not a rotation.s2 )∗ (Λ) ∼ D(s2 . and the boosts.s2 ) R(eθ) The representation of a boost however is screwed up in a way that cannot be undone by some equivalence.s2 )∗ (A(eφ)) = e−(J ∼ e−(J (+) −J (−) )∗ · eφ (−) −J (+) )·eφ The roles of J (−) and J (+) have ben exchanged. 2.s1 ) (unless s2 = s1 ).s2 ) .. 28 S −1 The sketch of fact (4) can be made more concrete by giving a name S to the matrix that satisfies J (s) S = −J (s) and using it to display the similarity transformation between D(s1 . (Dec. infinitely many particles would somehow have to be hidden. One can consider ∞ dimensional unitary reps. Eq. The range of boosts is infinite: 0 ≤ φ < ∞. Since J (±) ∼ −J (±)∗ (follows from their expressions in terms of J (s1 ) and J (s2 ) and the properties of J (s) . . The finite dimensional ones are never unitary. in the exponential are not purely imaginary.s2 )∗ is not equivalent to D(s1 .

Mi → −Mi .s2 ) “induces” a rep of SO(3) which is s1 +s2 D (s1 ) ⊗D (s2 ) ∼⊕ s=|s1 −s2 | D(s) . you have a representation of any of its subgroups (up to a phase). then in the tensor product – D(2s. we get a representation of SO(3) which is in general reducible.2s) is symmetric.s2 ) ⊗D (s1 .2s−2) is antisymmetric. and for those coefficients. I expect with some thought they can be derived from the analogous statements about SO(3) in short order. In fact we are down to the case where the coefficients of J (+) and J (−) are identical. D(2s−1. If you understood the distinction between D(s1 . we have exactly the same possible representation matrices as if we had taken D(s1 ) ⊗ D(s2 ) . D (s1 . (7).s2 ) ∼ D(s2 . D(2s. – D(2s. the representation of the subgroup may be reducible. D(2s. – D(2s−1.2s−1) is symmetric. a subgroup of SO(3. The proof of these statements would probably be instructive to construct.s2 ) =⊕ s1 =|s1 −s1 | s2 =|s2 −s2 | D(s1 . – D(2s−2.2s−2) is symmetric. etc. If s1 = s1 = s1 and s2 = s2 = s2 .20.2s−1) is antisymmetric. etc. Even if the representation of the group is irreducible. 1).2s−2) is symmetric. So D(s1 . By restricting ourselves to the rotations. is also interpretable as J (+) ↔ J (−) so (s Dp 1 .2s−3) is antisymmetric.s2 ) .s1 ) s1 +s1 s2 +s2 (6). this fact about D(s1 . The effect of parity on a rep was already discussed Li → Li . December 4 Notes from Sidney Coleman’s Physics 253a 234 (5).s2 ) will be easy to understand. Any time you have a representation of a group (up to a phase).s2 ) and D(s1 ) ⊗ D(s2 ) .

D(3/2.3/2) which is inequivalent to D(3/2. 0) and (0.1/2) ⊗ D(1/2.0) . If you restrict yourself to the rotation subgroup of SO(3.3/2) .1/2) ∼ D(1.0) however remains irreducible under this restriction. 1.20. What about rank 2 tensors Usually we get a rank 2 tensor by taking the tensor product of two vectors. D(3/2. 3.0) ⊗ D(0. because in the usual basis.1/2) must be a tensor. It looks funny. Aµ B ν transforms like a tensor. 1) and SO(3).1) ⊕ D(0. 1). Now ν D(1/2.0) is D(0. Later we’ll make a vector out of D(1/2. which is equivalent to D(1/2. It certainly can’t be ruled out along the lines of above three arguments. the matrices that transform a vector are purely real.1/2) (2 · 3/2 + 1)(0 + 1) (2 · 0 + 1)(2 · 3/2 + 1) (2 · 1/2 + 1)(2 · 1/2 + 1) 4 4 4 (these are only three ways to factor 4).1/2) . Identical arguments rule out D(0. If D(1/2. Examples Where is the vector? If the vector representation which we all know and love. but it must be right. A similar argument applies by considering the effect of parity. If Aµ and B transform like vectors. D(1/2. it must be equivalent to one of the four dimensional representations we have constructed.1/2) is a vector then D(1/2. However the vector is equivalent to its complex conjugate.1/2) ⊗ D(1/2. 2. It is a spinor with spin 3/2 under rotation. December 4 Notes from Sidney Coleman’s Physics 253a 235 (s. The only irreducible four dimensional reps on our list are D(3/2.0) S A A S dim 9 dim 3 dim 3 dim 1 .1/2) .0) ⊕ D(0. the four vector representation is reducible into a three vector and a rotational scalar. The complex conjugate of D(3/2.3/2) and D(1/2.1) ⊕ D(1.0) D(0. s) are irreducible reps of SO(3.0) cannot be the vector for any of several reasons.1/2) must be the vector. Manifestly. is irreducible (I’ve never been able to reduce it).

December 4 Notes from Sidney Coleman’s Physics 253a 236 There must be a way to reduce the tensor T µν into 9. 1).n/2) D is the symmetric traceless tensor of part n. 3. We can rewrite 1 1 T µν = (T µν + T νµ ) + (T µν − T νµ ) 2 2 ≡ S µν + Aνµ 10 dim 6 dim and from our general arguments about symmetric and antisymmetric tensor products. these two subspaces must transform independently. In general. so that 1 dimensional subspace transforms independently of the other 9 components of S which are in 1 S µν − gµν S µν 4 the “traceless” part of S. 1 1 Aµν = (Aµν + iADµν ) + (Aµν − iADµν ) 2 2 (+)µν (−)µν ≡A +A . Indeed. the symmetric traceless tensor.1) . This is D(1. Given any antisymmetric two index tensor. the Lorentz transformations transform the subspace of each eigenvalue of the dualing operation independently. and the eigenvalues must be ±i. The 10 dim symmetric subspace must contain a 9 dim and 1 dim subspace which transform independently. you can define a new antisymmetric tensor by AD µν = AD 01 = A23 the AD01 1 2 µναβ A αβ 0123 = +1 1 was inserted to avoid AD 01 = 2A23 2 = −A23 .20. Thus. (n/2. let’s find the two irreducible parts of Aµν . AD23 = A01 The cute thing about this operation is that ADD µν = −Aµν The square of the dualing operation is −1. and 1 dimensional subspaces that transform independently under SO(3. µν Just for completeness. 3. the linear combination gµν S µν is a Lorentz invariant. The dualing operation also commutes with Lorentz transformations (that have det (Λ) = +1).

5). this is really the only application of the last couple lectures that is going to be used. We’ll study D(0. We’ll be using. Spinors This is the example we are really interested in. 2 and using σ J (0) = 0. u+ . a rotation is given by D(R(e θ)) = e−i(J (+) +J (−) )·e θ = e−iσ·e θ/2 A boost is represented by D(A(eφ)) = e−(J (+) −J (−) )·eφ = e+σ·eφ/2 Because of the plus sign in e+σ·eφ/2 . (20. fields transforming like D(1/2.1/2) .20. December 4 Notes from Sidney Coleman’s Physics 253a 1 Dµν A ± iADDµν 2 1 = ADµν iAµν 2 = iA(±)µν 237 A(±)Dµν = The six dimensional antisymmetric part of the tensor product has been broken into two 3 dimensional parts. J (−) = 2 In this representation. with s1 = 0.0) and D(0. . let’s call a field that transforms under D(0.0) and D(0.) So if you get a good handle on how to manipulate these two representations.1/2) .1) .1) are not equivalent to their complex conjugates.0) and D(0. (The scalar and vector reps you (presumably) already understood.4) and (20. you don’t really have to understand all the representation theory that has gone before. J (1/2) = 2 we have σ J (+) = 0.1/2) . It is no surprise that this splitting required taking complex combinations because D(1. In fact. s2 = 1 . which must be D(1. for the rest of the course. Using the formulas in Eqs.

but a boost has a minus sign in the exponential. we’ll write u† . v 0 = u† u+ . Both u+ and u† transform as + rotational spinors. and a rotational scalar can be made by taking u† u+ + and a rotational vector is u† σu+ + The four vector must be (u† u+ . α is unknown from this argument. and if we can find the right combination. An easy way to find the right combinations is to recall the transformations of spinors under the subgroup of rotations. a rotation is represented the same way 29 .1/2)∗ ∼ D(1/2. 29 .0) rep. Since + D(1/2. Up to equivalences there is only one way for an irreducible rep of a given dimension to transform under rotations. i. but we can + + find it by looking at a boost along ez with rapidity φ. A(ez φ) : u+ → e+σz φ/2 u+ and u† → u† (eσz φ/2 )† = u† eσz φ/2 + + + so u† u+ → u† eσz φ u+ + + = u† (cosh φ + σz sinh φ)u+ + = cosh φ u† u+ + sinh φ u† σz u+ + + If we take α = 1. this says + + v 0 −→ cosh φv 0 + sinh φvz Let’s see what vz (v 3 if you want to keep your indices up) and vx (or vy ) transform into vz = u† σz u+ −→ u† eσz φ/2 σz eσz φ/2 u+ + + = u† (σz cosh φ + sinh φ)u+ + = cosh φ vz + sinh φ v 0 vx = u† σx u+ −→ u† eσz φ/2 σx eσz φ/2 u+ + + It better be.1/2) The product of u and u† must be a four vector.0) .20.0) will be called u− .0) . we can make that explicit.0) ⊗ D(0. Because D(0.e. so the field transforming as D(1/2. v = u† σu+ . αu† σu+ ).1/2) ∼ D(1/2. December 4 Notes from Sidney Coleman’s Physics 253a 238 In the D(1/2. u∗ must transform like D(1/2. If we want to think of u∗ + + as a new vector.

20. December 4 Notes from Sidney Coleman’s Physics 253a φ φ + sinh σz and σz σx = −σx σz . . −u† σu− ) − − It looks like the two component fields we have found have a fighting chance of describing spin 1/2 particles. If you had gone through this procedure for u− . you would have found that the vector is wµ = (u† u− . we see that we have found 2 2 vx −→ u† e+σz φ/2 e−σz φ/2 σx u+ = vx + 239 But using eσz φ/2 = cosh Similarly vy −→ vy .

nor can a scalar be + built from it with an even number of derivatives. a scalar can be built from it and one derivative. Criteria for a free theory made up of a u+ field. ∂µ u+ . S = d4 x L had better be real. Property (v) with property (iii) force us to have one u+ and one u† factor in each term. u† . So (ii) and (iv) imply L ∝ (u† ∂0 u+ + u† σ · + + u+ ) The coefficient of proportionality must be purely imaginary to satisfy (i) (proof involves a parts integration). (ii). L(u+ . However.21. (iii). December 9 Notes from Sidney Coleman’s Physics 253a 240 21 December 9 Promote the two component objects u± into two component functions of space time. + Now the product u+ u† is a four vector. S = S ∗ . which is neither a scalar itself. ∂µ u† ) + + (i). L bilinear in the fields so we get a linear equation of motion and a free field theory (iv). Want the theory to have a conserved charge u+ −→ eiθ u+ u† −→ e−iθ u† + + (Rule out Majorana neutrinos) because all known spin 1/2 particles in the world do carry some conserved quantum number. If we can’t construct anything of this type we’ll go to three. like baryon number. but not necessarily parity invariant. No more than two derivatives in L. L must be a Lorentz scalar. four or more derivative (v). spinor fields. By rescaling u+ and u† we have + L = ±i u† ∂0 u+ + u† σ · + + u+ u− ) (For u− we would have arrived at L = ±i u† ∂0 u− − u† σ · − − .

(no need yet for k 0 > 0) where u+ is some constant two component column vector.t. for a given k µ . which leads to u− (x) = u− e−ik·x ) Now we plug these potential solutions back into the equation to get the condition on the constant spinors ( k 0 −σ · k)u+ = 0 2×2 identity is understood Let’s take k = k 0 ez . This would not be possible if these particles weren’t massless.) 1 1 0 has the interpretation of being the spin along the direction of motion (at least for k 0 > 0). (An identity matrix has been suppressed). So all the solutions of the Weyl equation satisfy the wave equation. where a 2 × 2 identity is again suppressed in δij . (For u− we would have gotten ∂0 u− − σ · u− = 0.21. Unlike the normal theory of spinors. we only have one solution. If they were massive. and they must be of the form u+ (x) = u+ e−ik·x k 2 = 0. one direction of spin. December 9 Notes from Sidney Coleman’s Physics 253a 241 This is called the Weyl Lagrangian. turn their spin around and then boost . Let’s derive the equation of motion by varying w. To derive this you need to use equality of mixed particles and σi σj = i ijk σk + δij . This then says (1 − σz )u+ = 0 (At this point we would have gotten u− = In general u+ is the eigenstate of σ·k k0 with eigenvalue 1 which since |k| = k 0 u+ = 0 . you could always boost to their rest frame.r. u† + ∂0 u+ + σ · u+ = 0 We can see that any solution of this equation is a solution of the Klein-Gordon equation by acting on it with ∂0 − σ · to get 2 (∂0 − 2 )u+ = 0.

1 0 kz = k 0 kx = ky = 0.1/2) R(ez θ) u+ (0) and this equation can be used by taking the 0| − |k matrix elements. More physically. (0.1/2) there is no parity transformation in this theory. the component of angular momentum along the direction of motion. By the known transformation properties of the solution of the field equation. Since parity reverses linear momentum. December 9 Notes from Sidney Coleman’s Physics 253a 242 back. Because DP = D(1/2. It is the angular momentum in the rest frame. . Spin is usually reserved for massive particles.21. and you’ll have a particle with spin pointed the opposite direction.1/2) Some guesses about the quantum field u+ and the particles it will annihilate and create The solution of the Weyl equation going like e−ik·x with k 0 > 0 will probably multiply an annihilation operator in the expansion of the quantum field u+ . or 0|u+ (0)|k ∝ We are going to find the Jz value of |k : Jz |k = λ|k . we can obtain the transformation properties of the states it annihilates. We get e−iλθ 1 0 ∝ e−iσz θ/2 1 0 = e−iθ/2 1 0 find λ 1 0 k 0 > 0.0) which is inequivalent to D(0. we have found a theory with a single u+ spinor has only one helicity. The spin of a massless particle is usual referred to as helicity. In the quantum theory U (R(ez θ)) = e−iJz θ and thus U (R(ez θ))|k = e−iλθ |k U (R(ez θ))|0 = |0 On the other hand U † R(ez θ) u+ (0)U R(ez θ) = D(0. We expect 0|u+ (x)|k ∝ e−ik·x k 2 = 0. but leaves angular momentum unaffected. a theory with particles of only one helicity can’t be parity invariant.

n/2 . The Lagrangian has a U (1) symmetry under which u+ and u− transform the same way: u± −→ eiθ u ± u† −→ e−iθ u† ± ± . they turn out to be rank n symmetric traceless tensors under SO(3. and the creation operators create particles with helicity + 1 .88). ones whose angular momentum is along the direction of motion (agrees with I+Z. a set of two complex doublets. Instead of canonically quantizing this theory now. u† . p. k0 > 0. at least S = S ∗ (iii). u+ and u− . . The simplest parity invariant reducible rep with spinors is D(1/2. 2 (When Weyl came up with this theory. the world thought the world was parity invariant. these reps contain do not contain spinors. like neutrinos are observed to have. which describes neutrinos. The representation the fields are in must be equivalent to the representation obtained by parity. L(u+ . such a field is called “left handed”. People thought he was just playing with irrelevant mathematics).21. Conventionally. u† . No more than one derivative. By the right hand rule.0) ⊕ D(0. we are going to move on and find a Lagrangian for massive particles that can include parity. The field u† will annihilate particles with + 1 1 helicity λ = − 2 and create particles with λ = + 2 . We could look at the irreducible reps of this type. . in fact as already stated. 1 The annihilation operators in the u− field annihilate particles with helicity − 2 . December 9 Notes from Sidney Coleman’s Physics 253a 243 1 i. The annihilation operator multiplying u+ e−ik·x . and his theory was dismissed quickly. a “right handed” field annihilates “right handed” particles. . L is a Lorentz scalar. ∂µ u+ . but when you restrict to rotations. The field 2 always changes the helicity by the same amount. will annihilate parti2 cles with helicity 1 along the direction of motion. λ = . Bilinear (ii).) must be + − (i). 2 the antineutrinos.e. (v). that is.1/2) . Creation operators in the expansion of u+ (x) will create λ = − 1 particles. (iv). We’ll restrict the possible Lagrangian by set of conditions like those at the start of this lecture. Real. 1). u− . Dn/2.

you also get |b|2 = 1 and ab∗ = 1 The conditions on a and b can be summarized by saying a = b = eiλ P : u± (x. t) ∂ u− (−x. t) under which L is invariant. In this case. t) −→ b u+ (−x. t) −→ eiλ u (−x. t) −→ a u− (−x.0) (Λ) ∝ u† u− ) + After rescaling u+ and u− . compose parity with the symmetry u± (x) −→ e−iλ u± (x) . then I can redefine parity to be the old parity composed with any element of the internal symmetry group. we can always take m to be real and nonnegative.0) (Λ) ⊕ D(0. By considering the effect of parity on the other terms.0) ⊗ D(1/2. and I’ll have just as good a definition of parity. Let’s find out if we can define a parity of the type in condition (vi). t) a∗ (∂0 + σ · − ) a u− (−x.0) ∼ D(1. t) = |a|2 u† (−x. December 9 Notes from Sidney Coleman’s Physics 253a 244 (vi). u† (∂0 + σ · + )u+ −→ u† (−x. the only Lagrangian satisfying the first five conditions is L = ± iu† (∂0 + σ · + where = ±1. The theory has a parity operation u+ (x.21. t) u− (x. (D(1/2. t) Now if a theory has an internal symmetry (and this one has an internal U (1) symmetry). t) ∂0 − σ · − ∂(−x) )u+ + i u† (∂0 − σ · − )u− − mu† u− − m∗ u† u+ + − If this term in the parity transformed Lagrangian is going to equal the term in the original Lagrangian i u† (∂0 − σ · )u− − we must have = +1 and |a|2 = 1. By adjusting the relative phase of u− and u+ .

t) The Lagrangian we have found is the Dirac Lagrangian. although it doesn’t look like it yet.21. you get + i(∂0 + σ · and by varying u± i(∂0 − σ · )u− = mu+ This is Dirac’s equation although it doesn’t look like it yet. ψ≡ u+ u− (This is not the only way to do this. The solutions of the Dirac equation are solutions of the Klein-Gordon equation. see below). Let’s derive the equations of motion. t) −→ u (−x. Assemble the two two-component fields into a single four component one. December 9 Notes from Sidney Coleman’s Physics 253a 245 Then the new parity has the effect P : u± (x. Then L = ± iu† (∂0 + σ · + can be rewritten as L = ± i ψ † ∂0 ψ + ψ † α · where α’s and β are 4 × 4 hermitian matrices α= σ 0 0 −σ β= 0 1 1 0 ψ − mψ † βψ )u+ + iu† (∂0 − σ · − )u− − mu† u− − mu† u+ + − . and not m/2 or m or whatever. By varying u† . To see this. multiply the first equation by −i(∂0 − σ · ) and use the second equation to get 2 (∂0 − 2 )u+ = mu− )u+ = −m2 u+ or (2 + m2 )u+ = 0 This verifies that the thing we have been calling m actually is a mass. 3/5 We are now going to modify the equations in order to make them more obscure and sophisticated looking. Actually we’ll be building a machinery which will speed up calculations.

t) −→ βψ(−x.21. The equations Λ : u+ −→ eσ·eφ/2 u+ u− −→ e−σ·eφ/2 u− can be assembled into Λ : ψ −→ eα·eφ/2 ψ The generator of boosts (what is dotted into −ieφ in the exponential) is M= iα 2 )ψ = βmψ We can get the generators of rotations quickly by using [Mi . t) The equation of motion obtained by varying ψ † is i(∂0 + α · This is the Dirac equation (1929). December 9 Notes from Sidney Coleman’s Physics 253a 246 each entry represents a 2 × 2 block. In this spiffy notation. we can also write the effect of parity as P : ψ(x. Another way is ψ= u+ + u− u+ − u− 1 √ 2 If we had done that. the α’s and β would have come out differently! In fact this second way is the way Dirac originally did it. . Mj ] = −i L= I said ψ = 1 2 ijk Lk σ 0 0 σ u+ is not the only way of making a four component spinor out of two u− two-component spinor. A pleasant surprise is that in this notation we can also give the effect of a Lorentz boost without defining a whole bunch more matrices.

31 30 . but the terminology is incorrect There are extensive discussions of the solutions of the Dirac equation in the literature. doesn’t mean you have to learn it that way today. Of course. in the first basis. December 9 Notes from Sidney Coleman’s Physics 253a 247 To summarize. or p0 ≡ p2 + m2 p is a forward pointing vector on the mass shell. All you have to do is ignore all references to holes and negative solutions. usually called the Weyl representation. Mj ] = −i 0 1 1 0 ijk Lk ψ 1 2 u+ u− σ 0 0 σ =⇒ L = DIRAC (OR STANDARD) BASIS α= 0 σ σ 0 β= iα 2 1 0 0 −1 L= 1 2 1 ψ=√ 2 σ 0 0 σ u+ + u− u+ − u− M= In either basis. called the WEYL BASIS 30 σ 0 α= β= 0 −σ M= iα 2 [Mi .21. PLANE WAVE SOLUTIONS OF THE DIRAC EQUATION We are going to look for solutions31 of the form ψ = up e−ip·x or ψ = vp eip·x where up and vp are space-time independent four component spinor. Just because something was understood in a poor way 50 years ago. to have a chance of satisfying the Dirac equation a proposed solution must satisfy the Klein-Gordon equation. the Dirac Lagrangian is L = ± iψ † ( ∂0 +α · 4×4 identity matrix suppressed )ψ − mψ † βψ The Dirac basis is the standard basis because the solutions of the Dirac equation become especially simple in the nonrelativistic limit in this basis. so p2 = m2 .

p0 = − p2 + m2 . a convention that goes all the way back to Hψ = i ψn = e−iωn t ) into the Dirac equation. they are of the form   a b u0 =   0 0 In the standard basis. ψ = up e−ip·x (called “positive frequency”. p0 = m. Then the equation says u0 = βu0 1 0 (each entry represent a 2 × 2 block). β = )ψ = βmψ ∂ψ dt −→ . and there are 0 −1 two solutions linearly independent to this equation. is taken care of in the way we have busted up the problem into two cases: one with e−ip·x dependence and one with e+ip·x dependence. i(∂0 + α · you get (p0 − α · p)up = βmup .) If you plug the first type of solution. December 9 Notes from Sidney Coleman’s Physics 253a 248 (The other solution. Let’s look at a special case: p = 0.21.

as well as their total normalization. an annihilation operator that annihilates Jz = + 1 2 electrons with momentum p will multiply up e−ip·x 1 and an annihilation operator that annihilates electrons with Jz = − 2 will multiply (1) (r)† (s) (r)† + u0 = 2mδrs (s) up e−ip·x The 2m in the normalization of the u’s is there to agree with conventional normalization for relativistic states: p |p = (2π)3 2 ωp δ (3) (p − p ) reduces to 2m when p=0 (2) In many equations it will allow us to take a smooth m → 0 limit. In the standard basis it connects the upper two components with the lower two.22. make it satisfy u0 and u0 αu0 = 0 The second condition follows fairly easily from the form of α. By arguments like those used for the solutions of the Weyl equation. . any way I like. A convenient choice is   1 √ 0 (1) u0 = 2m   0 0   0 √ 1 (2) u0 = 2m   0 0 Whatever basis you choose. we expect that in the expansion of the quantum field ψ. December 11 Notes from Sidney Coleman’s Physics 253a 249 22 December 11 I can choose the direction my two standard linearly independent solutions point in the subspace they are allowed in. (Good for neutrinos or extremely high energy physics).

with p (hence e) pointing in the z direction. this is √    E+m √ 0    E+m  0 (1) (2)   up = √ up =   E − m   0 √ 0 − E−m √ (The normalization is already proving useful. and you know it is (2mδrs . (This proof is sweet because it is basis independent). m you can rewrite up = (r) E+m + 2m E−m (r) α · e u0 2m In the standard basis.22. I ought to get another solution by Lorentz transforming it. December 11 Notes from Sidney Coleman’s Physics 253a 250 The Dirac Lagrangian was constructed to be a Lorentz scalar. this doesn’t blow up when m → 0.) . Thanks to that factor of 2m. 0) when p = 0: that determines it completely for arbitrary p. (up up . up αup ) is a four vector. We will flaunt Lorentz invariance by getting solutions with momentum p by boosting those with momentum 0. you can rewrite up = (r) (r)† (s) (r)† (s) φ φ (r) cosh +α · e sinh u0 2 2 4×4 identity suppressed Using cosh φ = 2 1+cosh φ 2 and sinh φ = 2 cosh φ−1 2 and cosh φ = E . Using (α · e)2 = 1. so you expect that the equations of motion derived from it are Lorentz invariant in the sense that given one solution of the Dirac equation. up = eα·eφ/2 u0 where e = p |p| (r) (r) and sinh φ = |p| m (cosh φ = E ) m You can mechanically verify that the conditions up up = 2p0 δrs (r)† (s) up αup = 2pδrs (r)† (s) are satisfied. but it is actually not necessary. So while we could just go ahead and solve the Dirac equation for arbitrary p (it’s just the problem of finding the eigenvalues and eigenspinors of some p dependent 4 × 4 matrix).

“Anything” means any equation that is unaffected by a similarity transformation. 2pδrs ) In the standard basis. The theorem says “everything is in here. An example of a statement that is not basis independent is β† = β . αi = 0 Every α and β squares to 1 and anticommutes with all three others.22. αj =0 i=j β2 = 1 β. December 11 Notes from Sidney Coleman’s Physics 253a (r) 251 A similar set of relations holds for the vp   0 √ 0 (1) v0 = 2m   1 0 (r) v0 (2)   0 √ 0 = 2m   0 1 vp = eα·eφ/2 v0 (r)† (s) (r)† (s) (r) (vp vp .” Anything we get by manipulating a set of 4 × 4 matrices satisfying this algebra can be obtained by manipulating the algebra.    √ E−m √0  − E − m   0 (1) (2)   vp =  vp = √   E + m  0 √ 0 E+m The α’s and β satisfy a simple algebra. 2 αi = 1 αi . or any result that is basis independent. like a cross section summed over final spin states and averaged over initial ones. A famous theorem due to Pauli Any set of 4 4 × 4 matrices obeying these equation is equivalent to any other set. with p in the z direction. vp αvp ) = (2p0 δrs .

Mj = i ijk Mk and Li .0) ⊕ D(0. it is easy to show that Mi .1) Recall the stuff about parity: Given a representation of the Lorentz group.1/2) . Furthermore.1/2) ⊕ D(0. and the rotation generators square to 1 . The generators of boosts in DP were minus the generators of boosts in D. They are called “unitarily equivalent”. D(0.22. In general the representation obtained from D(n/2. Lj = i ijk Lk Thus we have defined a representation of the Lorentz group. I can define a new rep DP by DP (Λ) = D(ΛP ) The generators of the rotations in this new rep were the same. Coleman’s proof uses something we already believe (but in fact takes some effort to prove): That up to equivalence we have found all the finite dimensional reps of the Lorentz group. it is a four 1 dimensional representation. Mj = −i ijk Lk (Lk = Li . Thus it must i 4 be made of just spin 1 reps when you restrict this four dimensional rep of the Lorentz group 2 to the rotation subgroup.0) But now the existence of β can be used to rule out the second two possibilities. Start by constructing a representation of the Lorentz group from the α’s and β. D. L2 = 4 . Let Mi ≡ i αi 2 Define Lk by i ijk Mi . Mj ) 2 Using the algebra the α’s are supposed to obey. it does not follow that S −1 βS is hermitian It is true though if S is unitary. What we have made must be equivalent to D(1/2.m/2) was D(m/2. Note that β 2 = 1 ⇐⇒ β = β −1 so using the algebra β −1 αβ = βαβ = −α For our generators this implies β −1 M β = −M and β −1 Lβ = L (22. December 11 Notes from Sidney Coleman’s Physics 253a 252 This is true in the Weyl or standard basis but in general.0) ⊕ D(1/2.n/2) . just because β is hermitian.1/2) or D(1/2. Sometimes we’ll restrict ourselves to bases that are related a unitary S.

0) ⊕ D(1/2. B.) (This is fairly easy to show. These (reducible) reps are not equivalent to their parity transformed reps. C and D. Write β = The other condition on the β from the algebra is β 2 = 1. and the only matrix that commutes with all three of them is the identity. The parity transform of D(1/2. and there is only one such rep up to equivalence transformation.1) is such an equivalence. A B and find out what the conditions are C D on each of the 2 × 2 matrices A. but brings β into standard form? If I could. since they are all equivalent to a standard form. We have shown the α’s are always equivalent. December 11 Notes from Sidney Coleman’s Physics 253a 253 There are equivalent only if m = n.1/2) ⊕ D(0. by Eq. β = 0 =⇒ β = 0 λ2 λ1 0 where the λ1 . but it remains to be shown that β is equivalent by the same transformation. So suppose I have found a similarity transformation that puts the α’s into standard form α= σ 0 0 −σ Can I find a further similarity transformation that leaves the α’s unchanged. We must have constructed the rep D(1/2.1/2) and vice versa. 1) plus parity that is four-dimensional and only contains spin 1 particles. λ2 are blocks proportional to a 2 × 2 identity. No matrix anticommutes with all three Pauli matrices. A jargony way of saying what we have found is: There is only one rep of SO(3.0) ⊕ D(0.22. With α = . These reps cannot be candidates for what we have constructed. But β.0) is D(0. which implies λ1 λ2 = 1 or λ = λ1 = λ−1 2 . (22. this would show that any set of α’s and β is equivalent to σ 0 any other.1/2) (it is equivalent to its parity transform). 2 A little bit of the proof remains to be done. the algebra 0 −σ α.

32 due to Pauli . All popular representations satisfy these relations. DIRAC32 ADJOINT. we are going to assume we are in a basis where α = α† and β = β† A basis which is obtained from this basis by a unitary transformation will also satisfy these relations. This must have been unnecessary since the whole thing is tied up in the commutation relations.22. December 11 Notes from Sidney Coleman’s Physics 253a 254 σ 0 0 λ−1 and we need to find a similarity transformation and β = λ 0 0 −σ that leaves the α’s unaffected but puts β into the standard form So α = β= 0 1 1 0 The similarity transformation is S = S −1 βS = = λ 0 0 1 λ 0 0 1 λ−1 0 0 1 0 λ−1 λ 0 0 λ−1 1 0 λ−1 0 0 1 = 0 1 1 0 A lot of people including some Nobel laureates did lengthy calculation using explicit representations of the Dirac algebra in the 1930’s. PAULI-FEYNMAN NOTATION Since ψ † βψ is a Lorentz invariant (−mψ † βψ appears in the Lagrangians). From now on. that is since under Λ : ψ −→ D(Λ)ψ ψ † βψ −→ ψ † βψ We are going to define a new adjoint ψ = ψ†β This new adjoint has every property you’d like an adjoint to have except that ψψ is not always greater than zero.

The usual inner product between two four vectors yT x = µ y µ xµ is not a Lorentz invariant. oh so slickly. and the proofs are the same since it all comes from the definition of the adjoint. 1). Similarly. oh so slickly Λ : ψψ −→ ψψ (or Λ : χψ −→ χψ for two Dirac spinors ) [ The situation is a lot like SO(3. The combination that is is   1 0 0 0 0 −1 0 0  y T gx g= 0 0 −1 0  0 0 0 −1 So we define a “new transpose” y T g and call it a covariant vector. αA + βB = α∗ A + β ∗ B . the Dirac adjoint of an operator is obtained by φ Aψ ≡ (ψAφ)∗ It is the work of a moment to show that A = βA† β All your favorite equations for the adjoint of an operator follow for the Dirac adjoint. Then we can write.22. Λ: µ yµ xµ −→ µ y µ xµ . and put its indices down. December 11 Notes from Sidney Coleman’s Physics 253a 255 Then we can write. As a statement about the 4 × 4 matrices Λ this is ΛT gΛ = g or gΛT gΛ = 1] The definition of the adjoint of an operator is obtained from the definition of the adjoint of a vector by φ† A† ψ ≡ (ψ † Aφ)∗ That tells you an arbitrary matrix element of A† .

Here comes some more notation to make this look slick too. . µ. no more. We can rewrite V µ as V µ = (χβψ. More generally.3) This is the transformation law for a two index tensor. but they are “Dirac unitary”.22. The proof is: D(Λ)γ µ γ ν D(Λ) = D(Λ)γ µ D(Λ)D(Λ) γ ν D(Λ) = = fancy way of inserting 1 σ ν Λ σγ Λ τ γτ Λµ σ Λν τ γ σ γ τ µ (22. December 11 Notes from Sidney Coleman’s Physics 253a 256 AB = B A Aψ = ψ A This last relation implies that Λ : ψψ −→ ψ D(Λ)D(Λ)ψ but this equals ψψ so D(Λ)D(Λ) = 1 is the statement about the 4 × 4 matrices D(Λ). What is meant by this. . we can say that the γ matrices transform like a vector. is that D(Λ)γ µ D(Λ) = Λµ ν γ ν We can also say that the product γ µ γ ν transforms like a tensor. They can all be summed up in γ µ . χβαψ) = χγ µ ψ where γ µ = (β. 33 .33 If in some math book you start reading about Clifford algebras. ν = 1. no less. it is a special case of them we are studying. Of course. N and any # of diagonal components of g µν can be −1. βα) These are the famous Dirac γ matrices. χ† αψ) transforms like a four-vector. they don’t transform at all. . . With a slight abuse of language. The anti-commutation relations for the γ matrices follow from those for the α’s and β. γ ν = 2g µν which you should check.2) (22. They aren’t unitary. Remember that V µ = (χ† ψ.

(22.2) and (22. (The RHS we have already accepted this for. It is real.22. b &¡ = 2a · b (&2 = a2 ) a The Dirac Lagrangian and equation of motion are L = ±ψ(i − m)ψ ∂ (i − m)ψ = 0 ∂ . To see if it is correct we only have to check one of its components. say µ = 0.3). both sides of this equation transform like a four-vector. Let a & ≡ aµ γ µ (= aµ γµ = a · γ) The algebra of the γ matrices can be summarized in this notation as a. and it is not transposed because we are transposing only in the spinor indices. but here is a high-powered proof instead: In the same sense as in Eqs. we can rewrite the Dirac Lagrangian as L = ± iψγ µ ∂µ ψ − mψψ The equation of motion is iγ µ ∂µ ψ − mψ = 0 (from varying ψ). December 11 Notes from Sidney Coleman’s Physics 253a 257 Also. . We can make this look even more sophisticated and obscure by introducing a super compact notation due to Feynman. D(Λ)γ µ D(Λ) = Λµ ν γ ν .) So γ µ = γ µ is a Lorentz covariant equation. The LHS transforms the same way as you can see by taking the bar of this equation to get D(Λ) γ µ D(Λ) = Λµ ν γ ν ) (Λµ ν : This matrix is unaffected. for any given µ. Now that we have these Dirac γ matrices. Maybe I should first say. this is just a set of 4 real coefficients. in the special class of bases we have restricted ourselves to γ 0† = γ 0 γ i† = −γ i An elegant way of summarizing these four relations is γµ = γµ Which you could check. For µ = 0 it reduces to γ 0 γ 0† γ 0 = γ 0 .

We can’t proceed on building tensors of higher and higher rank. Define 1 µ ν σ µν = σ µν σ µν = γ . Actually we have obtained nothing new from the 00 and ii components because γ µ 2 = 1. December 11 Notes from Sidney Coleman’s Physics 253a 258 The proof that each component of every solution of the Dirac equation satisfies the Klein-Gordon equation is (i − m)ψ = 0 =⇒ (−i − m)(i − m)ψ = 0 ∂ ∂ ∂ =⇒ (2 + m2 )ψ = 0 Parity and γ5 P : ψ(x. P : ψγ µ ψ −→ ψβγ µ βψ = ψγ 0 ψ −ψγ i ψ µ=0 µ=i This is how you expect a vector to transform under parity. while the 0i components will go into minus themselves.22.γ 2i ψσ µν ψ is an antisymmetric tensor under L. What about ψγ µ γ ν ψ. t) = ψ(−x. So far we have found 1 + 4 + scalar vector 6 antisymmetric tensor = 11 of them .T. It’s clear that the 00 component will be unaffected by parity. as will the ii components. t) −→ ψψ(−x. t) −→ β ψ(−x. and parity. The only new quantities we have are the antisymmetric parts. t)β = ψ(−x. and parity. t) could also be written γ 0 ψ(x. This is a tensor under L.T. t) −→ βψ(−x. t) Not only is ψψ a scalar under the Lorentz transformations connected to the identity. t)β So P : ψψ(x. it is a scalar under parity. In the product of two four component objects there are only 16 possible bilinears.

because iγ5 = iγ5 . There is only one possibility if all four indices have to be different.22. but goes into minus itself under parity.T. Now µναβ γ µ γ ν γ α γ β transforms like a scalar under L. Except for γ5 = −γ5 . but if any two of the indices are the same. it is γ 0γ 1γ 2γ 3 It is conventional to define γ5 ≡ iγ 0 γ 1 γ 2 γ 3 (≡ γ 5 ) i µ ν α β = ( µναβ γ γ γ γ 4! 0123 = +1) † † 2 Unfortunately. ψγ µ γ5 ψ is also hermitian and P : ψγ µ γ5 ψ −→ −ψγ 0 γ5 ψ ψγ i γ5 ψ µ=0 µ=i an axial vector So now we have found a total of 16 bilinears transforming in distinct ways under parity and L. but γ5 = γ5 and γ5 = γ5 γ5 = +1. γ5 = −γ5 . γ µ = 0 We’ll write iγ5 . S P V A T 1 1 4 4 6 scalar pseudoscalar vector axial vector antisymmetric tensor . December 11 Notes from Sidney Coleman’s Physics 253a 259 Let’s jump up to tensors of the fourth rank and see what we can make ψγ µ γ ν γ α γ β ψ is a fourth rank tensor. because of the i. that is γ5 . this reduces to something we have already found. that is is a pseudoscalar P : ψiγ5 ψ −→ −ψiγ5 ψ ψiγ5 ψ is hermitian.T. it can appear in a Lagrangian with a real coefficient. γ5 is a lot like a fifth γ matrix.

no choice of parity is possible). and we can find what it is by looking at v0 γ 0 u0 = v0 u0 = 0 so vp γ µ up = up γ µ vp = 0 (r) (s) (r) (s) (r) (s) (r)† (s) . December 11 Notes from Sidney Coleman’s Physics 253a 260 (Any other bilinear we might construct must be expressible in terms of these.) We could start building Lagrangian with interactions (but we are going to proceed with canonical quantization) like • gφψiγ5 ψ (to conserve parity φ must be a pseudoscalar) • or gφψψ or g∂µ φψγ µ ψ (under parity φ would be a scalar) • or gφψiγ5 ψ + hφψψ (parity violating.22. Some things that are very useful when deriving any basis independent result are orthogonality and completeness conditions for the u(r) and v (r) The up and vp satisfy (p − m)up = 0 ¡ (r) (r) (r) (p + m)vp = 0 ¡ (r) Taking the bar of these equations we also have up (p − m) = 0 ¡ (r) vp (p + m) = 0 ¡ (r) ORTHOGONALITY CONDITIONS We have already derived up γ µ up = 2pµ δrs (r) (s) (r) (s) and vp γ µ vp = 2pµ δrs (r) (s) Now vp γ µ up is also a four vector.

n Then n e(r) e(r)T is the identity matrix. up up = 2mδrs (r) (s) (r) (s) vp vp = −2mδrs (r) (s) (r) (s) up vp = vp up = 0 COMPLETENESS RELATIONS Suppose I have a orthogonal normalized basis for Rn . and since they are scalar. We are going to get the analog of this for our 4 solutions of the Dirac equation for any p. Define A= r up up (r) (r) Let’s see what A is by seeing what it does to a basis for our 4-dim spinor space. so ¡ up up = p + m ¡ r (r) (r) Similarly.22. v’s are β = −1 eigenstates). e(r) r = 1. that gives their value for general p (u’s are β = +1 eigenstates. · · · . vp vp = p − m ¡ r (r) (r) . December 11 Notes from Sidney Coleman’s Physics 253a 261 There are also the scalars up up (r) (s) vp up (r) (s) and vp vp (r) (s) Evaluating them for p = 0 is easy. Aup = r (s) up 2mδrs = 2mup up · 0 = 0 r (r) (r) (s) Avp = (s) But we already know a matrix that has this effect on the basis p + m.

This sheet has been prepared to help you. σ 0 β= 1 0 . Here ψ is a set of four complex fields. arranged in a column vector (a Dirac bispinor). Dirac Matrices The theory is defined by the Lagrange density.22. Dirac Lagrangian. The equation of motion (the Dirac equation) is (i∂0 + iα · σ − βm)ψ = 0. . β = 0. αj = 2δij . (1). α= σ 0 .) Space-Time Symmetries The Dirac equation is invariant under both Lorentz transformation and parity. and the α’s and β are a set of four 4 × 4 hermitian matrices (the Dirac matrices). December 11 Notes from Sidney Coleman’s Physics 253a 262 What Every 253a Students Needs to Know about The Dirac Equation I have heard that some of you have had trouble keeping the Dirac equation in view through a cloud of SO(3. 0 −1 (2. αi . 1) representation theory. 1 0 and the standard representation. Any set of 4 × 4 matrices obeying this algebra is equivalent to any other set. Dirac Equation. 0 −σ β= 0 1 . α= 0 σ . It contains all all results we have derived to date that we will need in the remainder of the course. without proofs. The Dirac matrices obey the Dirac algebra. Two representations of the Dirac algebra that will be useful to us are the Weyl representation. αi . L = ψ † [i∂0 + iα · σ − βm]ψ. β 2 = 1.

Λ. 0 σ (3. In both the Weyl and standard representations L= Under parity. where the matrix D is defined from the α’s by the following rules: For an acceleration by rapidity φ in a direction e.34 D A(eφ) = eα·eφ/2 . γ i = βαi and β 2 = 1 i Lk = 4 kij γ i γ j . where L is defined by35 αi . 34 35 αi = γ 0 γ i . For a rotation by angle θ about an axis e D R(e θ) = e−iL·eθ . Λ: ψ(x) −→ D(Λ)ψ(Λ−1 x). t) −→ βψ(−x.22. 1 2 σ 0 . αj = 4i ijk Lk .) Dirac Adjoint. γ Matrices The Dirac adjoint of a Dirac bispinor is defined by ψ = ψ † β. of a 4 × 4 matrix by A = βA† β. follows from γ 0 = β. P : ψ(x. t). December 11 Notes from Sidney Coleman’s Physics 253a 263 Under a Lorentz transformation characterized by a 4 × 4 Lorentz matrix.

22. In this notation. The γ matrices obey the γ algebra. γ µ† = γµ ≡ gµν γ ν . γ µ. γν They also obey D(Λ)γ µ D(Λ) = Λµν γ ν . These are not all hermitian. the Dirac Lagrange density is ψ(i − m)ψ.g. . ∂ . e. For any vector. µ γ i = βαi . but they are self-Dirac adjoint (“self-bar”). (ψAφ)∗ = φ Aψ. The γ matrices are defined by γ 0 = β. = 2g µν .. a. γ µ = γ µ. we define a & = aµ γ It follows from the γ algebra that a¡ ba &b + ¡& = 2a · b. ∂ and the Dirac equation is (i − m)ψ = 0. December 11 Notes from Sidney Coleman’s Physics 253a 264 These obey the usual rules for adjoints.

The tensor is T µν = ψσ µν ψ. The negative-frequency solutions are of the form ψ = veip·x . γ µ = 0. 2i γ5 . (5). The scalar is S = ψψ. We can choose these sixteen to form the components of objects that transform in simple ways under the Lorentz group and parity.22. γ5 is in many ways “the fifth γ matrix”. The vector is V µ = ψγ µ ψ. where γ5 = iγ 0 γ 1 γ 2 γ 3 ≡ γ 5 . † γ5 = γ5 = −γ5 . 1 µ ν γ . December 11 Notes from Sidney Coleman’s Physics 253a 265 (4). . The axial vector is Aµ = ψγ µ γ5 ψ. It obeys (γ5 )2 = 1.γ . Bilinear Forms There are sixteen linearly independent bilinear forms we can make from a Dirac bispinor and its adjoint. Plane-wave Solutions The positive-frequency solutions of the Dirac equation are of the form ψ = ue−ip·x . where σ µν = The pseudoscalar is P = ψiγ5 ψ. where p2 = m2 and p0 is positive.

(r) (s) (r) (s) . They obey the completeness relations 2 (r) (r) up up r=1 2 (r) (s) (r) (s) and vp . The Dirac equation implies that (p − m)u = 0 = (p + m)v. (r) (s) = p + m. √    0 2m √  0   2m (1) (2)  . This form has a smooth limit as m goes to zero. ¡ ¡ For a particle at rest.  2m 0 v0 (2)  0  0   =  0 . p = (m. 0). These solutions are normalized such that up up = 2mδ rs = −vp vp . √ 2m  (r) We can construct the solutions for a moving particle. ¡ (r) (r) Another way of expressing the normalization condition is up γ µ up = 2δ rs pµ = vp γ µ vp . . ¡ r=1 vp vp = p − m.22. up Lorentz acceleration (see (2)). by applying a (r) up vp = 0.  u0 =  u0 =  0  0  0 0 and the two independent v’s to be   0  0  (1) v0 = √  . we can choose the two independent u’s in the standard representation to be. December 11 Notes from Sidney Coleman’s Physics 253a 266 There are two positive-frequency and two negative-frequency solutions for each p.

23. December 16

Notes from Sidney Coleman’s Physics 253a



December 16
CANONICAL QUANTIZATION OF DIRAC LAGRANGIAN L = ± ψ + (i∂0 + iα · πψ ≡ ∂L = ±iψ † ∂(∂0 ψ)

− βm)ψ

ψ, ψ † completely characterize system ∂L † = ±iψA ) ∂∂0 ψA using E-L Eq

(more generally πA = H = ±iψ † ∂0 ψ − L = ± ψ † (−iα ·

+ βm)ψ = ±iψ † ∂0 ψ

† ±i ψα (x, t), ψβ (y, t) = iδ (3) (x − y)δαβ

α, β = 1, 2, 3, 4

suppress α, β indices

ψ(x, t), ψ † (y, t) = ±1δ (3) (x − y)

ψ(x, t), ψ(y, t) = 0 = ψ † (x, t), ψ † (y, t) ψ(x, t), ψ(y, t) = ±γ 0 δ (3) (x − y)

easily show by reinserting indices 1 (r) (r) (r)† (r) bp up e−ip·x + cp vp eip·x 2Ep 1 (r)† (r)† (r) (r)† bp up eip·x + cp vp e−ip·x 2Ep 1 (r)† (r) (r) (r) bp up eip·x + cp vp e−ip·x 2Ep

ψ(x) =
r=1 2

d3 p

1 (2π)3/2 1 (2π)3/2 1 (2π)3/2

ψ † (x) =
r=1 2

d3 p

ψ(x) =

d3 p

Ansatz: (To avoid doing Fourier inversion): bp , bp
(r)† (r) (s)†

= δ rs δ (3) (p − p ) B = δ rs δ (3) (p − p ) C

cp , cp


23. December 16

Notes from Sidney Coleman’s Physics 253a


b, b = c, c = b, c = 0 =⇒ ψ, ψ = ψ † , ψ † = ψ, ψ = 0 1 (r) (r)† (r) (r)† Be−ip(x−y) up up + Ce+ip·(x−y) vp vp 3 2E (2π) p

ψ(x, t), ψ † (y, t)
Formula also true with ψ † replaced by ψ and u† , v † replaced by u, v


d3 p

up up

(r) (r)†


up up β = (p + m)β ¡

(r) (r)

= (Ep β − p · βα + m)β = Ep + p · α + βm Similarly, vp vp = (p − m)β = Ep + p · α − βm ¡
r (r) (r)

If B = C, terms ∝ α, β vanish, so that we may have integral proportional to 1, choose B = C = ±1, and we have canonical quantization relations: ψ(x, t), ψ † (y, t) = ± d3 p −ip·(x−y) e (2π)3

However, our expression for ψ has two annihilation or creation operators leading to problems with H:
δrs 2Ep 3 δrs 2Ep

H= H=±

d3 x H = ±

d p (r)† (s) (r)† (s) (r) (s)† (r)† (s) (b b u u Ep − cp cp vp vp Ep 2Ep p p p p
(r)† (r) (r) (r)†

d3 p Ep bp bp − cp cp

which is an unbounded below energy. For (+) sign c-type quanta carry negative energy.

5 TOPICS FOR THE REST OF THIS LECTURE 1. Canonical Anticommutation 2. Solves energy crisis 3. Fermi-Dirac statistics

23. December 16

Notes from Sidney Coleman’s Physics 253a


4. Fields as observables 5. Classical Limit ...


Fermi Bose

p’s and q’s have

1 odd 2




At equal times – Bose-Bose [pa , pb ] = [q a , q b ] = 0, [q a , pb ] = iδ ab . – B-F everything commutes. – Fermi-Fermi {pa , pb } = {q a , q b } = 0, {q a , pb } = iδ ab . Hence we make the changes bp , bp = δ rs δ (3) (p − p )B etc. 1 dtp ψ(x, t), ψ † (y, t) = ± ··· (2π)3 r
(r) (s)†


fr (p) bp {A, A† } = ±


|fr (p)|2

LOOK AT QUANTITY OF THE FORM φ| A, A† |φ = φ|AA† |φ + φ|A† A|φ ≥ 0

Hence we must choose + sign

(2) L = ψ(i − m)ψ ∂ bp , bp
(r) (s)†

= cp , cp



= δ rs δ (3) (p − p )

all others zero

23. December 16

Notes from Sidney Coleman’s Physics 253a



d3 p Ep bp bp + cp cp

(r)† (r)

(r)† (r)

− ¨¨(0) δ (3) ¨


from anticommutation of C and C †

Which is bounded below.

(3) Consider pedagogical simplification of Hamiltonian for a moment. H=
p † Ep bp bp † bp , bp = δp,p ,

b, b = b† , b† = 0

AB, C = A B, C − A, C B H, bp = −Ep bp
energy lowering † † H, bp = +Ep bp energy raising


bp |0 = 0 all p,

0|0 = 1
† bp |0 = |p

H|0 = 0 H|p = Ep |p

† † p |p = 0|bp bp |o = − 0|bp bp |0 + δp,p = δp .p † † |p1 , p2 = bp1 bp2 |0 = −|p2 , p1

H|p1 , p2 = (Ep1 + Ep2 )|p1 , p2 Pauli Exclusion principle
† † bp |p = (bp )2 |0 = 0

23. December 16

Notes from Sidney Coleman’s Physics 253a


(4) Observable made out of Fermi fields. Recall Bose Fields φ(x), φ(y)


and by Lorentz invariance this is true for all spacelike separated x and y. With Fermi fields ψα (x), ψβ (y) = 0 If ψ(x) were an observable, we would have observables that did not commute at spacelike separation. Observables can only be made of products with an even number of Fermi Fields. This is also necessary for just rotational properties. Under a rotation by 2π, ψ(x) −→ −ψ(x) No meter on any experimental apparatus ever gives a different reading when the experiment is rotated by 2π. All observables are in single valued representations of the Lorentz group. (4) Classical Limit ( −→ 0) (a) Two classical limits physically. Take some physical situation. N particles in a box all in the same energy and momentum eigenstate. E=N ω (i) −→ 0 N, E and p fixed ω, k −→ ∞ wavelength −→ 0

No diffraction. This is the classical particle limit. (ii) −→ 0 E, ω and p, k fixed. N −→ ∞ Lots of wavy behavior, but lose quantum granularity.

For fermions we can only do the first limit because of the Pauli exclusion principle. There is no analog of the wave limit. There will never be a competing theory as there was for light with the corpuscular and wave theories.

23. December 16

Notes from Sidney Coleman’s Physics 253a


(b) Classical limits formally. −→ 0 in canonical algebras

Bose fields −→ commuting quantities (numbers) Fermi fields −→ anticommuting quantities (Grassmann variables)

Working with classical Fermi Fields or Grassmann variables Never exchange the order of two terms in the classical field equations without a compensating minus sign or you would have no hope of the classical limit of the quantum theory agreeing with the classical theory even at order 0 . Derivation of the Euler-Lagrange equations: dL = ∂L a ∂L a dq + a dq ˙ ∂ qa ˙ ∂q

If both the derivatives are kept to the same side of the differentials then I can integrate by parts in the action and get the usual E-L equations. pa = ˙ rather than something else. Define H = pa q a − L ˙ Not H = q a pa − L, for example, which would give something different. ˙ dH = dpa q a − pa dqa − dL ˙ ˙ ∂H = qa ˙ ∂pa ∂H = −pa ˙ ∂q a ∂L ∂q q

Try using the quantum relations pa = −i pa , H , q a = −i q a , H with the canonical ˙ ˙ anticommutation relation, and see if you can reproduce the Heisenberg equation of motion in the Dirac theory.

this definition of time ordering is a failure. December 18 Notes from Sidney Coleman’s Physics 253a 273 24 December 18 Perturbation theory for spinors Because scalar fields commute at spacelike separations. In one frame ¯ ¯ T (ψα (x)ψβ (0)) = ψα (x)ψβ (0) . When x2 > 0. the idea of time ordering is Lorentz invariant. T (φ(0)φ(x)) = φ(x)φ(0) In the situation pictured.24. For spinor fields. whose axes are represented by dotted lines x0 is less than 0 and x T (φ(0)φ(x)) = φ(0)φ(x) This ambiguity is not a problem for scalar fields since [φ(0). That is x If x2 < 0. φ(x)] = 0 when x2 < 0. there is no ambiguity. but in another frame.

we used Wick’s theorem to turn the time ordered products into normal ordered products. so you always move spinor fields around in pairs. −∞) in the formalism with the turning on and off function. Assertion: Most of the derivations we did. For example if A1 and A2 are Fermi fields : A1 A2 := A1 A2 (+) (+) + A1 A2 (−) (+) + A1 A2 (−) (−) − A2 A1 (−) (+) . A way of seeing that this is probably true is to think about how we obtained S = UI (∞. expressing S matrix elements in terms of physical vacuum expectation values of time ordered products of renormalized Heisenberg picture fields. December 18 Notes from Sidney Coleman’s Physics 253a 274 and if x2 < 0. About the only place we could have problem is in obtaining the expression UI (∞. Once we have that big messy expression on the RHS above. −∞) = T e−i ∞ ∞ d 4 x HI But you expect no problem there because the Hamiltonian is quadratic in spinor fields.24. and showing that 0|T (φH (x1 ) · · · φH (xn ))|0 0|T φI (x1 ) · · · φI (xn )e−i 0|T e−i ∞ ∞ ∞ ∞ d4 xHI |0 p p = d4 xHI |0 bare vacuum are unaffected by the fact that the fields may now be spinors and the time ordered product now has a (−1)p in it. and then wrote down Wick diagrams to represent operations in the Wick expansion. in another frame it may be that ¯ ¯ ¯ T (ψα (x)ψβ (0)) = ψβ (0)ψα (x) = −ψα (x)ψβ (0) OH OH! The way to patch this up is to put an extra minus sign into the definition of the time ordered product whenever the number of permutations of Fermi fields required to turn a product into a time ordered product is odd. and the permutation is thus always even. and Feynman diagrams to represent S matrix elements. Assertion: Wick’s theorem can be proven for spinor fields and with the extra minus sign in the time ordered product provided you also put an extra minus sign in the normal ordered product. The new minus sign in the time ordered product doesn’t matter.

Here is a proof by cases. take . Take x0 > x0 . 1 2 The case x0 > x0 clearly goes through too. we can use the same trick for evaluating it as we did when we calculated the contraction of two scalar fields. The contraction A1 A2 is defined as usual to be the time ordered product minus the normal ordered product A1 A2 = T (A1 A2 )− : A1 A2 : The contraction is a c-number. i. A2 } NOTATION: : A1 A2 A3 A4 :≡ − A1 A3 : A2 A4 : − sign for odd permutations is needed to prove Wick’s theorem. Things are looking good though. December 18 Notes from Sidney Coleman’s Physics 253a 275 Note that : A1 A2 = − : A2 A1 : also T (A1 A2 ) = −T (A2 A1 ) (FERMI FIELDS ANTICOMMUTE INSIDE THE TIME ORDERED AND NORMAL ORDERED PRODUCTS). A2 ] For x0 > x0 . 2 1 (+) (−) (+) (−) (−) Calculation of the contraction (propagator) Since the contraction of two Fermi fields is a c-number. {A1 .e. then 1 2 T (A1 A2 ) = A1 A2 = A1 A2 and A1 A2 = A1 A2 (+) (+) (−) (+) (+) + A1 A2 (+) (−) + A1 A2 (−) (+) + A1 A2 (+) (−) + A2 A1 (−) (+) = {A1 . If we hadn’t stuck that extra − sign into the definition of the ordered product we would have gotten A1 A2 = [A1 . which is not a c# .24. A2 } is a c# .

In either case ψ(x)ψ(y) = (ix + m) φ(x)φ(y) ∂ contraction of a scalar field of mass m . you get dp e−ip·(x−y) (2π)3 2Ep ψ(x)ψ(y) = − d3 p eip·(x−y) (p − m) ¡ (2π)3 2Ep d3 p ∂ = (ix + m) eip·(x−y) (2π)3 2Ep (∗) (*): CRITICAL MINUS SIGN OUT FRONT IS THE ONE WE PUT INTO OUR TIME ORDERED PRODUCT FOR FERMI FIELDS The nice thing about this is that it is the same for x0 less than or greater than y 0 (the sign of p in the exponential no longer matters once p is turned into a derivative and pulled ¡ out). ψ(x) = r d3 p (2π)3/2 d3 p r 2Ep 2Ep bp up e−ip·x + cp vp eip·x bp (r )† (r ) ip ·y up e (r) (r) (r)† (r) ψ(y) = For x0 > y 0 . (2π)3/2 + cp vp e−ip ·y (r ) (r ) ψ(x)ψ(y) = 0| ψ(x)ψ(y) |0 = 0| T ψ(x)ψ(y) − $$$$ : ψ(x)ψ(y) : |0 0 $ $$ = 0|ψ(x)ψ(y)|0 d3 p d3 p (r) (r )† (r) (r ) = 0|bp bp |0 e−ip·x eip ·y up up 3/2 3/2 (2π) 2Ep (2π) 2Ep rr δrr δ (3) (p−p ) = d3 p e−ip·(x−y) (2π)3 2Ep 3 up up r p+m ¡ (r) (r) = (ix + m) ∂ For y 0 > x0 . December 18 Notes from Sidney Coleman’s Physics 253a 276 vacuum expectation value.24.

1) Both sides of this equation are 4 × 4 matrices. ory for Dirac fields as 2 p − m2 + i Recall though that when you wrote down p2 charged scalar ←p ¦ which stands for p2 i − m2 + i it did not matter whether p was routed in the same direction or the opposite direction as charge flow.24. December 18 Notes from Sidney Coleman’s Physics 253a 277 We’ve already massaged φ(x)φ(y): φ(x)φ(y) = d4 p −ip·(x−y) i e 4 2 − m2 + i (2π) p So without further ado we can rewrite ψ(x)ψ(y) = (ix + m) ∂ = d4 p −ip·(x−y) i e (2π)4 p2 − m 2 + i 4 d p i(p + m) −ip·(x−y) ¡ e 4 p2 − m2 + i (2π) (24. ψ α (x)ψβ (y) = d4 p i(pαβ + mIdαβ ) −ip·(x−y) ¡ e 4 p2 − m2 + i (2π) 3 Comments on the Propagator (1). If you prefer. charged fermion ←p while charged fermion →p ¦ ¦ will stand for p2 i(p + m) ¡ − m2 + i will stand for i(−p + m) ¡ p2 − m2 + i The propagator is a kind of projection operator. . The propagator i(p + m) ¡ is going to play the same role in the perturbation the− m2 + i i played in the perturbation theory for scalar fields. simply because (−p)2 i i = 2 2+i −m p − m2 + i Now however the propagator is not even in p.

the action is S= ∂ d4 x ψ(i − m)ψ In momentum space i−m becomes p−m. 1 Feynman diagrams in L = ψ(i − m)ψ + 2 (∂µ φ)2 − ∂ µ2 2 φ 2 − gψΓψφ (Γ = 1 (meson is a scalar) or iγ5 (meson is a pseudoscalar).) . ¡ ¡ (3). more common because it is shorter. i. For bosons the action is S= parts integration = d4 x (∂µ ψ ∗ ∂ µ ψ − m2 ψ ∗ ψ) d4 x ψ ∗ (−2 − m2 )ψ In momentum space. m > 0). Then adding this here does nothing to location of pole p2 i(p + m) i(p + m − i ) i ¡ ¡ = = 2+i −m (p − m + i )(p + m − i ) p−m+i ¡ ¡ ¡ There is no danger that these hokey looking matrix manipulations are wrong because p − m + i commutes with p + m − i .24. For fermions. −2 − m2 becomes p2 − m2 . There is more common way of writing the fermion propagator. December 18 Notes from Sidney Coleman’s Physics 253a 278 (2).e. Rewrite p2 − m2 + i = (p − m + i )(p + m − i ) ¡ ¡ (this still gives the right prescription for the location of poles as → 0. when acting on e−ip·x . The fact that the charged boson propagator ∂ ¡ i i came out to be p2 −m2 while the charged fermion propagator came out to be p−m is at ¡ least an interesting coincidence.

24. that is what the r does. and this expression clearly differs from the one in the first term by an exchange of dummy indices 1 ←→ 2.e. r. q Notice that the incoming and outgoing electrons have to have their spin specified as well as their momentum. I’ll just look at how the first term contributes to N + φ −→ N + φ (N : an electron or “nucleon”) since the second one is identical and serves only to cancel the 2!. December 18 Notes from Sidney Coleman’s Physics 253a 279 We’ll “derive” the Feynman rules by looking at a couple of processes and hoping that the general process at general orders has an obvious generalization. Is the second operator the same as the first? Well. Let’s look at the order 4 g term in T e−i d xHI . let’s look at how it contributes to N +φ −→ N +φ. [Alternatively. q |(−ig)2 d4 x1 d4 x2 : ψΓ ψφ(x1 )ψ Γψφ(x2 ) : |p. The relevant terms in the Wick expansion of the time ordered product are (−ig)2 2! d4 x1 d4 x2 : ψΓ ψφ(x1 )ψ Γψφ(x2 ) : + : ψΓψφ(x1 )ψΓψ φ(x2 ) : The Feynman diagrams for these two terms are 1 2 and 2 1 respectively. you could just give an arbitrary . p . r . (−ig)2 2! d4 x1 d4 x2 T ψΓψφ(x1 )ψΓψφ(x2 ) This can contribute to many processes. : ψΓψφ(x1 )ψΓψ φ(x2 ) :=: ψΓ ψφ(x2 )ψ Γψφ(x1 ) : because interaction Hamiltonians are made out of fermion bilinears which commute inside the normal ordered product. i. We’ll look at the matrix element of the operator between relativistically normalized states. The picture for the second term looks identical to the first picture (they are of the same pattern) except for an exchange of the dummy variable 1 ←→ 2.

say u and u . which I’ll use and our matrix element simplifies to (−ig)2 d4 x1 d4 x2 i d4 l ip ·x1 −ip·x2 −il·(x1 −x2 ) (r ) (r) e e e up Γ Γup q | : φ(x1 )φ(x2 ) : |q 4 (2π) l−m+i £ eiq ·x1 e−iq·x2 + x1 ↔ x2 Let’s just consider the first one for a second. The exponentials go along with an interpretation.24. (−ig)2 d4 l i (r ) (r) (2π)4 δ (4) (p − l + q )(2π)4 (−p + l − q)up Γ Γup 4 (2π) l−m+i £ The l integration can be done because the energy momentum of the internal electron is fixed by the δ function. This can happen at any space-time points x1 and x2 . so they are integrated over. At x1 an electron with momentum p and a meson with momentum q are created and a virtual electron with momentum l is absorbed (+ signs go with creation. ψ(x2 ). r (2π)3 2ωl d3 l (s) u e−il·x2 (2π)3 2ωl l 0|b(s) (l)|p. We also have derived an expression for ψ(x1 )ψ(x2 ).] There is only one field in the normal ordered product that has an annihilation operator that can annihilate the incoming electron. r = 0| s d3 l (s) b(s) (l)ul e−il·x2 |p. and in fact the integrals are trivial to do. − with absorption). The integral is (−ig)2 d4 x1 d4 x2 i d4 l i(p −l+q )·x1 i(−p+l−q)·x2 (r ) (r) e Γup e up Γ 4 (2π) l−m+i £ (r ) There are two terms in q | : φ(x1 )φ(x2 ) : |q . r (2π)3 2ωp δ (3) (p−l)δsr = s (r) = up e−ip·x2 Similarly when ψ(x1 ) is used to create the outgoing electron it becomes up e+ip ·x1 because that is the coefficient of b(r ) (p )† in the expansion of ψ(x1 ). December 18 Notes from Sidney Coleman’s Physics 253a 280 spinor for the incoming and outgoing electrons. We get (−ig)2 (2π)4 δ (4) (p + q − (p + q))up Γ (r ) i (r) Γup p+q−m+i ¡ ¡ . 0|ψ(x2 )|p. while a virtual electron with momentum l is created. At x2 an electron with momentum p is absorbed and a meson with momentum q is absorbed.

December 18 Notes from Sidney Coleman’s Physics 253a 281 The Feynman diagram for this contribution to N + φ −→ N + φ is What about the second term in q | : φ(x1 )φ(x2 ) : |q ? It contributes to the same process. But. q (r ) (r ) The ψ field then only has to get by two (0) Fermi fields to annihilate the positron on the (r) (r) right. it has to be anticommuted past (r )† three (actually 1: two have been contracted) fermionic fields to do it. r . The coefficient of cp in ψ(x2 ) is eip·x2 vp . r. q incoming positron and meson The field ψ(x2 ) has to create the outgoing positron. p . q | (−ig 2 ) outgoing positron and meson d4 x1 d4 x2 : ψΓ ψφ(x1 )ψ Γψφ(x2 ) : |p. The coefficient of cp in ψ(x1 ) is e−ip·x1 bp so I get −(−ig)2 d 4 x1 d 4 x2 d4 l −ip·x1 ip ·x2 −il·(x1 −x2 ) (r) i (r ) e e e vp Γ Γvp q | : φ(x1 )φ(x2 ) : |q 4 (2π) l−m+i £ . r. but the diagram is different: The contribution is (−ig)2 (2π)4 δ (4) (p + q − (p + q))up Γ (r ) i (r) Γup p−q −m+i ¡ ¡ Let’s look at another process N + φ −→ N + φ that the exact same operator in the Wick expansion can contribute to.24. so we get − q |(−ig)2 d4 x1 d4 x2 : ψΓ ψφ(x1 )ψ Γvp eip ·x2 φ(x2 ) : |p.

If the initial state had been specified simply by some spinor v (rather than a type r. that is linear in v. and the change from u’s to v’s. →p f p q p Feynman Rules . There are still going to be two Feynman diagrams from the two terms in q | : φ(x1 )φ(x2 ) : |q . I’ll just write down the result of doing the x1 . x2 and l integrals. December 18 Notes from Sidney Coleman’s Physics 253a 282 The main difference to notice for this process is the overall minus sign. the amplitude is antilinear in v. −(−ig 2 )(2π)4 δ (4) (p + q − (p + q ))vp Γ (r) i i + −p − q − m + i −p + q − m + i ¡ ¡ ¡ ¡ (r) Γvp (r ) The initial antinucleon state gave us a vp .24. (Why do we expect this?) i internal nucleon p−m+i oriented along arrow (along charged flow for positively ¡ charged particles). one of our basis spinors).Factors internal meson i q 2 − µ2 + i p Integrate over internal momenta ¤f i ←q −igΓ(2π)4 δ (4) (p + q − p ) • For every incoming nucleon (annihilated by a ψ) get a u • For every incoming antinucleon (annihilated by a ψ) get a v • For every outgoing nucleon (created by a ψ) get a u • For every outgoing antinucleon (created by a ψ) get a v .

) Working against the arrows. . December 18 Notes from Sidney Coleman’s Physics 253a 283 Because fermions always appear bilinearly (we’ll soon see quadrilinears are ruled out) in a Lagrangian.24. Feynman Rules .Matrix multiplication Go to the head of any fermion line that goes all the way through a graph. EXAMPLE (2) l1 (4) (3) (5) (6) (8) u' (1) l2 l2 (7) (9) u (6): mass renormalization. the next thing you get is a vertex. you’ll either be at an incoming nucleon. so I start with the row vector then go through the matrices and finish with a column vector. comes from −Bψψ term in LCT . Write down the matrix for the vertex. a fermion line either goes all the way through a graph or in a loop. You will either be at an incoming antinucleon. When you get to the tail end of the line. Write down the propagator matrix. u (1) Γ (2) i l1 − m + i £ (3) Γ (4) i 21 − m + i ¡ (5) Id (6) i 21 − m + i ¡ (7) Γ (8) u (9) What you finally obtain is a number. we’ll just do the matrix multiplication rules for a line going all the way through a graph first. (I start at the head of the line because I habitually write from left to right. then the interaction matrix each time. Since we haven’t done any examples with a fermion loop yet. in which case you write down a v. You get another product like this for each Fermi line that goes through a graph. in which case write down a u. or at an outgoing nucleon in which case you write down a u. in which case write down a v. or at an outgoing antinucleon. Then you’ll get an internal line followed by another vertex some number of times.

r. You get a factor like this for each Fermi loop. To conclude the matrix multiplication rules the rule is thus: If a Fermi line goes in a loop. and working against the arrows write down vertex and propagator matrices until you get back to where you started. q . This is why there is a minus sign out front. 28. each Fermi loop gives you a minus sign. First ψ(x4 ). fdy#df d¦e ¢pf e d Feynman Rules + Fermi minus signs One thing is for sure.1) is the expression whose spinor analog is going to give us some troublesome minus signs. To get the rest of the minus signs. Back on Oct. s 2! . I not only had to move ψ(x4 ) all the way to the left and write the sum on α p−m ¡ in Γβα ψα (x4 ) as a trace. q. December 18 Notes from Sidney Coleman’s Physics 253a 284 What about Fermi lines that go in loops? Here is a graph: Another is This is an order g 4 diagram for 2 meson −→ 2 meson scattering. I had to anticommute ψ(x4 ) by an odd number of Fermi fields to get it there.24. To put this in standard form for replacing ψψ by our integral i over . We need to simplify: 1 p . then a bunch of bilinears. start anywhere in the loop. Then take the trace to get a number. The factor of interest in this contribution to the process is ψΓψ(x1 )ψΓψ(x2 )ψΓψ(x3 )ψΓψ(x4 ) I’ll rewrite this as −Tr ψ(x4 )ψΓψ(x1 )ψΓψ(x2 )ψΓψ(x3 )ψ(x4 )Γ Notice the minus sign. I’ll do an example. There are no u’s or v’s in the factor coming from a Fermi loop. Eq. r . s | : ψΓψ(x1 )ψΓψ(x2 ) : |p. We’ll do the analogous calculation for nucleon-nucleon scattering (no quotes). (11. we did “nucleon”“nucleon” scattering at O(g 2 ).

s | = (|p . q. not negative. so what I have is (2π)3 2Eq 0| br (p ) us Γus ψ(x1 ) Γur |0 e−iq·x2 e−ip·x1 eiq ·x2 q p q . q . If ψ(x1 ) annihilates the nucleon with momentum q. I’ll e−iq·x2 1 . s | to be the corresponding bra. r . These two possibilities are different because a distinction between x1 and x2 has now been made by the way we annihilate the incoming nucleon. s and p .as long as you choose p . Suppose ψ(x2 ) takes care of the outgoing nucleon with momentum q . The coefficient of bs (q) in ψ(x2 ) is us q (2π)3 2Eq what I have is 0|br (p )bs (q ) : ψΓψ(x1 )ψ(x2 )Γus : br (p)† |0 e−iq·x2 q Move the bilinear past ψ(x2 ) and let ψ(x1 ) annihilate the remaining incoming nucleon: 0|br (p )bs (q ) : ψ(x2 )Γus ψ(x1 )Γur : |0 e−iq·x2 e−ip·x1 q p Now either ψ(x2 ) or ψ(x1 ) can take care of the outgoing nucleon with momentum p . p . r. q . q. So what we have to simplify is 1 0|br (p )bs (q ) : ψΓψ(x1 )ψΓψ(x2 ) : bs (q)† br (p)† |0 2! ψ(x1 ) and ψ(x2 ) both contain operators which could annihilate either of the incoming nucleons. r . s = b(p)† bs (q)† |0 and {br (p). r . r. December 18 Notes from Sidney Coleman’s Physics 253a 285 It is slightly ambiguous to write |p. but this does not just give us another factor of 2. q.24. and the only difference will be x1 ←→ x2 . bs (p)† } = (2π)3 δ rs δ (3) (p − p )Ep |p. r. we can rewrite what follows. r . s = bs (q)† br (p)† |0 = −br (p)† bs (q)† |0 It doesn’t matter which choice you take . The two possibilities for the ket are relativistically normalized |p. s )† = 0|br (p )bs (q ) That way the forward scattering amplitude when there is no interaction (or at zeroth order when there is) is positive.I’ll take the second . Let’s say ψ(x2 ) annihilates the nucleon with momentum q. q . s |. Since these are dummy variables in an otherwise symmetric integration. q . so ignore this second case and cancel the 2! . The coefficient of us e+iq ·x2 q bs (q )† in ψ(x2 ) is .

but of course you won’t have to be so detailed when you are just checking how many reorderings of Fermi fields it takes to get a given contribution to a matrix element.r us eiq ·x1 q . you can continue on by doing the x integrations in the fashion leading up to Eq. The graph is p'. but the graph these factors are from is p'.24. (11. . q is absorbed at the same spacetime point as q is created.r' p.s' q.s' q.s p is absorbed at the same spacetime point as p is created. The exponential factors are different in the expected way. This was how reordering Fermi operators gives minus signs. December 18 Notes from Sidney Coleman’s Physics 253a 286 The final simplification gives us Γ us ur Γ ur e−iq·x2 e−ip·x1 eiq ·x2 eip ·x1 q p p q There are other factors. So I get (2π)3 2Eq q'.6) for the “nucleon”-“nucleon” scattering. What about the contribution where ψ(x) takes care of the outgoing nucleon with momentum q? First I’ll anticommute the two ψ fields to get: − 0||br (p )bs (q ) : ψ(x1 )Γur ψ(x2 )Γus |0 e−iq·x2 e−ip·x1 p q Now the coefficient of bs (q )† in ψ(x1 ) is The final simplification gives −us Γur ur Γus e−iq·x2 −ip·x1 +iq ·x1 +ip·x2 p p q q The differences with the previous expression worth noting are the minus sign and the different spinor structure.r q'.r' p. I don’t have a tidy little rule to summarize the sign of the general case.s Having obtained the spinor factor and the minus sign.

u q p'.u' p-q p.24.u q' . December 18 Notes from Sidney Coleman’s Physics 253a 287 Γ = iγ 5 EXAMPLE: COMPLETE EXPRESSION FOR N + N −→ N + N p'u' 1 1 p u 1 1 p'u' 2 2 + crossed graph p2u 2 iA = (−ig)2 u1 iγ 5 u1 u2 iγ 5 u2 i (p1 − p1 )2 − µ2 (+1) + u1 iγ 5 u2 u2 iγ 5 u1 i (p1 − p2 )2 − µ2 (−1) EXAMPLE: COMPLETE CALCULATION OF N + φ −→ N + φ q' q Γ = iγ 5 p'.u' p+q p.

UNKNOWN. p} = 0 ¡ USING pu = mu AND u p = u m ¡ ¡ p −q+m p+q+m ¡ ¡ ¡ γ 5u + ¡ 2 − m2 (p + q) (p − q)2 − m2 −p − q + m −p + q + m ¡ ¡ ¡ ¡ u + 2 − m2 (p + q) (p − q)2 − m2 1 1 − 2 − m2 (p − q) (p + q)2 − m2 = g 2 u qu ¡ SPIN AVERAGING AND SPIN SUMMING INITIAL (RESP. SUM) TRANS.24. u2 δ ] using spin 1 I : collective indices u1 α u1 β = (p1 + m)αβ ¡ u2α u2β = (p2 + m)αβ ¡ |M |2 = u2 γ γ 0 O† γ 0 u1 u1 O p1 + m O = γ 0 O† γ 0 ¡ 2m = Tr γδ u2 δ u2 δ [ using spin 2 p1 + m p2 + m O¡ γ0 O† γ 0 ¡ 2m 2m In other calculations you will need v3 v 3 = p3 − m ¡ spins 1 2 r s 1 |ars |2 = |F |2 2 Tr up qup up qup ¡ ¡ r.e. Average or sum probability. FINAL) SPINS IN EXPT.s (s) (s) (r) initial av. AVERAGE (RESP. final sum 1 = |F |2 Tr q(p + m)q(p + m) ¡ ¡ ¡ ¡ 2 1 = |F |2 (4m2 µ2 + 8p · qp · q − 4p · p µ2 ) 2 36 Do not sum or average the Feynman amplitude. PROB OVER THEM36 MATRIX KINEMATIC FACTORS M = u1 OI u2 ( )I γδ i. December 18 Notes from Sidney Coleman’s Physics 253a 288 iA = (−ig)2 u iγ 5 A = g2u γ 5 A = g2u i i iγ 5 u + u iγ 5 iγ 5 u p+q−m+i p −q−m+i ¡ ¡ ¡ ¡ USING (γ 5 )2 = 1 AND {γ 5 . A = u1 γ µ u2 qµ . .

It has no spinor indices. βu0 = u0 comes right from (p − m) = 0 ¡ for p = (m. t) −→ UP ψ(x. January 6 Notes from Sidney Coleman’s Physics 253a 289 25 January 6 P . t)UP = βψ(−x. C. t) UP is a unitary operator in Hilbert space. it reverses the direction of motion but doesn’t do anything to the spin. So the effect of UP on bp must be † UP bp UP = b−p (r) (r) (r) . ψ(x. But that’s easy. t). (A Dirac field has a spinor index. and P T for spinor fields We know what P does to a Dirac’s field P : ψ(x. we get the effect of UP on the creation and annihilation operators. βup = βeα·e φ/2 u0 (r) (r) (r) (r) (r) (r) (r) (r) e= p |p| sinh φ = |p| m = e−α·e φ/2 βu0 = = (r) e−α·e φ/2 u0 (r) u−p (r) (because {β. t) −→ βψ(−x. αi } = 0) We have shown that parity does to positive frequency solutions of the Dirac equation what you would expect it to. The first step is to evaluate βup and the first step to evaluating that is to find βu0 . 0). t) We expect (general QM theorem coming from uniqueness of canonical commutation relations) that in the quantum theory there is a unitary operator effecting this change P : † ψ(x. each entry in the Dirac field is itself an operator in Hilbert space.25.) From the effect of UP on ψ and the expansion of ψ in terms of creation and annihilation operators. To get the effect of β on up we now use up = eα·e φ/2 u0 . t) = r d3 p (r) (r) (r)† (r) bp up e−ip·x + cp vp eip·x 3/2 2E (2π) p positive frequency negative frequency solution of Dirac equation (r) We need an expression for βψ(x.

UP cp UP = −C−p . There are two possible l = 0 states. Is this just an artifact of some unfortunate convention37 . . UP acting on a state with n elections does nothing. Fermion and antifermion have opposite intrinsic parity. and l = 1 is an odd function of relative momentum so it has negative parity. The nucleon and antinucleon are taken to be at rest. equation UP bp UP = b−p ) A very similar argument goes through for the cp which are multiplied by vp except for one thing (r) (r) βv0 = −v0 Because of that minus sign the effect of UP on cp † h. eqn).c.) The fact that the pion is pseudoscalar doesn’t affect the outcome because there are two (an even number) of them. January 6 Notes from Sidney Coleman’s Physics 253a (r)† (r)† 290 † (and the h. ¯ Consider the process N + N −→ 2π. in general you get (−1)l . either s = 0 or s = 1. because this implies (no momentum −→ no angular momentum) that they are in an l = 0 state. but the strong interactions which 37 i. but the c’s have negative intrinsic parity.e.25. Looking at the possibilities. (r) (r) (r)† † is UP cp UP = −C−p (and the (r)† (r)† (r)† (r) The b’s have positive intrinsic parity. l=0 l=0 s=0 s=1 J =0 J =1 P = −1 P = −1 The total angular momentum J is absolutely conserved. could I redefine P to be P x some internal symmetry which would not have this relative minus sign. This is unlike the charged scalar. The two pseudoscalar pions either have l=0 l=1 J =0 J =1 P = +1 P = −1 (l = 0 is an even function of relative momentum so it is positive parity. where both particle and antiparticle were scalar or pseudoscalar. in particular at the J and P . UP acting on a state with n positrons gives (−1)n . (Except by the P violating weak interaction. you see that this process is forbidden. and the parity is −1 because of what we have just found. or does this counterintuitive result have observable consequences.c.

so there are lots of possibilities. You already know four matrices with square one that anticommute with each other: 1 0 0 σ and 0 −1 σ 0 Of these four. and the others are pure real.) In the Dirac theory $ (i − m)ψ = 0$$  − m)ψ ∗ = 0 ∂ ⇐⇒(i∂ Unless we are in a representation in which all of the γ µ are purely imaginary. it just means it takes a more complicated form.] I’ll show that a representation in which all of the γ µ are purely imaginary exists by constructing one. In the free case: (2 + m2 )ψ = 0 ⇐⇒ (2 + m2 )ψ ∗ = 0 This follows because 2 + m2 is real. [This doesn’t mean that a symmetry worth calling charge conjugation does not exist in a general basis. not create them. you could see that ψ ∗ (x) was a solution of the equation of motion whenever ψ was. Any real change of coordinates preserves this property. Alternatively. when you add interactions you break symmetries. So σy 0 γ2 = i 0 σx σx 0 γ3 = 0 σz σz 0 1 0 0 −1 . This is a convention independent consequence of the opposite intrinsic parity of N and ¯ N. one is pure imaginary γ0 = 0 σy σy 0 γ1 = i 0 σy . (The free case is worth looking at because in general.25. January 6 Notes from Sidney Coleman’s Physics 253a 291 are P conserving and which would make this process occur very quickly compared to the weak interaction aren’t allowed to do it). C Charge conjugation Recall how charge conjugation acted on a charged scalar: C: ψ(x) −→ ψ ∗ (x) (ψ ∗ (x) −→ ψ(x)) was a symmetry of the action.

Let ψs = Sψm ∗ C : ψm −→ ψm ψm : Dirac spinor in Majorana basis. not only have we been using † to mean complex conjugation of numbers and hermitian conjugation of operators. If you like. ψs : some other basis like the standard one. γ ν = 2g µν and γ µ ∗ = −γ µ . What does C do to ψs ? C: ∗ ∗ ψs −→ Sψm = SS ∗−1 S ∗ ψm = SS ∗−1 (Sψm )∗ ∗ = SS ∗−1 ψs The matrix SS ∗−1 is usually denoted C. and so as not to completely jettison the approach that most books take to charge conjugation. UC . so we’ll use ψ ∗ (x) to mean hermitian conjugation without transposition in spinor indices. The representation dependent computations that we’ll do are so much simpler. will exist such that † UC ψ(x)UC = ψ ∗ (x) Why write ψ ∗ (x) and not ψ † (x)? Because we have been thinking of ψ † as a row vector. I’ll show what charge conjugation looks like in a general basis. we expect a unitary operator. We don’t always want to do a transpose when we hermitian conjugate a spinor. the “Majorana” condition. January 6 Notes from Sidney Coleman’s Physics 253a 292 satisfies γ µ .] LORENTZ TRANSFORMATION IN A MAJORANA BASIS iγ 0 γ i iαi = 2 2 Lk = −i ijk Mi Mj Mi = so Mi = −Mi∗ so Lk = −L∗ k D(A(eφ)) − e−iM ·eφ = D(A)∗ . In this basis C : ψ(x) −→ ψ ∗ (x) is a symmetry of the free equation of motion and the free Dirac action (provided the classical field is thought of as anticommuting). [There is always a similarity transformation between any two sets of γ µ (that’s Pauli’s theorem).25. Thinking quantum mechanically. that this is all we’ll have to say about C. but transpose in spinor indices. ψ ∗ (x) = ψ † T (x).

¡ complex conjugated in a Majorana basis. January 6 Notes from Sidney Coleman’s Physics 253a D(R(e θ)) = e−iL·eθ = D(R)∗ i. D(Λ) = D(Λ)∗ real 293 Charge conjugation commutes with Lorentz transformations. i. (r) (p − m)up = 0. So to make the action of charge conjugation simple. Let’s see what this implies about the effect of charge conjugation on spins. This says that the complex conjugate of a solution of the Dirac equation that has positive frequency is a solution of the Dirac equation with negative frequency. Thus we need to know what up and vp are.25. From the eqn.e.e. UC ψ(x)UC = ψ ∗ (x).) We want to find the effect of UC on creation and annihilation operators. It shouldn’t have any effect on spin. We are free to choose the up and vp any way we want. (The utility of this simple little result will become clear in the following.e. 2 1 (1) (1) Lz v0 = − v0 2 Since u multiplies an annihilation operator and v multiples a creation operator. we have (−p − m)up ¡ (r)∗ = 0 i. this is exactly what we expect for charge conjugation. You get that from the expansion of ψ in terms of creation and annihilation operators and the assumed † effect of UC on ψ. (p + m)up ¡ (r)∗ = 0. choose (r) (r)∗ vp = up This is consistent with up = D(A(eφ))u0 vp = D(A(eφ))v0 (r) (r) (r) (r) (r) (r) because charge conjugation commutes with Lorentz transformation (in a Majorana basis). ψ(x) = r d3 p (2π)3/2 2Ep bp up e−ip·x + cp vp eip·x (r) (r) (r)† (r) . From the complex conjugate of 1 (1) (1) Lz u0 = u0 2 (1)∗ (1)∗ we have −Lz u0 = 1 u0 . We need to get the expansion of ψ ∗ (x) in terms of (r)∗ (r)∗ creation and annihilation operators explicitly.

p )cp bp |0 if F (p p) = ±F (p. Thanks to the way we set up the correspondence. January 6 Notes from Sidney Coleman’s Physics 253a 294 So. Construction of nucleon-antinucleon state Scalar case (“nucleon” -“antinucleon” state for warm up and comparison) creates “nucleon” † cp † bp |0 adds an “anti-nucleon” † † † † UC cp bp |0 = bp cp |0 |ψ ≡ Then UC |ψ = ±|ψ † † d3 p d3 p F (p. Spin ups transform into spin ups and spin downs transform into spin downs. equations are † UC bp UC = cp (r)† (r)† † and UC cp UC = bp (r) (r) This couldn’t be simpler. p ) . exactly as if the spin up electron was a boson whose antiparticle is a spin up positron. ψ (x) = r ∗ d3 p (2π)3/2 d3 p r 2Ep 2Ep bp up eip·x + cp vp e−ip·x bp vp eip·x + cp up e−ip·x (r)† (r) (r) (r) (r)† (r)∗ (r) (r)∗ = We equate this with † UC ψ(x)UC = r (2π)3/2 d3 p (2π)3/2 2Ep † † [UC bp UC up e−ip·x + UC cp UC vp eip·x (r) (r) (r)† (r) † ∗ Matching coefficients gives (UC ψ(x)UC = ψ ∗ (x)) † UC bp UC = cp (r) (r) † and UC cp UC = bp (r)† (r)† The h. complex conjugation does not mix up the spin ups and spin downs.c.25.

† UC BUC = −B ∗ ∗ ¯ BM ∗ A: ψγ µ ψ −→ −ψγ µ ψ ψσ µν ψ −→ −ψσ µν ψ ψiγ5 ψ −→ ψiγ5 ψ ψγµ γ5 −→ ψγµ γ5 ψ . So we can anticommute A and B without worrying about the anticommutator. January 6 Notes from Sidney Coleman’s Physics 253a 295 Fermionic case |ψ = rs d3 p d3 p Frs (p. Under charge conjugation † UC AUC = A∗ † UC BUC = B ∗ From this † UC A∗ UC = A † UC B ∗ UC = B † UC B † UC = B T (or † UC A† UC = AT same statement as a row vector. either consider the Fermi fields A and B to be classical Fermi fields. or consider the normal ordered product. p ) = ±Frs (p . p) (s)† (r)† UC |ψ = An antisymmetric state of fermion and antifermion is charge conjugation even! Came from (anti-)commutation relations. M is just some matrix in spinor space.25. p )cp bp |0 |ψ if Fsr (p. like iγ5 .) Now A = A† γ 0 so † † UC AUC = UC A† UC γ 0 = AT γ 0 = A†∗ (−γ 0∗ ) = −(A† γ 0 )∗ = −A also. : AM B :. Charge conjugation properties of fermion bilinears Consider AM B.

) So ψψ is charge conjugation invariant. That’s good. January 6 Notes from Sidney Coleman’s Physics 253a 296 † † Exercise In quantum mechanics you frequently use (θ1 θ2 )† = θ2 θ1 . it would be bad to find out that our mass term breaks C.25. The idea of the next step is to write this as the complex conjugate of something. What I have shown is † UC : AM B : UC =: AM ∗ B :∗ =: B M ∗ A look bac when we introduced the bar of a matrix : So all you have to do to calculate the effect of C on our 16 bilinears is to calculate things like 1 =1 ∗ γ µ = −γ µ iγ5 = iγ5 ∗ γ5 γ µ = (γ5 γ µ )∗ = γ5 γµ ∗ σ µν = −σ µν ∗ ∗ in a Majorana basis in a Majorana basis (You can be sloppy and not distinguish between M and M ∗ in a Majorana basis. ∗ . θ† ψ) = (ψ. Is this formula changed if θ1 and θ2 are Fermi? No. We have ∗ − : A M B ∗ : = − : Aα Mαβ Bβ : ∗ = − : Bβ Mαβ Aα :∗ ∗ = + : Aα Mαβ Bβ :∗ remember ∗ means adjoint without transverse and adjoint reverses order Fermi fields anticommute inside normal ordered product ∗ ∗ ∗ This last anticommutation puts things back in the right order to use the conventions of spinor matrix multiplication. † † UC : AM B : UC =: UC AM BUC : This step is allowed because UC does not mix up creation and annihilation operators † † =: UC AUC M UC BUC : = − : A M B∗ : To do the next step. Prove this from the definition of the adjoint (φ. θφ)∗ . I am going to explicitly display the spinor matrix multiplications so I don’t have to worry about keeping matrices and spinors in a given order.

If J µ is charge conjugation odd and if the Lagrangian is to preserve charge conjugation. p ) = ±f (p .1/2) ⊗ D(1/2. but it preserves CP . p) Example of Charge Conjugation in QED The QED Lagrangian contains Aµ J µ where J µ = eψγ µ ψ and Aµ is the photon field. p ) = f (p . then Aµ must be charge conjugation odd and a state with N photons satisfies: UC |N γ = (−1)N |N γ . vector mesons : ψσ µν ψ :−→ − : ψσ µν ψ : 10 odd fermion bilinears from symmetric combinations and 6 even ones from antisymmetric combination D(1/2.0) Bosons: |ψ = UC ψ = ±|ψ Fermions: UC |ψ = ±|ψ if f (p. January 6 Notes from Sidney Coleman’s Physics 253a 297 g1 φψψ + g2 φψγ5 ψ is charge conjugation invariant −e Aµ ψγ µ ψ is charge conjugation invariant only if Aµ is charge conjugation odd.1) ⊕ D(1.1/2) = D(1.1) ⊕ D(0.25.0) + D(0. a Wµ ψγ µ γ 5 ψ + vWµ ψγ µ ψ is parity violating and C violating. p) † † d3 p d3 p f (p. p )bp cp |0 if f (p. vector mesons † UC Aµ UC = −Aµ .

Now what are the possible decay products? Electron and positron are the lightest charged particle antiparticle pairs so the decay must be into n photons. The decay that goes fastest will be the one that goes into the lowest number of photons. assuming they are both allowed. The lowest possibilities are 2 photons c = +1 3 photons c = −1 ortho : para : s = 1 J = 1 C = −1 s = 0 J = 0 C = +1 para −→ 2γ ortho −→ 2γ ortho −→ 3γ allowed not allowed allowed For another explanation along these lines. you get a spin 1 state. When they are put together antisymmetrically. 1 the partial decay rate into n + 1 photons should be down by a factor of e2 ≈ 137 compared to the partial decay rate into n photons. January 6 Notes from Sidney Coleman’s Physics 253a 298 We’ll use this to evaluate the relative decay rates of ortho and para positronium. When two spin 2 are put together in a symmetric combination. you get a spin 0 state. The orbital wave function when 1 l is even is symmetric. one photon is not allowed. but each additional photon comes with a factor of e in the amplitude. where the experimental values are also given. Assuming there are no numerical surprises (without doing some calculations there is no way to rule out factors like (2π)4 in relative amplitudes). the two lowest nearly degenerate hydrogen like bound states of an electron and a positron l=0 (“para” means opposite to) The charge conjugation properties are deduced: a state of one electron and one positron stands a chance of being a charge conjugation eigenstate. see I+Z p. and the crucial Fermi minus sign from anticommuting particle and antiparticle creation operators gives the charge conjugation eigenvalue.154.25. or e2 in the probability. By kinematics alone. Two or more photons are kinematically allowed. These facts. UC UP = UP UC (−1)NF = UP UC U (R(2πe)) Show: UC UP = UP UC U (R(e2π)) U (R(e2π)) is a fancy way of writing the operator (−1)NF (NF = # of fermions) .

M must anticommute with γ µ . January 6 Notes from Sidney Coleman’s Physics 253a 299 Write (−1)NF this way because it reminds you that it is a symmetry of the theory. or at least a basis. call it M . M M must be proportional to the identity −1 and M must be proportional to γ5 = γ5 ] PT : ψ(x) −→ iγ5 ψ(−x) = Ω−1 ψ(x)ΩP T PT .e. Lorentz invariance makes it easier to consider P T than T . −t) = 0 so we expect there to be an antiunitary operator having the effect Ω−1 φ(x)ΩP T = φ(−x) PT In the Dirac theory $ (i − m)ψ(x) = 0$$  − m)φ(−x) = 0 ∂ =⇒(i∂ Now an operation worth calling P T can have a more general form. Then M M commute with every one of the 16 Γ matrices.25. P T : ψ(x) −→ M ψ(−x) where M is some four-by-four matrix in spinor space. PT Just as in a scalar theory. What we need is (i − m)ψ(x) =⇒ (i − m)M ψ(−x) = 0 ∂ ∂ i. Up to a factor this choice is unique. Consider † † † † bp1 · · · bpn dp1 dpn |0 States of this form are a basis and states of this form it is easy to convince yourself the identity is true. We’ll take M = iγ5 . Recall P T in the classical scalar theory (2 + m2 )φ(x) = 0 =⇒ (2 + m2 )φ(−x. [ Proof: Suppose there is a second M anticommuting with the γ µ . One way to show this is to show it is true when acting on an arbitrary state.

Ω−1 A† Ωa) = (b. Ω−1 A† Ωa) ] definition of adjoint of Ω−1 AΩ antiunitary of Ω definition of adjoint of A antiunitarity of Ω−1 Ω−1 ψ † (x)ΩP T = (iγ5 ψ(−x))† PT = −ψ † (−x)iγ5 This tells how ψ = ψ † β transforms −β Ω−1 ψΩP T PT = = −ψ (−x)iγ5 (−β) = ψ(−x)iγ5 So PT : ψ(x)ψ(x) −→ ψ(−x)ψ(−x) Ω−1 ψ † ΩP T PT † β∗ . A† Ωa)∗ = (Ω−1 Ωb.25. January 6 Notes from Sidney Coleman’s Physics 253a 300 Consider applying P T twice Ω−2 ψ(x)Ω2 T = Ω−1 iγ5 ψ(−x)ΩP T P PT PT = iγ5 Ω−1 ψ(−x)ΩP T PT = iγ5 iγ5 ψ(x) = −ψ(x) = U (R(e2π))ψ(x) iγ5 is real so it goes through 2ΩP T 2 γ5 = 1 P T is a rotation half way around the rotation group. Ω−1 AΩb)∗ = (Ωa. ΩΩ−1 AΩb) = (Ωa. This proof is unaffected by giving M an arbitrary phase (Ω2 T = U (R(e2π)) P Using (Ω−1 AΩ)† = Ω−1 A† Ω [ Proof: (b. (Ω−1 AΩ)† a) = (a. AΩb) = (Ωb.

The spin of a particle is affected though. January 8 Notes from Sidney Coleman’s Physics 253a 301 26 January 8 P T (cont’d) (P T commutes with Lorentz transformations) ΩP T on states. (r) . S −→ S −→ −S P T Thus we can’t expect to find a basis where single particle states are left unchanged. Ω−1 PT where bp (1) (2) bp (r) cp (r) ΩP T = bp (r) cp (r) = some phase × bp (2) (1) bp = some other phase × bp (some other phase = . let cp be associated with vp .e.26. (r) (r) is associated with vp . Tricky choice of spinor basis If bp is associated with the solution of the Dirac equation up .some phase∗ ) So that ΩP T applied twice gives − bp (r) cp (r) . let bp (r) with up (r) If cp (r) (r) (r) be associated ≡ (r)∗ −iγ5 up . We expect P T to do nothing to momenta k −→ −k −→ k P T Reflection turns the momenta around but summing the movie backward does it again. I.

If (p − m)u = 0.). Lz u0 (1) (1) (1)∗ (1)∗ iγ5 u0 (1)∗ iγ5 u0 = − 1 u0 2 (1)∗ and Lz − = −1 2 − Because Lorentz transformations are real in a Majorana basis.26. we . Writing out the LHS.c. The action of ΩP T on creation and annihilation operators is derived from the action of ΩP T on the field. ¡ ¡ ¡ 1 Given Lz u0 = + 2 u0 . January 8 Notes from Sidney Coleman’s Physics 253a 302 What’s good about this choice? It makes the action of ΩP T on creation and annihilation operator simple. and the expansion of the field (r) (r)∗ (vp ≡ −iγ5 vp ) So now we’ll see that the definitions of bp Ω−1 PT are consistent with Ω−1 ψ(x)ΩP T = iγ5 ψ(−x) PT which is equivalent to −iγ5 Ω−1 ψ(−x)ΩP T = ψ(x) PT We can write the expansion of ψ(x) two ways ψ(x) = (· · · ) bp up e−ip·x + cp (r)† (r) and cp (r) and (r) bp (r) cp (r) ΩP T = bp (r) cp vp eip·x kinematic factors unimportant to the argument or ψ(x) = (· · · ) bp up e−ip·x + cp (r) (r) (r) † (r) vp eip·x Use the first way in the LHS and the second way in the right. Furthermore it agrees with the expectations of the previous page.). (−p − m)u∗ = 0 and (p − m)(−iγ5 u∗ ) = 0.c. this generalizes to moving states. then (taking c. then (taking c.

· · · . What is the corresponding statement when there are Dirac particles in the theory. We’ll simplify by looking only at 1 fermion + any number of mesons −→ 1 fermion + any other number of mesons u’ u a is of the form a = u M (p1 . −pn ) This says that the amplitude with all incoming particles turned into outgoing antiparticles with the same 3-momentum is the same. the CPT reversed process has an incoming antifermion with the opposite spin characterized by −iu γ5 . January 8 Notes from Sidney Coleman’s Physics 253a 303 have LHS = −iγ5 Ω−1 PT = −iγ5 = (· · · ) bp up eip·x + cp vp e−ip·x ΩP T (r) (r)∗ (r)† (r)∗ (r) (r) (r)† (r) (· · · ) Ω−1 bp ΩP T up e−ip·x + Ω−1 cp ΩP T vp eip·x PT PT (r) (r) (r) † (r) vp (· · · ) bp up e−ip·x + cp eip·x = RHS Proof of P CT within perturbation theory For scalars T CP invariance of the S matrix was equivalent to a(p1 . · · · . pn )u Instead of having an outgoing fermion characterized by u . · · · . Instead of an .26. pn ) = a(−p1 .

[ If LHS(φ) = RHS(φ) for real φ.26. If you want to understand this in two steps u describes an incoming fermion with some spin −→ −iγ5 u∗ PT describes an outgoing fermion of the opposite spin −→ C −iγ5 u describes an outgoing antifermion of the opposite spin (This is a definite choice for the transformed spin. · · · . it is still a polynomial. a choice of another phase would screw up the CPT theorem. However.I. and if both sides are analytic functions of φ in some domain of the complex plane containing a segment of the real line. and using that D(A(eφ)) = eα·eφ/2 We see that both D(Λ) and D(Λ) are analytic. and we can use the equation for complex φ. In this form. · · · . · · · . L. km ) The denominator is Lorentz invariant. pn )u = − (−i)2 u γ5 M (−p1 . · · · . · · · . · · · . they are just exponentials. of the Feynman rules. pn )u = u D(Λ)M (Λp1 . pn . Λpn ). but at any finite order in perturbation theory. Equality of the amplitude for this process and the CPT transformed process is thus u M (p1 . k1 . k1 . The numerator may be an unbelievably complex matrix.) There is also an additional minus sign in the amplitude. Is the RHS an analytic function of φ? D(Λ) contains complex conjugation. −pn )γ5 u switching the operators The proof that these two are equal only uses L. km ) D(p1 . so we’re off to a bad start. D(Λ) = D(Λ)−1 . Λpn )D(Λ)u Consider the case when Λ is a boost in any direction by an angle φ. So the whole RHS is an analytic function of φ.I. January 8 Notes from Sidney Coleman’s Physics 253a 304 incoming fermion characterized by u the CPT reversed process has an outgoing antifermion characterized by −iγ5 u. What about M (Λp1 . Whatever the Feynman rules are. Because an operator has to have an odd number of reordering of Fermi fields to contribute to this CPT reversed process. M is of the form M= d4 k1 · · · d4 kn N (p1 . tells us that u M (p1 . · · · . · · · . pn . · · · . then both sides are equal .

What is D(Λ)? Lz = i 4 3ij γ i j i i γ = (γ 1 γ 2 − γ 2 γ 1 ) = γ 1 γ 2 4 2 D(R(ez π)) = e−iLZ π = e2γ = cos π 1γ2 π π + γ 1 γ 2 sin 2 2 = γ 1γ 2 .26. January 8 Notes from Sidney Coleman’s Physics 253a 305 in that domain. A rotation by π in the z. y plane.] Consider the Lorentz transformation Λ = R(ez π)A(ez iπ) A(ez φ) : p0 p1 p2 p3 −→ p0 cosh φ + p3 sinh φ −→ p1 −→ p2 −→ p3 cosh φ + p3 sinh φ so A(ez iπ) : p0 p1 p2 p3 −→ −p0 −→ p1 −→ p2 −→ −p3 while R(ez π) : −p0 p1 p2 −p3 −→ −p0 −→ −p1 −→ −p2 −→ −p3 so Λ : pµ −→ −pµ . it plane and a rotation by π in the x.

You just compute D(Λ) for the funny Lorentz transformation with complex rapidity. Only analyticity of Feynman amplitude was used in the proof of this theorem. A scalar meson may be forced to be scalar or pseudoscalar. pn )u = u D(Λ)−1 M (Λp1 .26. When we talked about parity invariance we had to hunt for the correct transformation of the field. you don’t have to hunt for the right transformation. This proof easily generalizes to higher spin: You just compute D(Λ) for the higher spin field. This suggests that the theorem has very little to do with perturbation theory. · · · . · · · .) NEXT: renormalization of spinor fields. For CPT invariance. · · · . (The restriction to one incoming and one outgoing fermion was totally inessential. −pn )γ5 u and this exactly the statement of equality between an amplitude and the CPT transformed amplitude. Depending on the interactions that transformation may have to be chosen in various ways. Λpn )D(Λ)u = u γ5 M (−p1 . January 8 Notes from Sidney Coleman’s Physics 253a D(A(ez iπ)) = eαz iπ/2 = ei 2 γ γ π π = cos + iγ 0 γ 3 sin 2 2 = iγ 0 γ 3 π 0 3 306 So D(Λ) = D(R(ez π))D(A(ez iπ)) = γ 1 γ 2 iγ 0 γ 3 = γ 5 (P CT is the analytic continuation of the Lorentz-transformation) Since 2 γ5 = 1 D(Λ)−1 = γ5 also Lorentz invariance says u M (p1 . . It will be a symmetry of the Lagrangian as long as the Lagrangian is Lorentz invariant and hermitian.

We’ll just study 0|ψ(0)|r. Let |r. Anything else can be obtained by a Lorentz transform physical vacuum 0| ψ(x) |p.26. label the Jz = + 1 state by r = 1 and Jz = − 2 by r = 2. s physical nucleon (Matrix elements of ψ are related to matrix elements of ψ by CPT or just by C if the theory had that invariance. 0|ψ(x)|0 = 0 by Lorentz invariance. p (relativistically normalized so as to make L.) 1 For definiteness.T. φ and ψ are not necessarily good fields from the standpoint of the LSZ reduction formula. Not so if the meson nucleon interaction is g0 ψψφ. the various components of this field may need different rescalings. This ensures 0|φ(x)|0 = 0. January 8 Notes from Sidney Coleman’s Physics 253a 307 Renormalization of spinor theories To have a simple example in mind 1 1 L = (∂µ φ)2 − µ2 φ2 + ψ(i − m0 )ψ − λ0 φ4 − g0 ψiγ5 ψφ ∂ 2 2 0 The meson is pseudoscalar. properties simple) be a one fermion state with momentum p and spin labelled by r. g0 and λ0 have no necessary connection with the couplings measured in the standard scattering process. Let 2 u0 ≡ 0|ψ(0)|1. m0 and µ0 have no necessary connection with physical masses. However. So it also only needs rescaling to get a good field for LSZ. p in the rest frame of p. Define Z3 by one meson 1/2 0|φ(0)|k ≡ Z3 φ ≡ Z3 −1/2 φ 0|φ (0)|k = 1 φ is a good field from the standpoint of the LSZ reduction formula (but of course it does not have conventionally normalized equal time commutation relations). p .

p remember we are in the rest frame of p Now this is an assumption38 about the parity transformation properties of a physical nucleon. January 8 Notes from Sidney Coleman’s Physics 253a 308 We can obtain some restrictions on the form of u0 by using Lz conservation u0 ≡ 0|ψ(0)|1. p = e−iLz θ eiθ/2 u0 In the standard basis Lz = the form 1 2 σz 0 0 σz so in the standard basis this restricts u0 to be of   a 0 u0 =   b 0 A less formal way of getting what we have just shown is to that of the four composay x . Turn up the coupling and it may become lighter than the nucleon.p = e−iLz θ eiθ/2 0|ψ(0)|1. ASIDE: Strong coupling scenarios could violate the assumption that the parity transformation property of the physical nucleon are the same as that of the bare nucleon. p |0 e−iLz θ ψ(0) eiθ/2 |1. the transformation properties of the physical nucleon should be the same as those of the bare nucleon for whatever symmetries are not broken by the interaction. To simplify life. Starting with weak coupling. p = 0|e−iJz θ eiJz θ ψ(0)e−iJz θ eiJz θ |1. 38 .26. the s wave bound state will not have the same parity the perturbation theory nucleon did. but in perturbation theory. as you turn up the coupling a nucleon meson bound state may form. 1 nents of ψ. p = |1. If the meson is a pseudoscalar. in the standard basis these two (marked with x)   lower Jz by 2 and the  x . 1 1 other two raise Jz by 2 so only the first two can have a nonzero Jz = 2 to zero matrix element. let’s also assume the theory has parity invariance † Up ψ(0)Up = βψ(0) Up |1. What you had called the nucleon is now unstable.

p 0| βψ(0) |1.p = β 0|ψ(0)|1. p = e−ip·x up (r) which has been arranged to be exactly like the free theory. p =   0  0 For general p. r = 1. 2. so in the standard basis is now restricted to be 0 −1   a 0 u0 =   0 0 1/2 Define Z2 1/2 by a = z2 1/2 √ 2m and ψ by ψ = z2 ψ then √  2m  0   0|ψ (0)|1. p = βu0 In the standard basis β = 1 0 . p † † = 0|Up Up ψ(0)Up Up |1. The Lagrangian you proceed from to do renormalized perturbation theory is 1 µ2 L = (∂µ φ )2 − φ 2 + ψ (i − m)ψ ∂ 2 2 − λφ 4 − gψ iγ5 ψ φ 1 1 + A(∂µ φ )2 − Bφ 2 + Cψ iψ − Dψ ψ ∂ 2 2 − Eψ iγ5 ψ φ − F φ 4 . The LSZ reduction formula goes through as before. January 8 Notes from Sidney Coleman’s Physics 253a 309 This assumption simplifies the possible form of u0 u0 = 0|ψ(0)|1.26. and any x we then have 0|ψ (x)|r.

(24.) The field ψ (x) = aψ(x) − bγ5 ψ(x) is the one satisfying a2 − b 2 0|ψ (x)|r.26. On December 18. 0|T (ψ (x)ψ (y)|0 is a function of x − y alone) Let’s check that the conventions are right by comparing with the free field theory result.1)) ψ(x)ψ(y) = 0|T (ψ(x)ψ(y))|0 i d4 q −iq·(x−y) e = (ix + m) ∂ 4 2 − m2 + i (2π) q 4 d q −iq·(x−y) i(q + m) ¡ = e (2π)4 q 2 − m2 + i d4 q −iq·(x−y) i = e 4 (2π) q−m+i ¡ . we’ll show how C and D are determined order by order in perturbation theory. p = e−ip·x up (r) In parallel with the method for determining A and B order by order in perturbation theory done on November 18 and 20. In the standard rep γ5 = so 1 0   b 0 0|γ5 ψ(0)|1. we calculated (Eq. January 8 Notes from Sidney Coleman’s Physics 253a 310 Digression on spinor renormalization in parity nonconserving theories. Define ¦p¦ ←p ←p = d4 xd4 y eip ·x e−ip·y 0|T (ψ (x)ψ (y)|0 ≡ (2π)4 δ (4) (p − p)S (p) where S (p) is some 4 × 4 matrix function of p (this form is dictated by translational invariance. γ5 commutes with Lorentz transformation. so γ5 ψ(x) transforms in the same way as ψ(x) 0 1 under Lorentz transformations.p =    a at rest 0 (ψ and γ5 ψ have opposite parity transformation properties.

Then because p2 = p2 ¡ S (p) = S (p) ¡ .26. a function of a single complex variable. January 8 Notes from Sidney Coleman’s Physics 253a 311 in free field theory. Putting this in above we have = i d4 q d4 xd4 y e−iq·(x−y) eip ·x e−ip·y 4q−m+i (2π) ¡ d4 q i = (2π)4 δ (4) (p − q)(2π)4 δ (4) (p − q) 4q−m+i (2π) ¡ i = (2π)4 δ (4) (p − p) p−m+i ¡ The conventions are right. this is what we write down upon seeing Lorentz invariance restricts the form of S (p) = = d4 x eip·x 0|T ψ (x)ψ (0) |0 ←p ¦ ←p d4 x eip·x 0|U (Λ) U (Λ)† T ψ (x)ψ (0) U (Λ) U (Λ)† |0 0| |0 = = d4 x eip·x 0|T U (Λ)† ψ (x)U (Λ)U (Λ)† ψ (0)U (Λ) |0 d4 x eip·x D(Λ) 0|T ψ (Λ−1 x)ψ (0) |0 D(Λ) d4 x eip·Λx 0|T ψ (x)ψ (0) |0 D(Λ) d4 x eiΛ −1 p·x = D(Λ) = D(Λ) 0|T ψ (x)ψ (0) |0 D(Λ) = D(Λ)S (Λ−1 p)D(Λ) You can use this and the Lorentz transformation properties of the 16 Γ matrices (which are a complete set of 4 × 4 matrices) to get Ruled out if we assume parity invariance Ruled out if we assume parity invariance S (p) = a(p2 ) 2 + 2 $ b(p2 )γ $$$ 5 + c(p2 )γ µ pµ + $ $ µ$ d(p2 )γ $ $$$ 5 γ pµ + $ $ e(p2$$$µ ν $$ )σµν p p 0 by antisymmetry = a(p ) + c(p )p ¡ Define a new function S (z) = a(z 2 ) + zc(z 2 ).

26. a function of one variable! ¡ (There is a one-to-one correspondence between functions of one variable and functions of 1 matrix. A function of two matrices is far more complicated than a function of two numbers unless the two matrices commute) . January 8 Notes from Sidney Coleman’s Physics 253a 312 The propagator is characterized by a single function of p.

¡ ¦p¦ ¦ ¦p¦ ¦p¦p¦ ¦ ¦p¦ 1P I 1P I 1P I = +   + + ···   = 1 1P I 1− Mathematically.e. if a term in the Lagrangian is −δmψψ There is a contribution to −iΣ (p) of ¡ ¦x¦ ←p − iδm i. S (p) = ¡ i i i + (−iΣ (p)) ¡ p−m+i p−m+i p−m+i ¡ ¡ ¡ i i + (−iΣ (p)) ¡ p − m + i (−iΣ (p)) + · · · ¡ p−m+i ¡ ¡ which sums to S (p) = ¡ = i p−m+i ¡ 1+ Σ (p) ¡ + p−m+i ¡ Σ (p) ¡ p−m+i ¡ 2 + ··· 1 i p) p − m + i 1 − Σ (¡ ¡ p−m+i ¡ i = p − m − Σ (p) + i ¡ ¡ .27. i. to Σ (p) of δm ¡ The nice thing about the 1P I function is that it gives us an expression for the full Green’s function (without the (2π)4 δ (4) (p − p)).e. January 13 Notes from Sidney Coleman’s Physics 253a 313 27 January 13 So far we have found ¦p¦ ←p ←p = (2π)4 δ (4) (p − p)S (p) ¡ Now we define a one particle irreducible Green’s function (defined to not include (2π)4 δ (4) (p − p) or external propagators) 1P I = −iΣ (p) ¡ ¦p¦ ←p For example. it gives us S (p).

One positron does not contribute because one positron|ψ |0 = 0 (fermion # conservation). January 13 Notes from Sidney Coleman’s Physics 253a 314 (Remember p is the only matrix in the problem so it commutes with “every other matrix” ¡ and manipulations in which p is treated like a number are correct.1). ¡ 0|ψ (x)|n n|ψ (y)|0 = complete set |n of momentum eigenstates |n e−iPn ·(x−y) 0|ψ (0)|n n|ψ (0)|0 Now we break the |n up into physical vacuum. The parity − state only give nonzero components to the lower two components of ψ. This simplification does ¡ not persist in the spin 3 problem. 2 The parity + states only give nonzero contributions to the upper two components 0|ψ (0)|n (in a standard basis which is easiest for states at rest to work with).27. we use the renormalization conditions to eliminate the physical vacuum contribution and to simplify the one electron contribution. (16. like a nucleon and meson in 2 − a p wave. r q. (Eq. Furthermore the contribution 1 of a Jz = + 2 state to the top component is the same as a contribution to the second component of the same state hit with the lowering operator Jx − iJy .) 2 To get a spectral representation for S (p) we insert a complete set into 0|ψ (x)ψ (y)|0 . physical one electron. ¡ 2×2 blocks . r|ψ (0)|0 (2π)3 2ωq e−iPn ·(x−y) 0|ψ (0)|n n|ψ (0)|0 all other states |n one electron 0|ψ (x)ψ (y)|0 = r + = r d3 q (r) (r) e−iq·(x−y) uq uq (2π)3 2ωq d4 p −ip·(x−y) e (2π)4 δ (4) (p − Pn ) 0|ψ (0)|n n|ψ (0)|0 + all other |n 1 The sum on |n only contributes when the state |n has spin 2 in its rest frame. As on November 18. like a nucleon and meson in an s wave. In a + parity invariant theory. one positron and all other states. a matrix which reduces to  1 0 2 p2  0 0 (in the standard basis) when p is at rest and is covariant in p + p2 . d3 q e−iq·(x−y) 0|ψ (0) |q. and J p = 1 . we can split these states into J p = 1 .

a2 ) ∂ da [ρ+ (a)(ix + a)∆(x − y) + ρ− (a)(ix − a)∆(x − y)] ∂ ∂ = 0 For compactness in the last step I have introduced ρ+ (a) = σ+ (a) + δ(a − m) Can shortcut some work by noticing that ψ = γ5 ψ has the same matrix elements with a − − 1+ ψ has with a 1 state. January 13 Notes from Sidney Coleman’s Physics 253a 315 so δ (4) (p − Pn ) 0|ψ (0)|n n|ψ (0)|0 = all other |n + with J p = 1 2 θ(p0 ) σ+ ( (2π)3 p2 )(p + ¡ p2 ) Similarly 39 δ (4) (p − pn ) 0|ψ (0)|n n|ψ (0)|0 = all other |n 1− with J p = 2 θ(p0 ) σ− ( p2 )(p − m) ¡ (2π)3 σ+ and σ− .” 2 2 39 1+ 2 state as . we have 0|ψ (x)ψ (y)|0 = d3 q e−iq·(x−y) (q + m) ¡ (2π)3 2ωq d4 p −ip·(x−y) + e θ(p0 )σ+ ( p2 )(p + p2 ) + θ(p0 )σ− ( ¡ (2π)3 = (ix + m)∆(x − y) ∂ ∞ d4 p θ(p0 )δ(p2 − a2 )(p + a)e−ip·(x−y) + da σ+ (a) ¡ (2π)3 0 ∞ d4 p θ(p0 )δ(p2 − a2 )(p − a)e−ip·(x−y) + da σ− (a) ¡ (2π)3 0 ∞ p2 )(p − ¡ p2 ) = (ix + m)∆(x − y) + ∂ ∞ 0 da σ+ (a)(ix + a)∆(x − y. σ+ = σ− = 0. the 1 states are to γ5 ψ. when p2 < (m + µ)2 . In perturbation theory.27. but it should be because 0|ψ (0)|n n|ψ (0)|0 differs from 0|ψ (0)|n n|ψ † (0)|0 by the matrix γ 0 which is negative in its lower two components. You might worry that p − m is negative in the rest ¡ frame of p. “As the 2 states are to ψ. Putting this together. a2 ) ∂ + ∞ 0 da σ− (a)(ix − a)∆(x − y. which are defined by these equations are both positive semidefinite by the positivity of the norm on Hilbert space.

∞ 0|T (ψ (x)ψ (y))|0 = θ(x0 −y 0 ) ∞ 0 da [ρ+ (a)(ix + a)∆(x − y) + ρ− (a)(ix − a)∆(x − y)] ∂ ∂ ∗T +θ(y 0 −x0 ) 0 da ρ+ (a)γ5 γ 0 (ix + a)∆(x − y)γ 0 γ5 + ρ− (a)γ5 γ 0 (ix − a)∆(x − y)γ 0 γ5 ∂ ∂ (iγ5 γ 0 γ µ γ 0 γ5 )∗T = iγ µ Now and (γ5 γ 0 1γ 0 γ5 )∗T = 1 so (in this case we can pull the time derivative through the time ordered product) ∞ 0|T (ψ (x)ψ (y)|0 = 0 da (ρ+ (a)(ix + a) + ρ− (a)(ix − a)) ∂ ∂ acting on θ(x0 − y 0 )∆(x − y) + θ(y 0 − x0 )∆(y − x) (Using that ρ+ (a) and ρ− (a) are real and that ∆(x − y)∗ = ∆(y − x)) . but I’ll be more general. I can get 0|ψ β (y)ψα (x)|0 from what we’ve just calculated using ΩCP T (= Ω). we actually need this thing transposed. (In a theory with C invariance.27. a2 in ∆ 316 ρ− (a) = σ− (a) Rather than redoing a lot of steps. January 13 Notes from Sidney Coleman’s Physics 253a and dropped the . α come out in the wrong order. so to think of 0|T (ψ (x)ψ (y))|0 as a matrix.) Then we’ll be set to write down the time ordered product 0|T (ψα (x)ψ β (y))|0 =θ(x0 − y 0 ) 0|ψα (x)ψ β (y)|0 − θ(y 0 − x0 ) 0|ψ β (y)ψα (x)|0 I’ll do the calculation in a Majorana basis 0|ψ β (y)ψα (x)|0 = 0|ΩΩ−1 ψ β (y)ΩΩ−1 ψα (x)ΩΩ−1 |0 = ( 0| Ω−1 ψ β (y)Ω Ω−1 ψα (x)Ω |0 )∗ −i(γ5 γ 0 ψ (−y))β −i(ψ (−x)γ 0 γ5 )α In this step when 0|Ω is simplified to 0| the resulting matrix element must be complex conjugation because Ω is antiunitary = −i(γ5 γ 0 )βσ 0|ψσ (−y)ψ τ (−x)|0 (γ 0 γ5 )τ α Notice the indices β. it would be easier to just use Uc .

So the result is ∂ ¡ d4 p −ip·(x−y) ∞ e da 0|T (ψ (x)ψ (y))|0 = (2π)4 0 Or in a more suggestive form 0|T (ψ (x)ψ (y))|0 = = where d4 p −ip·(x−y) ∞ da e (2π)4 0 d4 p −ip·(x−y) e S (p) ¡ (2π)4 ∞ ρ+ (a) p2 i(p − a) i(p + a) ¡ + ρ− (a) 2 ¡ 2 2+i −a p −a +i ρ+ (a) i i + ρ− (a) p−a+i p+a+i ¡ ¡ i i + ρ− (a) z−a+i z+a+i 0 This result for S (z) has the renormalization conditions built in. January 13 Notes from Sidney Coleman’s Physics 253a 317 The object in brackets is d4 p i e−ip·(x−y) 4 p 2 − a2 + i (2π) ix hitting this gives p.27. we see the renormalization conditions are Σ (m) = 0 dΣ =0 dz z=m pole is at m residue is i (often written dΣ dp ¡ = 0) p=m ¡ . They say S has a pole at z = m with residue i. S (z) = da ρ+ (a) z -(m+u) m m+u Figure 1: Analytic structure of S in perturbation theory Compare this with our other expression for S S (z) = i z − m − Σ (z) + i In terms of Σ .

D : ∞ power series in C (n) ∝ g n Get Σ (p) to order g 2 ¡ ¦do#d¦ + ¦x¦ (2) −iΣ (p) = −iΣ+ (p) + iC (2) p − iD(2) ¡ ¡ ¡ dΣ+ Σ (p) = Σ+ (p) − Σ+ (m) − (p − m) ¡ ¡ dp m ¡ ¡ Only knocks off 1 power of p ¦do#d¦ k→ p p L = −gψiγ5 ψφ ←− a 4 × 4 matrix. p+k −iΣ+ = −(−ig)2 −ig 2 (2π)4 −ig 2 = (2π)4 = −ig 2 (2π)4 i p +− m d4 k ¡ k iγ5 i iγ5 4 k 2 − µ2 + i (2π) (p + k)2 − m2 + i 1 −p −  + m k ¡ d4 k 2 2 + i (p2 + 2kp + k 2 − m2 + i ) k −µ 1 p −− m k ¡ d4 k dx 2 [k + 2kpx + p2 x − m2 x − µ2 (1 − x) + i ]2 0 ODD 1 k = k + px = d4 k 0 dx [k + 2 p2 x(1 −p(1 − x) + m − k ¡ − x) − m2 x − µ2 (1 − x) + i ]2 .27. this is a propagator without external propagators included. January 13 Notes from Sidney Coleman’s Physics 253a 318 In the model with C and L = −gψiγ5 ψφ + Cψiψ − Dψψ ∂ C.

unlike scalar case where diagram was only log divergent. January 13 Notes from Sidney Coleman’s Physics 253a 319 −ig 2 Σ = (2π)4 1 dk 0 4 dx [k + m2 x(1 2 p2 x(1 −p(1 − x) + m ¡ − x) − m2 x − µ2 (1 − x) + i ]2 − − DIVERGENT PART [k + 2 −m(1 − x) + m − x) − m2 x − µ2 (1 − x) + i ]2 −m(1 − x) + m 2mx(1 − x) + m(1 − x) + m +2 2 [ ] [ ]2 ∝ =0 d4 k [−p(1 − x) + m + m(1 − x) − m − (p − m)(1 − x)] ¡ ¡ k4 d2 Σ . Σ is completely deterdp2 ¡ A quicker way of seeing if the result is finite is to compute mined by this second derivative. Two does the job =⇒ Need two subtractions to remove ∞’s. Note that one derivative is not enough to give a finite integral. p) some awful 4×4 matrix q=p −p hf ¥ + + hxf ¥ (3) You might give your renormalization condition as Γ = igγ5 . Coupling constant renormalization in spinor theory (parallels scalar case) p Contributions up to order g 3 are h pf ¥ 1P I p = −i Γ (p .27.

The two spin 1 ’s in the final state can make S = 1 or S = 0.27. Consider this graph as contributing to φ(off shell) −→ N + N and look at the process in the COM frame where q = (q 0 . p)(p + m)|p2 =p 2 =m2 ¡ ¡ p+m must be ∝ to (p + m)γ5 (p + m). We’ll show (p + m)Γ(p . (It may vary with q 0 ) (p + m)γ5 (p + m) ¡ ¡ is nonzero when sandwiched between a u and a v (remember p0 < 0 for this process with ¯ the momentum conserving conditions So We’ll take G(q 2 = µ2 ) ≡ g p ¤pf i p p2 =p 2 =m2 ) (p + m)Γ(p. However. p 2 )(p + m) ¡ ¡ = (p + m)iγ5 (p + m)G(q 2 ) ¡ ¡ . ¡2m projects onto incoming nucleons or outgoing antin¡ ¡ ucleon. Γ may not be proportional to γ5 It may have pγ5 or pµ pν σ µν γ5 . To get J = 0 the only 2 possible final states are l=0 l=1 S=0 S=1 which has which has P = −1 P = +1 Only the first final state is allowed. There is only one amplitude. January 13 Notes from Sidney Coleman’s Physics 253a 320 at some astutely chosen value of p and p. We can remedy this by cutting down the 4 × 4 matrix by ¡ sandwiching it between projectors. 0) The initial state is J p = 0− .

Utility of this choice q' q q q p'u' p u = p 1PI 1PI p + + graphs with no pole at (p + q)2 = s = µ2 Pole piece in ia − u Γ(p .27. January 13 Notes from Sidney Coleman’s Physics 253a 321 as our renormalization conditions. p)u p +m = −u Γ(p . + hxf ¥ (3) . we have Γ sandwiched between projection operators and we can use the renormalization condition to get that the pole piece in ia is −u iγ5 g i(p + q + m) ¡ ¡ iγ5 gu (p + q)2 − m2 This simplification allows for unambiguous comparison with experiment to set g. Is renormalization necessary and sufficient to get rid of ∞’s ? Let’s look at the contributions to Γ at O(g 3 ). p) ¡ u 2m insert £ £ +analytic So near s = µ2 . p + q)S (p + q)Γ(p + q. p + q) 2m insert it is the identity on u S (p + q) near s = µ2 this is i(p+q+m) (p+q)2 −m2 +i p+m Γ(p + q.

27. and discovering they always end up in convergent combinations isn’t good enough. The infinities came because the theory has an infinite # of degrees of freedom. the integral for the first Feynman graph goes like d4 k 1 1 1 γ5 γ5 γ5 ∼ γ5 k2   k k d4 k 1 k4 This is divergent. . January 13 Notes from Sidney Coleman’s Physics 253a 322 at high k. both from the ∞ extent of spacetime (which gives IR ∞’s) and from the fact that in any given volume there is an ∞ # of degrees of freedom (which gives UV ∞’s). Regularization and renormalization Throwing around ill-defined quantities. So far our slovenliness has been good enough. but only logarithmically divergent. Next lecture we’ll talk about ways to cut down the # degree of freedom in a given volume. So the second graph cancels the divergent part. and it multiplies γ5 .

Regulator fields (Feynman) B. January 15 Notes from Sidney Coleman’s Physics 253a 323 28 January 15 This lecture I. In all theories worth studying. Regularization A. so as not to be making ad hoc cancellations of infinities with infinities.) . Regulator fields or Propagator Modification A good regularization method should (1) be analytically tractable (2) ruin as few properties of the theory as possible. The big job is to prove that the properties you expect of the theory (Lorentz invariance. Instead of making subtractions of ∞’s from ∞’s to satisfy the renormalization conditions. Then you would renormalize as usual. For example. Then you try to undo your hatchet job by sending Λ to ∞. you have to hack up the theory in some way to make it finite. positivity of the Hilbert space inner product) are recovered as Λ −→ ∞. you will be subtracting finite (but big. you could throw away all Fourier components in the Feynman integrals with momentum greater than some cutoff value Λ. BPHZ renormalization I. (The less you ruin the less you have to laboriously prove you recover in the Λ −→ ∞ limit. and that nothing depends on Λ in this limit.28. Regularization No one knows of a quantum field theory that is nontrivial and finite. Λ2 or ln Λ) things from other finite things to satisfy the renormalization conditions. Dimensional Regularization (’t Hooft-Veltman) II. A scattering amplitude should not depend on the method some theorist used to make an infinity large but finite. proportional to Λ. gauge invariance.

1) k2 M plays the role of the cutoff. Here’s k 1 how to make them go like 2n for as big as you need. January 15 Notes from Sidney Coleman’s Physics 253a 324 Regulator fields (or at least a variant we’ll discuss March 3 called Pauli-Villars) only wreck positivity of the Hilbert space metric in QED with massive charged particles. To do this we’ll let k2 i − m2 become a combination of propagator. The idea is to replace propagators in the Feynman integrals by propagators that fall off faster at high momentum so that loop integrals will be finite. Let k i i −→ 2 + 2 − m2 k k − m2 n 2 iCr 2 k 2 − Mr r=1 2 2 (I write the coefficient as Cr . but don’t let me mislead you into thinking Cr > 0) We can look at the behavior of this for high k 2 by expanding 1 1 1 = 2 k 2 − m2 k 1 − m22 k 1 m2 m2 = 2 1+ 2 + k k k2 2 + ··· . For k 2 1 k4 Similarly instead of i i i 1 −→ − ∝ 2 p−m p−m p−M p ¡ ¡ ¡ After modifying the propagators enough to make the diagrams convergent.28. and the only kinds of integrals that have to be evaluated are of the same type we have already studied. you adjust the counterterms to satisfy the renormalization conditions. and then send M −→ ∞ 1 Making a propagator go like 4 may not be enough to make diagrams convergent. For example i i i −→ 2 − 2 2 2 −m k −m k − M2 M 2 this combination falls off like 1 k2 at high p (28.

(28.) Of course 2 some of the Cr are going to have to be less than zero.1)..28. Suppose the original theory had 1 µ2 L = (∂µ φ)2 − φ2 + L (φ ) 2 2 The unrealistic theory that has these propagators is 1 m2 2 L = (∂µ φ)2 − φ 2 2 n 1 M2 + (∂µ φr )2 − r φ2 2 2 r r=1 n + L (Φ) Φ= r=1 Cr φr I talk about why this gives the right propagator combination on March 3. m2 + r=1 n 2 2 Mr Cr = 0 m4 + r=1 n 4 2 Mr Cr = 0 m6 + r=1 6 2 Mr Cr = 0 (n = 1. and still have freedom to send all the Mr −→ ∞. There is no way this can happen in any realistic 2 theory of the world. (The Cr must remain finite. a e We can construct an operator theory that is unrealistic that has these sicko propagators though. Cr > 0 is a consequence of the Lehmann-K¨ll´n spectral representation. or you are just going to have i times a sum of things with the same sign at large k 2 . you can clearly make the propagator fall off as fast as you like.. January 15 Notes from Sidney Coleman’s Physics 253a 325 By choosing n 1+ r=1 n 2 Cr = 0 makes propagator makes propagator makes propagator makes propagator ∼ ∼ ∼ ∼ 1 k4 1 k6 1 k8 1 k 10 and and and etc.) By making n large enough. . M1 = M creates the simplest example in Eq. c1 = i.

and the new probability is conserved. 2 Because some of the Cr are less than zero. (φ1 )† = −φ1 new because φ1 anticommutes with (−1)N1 . . we have a new definition of probability. m2 2 1 M2 2 1 φ + (∂µ φ1 )2 − φ + L (Φ) L = (∂µ φ)2 − 2 2 2 2 1 define a new inner product a|b new Φ = φ + iφ1 = a|(−1)N1 |b N1 counts the number of mesons of the sicko type. and the Hamiltonian is not Hermitian. The only initial and final states possible with thus be the ones with sensible particles in them. Here is why you might hope that a sensible theory will be recovered when the M −→ ∞ limit is taken. and when M −→ ∞. the inner product is normal. just for lack of energy. In the theory that embodies the simplest propagator modification. some of the Cr are imaginary. In the old metric. We can gain some insight into what is going on by defining a new inner product. the Hamiltonian wasn’t hermitian and thus didn’t conserve probability. In the new metric. it will be impossible to produce them in the final state.28. Answer: It helps you ascertain what properties of the theory you have or haven’t ruined. a|a new <0 if |a has an odd number of φ1 mesons in it The great thing about this metric is that in it Φ is hermitian. January 15 Notes from Sidney Coleman’s Physics 253a 326 About this point you may be wondering why we are trying to construct a Lagrangian that reproduces our hatchet job. We won’t be interested in amplitudes that have those phony particles in the initial states. which was positive definite. and for them. So (Φ)† = (φ + iφ1 )† = Φ new new To summarize. and although it is not always greater than zero. the Hamiltonian is hermitian. This metric is not positive definite.

You just work along with each propagator separately. internal symmetries in theories with massive particles (they spoil symmetries that depend on masslessness). Dimensional Regularization Begin with an example.28. Start with ∞ Γ(n) = 0 tn−1 e−t dt (which is sometimes taken as the definition of the Γ function). with some modification. α real > 0. and use the integral tables that are valid when you work with a convergent combination. Here is a trick to turn a denominator into an exponential. and they are computationally easy to introduce. Let’s evaluate I= dd k (k 2 + a2 )n which is convergent if n > d 2 in an arbitrary number of Euclidean space dimensions d. A note on computation: In practice you don’t try to combine the various propagators to make the integrals manifestly convergent. Change variables in the integrand to λ given by αλ = t. conserve probability at energies low compared to the cutoff. will be seen to preserve gauge invariance in QED. ∞ Γ(n) = 0 (αλ)n−1 e−αλ d(αλ) ∞ or 1 1 = n α Γ(n) λn−1 e−αλ dλ 0 . January 15 Notes from Sidney Coleman’s Physics 253a 327 The good things about regulator fields are that they preserve Lorentz invariance.

. It has to acquire dimension. It cannot remain constant as we change d. If you stay away from even integer d ≥ 2n. You might think that after renormalizing. As you head toward d = 4. 1 e = in an arbitrary number of dimensions because only in You can’t just maintain 4π 137 four dimensions is e dimensionless. January 15 Notes from Sidney Coleman’s Physics 253a 328 Our (Euclidean space) integral becomes I= 1 Γ(n) ∞ λn−1 dλ 0 dd k e−λ(k 2 +a2 ) e−λa2 ( π )d/2 λ π d/2 ∞ n− d −1 −λa2 = λ 2 e dλ Γ(n) 0 π d/2 Γ(n − d ) 2 = Γ(n) a2n−d Here is ’t Hooft and Veltman’s whammy: adopt this formula for arbitrary complex d.28. we must have d = [λ] + 4[φ ] [λ] = d − 2(d − 2) = −d + 4 Only in four dimensions is λ dimensionless. 2 But to keep the Lagrangian having dimension d. you approach poles in the Γ function. So we rewrite the interaction: λ µ4−d φ4 4! Now dimensionless for any d where µ is a parameter that has appeared uninvited into the theory. when we set d = 4. You do your renormalization in arbitrary d and only after you have your expressions for the graphs plus counterterms in convergent combinations (that is with poles in d − 4 cancelling) you send d −→ 4.t. this expression is well defined. There are simpler examples than QED to demonstrate the effect of this. L = (∂µ φ )2 − 2 2 [φ ] = d−2 . You have to be careful formulating a theory in an arbitrary # of dimensions. Take m2 2 1 φ − λφ 4 + Lc. all µ dependence will go away. so m is a mass as it appears.

and I have suppressed the Feynman parameter integral and a lot of factors. d Let’s see what the finite part is. Our result for this integral is (λµ4−d )2 d d−4 π d/2 Γ 2− a Γ(2) 2 Γ has a pole piece near d = 4. That would give you the right coefficient of the pole. but the wrong finite part. 1981)] p. 2 We’ll pull out the a factor of µ4−d since that is the dimension of this Green’s function (the lowest order contribution is proportional to λµ4−d ). For integer n Γ(−n + ) = (−1)n 1 + ψ(n + 1) +O( ) n! some number like π except more complicated see Ramond. [Field Theory: A Modern Primer (1st edition.28. 152. external momenta and Feynman parameters. An O(λ2 ) diagram It leads to an integral like (λµ 4−d 2 ) dd k 1 d (k 2 + a2 )2 (2π) where a contains masses. You can’t just set d = 4 in the rest of the expression. January 15 Notes from Sidney Coleman’s Physics 253a 329 We’ll look at a contribution to the four point function. We expand the dimensionless thing . In our example n = 0 and = 2 − .

I’ll do it in our example. that is. It amounts to just chucking the pole terms in the dimensionally regularized integrals. µ4−d λ2 π 2 1 φ4 2 − d 4! 2 (up to is and minus signs) . πa2 Now you can renormalize as usual although you need an extension of the renormalization conditions for arbitrary dimension. January 15 Notes from Sidney Coleman’s Physics 253a 330 that is left λ2 µ4−d π d/2 d d−4 d µ 4−d d −2 Γ 2− a = λ2 π 2 Γ 2 − π2 1! 2 2 a d µ2 4−d 2 = λ2 π 2 Γ 2 − 2 2 πa 0 1 (−1) + ψ(1) + O(d − 4) = λ2 π 2 0! 2− d 2 1 2− µ2 + O(d − 4) πa2 e 4−d 2 ln µ2 πa2 µ2 πa2 1+ 4−d ln 2 = λ2 π 2 d 2 + ψ(1) + ln You would have lost the ln µ2 piece of you prematurely set d = 4. Theorists like it because they no longer make comparison with experiment and the minimal subtraction renormalization prescription is easy. It makes no reference to the physical mass and coupling so it is not good for comparison with experiment. we have found = µ4−d λ2 π 2 1 2− d 2 + finite as d → 4 The coefficient of the pole is unambiguous.28. a way of determining counterterms. It is a companion to dimensional regularization. Minimal subtraction says introduce a counterterm to exactly cancel it. Again suppressing the Feynman parameter integral and whatever else. In this example we need a term in Lc.t. Minimal Subtraction (or MS) MS is another renormalization prescription.

Keck41 =c=1 [M L] = 1 [S] = [ ] = 1 [L] = M 4 [φ] = M [ψ] = M 3/2 [∂µ ] = M dim L (in mass units) = bi + 3 fi + di = δi + 4 2 Given a set of bosons + fermions the most general interaction of renormalizable type defines a renormalizable theory. 1) [BGC: Annotated photocopies of Coleman’s lecture from Aspects of Symmetry were attached here] [BGC: included in the notes was a photocopy of the discussion from the original publication of the lecture “Renormalization and symmetry” in Properties of the Fundamental Interactions (Editrice Compositori. Renormalization and symmetry: a review for non-specialists (1971)40 Discussion Chairman: Prof. 2. The same is true if we restrict the theory to be invariant under parity and internal symmetry. It also makes no reference to physical masses and coupling either. Coleman Scientific Secretary: B. It’s called BPH. January 15 Notes from Sidney Coleman’s Physics 253a 331 In what follows.W. 1973)] 41 40 . another renormalization prescription is heavily used. It is useful for proving that renormalization removes the ∞’s . S.28. The same is true if we allow symmetry breaking interaction if we allow all sym-breaking interaction with dim ≤ n (n = 3. Bologna.

28. u . As equation (9) 2 in “Renormalization and Symmetry” suggests we add 1 J (x)φ2 (x) 2 to L. This gives us a new Feynman rule p+k k ⇐⇒ iJ (p) To O(g 2 ).u = iJ (l − l )(−ig)2 = −J (l − l )g 2 = −2g 2 J (l − l ) 1 1−x i d4 k i i u u k (2π)4  − m (k − l )2 − µ2 (k − l)2 − µ2 4 1 1 k d k u ( + m)u 4 k 2 − m2 (k − l )2 − µ2 (k − l)2 − µ2 (2π) d4 k u ( + m)u k 4 {(1 − x − y)(k 2 − m2 ) + x[(k − l )2 − µ2 ] + y[(k − l)2 − µ2 ]}3 (2π) dx 0 2 0 dy = −2g J (l − l ) 1 1−x dx 0 0 d4 k u ( + m)u k (2π)4 [k 2 − (1 − x − y)m2 − 2xk · l − 2yk · l − (x + y)µ2 + xl 2 + yl 2 ]3 (k = k − xl − yl) dy d4 k u ( + xl + y l − m)u k £ £ (2π)4 [k 2 + x(1 − x)l 2 + y(1 − y)l2 − 2xyl · l − (1 − x − y)m2 − (x + y)µ2 ]3 = −2g 2 J (l − l ) 1 1−x dx 0 0 dy . u | 1 φ2 (x)|l. January 15 Notes from Sidney Coleman’s Physics 253a 332 As an example of a matrix element of a composite operator. is calculated by evaluating 2 k-l' l'u' k k-l l. l . let’s calculate the matrix element of 1 φ2 between single nucleon states in our meson nucleon theory.

Redefinitions of the composite operator are necessary and additional renormalization conditions to make these redefinitions definite are needed. . u | 2 φ2 (x)|l. the matrix elements of composite operators will not necessarily be finite to that order. We have (dropping prime on k) £ 1 1−x −2g 2 J (l − l ) 0 dx 0 dy d4 k (x + y − 1)mu u (2π)4 [k 2 − M 2 (x. Also l2 = l 2 = m2 . u | φ2 (x)|l. u from this we have to write J (l − l ) in terms of J (x) divide by δ i and then take δJ (x) δ g2 1 d4 x e−i(l−l )·x J (x)mu uF l . January 15 Notes from Sidney Coleman’s Physics 253a 333 The  term in the numerator is seen to be odd. µ2 . y) = [−x(1 − x) − y(1 − y) + 1 − x − y]m2 + (x + y)µ2 + 2xyl · l The k integration is in our tables. (l − l )2 ) 16π 2 Renormalization of composite operators Unfortunately even in a theory that was finite to some order in perturbation theory.28. (l − l )2 ) 2 16π 1 To get l . y) Let’s just call the result of the Feynman parameter integrations F (m2 . lu = mu and k £ u l = mu are simplifications. y)]3 where M 2 (x. (l − l )2 ) so what we have is g2 iJ (l − l )mu uF (m2 . µ2 . u = 2 δJ (x) 16π 2 g 2 −i(l−l )·x = e mu uF (m2 . 2ig 2 J (l − l )mu u 1 32π 2 1 1−x dx 0 0 dy 1 M 2 (x. µ2 .

q1 .2) For n = 2 the contributions to O(λ) are 42 L −→ L + J(x) O(x) − O all ops with right sym prop of lower dim . qn ) = = qn q3 q3 d4 x d4 y1 · · · d4 yn e−ip·x e−i(q1 ·y1 +···+qn ·yn ) 0|T (φ2 (x)φ (y1 ) · · · φ (yn ))|0 R (28. For example we will see that at order λ in a theory with a φ4 interaction it is necessary to add to L in addition to 1 J (x) φ2 (x) 2 further terms A J (x)( φ2 (x) + B) 2 1 2 1 φ (x)(1 + A) + B ≡ φ2 2 2 R We’ll have cutoff independent matrix elements in the limit of large cutoff. to order λ. let’s calculate 2 R 0|T (φ2 (x)φ (y1 )φ (y1 )φ (y2 ) · · · φ (yn ))|0 R at least for n = 0 and n = 2. January 15 Notes from Sidney Coleman’s Physics 253a 334 In the method of getting Feynman rules for composite operators by adding a source coupled to the operator to L the redefinitions come as further additions multiplied by the same source. We’ll just calculate the Fourier transform p The total coefficient of J (x) G(p. · · · . The finite parts of A and B will be determined by renormalization conditions42 Rather than calculate a matrix element of 1 φ2 .28.

q1 .28. January 15 p q 1 Notes from Sidney Coleman’s Physics 253a p + q2 q 1 p q2 + q 1 q2 + q 1 q2 335 p + more graphs that aren't connected graphs I’ll discuss these contribution in reverse order. · · · . The disconnected graphs will be disposed of by only computing the connected part of G. It is (−iλ) i i (2π)4 δ (4) (q1 + q2 − p) · 2 2 q 1 − µ2 q2 − µ2 2 d4 k i i p 2 p 2 (2π)4 (k − 2 ) − µ2 (k + 2 ) − µ2 This integral is logarithmically divergent which is why we need the O(λ) graph coming from the O(λ) part of Aφ2 in φ2 . qn ) The second and fourth graphs will be exactly cancelled by the O(λ) mass renormalization counterterm graphs (1) q 1 p q2 and q 1 p (1) q2 The third graph is moderately interesting. denoted Gc (p. I’ll denote that A(1) φ2 and we get one more graph R p q 1 This graph and the first graph give (1 + A(1) ) 2 q1 (1) q2 i i (2π)4 δ (4) (q1 + q2 − p) 2 − µ2 q2 − µ2 .

Apparently the counterterms for a conserved current can usually be pinned down by calling upon cherished properties such as T µν = −T νµ and ∂µ T µν = 0 B (1) could be chosen so that A stupid example to make sure I have my F. A logical one is that Gc at zero momentum be given exactly by its lowest order contribution. It is sufficient to do that. A superficially log divergent graph only needs one subtraction (in any order of perturbation theory) according to BPHZ. conventions right. q2 ) =(2π)4 δ (4) (p + q1 + q2 ) · 1− iλ 2 2 q1 (−iλ) 2 d4 k 1 =0 4 (k 2 − µ2 )2 (2π) i i 2 q 2 − µ2 −µ 2 i d4 k i i − 2 4 (k − p )2 − µ2 (k + p )2 − µ2 (2π) (k − µ2 )2 2 2 1 0| φ2 (x)|0 = 0 2 R It is surprising to me that there is so much arbitrariness in the definition of Green’s function with a composite operator that has to be fixed by renormalization conditions.28. Let’s compute 0|T (ψ(x)ψ(y))|0 in free Dirac theory . We need a renormalization condition to determine the finite part of A. January 15 Notes from Sidney Coleman’s Physics 253a 336 A(1) is a divergent constant chosen to cancel the logarithmically divergent part of the second graph. This means A(1) + So finally to O(λ) G(c) (p. If someone hands you T µν = ∂ µ φ∂ ν φ − g µν L (the energy-momentum tensor obtained through Noether’s theorem). q1 .T. you’d think something like 0|T µν |0 being such a physical thing would not be susceptible to redefinition.

(24. (28.2) this should be d4 q1 d4 q2 iq1 ·x+iq2 ·y q q2 1 e 4 (2π)4 (2π) 4 d q1 d4 q2 −iq1 ·x+iq2 ·y i e (2π)4 δ (4) (q1 + q2 ) · 4 (2π)4 (2π) q2 − m + i ¡ 4 d q −iq·(x−y) 1 e · (2π)4 q−m+i ¡ = = In agreement with Dec. Eq. 18.1) and following “3 comments”. .28. January 15 Notes from Sidney Coleman’s Physics 253a 337 According to the equation after Eq.

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->