GRADUATE STUDENT SERIES IN PHYSICS Series Editor: Professor Douglas F Brewer, MA, DPhil

Emeritus Professor of Experimental Physics, University of Sussex

Volume 2 Non-Abelian Gauge Theories: QCD and the Electroweak Theory IAN J R AITCHISON
Department of Physics University of Oxford

Department of Electronics and Computer Science University of Southampton

Copyright 2004 IOP Publishing Ltd

c ­ IOP Publishing Ltd 2004 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. Multiple copying is permitted in accordance with the terms of licences issued by the Copyright Licensing Agency under the terms of its agreement with Universities UK (UUK). British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. ISBN 0 7503 0950 4 Library of Congress Cataloging-in-Publication Data are available Front cover image: Simulation by the ATLAS experiment of the decay of a Higgs boson into four muons (yellow tracks). c CERN Geneva.


Commissioning Editor: John Navas Production Editor: Simon Laurenson Production Control: Sarah Plenty and Leah Fielding Cover Design: Victoria Le Billon Marketing: Nicola Newey and Verity Cooke Published by Institute of Physics Publishing, wholly owned by The Institute of Physics, London Institute of Physics Publishing, Dirac House, Temple Back, Bristol BS1 6BE, UK US Office: Institute of Physics Publishing, The Public Ledger Building, Suite 929, 150 South Independence Mall West, Philadelphia, PA 19106, USA
A Typeset in LTEX 2 by Text 2 Text Limited, Torquay, Devon Printed in the UK by MPG Books Ltd, Bodmin, Cornwall

Copyright 2004 IOP Publishing Ltd

To Jessie and to Jean, Katherine and Elizabeth

Copyright 2004 IOP Publishing Ltd


Preface to Volume 2 of the Third Edition Acknowledgments

PART 5 NON-ABELIAN SYMMETRIES 12 Global non-Abelian Symmetries 12.1 The flavour symmetry SU(2)f 12.1.1 The nucleon isospin doublet and the group SU(2) 12.1.2 Larger (higher-dimensional) multiplets of SU(2) in nuclear physics 12.1.3 Isospin in particle physics 12.2 Flavour SU(3)f 12.3 Non-Abelian global symmetries in Lagrangian quantum field theory 12.3.1 SU(2)f and SU(3)f 12.3.2 Chiral symmetry Problems 13 Local non-Abelian (Gauge) Symmetries 13.1 Local SU(2) symmetry: the covariant derivative and interactions with matter 13.2 Covariant derivatives and coordinate transformations 13.3 Geometrical curvature and the gauge field strength tensor 13.4 Local SU(3) symmetry 13.5 Local non-Abelian symmetries in Lagrangian quantum field theory 13.5.1 Local SU(2) and SU(3) Lagrangians 13.5.2 Gauge field self-interactions 13.5.3 Quantizing non-Abelian gauge fields Problems
Copyright 2004 IOP Publishing Ltd

PART 6 QCD AND THE RENORMALIZATION GROUP 14 QCD I: Introduction and Tree-Graph Predictions 14.1 The colour degree of freedom 14.2 The dynamics of colour 14.2.1 Colour as an SU(3) group 14.2.2 Global SU(3)c invariance and ‘scalar gluons’ 14.2.3 Local SU(3)c invariance: the QCD Lagrangian 14.3 Hard scattering processes and QCD tree graphs 14.3.1 Two-jet events in pp collisions ¯ 14.3.2 Three-jet events 14.4 Three-jet events in e+ e− annihilation Problems 15 QCD II: Asymptotic Freedom, the Renormalization Group and Scaling Violations in Deep Inelastic Scattering 15.1 QCD corrections to the parton model prediction for σ (e+ e− → hadrons) 15.2 The renormalization group and related ideas 15.2.1 Where do the large logs come from? 15.2.2 Changing the renormalization scale 15.2.3 The renormalization group equation and large −q 2 behaviour in QED 15.3 Back to QCD: asymptotic freedom 15.4 A more general form of the RGE: anomalous dimensions and running masses 15.5 Some technicalities 15.6 σ (e+ e− → hadrons) revisited 15.7 QCD corrections to the parton model predictions for deep inelastic scattering: scaling violations 15.7.1 Uncancelled mass singularities 15.7.2 Factorization and the DGLAP equation 15.7.3 Comparison with experiment Problems 16 Lattice Field Theory and the Renormalization Group Revisited 16.1 Introduction 16.2 Discretization 16.3 Gauge invariance on the lattice 16.4 Representation of quantum amplitudes 16.5 Connection with statistical mechanics 16.6 Renormalization and the renormalization group on the lattice 16.6.1 Introduction 16.6.2 The one-dimensional Ising model
Copyright 2004 IOP Publishing Ltd


16.6.3 Further developments and some connections with particle physics Numerical calculations Problems

PART 7 SPONTANEOUSLY BROKEN SYMMETRY 17 Spontaneously Broken Global Symmetry 17.1 Introduction 17.2 The Fabri–Picasso theorem 17.3 Spontaneously broken symmetry in condensed matter physics 17.3.1 The ferromagnet 17.3.2 The Bogoliubov superfluid 17.4 Goldstone’s theorem 17.5 Spontaneously broken global U(1) symmetry: the Goldstone model 17.6 Spontaneously broken global non-Abelian symmetry 17.7 The BCS superconducting ground state Problems 18 Chiral Symmetry Breaking 18.1 The Nambu analogy 18.1.1 Two flavour QCD and SU(2)f L ×SU(2)f R 18.2 Pion decay and the Goldberger–Treiman relation 18.3 The linear and nonlinear σ -models 18.4 Chiral anomalies Problems 19 Spontaneously Broken Local Symmetry 19.1 Massive and massless vector particles 19.2 The generation of ‘photon mass’ in a superconductor: the Meissner effect 19.3 Spontaneously broken local U(1) symmetry: the Abelian Higgs model 19.4 Flux quantization in a superconductor 19.5 ’t Hooft’s gauges 19.6 Spontaneously broken local SU(2)×U(1) symmetry Problems
Copyright 2004 IOP Publishing Ltd

PART 8 WEAK INTERACTIONS AND THE ELECTROWEAK THEORY 20 Introduction to the Phenomenology of Weak Interactions 20.1 Fermi’s ‘current–current’ theory of nuclear β -decay and its generalizations 20.2 Parity violation in weak interactions 20.3 Parity transformation of Dirac wavefunctions and field operators 20.4 V − A theory: chirality and helicity 20.5 Charge conjugation for fermion wavefunctions and field operators 20.6 Lepton number 20.7 The universal current–current theory for weak interactions of leptons 20.8 Calculation of the cross-section for νµ + e− → µ− + νe 20.9 Leptonic weak neutral currents 20.10 Quark weak currents 20.11 Deep inelastic neutrino scattering 20.12 Non-leptonic weak interactions Problems 21 Difficulties with the Current–Current and ‘Naive’ Intermediate Vector Boson Models 21.1 Violation of unitarity in the current–current model 21.2 The IVB model 21.3 Violation of unitarity bounds in the IVB model 21.4 The problem of non-renormalizability in weak interactions Problems 22 The Glashow–Salam–Weinberg Gauge Theory of Electroweak Interactions 22.1 Weak isospin and hypercharge: the SU(2) × U(1) group of the electroweak interactions: quantum number assignments and W and Z masses 22.2 The leptonic currents (massless neutrinos): relation to current– current model 22.3 The quark currents 22.4 Simple (tree-level) predictions p 22.5 The discovery of the W± and Z0 at the CERN p¯ collider 22.5.1 Production cross-sections for W and Z in p¯ colliders p 22.5.2 Charge asymmetry in W± decay p 22.5.3 Discovery of the W± and Z0 at the p¯ collider and their properties 22.6 The fermion mass problem 22.7 Three-family mixing 22.7.1 Quark flavour mixing
Copyright 2004 IOP Publishing Ltd

22.7.2 Neutrino flavour mixing 22.8 Higher-order corrections 22.9 The top quark 22.10 The Higgs sector 22.10.1 Introduction 22.10.2 Theoretical considerations concerning m H 22.10.3 Higgs phenomenology Problems Appendix M Group Theory M.1 Definition and simple examples M.2 Lie groups M.3 Generators of Lie groups M.4 Examples M.4.1 SO(3) and three-dimensional rotations M.4.2 SU(2) M.4.3 SO(4): The special orthogonal group in four dimensions M.4.4 The Lorentz group M.4.5 SU(3) M.5 Matrix representations of generators and of Lie groups M.6 The Lorentz group M.7 The relation between SU(2) and SO(3) Appendix N Appendix O Dimensional Regularization Grassmann Variables

Appendix P Majorana Fermions P.1 Spin- 1 wave equations 2 P.2 Majorana quantum fields Appendix Q Feynman Rules for Tree Graphs in QCD and the Electroweak Theory Q.1 QCD Q.1.1 External particles Q.1.2 Propagators Q.1.3 Vertices Q.2 The electroweak theory Q.2.1 External particles Q.2.2 Propagators Q.2.3 Vertices References

Copyright 2004 IOP Publishing Ltd


Volume 1 of our new two-volume third edition covers relativistic quantum mechanics, electromagnetism as a gauge theory, and introductory quantum field theory, and leads up to the formulation and application of quantum electrodynamics (QED), including renormalization. This second volume is devoted to the remaining two parts of the ‘Standard Model’ of particle physics, namely quantum chromodynamics (QCD) and the electroweak theory of Glashow, Salam and Weinberg. It is remarkable that all three parts of the Standard Model are quantum gauge field theories: in fact, QCD and the electroweak theory are certain generalizations of QED. We shall therefore be able to build on the foundations of gauge theory, Feynman graphs and renormalization which were laid in Volume 1. However, QCD and the electroweak theory both require substantial extensions of the theoretical framework developed for QED. Most fundamentally, the discussion of global and local symmetries must be enlarged to include nonAbelian symmetries, and spontaneous symmetry breaking. At a somewhat more technical level, the lattice (or path-integral) approach to quantum field theory, and the renormalization group are both needed for access to modern work on QCD. For each of these theoretical elements, a self-contained introduction is provided in this volume. Together with their applications, this leads to a simple four-part structure (the numbering of parts, chapters and appendices continues on from Volume 1): Part 5 Non-Abelian symmetries Part 6 QCD and the renormalization group (including lattice field theory) Part 7 Spontaneous symmetry breaking (including the spontaneous breaking of the approximate global chiral symmetry of QCD) Part 8 The electroweak theory. We have already mentioned several topics (path integrals, the renormalization group, and chiral symmetry breaking) which are normally found only in texts pitched at a more advanced level than this one—and which were indeed largely omitted from the preceding (second) edition. Nor, as we shall see, are these topics the only newcomers. With their inclusion in this volume, our book now becomes a comprehensive, practical and accessible introduction to the major theoretical and experimental aspects of the Standard Model. The emphasis is crucial: in once again substantially extending the scope of the book, we have tried hard not to
Copyright 2004 IOP Publishing Ltd

We decided against banishing these matters to an appendix. the second in part 5. On the mathematical side. ‘parallel transport’. as before. and also section 13. Much of the necessary mathematics already appears in the simpler case in which the symmetry is a global. which collects together the main relevant ideas. In chapter 12 we introduce global non-Abelian symmetries via the physical examples of the (approximate) SU(2) and SU(3) flavour symmetries of the strong interactions. namely general relativity. to make the chosen material accessible to the wide readership which the previous editions evidently attracted. which is expected to be relevant if the quark masses are substantially less than typical hadronic scales. We hope that this compact summary will be of use to those readers who want a sense of the mathematical unity behind the succession of specific calculations provided in the main text. being directly relevant to the electroweak theory and to QCD respectively. This refers to the fact that the symmetry transformations are matrices (acting on a set of fields). we have provided a new and fairly serious appendix (M) on group theory. since they are such a significant part of the conceptual structure of all gauge theories.5. ‘connection’. not present in QED—for example. so that no prior knowledge of group theory is necessary. we take care to develop everything we need on a ‘do-it-yourself ’ basis as we go along. A further important global non-Abelian symmetry is also introduced in chapter 12—that of chiral symmetry.5 at the corresponding nonAbelian gauge field theories which are the main focus of the book.2 and 13. the algebraic (or group-theoretic) aspects developed in chapter 12 carry through unchanged into chapter 13. the self-interactions among the gauge field quanta. we move on to the local versions of SU(2) and SU(3) symmetry.3. while at the same time highlighting those items which are new to this edition.3. All the same. The apparent non-observation of this expected symmetry creates a puzzle. so that the order in which they are applied makes a difference. but the ‘gauging’ of the symmetry brings in some new geometrical concepts. It may therefore be helpful to offer some more detailed guidance. any two of which will generally not commute with each other. not all of the topics are likely to be of equal interest to every reader. A glance at the contents will suggest that we have set ourselves a considerable challenge. such as ‘covariant derivative’. rather than a local one. the resolution of which has to be deferred until part 7. Nevertheless. arriving in section 13. non-Abelian symmetry.compromise the title’s fundamental aim—which is. Crucial new physical phenomena appear. practicallyminded readers may want to pass quickly over sections 13. The underlying mathematics involved here is group theory. However. which explains why obtaining the correct Feynman rules for loops Copyright 2004 IOP Publishing Ltd . First. as is indeed the case. and shows how they apply to the groups we are dealing with (including the Lorentz group). and ‘curvature’. On the other hand. and moreover their inclusion allows instructive reference to be made to a theory otherwise excluded from mention. In chapter 13. then.

2 we have tried to provide an elementary introduction to the RG. We do not expect the majority of our readers to find chapter 15 easy going. At this half-way stage. and it is described in chapter 15. In fact. But of course a fundamental question immediately arises: how can such an approach. Some simple tree-graph applications are then described. possibly apply to QCD which. In section 15.5 are less central to the main argument. The technique required to do this goes by the name of the renormalization group (RG). These provide a good first orientation to a non-Abelian gauge theory is such a difficult problem. in section 16. In chapter 16. using the techniques learned for QED. describes the strong interactions between quarks? The answer lies in the profound property (possessed only by non-Abelian gauge theories) called ‘asymptotic freedom’—that is.7. and leads on to the QCD Lagrangian. perturbation theory becomes useful at high energies only after an infinite series of loop contributions has been effectively re-summed. which form part 6. this property cannot be understood in terms of tree graphs: loops must be studied. based on the discretized (lattice) version of Feynman’s path-integral formulation of quantum field theory. as is an appendix (N) on dimensional regularization. by considering in detail the much simpler case of QED.4 and 15. and to the calculation of scaling violations in deep inelastic scattering. using no more theory than is contained in chapter 11 of volume 1. the third of part 6. Immediate application of the formalism can now be made to QCD. In particular. after all. Copyright 2004 IOP Publishing Ltd . to which we provide a simple introduction in section 16. based on perturbation theory. nor its direct relevance to experiment. and this occupies most of the next three chapters. An appendix (O) on Grassmann variables is provided for those interested in seeing how the path-integral formalism can be made to work for fermions. by the success of both RG-improved perturbation theory and nonperturbative numerical computations. following ‘parton model’ and ‘scaling’ ideas. and this immediately involves renormalization. The chapter ends with some illustrative results from lattice QCD calculations. leading to an effective theory valid at longer distances.4. Chapter 14 introduces ‘colour’ as a dynamical degree of freedom. But there is no denying the central importance of the RG in modern field theory. within the ‘canonical’ approach to quantum field theory as developed in volume 1. A substantial bonus of this formulation is that it allows fruitful analogies to be drawn with the statistical mechanics of spin systems on a lattice. we turn to the problem of how to extract predictions from a quantum field theory (in particular. Crucially. QCD) in the nonperturbative regime. Sections 15. we hope that readers who may have struggled with the formal manipulations of chapter 15 will be refreshed by seeing RG ideas in action from a different and more physical point of view—that of ‘integrating out’ short distance degrees of freedom. the decrease of the effective interaction strength at high energies or short distances. along with applications to asymptotic freedom. The available technique is computational. QCD has been established as the theory of strong interactions.

and the BCS superconductor in section 17. is rather more optional. we consider the spontaneous breaking of local (gauge) symmetries. and aspects of Higgs phenomenology. in support of which we provide an appendix (P) on Majorana fermions. In chapter 19. and with theories in which the W and Z bosons are given a ‘naive’ mass. Section 18. and also considers two physical examples in considerable detail. and to the non-Abelian case required for the electroweak theory. The remarkably precise agreement—thus far—between theory and experiment. However. Copyright 2004 IOP Publishing Ltd .4. together with more detail on higher order (one-loop) corrections.2 and 19. We consider applications both to the Abelian case of a superconductor (sections 19. in part 8. in chapter 22.Further progress requires one more fundamental idea—the subtle concept of spontaneous symmetry breaking. spontaneously broken. New additions here include threefamily mixing via the CKM matrix. our presentation of chiral symmetry breaking in chapter 18 follows Nambu’s remarkable original analogy between fermion mass generation and the appearance of an energy gap in a superconductor. and of lepton number conservation taking into account the possibility that neutrinos may be Majorana particles. But the story of the Standard Model is not yet quite complete. Chapter 21 describes what goes wrong with the current–current model. the characteristic methods and concepts developed for such systems provide valuable background for the particle physics applications of the idea. which depends upon the inclusion of one-loop effects. Finally. and a lively interest in their outcome. a valuable working model of the physics). on linear and nonlinear sigma models. requires input from the next generation of experiments. Chapter 17 sets out the basic theory of spontaneously broken global symmetries. Further progress in understanding the mechanism of electroweak symmetry breaking. namely the Bogoliubov superfluid in section 17.3. the top quark. primarily at the LHC. In particular. Here the fundamental point is that it is possible for gauge quanta to acquire mass. when interacting weakly. Chapter 20 is a self-contained review of weak interaction phenomenology. and suggests why a gauge theory is needed to avoid these difficulties. based on Fermi’s ‘current–current’ model.3.7. the third in part 7. which forms the subject of part 7. all the pieces are put together in the presentation of the electroweak theory. which follow in the next two chapters. as is our brief introduction to chiral anomalies in section 18. New material here includes discussion of the discrete symmetries C and P. makes it hard to deny that. It is of course true that these systems are not part of the standard model of particle physics. We hope that we leave our readers with a sound grasp of what is at stake in these experiments. Nature has indeed made use of the subtle intricacies of a renormalizable. One vital part—the Higgs sector—remains virtual. while still preserving the local gauge symmetry of the Lagrangian. and phenomenological. The way is now clear to develop the electroweak theory. non-Abelian chiral gauge theory. and of mass generation.4—once again.

for section 16. In addition. made in the first volume. old and new. we are delighted to thank two new correspondents. there still will be. and Nikki Fathers again provided essential help in making the electronic version. as well as electronically. whom we look forward to greeting physically.7. Paolo also read drafts of chapters 12 and 14. of course: we hope readers will draw them to our attention. and we thank him for his careful reading and encouraging comments. Errors. he made numerous excellent suggestions for improving our treatment of weak interactions in the second edition (parts 4 and 5). and between them found a good many misprints and been invaluable. More particularly: Keith Hamilton was a willing ‘guinea pig’ for chapters 14–17. once again.htm?&book=1130p. Above all.Acknowledgments Our expression of thanks to friends and colleagues. Ian J R Aitchison and Anthony J G Hey October 2003 Copyright 2004 IOP Publishing Ltd . the constant and unstinting help of our good friend George Emmons throughout the genesis and production of the book has. Paolo Strolin and Peter Williams both worked very carefully through Volume 1. and—where feasible—many of them have been incorporated into the present part 8. An up-to-date list will be posted on the book’s website: http://bookmarkphysics. and made many useful comments. We wish it had been possible to send him more chapters: however.iop. applies equally to this one. we are grateful to Chris Allton for advice on lattice gauge theory results.

PART 5 NON-ABELIAN SYMMETRIES Copyright 2004 IOP Publishing Ltd .

since the phase factors commuted. and the flavour independence of the QCD interquark forces. together with flavour independence of QCD.2 is then reformulated in field-theoretic language in section 12. one is led to a nonAbelian gauge theory. This may be expected to be a good approximate symmetry for the u and d quarks. the symmetry is a ‘global non-Abelian’ one.12 GLOBAL NON-ABELIAN SYMMETRIES In the preceding volume. d and s quark masses. which generally do not commute with each other. in which the transformations involve more than one state. which lead to useful conservation laws but not to any specific dynamical theory. as reflecting the rough equality of the u. the hadronic isospin ‘SU(2) symmetry’ proposed by Heisenberg (1932) in the context of nuclear physics. In this chapter we shall be mainly concerned with two global non-Abelian symmetries. When the phase factors are independent of the spacetime coordinate x . providing generalizations of the Abelian U(1) gauge theory which is QED. and the associated symmetry is called a ‘non-Abelian’ one. based on the remarkably simple gauge principle: namely that the theory should be invariant under local phase transformations on the wavefunctions (chapter 3) or field operators (chapter 7) of charged particles.2 we extend this to SU(3)f flavour symmetry. Both QCD and the electroweak theory are of the latter type. In section 12. QCD and the electroweak theory. Such transformations were characterized as Abelian in section 3.1 and 12. which is a symmetry of theories with massless fermions. when they are allowed to depend on x . as was first done by Gell-Mann (1961) and Ne’eman (1961)—an extension seen. at a time. In that case.1 with the first non-Abelian symmetry to be used in particle physics. The second volume of this book will be largely concerned with the formulation and elementary application of the remaining two dynamical theories within the Standard Model—that is. It is a striking fact that all three dynamical theories in the Standard Model are based on a gauge principle of local phase invariance. They are built on a generalization of the gauge principle. nucleon Copyright 2004 IOP Publishing Ltd . But the anticipated observable consequences of this symmetry (for example. In the last section of this chapter. a very successful dynamical theory—QED—has been introduced. The ‘wavefunction’ approach of sections 12. We begin in section 12. we shall introduce the idea of a global chiral symmetry.6. and now seen as following from the near equality of the u and d quark masses (on typical hadronic scales). or field. the ‘phase factors’ become matrices.3. in its turn.

gauge invariance forbids a mass term in the Lagrangian for non-Abelian gauge fields. far from obvious why either the strong interactions between quarks or the weak interactions should have anything to do with gauge theories at all. it describes the interactions between the constituents of the hadrons. s (see chapter 14). this constitutes the electroweak theory (see part 8). pp 474-82. But this. In the case of the strong quark interactions. However. Copyright 2004 IOP Publishing Ltd . In the weak interaction case. This puzzle will be resolved in part 7. The formalism introduced in this chapter for SU(2) and SU(3) will be required again in the following one. due to exchange of these massless quanta—and of course in neither the strong nor the weak interaction case is that what is observed. physically quite distinct from hadronic spin. together with QED.1 As regards the former. as we indicated in chapter 2. But we must always remember that it is weak isospin. In fact. We shall discuss this further in chapter 16. hence. a local SU(3) symmetry (called SU(3)c ) governs the strong interactions of quarks. the physics involved is completely different. Furthermore. would imply that the associated forces must have a long-range (Coulombic) part. when we consider the local versions of these non-Abelian symmetries and the associated dynamical gauge theories. particularly since flavour degrees of freedom are involved. Second. The whole modern development of non-Abelian gauge theories began with the attempt by Yang and Mills (1954) (see also Shaw 1955) to make hadronic isospin into a local symmetry. it is natural to use ‘isospin language’ in talking about it. as we shall see in chapter 13. namely quarks—and this in two respects. Indeed it is worth emphasizing very strongly that it is. a third realization appears: the gauge quanta acquire mass via (it is believed) a second instance of spontaneous symmetry breaking. Despite the attractive conceptual unity associated with the gauge principle. d. Thus it would seem that gauge field quanta are necessarily massless. despite the fact that each of these two local symmetries is based on the same group as one of the earlier global (flavour) symmetries. via the profoundly important concept of ‘spontaneous symmetry breaking’. the beautiful formalism developed by these authors turned out not to describe interactions between hadrons. the way in which each of QCD and the electroweak theory ‘works’ is actually quite different from QED and from each other. a local SU(2) symmetry (called weak isospin) governs the weak interactions of quarks (and leptons). which (as we shall see in chapter 20) is an attribute of leptons as well as of quarks and. in turn. Instead. since the group is an SU(2). as will be explained in part 7. It is important to realize that. First. In weak interactions. the gluon quanta are indeed massless but the contradiction is resolved by non-perturbative effects which lead to confinement.parity doublets) appear to be absent. Just as in the U(1) (electromagnetic) case. a further application of this idea is required in the electroweak theory because of 1 Pauli had independently developed the theory of non-Abelian gauge fields during 1953 but did not publish any of this work because of the seeming physical irrelevancy associated with the masslessness problem (Enz 2002. binding them into hadrons (see part 6). Pais 2002. a priori. it is a parity-violating chiral gauge theory. pp 242-5). the SU(3)c group refers to a new degree of freedom (‘colour’) which is quite distinct from flavour u.

Quite generally in quantum mechanics. In particular. Thus redefinitions of neutron and proton wavefunctions could be allowed. the chiral nature of the gauge symmetry in this case: the quark and lepton masses also must be ‘spontaneously generated’. provided the normalization conditions on the states are still satisfied. as far as this force was concerned. as far as the strong nuclear forces were concerned (electromagnetism being negligible by comparison). the neutron and proton (figure 12.4) (12.1). we know that whenever we have a set of states which are degenerate in energy (or mass) there is no unique way of specifying the states: any linear combination of some initially chosen set of states will do just as well. we have H ψp = Eψp from which it follows that H ψp = H (αψp + βψn ) = α H ψp + β H ψn = E(αψp + βψn ) = Eψp Copyright 2004 IOP Publishing Ltd H ψn = Eψn (12. or state.5) (12. (12.3) for complex coefficients α. so that any arbitrary linear combination of neutron and proton wavefunctions would be entirely equivalent.Figure 12.2) (12.1) The generalization to non-Abelian invariances comes when we take the simple step—but one with many ramifications—of considering more than one wavefunction. the simplest case of just two such states—to be specific. This hypothesis became known as the ‘charge independence of nuclear forces’. Consider.1.1. β.6) . for example. This single near coincidence of the masses was enough to suggest to Heisenberg (1932) that. of the form ψp → ψp = αψp + βψn ψn → ψn = γ ψp + δψn (12.1 The flavour symmetry SU(2)f 12. for a single ‘neutron’ or single ‘proton’ wavefunction. γ and δ. 12. the two states could be regarded as truly degenerate. since ψp and ψn are degenerate.1 The nucleon isospin doublet and the group SU(2) The transformations initially considered in connection with the gauge principle in section 3. Early evidence for isospin symmetry. at a time.5 were just global phase transformations on a single wavefunction ψ = eiα ψ.

this terminology arises. even though both the proton and the neutron have (true) spin. however. an intrinsically approximate symmetry. Indeed.2) and (12. a symmetry was involved. was that the physics of strong interactions between nucleons remained the same under the transformation (12.59) corresponding to eigenvalues ± 1 h of (true) spin: compare 2¯ also (4.7) showing that the redefined wavefunctions still describe two states with the same energy degeneracy.81).10) where V is the indicated complex 2 × 2 matrix.and.9) and the two-component eigenvectors (4.2) and 2 (12. H ψn = Eψn (12. We must emphasize that such a symmetry can only be exact in the absence of electromagnetic interactions: it is.81). It is important to be clear. As far as the states are concerned.1 is suggestive of that found in spin. we are suppressing the spinor part of both wavefunctions altogether (they are of course 4-component Dirac spinors). as constrained by various relevant restrictions: quite remarkably.8). then. therefore.3) leave the (true) spin part of their wavefunctions completely untouched. The two-fold degeneracy seen in figure 12.1 ’ analogy will become clear. of course. which we encountered in the discussion of the transformation of (real) spin wavefunctions under rotations of the (real) space axes. in particular. As we proceed. Copyright 2004 IOP Publishing Ltd . that the degrees of freedom involved in the two cases are quite distinct.10): in other words. We now consider the general form of the matrix V. 2 Equations (12. ψp is the amplitude for the nucleon to have ‘isospin up’ and ψn is that for it to have ‘isospin down’. we shall discover that (after extracting an overall phase) V has essentially the same mathematical form as the matrix U of (4. from the formal identity between the ‘isospinors’ of (12. similarly. Heisenberg’s proposal.9) ψp ψn ≡ ψp χp + ψn χn (12. the precise mathematical nature of this ‘spin. though presumably quite a useful one in view of the relative weakness of electromagnetic interactions as compared to hadronic ones. (12.8) In ψ (1/2) .3) can be compactly written in terms of ψ (1/2) as ψ (1/2) → ψ (1/2) = Vψ (1/2) V= α γ β δ (12. The analogy can be brought out by introducing the two-component nucleon isospinor ψ (1/2) ≡ where χp = 1 0 χn = 0 1 .1 .1 systems in the absence of any magnetic field: the sz = ± 1 components are 2 2 degenerate. the transformations (12. It will be instructive to see how the present discussion leads to the same form (4.60) and (12.

2. for which. (12.1) with θ = 2α . In the field-theoretic formalism of section 7.6. therefore. (12. finally. because it corresponds to a rotation of the phase of both p and n wavefunctions by the same amount: ψp = eiα ψp ψn = eiα ψn .1). in order to preserve the normalization of ψ (1/2) . V reduces to a single phase factor as in (12.81) depends on only three real parameters: two to describe the axis of rotation represented by the unit vector n.15) (12. of the form (12. V is subject to certain restrictions and these reduce the number of free parameters in V to three. as we noted in section 3. together with a third for the angle of rotation θ .11) where 12 is the unit 2 × 2 matrix. alternatively. the elements of the group are all such 2 × 2 matrices and the ‘law of combination’ is just ordinary Copyright 2004 IOP Publishing Ltd . the matrix U of (4. we require ψ (1/2) † ψ (1/2) = ψ (1/2)† V† Vψ (1/2) = ψ (1/2)† ψ (1/2) which implies that V has to be unitary: V† V = 1 2 (12.We first note that V of (12. Indeed. such a symmetry can be shown to lead to the conservation of baryon number Nu + Nd − Nu − Nd . By contrast. we have det V† V = det V† · det V = det V∗ · det V = | det V|2 = 1 so that det V = exp(iθ ) (12. Thus. as we now discuss.10) depends on four arbitrary complex numbers or. We can separate off such an overall phase factor from the transformations mixing ‘p’ and ‘n’. on eight real parameters.16) Such a matrix is said to be a special unitary matrix—which simply means it has unit determinant. The general defining properties of a group are given in appendix M. where bar denotes the anti-particle. from the elementary properties of determinants. In the present case. 2 × 2 matrices.15) is V = eiα 12 . The set of all such matrices form a group. A trivialization is the case n = 1. unitary.13) The V corresponding to (12. Clearly this unitarity property is in no way restricted to the case of two states—the transformation coefficients for n degenerate states will form the entries of an n × n unitary matrix.12) (12. which has determinant exp(2iα) and is. indicating how all the previous work is going to be contained as a special case of these more general transformations. However.14) where θ is a real number. First. ¯ ¯ The new physics will lie in the remaining transformations which satisfy det V = +1. the V’s we are dealing with are special.

21) where a. The condition that Vinfl be unitary. 3) · τ /2 (12. For such an infinitesimal SU(2) matrix Vinfl . Such groups have the important property that their physical consequences may be found by considering ‘infinitesimal’ transformations. The condition det Vinfl = 1 now reduces. to the condition (see problem 12.1) that all the defining properties are satisfied here.23) (12. on neglect of second-order terms O(ξ 2 ).24) which are all first-order small. which means it must have the form ξ= a b + ic b − ic −a (12. the ‘U’ for ‘unitary’ and the ‘2’ for ‘2 × 2’. The three matrices τ are just the familiar Hermitian Pauli matrices τ1 = 0 1 1 0 τ2 = 0 −i i 0 τ3 = 1 0 0 −1 (12.20) (12. 2.22) (12.2) Tr ξ = 0. i.21) can be put in the more suggestive form ξ= where stands for the three quantities = ( 1. that is—in this case—matrices V which differ only slightly from the ‘no-change’ situation corresponding to V = 12 .e. (12 + iξ )(12 − iξ † ) = 12 similarly reduces (in first order) to the condition ξ = ξ †. (12.25) Copyright 2004 IOP Publishing Ltd .18) Thus ξ is a 2 × 2 traceless Hermitian matrix.19) (12. we may therefore write Vinfl = 12 + iξ (12. Writing a= 3 /2 b= 1 /2 c= 2 /2 (12. It is straightforward to verify (problem 12.matrix multiplication. SU(2) is actually an example of a Lie group (see appendix M). c are infinitesimal parameters.17) where ξ is a 2 × 2 matrix whose entries are all first-order small quantities. b. the ‘S’ standing for ‘special’. the group is called ‘SU(2)’.

here called ‘tau’ precisely in order to distinguish them from the mathematically identical ‘sigma’ matrices which are associated with the real spin degree of freedom.1 In that case. introducing T(1/2) ≡ τ /2.23). Tj ] = i i j k Tk . j and k run from 1 to 3 and a sum on the repeated index k is understood as usual.26) and an infinitesimal SU(2) transformation of the p–n doublet is specified by ψp ψn = (12 + i · τ /2) ψp ψn . Of course. We note that they do not commute amongst themselves: rather.1 ’ one. (12. the choice Ji = σi /2 ≡ Ji 2 system.28) as being precisely the same as those of angular momentum operators in quantum mechanics: [ Ji . the matrices τ /2 provide a particular matrix representation of the generators. (12. They are called the generators of infinitesimal SU(2) transformations.31) Copyright 2004 IOP Publishing Ltd . we find (see problem 12. Let = α/n. J j ] = i i j k Jk . more precisely. (12. namely the twodimensional or ‘fundamental’ one (see appendix M).29) would correspond to a (real) spin.28) [Ti where i. where α = (α1 . since they determine the forms of the three independent infinitesimal SU(2) transformations. Hence. We obtain V = exp(iα · τ /2) so that ψ (1/2) ≡ ψp ψn = exp(iα · τ /2) ψp ψn = exp(iα · τ /2)ψ (1/2) . The reader will recognize the commutation relations (12. (12. apply the infinitesimal transformation n times and let n tend to infinity.30) generalized to matrices.27) The τ -matrices clearly play an important role. a general infinitesimal SU(2) matrix takes the form Vinfl = (12 + i · τ /2) (12. Here the identity between the tau’s and the sigma’s gives us a good reason to regard our ‘p–n’ system as formally analogous to a ‘spin. The form for a finite SU(2) transformation V may then be obtained from the infinitesimal form using the result e A = lim (1 + A/n)n n→∞ (1/2) (12.32) (12. α3 ) are three real finite (not infinitesimal) parameters. the 2 ‘analogy’ was made into a mathematical identity by the judicious way in which ξ was parametrized in (12.3) (1/2) (1/2) (1/2) . α2 .

We may now put our initial ‘spin. and n: for further discussion. again in section 12. it is easy to see what these operators are.1).6) and (12. Indeed. Equation (12. In the case of the p–n doublet. in the case of electric charge or particle number.7. (12.3.37) 0 ≤ θ ≤ 2π . (12. As promised. The last fact is the reason for the description ‘non-Abelian’ phase invariance. We shall meet generalizations of this structure in the following section for SU(2).36) and H2ψ (1/2) = Eψ (1/2) (12. three real parameters appear: in (4. In each case.35) where the ‘Hermitian matrix’ is composed of the generators and the transformation parameters. By contrast.34) where we used the Hermiticity of the tau’s. in (12.32) is of essentially the same mathematical form as (4. commute. In quantum mechanics.33) has the general form unitary matrix = exp(i Hermitian matrix).Note that in the finite transformation. as noted in section 3. As the commutation relations for the τ matrices show.2 for SU(3).31) has the form V = exp(iG) (12. ˆ In the form (12. (12. (12.32).4). ˆ 2 It is not completely obvious that the general SU(2) matrix can be parametrized by an angle θ with Copyright 2004 IOP Publishing Ltd . We may write (12. successive transformations clearly commute: this corresponds to an Abelian phase invariance and. it is clear that our 2 × 2 isospin transformation is a generalization of the global phase transformation of (12. SU(2) is a non-Abelian group in that two SU(2) transformations do not.81) there are three parameters to describe the axis n and angle θ of rotation. in general.33) where G = α · τ /2.6.32) there are just the three ˆ components of α . to an Abelian U(1) group.81). and (b) there are non-commuting matrix operators (the τ ’s) appearing in the exponent. and a field-theoretic version of it in section 12. see appendix M. states within a degenerate multiplet may conveniently be characterized by the eigenvalues of a complete set of Hermitian operators which commute with the Hamiltonian and with each other. (12. from which the unitary property of V easily follows: V† = exp(−iG † ) = exp(−iG) = V−1 . section M. the generators appear in the exponent.7) as H2ψ (1/2) = Eψ (1/2) (12.1 ’ analogy on a more precise mathematical 2 footing. except that (a) there are now three ‘phase angles’ α . We can always2 write α = |α|α and identify |α| with θ and α ˆ ˆ with n.

43) The reason for the ‘spin’ part of the name ‘isospin’ should by now be clear: the term is actually a shortened version of the historical one ‘isotopic spin’. as we would expect. according to the general principles of quantum mechanics. The corresponding eigenfunctions are just the χp ’s and χn ’s of (12. equivalently.where H2 is the 2 × 2 matrix H2 = H 0 0 H . Equation (12.40). VH2V−1 = H2 (12. it does not commute with either τ1 or τ2 . τ ] = 0. electromagnetic corrections to the strong interaction Hamiltonian will violate SU(2) symmetry. The answer is3 that the total squared ‘spin’ 2 2 2 (T(1/2) )2 = ( 1 τ )2 = 1 (τ1 + τ2 + τ3 ) = 3 12 2 4 4 (1/2) (12. Thus.39) which is the statement that H2 is invariant under the transformation (12.41) and one component of spin.39). can be given definite values 2 simultaneously. 2 (12. (12. Mandl (1992). which satisfy 1 2 4 τ χp 1 2 4 τ χn = 3 χp 4 = 3 4 χn 1 2 τ 3 χp 1 2 τ 3 χn = 1 χp 2 = − 1 χn . correspond to possible observables.32). associated with the invariance (12. In concluding this section we remark that.e. given the commutation relations (12. Copyright 2004 IOP Publishing Ltd .38) Hence H2 is proportional to the unit matrix in this two-dimensional space and it therefore commutes with the tau’s: [H2. in this two-dimensional p–n space.9). for example.29)).40) (12.28).44) It is clear that although Qem commutes with τ3 . 3 See. 2 (12. is easily solved by making use of the formal identity between the operators τi /2 and angular momentum operators Ji (cf (12. Now the tau’s are Hermitian and. say T3 = 1 τ3 . But the tau’s do not commute amongst themselves and so. the electromagnetric charge operator is represented by the matrix Qem = 1 0 0 0 = 1 (12 + τ3 ).39) implies that their eigenvalues are constants of the motion (i. It then also follows that H2 commutes with V or. conserved quantities). hence.42) (12. we cannot give definite values to more than one of them at a time. The problem of finding a classification of the states which makes the maximum use of (12.

. .28) for each of the individual τ ’s imply (see problem 12. . which can be simultaneously diagonalized along with H . −T + 1. and for the neutron–proton mass difference) provide clear evidence of T = 1 (doublet).. For A nucleons in the nucleus. Further. which are simply the standard angular momentum commutation relations.40) for all the nucleons independently. 3. the eigenvalues of the T operators are constants of the motion. . . we introduce three ‘total isospin operators’ T = (T1 . j.1. ‘good quantum numbers’ for nuclei. What are the possible eigenvalues? We know that the T’s are Hermitian and satisfy exactly the same commutation relations (12. (12. and correcting for the slight neutron–proton mass difference—by the eigenvalues of T2 and T3 . Thus the energy levels of nuclei ought to be characterized—after allowance for electromagnetic effects.8) which we reproduce here as figure 12. . if the assumed isospin invariance is true. T] = 0. and since T3 counts + 1 for every proton and − 1 for every neutron. there are 2T + 1 degenerate states for a given T . 2. it is clear that successive 2 2 values of T3 correspond physically to changing one neutron into a proton or vice versa. T . the commutation relations (12. These conditions are all that are needed to show that the eigenvalues of T2 are of the form T (T + 1). T2 . T = 1 (triplet) and T = 3 (quartet) multiplets. and that for a given T 2 the eigenvalues of T3 are −T. These level schemes (which have been adjusted for Coulomb energy differences. the foregoing is really nothing more than the general quantum mechanics of a two-state system.4) that the components of T defined by (12.46) Thus. once more. It then follows that [ H.45) which are Hermitian. since the isospin operators for different nucleons commute with each other (they are quite independent).47) for i.45) satisfy the commutation relations [Ti . where T = 0.2 for convenience. Thus we expect to see ‘charge multiplets’ of levels in neighbouring nuclear isobars.2 Larger (higher-dimensional) multiplets of SU(2) in nuclear physics For the single nucleon states considered so far. phrased in ‘spin1 2 ’ language. These are precisely the multiplets of which we have already introduced examples in chapter 1 (see figure 1. The real power of the isospin (SU(2)) symmetry concept becomes more apparent when we consider states of several nucleons.47) as the angular momentum operators. k = 1. that is. 1. These states all have the same A value. T − 1. 1 . It is important to 2 2 note that states in the same T -multiplet must have the same J P quantum numbers Copyright 2004 IOP Publishing Ltd . These eigenvalues should then be. say. to a good approximation.12. T3 ) via T = 1 τ (1) + 1 τ (2) + · · · + 1 τ ( A) 2 2 2 (12. The Hamiltonian H describing the strong interactions of this system is presumed to be invariant under the transformation (12. T j ] = i i j k Tk (12. .

in fact. . the same. Thus. . which themselves correspond to T = 1 . the assumed invariance of the nucleon–nucleon force produces a richer nuclear multiplet structure. nevertheless. These higher-dimensional multiplets (T = 1.2. Energy levels (adjusted for Coulomb energy and neutron–proton mass differences) of nuclei of the same mass number but different charge. Copyright 2004 IOP Publishing Ltd .47) are called the Lie algebra of SU(2)4 (see appendix M) and the general group-theoretical problem of understanding all possible multiplets for SU(2) is equivalent to the problem of finding matrices which satisfy these commutation relations. 3 .Figure 12. going beyond the original p–n doublet. . section M. see appendix M. (b) triplets and (c) doublets and quartets. the angular momentum commutation relations (12. therefore. the groups do differ.29) are the Lie algebra of the rotation group SO(3). These are. (these are indicated on the levels for 18 F).7. The Lie algebras of the two groups are. precisely the angular momentum matrices of dimension (2 T + 1) × (2 T + 1) which are generalizations of the τ /2’s. showing (a) ‘mirror’ doublets. The commutation relations (12. 2 4 Likewise.) are called ‘irreducible 2 representations’ of SU(2). obviously the nuclear forces will depend on the space and spin degrees of freedom of the nucleons and will only be the same between different nucleons if the space-spin part of the wavefunction is the same. For an indication of how.

and similarly for ‘d’. The u–d isospin doublet plays the role of the p–n doublet in the nuclear case and this degree of freedom is what we now call SU(2) isospin flavour symmetry at the quark level. and 2 in the meson sector.48) where the numbers −i i j k are deliberately chosen to be the same numbers (with a minus sign) that specify the algebra in (12. just as in the nuclear (p. denoted by SU(2)f .49) which has the general form of (12.5).3. doublets). We shall denote the u–d quark doublet wavefunction by u q= (12. corresponding to T = 3 .1. which is particle physics. The generalization of (12. the π’s appear as the lowest states of a sequence of mesonic triplets (T = 1). as shown in figure 1. We now proceed to consider isospin in our primary area of interest. For example.n) case. sections indicated in the notation T( 2 ) . These entities are.11.8). 2 The most natural interpretation of these facts is that the observed states are composites of more basic entities which carry different charges but are nearly degenerate in mass.10(a)). there will be (2 T + 1) × (2 T + 1) matrices T(T ) which satisfy (12. the matrices T(T ) provide a (2 T + 1)-dimensional matrix representation of the generators of SU(2). Many other examples also exist but with one remarkable difference as compared to the nuclear physics case: no baryon states are known with T > 3 nor any meson states with T > 1.47): the latter are called the structure constants of the SU(2) group (see appendix M. 1 (12. Then.50) d omitting now the explicit representation label ‘( 1 )’ and shortening ‘ψu ’ to just 2 ‘u’. as 2 noted in chapter 1 (see figure 1.51) .10(b). ddd). Copyright 2004 IOP Publishing Ltd (12.35). q → q = Vq = exp(iα · τ /2) q.32) to these 2 higher-dimensional multiplets is then ψ (T ) = exp(iα · T(T ) )ψ (T ) (12. while the forces between these entities are chargeindependent.e. shown in figure 1. uud.3 Isospin in particle physics The neutron and proton states themselves are actually only the ground states of a whole series of corresponding B = 1 levels with isospin 1 (i. Another series of baryonic levels comes in four charge states. udd. the T = 1 matrices are 3 × 3 and can be compactly summarized by (problem 12. In general. of course.3–M. We shall meet field-theoretic representations of the generators in section 12.5) (Ti(1) ) j k = −i ijk.47) and correspondingly (2T + 1) dimensional wavefunctions ψ (T ) analogous to the twodimensional (T = 1 ) case of (12. 12. In this case. under an SU(2)f transformation. the p is (uud) and the -quartet is (uuu. the quarks: the n contains (udd).

provide a valid matrix representation of the SU(2) generators. T = 1 quark has B = − 3 . a challenge for QCD to explain why—for example—states with four or five quarks should not exist (nor states of one or two quarks!) and why a state of six quarks. Note that. +τ2 /2 and −τ3 /2 do indeed satisfy the required commutation relations (12.The limitation T ≤ 3 for baryonic states can be understood in terms of their being 2 composed of three T = 1 constituents (two of them pair to T = 1 or T = 0 and 2 the third adds to T = 1 to make T = 3 or T = 1 and to T = 0 to make T = 1 .25). Similarly. appears as the deuteron. identifying ¯ u ≡ u ∗ and d ≡ d ∗ .32)) that the three matrices − 1 τ ∗ must represent 2 the generators of SU(2)f in the 2∗ representation (i. 5 The overbar (u etc) here stands only for ‘anti-particle’. we have5 ¯ q ∗ = V∗ q ∗ or u ¯ ¯ d = exp(−iα · τ ∗ /2) u ¯ ¯ d (12. which is a loosely bound state of n and p.4. just as u and d themselves do. since the third component of isospin ∗ is here represented by −τ3 /2 = −τ3 /2.51). how does the (u. the d 3 2 and T3 = 1 . It is then straightforward to check that the three matrices −τ1 /2. it is ‘vectorially’ conserved according to the rules of combining angular-momentum-like quantum ¯ numbers.e. d) and anti-quark doublet (u.52) for the SU(2)f transformation law of the anti-quark doublet. T = 2 . the desired reversal in sign of the additively conserved eigenvalue does occur. Thus if a u-quark has B = 1 . d) do transform ¯ ¯ differently under SU(2)f transformations. as we have seen.28) and. appropriate at this point to explain how anti-particles are described in isospin terms. Thus. Referring ∗ ∗ ∗ to (12. however. relative to those of the corresponding particle. Also. there is nevertheless a sense in which the 2∗ and 2 representations are somehow the ‘same’: after all. d) doublet transform? ¯ ¯ The answer is that anti-particles are assigned to the complex conjugate of the representation to which the corresponding particles belong. the quantum ¯ conjugate ψ introduced in section 4. thus. In mathematical terms. Although the quark doublet (u. T3 = − 2 . T = 1 . d) doublet transforms as in (12. Meson states such as the pion are formed from a quark and an anti-quark and it is. therefore. while T3 is an additively conserved quantum number. we see that τ1 = τ1 . the anti-quarks d and u form the T3 = + 1 and ¯ 2 1 T3 = − 2 members of an SU(2)f doublet. 2 the magnitude of the isospin is not additively conserved: rather. and has nothing to do with the Dirac ¯ Copyright 2004 IOP Publishing Ltd . rather than as a compact B = 2 analogue of the n and p themselves. and the question arises: given that the (u. An anti-particle is characterized by having the signs of all its additively conserved quantum numbers reversed. for example. a u¯ 3 2 2 1 1 1 ¯ has B = − 1 . by 2 2 2 the usual angular momentum addition rules). τ2 = −τ2 and τ3 = τ3 . T3 = 1 . this means (compare (12. It is. Thus. which we will now call 2). the complex conjugate of the original two-dimensional representation.

57) transforms in exactly the same way as (u. d ≡ d ∗ . the corresponding wavefunction is 1 ¯ 1 ¯ √ (dd − (−u)u) = √ (dd + uu). d). d) states to form the isoscalar ω-meson. Note that. it is clear that we can choose UC proportional to τ2 . and the fact that τi−1 = τi . T3 = 0 state describing ¯ ¯ 2 2 1 ¯ 1 ¯ 0 is √ (dd + (−u)u) = √ (dd − uu). (12. the π ¯ ¯ 2 2 2 There is a very convenient alternative way of obtaining these wavefunctions. T3 = ± 1 describe them both. and the combination of (u. we see that the T = 0 ¯ ¯ can be written as u ∗ u + d ∗ d which is just q † q. because it means that we can use the familiar tables of (Clebsch–Gordan) angular momentum coupling coefficients for combining quark and anti-quark states ¯ together. the isospin ¯ ¯ 1 part of the wavefunction is √ (ψp ψn − ψn ψp ).53) becomes −1 UC (−τ ∗ )UC = τ (12.54) or −1 UC τ1UC = −τ1 −1 UC τ2 U C = τ2 −1 UC τ3UC = −τ3 . which we include here because it generalizes straightforwardly to SU(3): its advantage is that it avoids the use of the explicit C–G coupling coefficients and of their (more complicated) analogues in SU(3).1 particles in quantum mechanics given by √ (| ↑ 2 2 | ↓ − | ↓ | ↑ ).55) Bearing the commutation relations (12. In fact.53) This requirement is easier to disentangle if we consider infinitesimal transformations. This result is useful. as expected. (12. d) and (u. In the first. let us compare the T = 0 combination of the p and n states to form the (isoscalar) deuteron. (recall qq combination uu + dd ¯ ¯ Copyright 2004 IOP Publishing Ltd . But.56) to obtain a convenient unitary form. for which (12. provided we include the relative minus sign between the d and u ¯ ¯ components which has appeared in (12.57). in the second case. the two representations 2 2 are ‘unitarily equivalent’. This implies that the doublet UC u ¯ ¯ d = ¯ d −u ¯ (12. Similarly. corresponding to the S = 0 1 combination of two spin.numbers T = 1 . and set UC = iτ2 = 0 1 −1 0 (12. the d is in the T3 = + 1 position and the u is in the T3 = − 1 position. the T = 1. ¯ Bearing in mind the identifications u ≡ u ∗ . ¯ 2 2 As an application of these results.28) in mind. in that we can find a unitary matrix UC such that −1 UC exp(−iα · τ ∗ /2)UC = exp(iα · τ /2).

Equation (12. which means it has T = 0 (no multiplet partners). In this case.58) using V† V = 12 . (12. q → q = Vq . 3 is understood.48).63) ijk j q τk q i j k j vk which may also be written in ‘vector’ notation as v = v − × v. (12. 2. 3) (12.64) Equation (12. 2. Copyright 2004 IOP Publishing Ltd . Consider the ¯ three quantities vi defined by vi = q † τi q i = 1. 2.28) we know the commutator of two τ ’s. the three quantities vi (i = 1.62) where the sum on j = 1. We may also construct the T = 1 q –q states in a similar way. so that (12. so q † → q † = q † V† and q † q → q † q = q † V† Vq = q † q (12. as determined by the coefficients i j k (the ’s are just the parameters of the infinitesimal transformation). This is precisely what is needed for a set of quantities to form the basis for a representation.59) Under an infinitesimal SU(2)f transformation q = (12 + i · τ /2)q the three quantities vi transform to vi = q † (12 − i · τ /2)τi (12 + i · τ /2)q (12. Retaining only the first-order terms in gives (problem 12.63) states that.6) vi = vi + i j 2 q † (τi τ j − τ j τi )q (12. but we can also directly verify it as follows. But from (12. 3) transform into specific linear combinations of themselves. 2.that † means transpose and complex conjugate). Under an SU(2)f transformation. q † q is indeed an SU(2)f invariant. it is the T = 1 representation as we can guess from the multiplicity of three.62) becomes vi = vi + i = vi − = vi − j 2 q † 2i † i j k τk q (sum on k = 1. under an (infinitesimal) SU(2)f transformation.49) with T = 1.60) where we have used q † = q † (12 + i · τ /2)† and then τ † = τ . 3. thus.61) (12. together with (12.

however: rather.65) which is exactly the same as (12. the ¯ four combinations ψγ µ ψ behave as a 4-vector. we need to consider the linear combinations 1 2 (v1 ¯ v1 = ud + du ¯ + iv2 ) = ud ¯ ¯ − iv2 ) = du π− T3 = −1 T3 = +1 π +.48) j . in which strange particles are grouped with non-strange ones. therefore.59). when the transformations are extended to include Lorentz (velocity) transformations. we find explicitly that ¯ ¯ v2 = −i ud + i du ¯ v3 = uu − dd. inserting (12.63) that if φ is a twocomponent (Pauli) spinor.59)–(12.tell us how a T = 1 triplet should transform: namely. ψi (1) = (13 + i · T(1) )ik ψk = (13 + i = (δik + i = (δik + i = ψi(1) − (1) (sum on k = 1. be identified with the T3 = 0 member of the T = 1 triplet. 12.68) and 1 2 (v1 and The use of v1 ± iv2 which have the quantum numbers of the here is precisely analogous to the use of the ‘spherical basis’ wavefunctions x ± iy = r sin θ e±iφ for = 1 states in quantum mechanics. Returning to the physics of vi . As an aside. We are now ready to proceed to SU(3). Neither v1 nor v2 has a definite value of T3 . we can at once infer from (12. − i j ik )ψk (1) using the anti-symmetry of i j k . (12. The spatial part ψγ ψ generalizes this to four-component spinors (the µ = 0 part ψ † ψ is rotationally invariant.50) into (12. 2. i j k j ψk (12. φ † σ φ behaves as a vector (SO(3)-triplet) under real¯ space rotations.66) √ Apart from the normalization factor of 1/ 2.2 Flavour SU(3)f Larger hadronic multiplets also exist. recalling the close connection between SU(2) and SO(3). 2. Indeed.63). ¯ (12. 3) (1) (1) (sum on j = 1. having the quantum numbers of the π 0 . v3 may. Gell-Mann (1961) and Ne’eman (1961) (see also Gell-Mann Copyright 2004 IOP Publishing Ltd . analogous to q † q). 3) j T j )ik ψk (1) (1) j (T j )ik ψk (1) using (12.67) (12. under an infinitesimal transformation (with 13 the unit 3 × 3 matrix). as we have seen in volume 1. the reader may have been struck by the similarity between the ¯ Lorentz 4-vector combination of Dirac spinors given by ‘ψγ µ ψ’ and the present triplet combination ‘q † τ q’. rather than the ‘Cartesian’ ones x and y.

A finite SU(3) transformation on the quark triplet is then (cf (12. we regard the three quark wavefunctions u. .7)) and can be written as χ = η · λ/2 (12. . Like SU(2). along with the SU(3) structure constants f abc .32)) q = exp(iα · λ/2)q Copyright 2004 IOP Publishing Ltd (12. An infinitesimal SU(3) matrix has the form Winfl = 13 + iχ (12. In more physical terms. λb /2] = i f abc λc /2 (12. are given in appendix M. η8 ) and the λ’s are eight matrices generalizing the τ matrices of (12. We can. an overall phase has been extracted). b and c run from 1 to 8.72) where η = (η1 . the generalization of the twocomponent isospinors of SU(2). in fact.28) for SU(2)): [λa /2.70) and W is a 3 × 3 unitary matrix of determinant 1 (again. The representation provided by this triplet of states is called the ‘fundamental’ representation of SU(3)f (just as the isospinor representation is the fundamental one of SU(2)f ).74) .25).71) where χ is a 3 × 3 traceless Hermitian matrix. we follow exactly the same steps as in the SU(2) case. Such a matrix involves eight independent parameters (problem (12. The general group-theoretic analysis of SU(3) is quite complicated but is fortunately not necessary for the physical applications we require.69) (12. . They are the generators of SU(3) in the three-dimensional fundamental representation and their commutation relations define the algebra of SU(3) (compare (12. . develop all the results needed by mimicking the steps followed for SU(2). SU(3) is a group whose elements are matrices—in this case.73) where a. d and s as being approximately degenerate and we consider unitary 3 × 3 transformations among them via q = Wq where q now stands for the three-component column vector   u q = d  s (12. The λ-matrices (often called the Gell-Mann matrices).and Ne’eman (1964)) were the first to propose SU(3)f as the correct generalization of isospin SU(2)f to include strangeness. To determine the general form of an SU(3) matrix W. unitary 3 × 3 ones. Such matrices obviously act on three-component column vectors. We start by finding the general form of an SU(3) matrix. of unit determinant.

Here it is more involved—for a start. section M. As in the case of SU(2)f .35). there are eight matrices λa .4(v).75) as can be verified from the explicit forms of the λ-matrices given in appendix M. which is diagonalizable simultaneously with τ3 in the case of SU(2)? Indeed. In just the same way.4(v). while the u and d quark masses are of order 5–10 MeV. but because of the commutation relations (12. Such a symmetry implies that the eigenvalues of the λ’s are constants of the motion. If we denote the fundamental three-dimensional representation accommodating the quarks by 3. of course. where B is the baryon number ( 1 for each 3 quark) and the strangeness values are S(u) = S(d) = 0. A glance at appendix M. it is still possible to regard this as relatively small on a typical hadronic mass scale of ∼ 1 GeV. We may now consider other representations of SU(3)f . This means physically that for SU(3) there are two additively conserved quantum numbers. d and s were degenerate in mass. This happened for SU(2) too. (cf (12. we find that the physically required eigenvalues imply that the matrix representing the 1 hypercharge operator is Y (3) = √ λ8 . Thus. in this fundamental (three-dimensional) representation. by the eigenvalues of the relevant complete set of mutually commuting operators. Actually. which in this case are of course the third component of hadronic isospin (since λ3 is simply τ3 bordered by zeros). so we may proceed to explore the physical consequences of this (approximate) SU(3)f flavour symmetry. Defining the hadronic hyperchange Y by Y = B + S .73) only a subset of these operators has simultaneous eigenstates. λ3 and λ8 . the s-quark mass is much greater. section M. Identifying T3(3) = 1 λ3 then gives the 2 Gell-Mann–Nishijima relation Q = T3 + Y/2 for the quark charges in units of |e|. now with eight ‘phase angles’. as shown in figure 12. one finds that λ2 is also proportional to the unit 3 × 3 matrix: 8 3 (λ/2)2 = i=1 (λa /2)2 = 4 13 3 (12. τ 2 is a multiple of the 2 × 2 unit matrix. denoted by the symbol 3.3. and a quantity related to strangeness. we may characterize the ‘fundamental triplet’ (12. So λ3 and λ8 are analogous to τ3 : what about the analogue of τ 2 . of order 150 MeV.which also has the ‘generalized phase transformation’ character of (12. W is parametrized as W = exp(iα · λ/2).70) by the eigenvalues of (λ/2)2 . but there the very close analogy with SO(3) told us how the states were to be correctly classified. S(s) = −1. namely λ3 and λ8 . Nevertheless. shows that two of the λ’s are diagonal (in the chosen representation). Thus. then the antiquarks have quantum numbers appropriate to the ‘complex conjugate’ of this Copyright 2004 IOP Publishing Ltd . The first important one is that to which the anti-quarks belong.41)). The conventional way of representing this pictorially is to plot the states in a Y –T3 diagram. this is not the case: in particular. exact symmetry under SU(3)f would imply that the three states u.

figure 12.76) means that the eight quantities −λ∗ /2 represent the SU(3) generators in this 3∗ representation. denoted by 3∗ just as in the SU(2) case. For our present purposes. one quickly sees that λ3 ∗ and λ8 are real. Referring to appendix M. d ≡ d ∗ and s ≡ s ∗ then transform by ¯   u ¯ ¯ ¯ ¯ (12.Figure 12.74).3.3. as expected for anti-particles. (12. an important one is the eightdimensional (‘octet’) representation which appears when one combines the 3∗ Copyright 2004 IOP Publishing Ltd . d and s states may also be plotted on the ¯ ¯ ¯ Y –T3 diagram. so that the eigenvalues of the physical observables T3(3 ) = −λ3 /2 ∗ 1 and Y (3 ) = − √ λ8 /2 (in this representation) are reversed relative to those in the 3 3. representation.76) q =  d  = W∗ q = exp(−iα · λ∗ /2)q ¯ s ¯ instead of by (12. Larger SU(3)f representations can be created by combining quarks and antiquarks.4(v). The Y –T3 quantum numbers of the fundamental triplet 3 of quarks and of the anti-triplet 3∗ of anti-quarks.3. The u. as shown. where it is clear that the extra quantum number Y distinguishes the two representations. As for the 2∗ representation of SU(2). This follows immediately from figure 12. section M. as in SU(2)f . The q wavefunctions ¯ ¯ ¯ identified as u ≡ u ∗ . Here is already one important difference between SU(3) and SU(2): the fundamental SU(3) representation 3 and its complex conjugate 3∗ are not equivalent.

in a way which is very analogous to the three-dimensional (‘triplet’) representation obtained by combining the 2∗ and 2 representations of SU(2).81) w = exp(iα · G(8) )w where w is an eight-component column vector   w1  w2    w= . As in the SU(2) case.63). . .and 3 representations. 2.80) shows that. G b ] = i f abc G c . as determined by the coefficients f abc (the η’s are just the parameters of the infinitesimal transformation).62)). . equation (12. . under an SU(3)f transformation. . . and where (cf (12.  . 8. a = 1. G 2 . This is. the G(8) ’s satisfy (cf (12. . which is invariant under q → q = Wq since W† W = 13 .78) where the sum on b = 1–8 is understood. ¯ Consider first the quantity uu + dd + s s.77) wa = q † λa q Under an infinitesimal SU(3)f transformation (compare (12.   . .80) wa = wa − f abc ηb wc which may usefully be compared with (12.49) for SU(2))f ) the quantities (8) (8) (8) G(8) = (G 1 . .59). again. So this combination is an SU(3) singlet. the eight quantities wa (a = 1.83) Copyright 2004 IOP Publishing Ltd . (8) (12. 2. we can ‘exponentiate’ (12. (12. an eight-dimensional representation of SU(3)f .79) 2 or (12.73) for the commutator of two λ’s we find that ηb wa = wa + i q † · 2i f abc λc q (12.82) such that wa = q † λa q.61) and (12. 8) transform with specific linear combinations of themselves. wa → wa = q † (13 − iη · λ/2)λa (13 + iη · λ/2)q ηb ≈ q † λa q + i q † (λa λb − λb λa )q 2 (12. G 8 ) are 8 × 8 matrices. . Just as in the SU(2)f triplet case. w8 (12. this can ¯ ¯ be written equivalently as q † q. acting on the eight-component vector w and forming an eight-dimensional representation of the algebra of SU(3): that is to say. . .80) to obtain (12. precisely what is needed for a set of quantities to form the basis for a representation—in this case. Using (12. For a finite SU(3)f transformation. The octet coupling is formed by a straightforward generalization of the SU(2) triplet coupling q † τ q of (12.73)) (8) (8) [G a .

K∗ ¯ and K∗ ) are all much closer in mass. where it is also verified that the matrices specified by (12. while the orbital angular momentum is still zero. seem to be ‘anomalously light’ for Copyright 2004 IOP Publishing Ltd .4. which is orthogonal to ¯ s 0 state and the SU(3) singlet. while the η ¯ (in this simple model) would be w8 ∼ (uu + dd − 2¯ s).8. The Y –T3 quantum numbers of the pseudoscalar meson octet. The ¯ pion. averaging around 800 MeV: in these states the qq spins add to S = 1. for example. a similar octet of vector (J P 1− ) mesons (the ω.83). Similarly the K0 . the eight states generated by the combinations q † λa q are not necessarily the ones with the physically desired quantum numbers. and to a much lesser extent the kaons. as shown on the Y –T diagram of figure 12. By contrast. As in the SU(2)f case. To get the π + .4.84) do obey the commutation relations (12. Similarly. ρ. we again need to form (w1 ± iw2 )/2. We say ‘octet of π-partners’ but a reader knowing the masses of these particles might well query why we should feel justified in regarding them as (even approximately) degenerate.81) with (12. so the K± states are ¯ ¯ ¯ ¯ ¯ w4 ∓ iw5 . The actual form of the G a matrices is given by comparing the infinitesimal version of (12.80): ( G a8) bc (8) = −i f abc (12. K0 states are w6 − iw7 and w6 + iw7 . w4 produces us + s u and w5 the combination −i us + i s u.Figure 12. In this way all the pseudoscalar octet of both the π f π-partners has been identified.84) as may be checked in problem 12.

1 and 7.3 Non-Abelian global symmetries in Lagrangian quantum field theory 12.89) q = exp(−iα · τ /2)q ˆ Copyright 2004 IOP Publishing Ltd . Lagrangian quantum field theory is a formalism which is especially well adapted for the description of symmetries. This will also prepare the way for the (local) gauge case. ˆ ˆ (12. as we will see.84) and (12. the associated Hamiltonian À—are invariant under the global U(1) transformation q = eiα q ˆ ˆ (12. since gauge quanta always belong to the adjoint representation of the gauged group (for example.87) ˆ ˆ In this form it is obvious that Ä—and. to be considered in the following chapter. In both cases. 12.1)) which is associated with baryon number conservation. Instead. we shall now give a few examples showing how global flavour symmetry is very easily built into a Lagrangian.85) describing two free fermions ‘u’ and ‘d’ of equal mass m. for example.50). It is also invariant under global SU(2)f transformations acting in the flavour u–d space (cf (12. generalizing in a simple way the global U(1) symmetries considered in sections 7. Consider. Without going into any elaborate general theory.48).2. the Lagrangian ¯ / ˆ ˆ ˆ Ä = u(i∂ − m)u + d(i∂ − m)d ˆ ˆ ¯ / (12. Further flavours c. section M.3. Let us introduce ˆ q= ˆ ˆ so that Ä can be compactly written as ˆ ¯ / Ä = q(i∂ − m)q.5). or regular. we now move on to consider the field-theoretic formulation of global SU(2)f and SU(3)f .88) u ˆ ˆ d (12.32)): ˆ (12. As in (12. ˆ ˆ ˆ we are using the convenient shorthand ψu = u and ψd = d. There is a deep similarity between (12. the eight gluons in SU(3)c ). Such a representation is always possible for a Lie group and is called the adjoint.1 SU(2)f and SU(3)f As may already have begun to be apparent in chapter 7.some reason: we shall learn the likely explanation for this in chapter 18. with the overbar now meaning the Dirac conjugate for the four-component spinor fields. hence. These representations are of particular importance in gauge theories. b and t of course exist but the mass differences are now so large that it is generally not useful to think about higher flavour groups such as SU(4)f etc. representation (see appendix M.86) (cf (12. a representation has been found in which the matrix element of a generator is minus the corresponding structure constant.

89)? Let us recall the discussion of the simpler U(1) cases studied in sections 7.91) are then ˆ ˆˆ q = U ( 2 ) q U ( 2 )† ˆ 1 1 (12.89).95) Since the single U(1) parameter is now replaced by the three parameters ˆ = ( 1 . (12.96) corresponding to the three independent infinitesimal SU(2) transformations. we have ˆ ˆ φ ≈ (1 − i )φ so that (12. T3 2 ) 1 1 1 1 (12. the analogue ˆ ˆ ˆ of (12.98) Copyright 2004 IOP Publishing Ltd .35) but note that here Nφ is a quantum field operator.(for the change in sign with respect to (12.92) ˆ ˆ U = exp(iα Nφ ). hence. we require ˆ ˆ ˆ [ Nφ .1 and 7. ˆ ˆ (12.1. Considering the complex scalar field of section 7. The generalizations of (12.27) and ˆ (12. T2 2 .91) (12. not a matrix).2 in the U(1) case).97) and ˆ ˆ( ) U ( 2 ) = exp(iα · T 2 ) 1 1 (12. 3 ). 2 .89). Insofar as Nφ determines the form of an infinitesimal version of ˆ the unitary transformation operator U .90) becomes ˆ ˆ ˆ ˆ ˆ ˆ ˆ (1 − i )φ = (1 + i Nφ )φ(1 − i Nφ ) ≈ φ + i [ Nφ . In (12. Consider now the SU(2)f transformation (12.31).89) was just φ → φ = e−iα φ and the conserved quantity was the ˆ Hermitian operator Nφ which appeared in the exponent of the unitary operator ˆ ˆ ˆ U that effected the transformation φ → φ via ˆ ˆˆ ˆ φ = U φU † with For an infinitesimal α.94) ˆ for consistency. φ] = −φ (12.1 and 7. in the infinitesimal case: q = (1 − i · τ /2)q. ˆ What are the conserved quantities associated with the invariance of Ä under (12.93) ˆ ˆ U ≈ 1 + i Nφ (12. which we call ˆ( ) ˆ( ) ˆ( ) ˆ( ) T 2 = (T1 2 . compare sections 7.90) (12. it seems reasonable to call it the generator of these global U(1) transformations (compare the discussion after (12. φ]. the three parameters α are independent of x.90) and (12. we shall need three analogues of Nφ .2.

follows from the fact that they obey the SU(2) algebra (problem 12.101) reduces to the sum of the u and d number operators. 1 ˆ( ) It is possible to verify that these T 2 ’s do indeed commute with the ˆ Hamiltonian H : 1 1 ˆ( ) ˆ ˆ( ) (12. Tj 2 ] = i 1 1 ˆ(2) i j k Tk .35)). Note that if ‘τ /2’ is replaced by 1.105) Copyright 2004 IOP Publishing Ltd .11): 2 ˆ [Ti 1 (2) 1 ˆ( ) .94). and equating coefficients of on both sides.101) T 2 = q † (τ /2)q d3 x as can be checked (problem 12.3.100) expresses a very specific 1 ˆ( ) commutation property of the operators T 2 .9) 1 ˆ( ) ˆ [ T 2 .100) which is the analogue of (12. which turns out to be satisfied by the expression 1 ˆ( ) ˆ ˆ (12.99) reduces to (problem 12. It would 1 ˆ( ) seem reasonable in this case too to regard the T 2 as providing a field-theoretic representation of the generators of SU(2)f . (12. q] = −(τ /2)q ˆ (12.99) to first order in .ˆ( ) ˆ 1 where the T 2 are Hermitian. appropriate to the case T = 1 .1. H ] = 0 ˆ( ) so that their eigenvalues are conserved. we easily find ˆ( ) T+2 = 1 u † d d3 x ˆ ˆ (12.104) For example. Equation (12. This means that the three operators T 2 themselves belong to a T = 1 triplet of SU(2)f . 1 1 1 (12.99) ˆ( ) using the Hermiticity of the T 2 ’s. as already suggested. Expanding the right-hand side of (12.97) and (12. so that U ( 2 ) is unitary (cf (12. (12. as required for the one-parameter U(1) case. That the T 2 ’s are. (12.103) For many purposes it is more useful to consider the raising and lowering operators ˆ( ) ˆ( ) ˆ( ) T±2 = (T1 2 ± iT2 2 ). We shall derive (12.10) from the anti-commutation relations of the fermionic fields in q . (12. The ‘q † τ q ’ ˆ ˆ combination is precisely the field-theoretic version of the q † τ q coupling we 1 ˆ( ) discussed in section 12. an interpretation we shall shortly confirm. In the infinitesimal case.98) become ˆ (1 − i · τ /2)q = (1 + i · T ˆ 1 1 (2) 1 ˆ )q(1 − i · T ˆ (1) 2 ) (12. a field-theoretic representation of the generators of SU(2).101) from Noether’s theorem in a little ˆ while.102) d T 2 /dt = −i[ T 2 .

108) The obvious generalization appropriate to (12. ˆ 2 1 (12. ˆ ˆ ∂Ä ∂Ä ˆ ˆ ˆ 0 = δÄ = δ ψr + ∂ µ (δ ψr ). in either ¯ (1) ˆ case raising the T 2 eigenvalue by +1.8).¯ which destroys a d quark and creates a u. since 3 1 1 ˆ( ) T3 2 = 2 ˆ ˆ (u † u − d † d) d3 x ˆ ˆ (12. ˆ ∂ ψr ∂(∂ µ ψˆ ) r Copyright 2004 IOP Publishing Ltd (12. these operators ¯ 2 2 certainly ‘do the job’ expected of field-theoretic isospin operators. or destroys a u and creates a d. ˆ( ) T 2 = 1 1 (12. considering now the fermionic example of section 7. Equation (12.6)–(7.112) . (12. via the (suppressed) dependence of the q-fields on x. d) system: it transforms as a 4-vector under Lorentz transformations and as a T = 1 triplet under SU(2)f transformations.109) ˆµ ˆ ( )µ Note that both Nψ and T 2 are of course functions of the spacetime coordinate x. we could go further and associate the conserved operator Nψ with µ ˆ a conserved current Nψ : ˆ Nψ = where ˆ0 Nψ d3 x µ ¯ ˆµ ˆ ˆ Nψ = ψγ µ ψ (12. in equations (7. Thus.101) is ˆ ( )0 T 2 d3 x 1 ¯ τˆ ˆ ( )µ T 2 = qγ µ q.111) for some set of numerical coefficients Trs .2 ˆ for variety. in the U(1) case.110) ˆ ( )µ Thus T 2 is a conserved isospin current operator appropriate to the T = 1 2 (u.107) ˆ ∂µ Nψ = 0. Suppose the Lagrangian involves a set of ˆ fields ψr (they could be bosons or fermions) and suppose that it is invariant under the infinitesimal transformation ˆ ˆ δ ψr = −i Trs ψs (12. Clearly there should be some general formalism for dealing with all this more efficiently and it is provided by a generalization of the steps followed. Indeed one can verify ˆ from the equations of motion that ˆ ∂µ T 1 1 ( 2 )µ = 0. In the U(1) case. in this isospin1 2 case.106) ¯ which counts + 1 for each u (or d) and − 1 for each d (or u).111) generalizes (7. ˆ Then since Ä is invariant under this change.5).

for example 1 ˆ ∂ Ä τ1 ¯ τ1 ˆ ˆ(2) q = qγµ q ˆ ˆ T1µ = −i µ q) 2 ∂(∂ ˆ 2 (12.But ˆ ∂Ä = ∂µ ˆ ∂ ψr ˆ ∂Ä ˆ ∂(∂ µ ψr ) (12. there are now three ’s and so three Trs ’s.117) (12. ˆ ∂(∂ µ ψr ) Let us try this out on (12. there follows the conservation of an associated symmetry current.109). disregarding the irrelevant constant small parameter . Thus. ˆ ˆ As we know already. Suppose we have three spin-0 fields all with the same mass.115). the conserved current is ˆ ∂Ä ˆ ˆ (12.119) ˆ ˆ It is obvious that Ä is invariant under an arbitrary rotation of the three φ’s among ˆ themselves. this time involving spin-0 fields.1. and take ˆ ˆ2 ˆ2 ˆ2 ˆ ˆ ˆ ˆ ˆ ˆ Ä = 1 ∂µ φ1 ∂ µ φ1 + 1 ∂µ φ2 ∂ µ φ2 + 1 ∂µ φ3 ∂ µ φ3 − 1 m 2 (φ1 + φ2 + φ3 ). From the invariance of the Lagrangian under the transformation (12.111)) to ‘current’ conservation laws (of the form (12.113) from the equations of motion. 2 (τ3 )rs . 2 (τ2 )rs . and it constructs the actual currents for us. An infinitesimal such rotation is (cf (12.118) and similarly for the other τ ’s and so we recover (12. 2 2 2 2 (12. ∂µ ˆ ∂Ä ˆ δ ψr µψ ) ∂(∂ ˆ r =0 (12.114) which is precisely a current conservation law of the form ˆ ∂ µ jµ = 0.116) jµ = −i Trs ψs . This theorem is of fundamental significance as it tells us how to relate symmetries (under transformations of the general form (12. generalizing the ‘rotation about the three-axis’ considered for the φ1 – ˆ 2 system of section 7.115) Indeed. (12. This is the quantum field theory version of Noether’s theorem (Noether 1918). For each one we have a current.64). the dynamics is generated from a symmetry. in the sense that (as we have seen in the local U(1) of electromagnetism) the symmetry currents are the dynamical currents that drive the equations for the force field. In gauge theories. Let us look at another example.87) with δ q = (−i · τ /2)q. the symmetries of the Lagrangian are basic to gauge field theories. Hence. and noting φ Copyright 2004 IOP Publishing Ltd .117). namely 1 1 1 2 (τ1 )rs .

and for an isospin T -multiplet of fermions we shall have ¯ ˆ ˆ ψ (T ) γ µ T(T ) ψ (T ) . of course.47). In all cases. we shall have the form ˆ ˆ ˆ ˆ i(φ (T )† T(T ) ∂ µ φ (T ) ) − (∂ µ φ (T ) )† T(T ) φ (T ) ) (12. ˆ (1)µ = φ2 ∂ µ φ3 − φ3 ∂ µ φ2 .126) where we have put the † to allow for possibly complex fields. Copyright 2004 IOP Publishing Ltd .120) (12. ˆ ˆ ψ (T ) = exp(−iα · T(T ) )ψ (T ) (12.127) ˆ ˆ where. the integral over all space of the µ = 0 component of these currents results in a triplet of isospin operators obeying the SU(2) algebra (12. (12. as in (12.103). For an isospin T -multiplet of bosons.e.48).123) Cyclic permutations give us the other components which can be summarized as ˆ ˆ ˆ ˆ ˆ (1)µ = i(φ (1)tr T(1) ∂ µ φ (1) − (∂ µ φ (1) )tr T(1) φ (1) ) T where we have written ˆ φ (1)  ˆ φ1 ˆ =  φ2  ˆ φ3  (12.128) ˆ and similarly for φ (T ) . three conserved T operators again and three µ (1) (1)µ ˆ ˆ ˆ respectively. which we call T and T T = 1 isospin case. The a = 1 component of the conserved current in this case is.125) and tr denotes transpose. The general form of such SU(2) currents should now be clear. Equation (8.122) ars ˆ as in (12. from (12. i.76) has the form expected of a bosonic spin0 current but with the matrices T(1) appearing. in each case. There are.121) (12. since we are now dealing with a T ’s.124) (12.the sign change in the field theory case) ˆ ˆ ˆ φ =φ+ ×φ which implies with (1) ˆ ˆ δ φr = −i a Tars φs (1) Tars = −i (12. where T(T ) are the 2T + 1 × 2T + 1 matrices representing the generators of SU(2)f in this representation. the (2T + 1) components of φ or ψ transforms as a T multiplet under SU(2). ˆ ˆ ˆ ˆ T1 (12.116). appropriate to the T = 1 (triplet) representation of SU(2)f .

8 (12. . 2. Ä is clearly invariant under global SU(3)f transformations q = exp(−iα · λ/2)q ˆ ˆ (12. . .1 ) 2 ˆ ˆ field ψ and a T = 1 isotriplet (Lorentz scalar) φ.133) ˆ G a = qγ µ q 2 (note that there are eight of them) and the associated conserved ‘charge operators’ are λa ˆ (q) ˆ (q)0 G a = G a d3 x = q † q ˆ a = 1.132) as well as the usual global U(1) transformation associated with quark number conservation. . . 2. For example.131) ˆ describes free u. allowing for the pseudoscalar nature of the pions (we shall see in the following section that ¯ ˆ ˆ ψγ5 ψ is a pseudoscalar. which has the form ¯ τˆ ˆ g qγ µ q · W µ ˆ 2 (12. .134) ˆ 2 which obey the SU(3) commutation relations ˆ (q) ˆ (q) ˆ (q) [G a .130) with q now extended to ˆ  u ˆ ˆ q = d  ˆ s ˆ  (12. . An effective interaction ¯ ˆ ˆ ˆ between pions and nucleons could take the form gπ ψτ γ5 ψ · φ. so the product is a true scalar as is required for a parityconserving strong interaction). For example. G b ] = i f abc G c . In these examples the ‘vector’ analogy for the T = 1 states allows us to see that the ‘dot product’ will be invariant. The associated Noether currents are (in somewhat informal notation) ˆ (q)µ ¯ λa ˆ a = 1. the Lagrangian ˆ ¯ / Ä = q(i∂ − m)q ˆ ˆ (12. d and s quarks of equal mass m. Copyright 2004 IOP Publishing Ltd (12. This is just the SU(2) dot product of the symmetry current (12. 8 (12. A similar ˆ 1 dot product occurs in the interaction between the isospinor ψ ( 2 ) and the weak ˆ SU(2) gauge field W µ .The cases considered so far have all been free field theories but SU(2)¯ ˆ ˆ invariant interactions can be easily formed.129) as will be discussed in the following chapter. All of the foregoing can be generalized straightforwardly to SU(3)f . the interaction g1 ψτ ψ · 1 ˆ φ describes SU(2)-invariant interactions between a T = 2 isospinor (spin.135) .109) and the gauge field triplet. both of which are in the adjoint (T = 1) representation of SU(2). .

tˆ. this expectation leads to a puzzle. Since we are interested in the ‘small mass’ regime.15): σ 0 0 1 α= β= . 8) gluon fields.138) 0 −σ 1 0 Writing the general Dirac spinor ω as ω= we have (as in (4. and possibly also of the larger symmetry holding when m u ≈ m d ≈ m s ≈ 0.136) In (12. 12. we shall introduce the idea of chiral symmetry.140) (12.97) of the Dirac matrices.136). ˆ 2 (12.SU(3)-invariant interactions can also be formed. however. .139) (12. it is sensible to use the representations (4. c. we may certainly expect some physical signs of the symmetry associated with m u ≈ m d ≈ 0.98).137) ˆ ˆµ where ‘ fˆ’ is any of the six quark flavour fields u. d. A particularly important one is the ‘SU(3) dot-product’ of two octets (the analogues of the SU(2) triplets). (12. the resolution of which will have to be postponed until the concept of ‘spontaneous symmetry breaking’ has been developed in part 7. Copyright 2004 IOP Publishing Ltd φ χ (12. (12. which is an exact symmetry for fermions in the limit in which their masses may be neglected. 2. We have seen that the u and d quarks have indeed very small masses (≤ 10 MeV) on hadronic scales and even the s quark (∼150 MeV) is relatively small. b and Aa are the eight ˆ ˆ ˆ ˆ (a = 1. in which the momentum part of the Dirac Hamiltonian is ‘diagonal’ and the mass appears as an ‘off-diagonal’ coupling (compare problem 4.99)) Eφ = σ · pφ + mχ Eχ = − σ · pχ + mφ. Thus.3. which arises in the quark–gluon vertex of QCD (see chapters 13 and 14): −igs f λa ˆ ¯ ˆ µ q f γ µ qf A a . Once again. . . s .136) has the form ‘symmetry current×gauge field’ characteristic of all gauge interactions. We begin with the simplest case of just one fermion. . qf stands for the SU(3)c colour triplet ˆ   fˆr qf =  fˆb  ˆ fˆg (12.2 Chiral symmetry As our final example of a global non-Abelian symmetry. (4. As we shall see.141) .

reflecting the fact that no Lorentz transformation can reverse the helicity of a massless particle. the two-component spinors φ. the two helicity spinors are decoupled. Also in this limit. for these states chirality equals helicity. Similarly. With the choice (12.148) which is easily seen to commute with γ5 . A physical anti-fermion of energy E and momentum p Copyright 2004 IOP Publishing Ltd . as we shall see. Thus. γ µ } = 0. As m → 0. the ‘u’-spinor 0 0 a fermion with negative helicity is described by which has γ5 = −1. We have to be more careful for anti-fermions.143) can easily be verified.142) The matrix γ5 plays a prominent role in chiral symmetry. in this representation. we have E → | p| and (12. and χ negative helicity (cf ˜ (4. the Dirac energy operator is α· p= σ·p 0 0 −σ · p (12. the massless states may equivalently be classified by the eigenvalues of γ5 .15). χ ˜ Thus. which are clearly ±1 since 2 γ5 = I .142) in the present case.We now introduce the matrix γ5 defined.141) reduce to ˜ ˜ (σ · p/| p|)φ = φ (σ · p/| p|)χ = − χ ˜ ˜ (12.140) and (12.68)). as γ5 = 1 0 0 −1 . however. Its use here stems from the fact that.138) for α and β.147) ˜ so that the limiting spinor φ has positive helicity.145) which reduces to (12.67) and (4. In this m → 0 limit.146) (12.144) (12. ‘Chirality’ means ‘handedness’ from the Greek word for hand. It is described by ˜ φ which is an eigenstate of γ5 with eigenvalue +1. γ5 is defined by γ5 = iγ 0 γ 1 γ 2 γ 3 (12. in the limit m → 0. (12. In a general representation. χ iρ.143) and (12. χ become helicity eigenstates (cf problem 4. Consider then a massless fermion with positive helicity. having definite ‘handedness’. Its defining property is that it anti-commutes with the γ µ matrices: {γ5 . we have γ = 0 σ −σ 0 γ0 = β = 0 1 1 0 (12.

therefore. with p along the z-axis say.151) 1 + γ5 2 ω+ 1 − γ5 2 ω ≡ PR ω + PL ω ≡ ωR + ωL (12.140) and (12. the chirality of φ and χ is a good quantum number (γ5 ˜ commuting with the energy operator) and we may say that ‘chirality is conserved’ in this massless limit. ‘L’ is used for the chirality eigenvalue. however. The corresponding v-spinor 2 must then have sz = − 1 (see section 4. ˜ In the massless limit. So the v-spinor for this anti-fermion of positive helicity is χ ˜ which has γ5 = −1. The Dirac Lagrangian for a single massless fermion is ¯ /ˆ ˆ ˆ Ä0 = ψi∂ ψ. − p as for E.147). but with m = 0 in (12. for fermions the γ5 eigenvalue is equal to the helicity and for anti-fermions it is equal to minus the helicity. γ5 } = 0. Copyright 2004 IOP Publishing Ltd (12. p. It is usual to introduce operators PR.150) Then clearly γ5 ωR = ωR and γ5 ωL = −ωL . (12. (12. The invariance is easily verified: using {γ 0 .3) and described by a ‘v’-spinor corresponding to −E and − p.5.141) the equations for φ and χ remain the same for −E.154) . We now reformulate this in field-theoretic terms. the notation ‘R’. (12. slightly confusingly. If the physical anti-particle has positive helicity. In summary. However. we may therefore regard the mass terms as ‘coupling the states of different chirality’.L = ((1 ± γ5 )/2) which ‘project’ out states of definite chirality from ω: ω= so that ωR = 1 0 0 0 φ χ = φ 0 ωL = 0 χ . we have ¯ ¯ ˆ ˆ ˆ ˆ ˆ ψ = ψ † γ 0 = ψ † eiβγ5 γ 0 = ψ † γ 0 e−iβγ5 = ψe−iβγ5 . Consider the spin.152) This is invariant not only under the now familiar global U(1) transformation ˆ ˆ ˆ ψ → ψ = e−iα ψ but also under the ‘global chiral U(1)’ transformation ˆ ˆ ˆ ψ → ψ = e−iβγ5 ψ (12.141). It is the γ5 eigenvalue that is called the ‘chirality’. then sz = + 1 .153) where β is an arbitrary (x-independent) real parameter. be of χ type ˜ 2 0 (12.149) −χ χ Referring to (12. the massive spinor ω is clearly not an eigenstate of chirality: φ φ γ5 ω = =λ .140) and (12.

Hence. χ → φ. The spatial parts ψγ 5 ˆ vector rather than a normal (polar) vector under parity. (12.157) We denote this chiral U(1) by U(1)5 . ˆ ˆ (12. as we have seen. remaining the same in the ˆ the non-chiral operator ψ parity-transformed frame. the appearance of the γ5 in the current operator j5 = ¯ ˆ ˆ ˆ ˆ ψγ µ γ5 ψ affects its parity properties: for example. this will ensure (as we saw at the end of section 4.4. ˆ ˆ ˆ with ψP = γ 0 ψ. rather than like r. χP = φ. Copyright 2004 IOP Publishing Ltd . namely parity. The corresponding Noether current is ¯ ˆ ˆ ˆµ j5 = ψγ µ γ5 ψ (12. In contrast. All this carries over to the field-theoretic case.141) to be covariant under parity. Under the parity transformation p → − p and thus.and then using {γ µ . for (12. the µ = 0 component ψ † γ5 ψ ¯ γ ψ behave as an axial ˆ is a pseudoscalar. ˆ ˆ † ψd3 x is a (true) scalar.4) that the Dirac equation in the parity-transformed frame will be consistent with the one in the original frame.156) and the spatial integral of its µ = 0 component is the (conserved) chirality operator ˆ Q5 = ˆ ˆ ψ † γ5 ψd3 x = ˆ ˆ (φ † φ − χ † χ ) d3 x. in that they do not reverse sign under parity. Consider. Q 5 is a ‘pseudoscalar’ operator. this is equivalent to saying that the spinor ωP in the parity-transformed frame is given by ωP = γ 0 ω. ¯ ¯ ˆ ˆ ˆ ˆ ψ γ µ ∂µ ψ = ψe−iβγ5 γ µ ∂µ e−iβγ5 ψ ¯ ˆ ˆ = ψγ µ eiβγ5 ∂ e−iβγ5 ψ µ ¯ ˆ ˆ = ψγ µ ∂µ ψ as required. they behave like r × p for example.158) which implies φP = χ. ˆ It is interesting to compare the form of Q 5 with that of the corresponding ˆ ˆ operator ψ † ψd3 x in the non-chiral case (cf (7.157). γ5 } = 0 and (γ 0 )2 = 1. the operator Q 5 in the parity-transformed frame: ˆ ( Q 5 )P = ˆ† ψP γ5 ψP d3 x = ˆ ˆ ψ † γ 0 γ5 γ 0 ψ d 3 x = − ˆ ˆ ˆ ψ † γ5 ψ d 3 x = − Q 5 (12. In the representation (12. that is. The difference has to do with their behaviour under a transformation briefly considered in section 4. γ5 } = 0.138).48)). We can ˆ also see this directly from (12.159) ˆ where we used {γ 0 . meaning that it changes sign in the parity-transformed frame. we require φ → χ. then. making the interchange φ ↔ χ. ˆµ In a similar way.155) (12.140) and (12.

γ ) symmetry ˆ ∂ 5 of the kinetic part will be preserved when the gauge interaction is included.g. (12.165) with quark number conservation.(Polar and axial vectors will be discussed again in section 20.164) as m → 0. such that states of definite parity |+ . We have Hamiltonian H ˆ ˆ ˆ ˆ ˆ H Q 5 |ψ = Q 5 H |ψ = E Q 5 |ψ .1. the operator Q 5 changes the parity of any state on which it acts. this is just where the ˆ analysis begins to get interesting.87) becomes ˆ ¯ /ˆ Ä = q i∂ q ˆ (12. which is invariant under the γ5 -version of (12.89). A very important physical consequence now follows from the fact that (in ˆ this simple m = 0 model) Q 5 is a symmetry operator commuting with the ˆ .159). Äint is also invariant under the chiral transformation (12. We can see this formally by introducing the (unitary) parity ˆ operator P in field theory. following the normal Equation (12.161) ˆ ˆ showing that Q 5 |+ is an eigenstate of P with the opposite eigenvalue. Of course. Thus chirality remains a ‘good symmetry’ in the presence of a U(1)-gauge interaction. ˆ 6 (12.2. The Lagrangian (12. We have ˆ ˆ ˆ ˆ ˆ ˆ P Q 5 |+ = ( P Q 5 P −1 ) P|+ ˆ = − Q 5 |+ (12.160) ˆ ˆ ˆ = − Q 5 .) Such a current is referred to generally as an ‘axial vector current’. ˆ As a consequence of (12. The generalization of this to the more physical m u ≈ m d ≈ 0 case is quite straightforward.163) is exactly the ¯/ˆ ˆ same as that of the free kinetic term ψ ∂ ψ : the ‘covariant derivative’ prescription µ → D µ = ∂ µ + iq A µ automatically means that any ‘Dirac’ (e. Ä ˆ 0 is also invariant under q = e−iβγ5 q which is an ‘axial’ version of the global U(1) associated ˆ ˆ Copyright 2004 IOP Publishing Ltd .153). Consider now the state ˆ Q 5 |+ .163) ˆ Remarkably enough. |− satisfy ˆ P|+ = |+ ˆ P|− = −|− . there should exist a state ˆ Q 5 |ψ with the same eigenvalue E and the opposite parity.1. for the simple reason that the ‘Dirac’ structure of (12. (12. it may reasonably be objected that all of this refers not only to the massless but also the non-interacting case.159) then implies that P Q 5 rule for operator transformations in quantum mechanics. Suppose we allow the fermion field ψ to interact ˆ µ via the standard electromagnetic coupling with a U(1)-gauge field A ¯ ˆ ˆ ˆ ˆ Äint = q ψγ µ ψ Aµ . as opposed to the ordinary vector currents with no γ5 . for every state |ψ with energy eigenvalue E . We shall discuss this additional U(1)-symmetry in section 18. chiral symmetry apparently implies the existence of ‘parity doublets’.6 namely ˆ q = exp(−iβ · τ /2γ5 )q. that is.162) Hence. ˆ P −1 (12. However. −1.

not in flavour space. somewhat confusingly. and will be degenerate with |u . these chiral charges have exactly the same effect as the ordinary flavour isospin operators of (12. which we shall denote by SU(2)f 5 . In fact.109)) τ ¯ ˆ ( )µ ˆ ˆ T 52 = qγ µ γ5 q 2 1 (12.171) ˆ ug ˆ dg and the 3 × 3 λ’s act in the r–b–g space. H ] = 0. since [T+5 . But they are pseudoscalars rather than scalars and. As far as their action in the isospinor u–d space is concerned.170) uγ µ u Aa + dγ µ d Aa ˆ µ µ 2 2 where     ˆ dr ur ˆ ˆ ˆ ˆ u =  ub  ˆ d =  db  (12.170) is not a simple ‘gauging’ of (12. whereas the isospin-raising 1 ˆ( ) operator T+2 is such that ˆ( ) T+2 |d = |u 1 (12. hence. Further.167) which are pseudoscalars. ˜ ˜ this state |u will be degenerate with |d . Thus. |d of negative parity. acting in the u–d space. Similarly. the interaction (12. the interaction terms have the form ¯ λa ˆ ˆ ¯ λa ˆ ˆ ˆ (12. ˜ ˜ Suppose we now let the quarks interact. so that it is invariant under Copyright 2004 IOP Publishing Ltd . for example by an interaction of the QCD type already indicated in (12.169) ˆ (2) ˆ but the |u state will have opposite parity from |u .109). the state |d produced via ˜ 1 (2) ˆ T−5 |u will have opposite parity from |d . |d of (say) positive parity and a further two massless states |u .168) ˆ T+5 will also produce a u-type state from a d-type one via ˆ (2) ˜ T+5 |d = |u 1 1 1 (2) (12.170) is invariant under the global SU(2)f 5 chiral symmetry (12.164): a covariant derivative is being introduced but in the space of a new (colour) degree of freedom. In that case. (12.170).There are three associated Noether currents (compare (12. in this simple model. Note that. called chiral SU(2)f . We have a new non-Abelian global symmetry.166) which are axial vectors and three associated ‘charge’ operators ˆ( ) T 52 = 1 τ ˆ q † γ5 q d 3 x ˆ 2 (12. Just as in the previous U(1) case. The upshot is that we have two massless states |u .165). belonging to the T = 1 representation of SU(2). the flavour degrees of freedom are ‘inert’ in (12.136). they flip the parity of a state on which they act.

the law of combination being matrix multiplication.4 Show that the Ti ’s defined by (12. Are the baryons then supposed to be massless too? If so. in part 7.3 Check the commutation relations (12. perhaps the discussion is idle. In short. we have here a situation in which a symmetry of the Lagrangian (to an apparently good approximation) does not seem to result in the expected multiplet structure of the states. One might feel that this whole discussion is unrealistic. while the Dirac structure implies that it is also invariant under chiral SU(2)f 5 transformations (12. 12. Although these T currents have been introduced entirely within the context of strong interaction symmetries. based as it is on massless quarks.165). as they are evidently by no means massless. Yet such ‘parity doublet’ partners of the physical p and n are not observed and so we have a puzzle.28). But this seems very implausible in view of the actual magnitudes of m u and m d compared to the nucleon masses.18). The effect of the QCD interactions must be to bind the quark into nucleons such as the proton (uud) and neutron (ud d). Copyright 2004 IOP Publishing Ltd . 12. promoting the negative-parity ‘nucleon’ state to an acceptably high mass. from the interaction energy in the fields.2 Derive (12.1 Verify that the set of all unitary 2 × 2 matrices with determinant equal to +1 form a group. But it is not necessary to suppose that the mass of a relativistic bound state has any very simple relation to the masses of its constituents: its mass may derive. Alternatively. Problems 12. as we shall see in chapter 20. splits the degeneracy of the nucleon parity doublets.45) satisfy (12. it is a remarkable fact that exactly these currents also appear in strangeness-conserving semileptonic weak interactions such as β -decay.SU(2)f transformations. which of course breaks the chiral symmetry. But what about the equally possible states (u u d) and (u d d). for example? These would have to be degenerate in mass ˜˜ ˜ ˜˜˜ with (uud) and (ud d). in part at least. All the foregoing can be extended unchanged to chiral SU(3)f 5 . one might suppose that somehow the finite mass of the u and d quarks.) Thus some of the physical consequences of ‘spontaneously broken chiral symmetry’ will involve weak interaction quantities. (The fact that both appear is precisely a manifestation of parity violation in weak interactions. given that QCD is ‘flavour blind’ and supposing that m s ≈ 0. The resolution of this puzzle will have to await our discussion of ‘spontaneous symmetry breaking’. 12. and of opposite parity. In conclusion.47). we note an important feature of the flavour symmetry currents 1 1 ( 2 )µ ˆ ( )µ ˆ and T 52 discussed in this and the preceding section.

83).) ˆ( ) 12.100).8 Check that (12. 12.81) and verify that the matrices G a defined by (12.99).101) satisfy (12. 12.62). 1 Copyright 2004 IOP Publishing Ltd .5 Write out each of the 3 × 3 matrices Ti(1) (i = 1.7 Show that a general Hermitian traceless 3×3 matrix is parametrized by eight real numbers.6 Verify (12.84) satisfy the commutation relations (12.99). (1) 1. (Note: use the anti-commutation relations of the fermionic operators.103).101) satisfy the commutation relations (12.11 Verify that the operators T 2 given by (12. 12. 3) whose matrix elements are given by (12.10 Verify that the operators T 2 defined by (12.80) and the infinitesimal form (8) of (12. 12.12. ˆ 12. 2 and 3 on both sides of (12.48) and verify that they satisfy the SU(2) commutation relations (12.9 Verify.100) follows from (12.84) is consistent with (12.47). 2. by comparing the coefficients of that (12.

In the present paper we wish to explore the possibility of requiring all interactions to be invariant under independent rotations of the isotopic spin at all spacetime points. one is then not free to make any choices at other spacetime points. . It seems that this is not consistent with the localized field concept that underlies the usual physical theories. The dependence of ψ ( 2 ) (x) on the spacetime coordinate x has now been included explicitly but the parameters α are independent of x. . what a neutron. As we have seen in the previous chapter. n). As usually conceived. x) as indicated.13 LOCAL NON-ABELIAN (GAUGE) SYMMETRIES The difference between a neutron and a proton is then a purely arbitrary process. Yang and Mills (1954) Consider the global SU(2) isospinor transformation (12. ψ ( 2 ) (x) = exp(iα · τ /2)ψ ( 2 ) (x) 1 1 1 1 (13. therefore. . only globally. . To Yang and Mills (1954) (cf the quotation above) this seemed somehow an unaesthetic limitation of symmetry: ‘Once one chooses what to call a proton.1) for an isospin doublet wavefunction ψ ( 2 ) (x). one is then not free to make any choices at other spacetime points. However. invariance under this transformation amounts to the assertion that the choice of which two base states—( p. written here again. Note that we have inserted a parameter g in the exponent to make the Copyright 2004 IOP Publishing Ltd . which is why the transformation is called a ‘global’ one. the choice cannot be made independently at all spacetime points.’ They even suggested that this could be viewed as ‘inconsistent with the localized field concept’ and they. at one spacetime point.32). . however. (u. d). what a neutron. at one spacetime point.—to use is a matter of convention: any such non-Abelian phase transformation on a chosen pair produces another equally good pair.2) in which the phase parameters α(x) are also now functions of x = (t. ‘explored the possibility’ of replacing this global (spacetime independent) phase transformation by the local (spacetime dependent) one 1 1 ψ ( 2 ) (x) = exp[igτ · α(x)/2]ψ ( 2 ) (x) (13. this arbitrariness is subject to the following limitations: once one chooses what to call a proton.

We shall also start with the ‘wavefunction’ formalism.1 Local SU(2) symmetry: the covariant derivative and interactions with matter In this section we shall introduce the main ideas of the non-Abelian SU(2) gauge theory which results from the demand of invariance. see also Shaw (1955). analogous to the electromagnetic charge q . as mentioned in the introduction to chapter 12. bearing in mind that this will eventually mean weak isospin. it is not the flavour SU(3) of section 12. This will be introduced in the following chapter. we shall delay the discussion of this application until after QCD—which is the theory of strong interactions but at the quark rather than the composite (hadronic) level.4 how the ‘requirement’ of local U(1) phase invariance led almost automatically to the local gauge theory of QED. However. Global symmetries and their associated (possibly approximate) conservation laws are certainly important but they do not have the dynamical significance of local symmetries. We saw in section 7. under transformations such as (13. The consideration of theories based on (13.2) was the fundamental step taken by Yang and Mills (1954). since the group theory is more familiar. QCD is based on the local form of an SU(3) symmetry—once again.5. in ¯ ˆ ˆ which the conserved current ψγ µ ψ of the global U(1) symmetry is ‘promoted’ to ˆ the role of dynamical current which. colour. when dotted into the gauge field Aµ . consequently. In the event. the original Yang– Mills attempt to get a theory of hadronic interactions by ‘localizing’ the flavour symmetry group SU(2) turned out not to be phenomenologically viable (although a remarkable attempt was made to push the idea further by Sakurai (1960)). But this is complicated by the fact that the symmetry is ‘spontaneously broken’ and. which are the subject of the present one. gave the ˆ interaction term in ÄQED . however.2).analogy with the electromagnetic U(1) case ψ (x) = exp[iqχ(x)]ψ(x) (13.2 but a symmetry with respect to a totally new degree of freedom. We shall generally use the language of isospin when referring to the physical states and operators. A similar link between symmetry and dynamics appears if—following Yang and Mills—we generalize the non-Abelian global symmetries of the preceding chapter to local non-Abelian symmetries. 13. Although the application of local SU(2) symmetry to the weak interactions will follow that of local SU(3) to the strong. we shall begin our discussion of local non-Abelian symmetries with the local SU(2) case. or covariance. the successful application of a local SU(2) symmetry was to the weak interactions. Copyright 2004 IOP Publishing Ltd .3) even stronger: g will be a coupling strength. deferring the field-theory treatment until section 13.

there is an extra term in (13. in addition to the expected first term on the right-hand side of (13.5).and 4-vectors. covariance under local non-Abelian phase transformations requires the introduction of a definite force field.5). (13. We need the local SU(2) generalization of (13. In fact.4) For convenience. while ∂ µ ψ does not. wave equations with a suitably defined covariant derivative can be covariant under (13.6) Comparing (13. appropriate to the local SU(2) transformation (13. In appendix D of volume 1 we introduced the idea of ‘covariance’ in the context of coordinate transformations of 3. physically this means that. Note that D µ has to carry a prime also. nevertheless. The property (13.5).6) with (13. The essential notion was of something ‘maintaining the same form’ or ‘transforming the same way’.7) ensures the gauge covariance of wave equations in the U(1) case. the ordinary gradient Copyright 2004 IOP Publishing Ltd . a wavefunction transforms as (cf (13.5 of volume 1.4).4 and 3. we see that. with no additional term on the right-hand side. which we did not present in volume 1 but shall discuss in the following section.We shall mimic as literally as possible the discussion of electromagnetic gauge covariance in sections 3.4 of volume 1) D µ ψ(x) → D µ ψ (x) = exp(iqχ(x))D µ ψ(x) (13. The transformations being considered here are gauge transformations rather than coordinate ones. which has the same form as the right-hand side of (13.2). just as for electromagnetism. However. the covariant derivative is D µ = ∂ µ + iq Aµ (x). since the gradient terms in the equation will act on the phase factor α(x). under them. By contrast. there is a much closer analogy between the ‘coordinate’ and the ‘gauge’ cases.6).5). the similar property in the quantum field case meant that a globally U(1)-invariant Lagrangian could ˆ be converted immediately to a locally U(1)-invariant one by replacing ∂ µ by D µ (section 7.5) from which it easily follows that the derivative (gradient) of ψ transforms as ∂ µ ψ(x) → ∂ µ ψ (x) = exp(iqχ(x))∂ µ ψ(x) + iq∂ µ χ(x) exp(iqχ(x))ψ(x).2) (taking the isospinor example for definiteness). the covariant derivative of ψ transforms as (see section 3. (13.4).2). the term covariant derivative seems appropriate. Under a local U(1) phase transformation.6). Just as in the U(1) case (13. no freeparticle wave equation can be covariant under the transformation (13. D µ ψ transforms in the same way as ψ.6).7) exactly as in (13. In the electromagnetic case. Thus. As in that case. since it contains Aµ which transforms to A µ = Aµ − ∂ µ χ(x) when ψ transforms by (13.3)) ψ(x) → ψ (x) = exp(iqχ(x))ψ(x) (13. we recall here the crucial property of D µ . it is true that.

1 1 (13. W2 . So we require (D µ ψ ( 2 ) (x)) = exp[igτ · α(x)/2](D µ ψ ( 2 ) (x)). By analogy with (13. We now discuss (13.8) exp[A] = n=0 An /n! and differentiating term by term. without the second term in (13. as indicated on the left-hand side.8). Yang–Mills fields.4) so as to fulfil this requirement is D µ (acting on an isospinor) = ∂ µ + igτ · W µ (x)/2.9) The definition of D µ which generalizes (13. ψ ( 2 ) transforms by 1 1 ψ ( 2 ) = (1 + igτ · (x)/2)ψ ( 2 ) (13. which are 1 of course functions of x since the transformation is local.9) and (13.44)).acting on ψ ( 2 ) (x) does not transform in the same way as ψ ( 2 ) (x): taking ∂ µ of (13. is only appropriate for 1 isospinors ψ ( 2 ) : it has to be suitably generalized for other ψ (t ) ’s (see (13. The term τ · W µ is then the 2 × 2 matrix µ µ µ W1 − iW2 W (13.10) The definition (13.25): the x-dependence of the W µ ’s is understood. (13.2) leads to ∂ µ ψ ( 2 ) (x) = exp[igτ · α(x)/2]∂ µψ ( 2 ) (x) + igτ · ∂ µ α(x)/2 exp[igτ · α(x)/2]ψ ( 2 ) (x) as can be checked by writing the matrix exponential exp[A] as the series ∞ 1 1 1 1 1 (13.11) generalizing the single electromagnetic gauge field Aµ .10) in detail.14) . W3 ) µ µ µ (13.9). for the algebraically simpler case of an infinitesimal local SU(2) transformation with parameters (x). The W µ (x) are three independent gauge fields W µ = (W1 .10). In this case.13) and the ‘uncovariant’ derivative ∂ µ ψ ( 2 ) transforms by ∂ µ ψ ( 2 ) = (1 + igτ · (x)/2)∂ µ ψ ( 2 ) + igτ · ∂ µ (x)/2 ψ ( 2 ) Copyright 2004 IOP Publishing Ltd 1 1 1 1 (13. The ∂ µ is multiplied implicitly 1 by the unit 2 matrix and the τ ’s act on the two-component space of ψ ( 2 ) .7).e. the key property we 1 demand for our SU(2) covariant derivative D µ ψ ( 2 ) is that this quantity should 1 transform like ψ ( 2 ) —i. Let us ‘decode’ the desired property (13. They are called SU(2) gauge fields or.12) τ · Wµ = µ 3 µ µ W1 + iW2 −W3 using the τ ’s of (12. more generally.

4(b)) σ · a σ · b = a · b + iσ · a × b this yields τ · δW µ = −τ · ∂ µ (x) − gτ · ( (x) × W µ ). we would need to know W µ —that is.8) 1 with α → . for the left-hand side in the infinitesimal case. 1 1 µ = W µ + δW µ . Instead.16) to determine the transformation law of W µ . (13.9) yields. 2 (13.18) Multiplying out the terms. D µ ψ ( 2 ) = (∂ µ + igτ · W µ /2)[1 + igτ · (x)/2]ψ ( 2 ) 1 1 (13.20) Using the identity for Pauli matrices (see problem 4. and use the imposed equality between (13. the transformation law for the three W µ fields.13) into our covariant derivative requirement (13.21) (13.17) (13. as compared to (13.22) Copyright 2004 IOP Publishing Ltd .10) and (13. we shall proceed ‘in reverse’.16) In order to verify that these are the same. 1 (13. under this infinitesimal transformation.14) exhibits again an ‘extra piece’ on the right-hand side. however. (13. Equation (13.where we have retained only the terms linear in from an expansion of (13. However.15) while the right-hand side is [1 + igτ · (x)/2](∂ µ + igτ · W µ /2)ψ ( 2 ) . Wµ → W Then the condition of equality is [∂ µ + igτ /2 · (W µ + δW µ )][1 + igτ · (x)/2]ψ ( 2 ) = [1 + igτ · (x)/2](∂ µ + igτ · W µ /2)ψ ( 2 ) . inserting (13. neglecting the term of second order involving the product of δW µ and and noting that ∂ µ ( ψ) = (∂ µ )ψ + (∂ µ ψ) we find that many terms cancel and we are left with ig τ · ∂ µ (x) τ · δW µ = − ig 2 2 τ · (x) + (ig)2 2 (13.13).19) τ · Wµ 2 − τ · Wµ 2 τ · (x) .15) and (13. We have now dropped the x -dependence of the ψ ( 2 ) ’s but kept that of (x) and we have used the simple ‘1’ for the unit matrix in the two-dimensional isospace. Suppose that.

65) in section 12.3.10) for the covariant derivative has been assumed and only the infinitesimal version of (13.25) the gauge fields transform according to W µ = W µ − ∂ µ (x) − g[ (x) × W µ ]. (13. In exactly the same 1 way. the W µ behave as a normal triplet. But there is the extra term −∂ µ (x). the form (13.1.26). in general. which also involves 1 ∂ µ (x). Thus. we deduce δ W µ = −∂ µ (x) − g[ (x) × W µ ]. ∂ µ ψ ( 2 ) did not transform ‘properly’ as an SU(2) doublet.23) would be simply the (infinitesimal) transformation law for the T = 1 triplet representation of SU(2)—see (12. representation of SU(2) and this is the one to which gauge fields belong. (13. thus. we can say that under global SU(2) transformations. It turns out that (13. As mentioned at the end of section 12.24) (where D µ is given by (13.1. no longer transform ‘properly’ as an SU(2) triplet. The remarkable result behind the fact that D µ ψ ( 2 ) does transform ‘properly’ under local SU(2) transformations is that the extra term in (13.2) has been treated explicitly. or ‘regular’. Clearly this is directly analogous to the −∂ µ χ(x) term in the transformation of the U(1) gauge field Aµ . under a local SU(2) transformation. then ∂ µ (x) would of course vanish and the transformation law (13.23) The reader may note the close similarity between these manipulations and those encountered in section 12.14)! To summarize progress so far: we have shown that.26) In obtaining these results.10)) holds true if in addition to the infinitesimal local 1 SU(2) phase transformation on ψ ( 2 ) ψ ( 2 ) = [1 + igτ · (x)/2]ψ ( 2 ) 1 1 (13. the relation (D µ ψ ( 2 ) ) = [1 + igτ · (x)/2](D µ ψ ( 2 ) ) 1 1 (13.23) precisely cancels that in (13.14). an independent µ infinitesimal function i (x) is required for each component Wi (x).Equating components of τ on both sides. because of the second term in (13.10) is still appropriate for the finite (noninfinitesimal) transformation (13.2) but the associated transformation law for the gauge fields is then slightly more complicated than (13. for infinitesimal transformations. Equation (13.3.27) . Let us write U(α(x)) ≡ exp[igτ · α(x)/2] Copyright 2004 IOP Publishing Ltd (13. But under local SU(2) transformations they acquire the additional −∂ µ (x) piece and.23) would indeed be just that of an SU(2) triplet.23) defines the way in which the SU(2) gauge fields W µ transform under an infinitesimal SU(2) gauge transformation.64) and (12. here. If it were not for the presence of the first term ∂ µ (x) on the right-hand side. (13.2. the T = 1 representation is the ‘adjoint’. If the ’s were independent of x.

The left-hand side is (∂ µ + igτ · W µ /2)U(α(x))ψ ( 2 ) 1 1 1 1 1 1 (13.33) = i µ 1 (∂ U)U−1 + U τ · W µ U−1 g 2 (13.28) and (13. Then we require D µ ψ ( 2 ) = U(α(x)) D µ ψ ( 2 ) .30) (13. 1 τ·W 2 µ (13.10). Indeed. 1 1 1 1 ( that ψ ( 2 ) transforms by ψ ( 2 ) = U(α(x))ψ ( 2 ) . 1 Suppose now that we consider a Dirac equation for ψ ( 2 ) : (iγµ ∂ µ − m)ψ ( 2 ) = 0 1 1 (13. 1 1 1 1 (13.29). we have U(α(x))[iγµ D µ − m]ψ ( 2 ) = iγµ U(α(x))D µ ψ ( 2 ) − mU(α(x))ψ ( 2 ) = iγµ D µ ψ ( 2 ) − mψ ( 2 ) 1 1 1 1 1 (13. if (iγµ D µ − m)ψ ( 2 ) = 0 Copyright 2004 IOP Publishing Ltd 1 (13. We assert that we can ensure local SU(2) gauge covariance by replacing ∂ µ in this equation by the covariant derivative of (13. equivalently.31) The U∂ µ ψ ( 2 ) terms cancel leaving 1 (∂ µ U)ψ ( 2 ) + igτ · W µ /2 Uψ ( 2 ) = Uigτ · W µ /2 ψ ( 2 ) .34) which defines the (finite) transformation law for SU(2) gauge fields. Problem 13.32) Since this has to be true for all (two-component) ψ ( 2 ) ’s.37) .1 verifies that (13.26) in the infinitesimal case α(x) → (x).29) = (∂ µ U)ψ ( 2 ) + U∂ µ ψ ( 2 ) + igτ · W µ /2 Uψ ( 2 ) while the right-hand side is U(∂ µ + igτ · W µ /2)ψ ( 2 ) .36) using equations (13.28) (13. Thus.34) reduces to (13. we can treat it as an 1 operator equation acting in the space of ψ ( 2 ) ’s to give ∂ µ U + igτ · W µ /2 U = Uigτ · W µ /2 or.35) where both the ‘isospinor’ components of ψ ( 2 ) are four-component Dirac spinors.

just as in the U(1) case.129). obeying equation (13.41) is a very economical generalization of rule (ii) in comment (3) of section 8. (13. 3) which tells us which of the three SU(2) W-states is participating.39) In lowest-order perturbation theory the one-W emission/absorption process is given by the amplitude (cf (8. then (iγµ D µ − m)ψ ( 2 ) = 0 1 (13. 1 ψ ( 2 ) . In the same way. −ig(τ r /2)γµ (13.Figure 13. The external W µ field is now specified by a spin-1 polarization vector µ . of course. The essential point here. The foregoing is easily generalized to SU(2) multiplets other than doublets.39)) for the electromagnetic case) −ig ( ) ¯( ) ψf 2 (τ /2)γµ ψi 2 · W µ d4 x 1 1 (13.40) exactly as advertised (for the field-theoretic vertex) in (12. therefore.38) proving the asserted covariance.1 is. 1 1 (13. 2. u f and dotted into µ ¯ and a r . The matrix degree 1 of freedom in the τ ’s is sandwiched between the two-component isospinors ψ ( 2 ) : 1 the γ matrix acts on the four-component (Dirac) parts of ψ ( 2 ) . like a photon. We shall change the notation slightly to use t instead of T for the ‘isospin’ Copyright 2004 IOP Publishing Ltd .41) which is to be sandwiched between spinors/isospinors u i . Consider again the case of an SU(2) isospinor fermion. Vertex for isospinor-W interaction. which are determined by the local phase invariance requirement (the ‘gauge principle’).38). Indeed. and by an ‘SU(2) polarization vector’ a r (r = 1.1. is that the locally covariant form includes 1 interactions between the ψ ( 2 ) ’s and the gauge fields W µ . This can be written as / / (i∂ − m)ψ ( 2 ) = g(τ /2) · Wψ ( 2 ) . any free-particle wave equation 1 satisfied by an ‘isospinor’ ψ ( 2 ) —the relevant equation is determined by the Lorentz spin of the particles involved—can be made locally covariant by the use of the covariant derivative D µ .3. we can already begin to find some of the Feynman rules appropriate to tree graphs for SU(2) gauge theories. The Feynman rule for figure 13.

[τ · . The crucial point is the appearance of the quadratic g 2 multiplying the commutator of the τ ’s. for example. τ · W]. 3) satisfy (cf (12. The gauge fields themselves are not ‘inert’ as far as the gauge group is concerned: in the SU(2) case they have isospin 1.40) by T (t ) .quantum number. This is not the same as in electromagnetism. so as to emphasize that it is not the hadronic isospin. in ‘grand unified theories’ of strong.47)) [Ti . −5e. Since it is the coupling constant of an Abelian group.T(t ) ]ψ (t ) . This is one of the motivations of attempts to embed the electromagnetic gauge transformations inside a larger non-Abelian group structure. while for a general group they belong to the regular representation Copyright 2004 IOP Publishing Ltd .23)—in the SU(2) case—then 1 the g’s appearing in ψ ( 2 ) (equation (13. each charged field interacts with the gauge field Aµ via a coupling whose strength is its charge (e. once δW µ is given by equation (13. The fermion vertex corresponding to (13. for which we retain T : t will be the symbol used for the weak isospin to be introduced in chapter 20.42)) must be the same as the one appearing in δW µ . −e.44) (t ) (t ) (t ) i j k Tk .43) which is a (2t + 1) × (2t + 1) matrix acting on the (2t + 1) components of ψ (t ) . (t ) (13.44) is obtained by replacing τ /2 in (13.20)). there is no such commutator—the associated U(1) phase group is Abelian. (c) Finally we draw attention to the extremely important physical significance of the second term δW µ (equation (13. in the W µ transformation (equation (13. 2e. (b) According to the foregoing argument. The general local SU(2) transformation for a t -multiplet is then ψ (t ) → ψ (t ) = exp[igα(x). T j ] = i The appropriate covariant derivative is D µ = ∂ µ + igT(t ) · W µ (13. and the scale of quantities appearing in such commutation relations is not arbitrary. Such is the case. a commutator is a nonlinear quantity. each charged field could have an arbitrary charge from this point of view: there are no commutators to fix the scale. ).13)) and ψ (t ) (via the infinitesimal version of equation (13.42) where the (2t + 1) × (2t + 1) matrices Ti (i = 1. We end this section with some comments: (a) It is a remarkable fact that only one constant g is needed.23)). 2. As signalled by the presence of g 2 . it is actually a mystery why electric charge should be quantized. It is an instructive exercise to check that. The gauge fields interact with such ‘isomultiplets’ in a universal way—only one g. There. In the electromagnetic case. (13. weak and electromagnetic interactions. the same for all the particles—which is prescribed by the local covariance requirement to be simply that interaction which is generated by the covariant derivatives.

There. we explore further the geometrical analogy.5. eI → eR . . Other such ‘vectors’ φ1 (x). other wavefunctions for particles of the same charge q) when evaluated at the same point x will have ‘components’ transforming the same as (13. and write α(x) = qχ(x) so that (13. normal to the eR –eI plane). . equations (13. already hinted at. they will necessarily interact with themselves. (i.3) becomes (cf (3.45) under the axis rotation eR . where the gauge field Aµ for the photon is of course uncharged: quite simply. we may perhaps expect something to ‘go wrong’ with the transformation law for the gradient. let us split ψ into its real and imaginary parts ψ = ψR + iψI . as shown in figure 13. which have been rotated by −α(x) about an axis in the direction eR × eI (i.2 Covariant derivatives and coordinate transformations Let us go back to the U(1) case.of the group. Thus. unlike the ordinary derivative ∂ µ ψ(x) which acquired an ‘extra’ piece when transformed.4)–(13. But the components of the vector ψ(x + dx) will behave differently.45) ψI (x) = sin α(x)ψR (x) + cos α(x)ψI (x). of course—but there is a slightly different way of thinking about it. eI . a non-Abelian gauge theory of gauge fields alone. 13. while that of ψ(x + dx) would involve the different function α(x + dx).7). The derivative involves not only ψ(x) at the point x but also ψ at the infinitesimally close. To bring out the geometrical analogy we are seeking. Thus.e. for the (gauge) covariant derivative. We shall examine the form of these ‘self-interactions’ in section 13.63)) ψR (x) = cos α(x)ψR (x) − sin α(x)ψI (x) (13. which transformed like ψ(x) under local U(1) phase transformations.45) when written at x + dx will involve α(x + dx).2.23) is absent for Aµ . φ2 (x). This is profoundly different from the electromagnetic case. e = 0 for a photon and the second term in (13. D µ ψ(x). since they are also the quanta of the force field. with no ‘matter’ fields. which Copyright 2004 IOP Publishing Ltd . The transformation law (13.e. First. This will ultimately lead us to an important new quantity in non-Abelian gauge theories. and the transformation law of ψ(x) involves α(x). If we think of ψR (x) and ψI (x) as being the components of a ‘vector’ ψ(x) along the eR and eI axes. then (13. has non-trivial interactions and is not a free theory. .45) would represent the components of ψ(x) as referred to new axes eR and eI . the introduction of the (gauge) covariant derivative D µ produced an object. point x + d x . This followed from simple calculus. the analogue of the Maxwell field strength tensor F µν .2. but different. respectively. The fact that non-Abelian (Yang–Mills) gauge fields carry non-Abelian ‘charge’ degrees of freedom means that.

chapter 5)) where just this kind of problem arises.3)! But that would mean that our ‘naive’ approach to rotations of the derivative of ψ(x) amounts to using one set of axes at x. Thus. x = r cos θ y = r sin θ (13. which is likely to lead to ‘trouble’. We have. Consider now an elementary example (from Schutz (1988.2.46) in (13.47) (13.Figure 13. namely the use of polar coordinate basis vectors er and eθ . Geometrical analogy for a U(1) gauge transformation.46) and in a (real!) Cartesian basis dr is given by dr = dx i + dy j. the rotation angle is α(x) + ∂µ α(x) d x µ rather than α(x). Using (13. (13. Now comes the key step in the analogy: we may think of the additional angle ∂µ α(x) d x µ as coming about because.47) we find dr = (dr cos θ − r sin θ dθ )i + (dr sin θ + r cos θ dθ ) j = dr er + dθ eθ where er = cos θ i + sin θ j Copyright 2004 IOP Publishing Ltd (13. which point in the r and θ directions respectively. and another at x + dx. (to first order in d x ) is α(x) + ∂µ α(x) dx µ . in going from x to x + dx . the coordinate basis vectors eR and eI have been rotated through +∂µ α(x) d x µ (see figure 13. for ψR (x + d x) and ψI (x + d x).48) eθ = −r sin θ i + r cos θ j.49) . as usual.

Similarly. Plainly. 2 with q 1 = r. But whatever the nature of the coordinate system. we were able to calculate ∂ eα /∂q β explicitly from (13.51) to be evaluated.Figure 13. eθ .50) V = V r er + V θ eθ ≡ V α eα and that we are also interested in the derivatives of V . Let us calculate ∂ V /∂r .49).49) we find ∂ eθ 1 ∂ er =0 = − sin θ i + cos θ j = eθ (13. as shown in figure 13.3.52).3.52) ∂r ∂r r which allows the last two terms in (13. In the present case. So at each point (r. by brute force: ∂Vr ∂V θ ∂ er ∂ eθ ∂V = er + eθ + V r + Vθ ∂r ∂r ∂r ∂r ∂r (13. θ ) (13. we can calculate ∂ V /∂θ . θ ) we have different axes er . θ . in this basis. we may write these results as ∂V ∂V α ∂ eα = e + Vα β β β α ∂q ∂q ∂q (13.54) . In general. Changes in the basis vectors er and eθ of polar coordinates. for eθ ) as we move about in the x–y plane. From (13. Now suppose that we wish to describe a vector field V in terms of er and eθ via (sum on α = r.51) where we have included the derivatives of er and eθ to allow for the fact that these vectors are not constant. er and eθ change direction (and even magnitude. for example. as in (13.53) where β = 1. q 2 = θ and α = r. ∂ eα /∂q β is some vector and must be expressible as a linear combination of the basis vectors via an expression of the form ∂ eα = ∂q β Copyright 2004 IOP Publishing Ltd γ αβ eγ (13.

As we have seen. the components of the ∂β derivative of V are ∂β V α + α γβ V γ ≡ Dβ V α . for which Dµ ψ = ∂µ ψ + iq Aµ ψ.where the repeated index γ is summed over as usual (γ = r. the analogy is even closer in the non-Abelian—e. But the gauge covariant 1 combination (∂ µ + igτ · W µ /2)ψ ( 2 ) does transform as an isospinor under local SU(2) transformations. also because of a ∂ µ term.56)) by Dβ V α or.54) into (13. There are two pieces. the components of the derivative of V are not simply ∂ V α /∂q β but contain an additional term: the ‘components of the derivative of a vector’ are not just the ‘derivatives of the components of the vector’. There is.55) ∂q β ∂q β This is a very important result: it shows that. for example. thus. ∂β V α by itself does not transform as a ‘Tβ ’ tensor and. This circumstance is highly reminiscent of the situation we found in the case of gauge transformations. one down’ indices: we shall not prove this here. The ‘i’ of course is a big difference. The quantity Dµ ψ transforms under a gauge transformation in the same way as ψ itself but ∂µ ψ does not.55) tells us that in the eα basis. the structure of Dµ ψ is very similar to that of Dβ V α . Copyright 2004 IOP Publishing Ltd . as used in (13. Indeed. whereas the components of V in the basis eα are just V α . and despite appearances. often. ∂ µ ψ ( 2 ) does not transform as an SU(2) isospinor because of the extra piece involving ∂ µ . meaning in this case essentially that it transforms the way its indices would have you believe it should. a close analogy between the ‘good’ transformation properties of Dβ V α and of Dµ ψ. Further.55).β (in the latter notation. the first of which is the straightforward derivative. by V α .β ). showing that in the gauge symmetry case the transformations mix the real and imaginary parts of the wavefunction. α By contrast.56) is called the ‘covariant derivative’ of V α within the context of the mathematics of general coordinate systems: it is denoted (as in (13. (13. ∂β V α is V α . Let us abbreviate ∂/∂q β to ∂β : then (13. the two ‘extra’ ∂ µ pieces cancelling each other out. local SU(2)— 1 case. α γ β is not a ‘T α γ β ’-type tensor: only the combined α object Dβ V α is a ‘Tβ ’. Inserting (13. θ ). rather than actual spatial components of a vector. nor do the gauge fields W µ transform as pure T = 1 states.g. (13.e.56) The expression (13. referring instead to Schutz (1988). Crucially. summed over) indices α and γ gives finally ∂V ∂V α = + α γ β V γ eα .53) and interchanging the ‘dummy’ (i. while the second involves a new field ( or A) and is also proportional to the original field. This property is the reason for the name ‘covariant derivative’. Consider the simplest case. it transforms as a tensor Tβ (see appendix D of volume 1) with the indicated ‘one up. in a similar way. The most important property of Dβ V α is its transformation character under α general coordinate transformations. that of U(1).

Such a definition of ‘no change’ of course implies that change has occurred.4). with respect to the original axes at q β . α (13.Figure 13. representing the change in V α in moving from q β to q β + dq β : dV α = [V α (q β + dq β ) − V α (q β )]. the quantities dV α do not form the components of a vector and the reason is that V α (q β + dq β ) are components with respect to axes at q β + dq β . A reasonable definition of such a ‘preserved’ vector field is one that is unchanged in length and has the same orientation relative to the axes at q β + dq β as it had relative to the axes at q β (see figure 13. (13. Parallel transport of a vector V in a polar coordinate basis. while keeping it somehow ‘the same’ as it was at q β .58) . while V α (q β ) are components with respect to different axes at q β . Again. To form a ‘good’ differential DV α . In other words.57) The first term on the right-hand side of (13. Then a reasonable definition of the ‘good’ differential of V α would be V α (q β + dq β ) − (V α (q β ) + δV α ) = dV α − δV α . make the identification δV α = − Copyright 2004 IOP Publishing Ltd α γβ V γ dq β .4. Let us denote by δV α the difference between the components of V after parallel transport to q β + dq β and the components of V at q β (see figure 13.56) by dq β and sum over β so as to obtain DV α ≡ ∂β V α dq β + α γβ V α dq β . We interpret this as the covariant differential DV α of (13. This means that we need some way of ‘carrying’ V α (q β ) to q β + dq β . in general. transforming as a vector. we must subtract quantities defined in the same coordinate system. despite appearances. V is ‘dragged around’ with the changing coordinate frame.4).57) and. accordingly.57) is ∂ Vβ dq β which is just the ∂q conventional differential dV α . There is a useful way of thinking about the two contributions to Dβ V α (or Dµ ψ). a process called parallel transport. Let us multiply (13.

63) to first order in dx.60) in a very satisfactory way to our original discussion of electromagnetism as a gauge theory in chapter 3 and.On this interpretation. Equations (13. On the right-hand side of (13.60) generalize 1 1 straightforwardly for Dψ ( 2 ) and δψ ( 2 ) .59) and (13. as indeed it is in some quarters. x) A·d ∇ψ( A = 0. after the vector has been carried by ‘parallel transport’ from one point to the other: they are often called ‘connection coefficients’ or just ‘the connection’. due to ‘parallel transport’ as prescribed by the gauge connection A. Copyright 2004 IOP Publishing Ltd .60) has a very similar structure to (13.63). suggesting that the electromagnetic potential Aµ might well be referred to as a ‘gauge connection’.63).61). the coefficients α γ β connect the components of a vector at one point with its components at a nearby point.62) A·d A·d ψ( A = 0. For transport restricted to the three spatial directions.2). we see (i) the change δψ of (13. Dψ ≡ D µ ψdx µ = ∂ µ ψdx µ + ie Aµ ψdx µ ≡ dψ − δψ with δψ = −ie Aµ ψ dx µ .60) Equation (13. (13.59) (13. in fact. the ‘integrated’ form of the small displacement law (13. x) (13. x + dx) x exp ie x −∞ −∞ A·d ψ( A = 0. in the absence of A.61) A·d ψ( A = 0. In an analogous way we can write. x) · dx (13.62) gives. x + dx) ≈ (1 + ie A · dx) exp ie + ∇ψ( A = 0. then. and (ii) the change in ψ viewed as a function of x.82) gives ψ(x) = exp ie replacing q by e. We can relate (13. So ψ(x + dx) = exp ie = exp ie x+dx −∞ x+dx x x −∞ (13. The solution (13. However. in particular. (13. in the U(1) gauge case.60) reduces to δψ(x) = ie A · dxψ(x).8. to (3. x) · dx] ≈ ψ(x) + ie A · dxψ(x) + exp ie x −∞ A·d [ψ( A = 0. the solution (3.58).

the vector is drawn ‘as parallel as possible’ to the previous one. C → A. the parallel transport of a vector round a flat triangle in the Euclidean plane leads to no such net change in the vector (figure 13. eθ example. j basis (which is constant throughout the plane) we would have had no such ‘trouble’. It is clear from the figure that the vector we end up with at A.5(a). eθ basis. Parallel transport (a) round a curved triangle on the surface of a sphere and (b) round a triangle in a flat plane. it has rotated by π/2 in this example. at each stage. It seems reasonable to suppose that the information about whether the space Copyright 2004 IOP Publishing Ltd . This is a fair point. after this circuit. going back to the er . By contrast. B → C. At this point the reader might object. with the rule that. in fact. and if we had simply used the i. no one forced us to use the er . an intuitively plausible definition of parallel transport is shown in figure 13.5(b)). in which one-eighth of the surface area of the unit sphere is enclosed by the triangle ABC. Then. in which transport is carried out around a closed path consisting of three great circle arcs A → B. provided that we somehow knew that we are really doing physics in a ‘flat’ space—such as the Euclidean plane.5. But suppose instead that our two-dimensional space was the surface of a sphere.Figure 13. is no longer parallel to the vector with which we started. that we had made a lot of fuss about nothing: after all.

67) . We want to calculate the net change (if any) in δV α as we parallel transport V around the loop. q 2 = b. Closed loop ABCD in q 1 –q 2 space.q 1 =a+δa q 2 =b+δb. B).64) to first order in δa. b + δb) ≈ V γ (a.66) and. remembering that we are parallel-transporting V .q 1 =a+δa α γ 1V γ dq 1 α γ 1 γ 1 V dq q 2 =b+δb. while that along C → D is (δV α )CD = − = + ≈ δa Now α γ 1 (a. b + δb).Figure 13. in the gauge case the analogy we have built up so far would lead us to expect that there are potentials Aµ which are somehow ‘flat’ ( E = B = 0) and others which represent ‘curvature’ (non-zero E. q 1 = a + δa. (13. b) − Copyright 2004 IOP Publishing Ltd γ δ2 V δ δb. we are dealing with is ‘flat’ or ‘curved’ is contained in the connection α γ β . b q 2 =b+δb. This is what we discuss next.q 1 =a δa α γ 1 (a. b + δb)V (a. b) + δb ∂ γ1 2 ∂q α (13. In a similar way.3 Geometrical curvature and the gauge field strength tensor Consider a small closed loop in our (possibly curved) two-dimensional space— see figure 13.65) + δb) ≈ α γ 1 (a.q 1 =a α γ γ 1 (a. q 2 = b + δb. (13. The change along A → B is (δV )AB = − ≈ − α α γ γ 1V q 2 =b. b)V γ (a.6—whose four sides are the coordinate lines q 1 = a. b) q 2 =b.6. 13.q 1 =a q 2 =b+δb.q 1 =a+δa dq 1 (13. V γ (a.

A non-zero value for any component of R α γ βσ means the space is curved.64) and (13. b)δa and (δψ)CD = + ie A1 (a. x 2 = b. b + δb)ψ(a.68) or. − ∂ α ∂q β γσ + α δσ δ γβ − α δβ δ γσ V γ d Aβσ (13.73) ≈ ie A1 (a. we find that (δV α )AB + (δV α )CD ≈ δaδb ∂ γ1 γ V 2 ∂q α − α γ1 γ δ2 V δ (13.6 but with q 1 replaced by x 1 and q 2 by x 2 . around the small two-dimensional rectangle defined by the coordinate lines x 1 = a.71) would take the form (δV α )loop ≈ ∂ γβ ∂q σ α ∂q 1 + α δ2 δ γ1 − α δ1 δ γ2 Vγ.65) to lowest order. (13. b) + ∂ A1 δb [ψ(a. (δV α )BC + (δV α )DA ≈ δaδb − ∂ α γ2 ∂ α γ1 ∂q 2 − α δ1 δ γ2 Vγ.70) and so the net change around the whole small loop is (δV α )ABCD ≈ δaδb ∂ γ1 ∂q 2 α − ∂ α γ2 (13. labelled as in figure 13.71) The indices ‘1’ and ‘2’ appear explicitly because the loop was chosen to go along these directions. b + δb)δa ≈ ie A1 (a.69) ∂q 1 + α δ2 δ γ1 Vγ (13. b) − ie A2 (a. b)δa + ie ∂ A1 ψ(a. b)ψ(a. We now follow exactly similar steps to calculate the net change in δψ as given by (13. which can clearly be calculated once the connection coefficients are known. (13. b) δaδb. b)ψ(a. interchanging dummy (summed) indices γ and δ in the last term.60). A flat space is one for which all components R α γ βσ = 0. (δV α )AB + (δV α )CD ≈ δaδb Similarly. The quantity in brackets in (13. the reader may verify that this is the case for our polar basis er . x 1 = a + δa. x 2 = b + δb. (13. b)ψ(a. b)ψ(a.Combining (13.74) ∂ x2 Copyright 2004 IOP Publishing Ltd . eθ in the Euclidean plane. In general. b) − ie A1 (a.72) is the Reimann curvature tensor R α γ βσ . b)A2(a.72) where d Aβσ is the area element. b)δb]δa ∂ x2 (13. Then (δψ)AB = −ie A1(a.

a curvature which we call an electromagnetic field and which has observable physical effects. suggest that F µν = 0 indicates ‘no physical effect’. Indeed. If Aµ is not expressible as the 4-gradient of a scalar. for spatial loops. and determining the trajectories of massive particles.83) can be regarded as the integrated form of (13.77).78).73) and (13. (δψ)BC + (δψ)DA ≈ −ie ∂ A2 ψ − e2 A1 A2 ψ δaδb ∂ x1 (13. consistently. In the former. in the U(1) case). charge is the source of curvature in an ‘internal’ space (the complex ψ-plane.61)–(13. in the latter. equation (3. (13. ∂ x2 (13.77) For a general loop. It will lead to an expression for the nonAbelian field strength tensor. there is a satisfying consistency between this ‘geometrical’ viewpoint and the discussion of the Aharonov–Bohm effect in section 3. (13. A closely related. while F µν = 0 implies the presence of a physical effect. The reader may consider repeating. for the local SU(2) case. Transport round such a loop results in a non-trivial net phase change if non-zero B flux is enclosed. Once again.78) where F µν = ∂ ν Aµ − ∂ µ Aν is the familiar field strength tensor of QED.63). the closed-loop transport calculation of (13. (13.77) is replaced by (δψ)loop = ie ∂ Aν ∂ Aµ ψ dx µ dx ν − ∂ xν ∂ xµ = ieF µν ψ dx µ dx ν .Combining (13. As in our remarks at the end of the previous section and equations (13. and this can be observed. therefore. because (∂ µ ∂ ν − ∂ ν ∂ µ )χ = 0). causing the spacetime axes themselves to vary from point to point. and (for the non-Abelian case) Copyright 2004 IOP Publishing Ltd .6.75) with the result that the net change around the loop is (δψ)ABCD ≈ ie ∂ A2 ∂ A1 − ∂ x2 ∂ x1 ψδaδb. The analogy we have been pursuing would. analogous to the spatial curvature revealed by Rγ βσ = 0. matter (or energy) is regarded as the source of curvature of spacetime. From this point of view there is undoubtedly a strong conceptual link between Einstein’s theory of gravity and quantum gauge theories.76) ∂ A1 ψ + e2 A1 A2 ψ δaδb. when Aµ has the ‘pure gauge’ form Aµ = ∂ µ χ the associated F µν is zero: this is because such an Aµ can clearly be reduced to zero by a gauge transformation (and also.74) we find (δψ)AB + (δψ)CD ≈ ie Similarly.73)–(13. then F µν = 0 and an electromagnetic α field is present.

) Equation (13. as an SU(2) triplet should. D ν ]ψ ( 2 ) (13.1) that the gauge fields themselves carry SU(2) ‘charge’ and act as sources for the field strength.64)) for an SU(2) triplet. Since it arises from the commutator of two gauge-covariant derivatives.82) the manipulations are very similar to those in (13. Consider first the U(1) case. or field strength tensor. D µ (on ψ ( 2 ) ) = ∂ µ + igτ · Aµ /2.23). confirming (cf comment (c) in section 13. to be given by F µν = ∂ µ W ν − ∂ ν W µ − g W µ × W ν or. as usual. (A similar result holds in the spacetime coordinate transformation case.1. (13.79) suggests that we will find the SU(2) analogue of F µν by evaluating 1 [ D µ . we accordingly expect the SU(2) ‘curvature’.23) for the W µ fields. ‘it transforms under local SU(2) transformations in the way its SU(2) structure would indicate’. Then [ D µ .83) = ∂ µ Wiν − ∂ ν Wi − g µ µ ν i j k W j Wk . This non-Abelian F µν is a much more interesting object than the Abelian µν (which is actually U(1)-gauge invariant. Noting the analogy between the right-hand side of (13. in component notation.79).2. of course: F µν = F µν ). Now F µν has clearly three SU(2) components and must be an SU(2) triplet: indeed.3 confirms that the result is [ D µ .83) and the transformation law (13.20)–(13. way of obtaining the result is to consider the commutator of twocovariant derivatives. Fi µν (13.81) (13. D ν ]ψ ≡ ( D µ D ν − D ν D µ )ψ = ieF µν ψ (13. Problem 13. 1 1 1 (13.85) which is the expected law (cf (12. we are guaranteed that it itself is gauge covariant—that is to say. Copyright 2004 IOP Publishing Ltd . it is true that under an infinitesimal local SU(2) transformation F µν = F µν − g (x) × F µν (13. D ν ]ψ ( 2 ) = igτ /2 · (∂ µ W ν − ∂ ν W µ − g W µ × W ν )ψ ( 2 ) . F F µν contains the gauge coupling constant g .slightly simpler.84) This tensor is of fundamental importance in a (non-Abelian) gauge theory. Problem 13.4 verifies that (13. without the ∂ µ part which appears in δ W µ . Note particularly that F µν transforms ‘properly’.80) where.85) follows from (13. where the curvature tensor appears on the right-hand side.82) and (13.79) as is verified in problem 13.

It is now straightforward to move to the quantum field case and construct ˆ ˆ the SU(2) Yang–Mills analogue of the Maxwell Lagrangian − 1 Fµν F µν . and the preceding discussion of how to make a global SU(2) into a local one. This is the adjoint or regular representation of SU(3)—as we have now Copyright 2004 IOP Publishing Ltd .10). The infinitesimal version of (13. when acting on an SU(3) triplet ψ . But before proceeding in this way. It is 4 µν 1 ˆ ˆ simply − 4 F µν · F . in view of the transformation law (13.87) where A1 . as will be further discussed in chapter 14. By contrast. . which transforms under a local SU(3) transformation according to ψ = exp[igsλ · α(x)/2]ψ (13.90) Comparing (13.86) which is the same as the global transformation (12. 13. the quanta of which are called gluons.2.6 verifies that this is fulfilled by having the gauge fields transform by Aaµ = Aµ − ∂ µ ηa (x) − gs fabc ηb (x)Aµ . the three degrees of freedom of the fundamental quark triplet now referring to ‘colour’.14). . The SU(3)-covariant derivative.90) with (12. A8 are eight gauge fields. The coupling is denoted by ‘gs’ in anticipation of the application to strong interactions via QCD. This is the gauge group of QCD. it is straightforward to develop the corresponding theory of local SU(3). A2 . Problem 13. it is clear that ∂ µ ψ will involve an ‘unwanted’ term ∂ µ η(x). As in (13.86) is (cf (13. the η’s now depending on x. a c (13.74) but with the eight constant parameters α replaced by x -dependent ones.80) we can identify the term in f abc as telling us that µ the eight fields Aa transform as an SU(3) octet. . and with a coupling strength gs inserted.88) where ‘1’ stands for the unit matrix in the three-dimensional space of components of the triplet ψ . the SU(2) ‘dot product’ ensuring SU(2) invariance (see problem 13.13)) ψ = (1 + igs λ · η(x)/2)ψ (13.4 Local SU(3) symmetry Using what has been done for global SU(3) symmetry in section 12. of course. is given by the indicated generalization of (13.89) D µ ψ = (1 + igs λ · η(x)/2)D µ ψ without the ∂ µ η(x) term. the desired covariant derivative D µ ψ should transform according to (13.5).85). We denote the basic triplet by ψ . we first need to introduce local SU(3) symmetry. even under local transformation. . namely D µ (acting on SU(3) triplet) = ∂ µ + igs λ/2 · Aµ µ µ µ (13.

89). it develops no ‘∂ µ ηa ’ part but transforms as a ‘proper’ octet: µν Faµν = Fa − gs f abc ηb (x)Fcµν . Once again.91) can be ‘promoted’ into one which is covariant under local SU(3) transformations by replacing ∂ µ by D µ of (13. −igs λa /2γ µ .133)—but this time in colour space—is ‘dotted’ with the gauge field. under local SU(3) transformations. As in the SU(2) case.1 for SU(2) but we shall not need the result here.96) This allows us to write down a locally SU(3)-invariant analogue of the Maxwell Lagrangian µν − 1 Fa Faµν (13. The Feynman rule for figure 13. But it is just what is needed to cancel the corresponding ∂ µ η(x) term in ∂ µ ψ .7 is. leaving D µ ψ transforming as a proper triplet via (13.7. (iγµ ∂ µ − m)ψ = 0 (13.95) which is closely analogous to the SU(2) case (13.94) The SU(3) field strength tensor can be calculated by evaluating the commutator of two D ’s of the form (13.Figure 13.97) 4 Copyright 2004 IOP Publishing Ltd .3.87).90) can be derived as in section 13. therefore.93) as already suggested in section 12. Quark–gluon vertex.1: the SU(3) current of (12.7) is µν Fa = ∂ µ Aν − ∂ ν Aµ − gs f abc Ab Aν a a c µ (13.87): the result (problem 13.84) ( the structure constants of SU(2) are given by the i j k symbol and of SU(3) by f abc ).7) ¯ −igs ψf λ/2γ µ ψi · Aµ d4 x (13. The finite version of (13.39)). This leads immediately to the one gluon emission amplitude (see figure 13. (13. leading to / / (i∂ − m)ψ = gs λ/2 · Aψ (13. (13.92) (compare (13. the free Dirac equation for an SU(3)-triplet ψ . come to expect for gauge fields. the ∂ µ ηa (x) piece spoils this simple transformation property under local transformations. the crucial µν property of Fa is that. However.

by dotting the two octets together. It is now time to consider locally SU(2)- and SU(3)-invariant quantum field Lagrangians and, in particular, the resulting self-interactions among the gauge quanta. 13.5 Local non-Abelian symmetries in Lagrangian quantum field theory 13.5.1 Local SU(2) and SU(3) Lagrangians We consider here only the particular examples relevant to the strong and electroweak interactions of quarks: namely, a (weak) SU(2) doublet of fermions µ interacting with SU(2) gauge fields Wi and a (strong) SU(3) triplet of fermions µ interacting with the gauge fields Aa . We follow the same steps as in the U(1) case of chapter 7, noting again that for quantum fields the sign of the exponents in (13.28) and (13.86) is reversed, by convention; thus (12.89) is replaced by its local version q = exp(−ig α(x) · τ /2)q ˆ ˆ ˆ (13.98) and (12.132) by q = exp(−igsα(x) · λ/2)q. ˆ ˆ ˆ (13.99)

The globally SU(2)-invariant Lagrangian (12.87) becomes locally SU(2)ˆ invariant if we replaced ∂ µ by D µ of (13.10), with W µ now a quantum field: ¯ ˆ ˆ ÄD,local SU(2) = q(i D − m)q ˆ / ˆ

¯ / ¯ = q(i∂ − m)q − g qγ µ τ /2q · W µ ˆ ˆ ˆ ˆ ˆ


with an interaction of the form ‘symmetry current (12.109) dotted into the gauge field’. To this we must add the SU(2) Yang–Mills term ˆ ˆ ÄY−M,SU(2) = − 1 F µν · F µν 4 (13.101)

to get the local SU(2) analogue of ÄQED . It is not possible to add a mass term for ˆµ ˆ the gauge fields of the form 1 W · W µ , since such a term would not be invariant 2 under the gauge transformations (13.26) or (13.34) of the W-fields. Thus, just as in the U(1) (electromagnetic) case, the W-quanta of this theory are massless. We presumably also need a gauge-fixing term for the gauge fields, as in section 7.3, which we can take to be1

Ägf = −

1 ˆµ ˆν (∂µ W · ∂ν W ). 2ξ


1 We shall see in section 13.5.3 that in the non-Abelian case this gauge-fixing term does not

completely solve the problem of quantizing such gauge fields; however, it is adequate for tree graphs.

Copyright 2004 IOP Publishing Ltd

Figure 13.8. SU(2) gauge-boson propagator.

The Feynman rule for the fermion-W vertex is then the same as already given in (13.41), while the W-propagator is (figure 13.8) i[−g µν + (1 − ξ )k µ k ν / k 2 ] i j δ . k2 + i (13.103)

Before proceeding to the SU(3) case, we must now emphasize three respects in which our local SU(2) Lagrangian is not suitable (yet) for describing weak interactions. First, weak interactions violate parity, in fact ‘maximally’, by ˆ which is meant that only the ‘left-handed’ part ψL of the fermion field enters the ˆ ˆ interactions with the W µ fields, where ψL ≡ ((1 − γ5 )/2)ψ ; for this reason the weak isospin group is called SU(2)L . Second, the physical W± are, of course, not massless and, therefore, cannot be described by propagators of the form (13.103). And third, the fermion mass term violates the ‘left-handed’ SU(2) gauge symmetry, as the discussion in section 12.3.2 shows. In this case, however, the chiral symmetry which is broken by fermion masses in the Lagrangian is a local, or gauge, symmetry (in section 12.3.2 the chiral flavour symmetry was a global symmetry). If we want to preserve the chiral gauge symmetry SU(2)L –and it is necessary for renormalizability—then we shall have to replace the simple fermion mass term in (13.100) by something else, as will be explained in chapter 22. The locally SU(3)c -invariant Lagrangian for one quark triplet (cf (12.137))   fˆr qf =  fˆb  ˆ (13.104) fˆg where ‘f’ stands for ‘flavour’, and ‘r, b, and g’ for ‘red, blue and green’, is 1 ˆ 1 ¯ / ˆ µν ˆa ˆa ˆ q f (i D − m f )qf − Faµν Fa − (∂µ Aµ )(∂ν Aν ) ˆ ˆ 4 2ξ


ˆ ˆ where D µ is given by (13.87) with Aµ replaced by A and the footnote before equation (13.102) also applies here. This leads to the interaction term (cf (13.93)) ¯ ˆ ˆ ˆ −gsq f γ µ λ/2qf · Aµ (13.106)

and the Feynman rule (13.94) for figure 13.7. Once again, the gluon quanta must be massless and their propagator is the same as (13.103), with δi j → δab (a, b = 1, 2, . . . , 8). The different quark flavours are included by simply repeating the first term of (13.105) for all flavours: ¯ ˆ q f (i D − m f )qf ˆ / ˆ


Copyright 2004 IOP Publishing Ltd

which incorporates the hypothesis that the SU(3)c -gauge interaction is ‘flavourblind’, i.e. exactly the same for each flavour. Note that although the flavour masses are different, the masses of different ‘coloured’ quarks of the same flavour are the same (m u = m d , m u,r = m u,b = m u,g ). The Lagrangians (13.100)–(13.102) and (13.105), though easily written down after all this preparation, are unfortunately not adequate for anything but tree graphs. We shall indicate why this is so in section 13.5.3. Before that, we want to discuss in more detail the nature of the gauge-field self-interactions contained in the Yang–Mills pieces. 13.5.2 Gauge field self-interactions We start by pointing out an interesting ambiguity in the prescription for ‘covariantizing’ wave equations which we have followed, namely ‘replace ∂ µ by D µ ’. Suppose we wished to consider the electromagnetic interactions of charged massless spin-1 particles, call them X’s, carrying charge e. The standard wave equation for such free massless vector particles would be the same as for Aµ , namely £ X µ − ∂ µ∂ ν X ν = 0. (13.108) To ‘covariantize’ this (i.e. introduce the electromagnetic coupling), we would replace ∂ µ by D µ = ∂ µ + ie Aµ so as to obtain D 2 X µ − D µ D ν X ν = 0. (13.109)

But this procedure is not unique: if we had started from the perfectly equivalent free particle wave equation

£X µ − ∂ν∂µ Xν = 0
we would have arrived at D2 X µ − Dν Dµ X ν = 0 which is not the same as (13.109), since (cf (13.79)) [D µ , D ν ] = ieF µν .




The simple prescription ∂ µ → D µ has, in this case, failed to produce a unique wave equation. We can allow for this ambiguity by introducing an arbitrary parameter δ in the wave equation, which we write as D 2 X µ − D ν D µ X ν + ieδ F µν X ν = 0. (13.113)

The δ term in (13.113) contributes to the magnetic moment coupling of the X-particle to the electromagnetic field and is called the ‘ambiguous magnetic moment’. Just such an ambiguity would seem to arise in the case of the charged
Copyright 2004 IOP Publishing Ltd

weak interaction quanta W± (their masses do not affect this argument). For the photon itself, of course, e = 0 and there is no such ambiguity. It is important to be clear that (13.113) is fully U(1) gauge-covariant, so that δ cannot be fixed by further appeal to the local U(1) symmetry. Moreover, it turns out that the theory for arbitrary δ is not renormalizable (though we shall not show this here): thus, the quantum electrodynamics of charged massless vector bosons is, in general, non-renormalizable. However, the theory is renormalizable if—to continue with the present ¯ terminology—the photon, the X-particle and its anti-particle, the X, are the members of an SU(2)-gauge triplet (like the W’s), with gauge coupling constant e. This is, indeed, very much how the photon and the W± are ‘unified’ but there is a complication (as always!) in that case, having to do with the necessity for finding room in the scheme for the neutral weak boson Z0 as well. We shall see how this works in chapter 19: meanwhile we continue with this X–γ model. We shall show that when the X–γ interaction contained in (13.113) is regarded as a 3–X vertex in a local SU(2) gauge theory, the value of δ has to equal one; for this value the theory is renormalizable. In this interpretation, the X µ wavefunction µ µ µ µ 1 1 ¯ is identified with ‘ √ (X 1 + iX 2 )’ and X µ with ‘ √ (X 1 − iX 2 )’ in terms of components of the SU(2) triplet X i , while Aµ is identified with X 3 . Consider then equation (13.113) written in the form2 ˆ £X µ − ∂ν ∂µ Xν = V X µ where ˆ V X µ = − ie{[∂ ν (Aν X µ ) + Aν ∂ν X µ ] − (1 + δ)[∂ ν (Aµ X ν ) + Aν ∂ µ X ν ] + δ[∂ µ (Aν X ν ) + Aµ ∂ ν X ν ]} (13.115) and we have dropped terms of O(e2 ) which appear in the ‘D 2 ’ term: we shall come back to them later. The terms inside the { } brackets have been written in such a way that each [ ] bracket has the structure ∂(AX) + A(∂ X) (13.116)
2 µ 2 µ


which will be convenient for the following evaluation. The lowest-order (O(e)) perturbation theory amplitude for ‘X → X’ under ˆ the potential V is then ∗ ˆ (13.117) −i X µ (f)V X µ (i) d4 x. Inserting (13.115) into (13.117) clearly gives something involving two ‘X ’µ wavefunctions and one ‘A’ one, i.e. a triple-X vertex (with Aµ ≡ X 3 ), shown
2 The sign chosen for V here apparently differs from that in the KG case (4.133) but it does agree ˆ

when allowance is made, in the amplitude (13.117), for the fact that the dot product of the polarization vectors is negative (cf (7.84)).

Copyright 2004 IOP Publishing Ltd

Figure 13.9. Triple-X vertex.

in figure 13.9. To obtain the rule for this vertex from (13.117), consider the first [ ] bracket in (13.115). It contributes −i(−ie)
ν ¯∗ X µ (2){∂ ν ( X 3ν (3) X µ (1)) + X 3 (3)∂ν X µ (1)} d4 x


where the (1), (2), (3) refer to the momenta as shown in figure 13.9, and for reasons of symmetry are all taken to be ingoing; thus, X 3 (1) =
µ µ 3 exp(−ik 3

· x)


for example. The first term in (13.118) can be evaluated by a partial integration ¯∗ to turn the ∂ ν onto the X µ (2), while in the second term ∂ν acts straightforwardly µ (1). Omitting the usual (2π)4 δ 4 energy–momentum conserving factor, we on X find (problem 13.8) that (13.118) leads to the amplitude ie



(k1 − k2 ) ·



In a similar way, the other terms in (13.117) give −ieδ( and +ie(1 + δ)(
2 1


3 2

· k2 −



3 1

· k1 ) · k1 ).



3 1

· k2 −



3 2


Adding all the terms up and using the 4-momentum conservation condition k1 + k2 + k3 = 0 we obtain the vertex +ie{
1 · 2 (k 1 − k 2 ) · 3 + 2 · 3 (δk 2 − k 3 ) · 1 + 3 · 1 (k 3 − δk 1 ) · 2 }.



It is quite evident from (13.124) that the value δ = 1 has a privileged role and we strongly suspect that this will be the value selected by the proposed SU(2) gauge symmetry of this model. We shall check this in two ways: in the first, we consider a ‘physical’ process involving the vertex (13.124) and show how
Copyright 2004 IOP Publishing Ltd

(a )


Figure 13.10. Tree graphs contributing to X + d → X + d.

requiring it to be SU(2)-gauge invariant fixes δ to be 1; in the second, we ‘unpack’ ˆ µν ˆ the relevant vertex from the compact Yang–Mills Lagrangian − 1 X µν · X . 4 The process we shall choose is X + d → X + d where d is a fermion (which we call a quark) transforming as the T3 = − 1 component of a doublet under the 2 SU(2) gauge group, its T3 = + 1 partner being the u. There are two contributing 2 Feynman graphs, shown in figures 13.10(a) and (b). Consider, first, the amplitude for figure 13.10(a√ We use the rule of figure 13.1, with the τ -matrix combination ). τ+ = (τ1 + iτ2 )/ 2 corresponding to the absorption of the positively charged X √ and τ− = (τ1 − iτ2 )/ 2 for the emission of the X. Then figure 13.10(a) is
1 1 τ− i τ+ ¯ (−ie)2 ψ ( 2 ) ( p2 ) /2 /1 ψ ( 2 ) ( p1 ), / 2 p / 1 + k1 − m 2


where ψ( 2 ) =

u d


and we have chosen real polarization vectors. Using the explicit forms (12.25) for the τ -matrices, (13.125) becomes 1 i 1 ¯ (−ie)2 d( p2 ) √ /2 √ / d( p1 ). /1 − m 2 1 p 2 /1 + k (13.127)

We must now discuss how to implement gauge invariance. In the QED case of electron Compton scattering (section 8.6.2), the test of gauge invariance was that the amplitude should vanish if any photon polarization vector µ (k) was replaced by k µ —see (8.168). This requirement was derived from the fact that a gauge transformation on the photon Aµ took the form Aµ → A µ = Aµ − ∂ µ χ, so that, consistently with the Lorentz condition, µ could be replaced by µ = µ + βk µ (cf 8.165) without changing the physics. But the SU(2) analogue of the U(1)-gauge transformation is given by (13.26), for infinitesimal ’s, and although there is indeed an analogous ‘−∂ µ ’ part, there is also an additional part (with g → e in our case) expressing the fact that the X’s carry SU(2) charge. However, this extra part does involve the coupling e. Hence, if we were to make the full change corresponding to (13.26) in a tree graph of order e2 , the extra part
Copyright 2004 IOP Publishing Ltd

Figure 13.11. Tree graphs contributing to γ + X → γ + X.

would produce a term of order e3 . We shall take the view that gauge invariance should hold at each order of perturbation theory separately; thus, we shall demand that the tree graphs for X–d scattering, for example, should be invariant under µ → k µ for any . The replacement 1 → k1 in (13.127) produces the result (problem 13.9) i ¯ (−ie)2 d( p2)/2 d( p1 ) 2 (13.128)

where we have used the Dirac equation for the quark spinors of mass m . The term (13.128) is certainly not zero—but we must of course also include the amplitude for figure 13.10(b). Using the vertex of (13.124) with suitable sign changes of momenta, and the photon propagator of (7.119), and remembering that d has τ3 = −1, the amplitude for figure 13.10(b) is ie[

· ×

2 (k 1

−ig µν q2

+ k 2 )µ +

· (−δ k2 − k2 + k1 ) + 1 ¯ γν d( p1 ) , −ied( p2 ) − 2
2µ 1

1µ 2

· (k2 − k1 − δ k1 )] (13.129)

2 2 where q 2 = (k1 − k2 )2 = −2k1 · k2 using k1 = k2 = 0, and where the ξ -dependent ¯ p2 ) / d( p1) = 0. We now leave it as an q part of the γ -propagator vanishes since d( exercise (problem 13.10) to verify that, when 1 → k1 in (13.129), the resulting amplitude does exactly cancel the contribution (13.128), provided that δ = 1. ¯ Thus, the X–X–γ vertex is, assuming the SU(2)-gauge symmetry,




2 (k 1

− k2 ) ·





3 (k 2

− k3 ) ·






(k3 − k1 ) ·

2 ].


The verification of this non-Abelian gauge invariance to order e2 is, of course, not a proof that the entire theory of massless X quanta, γ ’s and quark isospinors will be gauge invariant if δ = 1. Indeed, having obtained the X– X–γ vertex, we immediately have something new to check: we can see if the lowest order γ –X scattering amplitude is gauge invariant. The X–X–γ vertex will generate the O(e2 ) graphs shown in figure 13.11 and the dedicated reader may check that the sum of these amplitudes is not gauge invariant, again in the
Copyright 2004 IOP Publishing Ltd

Figure 13.12. γ –γ –X–X vertex.

Figure 13.13. 4–X vertex.

(tree graph) sense of not vanishing when any is replaced by the corresponding k . But this is actually correct. In obtaining the X–X–γ vertex we dropped an O(e2 ) term involving the three fields A, A and X , in going from (13.115) to (13.124): this will generate an O(e2 ) γ –γ –X–X interaction, figure 13.12, when used in lowest-order perturbation theory. One can find the amplitude for figure 13.12 by the gauge-invariance requirement applied to figures 13.11 and 13.12, but it has to be admitted that this approach is becoming laborious. It is, of course, far more efficient to deduce the vertices from the compact Yang–Mills ˆ µν ˆ Lagrangian − 1 X µν · X , which we shall now do; nevertheless, some of the 4 physical implications of those couplings, such as we have discussed earlier, are worth exposing. ˆµ The SU(2) Yang–Mills Lagrangian for the SU(2) triplet of gauge fields X is ˆ ˆ ˆ Ä2,YM = − 1 X µν · X µν (13.131) 4 where ˆµ ˆν ˆν ˆµ ˆ µν X = ∂µ X − ∂ν X − X × X . ˆ ˆ ˆ ˆ ˆν ˆν − 1 (∂µ X ν − ∂ν X µ ) · (∂ µ X ) + e( X µ × X ν ) · ∂ µ X 2 ˆµ ˆ ˆµ ˆν ˆ ˆ − 1 e2 [( X · X µ )2 − ( X · X )( X µ · X ν )]. 4 (13.133) The X–X–γ vertex is in the ‘e’ term, the X–X–γ –γ one in the ‘e2 ’ term. We give the form of the latter using SU(2) ‘i, j, k’ labels, as shown in figure 13.13: − ie2 [ i j + in +
Copyright 2004 IOP Publishing Ltd
im mn


ˆ Ä2,YM can be unpacked a bit into



1 1

3 2

jm nj

( (

· ·


− 1 · 4 2 · 3) 3 · 4 − 1 · 3 2 · 4)


4 2





2 3


4 )].


The reason for the collection of terms seen in (13.130) and (13.134) can be understood as follows. Consider the 3–X vertex k2 ,

j ; k3,

ˆ 3 , k|e( X µ

ˆ ˆν × X ν ) · ∂ µ X |k1 ,

1, i


ˆ for example. When each X is expressed as a mode expansion and the initial and final states are also written in terms of appropriate a’s and a † ’s, the amplitude ˆ ˆ will be a vacuum expectation value (vev) of six a’s and a † ’s; the different terms ˆ ˆ in (13.130) arise from the different ways of getting a non-zero value for this vev, by manipulations similar to those in section 6.3. We end this chapter by presenting an introduction to the problem of quantizing non-Abelian gauge field theories. Our aim will be, first, to indicate ˆ where the approach followed for the Abelian gauge field Aµ in section 7.3.2 fails and then to show how the assumption (nevertheless) that the Feynman rules we have established for tree graphs work for loops as well, leads to violations of unitarity. This calculation will indicate a very curious way of remedying the situation ‘by hand’, through the introduction of ghost particles, only present in loops. 13.5.3 Quantizing non-Abelian gauge fields We consider for definiteness the SU(2)-gauge theory with massless gauge fields ˆµ W (x), which we shall call gluons, by a slight abuse of language. We try to carry through for the Yang–Mills Lagrangian ˆ ˆ ˆ Ä2 = − 1 F µν · F µν 4 where ˆ ˆ ˆ ˆ ˆ F µν = ∂µ W ν − ∂ν W µ − g W µ × W ν , (13.136) (13.137)

the same steps we followed for the Maxwell one in section 7.3.2. We begin by re-formulating the prescription arrived at in (7.116), which we reproduce again here for convenience: ˆ ˆ ˆ Äξ = − 1 Fµν F µν − 4 ˆ Äξ leads to the equation of motion 1 ˆ ˆ ˆ £ Aµ − ∂ µ∂ν Aν + ξ ∂ µ∂ν Aν = 0. (13.139) 1 ˆ (∂µ Aµ )2 . 2ξ (13.138)

This has the drawback that the limit ξ → 0 appears to be singular (though the propagator (7.119) is well behaved as ξ → 0). To avoid this unpleasantness, consider the Lagrangian (Lautrup 1967) ˆ ˆ ˆ ˆ ˆ ˆ Äξ B = − 1 Fµν F µν + B∂µ Aµ + 1 ξ B 2 4 2
Copyright 2004 IOP Publishing Ltd


140) by means of (13. Copyright 2004 IOP Publishing Ltd (13. We may think of the ‘ B∂ · A’ term as a field-theory analogue of the procedure followed in classical Lagrangian mechanics. and (12. The obvious analogue of (13.142)) ˆ ˆ ˆ ( D ν )i j F j µν + ∂µ Bi = 0 (13. (13. which implies that ˆ ˆ ( D µ )ki ∂µ Bi = 0. (13.145) gives (cf (13.147) ∂µ W + ξ B = 0. Equation (13. it is also essential to prove that all physical results are independent of the gauge parameter ξ .136). Similarly.143) ˆ Eliminating B from (13. Taking ˆ ˆ ∂µ of (13. so that ∂ A µ is also a free massless ˆ £ to (13.144) Kugo and Ojima (1979) have shown that (13. k are the SU(2) matrix indices. It is possible to verify that ˆ ˆ ˆ ( D µ )ki ( D ν )i j F j µν = 0 where i.142) (13. we recover (13.ξ B = − 1 F µν · F µν + B · (∂µ W µ ) + 1 ξ B · B 4 2 (13. In this formulation.ˆ ˆ ˆ where B is a scalar field.143) becomes ˆ ˆµ (13.138).144) provides a satisfactory definition of the Hilbert space of states. so that B is a free massless field.44) with t = 1.140) would be to consider ˆ ˆ ˆ ˆ ˆ ˆ ˆ Ä2.108)) ˆ B (+) (x)| = 0. j.148) (13. the appropriate subsidiary condition for getting rid of the unphysical (non-transverse) degrees of freedom is (cf (7.141) π0 = B ˆ ˆ while the Euler–Lagrange equations for Aµ read as ˆ ˆ ˆ £ Aµ − ∂ µ ∂ν Aν = ∂ µ B ˆ and for B yield ˆ ˆ ∂µ Aµ + ξ B = 0.142) we learn that £ B = 0. Applying µ = 0. In addition to this.146) where the covariant derivative is now the one appropriate to the SU(2) triplet F µν (see (13.145) ˆ where B is an SU(2) triplet of scalar fields. We now try to generalize the foregoing in a straightforward way to (13.143). j are the SU(2) labels.143) then shows that £∂µ A µ ˆ field.149) . whereby ˆ a constraint (in this case the gauge-fixing one ∂ · A = 0) is brought into the ˆ Lagrangian with a ‘Lagrange multiplier’ (here the field B). The momentum ˆ conjugate to A 0 is now ˆ (13.48)) and i. (13.

As we saw in section 7. This relation determines the imaginary part of an amplitude via an equation of the form (cf (11. i. and the sum is over a complete set of physical intermediate states |n .108). We will see that this leads to physically unacceptable consequences. Now part of the power of the ‘Feynman rules’ approach to perturbation theory is that it is manifestly covariant.63)) 2 Im f|Å|i = n f|Å|n n|ņ |i dρn (13. Consider. dρn represents the phase space element for the general intermediate state |n . for example.144).147)) is ∂µ W . Consider now the possibility of gauge quanta appearing in the states |n . which can enter at the given energy.144) must hold for all times. they will not be restricted to having only two polarization states (as we shall see explicitly in a moment). this is a problem which threatens all gauge theories if the gauge field is treated covariantly.3. which has no counterpart at all in the physical unitarity relation.108) or (13.14): q 2 Im q¯ |Å|q¯ >= q q¯ |Å|gg gg|ņ|q¯ dρ2 q q (13. there is a real chance that when the imaginary part of such graphs is calculated. In fact.1).e. its violation is a serious disease indeed.3. the process q¯ → q¯ (where the ‘quarks’ are an q q SU(2) doublet) whose imaginary part has a contribution from a state containing two gluons (figure 13. and only if the field is free is its time variation known (and essentially trivial). but as we have seen the analogous conditions will not work in the non-Abelian case. this introduces unphysical degrees of freedom which must somehow be eliminated from the theory—or at least prevented from affecting physical processes.ˆ This is the crucial result: it implies that the auxiliary field B is not a free field in ˆµ this non-Abelian case and so neither (from (13. In consequence.150) where f|Å|i is the (Feynman) amplitude for the process i → f. But there is no completely covariant way of selecting out just the two physical components of a massless polarization vector µ . The reason is that a condition like (13. so that unitarity will not be satisfied. Since unitarity is an expression of conservation of probability. a contribution from the unphysical polarization states will be found. for example in the ‘sum over intermediate states’ which arises in the unitarity relation. such quanta can have only the two degrees of freedom (polarizations) allowed for a physical massless gauge field (cf section 7. Since unitarity deals only with physical states. the obvious generalizations of (7. In QED we do this by imposing the condition (7. or (13. In fact. from the four originally introduced precisely for reasons of covariance. namely to the violation of unitarity. as a 4-vector. when gauge quanta appear as virtual particles in intermediate states in Feynman graphs. Let us press ahead nevertheless and assume that the rules we have derived so far are the correct Feynman rules for this gauge theory. Hence.2.144) cannot be used to define the physical (transverse) states.151) Copyright 2004 IOP Publishing Ltd . and so unphysical states may make their presence felt.

151) is then to be performed over λ1 = 1. the coefficients of corresponding powers on each side should be equal. (ki . λi = −1) = (0.153).79): (ki . both sides are consistently q 4 .151) must have the form The two-gluon ŵ1 ν1 µ1 ν1 1 (k 1 . 2 and λ2 = 1. 1.λ2 =1. 0.14. which we call ŵν .153) may be tested is for the fourth-order amplitude (4) (4) Åq¯ →q¯ .15: all contain a loop.151) becomes 2 Im Åq¯ →q¯ = q q ׊ŵ1 ν1 µ1 ν1 1 (k 1 . The sum in (13. λ) is the polarization vector for the gluon with polarization λ and 4-momentum k . Since each emission or absorption of a gluon produces one power of the SU(2) coupling g .153) For later convenience we are using real polarization vectors as in (7. thus.2. λ2 ) λ1 =1. We first note that we want the unitarity relation (13. (13. of 2 2 course. Thus the lowestorder process in which (13.78) and (7.16: when of order g µν (4) these are placed in (13. and so each must represent the O(g 2 ) (2) contribution to q¯ → gg. k1 = k2 = 0. amplitudes in (13. 2 which are the physical polarization states (cf section 7. 0). each Å involves two polarization vectors. λ2 ) (13. λ1 ) 2 (k 2 .153) to be satisfied order by order in perturbation theory: that is to say.2 µ2 ν2 ∗ µ2 ν2 1 (k 1 .153) will hold when the Å’s represent some suitable Feynman graphs.152) where µ (k. We now wish to find out whether or not a result of the form (13. some of which q q q q are shown in figure 13.1). λ1 ) 2 (k 2 . There are three contributions to Å(2) shown in figure 13.Figure 13. Thus (13.153). 1. There are quite a number of contributions to Åq¯ →q¯ . Two-gluon intermediate state in the unitarity relation for the amplitude for q¯ → q¯ . q q where dρ2 is the two-body phase space for the g-g state. λ1 ) 2 (k 2 .153) involves at least the power g 4 . which should agree with the imaginary part of the total O(g 4 ) loopgraph contribution. On the right-hand side of (13. 0) and. contributions to the imaginary part of Åq¯ →q¯ are q q generated. λi = +1) = (0. the right-hand side of (13. λ2 ) dρ2 . Copyright 2004 IOP Publishing Ltd . Let us see if this works out. when the Å’s on both sides are expanded in powers of the coupling strengths (as in the usual Feynman graph expansion). 0.3.

instead of varying freely as the loop momentum varies.16. The extraction of the imaginary part of a Feynman diagram is a technical matter.15.g. q q Figure 13. According to the rules for propagators and vertices (4) already given.Figure 13. together with appropriate ‘γ ’ and ‘τ ’ vertex factors. the propagator is replaced by a condition stating that. q We choose to work in the gauge ξ = 1. O(g 2 ) contributions to q¯ → gg. the gluon’s mass is constrained to have the physical (free-field) value of zero. each of the loop amplitudes Åq¯ →q¯ (e. one replaces each gluon propagator of momentum k by π(−g µν )δ(k 2 )θ (k0 )δi j .15. to compute the imaginary part of the amplitudes of figure 13. the whole being integrated over the loop momentum. and its energy is positive.154) That is. those of figure 13.9) and in the present case the result is that. Rules for doing this exist (Eden et al 1966. section 2.155) Copyright 2004 IOP Publishing Ltd . so that the gluon propagator takes the familiar form −ig µν δi j /k 2 . so that eventually Im Åq¯ →q¯ = q q (4) (2) (2) ŵ1 ν1 (−g µ1ν1 )ŵ2 ν2 (−g µ2 ν2 ) dρ2 (13. in evaluating the imaginary part of the diagram. having to do with careful consideration of the ‘i ’ in the propagators.15) q q will be proportional to the product of the propagators for the quarks and the gluons. Some O(g 4 ) contributions to q¯ → q¯ . These conditions (one for each gluon) have the effect of converting the loop integral with a standard two-body phase space integral for the gg intermediate state. (13.

λ) (13. Taylor 1976. with all external legs satisfying the ‘mass-shell’ conditions. λ) µ (k. we have to investigate whether this difference matters.153) by the gluon polarization sum µ P µν (k) ≡ (k. λ2 respectively. it is precisely the fact that any given µ satisfying (13. (4) So.155).161) Copyright 2004 IOP Publishing Ltd . In order to calculate (13. the imaginary part of the loop contribution to Åq¯ →q¯ does seem to have q q the form (13. now two further difficulties have appeared: namely. k2 and λ = λ1 . This certainly fixes µ and enables us to calculate (13. A standard choice (see.158) where t is some 4-vector. λ) ν (k.160) (13. We note that whereas k µ Pµν = k ν Pµν = 0 (13. with |n the gg intermediate state as in (13.156).) Indeed. using two ’s specified by the conditions (cf (7.157) and (13.157) The trouble is that conditions (13.155) is taken in (13. (13. of course.155) and (13. we can calculate (13. so as to fix it uniquely. But there is one essential difference between (13.g.157) with the further condition t· =0 (13. λ ) = −δλλ · k = 0.157) can be replaced by µ + λkµ that both reduces the degrees of freedom to two (as we saw in section 7.16. the physical results seem to depend on tµ .158)..153): the place of the factor −g µν in (13.159) is to be seen in (13.156) λ=1.84)) µ (k.159) But only the first term on the right-hand side of (13.156) but. We might think of calculating the necessary sum over λ by brute force.154). it is helpful to have an explicit expression for P µν . t) ≡ − gµν − Pµν = [t 2 kµ kν − k · t (kµ tν + kν tµ )]/(k · t)2 .157). pp 14–15) is to supplement (13. (13.150) as required by unitarity.153). Thus.(2) where ŵ1 ν1 is the sum of the three O(g 2 ) tree graphs shown in figure 13. finding (problem 13.3. e. and have we not lost Lorentz covariance because the theory involves a special 4-vector tµ ? Setting these questions aside for the moment. To proceed further.2 for k = k1 .11) Pµν = −gµν − [t 2 kµ kν − k · t (kµ tν + kν tµ )]/(k · t)2 .156) using the conditions (13. (Note the δ(k 2 ) in (13. A crucial quantity is clearly Uµν (k. we need to put another condition on µ .1) and evinces the essential arbitrariness in the µ specified only by (13.157) do not fix the ’s uniquely if k 2 = 0.

neither of the gluons. The discrepancy between (13. working to this given order in perturbation theory. it can be interpreted here also as expressing gauge invariance in the nonAbelian case. Uµν characterizes the contribution of these unphysical polarization states. there is one crucial difference here. Indeed. Condition (13. in fact.155) and (13.164) must be true for each gluon. k µ Uµν = −kν (13. λ2 ) = g 2 v( p2 ) ¯ τj 1 τi a1i /1 u( p1) / a2 j / 2 2 p / 1 − k1 − m 2 τj τi 1 a2 j /2 u( p1) ¯ + g 2 v( p2) a1i /1 / 2 p / 1 − k2 − m 2 + (−i)g 2 ki j [( p1 + p2 + k1 )ν1 g µ1 ρ + (−k2 − p1 − p2)µ1 g ρν1 τk −1 + (−k1 + k2 )ρ g µ1 ν1 ] 1µ1 a1i a2 j 2ν1 v( p2 ) γρ u( p1 ) ¯ ( p 1 + p 2 )2 2 (13. As discussed there. satisfies · k = 0—see the sentence following (13.163) together with similar terms involving one P and one U . the same is not true of k µ Uµν —in fact. t2 )] dρ2 1 2 (13. This will also ensure that amplitudes are independent of tµ . therefore. the diagrams in figure 13.(from the condition k · = 0).164) and similarly for k2 . by contrast.16 are essentially ‘crossed’ versions of those in figure 13.163). already recalled in the previous section. their polarizations satisfying the condition · k = 0.165) Copyright 2004 IOP Publishing Ltd . (13.162). in the discrepant contribution (13. even when the other gluon has · k = 0. In physical terms.2.165). λ1 ) 2 (k 2 .3. In figure 13.10. we shall now see that whereas the (crossed) version of (13.153) is then (4) Im Åq¯ →q¯ = q q ( ( ŵ2)ν1 [U µ1 ν1 (k1 )]ŵ2)ν2 [U µ2 ν2 (k2 . And.1). It follows that these unwanted contributions will.164) did hold for our dX → dX amplitudes of section 13. In figure 13.162) where we have used k 2 = 0. Thus.3. both the X’s were physical.16 together yield (2) ŵ1 ν1 µ ν1 1 (k 1 . vanish if (2) k1 1 ŵ1 ν1 = 0 µ (13.10. the crucial point is that (13. It follows that Uµν may be regarded as including polarization states for which · k = 0.164) is apparently the same as the U(1)-gauge-invariance requirement of (8. However. a gluon appearing internally in a Feynman graph has to be regarded as existing in more than just the two polarization states available to an external gluon (cf section 7. The three graphs of figure 13.16.164) fails for states with · k = 0. in fact.

Such quantities can be calculated from expression (13. τ j /2]u( p1)a1i a2 j = ig 2 v( p2 )/2 ¯ i j k (τk /2)u( p1 )a1i a2 j (13. in fact. 0.166) using the SU(2) algebra of the τ ’s. more) gluons. This choice obviously satisfies (13. if constructed according to the simple rules for tree diagrams. 1 = k1 . 0.170) to vanish. 0. Let us work in the centre of momentum (CM) ¯ frame of the two gluons. −|k|). −|k|).167) (13. The third term in (13.168) can be removed.160) for the sum over unphysical polarization states Uµν and make a specific choice for t. Copyright 2004 IOP Publishing Ltd .168) 2k1 · k2 We see that the first part (13. by k1 . 0. since k2 = 0. which only vanishes if k2 · 2 = 0.where we have written the gluon polarization vectors as a product of a Lorentz 4-vector µ and an ‘SU(2) polarization vector’ ai to specify the triplet state label. 0. 2 = k2 for the first. with k1 = (|k|. say. We conclude that loop diagrams involving two (or. The correct rule for loops must be such as to satisfy unitarity. k) = (kµ kν + kν kµ )/(2|k|2 ) and unitarity (cf (13.167) certainly does cancel (13. 0. we return to expression (13. From (13. 0. |k|).128)) ¯ g 2 v( p2 )/2 [τi /2. which will cancel the contributions of unphysical polarization states. To get an intuitive idea of what such extra terms might be.165) gives − ig 2 + ig 2 ijk ¯ i j k v( p2 )/2 (τk /2)u( p1 )a1i a2 j / v( p2 )k 1 (τk /2)u( p1 )k2 · ¯ 2 a1i a2 j . and consider for definiteness the µ ν µ ¯ν (2) (2) (2) contractions with the ŵ1 ν1 term.165) by setting 1 = ¯ k1 . We take ¯ ¯ ¯ tµ = kµ . the ¯ first two terms reduce to (cf (13. and so on. where the 4-vector k is defined by k = (−|k|. |k|). we must infer that the rule for loops will have to involve some extra term.168). This is not sufficient to guarantee the absence of all unphysical contributions to the imaginary part of the two-gluon graphs. over and above the simple tree-type constructions.166) but there remains the second piece (13.168) it is clear that a 2 term in which 2 is replaced by k2 as well as 1 by k1 will vanish. Then ¯ ¯ ¯ Uµν (k. (13.158).169) Å (2) µ1 ν1 Å µ1 ¯ µ2 (2) (k 1 k 1 µ2 ν2 ¯ ¯ ¯ + k1 2 k1 1 ) (k21 k22 + k22 k21 ) dρ2 2|k1 |2 2|k2 |2 µ µ ν ν ν ν (13. ŵ1 ν1 k1 1 k21 etc. k2 = (|k|. will violate unitarity. |k|). as the preceding discussion shows. We have already obtained the result of putting 1 = k1 . k1 = ¯ (−|k|. or terms. 0. k) and k = (0.163)) requires (13. Since there seems no other way in which the offending piece in (13. k2 = (−|k|. Using the Dirac equation for u( p1) and v( p2). These are ŵ1 ν1 k1 1 k21 . Now replace 1 . but it does not. 2 = k2 for the second. 0.

where the dotted lines represent the scalar particles.172). For this cancellation to work. Thus.172) jµk = g v( p2)γµ (τk /2)u( p1 ) ¯ (13. which in retrospect was the cause of all the trouble.172) can be interpreted as the amplitude for the tree graph shown in figure 13.17 and these can be arranged to cancel the unphysical polarization pieces like (13.168) this reduces to ijk / v( p2 )k 1 (τk /2)u( p1 )a1i a2 j ¯ (13.18 must enter with the opposite sign from the three-gluon loop graph of figure 13. in fact.15. the scalar particle loop graph of figure 13.18. that the fourth-order graph we are looking for has the form shown in figure 13.15.172).172).171) as jµk where −g µν δk ig (k1 + k2 )2 ij a1i a2 j k1ν (13. Tree-graph interpretation of the expression (13.3 and Copyright 2004 IOP Publishing Ltd .172) can be eliminated if we adopt the following rule (on the grounds of ‘forcing the theory to make sense’). is of just the right magnitude to cancel all the contributions to (13. the gluon propagator appears in Feynman gauge ξ = 1 and the rest of the expression would have the interpretation of a coupling between the intermediate gluon and two scalar particles with SU(2) polarizations a1i . From (13. It seems plausible.172) has the appearance of a one-gluon intermediate state amplitude.Figure 13. The new scalar particles must be massless. µ ¯ν (2) A typical non-vanishing term is of the form ŵ1 ν1 k1 1 k21 /2| k|2 .173).18 is calculated in the usual way. moreover. When the imaginary part of figure 13. constructed according to the simple ‘tree’ prescriptions.173) is the SU(2) current associated with the q¯ pair.171) −i g 2 2k 1 · k 2 ¯ using k2 · k2 /2| k|2 = −1. it will involve contributions from the tree graph of figure 13. We may rewrite (13. Such a relative minus sign between single closed-loop graphs would be expected if the scalar particles in figure 13. In addition to the fourth-order diagrams of the type shown in figure 13. The q¯ → g vertex is q represented by the current (13. a2 j . there must exist a previously unknown fourth-order contribution. only present in loops.17. Now (13. therefore. (13.17. such that it has an imaginary part which is non-zero in the same physical region as the twogluon intermediate state and.170) from terms like (13. q The unwanted terms of the form (13. fermions! (Recall the rule given in section 11.18 were. so that this new amplitude has an imaginary part in the same physical region as the gg state.

been entirely heuristic.3 Using the expression for D µ in (13. in appendix Q. Such a term indeed exists and was first derived using the path integral form of quantum field theory (see chapter 16) by Faddeev and Popov (1967). 1977). but obeying anti-commutation relations: the covariant derivative is the one appropriate for an SU(2) triplet. The result is that the covariant gauge-fixing term (13. namely (from (13. Ghost loop diagram contributing in fourth order to q¯ → q¯ . We give the rules for tree diagrams. Clearly what is required is some additional term in the Lagrangian which will do the job in general.26) in the infinitesimal case. But. 13. The preceding discussion has. Such particles are called ‘ghosts’. in reality the ghosts are always confined to loops. unitarity alone is not a sufficient constraint to provide the prescription for more than one closed gluon loop.2. The result (13.44) and (12. 13.175) in this case. Problems 13. We should add the caution that the form of the ghost Lagrangian depends on the choice of the gauge-fixing term: there are gauges in which the ghosts are absent. The complete Feynman rules for non-Abelian gauge field theories are given in Cheng and Li (1984). Peskin and Schroeder (1995) or Ryder (1996). their function being to cancel unphysical contributions from intermediate gluons.102) must be supplemented by the ‘ghost Lagrangian’ ˆ Äg = ∂µ ηi† Diµj η j ˆ ˆ ˆ (13.174) where the η field is an SU(2) triplet and spinless. verify (13. for example Cheng and Li (1984). q q problem 11.) Thus.Figure 13.79). for which there are no problems with ghosts.174) is derived in standard books of quantum field theory.81).17. which apparently involves ghosts as external lines.1 Verify that (13. It can be followed through so as to yield the correct prescription for eliminating unphysical contributions from a single closed gluon loop.18. we appear to need scalar particles obeying Fermi statistics. as Feynman recognized (1963. for example.48)) ˆµ Di j = ∂ µ δ i j + g ˆµ ki j Wk (13. Copyright 2004 IOP Publishing Ltd .2 Verify equation (13.82). We must emphasize that although we have introduced the tree graph of figure 13. of course.34) reduces to (13.

13. Copyright 2004 IOP Publishing Ltd .95). 13.89). 13. verify (13. 13.5 Verify that F µν · F µν is invariant under local SU(2) transformations.120) (omitting the (2π)4 δ 4 factors). 13. provided that δ = 1.118) reduces to (13.156).158).8 Verify that (13.6 Verify that the (infinitesimal) transformation law (13.127) leads to (13. 13.85) of F µν under local SU(2) transformations. 13. the resulting amplitude cancels the contribution (13.10 Verify that when 1 is replaced by k1 in (13.129). with the ’s specified by the conditions (13.159).7 By considering the commutator of two D µ ’s of the form (13. is given by (13.9 Verify that the replacement of 1 by k1 in (13.128). 13.157) and (13.11 Show that P µν of (13.90) for the SU(3) gauge µ field Aa is consistent with (13.4 Verify the transformation law (13.128).87).


This is a consequence of a fundamental property—possessed only by non-Abelian gauge theories—whereby the effective interaction strength becomes progressively smaller in such regimes. In sections 14. and to the necessary techniques. the lowest-order perturbation theory amplitudes (tree graphs) provide a very convincing qualitative. and the QCD Lagrangian. We briefly review the evidence for the ‘colour’ degree of freedom in section 14. The main new concept will be the renormalization group (and related ideas). QCD will be our first application of the theory developed in chapter 13. and much of chapter 16.2. or even ‘semi-quantitative’.14 QCD I: INTRODUCTION AND TREE-GRAPH PREDICTIONS In the previous chapter we have introduced the elementary concepts and formalism associated with non-Abelian quantum gauge field theories. This theory is called quantum chromodynamics or QCD for short. However. most of the quantitative experimental support for QCD now comes from comparison with predictions which include higher-order QCD corrections. This property is called ‘asymptotic freedom’ and has already been mentioned in section 11. in section 14. This immediately involves all the apparatus of renormalization. and we shall devote the next two chapters. despite its being a theory of the strong interactions. acting on a degree of freedom called ‘colour’ (indicated by the subscript c). In the present chapter we introduce QCD and discuss some of its simpler experimental consequences. There are now many indications that the strong interactions between quarks are described by a theory of this type. by building on the discussion of renormalization given in chapters 10 and 11 of volume 1. short distances or high energies—in which it is effectively a quite weakly interacting theory. indeed. to it. and then proceed to the dynamics of colour.1. The necessary calculations quite rapidly become too technical for the intended scope of this book but in chapter 15 we shall try to provide an elementary introduction to the issues involved. Perhaps the most remarkable thing about the dynamics of QCD is that. therefore.3 of volume 1. the asymptotic freedom property itself emerges from summing a whole class of higher-order contributions. which Copyright 2004 IOP Publishing Ltd .3 and 14. orientation to the data. In appropriate cases. in which the gauge group is an SU(3)c . as we shall indicate at the beginning of chapter 15.5.4 we shall see how the tree-graph techniques acquired for QED in volume 1 produce more useful physics when applied to QCD. there are certain kinematic regimes—roughly speaking.

We introduce a colour wavefunction with colour index α: ψα (α = 1. by revisiting it in coordinate (rather than momentum) space. 3). (14.1 quarks. We are here writing the three labels as ‘1. In this regime. the wavefunction ψ3q should be anti-symmetric with respect to quark interchange. which involves replacing the spacetime continuum by a discrete lattice of points. the original non2 relativistic quark model wavefunction took the form ψ3q = ψspace ψspin ψflavour . 3’ but they are often referred to by colour names such as ‘red. namely the method of ‘lattice’ QCD. But we shall see that. In chapter 16 we work towards understanding some non-perturbative aspects of QCD. Chapter 16 therefore serves as a useful ‘retrospective’ on our conceptual progress thus far. At first sight.2. called colour. 14. For a baryon made of three spin. it should be understood that this is merely a picturesque way of referring to the three basic states of this degree of freedom and has nothing to do with real colour! With the addition of this Copyright 2004 IOP Publishing Ltd . the ++ state mentioned in + section 1. An alternative. with respect to which the 3q wavefunction can be antisymmetrized. it provides some powerful new insights into several aspects of quantum field theory in general. Dalitz 1965) that the product of these space. For example. spin and flavour wavefunctions for the ground-state baryons was symmetric under interchange of any two quarks. But we saw in section 7. approach is available however.g. unrevealed degree of freedom of matter came from baryon spectroscopy (Greenberg 1964. this may seem to be a topic rather disconnected from everything that has preceded it. green’. As a natural concomitant of asymptotic freedom. and in particular of renormalization. in fact. it is to be expected that the effective coupling strength becomes progressively larger at longer distances or lower energies. 2. ultimately being strong enough to lead (presumably) to the confinement of quarks and gluons: this is sometimes referred to as ‘infrared slavery’.e.3 is made of three u quarks (flavour symmetric) in the J P = 3 an essential tool in the modern confrontation of perturbative QCD with data. as follows. 2. 2 which has zero orbital angular momentum and is hence spatially symmetric and a symmetric S = 3 spin wavefunction. Some of the simpler predictions of the renormalization group technique will be compared with experimental data in the last part of chapter 15. A simple way of implementing this requirement is to suppose that the quarks carry a further degree of freedom. purely numerical. perturbation theory clearly fails. see also Han and Nambu 1965 and Tavkhelidze 1965). blue.1 The colour degree of freedom The first intimation of a new.1) It was soon realized (e.2 that quantum field 2 theory requires fermions to obey the exclusion principle—i.

There is an indication that the data tend to lie Copyright 2004 IOP Publishing Ltd . as we saw in section 9. b with respective charges 2 .7) for the two cases.2) and this must then be multiplied into (14. at high energies the ratio R= σ (e+ e− → hadrons) σ (e+ e− → µ+ µ− ) (14.3) √ (d1 u 1 + d2 u 2 + d3 u 3 ) 3 √ † which we write in general as (1/ 3)dα u α . As far as meson (¯ q) states are concerned. − 1 . d. To date.2.2) is actually the only way of making an anti-symmetric combination of the three ψ ’s. all known baryon states can be described in this way. In any process which we can describe in terms of creation or annihilation of quarks.1) is symmetric overall. For example.2).degree of freedom.5.4) will. namely ψ3q. the symmetry of the ‘traditional’ space–spin–flavour wavefunction (14. We shall shortly see the grouptheoretical significance of this ‘neutral superposition’. The data (figure 14. Meanwhile. reflect the magnitudes of the individual quark couplings to the photon: R= a 2 ea (14. we can certainly form a three-quark wavefunction which is anti-symmetric in colour by using the anti-symmetric symbol αβγ . − 1 . It is reassuring that there is only one way of doing this—otherwise. (14. this yields 3 3 3 3 3 Rno colour = and Rcolour = 11 3 11 9 (14. This would seem a somewhat artificial device unless there were some physical consequences of this increase in the number of quark types—and there are. we note that (14. in the quark parton model (see section 9. 2 . s. what was previously a π + wavefunction d ∗ u is q now 1 ∗ ∗ ∗ (14. As we shall see in section 14.1) to give the full 3q wavefunction. we would have obtained more baryon states than are physically observed.1) rule out (14.3) is also a colour singlet combination. the multiplicity of quark types will enter into the relevant observable cross-section or decay rate.5). For five quarks u. and of (14.6) (14.6) and are in good agreement with (14. it is therefore called a (colour) singlet.1. c. − 1 .colour = αβγ ψα ψβ ψγ (14. i.7) at energies well above the b threshold and well below the Z0 resonance peak. while the required anti-symmetry is restored by the additional factor (14.e.5) where a runs over all quark types.2).

τ − decays proceed via the weak process shown in figure 14. consider the branching fraction for τ − → e− νe ντ (i. µ− νµ . In chapter 9 we also discussed the Drell–Yan process in the quark parton model: it involves the subprocess q¯ → l¯ which is the inverse of the one in q l (14.4)) (Montanet et al 1994).2. For example. the last with multiplicity Nc . µ.2.Figure 14.1. as will be discussed in section 15. τ decay. or ud. The ratio R (see (14. Similarly. A number of branching fractions also provide simple ways of measuring the number of colours Nc . the ratio of the rate for τ − → e− νe ντ to that for all other ¯ ¯ decays). Experiment gives a ¯ value of 10. so again Nc ≈ 3. above the parton model prediction: this is actually predicted by QCD via higherorder corrections.e.4). where the ¯ ¯ ¯ final fermions can be e− νe .1. Figure 14. u and c).8) Experiments give B ≈ 18% and hence Nc ≈ 3. τ. Thus B(τ − → e− νe ντ ) ≈ ¯ 1 . 2 + Nc (14. We mentioned that a factor of 1 appears in this case: this arises because 3 we must average over the nine possible initial q¯ combinations (factor 1 ) and q 9 Copyright 2004 IOP Publishing Ltd .7%. the branching fraction 1 B(W− → e− νe ) is ∼ 3+2Nc (from f ≡ e.

1 Colour as an SU(3) group We now want to consider the possible dynamical role of colour—in other words. What symmetry group is involved? We shall consider how some empirical facts suggest that the answer is SU(3)c . They must all have the same mass or else we would observe some ‘fine structure’ in the hadronic levels. then sum over the number of such states that lead to the colour neutral photon. ¯ ¯ which is three (¯ 1q1 . the way in which the forces between quarks depend on their colours. . We have seen that we seem to need three different quark types for each given flavour. the parton model gives a good first approximation to the data. the original calculation of this rate by Steinberger (1949) used a model in which the proton and neutron replaced the u and d in the loop.Figure 14. and using quark distribution q functions consistent with deep inelastic scattering.9) needs to be multiplied by nine. Finally. The amplitude is proportional to the square of the quark charges but because the π 0 is ¯ an isovector.1. the rate contains a factor (( 2 )2 − ( 1 )2 )2 = 1 .3 (and the one with the γ ’s ‘crossed’). the contributions from the u¯ and dd states have opposite signs (see u section 12.4. With this factor.3). this process is entirely calculable from the graph shown in figure 14. ) degenerate Copyright 2004 IOP Publishing Ltd .3. Triangle graph for π 0 decay. Furthermore.2 The dynamics of colour 14. it is certainly clear that the interquark force must depend on colour. where ‘q’ is u or d. d3 u 1 . Experimentally the rate agrees well with Steinberger’s calculation. 14. q2 q2 and q3 q3 ). indicating that (14. .3. in which case the factor corresponding to (14. since we do not observe ‘colour multiplicity’ of hadronic states: for ∗ ∗ example we do not see eight other coloured π + ’s (d1 u 2 . As will be discussed in section 18.9) is just one (since the n has zero charge). Thus. which corresponds to Nc = 3 identical amplitudes of the form shown in figure 14. ‘colour’ must be an exact symmetry of the Hamiltonian governing the quark dynamics. 3 3 9 (14. To begin with.9) However. we mention the rate for π 0 → γ γ . . and for the same reason.2.

If we assume that only colour singlet states exist and that the strong ¯ interquark force depends only on colour. .2 that anti-quarks belong to the complex conjugate of the representation (or multiplet) to which quarks belong. which transforms as (1)∗ ∗ ∗ ∗ ψα → ψα = Vαβ ψβ .with the one ‘colourless’ physical π + whose wavefunction was given previously.8(c)—where the levels are very dense. 1 and 1 occur.48) and (12. and for 2 baryons only T = 0. the quark forces must ¯ have a dramatic saturation property: apparently no qqq. The anti-symmetric. . 3.) The same restriction holds for SU(3)f also—only 1’s and 8’s occur for mesons and only 1’s. . states. this of course is what is translated ¯ into the recipe: ‘mesons are qq. More particularly. The observed hadronic states are all colour singlets and the force must somehow be responsible for this. This may be written in matrix notation as ψ † ψ where the ψ † as usual denotes the transpose of the complex conjugate of the column vector ψ. Then. three-quark baryon wavefunction of (14. 1 . if a quark colour triplet wavefunction ψα transforms under a colour transformation as ψα → ψα = Vαβ ψβ (1) (14. (In fact. states ¯¯ exist. in nuclear physics. no qq or qq states exist either—nor. such nuclear states are hard to 2 identify experimentally. Consider the quantity (cf (14. . .49)). . Consider again the analogy drawn in section 1. colour singlet.2 between isospin multiplets in nuclear physics and in particle physics. we can now construct colour singlet wavefunctions for ∗ mesons. built from qq. the fact that qq states are seen but qq and qq are not. which seems satisfactory. no qqqq. we find that ψ † Copyright 2004 IOP Publishing Ltd .11) Given this information. All this can be summarized by saying that the quark colour degree ¯ of freedom must be confined. One simple possibility might be that the three colours correspond to the components of an SU(2)c triplet ‘ψ’. taking the transpose of (14.1. a property we shall now assume and return to in chapter 16.3)) α ψα ψα where ψ ∗ ¯ represents the anti-quark and ψ the quark.10) where V(1) is a 3 × 3 unitary matrix appropriate to the T = 1 representation of ∗ SU(2) (cf (12. Furthermore. There is one very striking difference in the latter case: for mesons only T = 0.2) is then just the triple scalar product ψ 1 · ψ 2 × ψ 3 . the force has to produce only those very restricted types of quark configuration which are observed in the hadron spectrum. baryons are qqq’. Thus. that only A = 2 and A = 3 nuclei exist! Thus. But what about the meson wavefunction? Mesons are formed of quarks and anti-quarks. . do single q’s or q’s. for that matter.11). gives us an important clue as to what group to associate with ¯¯ colour. because they occur at high excitation energy for some of the isobars—cf figure 1. 8’s and 10’s for baryons. qqqqq.3 and 12. 1 and 3 . It is as if we said. then the wavefunction for the ‘anti’-triplet is ψα . In quark terms. and we recall from sections 12. while in nuclei there is nothing in principle to 2 2 stop us finding T = 5 . (14.

transforms by so that the combination ψ † ψ ψ † → ψ † = ψ † V(1)† transforms as ψ † ψ → ψ † ψ = ψ † V(1)† V(1) ψ = ψ † ψ V(1) (14.1(a)) that the anti-symmetric 3q combination (14. and for ¯¯ qq states just as well as for qq states—indeed they are formally identical. The next simplest possibility seems to be that the three colours correspond to the components of an SU(3)c triplet.10) and (14. unlike the case of SU(2)c triplets. chapter 3). and so on).58)). All this may seem fine. It is not possible to choose the λ’s to be pure imaginary in (14.13) is unitary (compare (12.11)—that is. and see also problem 14. Thus. SU(3)c seems to be a possible and economical choice for the colour ¯¯ group.1(b)) that. ψ †ψ (14.2) is also an SU(3)c invariant. as required.74)) ψ → ψ = Wψ where W is a special unitary 3 × 3 matrix parametrized as W = exp(iα · λ/2). as we learned in section 12. This is the meaning of the superposition (14.3).14) goes through as in (14. The three-dimensional representation of SU(2)c which we are using here has a very special nature: the matrix V(1) can be chosen to be real. both the proposed meson and baryon states are colour singlets. so that there is a distinction between ψ and ψ ∗ . It is always possible to choose the wavefunctions ψ to be real and the transformation matrix V(1) to be real also. Thus. Indeed.12) (14.16) (14. chapter 8). Jones (1990. We know very well how real three-dimensional vectors transform—namely by an orthogonal 3 × 3 matrix. But such ¯ ‘diquark’ (or ‘anti-diquark’) states are not found and hence—by assumption— should not be colour singlets. Copyright 2004 IOP Publishing Ltd . it can be shown (Carruthers (1966. But now the basic difficulty appears: there is no distinction between ψ and ψ ∗ ! They both transform by the real matrix V(1) . this means that it must be orthogonal. Thus. It is the same in SU(2). it is not possible to form an SU(3)c colour singlet combination out of two colour triplets qq or antitriplets qq. it is a colour singlet.2.13). Since V(1) is. in general. unitary. This means that we can make SU(2) invariant (colour singlet) combinations for qq states. This can be understood ‘physically’ if we make use of the great similarity between SU(2) and the group of rotations in three dimensions (which is the reason for the geometrical language of isospin ‘rotations’.15) and thus the 3 × 3 W matrices of SU(3)c cannot be real. In this case the quark colour wavefunction ψα transforms as (cf (12. and it can be The proof of the invariance of shown (problem 14. and ψ† transforms as ψ † → ψ † = ψ † W† .15) (14. the where the last step follows since product is invariant under (14. but there is a problem.

λ2 8 2 a=1 (λa ) (14. A similar thing happens in SU(3)c . In the triplet state T2 = 2 and in the singlet state (τ 1 · τ 2 )T =1 = 1 (τ 1 · τ 2 )T =0 = − 3 T2 = 0. (14. analogous to that of the total spin/angular momentum in SU(2).20) (14.24) = = say.1(b).17) and if V is positive. chapter 8)). Here ≡ is found (see (12. Suppose this interquark force depended on the quark colours via a term proportional to λ1 · λ2 .2 Global SU(3)c invariance and ‘scalar gluons’ As previously stated. (14.75)) to have where the value 16/3 (the unit matrix being understood). They depend on the SU(3)c representation—indeed. combine to give a 6c -dimensional representation and ˆ a 3∗ (see problem 14. The eigenvalues of C2 play a very important role in SU(3)c . Thus. and Jones (1990. The total isospin is T = 1 (τ 1 + τ 2 ) and we have 2 T 2 = 1 (τ 2 + 2τ 1 · τ 2 + τ 2 ) = 1 (3 + 2τ 1 · τ 2 + 3) 2 4 1 4 whence τ 1 · τ 2 = 2 T 2 − 3. What sort of interquark force could produce this dramatic result? Consider an SU(2) analogy again: the interaction of two nucleons belonging to the lowest (doublet) representation of SU(2). There is a simple isospin-dependent force which can produce a splitting between these states—namely V τ 1 · τ 2 . we are assuming.18) (14.2. they are one of the defining labels of SU(3) representations in general (see section M.14. the possible T values for two nucleons are T = 1 (triplet) and T = 0 (singlet). that the only physically observed hadronic states are colour singlets—and this now means singlets under SU(3)c .21) Then.22) (14. Labelling the states by an isospin T . The operator F 2 commutes with all components of λ1 and λ2 (as T 2 does with τ 1 and τ 2 ) and represents the ˆ quadratic Casimir operator C2 of SU(3)c (see section M. each in the representation 3c . on empirical grounds.5). in the ˆ colour space of the two quarks considered here. where the ‘1’ and ‘2’ refer to the two nucleons. in just the same way. The value of C2 for the c Copyright 2004 IOP Publishing Ltd . the T = 0 state is pulled down. Two quarks.19) (14.5 of appendix M). we can introduce the total colour operator F = 1 (λ1 + λ2 ) 2 so that F 2 = 1 (λ2 + 2λ1 · λ2 + λ2 ) 2 4 1 and λ1 · λ2 = 2 F 2 − λ2 λ2 1 λ2 2 λ2 .23) (14.

due to the assumed scalar. the quark wave equation describing the colour interactions must be SU(3)c covariant. as our strict ‘colour singlets only’ hypothesis would demand. This is the singlet representation 1c . The maximum attraction will clearly be for states in which F 2 is zero. for simplicity. the SU(3)c structure of the amplitude will be λ1 λ2 λ1a λ2b δab = · 2 2 2 2 Copyright 2004 IOP Publishing Ltd (14. and which would be one (rather crude) way of interpreting confinement. built from three quarks. Of course. 2. one-half c of the attraction in the qq colour singlet state. . As in (13.4).92): in the latter. (14. the vertex corresponding to (14.94).26) 2 where gs is a ‘strong charge’ and Aa (a = 1. the ‘λ1 · λ2 ’ interaction will give a repulsive eigenvalue +2/3 in the 8c channel. Two quarks cannot combine to give a colour singlet state.27) (14. Aa appears on the right-hand side. and a ‘maximally attractive’ eigenvalue −16/3 in the 1c channel.25) where ‘1’ refers to the quark and ‘2’ to the anti-quark.28) . for which ˆ C2 = 3. rather than vector. because the gauge field quanta are vectors rather than scalars. . for a quark and an anti-quark. Equation (14. . and so the potentials may be taken to be scalars. but still strongly attractive. In the case of baryons. In this case (14.94) simply in the absence of the γ µ factor. not a local SU(3) gauge symmetry.2). with maximum binding. In (14. such a simple potential model does not imply that the energy difference between the 1c states and all coloured states is infinite. When we put two such vertices together and join them with a gluon propagator (figure 14.sextet 6c representation is 10/3 and for the 3∗ representation is 4/3.27) differs from (13.26) may be compared with (13. the eigenvalue of λ1 · λ2 is −8/3. The ¯ (qq) pair in the 3∗ state can then couple to the remaining third quark to make the c overall colour singlet state (14. we can ask: what single particle exchange process between quark (or anti-quark) colour triplets produces a λ1 · λ2 type of term? The answer is the exchange of an SU(3)c octet (8c ) of particles. but we have seen in section 12. 8) is an octet of scalar ‘gluon / potentials’. Thus the c ‘λ1 · λ2 ’ interaction will produce a negative (attractive) eigenvalue −8/3 in the 3∗ c states.26).24) is replaced by λ1 · λ2 = 2 F 2 − 1 (λ2 + λ2 ) 2 2 1 (14. Since colour is an exact symmetry.2 that a quark and an anti-quark can: they combine to give 1c and 8c . A simple such equation is λa Aa ψ (14. but a repulsive eigenvalue +4/3 in the 6c states for two quarks. we have seen that when two of them are coupled to the 3∗ state. nature of the ‘gluon’ here. . we are dealing at this stage only with a global SU(3) symmetry. Nevertheless. which (anticipating somewhat) we shall call gluons. Thus.26) is / (i∂ − m)ψ = gs −igsλa /2.

2. Scalar gluon exchange between two quarks.29) 2 as in (13.30) as in (13.27) becomes −igs λa µ γ 2 (14. This interaction has exactly the required ‘λ1 · λ2 ’ character in the colour space. Copyright 2004 IOP Publishing Ltd . and the vertex (14.92 ). the assumption of local SU(3)c covariance leads to a great deal more: for example.Figure 14.29) and the vertex (14. the δab arising from the fact that the freely propagating gluon does not change its colour. 14.26) are. in fact. One motivation for this is the desire to make the colour dynamics as much as possible like the highly successful theory of QED.3 Local SU(3)c invariance: the QCD Lagrangian It is.26) then becomes λa / / (i∂ − m)ψ = gs Aa ψ (14. This is implemented by the replacement ∂µ → ∂µ + igs λa Aaµ (x) 2 (14. As we have seen in the last chapter.32) (14.5.31) in the Dirac equation for the quarks.4. like the photons of QED. it implies that the gluons are massless vector (spin-1) particles and that they interact with themselves via three-gluon and four-gluon vertices. which leads immediately to (14. tempting to suppose that the ‘scalar gluons’ introduced in (14. Of course. this involves the simple but deep step of supposing that the quark wave equation is covariant under local SU(3)c transformations of the form ψ → ψ = exp(igs α(x) · λ/2)ψ. vector particles.30). and to derive the dynamics from a gauge principle.94). Equation (14. of course. which are the SU(3)c analogues of the SU(2) vertices discussed in section 13.2.

36) The factors in front of the ‘ F F’ are chosen conventionally. as usual.33) In (14.103) (with δi j replaced by δab here).33) we have relied essentially on the ‘gauge principle’ (invariance under a local symmetry) and the requirement of renormalizability (to forbid the presence of terms with mass dimension higher than four). summed over: α and β are SU(3)c triplet indices running from 1 to 3.174) (adapted to SU(3)c ).87).α (i D − m f )αβ qf. in fact. we know that strong interactions conserve both P and T to a high degree of accuracy.36) can actually be written as a total divergence: ˆ µν ˆ ρσ ˆµ (14.105) and (13. and by ˆ ˆ ( Dµ )ab = ∂µ δab + gs f ca b Acµ (14. and a . E → −E and B → B.84). most of which we have already introduced in chapter 13. such a term is proportional to E · B (problem 14. The Feynman rules for tree graphs following from (14.36). The covariant derivatives are defined by ˆ ˆ ( Dµ )αβ = ∂µ δαβ + igs 1 (λa )αβ Aaµ 2 when acting on the quark SU(3)c triplet. (14. namely ˆ Äθ = 2 θ gs 64π 2 µνρσ ˆ µν ˆ ρσ Fa Fa .2).β − 1 Faµν Fa ˆ ˆ 4 − 1 µ ˆa ˆa (∂ν Aµ )(∂ν Aν ) + ∂µ ηa Dab ηb . note that the matrices representing the SU(3) generators in the octet representation are as given in (12. where the summation on a is absent. repeated indices are. We remind the reader that the last two terms in (14. and these take the place of the ‘λ/2’ in (14. The renormalizability of such a theory was proved by ’t Hooft (1971a. appropriate to a gauge field propagator of the form (13. b). Experimentally.34) (compare (13. The answer to this is that (14. there is. b are SU(3)c -octet indices running from 1 to 8.37) µνρσ Fa Fa = ∂µ K Copyright 2004 IOP Publishing Ltd . E → E and B → −B.33). Gathering together (13. However.The most compact way of summarizing all this structure is via the Lagrangian. The reader may wonder whether the ‘θ term’ (14.175) in the SU(2) case). however. respectively.33) are given in appendix Q.36) should give rise to a new Feynman rule. The same is true of (14. For the second of these. In the U(1) case of QED. The coefficient θ is therefore required to be small. we write it out here for convenience: ÄQCD = flavours f ˆ ¯ ˆ ˆ µν / q f.33) are the gauge-fixing and ghost terms. ˆ† ˆ ˆ 2ξ (14. In arriving at (14. as in (13.35) (14. This violates both parity and time-reversal symmetry: under P .34) when acting on the octet of ghost fields. one more gauge invariant term of mass dimension four which can be written down. while under T .

And this brings us to a central difficulty. chapter 16). we can only use perturbation theory for relatively weak interactions. however. they differ in at least one crucial respect: the fundamental quanta of QED (leptons and photons) are observed as free particles but those of QCD (quarks and gluons) are not. making conventional assumptions about the fields going to zero at infinity. however. The resolution of this unprecedented mystery lies in the fundamental feature of non-Abelian gauge theories called ‘asymptotic freedom’. It seems that in order to compare QCD with data we shall inevitably have to reckon with the complex non-perturbative strong interaction processes (‘confinement’) which bind quarks and gluons into hadrons—and the underlying simplicity of the QCD structure will be lost.38) Any total divergence in the Lagrangian can be integrated to give only a ‘surface’ term in the quantum action. whereby the effective coupling strength becomes progressively smaller at short distances or high energies.where ˆ Kµ = 2 µνρσ ˆ ˆ ˆ ˆ Aaν (∂ρ Aaσ − 2 gs f abc Abρ Acσ ). any theory must ultimately stand or fall by its success. We must now recall from chapter 7 the very considerable empirical success of the parton model. and it enables a quantitative perturbative approach to be followed (in appropriate circumstances) even in strong interaction physics. There are. A proper understanding of how this works necessitates a considerable detour. for further discussion of the θ term. in explaining the experimental facts. In particular. or to Weinberg (1996. because we also know that the forces are indeed so strong that no-one has yet succeeded in separating completely either a quark or a gluon from a hadron. so that they emerge as free particles. Elegant and powerful as the gauge principle may be. or otherwise. field configurations (‘instantons’) such that the contribution of the θ term does not vanish. We have one well-understood and reliable calculational procedure. so no perturbative Feynman rules are associated with (14. whereas QCD is supposed to be a strong interaction theory. namely renormalized perturbation theory. This property is the most compelling theoretical motivation for choosing a non-Abelian gauge theory for the strong interactions. however. we need to understand the important group of ideas going under the general heading of the Copyright 2004 IOP Publishing Ltd .36). However. we are faced with an almost paradoxical situation. into the physics of renormalization. in which the interactions between the partons (now interpreted as quarks and gluons) were totally ignored! Somehow it does seem to be the case that in deep inelastic scattering—or. How can our perturbative QED techniques possibly be used for QCD? Despite the considerable formal similarities between the two theories. 3 (14. These configurations are not reachable in perturbation theory. However. which can usually be discarded. ‘hard’.6). wide-angle collisions—the hadron constituents are very nearly free and the effective interaction is relatively weak. which we have emphasized. more generally. section 23. highenergy. We refer the reader to Cheng and Li (1984.

and the following one. it is clear that collisions between the hadronic partons should by no means be limited to QED-induced processes. therefore. due to all these non-perturbative strong interactions. in section 9. however. This was a model rather than a theory: the theme of most of the rest of this chapter.‘renormalization group’ and this will be the topic of chapter 15. justification being provided by the assumed property of asymptotic freedom. for the following reason. determined by QCD.5 we briefly introduced the idea of jets in e+ e− physics: wellcollimated sprays of hadrons. Thus we might hope that the identification and analysis of short-distance parton–parton collisions will lead to direct tests of the tree-graph structure of QCD. when we look more closely at it (with larger Q 2 ). e+ e− → qq. and well described in the lowest order of perturbative QED. But the asymptotic freedom property (assumed for the moment) implies that at short distances we can use perturbation theory even for ‘strong interactions’. In section 9. will be the way in which the theory of QCD both justifies the parton model and predicts observable corrections to it. more and more ‘bits’ are revealed. apparently created as a quark–anti-quark pair. We shall usually continue to use the language of partons. we expect to see strong interactions between the partons. In other words. At relativistic energies and short distances. separate from each other at high speed.Q 2 data in terms of the free point-like hadronic ‘partons’. On the contrary.3. However. the data will be complicated and hard to interpret. It is only at relatively low energies and/or momentum transfers that the (essentially non-relativistic) concept of a fixed number of constituents in a bound state is meaningful. How are short-distance collisions to be identified experimentally? The answer is: in just the same way as Rutherford distinguished the presence of a Copyright 2004 IOP Publishing Ltd . how in ¯ hadron–hadron collisions the hadrons acted as beams of partons—quarks. rather than—say—that of hadronic ‘constituents’. We also saw. 14.3 Hard scattering processes and QCD tree graphs ¯ 14. For the moment we proceed with a discussion of the perturbative applications of QCD at the tree-level. anti-quarks and gluons). In this situation we prefer ‘partons’ to ‘constituents’.1 Two-jet events in pp collisions In chapter 9 of volume 1 we introduced the parton model and discussed how it successfully interpreted deep inelastic and large. The dynamics at the parton level. was governed by QED. since the latter term seems to carry with it more of the traditional connotation of a fixed number. pair creation and other fluctuation phenomena are so important that it no longer makes sense to think so literally of ‘what the bound state is made of ’—as we shall see. the partons are now to be identified precisely with the QCD quanta (quarks.4. In general. antiquarks and gluons—which could produce ¯ pairs by the inverse process qq → ¯ ll ¯ ll: the force acting here is electromagnetic.

leading in lowest-order perturbation theory to two wide-angle partons emerging at high speed from the collision volume. Alper et ¨ al 1973. Nevertheless. We now face the experimental problem of picking out from the enormous multiplicity of total events just these hard scattering ones. Collins and Martin 1984. Jets will be the observed hadronic manifestation of the underlying QCD processes—just as they are in the analogous QED process of e+ e− annihilation into hadrons. In the first stage (figure 14. It is much better to surround the collision volume with an array of calorimeters which measure the total energy deposited. chapter 5).6). If hadronic matter were smooth and uniform (cf the Thomson atom). The physical process is viewed as a two-step one. discussed in section 9. small heavy scattering centre (the nucleus) in the atom—by looking at secondary particles emerging at large angles with respect to the beam direction. This is just what is observed in the vast majority of events: the average value of pT measured for charged particles is very low ( pT ∼ 0. since the probability that a single hadron in a jet will actually carry most of the jet’s total transverse momentum is quite small (Jacob and Landshoff 1978.pT secondaries is observed. But it turns out that such triggering really reduces the probability of observing jets. which is called parton fragmentation. the effective interaction strength increases and the second stage is entered. the dominant processes in the fragmentation stage must involve small momentum transfer.5. Banner et al 1983) at the CERN ISR (CMS energies 30–62 GeV) and were interpreted in parton terms as follows. Early experiments used a trigger based on the detection of a single highpT particle. Parton–parton collision. in order to analyze them further. each having a total 4-momentum approximately equal to that of the parent parton (figure 14. They were first seen (Busser et al 1972.5. Wide-angle jets can then be identified by the occurrence of a large amount of total transverse energy deposited in a number of adjacent calorimeter Copyright 2004 IOP Publishing Ltd . 1973.5) a parton from one hadron undergoes a short-distance collision with a parton from the other. Thus we have a picture in which two fairly well-collimated jets of hadrons occur. As yet we do not have a quantitative dynamical understanding of this second (nonperturbative) stage.4 GeV) but in a small fraction of collisions the emission of high. we can argue that for the forces to be strong enough to produce the observed hadrons.Figure 14. This is a ‘hard-scattering’ process. that in which the coloured partons turn themselves— under the action of the strong colour-confining force—into colour singlet hadrons. the distribution of events in pT would be expected to fall off very rapidly at large pT values—perhaps exponentially. As the two partons separate. For each secondary particle we can define a transverse momentum pT = p sin θ where p is the particle momentum and θ is the emission angle with respect to the beam axis.

4. The importance of calorimetric triggers was first emphasized by Bjorken (1973). 0.40) . Parton fragmentation into jets. 0. showing clear evidence of the wide-angle collisions characteristic of hard processes. P) and the left-moving one x 2 p2 = x 2 (P. following earlier work by Berman et al (1971).5) in the pp CMS. The application of this method to the detection and analysis of wideangle jets was first reported by the UA2 collaboration at the CERN pp collider ¯ (Banner et al 1982). (14. 0.6.4 GeV. −P) (14. 0. which are approximately limited by the observed pT value (∼ 0. Thus. For each event the total transverse energy E T is measured where ET = i E i sin θi . there is plenty of energy available to produce such events. the right-moving parton has 4-momentum x 1 p1 = x 1 (P. As we shall see shortly. Figure 14. Consider the kinematics of the two-parton collision (figure 14. cells: this is then a ‘jet trigger’. An impressive body of quite remarkably clean jet data was √ subsequently accumulated by both the UA1 and UA2 collaborations (at s = 546 and 630 √ GeV) and by the CDF and D0 collaborations at the FNAL Tevatron collider ( s = 1. At s ≥ 500–600 GeV.Figure 14. As in the Drell–Yan ¯ process of section 9.41) √ where P = s/2 and we are neglecting parton transverse momenta. and thus negligible Copyright 2004 IOP Publishing Ltd (14.7 shows the E T distribution observed by UA2: it follows the ‘soft’ exponential form for E T ≤ 60 GeV but thereafter departs from it. the majority of ‘hard’ events are of two-jet type. with the jets sharing the E T approximately equally. √ The total s value is important for another reason. a ‘local’ trigger set to select events with localized transverse energy ≥ 30 GeV and/or a ‘global’ √ trigger set at ≥ 60 GeV can be used.39) E i is the energy deposited in the i th calorimeter cell and θi is the polar angle of the cell centre: the sum extends over all cells.8 TeV).

13). The parton distribution functions are large at these relatively small x values. If this is to be greater than 50 GeV. Returning to the kinematics of (14.41). x 1 will not. their relative azimuthal angle will be 180◦.7. Consider the simple case of 90◦ scattering. it is found that essentially all of the transverse energy is indeed split roughly equally between two approximately back-to-back jets. equal to x say. be equal to x 2 . What are the characteristics of jet events? When E T is large enough (≥ 150 GeV). E T observed in the UA2 central on this energy scale). the jets will be coplanar with the beam direction. It is strikingly evident that we are seeing precisely a kind of ‘Rutherford’ process or—to vary the analogy—we might say that hadronic jets are acting as the modern counterpart of Faraday’s iron filings.8. Figure 14. so that—as is apparent in figure 14.1 will contribute to the process. then partons with x ≥ 0. due to sea quarks (section 9.Figure 14.8— the jets will not be collinear. A typical such event is shown in figure 14. The total outgoing transverse energy is then 2 x P = x s . in rendering visible the underlying field dynamics! We may now consider more detailed features of these two-jet events—in Copyright 2004 IOP Publishing Ltd . Distribution of the total transverse energy calorimeter (DiLella 1985).9 shows a number of examples in which the distribution of the transverse energy over the calorimeter cells is analysed as a function of the jet opening angle θ and the azimuthal angle φ.3) and gluons (figure 14. to the extent that the transverse parton momenta can be neglected. which requires = (for massless partons) x 1√ x 2 . in general. and thus we expect to obtain a reasonable cross-section. However.e.40) and (14. i.

gq → gq and p q q gg → gg. with so many contributing parton processes.Figure 14. numerically. squared.6!). q¯ → q¯ . associated with two large clusters of calorimeter energy depositions (Geer 1986). all processes involving one gluon exchange will lead to essentially the same dominant angular distribution of Rutherford-type. gg → gg. in the parton model. which are characteristic of their localization to hadronic dimensions. 1 In the sense that the partons in hadrons have momentum or energy distributions. where θ is the parton CMS scattering angle (recall section 2. one-gluon-exchange matrix elements |Å|2 . which lists the different relevant spin-averaged. q q q The most important. such as qq → qq.42) is replaced by the various QCD subprocess cross-sections. Two tightly collimated groups of reconstructed charged tracks can be seen in the cylindrical central detector of UA1. for a p¯ collider are q¯ → q¯ . ∼ sin−4 θ/2. except that the electromagnetic annihilation cross-section σ (q¯ → µ+ µ− ) = 4πα 2 /3q 2 q (14. At first sight this seems to be a very complicated story. Two-jet event.8.1 (adapted from Combridge et al (1977)). But a significant simplification comes from the fact that in the CMS of the parton collision. This is illustrated in table 14. we shall have many parton subprocesses. anti-quarks and gluons. each one being weighted by the appropriate distribution functions. The cross-section will be given. q¯ → gg. particular. etc. The initial hadrons provide wide-band beams1 of quarks. thus. Copyright 2004 IOP Publishing Ltd . by a formula of the Drell–Yan type. the expectations based on QCD tree graphs.

t and u are the subprocess invariants. DiLella 1985). the initial parton configuration shown in figure 14.9.Figure 14. in the θ. Each bin represents a cell of the UA2 calorimeter. in Copyright 2004 IOP Publishing Ltd . so tˆ ∝ (1 − cos θ ) ∝ sin2 θ/2. the outgoing partons then emerging back-to-back at an angle θ to the beam axis.43) Here s . where the parton differential cross-section is given by (cf (6. We note that the s . Note that the sum of the φ ’s equals 180◦ (mod 360◦ ).129)) 2 παs dσ = |Å|2 . so that ˆ ˆ ˆ s = ( x 1 p 1 + x 2 p 2 )2 = x 1 x 2 s ˆ (cf (9.85)).44) Continuing to neglect the parton transverse momenta. Only ˆ ˆ ˆ the terms in (t )−2 ∼ sin−4 θ/2 are given in table 14.1. u dependence of these terms is the same for the three types of process (and is.5 can be brought to the parton CMS by a Lorentz transformation along the beam direction. Four transverse energy distributions for events with E T > 100 GeV. d cos θ 2ˆ s (14. tˆ. (14. φ plane (UA2.

Table 14.1. Spin-averaged squared matrix elements for one-gluon exchange (tˆ-channel) processes. Subprocess qq → qq q¯ → q¯ q q qg → qg gg → gg
9 4

4 9



s 2 +u 2 ˆ ˆ tˆ2

s 2 +u 2 + · · · ˆ ˆ tˆ2 s 2 +u 2 ˆ ˆ tˆ2

+ ···

fact, the same as that found for the 1γ exchange process e+ e− → µ+ µ− —see problem 8.18, converting dσ/dt into dσ/d cos θ ). Figure 14.10 shows the two-jet angular distribution measured by UA1 (Arnison et al 1985). The broken curve is the exact angular distribution predicted by all the QCD tree graphs—it actually follows the sin−4 θ/2 shape quite closely. It is interesting to compare this angular distribution with the one predicted on the assumption that the exchanged gluon is a spinless particle, so that the vertices have the form ‘uu’ rather than ‘uγµ u ’. Problem 14.3 shows that, in this case, ¯ ¯ ˆ the 1/t 2 factor in the cross-section is completely cancelled, thus ruling out such a model. This analysis surely constitutes compelling evidence for elementary hard scattering events proceeding via the exchange of a massless vector quantum. It is possible to go much further, in fact. Anticipating our later discussion, the small discrepancy between ‘tree-graph’ theory (which is labelled ‘leading-order QCD scaling curve’ in figure 14.10) and experiment can be accounted for by including corrections which are of higher order in αs . To study such deviations from the ‘Rutherford’ behaviour it is convenient (Combridge and Maxwell 1984) to plot the data in terms of the variable χ= which is such that dσ dσ d cos θ 1 dσ = = − (1 − cos θ )2 . dχ d cos θ dχ 2 d cos θ (14.46) 1 + cos θ 1 − cos θ (14.45)

The singular Rutherford term in dσ/d cos θ is therefore removed, and as θ → 0, dσ/dχ → constant. Figure 14.11 shows a jet–jet angular distribution from the D0 collaboration (H Weerts 1994), plotted this way. The broken curve is the ‘naive’
Copyright 2004 IOP Publishing Ltd

Figure 14.10. Two-jet angular distribution plotted against cos θ (Arnison et al 1985).

parton model prediction and is clearly not in agreement with the data. The full curve includes QCD corrections beyond the tree level (Ellis et al 1992), involving the ‘running’ of the coupling constant αs and ‘scaling violation’ in the parton distributions, both of which effects will be discussed later. The corrections lead to good agreement with experiment. The fact that the angular distributions of all the subprocesses are so similar allows further information to be extracted from these two-jet data. In general, the parton model cross-section will have the form (cf (9.92)) d3 σ = d x 1 d x 2 d cos θ Fa (x 1 ) Fb (x 2 ) x1 x2 dσab→cd d cos θ (14.47)



where Fa (x 1 )/x 1 is the distribution function for partons of type ‘a’ (q, q or g), and ¯ similarly for Fb (x 2 )/x 2 . Using the near identity of all dσ/d cos θ ’s, and noting the numerical factors in table 14.1, the sums over parton types reduce to
9 4 4 {g(x 1 ) + 9 [q(x 1 )

+ q(x 1 )]}{g(x 2) + 4 [q(x 2) + q(x 2 )]} ¯ ¯ 9


where g(x), q(x) and q(x) are the gluon, quark and anti-quark distribution ¯ functions. Thus, effectively, the weighted distribution function2 F(x) = g(x) + 4 [q(x) + q(x)] ¯ 9 x (14.49)

2 The 4 reflects the relative strengths of the quark–gluon and gluon–gluon couplings in QCD. 9

Copyright 2004 IOP Publishing Ltd

Figure 14.11. Distribution of χ from the D0 collaboration (Weerts 1994) compared with QCD predictions (Ellis et al 1992; figure from Ellis et al 1996).

is measured (Combridge and Maxwell 1984); in fact, with the weights as in (14.48), F(x 1 ) F(x 2 ) dσgg→gg d3 σ = . (14.50) d x 1 d x 2 d cos θ x1 x 2 d cos θ x 1 and x 2 are kinematically determined from the measured jet variables: from (14.44), x 1 x 2 = s /s ˆ where s is the invariant [mass]2 of the two-jet system and ˆ √ x 1 − x 2 = 2 PL / s (cf (9.83)) (14.52) (14.51)

with PL the total two-jet longitudinal momentum. Figure 14.12 shows F(x)/x obtained in the UA1 (Arnison et al 1984) and UA2 (Bagnaia et al 1984) experiments. Also shown in this figure is the expected F(x) based on contemporary fits to the deep inelastic neutrino scattering data at Q 2 = 20 GeV2 and 2000 GeV2 (Abramovicz et al 1982a, b, 1983)—the reason for the change with Q 2 will be discussed in section 15.7. The agreement is qualitatively very satisfactory. Subtracting the distributions for quarks and anti-quarks as found in deep inelastic lepton scattering, UA1 were able to deduce the gluon distribution function g(x) shown in figure 14.13. It is clear that gluon processes will dominate at small x —and even at larger x they will be important because of the colour factors in table 14.1.
Copyright 2004 IOP Publishing Ltd

Figure 14.12. Effective distribution function measured from two-jet events (Arnison et al 1984 and Bagnaia et al 1984). The broken and chain curves are obtained from deep inelastic neutrino scattering. Taken from DiLella (1985).

14.3.2 Three-jet events Although most of the high E T events at hadron colliders are two-jet events, in some 10–30% of the cases the energy is shared between three jets. An example is included as (d ) in the collection of figure 14.9: a clearer one is shown in figure 14.14. In QCD such events are interpreted as arising from a 2 parton → 2 parton + 1 gluon process of the type gg → ggg, gq → ggq, etc. Once again, one can calculate (Kunszt and Pietarinen 1980, Gottschalk and Sivers 1980, ´ Berends et al 1981) all possible contributing tree graphs, of the kind shown in figure 14.15, which should dominate at small αs . They are collectively known as QCD single-bremsstrahlung diagrams. Analysis of triple jets which are well separated both from each other and from the beam directions shows that the data are in good agreement with these lowest-order QCD predictions. For example, figure 14.16 shows the production angular distribution of UA2 (Appel et al 1986) as a function of cos θ ∗ , where θ ∗ is the angle between the leading (most energetic) jet momentum and the beam axis, in the three-jet CMS. It follows just the same sin−4 θ ∗ /2 curve as in the two-jet case (the data for which are also shown in the figure), as expected for massless quantum exchange—the particular curve is for the representative process gg → ggg. Another qualitative feature is that the ratio of three-jet to two-jet events is controlled, roughly, by αs (compare figure 14.15 with the one-gluon exchange
Copyright 2004 IOP Publishing Ltd

Figure 14.13. The gluon dustribution function g(x) extracted from the effective structure function F(x) by subtracting the expected contribution from the quarks and anti-quarks (Geer 1986).

amplitudes of table 14.1). Thus, a crude estimate of αs could be obtained by comparing the rates of three-jet to two-jet events in pp collisions (see figure 14.22 ¯ for a similar ratio in e+ e− annihilation). Other interesting predictions concern the characteristics of the three-jet final state (for example, the distributions in the jet energy variables). At this point, however, it is convenient to leave pp ¯ collisions and consider instead three-jet events in e+ e− collisions, for which the complications associated with the initial state hadrons are absent. 14.4 Three-jet events in e+ e− annihilation Three-jet events in e+ e− collisions originate, according to QCD, from gluon bremsstrahlung corrections to the two-jet parton model process e+ e− → γ ∗ → q¯ , as shown in figure 14.17.3 This phenomenon was predicted by Ellis et al q (1976) and subsequently observed by Brandelik et al (1979) with the TASSO detector at PETRA and Barber et al (1979) with MARK-J at PETRA, thus providing early encouragement for QCD. The situation here is, in many ways, simpler and cleaner than in the pp case: the initial state ‘partons’ are perfectly ¯ physical QED quanta and their total 4-momentum is zero, so that the three jets
3 This is assuming that the total e+ e− energy is far from the Z0 mass; if not, the contribution from the intermediate Z0 must be added to that from the photon.

Copyright 2004 IOP Publishing Ltd

Figure 14.14. Three-jet event in the UA1 detector, and the associated transverse energy flow plot (Geer 1986).

Figure 14.15. Some tree graphs associated with three-jet events.

have to be coplanar; further, there is only one type of diagram compared to the large number in the pp case and much of that diagram involves the easier vertices ¯ of QED. Since the calculation of the cross-section predicted from figure 14.17 is not only relevant to three-jet production in e+ e− collisions but also to QCD corrections to the total e+ e− annihilation cross-section, and to scaling violations in deep inelastic scattering as well, we shall now consider it in some detail. It is important to emphasize at the outset that quark masses will be neglected in this calculation. The quark, anti-quark and gluon 4-momenta are p1 , p2 and p3 respectively, as shown in figure 14.17; the e− and e+ 4-momenta are k1 and k2 . The crossCopyright 2004 IOP Publishing Ltd

Figure 14.16. The distribution of cos θ ∗ (•), the angle of the leading jet with respect to the beam line (normalized to unity at cos θ ∗ = 0), for three-jet events in pp collisions ¯ (Appel et al 1986). The distribution for two-jet events is also shown (◦). The full curve is a parton model calculation using the tree-graph amplitudes for gg → ggg, and cut-offs in transverse momentum and angular separation to eliminate divergences (see remarks following equation (14.68)).

Figure 14.17. Gluon brehmsstrahlung corrections to two-jet parton model process.

section is then (cf (6.110) and (6.112)) dσ = |Åq¯ g |2 d3 p1 d3 p2 d3 p3 1 q δ 4 (k1 + k2 − p1 − p2 − p3 ) (2π)5 2Q 2 2E 1 2E 2 2E 3 (14.53)

where (neglecting all masses)

Åq¯ g = q

p p ea e 2 g s λc ( / + / 3 ) v(k2 )γ µ u(k1 ) u( p1 )γν · 1 ¯ ¯ · γµ v( p2 ) 2 2 2 p1 · p3 Q p p λc ( / + / 3 ) · γν v( p2 ) ∗ν (λ)ac (14.54) −u( p1 )γµ · 2 ¯ 2 2 p2 · p3

and Q 2 = 4E 2 is the square of the total e+ e− energy, and also the square of
Copyright 2004 IOP Publishing Ltd

the virtual photon’s 4-momentum Q , and ea (in units of e) is the charge of a quark of type ‘a’. Note the minus sign in (14.54): the anti-quark coupling is −gs . In (14.54), ∗ν (λ) is the polarization vector of the outgoing gluon with polarization λ; ac is the colour wavefunction of the gluon (c = 1, · · · , 8) and λc is the corresponding Gell-Mann matrix introduced in section 12.2; the colour parts of the q and q wavefunctions are understood to be included in the u and ¯ p v factors; and ( / 1 + / 3 )/2 p1 · p3 is the virtual quark propagator (cf (L.6) in p appendix L of volume 1) before gluon radiation, and similarly for the anti-quark. Since the colour parts separate from the Dirac trace parts, we shall ignore them to begin with and reinstate the result of the colour sum (via problem (14.4)) in the final answer (14.68). Averaging over e± spins and summing over final-state quark spins and gluon polarization λ (using (8.170), and noting the discussion after (13.127)), we obtain (problem 14.5) 1 4 |Åq¯ g|2 = q
spins,λ 2 2 e4 ea gs µν L (k1 , k2 )Hµν ( p1 , p2 , p3 ) 8Q 4


where the lepton tensor is, as usual (equation (8.118)),
ν ν L µν (k1 , k2 ) = 2(k1 k2 + k1 k2 − k1 · k2 g µν ) µ µ


and the hadron tensor is Hµν ( p1 , p2 , p3 ) = 1 [L µν ( p2, p3 ) − L µν ( p1 , p1 ) + L µν ( p1 , p2 )] p1 · p2 1 + [L µν ( p1 , p3 ) − L µν ( p2 , p2 ) + L µν ( p1 , p2 )] p2 · p3 p1 · p2 [2L µν ( p1 , p2 ) + L µν ( p1 , p3 ) + ( p1 · p3 )( p2 · p3 ) + L µν ( p2 , p3 )] (14.57)

Combining (14.56) and (14.57) allows complete expressions for the five-fold differential cross-section to be obtained (Ellis et al 1976). Data are generally not extensive enough to permit such differential crosssections to be studied and so one integrates over three angles describing the orientation (relative to the beam axis) of the production plane containing the three jets. After this integration, the (doubly differential) cross-section is a function of two independent Lorentz invariant variables, which are conveniently taken to be two of the three si j defined by si j = ( pi + p j )2 . (14.58)

Since we are considering the massless case pi2 = 0 throughout, we may also write si j = 2 pi · p j .
Copyright 2004 IOP Publishing Ltd


Figure 14.18. Virtual photon decaying to q¯ g. q

These variables are linearly related by 2( p1 · p2 + p2 · p3 + p3 · p1) = Q 2 as follows from ( p 1 + p 2 + p 3 )2 = Q 2 and pi2 = 0. The integration yields (Ellis et al 1976) d2 σ 2 1 2 = α 2 ea αs 2 3 ds13 ds23 3 (Q ) s23 2Q 2 s12 s13 + + s23 s13 s13 s23 (14.62) (14.61) (14.60)

2 where αs = gs /4π. We may understand the form of this result in a simple way, as follows. It seems plausible that after integrating over the production angles, the lepton tensor will be proportional to Q 2 g µν , all directional knowledge of the k1 having been lost. Indeed, if we use −g µν L µν ( p, p ) = 4 p · p together with (14.57), we easily find that

s13 p1 · p3 p2 · p3 p1 · p2 Q 2 s23 2Q 2 s12 1 = + + + + − g µν Hµν = 4 p2 · p3 p1 · p3 ( p1 · p3 )( p2 · p3 ) s23 s13 s13 s23 (14.63) exactly the factor appearing in (14.62). In turn, the result may be given a simple physical interpretation. From (7.115) we note that we can replace −g µν by λ µ (λ ) ν∗ (λ ) for a virtual photon of polarization λ , the λ = 0 state contributing negatively. Thus, effectively, the result of doing the angular integration is (up to constants and Q 2 factors) to replace the lepton factor v(k2 )γ µ u(k1 ) by −i µ (λ ), so that F is proportional to the γ ∗ → q¯ g processes ¯ q shown in figure 14.18. But these are basically the same amplitudes as the ones we already met in Compton scattering (section 8.6). To compare with section 8.6.3, we convert the initial-state fermion (electron/quark) into a final-state anti-fermion (positron/anti-quark) by p → − p, and then identify the variables of figure 14.18 with those of figure 8.14(a) by p → p1 k → p3 − p → p2 s → 2 p1 · p3 = s13 (14.64) t → 2 p1 · p2 = s12 u → 2 p2 · p3 = s23 .

Remembering that in (8.180) the virtual γ had squared 4-momentum −Q 2 , we see that the Compton ‘ |Å|2 ’ of (8.180) indeed becomes proportional to the factor (14.63), as expected.
Copyright 2004 IOP Publishing Ltd

The kinematically allowed region in (xi ) is the interior of the equilateral triangle.19.67) ˜ The limits of the physical region are then clearly cos θ = ±1. If θ is the angle between 1 and 3 in this system.65) An event with a given value of the set x i can then be plotted as a point in an equilateral triangle of height 1. (This is the massless limit of the classic Dalitz plot. (14. and because it will be useful subsequently. (14. it is often convenient to preserve the symmetry between the si j ’s and use three (dimensionless) variables x i defined by s23 = Q 2 (1 − x i ) and cyclic permutations. then (problem 14.20). In three-body final states of the type under discussion here.66) (14.Figure 14. we see that the entire perimeter of the triangle in figure 14. we now transform from the overall three-body CMS to the CMS ˜ of 2 and 3 (figure 14. In order to find the limits of the allowed physical region in this x i space. By symmetry.19. These are related by (14.19 is the required boundary: physical events fall anywhere inside the triangle.6) ˜ x 2 = (1 − x 1 /2) + (x 1 /2) cos θ ˜ x 3 = (1 − x 1 /2) − (x 1 /2) cos θ . first introduced by Copyright 2004 IOP Publishing Ltd .60). as shown in figure 14. which becomes x 1 + x 2 + x 3 = 2. which correspond to x 2 = 1 and x 3 = 1.

No integral is involved here—the tree amplitude itself becomes singular on the phase space boundary.7).21(a).21(b)) and is called a ‘soft divergence’. The factor in large parentheses is (14. Definition of θ .4). equation (14.68) where σpt is the point-like e+ e− → hadrons total cross-section of (9. Now consider the distribution provided by the QCD bremsstrahlung process. It is apparent from these figures that in either of these two cases the observed final-state hadrons. as it would be absent if the quarks had mass. will in fact resemble a two-jet configuration. which can be written equivalently as 1 d2 σ 2αs = σpt d x 1 d x 2 3π 2 2 x1 + x2 (1 − x 1 )(1 − x 2 ) (14. In contrast to this. then the factor (2 p · p )−1 .) Lines of constant θ are shown in figure 14.69) (14. which is condition (b). p1 = m 2 = 0 1 2 = m 2 = 0. or both. after the fragmentation process. ˜ Dalitz (1953) for the analysis of K → 3π .54). for example. If we restore the quark masses.62).70) which are easily interpreted physically. We can trace the origin of the infinity back to the denominator factors ( p1 · p3)−1 ∼ (1 − x 2 )−1 and ( p2 · p3)−1 ∼ (1 − x 1 )−1 in (14.19. The 1 divergence of type (a) is therefore also termed a ‘mass singularity’.20. The most striking feature of (14.19.100).˜ Figure 14.63) written in terms of the x i (problem 14.68) is that it is infinite as x 1 or x 2 . These become zero in two distinct configurations of the gluon momentum: (a) (b) p3 ∝ p1 or p3 ∝ p2 (using pi2 = 0) p3 → 0 (14. Condition (b) corresponds to the emission of a very ‘soft’ gluon (figure 14. and a factor of four has been introduced from the colour sum (problem 14. tend to 1! This is a quite different infinity from the ones encountered in the loop integrals of chapters 10 and 11. the gluon momentum p3 in type (a) does not have to be vanishingly small. Such events will be found in the regions x 1 ≈ 1 and/or x 2 ≈ 1 of the kinematical plot shown in figure 14. which correspond to strips adjacent Copyright 2004 IOP Publishing Ltd . Condition (a) corresponds to a situation in which the 4-momentum of the gluon is parallel to that of either the quark or the anti-quark: this is called a ‘collinear divergence’ and the configuration 2 is pictured in figure 14. becomes and p2 1 3 2 (( p1 + p3 )2 − m 2 )−1 which only vanishes as p3 → 0.

which needs to be defined in such a way as to be free of soft and collinear divergences. A compilation of three-jet fractions at different e+ e− annihilation energies.15. Sterman and Weinberg (1977) defined twojet events to be those in which all but a fraction of the total available energy is contained in two cones of half-angle δ .21. For example. direct evidence for asymptotic freedom. we must keep away from the boundaries of the triangle. Adapted from Akrawy et al (OPAL) (1990).22.Figure 14.and two-jet events can be done satisfactorily. Gluon configurations leading to divergences of equation (14. Assuming such a separation of three.22 a compilation of data on the fraction of three-jet events at different e+ e− annihilation energies. αs —changes with energy. their ratio carries important information—namely. figure from Ellis et al (1996). hence. we show in figure 14. in fact. and to regions near the vertices of the triangle.68) over the relevant range of x 1 and x 2 . that this fraction—and. The most remarkable feature of this figure is. Glossing over a number of technicalities (for which the reader is referred to Ellis et al (1996. This is. Figure 14. section 3.and three-jet events is done by means of a jet measure. Events inside the rounded triangular region should be mostly three-jet events. The two-jet cross-section is then obtained by integrating the right-hand side of (14. The events are viewed in the overall CMS.68): (a) gluon emitted approximately collinear with quark (or anti-quark): (b) soft gluon emission. to two of the boundaries of the triangle. it should be proportional to αs ! This follows simply from the extra factor of gs associated with the gluon emissions in figure 14. To isolate them. The quantitative separation between two. decreasing as the energy increases. of course. Copyright 2004 IOP Publishing Ltd .3)).

Q α = αβγ pβ qγ .54) is ac (c3 )χ † (c1 ) λc χ(c2 ) 2 (14.71) where c1 . given by     p2 q3 − p3 q2 Q1  p3 q1 − p1 q3  ≡  Q 2  . anti-quark and gluon respectively. Check that the three components Q α transform as a 3∗ .It is now time to start our discussion of the theoretical basis for this fundamental property. Problems 14.2) is invariant under the transformation (14.14) for each colour wavefunction. 14.2 Verify that the Lorentz-invariant ‘contraction’ µνρσ F µν F ρσ of two U(1) (Maxwell) field strength tensors is proportional to E · B. [Note: you will need the explicit forms of the λ matrices (appendix M). transforming under an infinitesimal SU(3)c transformation via p = (1 + iη · λ/2) p and similarly for q . c2 and c3 label the colour degree of freedom of the quark. you need to verify the transformation law Q = (1 − iη · λ∗ /2)Q. and a sum on the repeated index c is understood as usual.3 Verify that the cross-section for the exchange of a single massless scalar ˆ gluon contains no ‘1/t 2 ’ factor. The χ’s are the colour wavefunctions of the quark and anti-quark and are represented by three-component column vectors: a convenient choice is       1 0 0 χ(r ) =  0  χ(b) =  1  χ(g) =  0  0 0 1 Copyright 2004 IOP Publishing Ltd . in the particular case for which only the parameters η1 . The ‘colour wavefunction’ part of the amplitude (14.1 (a) Show that the anti-symmetric 3q combination of equation (14. η3 and c η8 are non-zero. η2 .] ˆ ˆ 14. p1 q2 − p2 q1 Q3 that is. Consider the anti-symmetric combination of their components. (b) Suppose that pα and qα stand for two SU(3)c colour wavefunctions.68).4 This problem is concerned with the evaluation of the ‘colour factor’ needed for equation (14. 14.

Copyright 2004 IOP Publishing Ltd .60)). which have been indicated explicitly.c3 (λc )rs (λd )lm ∗ χs (c2 )χl† (c2 ) χm (c1 )ad (c3 ) 2 2 (14.72) to a trace and to evaluate it as follows: (i) Show that χs (c2 )χl† (c2 ) = δsl .c2 . 14.68) is four. c=1 and hence that the colour factor for ( analogy with the spin wavefunctions of SU(2).63) becomes the factor in large parentheses in equation (14.72) becomes 1 4 8 Tr(λc )2 .67).5.55). when expressed in terms of the x i ’s. In this form the expression is very similar to the spin summations considered in chapter 8 (cf equation (8.72) where summation is understood on the matrix indices on the χ ’s and λ’s.68). 14. c2 (ii) Assuming the analogous result ∗ ac (c3 )ad (c3 ) = δcd c3 show that (14. section M.71) and summing over the colour labels ci : ac (c3 )χr† (c1 ) c1 .5 Verify equation (14. The cross-section is obtained by forming the modulus squared of (14. We proceed to convert (14. show that 8 Tr(λc )2 = 16. 14. c=1 (iii) Using the λ’s given in appendix M.6 Verify equation (14.7 Verify that expression (14.

there are ambiguities between two-jet and three-jet events. the gluon loops of figure 15. but did not give any details there of how the problem is solved. which is relevant to the high-energy (ultraviolet) divergence of Feynman integrals. where ‘v’ stands for ‘virtual’. Renormalization has nothing to offer the infrared problem. Indeed. We ran into exactly the same trouble in chapter 11 for the case of the analogous QED corrections. The gluon loops of figure 15. discussed in the last section of the previous chapter. as 115 Copyright 2004 IOP Publishing Ltd . not unexpected: the gluon momenta in the loops run over all possible values. has amplitude Fγ . In fact. The total cross-section from these contributions is thus proportional to |Fγ + Fg.2. The O(αs ) QCD corrections to Fγ are shown in figure 15.1 (assuming we are far from the Z0 peak). O(αs ).2: we denote the amplitude for the sum of these processes by Fg. of course. in the limit of zero quark mass. The point is that ¯ an outgoing quark (or anti-quark) cannot be distinguished. since these involve the emission and then reabsorption of gluons.v |2 —and this leads to a problem. arising from ‘collinear’ configurations of the quarks and gluons in the loops. after all.2 would also. This is.15 QCD II: ASYMPTOTIC FREEDOM.17.1).1 QCD corrections to the parton model prediction for σ (e+ e− → hadrons) We begin by considering QCD corrections (at the one-loop. THE RENORMALIZATION GROUP AND SCALING VIOLATIONS IN DEEP INELASTIC SCATTERING 15. level) to the simple parton model prediction for the total e+ e− annihilation cross-section into hadrons (see figure 14. kinematically. But they turn out also to diverge as the (virtual) gluon momenta approach zero. from one which is accompanied by a soft or collinear gluon—just as. of course.2 contain. exhibit further (non-ultraviolet) divergences.v . Now we need to be more explicit. say. The cure lies in a careful analysis of what is actually meant by the total annihilation cross-section to qq. The parton model graph. Such ‘soft’ divergences are usually called ‘infrared’ when they occur in loops—and they are not cured by renormalization. the usual (‘ultraviolet’) divergences at large momenta. shown again in figure 15. it is the latter processes which hold the key to dealing with these troublesome soft and collinear divergences of figure 15. in the appropriate kinematic regions. including those which gave trouble in the real gluon emission processes of figure 14.

we saw in section 14.v |2 ) 2 α .v |2 : one of order α 2 (from |Fγ |2 ). (15.1. (15. Thus and an interference term of order α s at order α 2 αs . The complete cross-section at order α 2 αs is found to be (see.Figure 15. as defined by a jet measure of some kind.2) The remarkable cancellation of the soft and collinear divergences between the real and virtual emission processes is actually a general result. another of order α 2 αs (from |Fg. The Bloch– Nordsieck (1937) theorem states that ‘soft’ singularities cancel between real and virtual processes when one adds up all final states which are indistinguishable by virtue of the energy resolution of the apparatus.2. (15. A theorem due to Kinoshita (1962) and Lee and Nauenberg (1964) states. One-photon annihilation amplitude in e+ e− → qq.v |2 and one in σg. Virtual gluon corrections to figure 15. This refinement hardly seems to have helped: we now have two lots of ‘infrared’ type divergences to worry about.r .1) will diverge due to the vanishing denominators (soft and collinear divergences).1).4. ¯ Figure 15. Note that three terms appear when squaring out 2 |Fγ + Fg. This leads to a cross-section for the production of real soft and collinear gluons which is given by σg.v |2 should also be added the contribution to the total cross-section due to production of soft (and collinear) gluons in these dangerous kinematical regions.1.1) must be added precisely to this interference term—and the wonderful fact is that their divergences cancel. which is the same order as (15. to | Fγ + Fg.19 close to the triangular boundary.r = σpt 2αs 3π 2 2 x1 + x2 dx 1 dx 2. one in | Fγ + Fg. (1 − x 1 )(1 − x 2 ) η (15. This will entail integrating (14.1) Clearly. that mass Copyright 2004 IOP Publishing Ltd . for example. Pennington (1983) or Ellis et al (1996)) σ = σpt (1 + αs /π). Thus. as the region η (parametrized in some way) tends to zero. Yet now comes the miracle.68) precisely over some area (call it η) of figure 14. roughly speaking.

2). singularities are absent if one adds up all indistinguishable mass-degenerate states.e.3. when effectively a quark or anti-quark appears in the initial state. being essentially the origin of scaling violations in deep inelastic scattering. to the 2 order α 2 αs in the cross-section). We return to (15. which is dispelled as soon as we go to the next order in αs (i.Figure 15. But in the case of deep inelastic scattering. Some typical graphs contributing to this order of the cross-section are shown in figure 15. They are of great importance physically. this is an illusion. some graphs will contribute via their modulus squared and some via interference terms). Copyright 2004 IOP Publishing Ltd .7. In the case of e+ e− annihilation. However. The result was obtained numerically by Dine and Sapirstein (1979) and analytically by Chetyrkin et al (1979) and by Celmaster and Gonsalves (1980). as with the O(α 2 αs ) terms.1 so that (assuming the expansion parameter is αs /π) the implied perturbation series in powers of αs would seem to be rapidly convergent.1 would suggest that αs ∼ 0. but which for the moment may be thought of as related in some way to an average 1 A more precise extraction of α can be made from the value of R at the Z0 peak. The value of αs at the Z0 peak is approximately equal to 0. Comparison with the data shown in figure 14.3) where µ is a mass scale (about which we shall shortly have a lot more to say. At first sight. Some higher-order processes contributing to e+ e− → hadrons at the parton level. For our present purposes. as we shall see in section 15.5 or less.12. the crucial feature of the answer is the appearance of a term σpt · −b 2 αs s ln π µ2 (15.3 (note that. this result might appear satisfactory.1. such uncancelled mass singularities will occur. If an ‘inclusive’ final state is considered (as in e+ e− → hadrons). see Ellis et al s (1996). these do not arise since the gluon cannot be attached to the e+ or e− lines. then only mass singularities from initial lines will remain.

namely αs {1 − bαs ln(s/µ2 )}.6) σpt 1 + {1 − bαs ln(s/µ2 )} . of these higher-order corrections is to convert αs to an s-dependent quantity.3) raises the following problem.55) is positive.4. αs ≈ 0. 3 Pressing on with the next order (α 2 αs ) terms. for the case of QED.8)? And what exactly is the mass parameter µ? To address these questions we need to take a substantial detour. (15.7) and higher. GeV)2 . Gorishny et al 1991) σpt · αs b ln s µ2 2 αs .3).56)) and their effect is such as to reduce the effective value of αs as s increases—exactly the property of asymptotic freedom. one remarkable difference: here the coefficient of the ln is negative. at least.5) once again! We are now strongly inclined to suspect that we are seeing.4) ¯ where Nf is the number of ‘active’ flavours (e.2) and (15.quark mass). which is αs (15.3) and (15. however.5) If we take Nf = 5.g.5) is of order 1 and can. µ ∼ 1 GeV and s ∼ (10 (15.8) The ‘re-summation’ effected by (15.8) has a remarkable effect: the ‘dangerous’ large logarithms in (15.3) and (15. we consider the sum of (15. in equation (11. we encounter a term (Samuel and Surguladze 1991. 1 + αs b ln(s/µ2 ) (15.7) and the ratio between this and (15. If true. s is the square of the total e+ e− energy (which we called Q 2 in section 14.61)) σpt 1 + αs /π . this would imply that all terms of the form (15. Nf = 5 above the bb threshold). in no sense. Suppose that. The term (15. We have seen something very like this before. and the coefficient b is given by b= 33 − 2 Nf 12π (15. In (15. π (15.3) is precisely (15. an expansion of the form (1 + x)−1 = 1 − x + x 2 − x 3 · · ·.7) are now effectively in the denominator (cf (11. whereas that in (11. π This suggests that one effect. are we guaranteed that still higher-order terms will indeed continue to contain pieces corresponding to the expression (15. We hasten to say that of course this is not how the property was discovered! The foregoing remarks leave many questions unanswered—for example. be regarded as a small perturbation.55). however. sum up to give (cf (11. The ratio between it and the O(α 2 αs ) term is clearly −bαs ln(s/µ2 ). Apart from this (vital!) difference.3). nevertheless.4). we can reasonably begin to think in terms of an effective ‘s-dependent strong coupling constant αs ’. in this class of terms. There is. Copyright 2004 IOP Publishing Ltd .

We learned in chapter 11 that the renormalized charge e was given in terms of the ‘bare’ charge 1/2 e0 by the relation e = e0 (Z 2 /Z 1 )Z 3 (see (11. Let us go back to the renormalization of the charge in QED. so that only Z 3 is needed.6).9) where. As we have just mentioned.30)) [2] Z3 = 1 + [2] γ (0) (15.2 The renormalization group and related ideas 15. due to the 1/2 Ward identity Z 1 and Z 2 are equal (section 11. from (11. To order e2 in renormalized perturbation theory. no ghosts). so that it is only at quite unrealistically enormous |q 2 | 4 values that the corresponding factor (α/3π) ln(|q 2 |/ m 2) (where m e is the electron e mass) becomes of order unity. the factor corresponding to αs b ∼ 1 is α/3π ∼ 10−3 . and to which we refer for a more detailed discussion. fundamentally have to do with renormalization. where.Figure 15. the origin of the logarithmic term is essentially the same in both cases and the technicalities are much simpler for QED (no photon self-interactions.24). in fact. which we have found very illuminating. however. therefore. For the answer does. We shall.23) and (11. forget about QCD for a while and concentrate on QED.4. 15.10) and γ = m 2 − x(1 − x)q 2 with q 2 < 0. There. Nevertheless. the phenomenon of ‘large logarithms’ arises also in the simpler case of QED.6)). Indeed.1 in Weinberg (1996). in fact. Z 3 is given by (cf (11. the discussion of renormalization of QED given in chapter 11 will be sufficient to answer the question in the title of this subsection.4. One-loop vacuum polarization contribution to Z 3 . including only the e+ e− loop of figure 15.1 Where do the large logs come from? We have taken the title of this section from that of section 18. [2] 2 γ (q ) = 8e2 i 0 1 dx x(1 − x) d4 k 4 (k 2 − 2 (2π) γ +i ) (15.2. We regularize the k integral by a cut-off e Copyright 2004 IOP Publishing Ltd .

3π It is not a coincidence that the coefficient α/3π of the ultraviolet divergence is also the coefficient of the ln(|q 2 |/m 2 ) term in (11.12) α ln( /m e ).12)). So we must enquire: what determines the ‘some mass’? With this question we have reached the heart of the problem (for the moment).1) [2] 2 γ (q ) =− e2 π2 0 1    dx x(1 − x) ln   + 2 γ 1 2 +  γ −   ( 2 + γ )2 1  . we find that (15. Now. this is a perfectly ‘natural’ definition of the renormalized charge—but it is by no means forced upon [2] us.54)–(11. determines Z 3 . as explained in sections 10. (15.3.. which entered [2] via the quantity [2] (0) or. The answer is. via the renormalization constant Z 3 (cf γ (15. not immediately obvious: it lies in the prescription used to define the renormalized coupling constant—this prescription. the appearance of a singularity in Z 3 as m e → 0 suggests that it is inappropriate to the case in which fermion masses are neglected.30)) was determined from the requirement that the O(e2 ) corrected photon propagator (in the ξ = 1 gauge) had the simple form −igµν /q 2 as q 2 → 0. Now.2. In fact.36). in Copyright 2004 IOP Publishing Ltd .14) as in (11. We could.54) was arrived at. that mass scale is evidently m e . In the present case. that is. which is defined by the ‘subtracted’ form ¯ [2] (q 2 ) = γ [2] 2 γ (q ) − [2] γ (0). obtaining (problem 15.13) disappears and we In the process of subtraction. the dependence on the cut-off are left with ¯ [2] (q 2 ) = − γ 2α π 1 0 dx x(1 − x) ln m2 e m 2 − q 2 x(1 − x) e (15. whatever it is. Why is it m e that enters into [2] (0) or Z 3 ? Part of the answer— γ once again—is of course that a ‘ln ’ cannot appear in that form but must be ‘ln( /some mass)’. as the photon goes on-shell. this leads to the ‘large log’ term (α/3π) ln(|q 2 |/m 2 ).1 and 10. The value (15. We first recall how (11.3. For large values of |q 2 |.56): we need to understand e why. in order to form such a term.11) Setting q 2 = 0 and retaining the dominant ln [2] (Z 3 )1/2 = 1 − term. (15. it is obviously not e possible to have just ‘ln |q 2 |’ appearing: the argument of the logarithm must be dimensionless so that some mass scale must be present to which |q 2 | can be compared. in fact.9) (or (11. equivalently. This is the beginning of the answer to our questions. It refers to the renormalized selfenergy part.

however. a considerable arbitrariness in the way renormalization can be done—a fact to which we did not draw attention in our earlier discussions in chapters 10 and 11. that if the point of renormalization is to render amplitudes finite by taking certain constants from experiment. the logarithm is now ln(|q 2 |/µ2 ) which is e 2 | is of order µ2 . actually. we have forgotten that. We shall come back to this important point shortly.2 Changing the renormalization scale The recognition that the renormalization scale (−µ2 in this case) is arbitrary suggests a way in which we might exploit the situation so as to avoid large ‘ln(|q 2 |/m 2 )’ terms: we renormalize at a large value of µ2 ! Consider what e [2] happens if we define a new Z 3 by [2] Z 3 (µ) = 1 + [2] 2 γ (q = −µ2 ). therefore. that with this different small when |q renormalization prescription we have ‘tamed’ the large logarithms.15) Then for µ2 m 2 but µ2 e 2.16) [2] (Z 3 (µ)) 2 = 1 − and a new renormalized self-energy [2] 2 γ (q . Certainly the difference between [2] (q 2 = 0) and [2] (q 2 = −µ2 ) is finite as γ γ → ∞. Thus. it must somehow be the case that. the ‘charge’ defined at q 2 = −µ2 rather than at q 2 = 0. say q 2 = −µ2 . Nevertheless. (15. so such a redefinition of ‘the’ renormalized charge would only amount to a finite shift. despite this arbitrariness. even a finite shift is alarming to those accustomed to 1 a certain ‘sanctity’ in the value α = 137 ! We have to concede. in terms of e0 . the ‘e’ we should now be using is the one defined. via [2] eµ = (Z 3 (µ)) 2 e0 = 1 − 1 α ln( /µ) e0 3π (15.principle. 1 we have α ln( /µ) 3π (15.2.18) rather than [2] e = (Z 3 ) 2 e0 = 1 − 1 α ln( /m e ) e0 . for consistency. then any choice of such constants should be as good as any other—for example. Nevertheless. (15. 3π (15. It seems. there is.17) m 2 − q 2 x(1 − x) e = − e2 2π 2 dx x(1 − x) ln For µ2 and −q 2 both m 2 . However. choose a different value of q 2 .19) Copyright 2004 IOP Publishing Ltd . 15. µ) = [2] 2 γ (q ) − 1 0 [2] 2 γ (q = −µ2 ) m 2 + µ2 x(1 − x) e . the physical results remain the same. at which to ‘subtract’.

23) (15. invalidating the approach. Returning to (15. Then we can proceed to eµ for µ not too different from µ. Differentiating (15. these other variables are µ /µ and m e /µ. Symanzik 1970): it governs the change of the coupling constant eµ as the renormalization scale µ changes. a finite shift from ‘e’ to ‘eµ ’ but the problem with it is that a ‘large log’ has resurfaced in the form of ln(µ/m e ) (remember that our idea was to take µ2 m 2 ). 0) ≡ β(eµ ) dµ (15.6) and the correction analogous to (15. z=1 (15. m e /µ) = µ deµ = β(eµ .18). it is more convenient to consider infinitesimal steps—that is. Instead of making one jump from m 2 to a large e value µ2 . we regard eµ at the scale µ as being a continuous function of eµ at scale µ and of whatever other dimensionless variables exist in the problem (since the e’s are themselves dimensionless). a similar procedure applied to QCD will involve the larger coefficient bαs as in (15. The relation between eµ and e is then α (1 − 3π ln( /µ)) α e ≈ 1+ ln(µ/m e ) e eµ = (15.21) with respect to µ and letting µ = µ. we obtain µ where deµ = β(eµ .20) indeed represents.20) α (1 − 3π ln( /m e )) 3π to leading order in α.22) For µ ∂ E(eµ . m e /µ) dµ . m e /µ). m/µ) ∂z m e .20) is certainly small.22) reduces to β(eµ . Rather than thinking of such a process in discrete stages m e → µ → µ → · · ·. as anticipated. We have to be more subtle. it is easy to calculate the crucial quantity β(eµ ). z.24) which is a form of the Callan–Symanzik equation (Callan 1970. µ /µ. To this one-loop order. we need to proceed in stages. Equation (15. In the present case. and so on. so that eµ must have the form eµ = E(eµ .21) (15. We can calculate eµ from e as long as µ is not too different from m e .25) . e Although the numerical coefficient of the log in (15. we may write the bare coupling e0 as e0 = eµ 1 − ≈ eµ ≈ eµ Copyright 2004 IOP Publishing Ltd α ln( /µ) 3π α ln( /µ) 1+ 3π αµ ln( /µ) 1+ 3π −1 (15. equation (15.20) will be of order 1.working always to one-loop order with an e+ e− loop.

in terms of its value at some scale µ = M .29) ln(µ2 / M 2 ) αM ln µ2 / M 2 which is αµ = 1− αM 3π (15.20) can be brought under control via (15. The solution of (15. but small enough so that ln( M/ m e ) terms do not invalidate the perturbation theory calculation of e M from e.28) e2 M 1− e2 M 12π 2 (15. We note that the general solution of (15. (15.26) 3 Working to order eµ . β(e) (15. The crucial point is that the ‘large log’ is now in the denominator (and has coefficient α M /3π !). e0 (15. Hence. 2 eµ = 1 e2 M − 1 2 eµ (15.25).30). differentiating (15.2) ln(µ/ M) = 6π 2 or. we can drop the last term in (15. The final piece in the puzzle is to understand how this can help us with the large −q 2 behaviour of our cross-section.31) We have made progress in understanding how the coupling changes as the renormalization scale changes and in how ‘large logarithmic’ change as in (15.24) may be written as µ = M exp eµ eM de .27) to obtain eµ at an arbitrary scale µ.25) with respect to µ at fixed e0 . the problem from which we originally started. which would be a higher-order correction to (15.26).30) where α = e2 /4π . obtaining finally (to one-loop order) µ d eµ dµ = e0 3 eµ 12π 2 ≡ β [2] (eµ ) . Now the unrenormalized coupling is certainly independent of µ. chosen in practice large enough so that for variable scales µ greater than M we can neglect m e compared with µ. Copyright 2004 IOP Publishing Ltd . we find d eµ dµ − e0 2 eµ d eµ eµ αµ − ln( /µ) · 3πµ 4π 2 dµ = 0. equivalently.27) is then (problem 15.where the last step follows from the fact that e and eµ differ by O(e3 ).27) We can now integrate equation (15.

34) is then µ2 ∂ ∂µ2 + β(αµ ) ∂ ∂αµ S(|q 2 |/µ2 . this condition is represented by the equation ∂ ∂µ or µ ∂ ∂µ + β(eµ ) eµ + eµ deµ dµ e0 ∂ ∂eµ ∂ ∂eµ S(|q 2 |/µ2 . (15.1 and 15.15. so as to be consistent with all the foregoing discussion.36) Copyright 2004 IOP Publishing Ltd . who were the first to discuss the freedom associated with the choice of renormalization scale.3 The renormalization group equation and large −q 2 behaviour in QED To see the connection we need to implement the fundamental requirement. The ‘group’ connotation is a trifle obscure—but all it really amounts to is the idea that if we do one infinitesimal shift in µ2 and then another. Gell-Mann and Low (1954) who realized how equation (15. (15. eµ ).21). however. We need to work in the space-like region.2. and hence of σ . pretending that the one-loop corrections we are interested in are those due to QED rather than QCD. It was. It is convenient to work in terms of µ2 and α rather than µ and e.2. stated at the end of section 15. To make this clear. the result will be a third such shift. as we now explain.34) could be used to calculate the large |q 2| behaviour of S. it is a kind of ‘translation group’. S can only be a function of the dimensionless ratio |q 2 |/µ2 and of eµ : S = S(|q 2 |/µ2 . we shall now denote the 4-momentum of the virtual photon by q rather than Q and take q 2 < 0 as in sections 15.35) αµ where β(αµ ) is defined by β(αµ ) ≡ µ2 ∂αµ ∂µ2 .33) S(|q|2 /µ2 .34) Equation (15. The terminology goes back to Stueckelberg and Peterman (1953). αµ ) = 0. (15.34) is referred to as ‘the renormalization group equation (RGE) for S’. eµ ) = 0. Equation (15. Neglecting all masses.1. our annihilation cross-section σ for e+ e− → hadrons. But S must ultimately have no µ dependence. in other words. It follows that the µ2 dependence arising via the |q 2 |/µ2 argument must cancel that associated with eµ . e0 (15. Consider. that predictions for physically measurable quantities must not depend on the renormalization scale µ. Bearing in mind the way we used the ‘dimensionless-ness’ of the e’s in (15. In symbols. eµ ) = 0 (15. This is why the µ2 -dependence of eµ controls the |q 2 | dependence of S. for example. let us focus on the dimensionless ratio σ/σpt ≡ S.

36) and (15. (15. (15.44) It follows that S(1.27). From (15.43) the minus sign coming from the fact that αµ is the lower limit in (15.35) then becomes − ∂ ∂ + β(αµ ) S(et . we have to recall how to differentiate an integral with respect to one of its limits—or.40) with respect to t at fixed αµ : we obtain 1= 1 ∂α(|q 2 |) .31): α(|q 2 |) dα . 2 |)) ∂t β(α(|q (15.40) with respect to αµ at fixed t (note that α(|q 2 |) will depend on µ and hence on αµ ): we obtain 0= 1 ∂α(|q 2 |) 1 − ∂αµ β(α(|q 2 |)) β(αµ ) (15. differentiate (15. so that |q 2 |/µ2 = et .From (15.42) and (15. we deduce that. this result is only valid in a regime of −q 2 which is much greater than Copyright 2004 IOP Publishing Ltd .38) This is a first order differential equation which can be solved by implicitly defining a new function—the running coupling α(|q 2 |)—as follows (compare (15.41) First. Of course. the formula ∂ ∂a f (a) g(x) dx = g( f (a)) ∂f . ∂t ∂αµ (15.40) t= β(α) αµ To see how this helps. ∂a (15. to the one-loop order to which we are working.39) (15.37) β [2] (αµ ) = 4π 3π Now introduce the important variable t = ln(|q 2 |/µ2 ). α(|q 2 |)) is a solution of (15. Equation (15.40). 2 αµ eµ [2] β (eµ ) = . let us differentiate (15. αµ ) = 0. we find the result − ∂ ∂ + β(αµ ) α(|q 2 |) = 0.43).42) Next. This is a remarkable result. more generally.39). ∂t ∂αµ (15. It shows that all the dependence of S on the (momentum)2 variable |q 2| enters through that of the running coupling α(|q 2 |).

(15. But the logic is exactly the same. 3π (15. Copyright 2004 IOP Publishing Ltd . on plausibility grounds. say. It follows that the ‘leading log corrections’ at high −q 2 are summed up by replacing this expression by αs (|q 2 |)/π.3 Back to QCD: asymptotic freedom The reader may have realized. Hence. Provided αµ is small. This is why the technique applies only at ‘high’ −q 2 .45) in powers of αµ : S(1. In fact.47). αµ ) by S(1. the squares of all particle masses which do not appear in (15.37)—to which we now turn. which generates the terms 2 S2 α 2 (|q 2 |) = S2 αµ 1 + 2αµ t + ··· . All we need to do is solve (15. 2 The difference has to do.46).all quantities with dimension (mass)2 —for example. α(|q 2 |)) sums all ‘leading log terms’.4) must be precisely the coefficient of αs in the one-loop contribution to the of course.46) where t = ln(|q 2|/µ2 ). The result implies that if we can calculate S(1. where the running αs (|q 2 |) is determined by solving (15. then replacing αµ by α(|q 2 |) will allow us to predict the q 2 -dependence (at large −q 2 ).32). with the different renormalization prescriptions. corrections.2 Suppose now that the leading perturbative contribution to S(1. Equation (11. then. . integrating (15.47) comes with one more power of αµ than in (15. the leading terms in t. α(|q 2 |)) ≈ S1 α(|q 2 |) = S1 αµ 1 − = S1 αµ 1 + αµ t + 3π αµ t 3π αµ t 3π 2 −1 + ··· (15. It is in this sense that replacing S(1. we have seen that β [2] (α) = α 2 /3π. The next higher-order calculation of S(1. 15.46) and (15. for QED with one e+ e− loop. Indeed. The leading (O(αs )) perturbative contribution to S = σ/σpt at q 2 = −µ2 is given in (15.40) we obtain αµ αµ . the one-loop (and higher) corrections to S in which we are really interested are those due to QCD. αµ ) (i.40).56) is written in terms of an ‘α’ defined at q 2 = 0 and without neglect of m e .40) with the QCD analogue of (15.e. αµ ) is S1 αµ .46).56). α(|q 2 |)) in this approximation can be found by expanding (15. . are contained in (15. we see that each power of the large log factor appearing in (15. S at the point q 2 = −µ2 ) at some definite order in perturbation theory. rather than QED. of course. t 2 . .45) α(|q 2 |) = αµ = αµ 1 − 3π t 1 − 3π ln(|q 2 |/µ2 ) This is almost exactly the formula we proposed in (11. some time back. αµ ) would be 2 S2 αµ . that the quantity b introduced in 2 (15.2) as αs (µ2 )/π. Then the terms contained in S(1.47) Comparing (15.

Gross and Wilczek (1973) and ’t Hooft. in QCD.3 we gave the conventional physical interpretation of the way in which the running of the QED coupling tends to increase its value at distances |q|−1 short enough to probe inside the screening provided by e+ e− 3 Except that. The difficulty is evaded in the approach of section 15.8). with ∂αs ∂µ2 .51) leads to (15. it is wanted at large time-like values. or Peskin and Schroeder (1995) section 18. we now understand.8). the quantity b is positive.3 Thus. The resolution. and then their re-summation in terms of the running coupling. lies in quite subtle properties of renormalized quantum field theory. The justly famous result (15. at least at high |q|2 . we repeat.48) fixed bare αs [2] 2 βs = −bαs (15. In section 11.2.β -function of QCD defined by β s = µ2 that is to say. (15.3: how the successful parton model could be reconciled with the undoubtedly very strong binding forces between quarks. because it immediately explained the ‘paradoxical situation’ referred to at the end of section 14.2) by (15. the strong coupling runs in the opposite way to QED.49)) is opposite to that of the QED analogue.51).6 below. For example. (15. a conclusion reinforced by the demonstration by Coleman and Gross (1973) that no theory without Yang–Mills fields possessed the property of asymptotic freedom. in (15.45) is replaced by αs (µ2 ) αs (|q 2 |) = (15. was not published.37). at sufficiently high |q|2 . announced at a conference in Marseilles in 1972.50) 12π For Nf ≤ 16. The result of Politzer and of Gross and Wilczek led rapidly to the general acceptance of QCD as the theory of strong interactions.50) explain the success of the parton model: it also.5. (15.3. Readers troubled by this may consult Pennington (1983) section 2.2. ’t Hooft’s result.50) was first obtained by Politzer (1973).49) 33 − 2 Nf . involving first the exposure of ‘large logarithms’. opened the prospect of performing reliable perturbative calculations in a strongly interacting theory. α is evaluated at large space-like values of its argument. The published calculation of Politzer and of Gross and Wilczek quickly attracted enormous interest. Not only did the result (15. Copyright 2004 IOP Publishing Ltd . we can reliably compute the β function in perturbation theory.5. so that the sign of (15. whereas in s (15. Correspondingly. equation (15.51) [1 + αs (µ2 )b ln(|q 2 |/µ2 )] b= and then replacing αs in (15. becoming smaller at large values of |q 2 | (or small distances)—the property of asymptotic freedom.

in agreement with the second term in (15. Hence. β ) on the external legs of the diagram. The one-loop graphs contributing to the calculation of b are shown in figure 15. 2 the way the λ’s are conventionally normalized).50) via the term −2 Nf /12π . despite appearances. graphs involving the gluon self-coupling which is absent in QED and also (in covariant gauges) ghost loops. but there are also. It is simple to check that (15. It arises from the ‘q¯ ’ vacuum polarization diagram of figure 15. characteristically.5 is the presence of the 1 λ-matrices at 2 each vertex. the value of which can be quite easily understood. If ‘a ’ and ‘b ’ are the colour labels of the ingoing and outgoing gluons. This vacuum polarization screening effect is also present e in (15.48) is only a first-order differential equation.50). in fact. chapter 16. The all-important.5. for example) but it would be nice to have a simple intuitive picture of the ‘anti-screening’ result in terms of the gluon interactions. positive. By introducing ln Copyright 2004 IOP Publishing Ltd 2 QCD = ln µ2 − 1/(bαs(µ2 )) (15. the 1 λ-matrix factors must be 2 3 α.β=1 1 λa 2 αβ 1 λb 2 (15. We do not want to enter into the details of the calculation of β(αs ) here (they are given in Peskin and Schroeder 1995. no fully satisfactory simple explanation exists. The only new feature in figure 15.52) βα since there are no free quark indices (of type α.6. q pairs (|q|−1 m −1 ). we note that. Unfortunately. q¯ vacuum polarization correction to the gluon propagator. say. first term must therefore be due to the gluons.52) has the value 1 δab (this is. They include figure 15.Figure 15. (15.51). 1981) and Nielsen (1981) for a ‘paramagnetic’ type of explanation rather than a ‘dielectric’ one.5. of course. q which is precisely analogous to the e+ e− diagram used to calculate ¯ [2] (q 2 ) in γ QED. though the reader may be interested to consult Hughes (1980.5. it does not really involve two parameters—after all. for one quark flavour we expect ‘α/3π ’ to be replaced by ‘αs /6π’.53) . Returning to (15.

an integration constant representing the scale at which αs would diverge to infinity (if we extended our calculation beyond its perturbative domain of validity). while perturbative QED is characterized by a dimensionless parameter α. Note that this is a distance scale of order (200 MeV)−1 ∼ 1 fm.3) and a straight line a quark. just about the size of a hadron—a satisfactory connection.40)) ln |q 2 |/ 2 QCD =− ∞ αs (|q 2 |) dαs [2] βs (15. The extraction of a precise value of QCD is a complicated matter. More usefully.54) Equation (15. a broken line a ghost (see section 13.Figure 15. but a typical value is in the region of 200 MeV. QCD is a measure of the scale at which αs really does become ‘strong’.5. (15. Significantly. therefore.6. perturbative QCD requires a mass parameter (µ2 or QCD ). Copyright 2004 IOP Publishing Ltd .5. A curly line represents a gluon. as we shall briefly indicate in section 15.54) is equivalent to (cf (15.3) as αs (|q 2 |) = 1 b ln(|q 2|/ 2 QCD ) .51) can be rewritten (problem 15.55) [2] 2 with βs = −bαs . Graphs contributing to the one-loop β function in QCD. equation (15. QCD is.

although gauge invariance does imply generalizations of the Ward identity used to prove Z 1 = Z 2 (Taylor 1971.So far we have discussed only the ‘one-loop’ calculation of β(αs ). First we want to explore the RGE further. The three-loop result has been obtained by Tarasov et al (1980) and by Larin and Vermaseren (1993). for QCD.5.7).9) (this equality was discussed in section 11. because we are renormalizing at scale µ. The reason for the simplification in QED was the equality between the renormalization constants Z 1 and Z 2 . Slavnov 1972). let us return to the one-loop calculation of β for QED. the consequence is no longer the simple relation ‘Z 1 = Z 2 ’ in this case. 24π 2 (15. The two[3] 3 loop result (Caswell 1974. e0 (15. due essentially to the ghost contributions. all the Z i ’s depend on µ (as in (15.16)) but we shall now not indicate this explicitly.48) gives a corrected expression for αs (|q 2 |) in terms of αs (µ2 ) which has to be solved numerically for αs (|q 2 |). Taking logs and differentiating with respect to µ at constant e0 .4 A more general form of the RGE: anomalous dimensions and running masses The reader may be wondering why. In order to see what change Z 1 = Z 2 would make. pretending that Z 1 = Z 2 . We recall that Z 2 is the field strength renormalization factor for the charged fermion in QED and Z 1 is the vertex part renormalization constant: their relation to the counter terms was given in equation (11.6 are needed. We have Z1 − 1 e0 = Z 2 eµ (15. whereas for QED we got away with only figure 11. 15. Typically. For QCD. all the graphs of figure 15. the b coefficient is associated with ln ln(|q 2 |/µ2 ) terms. which therefore cancelled out in the relation between the renormalized and bare charges e and e0 . where bb = 153 − 19 Nf .56) [2] [3] Inserting βs and βs into (15.58) Hence. we obtain µ d dµ ln Z 1 − µ e0 d dµ ln Z 2 − e0 1 d µ 2 dµ ln Z 3 + e0 µ deµ eµ dµ = 0. Jones 1974) may be written as βs = −bb αs . β(eµ ) ≡ µ deµ dµ = eµ γ3 + 2eµ γ2 − eµ µ e0 d ln Z 1 dµ (15.6.59) Copyright 2004 IOP Publishing Ltd .57) Z2 3 where. We shall return to σ (e+ e− → hadrons) in section 15. as briefly stated before equation (15.6).

αµ ) = / SF (q 2 ). in addition to the vacuum polarization loop associated with Z 3 . just as Z 3 of (15.62) Suppose we now ask: what is the large −q 2 behaviour of (15. α ). the latter ˜ ˆ was dimensionless. Z 2 enters into the relation ¯ ˆ ˆ between the propagator of the bare fermion |T (ψ 0 (x)ψ0 (0))| and the renormalized one. the normalization of the ψ’s will change via the Z 2 factors—by a finite amount here—and we must include this change when writing down the analogue of (15. we need to introduce the contributions from loops determining the fermion field strength renormalization factor. the γ3 term in (15.61) is. as well as those related to the vertex parts (together with appropriate ghost loops). of course.2)) ¯ ˆ ˆ |T (ψ(x)ψ(0)| = 1 Z2 ¯ ˆ ˆ |T (ψ 0 (x)ψ0 (0))| (15. the other two terms in (15.2.3 for the quantity S(|q 2 |/µ2 . Thus. So if.16) is used for Z 3 . q Accordingly. We might guess that.62) for space-like m 2 where m is the fermion mass? This sounds very similar to the q 2 .61) where (cf section 10.61) will—as previously noted—depend on the renormalization scale µ. to get the leading large |q 2 | behaviour we will need to as for S(|q µ ˜ calculate RF to some order in αµ and then replace αµ by α(|q 2 |/µ2 ). when we change 1 2 ˆ µ.60) To leading order in eµ .58) cancelling via Z 1 = Z 2 . via (cf (11.where γ2 ≡ 1 d µ 2 dµ ln Z 2 e0 γ3 = 1 d µ 2 dµ ln Z 3 . e0 (15.34) for this case (i.27) when (15. whereas (recalling that ψ has mass dimension 3 ) SF (q 2 ) has 2 −1 . The factor Z 2 in (15. we extract this ( / )−1 factor (compare σ/σpt ) and consider the q ˜ dimensionless ratio RF (|q 2 |/µ2 . Quantities such as γ2 and γ3 have an interesting and important significance.60) reproduces (15. on changing µ.16) did. ∂αµ (15.1. eµ ).e. with −q 2 question answered in 15. Z 1 is not equal to Z 2 . just q˜ 2 |/µ2 .3) | is the vacuum of the interacting theory. the condition that the ‘total change. which we shall illustrate in the case of γ2 for QED. We can gain insight into the essential difference caused by the presence of γ2 by Copyright 2004 IOP Publishing Ltd . However. But this is not quite all. The Fourier transform of (15.63) is somewhat more complicated than that of (15.34). is zero’). The required equation is µ2 ∂ ∂µ2 + β(αµ ) ∂ ˜ + γ2 (αµ ) RF (|q 2 |/µ2 . This dimensionality is just what a propagator of the free-field dimension M form i /( / − m) would provide. as in the case of QCD. the Feynman propagator: ˜ SF (q 2 ) = d4 xeiq·x ¯ ˆ ˆ |T (ψ(x)ψ(0))| . (15.63) αµ The solution of (15. αµ ) = 0.

3. The full solution of (15. ˜ since its presence effectively means that the |q 2 | behaviour of SF is not determined −1 . (15.6.66) can be evaluated ˜ by setting α(t ) = α ∗ . in fact. we easily find ˜ RF (|q 2 |/µ2 .7. the coupling will evolve to the ‘fixed point’ α ∗ . near this point the evolution of α is given by (cf (15. Anomalous dimensions turn out to play a vital role in the theory of critical phenomena—they are. asymptotically for large |q 2 |.e. as we shall see in section 15. (15. (15. Scaling with anomalous dimensions is also exactly what occurs in deep inelastic scattering of leptons from nucleons.66) The first factor is the expected one from section 15. The behaviour of such an α is shown Copyright 2004 IOP Publishing Ltd . α(|q 2 |/µ2 )) exp t 0 dt γ2 (α(t )) . meaning that if we multiply |q 2 | by a ˜ scale factor λ.3 and Peskin and Schroeder 1995. then SF is multiplied by λγ2 (eµ )−1 rather than just λ−1 .considering the special case β(eµ ) = 0. in the vicinity of which β(α) ≈ −B(α − α ∗ ) with B > 0.2. that at ˜ 2 large |q large |q 2 | F 1 ˜ SF (|q 2|/µ2 . SF has an ‘anomalous’power law dependence 2 | (i. The behaviour (15. αµ ) = RF (1. In this case. and RF will scale with an anomalous dimension γ2 (α ∗ ) determined by the fixed point value of α. chapter 13). the integral in (15. which is controlled by the on |q / parameter γ2 . We quote it here: ˜ ˜ RF (|q 2 |/µ2 ). In this case. see also Peskin and Schroeder (1995) section 12. Suppose now that β(α) has a zero at some point α ∗ .67) α(|q 2 |) = α ∗ + constant × (µ2 /|q 2|) B . The latter is called the ‘anomalous dimension’ of the fermion field. at a zero of the β -function.64) ˜ But since RF can only depend on µ via |q 2 |/µ2 .3. Then. in other words. the second results from the addition of the γ2 term in (15. (15. in addition to the obvious q −1 factor). chapter 3).68) Thus. αµ ) ∝ q / |q 2 | µ2 γ2 (αµ ) .40)) ln(|q 2 |/µ2 ) = which implies α(|q 2 |) αµ dα −B(α − α ∗ ) (15.63) for β = 0 is elegantly discussed in Coleman (1985.63). closely related to ‘critical exponents’ (see section 16. αµ ) ∝ (µ2 )−γ2 (αµ ) .65) is often referred to by its ‘normal’ dimensionality M as ‘scaling with anomalous dimension’. at sufficiently large −q 2 .65) ˜ Thus. we learn that if β = 0 then the 2 | behaviour of R is given by (|q 2 |/µ2 ) 1 γ2 —or.

such as the ABC model or to gauge-invariant quantities. However. and so γ2 depends on the choice of gauge. (b) the slope is negative at the origin (as in QCD) and ∗ positive near αs = αs . renormalization is required (in the shape of a mass counter-term) to achieve a pole at the ‘right’ physical mass m. in thinking about how to re-introduce the quark masses into our formalism. a definition we followed in chapters 10 and 11. at a given order of α . But this prescription Copyright 2004 IOP Publishing Ltd . This is an appropriate point at which to consider the treatment of quark masses in the RGE-based approach. Possible behaviour of β functions: (a) the slope is positive near the origin (as in QED) and negative near α = α ∗ . the latter therefore being neglected. The quantity Z 2 is not gauge invariant in QED (or QCD).4) that αs is reached in the infrared limit q 2 → 0. in figure 15.Figure 15.7. Significantly. This is really no surprise because the full fermion propagator itself is not gauge invariant (the free-field propagator is gauge invariant. in this sense. so that the corresponding behaviour (with a zero at a αs = 0) would look like that shown in figure 15.7(a). Thus. This applies equally to a fixed point at the origin.7(b). the analysis given here really only applies—in this simple form—to non-gauge theories. In this case. The point α ∗ in figure 15. of course). so that QED is infrared stable at α = 0 while QCD is ultraviolet stable at αs = 0. the reader can check (problem ∗ ∗ 15. s. We emphasize that there is no reason to believe that the QED β function actually does behave like this. While this may be adequate for the light quarks u.e. it seems surely a progressively worse assumption for c.7(a) is called an ultraviolet-stable fixed point: α ‘flows’ towards it at large |q 2 |. What ultimately matters is that the complete physical amplitude for any process. b and t. d. In the case of QCD. Up to now we have simply assumed that the relevant |q 2| is very much greater than all quark masses. the β function starts out ∗ negative. the ‘on-shell’ value p2 = m 2 ). We must now point out to the reader an error in the foregoing analysis in the case of a gauge theory. we are at once faced with a difficulty: how are they to be defined? For an unconfined particle such as a lepton. Clearly it is the slope of β near the fixed point that determines whether it is ultraviolet or infrared stable. it seems natural to define ‘the’ mass as the position of the pole of the propagator (i. and so αs is called an infrared-stable fixed point. be gauge invariant.

which never ‘escapes’ beyond the range of the confining forces and whose propagator can. αs . the ABC coupling itself had mass ¯ ˆ ˆ dimension 1). just as the gauge (or other) couplings do. therefore.66) of (15. The RGE in such a scheme now takes the form ∂ R(|q 2 |/µ2 . m/|q|) = 0 ∂m i (15. this will get translated into a |q 2 |-dependence of the mass parameters. (15. just as for α(|q 2 |) and αs (|q 2|).71) d ln |q 2 |γm (αs (|q 2 |). Inserting the one-loop solution for αs in the form Copyright 2004 IOP Publishing Ltd .42) for the running αs . arising from finite shifts in the mass parameter when the scale µ2 is changed. the operator ψ(x)ψ(x) is gauge invariant.1). Effectively. ∂t Equation (15. never approach the free form ∼ ( / − m)−1 . as is any such local operator.71) has the solution m(|q |) = m(µ ) exp 2 2 |q 2 | µ2 ∂ ∂ + β(αs ) + ∂αs ∂µ2 (15. Then equation (15. Here the γi are the anomalous dimensions relevant to the quantity R. no necessity to define the QED coupling parameter e via an on-shell prescription. Formally. becomes ∂(ln m(|q 2 |)) = γm (αs (|q|2)). section 18. we can think of γm in (15. after all.69) where the partial derivatives are taken at fixed values of the other two variables.70) (15. the solution of (15. µ2 γi (αs ) + γm (αs )m ∂αs (|q 2 |) = β(αs (|q 2 |)) ∂t where t = ln(|q 2|/µ2 ).69) is given in terms of a ‘running mass’ m(|q 2 |). we ¯ ˆ ˆ are regarding the ‘m’ in a term such as −m ψ(x)ψ(x) as a ‘coupling constant’ having mass dimension 1 (and. In turn.69) as analogous to β(αs ) and ln m as analogous to αs . and γm is an analogous ‘anomalous mass dimension’.72) 1 To one-loop order in QCD. γm (αs ) turns out to be − π αs (Peskin and Schroeder 1995. p Our present perspective on renormalization suggests an obvious way forward. Just as with the solution (15. so here a mass parameter in the Lagrangian can be defined in any way we find convenient: all that is necessary is that it should be possible to determine its value from experiment. it is clear that a renormalization scale will be involved in such a general definition of mass and we must expect to see our mass parameters ‘evolve’ with this scale. Taking this point of view.63).cannot be used for a confined particle. Incidentally. Just as there was. in principle.

73). the preservation of unitarity). Further discussion of the treatment of quark masses is contained in Ellis et al (1996) section 2.73) where (π b)−1 = 12/(33 − 2 Nf). we rotate the contour CR in an anti-clockwise direction (thus avoiding the poles the location of which is determined by the ‘+i ’ term). so that it runs 4 Note that dimensional regularization can equally well be used to deal with infrared divergences. For example.54). We describe this method briefly here: some more details are given in appendix N.8. we saw in section 11. instead of evaluating it with a cut-off. at first.3 that this was a potentially dangerous procedure in a gauge theory.24)) [2] 2 γ (q ) = 8e 2 i 0 1 dx x(1 − x) d4 k . 4 (k 2 − 2 (2π) γ +i ) (15. Thus the quark masses decrease logarithmically as |q 2 | increases.4 Thus. rather like αs (|q 2 |). First. since it could spoil gauge invariance.74) would converge if the number of dimensions over which k was integrated were less than four. by allowing d to be greater than four. in general.e.23) and (11. While several gauge-invariant regularizations are available. due to ’t Hooft and Veltman (1972). as we did in figure 10. we have—for the sake of simplicity of exposition—conducted the entire discussion of renormalization effects in the framework of regularization by means of the ultraviolet cut-off . if one can somehow calculate the integral as an analytic function of the dimensionality d of spacetime. the one now used most widely is ‘dimensional regularization’. such as the photon vacuum polarization with amplitude (see (11. and so does the elimination of unphysical states (i. quark mass effects are suppressed both by explicit m 2 /|q 2 | factors and by the logarithmic decrease given by (15.5 Some technicalities We conclude our discussion of RGE’s by commenting on a number of technical matters to which we should draw the reader’s attention. The first step is to consider the k0 integral as a contour integral in the k0 plane. which the process of renormalization can remove.4. we find ln(µ2 / 2 ) m(|q |) = m(µ ) ln(|q 2 |/ 2 ) 2 2 1/π b (15. This time.1 can be handled this way. 15. the infrared divergence of section 15. The latter is vital for two reasons: renormalizability depends upon it. The basic idea is very simple (if. However. it will converge for d < 4 and have an identifiable singularity as d approaches four. Copyright 2004 IOP Publishing Ltd . rather strange). however.(15. It follows that. It is based on the observation that a typical logarithmically ultraviolet divergent one-loop diagram.

the logarithm in (15. Comparing (10.76) is indeed singular when d = 4. hence.76) where is the gamma function (see.75) where kE = (k4 .77) for general z and integer n .76).76) in powers of . where the new α is dimensionless in d dimensions and µ is ‘some mass’. using x = e ln x ≈ 1 + ln x + · · ·. one replaces the kE -integral in (15. the denominator is never zero. chapter 11).5).79) The appearance of the dimensional quantity γ inside the logarithm is undesirable. (15.77) that (z) has a pole at z = 0. Boas 1983. k ) is the ‘Euclidean’ 4-momentum.51) and (15. Then (15. the field dimension will change when we go to d dimensions and. therefore.5772. We obtain the result (problem 15. To isolate the singular behaviour we use the approximation (2 − d/2) = 2 − γ + O( ) (15. The reader Copyright 2004 IOP Publishing Ltd . Note that for q 2 < 0. so will that of the couplings if we want to keep the action dimensionless.along the imaginary axis: k0 → ik4 .78) where = 4 − d and γ (not to be confused with the anomalous dimensions!) is the Euler–Mascheroni constant. Integrals such as (15. We may rectify this by noting that although the fine structure constant α is dimensionless in four dimensions. having the value γ ≈ 0. In dimensional regularization.6) 1 2 − γ + ln 4π − ln (4π)2 γ + O( ) . to rewrite the original α as αµ . we can see that we may crudely identify ‘ 1 ∼ ln ’.76) but with powers of kE in the numerator can be evaluated similarly (see appendix N). is related to the factorial function for integer values of its argument and satisfies (z + 1) = z (z) (n) = (n − 1)! (1) = 1 (15. so that (15.7 shows that in d dimensions the coupling e has dimension (mass) /2.18) does indeed cancel in this regularization procedure (problem 15. In that case.74) becomes −8 e 2 0 1 dx d4 kE x(1 − x) 2 (2π)4 (kE + γ )2 (15. Using these results one finds that the nongauge-invariant part of (11. We may expand the right-hand side of (15. for example. It is clear from (15. It is natural. Problem 15. so there is no need to include any ‘i ’.79) becomes ln( γ /µ2 ).75) by dd k E 1 d 2+ (2π) (kE = 1 (2 − d/2) d/2 (2) (4π) 1 γ 2−d/2 γ) 2 (15.

though admittedly there are some odd-looking constants present. from the point of view of simplicity.79)? This makes the remainder perfectly finite as → 0. after all. Such a procedure is called ‘minimal subtraction’. at q 2 = 0) scheme. Hence. section 2. the relationship between the QCD ’s in different schemes is Copyright 2004 IOP Publishing Ltd .55) is scheme-dependent. We can. (15.81) since the left-hand side is independent of |q 2 |. as in γ γ (11. most conveniently. we know very well by now that this is not the only possibility and that we can instead ‘subtract’ at a different point. We make two immediate points. so that they are scheme-independent. From (15. would be to get rid of the −γ + ln 4π ≈ 1. which is. This scheme tends to reduce loop corrections to their simplest form. in addition to the (potential) dependence of calculated quantities on the renormalization scale µ2 .32). in fact. This is called ‘modified minimal subtraction’ or MS (‘em-ess-bar’) (Bardeen et al 1978).36). the two corresponding values of QCD are related by ln B A 1 = 2 c1 = 2b B αs (|q 2 |) A αs (|q 2 |) dx bx 2 (1 + · · ·) (15. just as before. the full RGE equations of Stueckelberg and Peterman (1953) include variation due to the (suitably parametrized) renormalization scheme. define the renormalized ¯ [2] (q 2 ) by subtracting from [2] (q 2 ) its value at q 2 = 0. The change from one scheme ‘A’ to another ‘B’ must involve a finite renormalization of the form (Ellis et al 1996.80) Note that this implies that the first two coefficients of the β function are unchanged under this transformation. Even better.will not need to be told that µ may be identified. but of course the resultant parameters may be less easily related to physically measurable quantities than in the ‘on-shell’ (i. But in this dimensional regularization approach. say q 2 = −µ2 .81) (15. the parameter QCD introduced in (15. This will. denoted by MS. It is now clear that. In fact. The question now arises of how renormalization is to be done in this approach—and the answer is.e. the γ and ln 4π terms in (15. with the renormalization scale introduced earlier. a very simple prescription is most appealing: why not agree to define the renormalized [2] by simply throwing away the singular term γ 2/ in (15.55). there will also be a (potential) dependence on the renormalization scheme which is used. not that small numerically.79) disappearing along with the 2/ singularity.5) B A A αs = αs (1 + c1 αs + · · ·).82) where we have taken |q 2| → ∞ in (15. First. MS is the most widely used scheme as far as RGE-type applications are concerned. if we wish. However.95 as well (this being a ‘finite renormalization’). give exactly the same result as (11. and express the ultimate independence of physical qualities both of the choice of scale and of the choice of scheme.

referring the interested reader to Pennington (1983) section 4. the less changes which are O(αs the effect of truncation should be—but.8 for five flavours. it can be shown that n changing the scale in something which has been calculated to O(αs ) induces n+1 ). in practice. and it has the value 1. We shall not pursue this particular technicality any further here.8) πβ(αs (µ2 )) dC2 µ2 2 = − (15.2) are written as σ = σpt 1 + αs (µ2 ) + π ∞ Cn (s/µ2 ) n=2 αs (µ2 ) π n (15.84) should be independent of the choice of scale µ2 (this is a ‘direct’ way of applying the RGE idea). the series (15. µ2 -independence will break down.6 σ (e+ e− → hadrons) revisited Armed with this new-found sophistication as regards renormalization matters.84) (see Ellis et al 1996. 15. In general. or the review of QCD by Hinchcliffe in Hagiwara et al 2002). The coefficient C2 (1) was calculated by Dine and Sapirstein (1979). Chetyrkin et al (1979) and by Celmaster and Gonsalves (1980).5). As soon as the perturbative expansion is truncated. For example. The higher-order (in αs ) corrections to (15. truncating at the n = 2 stage: σ ≈ σpt αs (µ2 ) 1+ + C2 (s/µ2 )(αs (µ2 )/π) π 2 . it is more common to 2 relate data to the value of αs at a particular |q 2 | value. The µ2 -dependence of the coefficients is now fixed by the requirement that. The more terms in the series one has. we may wonder if there is a ‘best’ renormalization prescription to use. and is equal to −12. section 3. for example.8) 2 MS = 2 MS exp(ln 4/π − γ ).411 for five flavours. section 2.82). The coefficient C3 (1) was calculated by Samuel and Surguladze (1991) and by Gorishny et al (1991).1.83) For this and other reasons (Ellis et al 1996. at any finite order. Consider. for example. we must stress that the µ holds when they are evaluated exactly. changing from MS to MS gives (problem 15.86) dµ (αs (µ2 ))2 Copyright 2004 IOP Publishing Ltd . order by order.2 or to Ellis et al (1996) section 3.85) Differentiating with respect to µ2 and setting the result to zero we obtain (problem 15. 2 -independence of physical quantities only Second. we may now return to the physical process which originally sparked this extensive detour. usually taken to be MZ . (15.determined by the one-loop calculation which gives c1 in (15.1. which minimizes the various ambiguities. (15.

3 where an O(αs ) term has been dropped. Substituting the one-loop result (15. The logarithmic terms violate this simple (power law) scaling. the latter would scale as λ−2 when the momenta are all scaled by a factor λ.9. This will be the topic of the final section of this chapter. The current fitted value of αs ( MZ ) is 0. The terms in (15. In practice.7) and (9. The obvious first QCD corrections will be due to real gluon emission by either the initial or final quark. 15. predict observable violations of scaling in deep inelastic scattering.1 and predicted to be scale-invariant functions in the free-parton model.4)). The parton model cross-section σpt is proportional to s −1 as could be predicted on simple dimensional grounds.49)—as is consistent to this order—we find C2 (s/µ2 ) = C2 (1) − π b ln(s/µ2 ). Thus. As we shall see. In this section we shall show how QCD corrections to the simple parton model. to a good first approximation.10 in order (see later) to get rid of Copyright 2004 IOP Publishing Ltd . calculated using RGE techniques.8). but to these we must add the one-loop virtual gluon processes of figure 15.2. comparison between the theoretical predictions and experimental measurements provides strong evidence for the correctness of QCD as the theory of nucleonic constituents. We should also wonder whether something similar will happen to the structure functions introduced in section 9. and Q 2 and ν are defined in (9. if this were the only contribution to σ .8 (cf figure 9. only on the dimensionless ratio x = Q 2 /2 Mν . 15.3). if all masses are neglected. rather than on Q 2 and ν separately: this behaviour is referred to as ‘scaling’. (15. Here M is the nucleon mass.1 Uncancelled mass singularities The free-parton model amplitudes we considered in chapter 9 for deep inelastic lepton–nucleon scattering were of the form shown in figure 15.7 QCD corrections to the parton model predictions for deep inelastic scattering: scaling violations As we saw in section 9.87) gives the contribution identified in (15. the parton model provides a simple intuitive explanation for the experimental observation that the nucleon structure functions in deep inelastic scattering depend.84) which involve logarithms are referred to as ‘scaling violations’ (compare the discussion of anomalous dimension in section (15. as shown in figure 15. corrections are necessary to account for effects due to the b and √ t quark masses (Chetyrkin and Kuhn 1993).012 with an error of less than 5% (Hagiwara et al 2002). and (at sufficiently high s ) for Z0 2 effects.87) The second term on the right-hand side of (15.4).7.

46) that. the same can be done here. and also the diagram of figure 15.8.29) and (9. we shall consider what is called a ‘non-singlet structure ep NS en function’ F2 . In principle.10.30) in the case of the free-parton diagram in figure 15.9 (leaving aside figure 15. We now want to perform. Figure 15. in general. Virtual single–gluon corrections to figure 15.8.Figure 15.2 we solved the first problem by explicitly calculating the parton level d2 σ i /dQ 2 dν i i and picking off the associated νW2 .9. such as F2 –F2 in which the (flavour) singlet gluon contribution cancels out. leaving only the diagrams of figures 15. Electron–quark scattering via one-photon exchange. However. infrared divergences similar to those encountered in section 15.10. a simpler—if somewhat heuristic—way is available.8. In section 9.2. F1 = M W1 is given by the transverse virtual photon Copyright 2004 IOP Publishing Ltd .1. for these diagrams. We note from (9. W1 . the other is to relate these parton W ’s to observed nucleon W ’s via an integration over momentum fractions. calculations analogous to those of section 9. which enabled us to find the e–N structure functions νW2 and M W1 from the simple parton process of figure 15. To simplify matters.10 for the moment)—cf equations (9. starting from the five-fold differential cross-section for our e− + q → e− + q + g process. Electron–quark scattering with single-gluon emission Figure 15.11 corresponding to the presence of gluons in the nucleon. There are two problems here: one is to find the parton level W ’s corresponding to figure 15.8.9 and 15.

11.91). for the parton processes of figure 15. where. and so F2 /x = 2 F1 = 2 M W1 = σT /(4πα 2 /2 M K ). the Callan–Gross relation is still true (the photon only interacts with the charged partons. Further. which are quarks with spin. and replaced α 2 by ei 2 ααs . renamed the variables t → u. (15. taking the same (Hand) convention for the individual photon flux factors.Figure 15. ¯ cross-section W1 = σT /(4πα 2 / K ) = 1 2 λ=±1 ∗ µν µ (λ) ν (λ) W (15. Also.180) to (15. ˆ Thus.90) d cos θ 4 π ei 2 ααs 3 s ˆ ˆ s ˆ t ˆ 2u Q 2 − − + ˆ s ˆ t s tˆ ˆ (15.91) where.88) where W µν was defined in (9. Hence. in going from (8. u → t in accordance with figure 15. we have inserted a colour factor 4 3 (problem 15.9).3).9. Before proceeding with (15. ˆ ˆ ˆ ˆ 2 F1 = σT /(4π 2 α/2 M K ) = s ˆ 4π 2 α 1 −1 (15.12. Electron–gluon scattering with qq production. as usual. ‘ˆ’ denotes kinematic quantities in the corresponding parton process.12. is—apart from a colour factor—just the virtual Compton crosssection calculated in section 8. This crosssection.91). however.6. divided by the factor 4π 2 α/2 M K . with ˆ ˆ appropriate changes for parton masses M .1 2 and charge ei ). ˆ ˆ 2M K = s. the parton level 2 F1 for figure 15.9 is just the transverse photon cross-section as calculated from the ˆ ˆ graphs of figure 15. it is helpful to consider the other part of ˆ the calculation—namely the relation between the nucleon F1 and the parton F1 .2 but with one significant difference: the quark ‘taken’ from the proton has momentum fraction y (momentum yp) but now Copyright 2004 IOP Publishing Ltd .89) These formulae are valid for both parton and proton W1 ’s and W µν ’s. We mimic the discussion of section 9.

13.3 for help). section 10. we let k.93) d y f i ( y) 0 1 ˆi dz 2 F1 δ(x − yz) (15. Figure 15. of (15.12. k . and the sum is over contributing partons. k be the magnitudes of the CMS momenta k. Then.Figure 15. viewed as a contribution to e− –nucleon scattering.14.94)).e.91) inserted into (15. Let us call the quark momentum after gluon emission zy p (figure 15. Referring to figure 15. we need to look at the kinematics of the γ q → qg process. The first process of figure 15. its longitudinal momentum must be degraded in the final state due to the gluon bremsstrahlung process we are calculating. in the CMS. The reader may enjoy checking that (15. so that 2 F1 = i ei2 f i (x).13). To proceed further with the calculation (i.31)) F2 = 2 F1 = x 1 i 0 x = Q 2 /2q · p q 2 = −Q 2 (15. Virtual photon processes entering into figure 15.94) does ˆi reduce to (9. we have (15. Then s = 4k 2 = (yp + q)2 = Q 2 (1 − z)/z ˆ Copyright 2004 IOP Publishing Ltd z = Q 2 /(ˆ + Q 2 ) s .34) for free partons by showing that in that case 2 F1 = ei2 δ(1 − z) free (see Halzen and Martin 1984. assuming as in section 9.92) q 2 + 2zyq · p = 0 or x = yz and we can write (cf (9.2 that it stays on-shell.12.9.2 (we often call them q(x) or g(x) as the case may be) for parton type i .94) where the f i ( y) are the momentum distribution functions introduced in section 9.

91) for F1 . ˆ t = (q − p )2 = −2kk (1 − cos θ ) = −Q 2 (1 − c)/2z u = (q − q )2 = −2kk (1 + cos θ ) = −Q 2 (1 + c)/2z. this QCD correction contributes (from (15.91) only the terms proportional to tˆ−1 : ˆi 2 F1 ≈ ei2 1 −1 dc 1−c αs 4 1 + z 2 2π 3 1 − z (15.4—or. If we simply replace the propagator factor tˆ−1 = [(q − p )2 ]−1 by [(q − p )2 − m 2 ]−1 . we have here an uncancelled mass singularity and it violates scaling. in the terminology of section 14. occurring in the zero quark mass limit. Thus m regulates the divergence. a ‘mass singularity’.98) and so.e. As anticipated in section 15. then (15. Kinematics for the parton process of figure 15.94)) a term ei2 αs 1 dy q(y){Pqq(x/y) ln(Q 2 /m 2 ) + C(x/y)} (15. when q and p are parallel). We may calculate the coefficient of the ln Q 2 term by retaining in (15. This crucial physical result is present in the lowest-order QCD correction to the parton model. ˆ c = cos θ (15.97) which will produce a factor of the form ln( Q 2 / m 2 ) as m 2 → 0.Figure 15.14. This is a divergence of the ‘collinear’ type.99) 2π x y Copyright 2004 IOP Publishing Ltd . for just one quark species.1. we shall obtain an infinite result ∼ 1 dc 1−c (15. in this case. where m is a quark mass.96) becomes ∼ 1 dc (1 + 2m 2 z/ Q 2 ) − c (15.96) ˆ associated with the vanishing of t in the ‘forward’ direction (i. when we integrate over c = cos θ .95) ˆ We now note that in the integral (15. Such logarithmic violations of scaling are a characteristic feature of all QCD corrections to the free (scaling) parton model. as there.13.

Williams 1934. The application of these techniques to QCD corrections to the free parton model is due to Altarelli and Parisi (1977) who thereby opened the way to the previous much simpler and more physical way of understanding scaling violations. physically. that the ‘free’ quark distribution function q(x). and hence the interaction between them is very strong. Copyright 2004 IOP Publishing Ltd . becomes modified to q(x) + αs 2π 1 x dy q(y){Pqq(x/y) ln(Q 2 /m 2 ) + C(x/y)} y (15. beyond the reach of a perturbative calculation. Q 2 ) to this order in αs . we have 2F1 (x.103) ≡ ei2 q(x. The function Pqq has an important physical interpretation: it is the probability that a quark.101). which depended only on the scaling variable x.102) y which we have to take from experiment. Q 2 ) = ei2 1 x αs dy q(y. 15.101) due to lowest-order gluon radiation. µ2 ) given by q(x.3. since it diverges as m 2 → 0. is left with the fraction z of its original momentum. In terms of this quantity. We must find some way of making sense. Note that we have also absorbed the non-singular term C(x/y) into q(x. having radiated a gluon. section 4.2) is to realize that when two partons are in the collinear configuration their relative momentum is very small. Clearly.2 Factorization and the DGLAP equation The key (following Ellis et al 1996. and for one quark type. µ2 ) = q(x) + αs 2π 1 x dy q(y)Pqq(x/y){ln(µ2 /m 2 ) + C(x/y)} (15. of this uncancelled mass divergence in (15. then. Similar functions arise in QED in connection with what is called the ‘equivalent photon approximation’ (Weizs¨ cker 1934. and C(x/y) has no mass singularity.100) is called a ‘splitting function’. where Pqq (z) = 4 3 1 + z2 1−z (15. which had previously been discussed mainly within the rather technical operator product formalism (Wilson 1969). therefore. µ2 ) δ(1 − x/y) + Pqq (x/y) ln(Q 2 /µ2 ) y 2π (15. Chen and Zerwas a 1975).7. this corrected distribution function violates scaling because of the ln Q 2 term but the result as it stands cannot represent a well-controlled approximation. This suggests that we should absorb such uncalculable effects into a modified distribution function q( 2F1 . Our result so far is. µ2 ).

1978b). beyond (15.e. Q 2 ) = ei2 1 x αs dy q(y. The scale µ can be thought of as one which separates the perturbative short-distance physics from the non-perturbative long-distance physics. Q 2 ) = ei2 q(x. We have demonstrated the possibility of factorization only to O(αs ) but proofs to all orders in perturbation theory exist: a review is provided in Collins and Soper (1987). It is the scale entering into the separation in (15. for most purposes the factorization scale is chosen to be the same as the renormalization scale. Just as for the renormalization scale. in which ultraviolet divergences are controlled by similarly importing some quantities from experiment. This is called the MS factorization scheme. of course. This is why we obtained the simple result 2F1 (x. µ2 ).103) partially with respect to µ2 . Different factorization schemes can also be employed. the more terms that can be included in the perturbative contributions to the mass-singular terms (i.105) to this order. Partons emitted at large transverse momenta contribute to the shortdistance (calculable) part of the cross-section. µ2). In fact. is here referred to as a ‘factorization scale’.104) between one (uncalculable) factor which depends on the infrared parameter m but not on Q 2 and the other (calculable) factor which depends on Q 2 . Naturally. In this case 2F1 (x. where CMS is a calculable function. is absorbed into the distribution. and setting the result to zero. in addition to the singularity. Returning now to (15. In this example. This scheme is called the ‘deep inelastic’ scheme (DIS) (Altarelli et al 1978a. Differentiating (15.104) The arbitrary scale µ. Q 2 ) δ(1 − x/y) + C (x/y) y 2π MS (15.103). we absorbed all of it into q(x. as the notation has already implicitly assumed.106) Copyright 2004 IOP Publishing Ltd . µ2) = ∂µ2 2π 1 x dy Pqq (x/y)q(y. we obtain (to order αs on the right-hand side) µ2 αs (µ2 ) ∂q(x. as pointed out.101)). the weaker the dependence on µ will be. in analysing data the same factorization scheme must be employed consistently throughout. very reminiscent of ultraviolet renormalization.e.102). µ2 ). the reader can guess what is coming next: we shall impose the condition that the physical quantity F1 (x. partons emitted at small transverse momenta < µ (i. approximately collinear processes) should be considered as part of the hadron structure and are absorbed into q(x. we have essentially made use of the simple fact that ln(Q 2 /m 2 ) = ln(Q 2 /µ2 ) + ln(µ2 /m 2 ). In (15. Q 2 ) will be given by an expression of the form F1 (x. (15. y (15. Q 2 ) must be independent of the choice of factorization scale µ2 . depending on how the non-singular part C(x/y) is treated. A more common scheme is that in which the mass singularity is regulated in dimensional regularization and only the ‘ln 4π − γ ’ bit. Q 2 ) in (15. Thus.This procedure is.103). which is analogous to a renormalization scale.

A more sophisticated treatment (Georgi and Politzer 1974. the nucleon may be pictured as q having more and more constituents. µ2 ) to get smaller at large x as µ2 increases. Q 2 = µ2 )/ei2 .11) in which quarks are generated from the gluon distribution. and larger at small x (due to the build-up of slower partons). The change in the distribution for a quark with momentum fraction x . 0 it can be used to predict how the distribution changes (or ‘evolves’) as µ2 varies. One (theoretically convenient) way is to consider ‘moments’ (Mellin transforms) of the structure Copyright 2004 IOP Publishing Ltd . In general.106) will have to be supplemented by terms (calculable from figure 15. µ2 ) can be found experimentally via 0 q(x.16 p shows the Q 2 -dependence of the proton structure function F2 (x. Thus. (We recall from (15. Clearly at larger x (x ≥ 0.103).15. there is a predicted tendency for the distribution function q(x. the right-hand side of (15. the radiated gluons produce more q¯ pairs at small x .3 Comparison with experiment Data on nucleon structure functions do indeed show such a trend. 15. after Dokshitzer (1977).106) shows that. as its structure is probed on ever smaller distance (larger µ2 ) scales. It is called the DGLAP equation. all contributing to its total momentum. The equations must then be closed by a corresponding one describing the evolution of the gluon distributions (Altarelli 1982). This probability is high for large momentum fractions: high-momentum quarks lose momentum by radiating gluons.7. although perturbation theory cannot be used to calculate the distribution function q(x. The derivation here is not rigorous—for example we have assumed that it is correct to use αs (µ2 ) on the right-hand side.This equation is the analogue of equation (15. Q 2 ) for various fixed x values. In addition. µ2 ) = 2 F1 (x. Thus. Gross and Wilczek 1974) confirms the result and extends it to higher orders. which radiated away (via a gluon) a fraction x/ y of its momentum with probability (αs /2π)Pqq (x/ y). The effect is illustrated qualitatively in figure 15. Gribov and Lipatov (1972) and Altarelli and Parisi (1977) (it is also often referred to as the Altarelli–Parisi equation).) Replacing µ2 by Q 2 then tells us how the 0 0 structure function evolves with Q 2 . and is a fundamental equation in the theory of perturbative applications of QCD. the function gets smaller as Q 2 increases. Such equations can be qualitatively understood as follows. µ2 ) at any particular value µ2 = µ2 . via (15.103) that q(x.13). while at smaller x it increases. is given by the integral over y of the corresponding distribution for a quark with momentum fraction y . Q 2 ) = ei2 x f i (x. as compiled by B Foster.36) describing the running of the coupling αs with µ2 . while maintaining the integral of the distribution over x as a constant. which absorbs the virtual photon. Figure 15. A D Martin and M G Vincter for the 2002 Particle Data Group review (Hagiwara et al 2002). Equation (15. Fits to the data have been made in various ways.

to this (one-loop) order. Evolution of the distribution function with µ2 .106) and interchanging the order of the x and y integrations. we find that n dMq (t) dt = αs (t) 2π 1 0 dy y n−1 q(y.107) where we have now introduced the variable t = ln µ2 .109) we obtain n dMq (t) dt = αs (t) n n γ M (t).112) d ln αs 8πb Copyright 2004 IOP Publishing Ltd . functions. =− (15.106)—which is of convolution type—has been reduced to product form by this transformation. y (15. 8π qq q (15.49) that dαs 2 = −bαs (15.110) Thus the integral in (15. defined by n Mq (t) = 1 0 dx x n−1 q(x.108) Changing the variable to z = x/y in the second integral and defining n γqq = 4 1 0 dzz n−1 Pqq (z) (15.15.111) dt with b = (33 − 2Nf )/12π as usual. t) (15.Figure 15.48) and (15. t) 0 y dx (x/y)n−1 Pqq (x/y). Now we also know from (15. Taking moments of both sides of (15.110) becomes n n d ln Mq γqq n ≡ −dqq . Thus (15.

This is clearly a standard infrared divergence (the quark momentum yzp after gluon emission becomes equal to the quark momentum yp before emission) and we expect that it can be cured by including the virtual gluon diagrams of figure 15.112) is found to be n Mq (t) = n Mq (t0 ) αs (t0 ) αs (t) n dqq . in such a way as to make the integrals (15.16. This has been verified explicitly by Copyright 2004 IOP Publishing Ltd .10. (15.100) is singular as z → 1. as indicated at the start of the section (and as was done analogously in the case of e+ e− annihilation). The function Pqq (z) of (15.109) for γn diverge.Figure 15.113) At this point we hit one more snag—but a familiar one. Q 2 -dependence of the proton structure function F2 for various fixed x values (Hagiwara et al 2002). p The solution to (15.

dv . we find (problem 15. b. Pqq is defined by 1 0 + dz f (z) Pqq (z) = 1 0 dz [ f (z) − f (1)] Pqq (z). 1979). at z → 1): for any function f (z) regular as z → 1.117) and then n dqq = 2 4 +4 1− 33 − 2Nf n(n + 1) j =2 1 . d. in the leading + log approximation. the number of quarks minus the number of anti-quarks) of any flavour is conserved as t varies: d dt From (15. (15. µ2 ) ¯ ¯ (where f = u v . Distributions of x times the unpolarized parton distributions f (x.118) Copyright 2004 IOP Publishing Ltd . s. u. g) using the MRST2001 parametrization (Martin et al 2002) 2 = 10 GeV2 (from Hagiwara et al 2002). including the effect of the gluon loops.106) this implies 1 0 + d x Pqq (x.116) Applying this prescription to γn .10) that n γqq = − 2 8 1− +4 3 n(n + 1) n j =2 1 j n (15.17. at a scale µ Kim and Schilcher (1978) and by Altarelli et al (1978a. Alternatively.114) (15.e. j (15. one can make the physical argument that the net number of quarks (i. This fixes the contribution of these loops (which only enter. t) = 0 1 0 d x q(x.115) + where Pqq is the complete splitting function.Figure 15. (15. c. t) = 0.

ln Mq .Figure 15. versus the logarithm of another. Data on the structure function F2 in muon–proton deep inelastic scattering. Equation (15. γgq and γgg . Pgq and Pgg will enter.18.106) is then replaced by a 2 × 2 matrix equation for the evolution of the quark n n and gluon moments Mq and Mg . The curves are QCD fits (Martin et al 1994) as described in the text. t). In the more general case. from BCDMS (Benvenuti et al 1989) and NMC (Amaudruz et al 1992).113). m n /d m ). folded appropriately with the gluon distribution n n n function g(x. one way of testing it is to plot the logarithm of n m one moment. analogous splitting functions Pqg . Returning to (15. Data support this values. These should give straight lines with slopes (dqq qq Copyright 2004 IOP Publishing Ltd . for different n. together with the related quantities γqg . We emphasize again that all the foregoing analysis is directly relevant only to distributions in which the flavour singlet gluon distributions do not contribute to the evolution equations. ln Mq .

more and more partons (gluons. by Nature. it is worth pausing to reflect on how far our understanding of structure has developed. from the simple ‘fixed number of constituents’ models which are useful in atomic and nuclear physics. First. the moments cannot be determined from the data without some additional assumptions. t0 ) in some way. due to Martin et al (1994).5 Verify using dimensional regularization that the non-gauge-invariant part of (11. A global numerical fit is then performed to determine the best values of the parameters. the region of small x (say x ≤ 10−2 ) requires special treatment and is the subject of ongoing theoretical and experimental interest (Ellis et al 1996 section 4.109)) is not chosen accidentally. We conclude this chapter with two comments. via quantum field theory. in a way quantitatively predicted by QCD. The same γ ’s are indeed anomalous dimensions (cf section 15. including QCD which enters into αs (t). to the desired scale.4) of certain operators which appear in Wilson’s ‘operator product’ approach to scaling violations. When nucleons are probed on finer and finer scales. Readers keen to pursue this may consult Peskin and Schroeder (1995.28). Figure 15. These may then be evolved numerically.17 shows a typical set of distributions at µ2 = 10 GeV2 (Martin et al 2002). to which reference was made earlier (Wilson 1969).2 Verify equation (15.83). Second.18) cancels (see the text following equation (11.1 Verify equation (15.1 prediction (Bosetti et al (1978)). chapter 18). via the DGLAP equations.3 Check that (15. Problems 15.54). Copyright 2004 IOP Publishing Ltd /2 in d = . 15. in relativistic quantum field theory.11). Finally.6 Verify equation (15. 15. applicable to the non-singlet case too of course. 15.7(b).4 Verify that for the type of behaviour of the β function shown in ∗ figure 15. since data do not exist for arbitrarily small x . q¯ q pairs) appear.79). 15. 15. For further details of QCD fits to deep inelastic data the reader is referred to Ellis et al (1996.22)). An example of such a fit. αs is reached as q 2 → 0. as discussed by Ellis et al (1996)) constitutes a remarkable vote of confidence. the ‘γ ’ notation for the moments of the splitting functions (as in (15. However. is shown in figure 15. The precise experimental confirmation of these predictions (and many others. 15. A more direct procedure.18. 15.51) can be rewritten as (15. chapter 4). is to choose a reference point µ2 and parametrize the parton distribution 0 functions fi (x.7 Check that the electromagnetic charge e has dimension (mass) 4 − dimensions.6).8 Verify equation (15.

117).15. Copyright 2004 IOP Publishing Ltd .9 Using the results of problem 14.10 Verify equation (15. explain why the colour factor for (15. 3 15.91) is 4 .6.

thus far. if we cannot get to grips with them via the specific divergent integrals supplied by perturbation Copyright 2004 IOP Publishing Ltd . one of the triumphs of theoretical physics over the past 50 years has been the success of renormalization techniques. As we have seen several times in this book. one soon encounters integrals that have to be evaluated numerically.1 Introduction Throughout this book. in electromagnetic (chapter 9) or weak (chapter 20) processes? It is unlikely that answers to all these questions are going to be found by performing calculations analytically—as is possible. via the RGE. that this result itself required the summation of an infinite series of perturbative terms. we have relied on perturbation theory as the calculational tool. however.3. then it should be able to explain. ultimately. justifying its use in the case of QCD by the smallness of the coupling constant at short distances: note. for the tree diagrams of perturbation theory. The standard way of doing this is to approximate the integral by some kind of discrete sum. first in ‘taming’ the ultraviolet divergences of quantum field theories and then in providing quantitative predictions for short-distance phenomena in QCD. Here QCD we have no prospect of getting useful results from perturbation theory: it is the non-perturbative regime. the vast amount of data that exists in low-energy hadronic physics. Thus. If QCD is indeed the true theory of the interaction between quarks. the concomitant of asymptotic freedom is that αs really does become strong at small Q 2 or at long distances of order −1 ∼ 1 fm. Even in perturbation theory.16 LATTICE FIELD THEORY AND THE RENORMALIZATION GROUP REVISITED 16. of course. But this is precisely the regime in which quarks bind together to form hadrons. But this immediately raises a question for any non-perturbative approach: how can we regulate the ultraviolet divergences and therefore define the theory. For example: what are the masses of mesons and baryons? Are there novel colourless states such as glueballs? Is SU(2)f or SU(3)f chiral symmetry spontaneously broken? What is the form of the effective interquark potential? What are the hadronic form factors. however. But there is a more fundamental point involved here. As remarked at the end of section 15. a mesh of points (in general multi-dimensional) has to be chosen: perhaps it would make sense to formulate the theory on such a mesh—or ‘lattice’—in the first place.

Of course. 1975) was the first to propose. The way this is done is briefly described in section 16. This relationship has proved to be extremely fruitful. which was briefly mentioned in section 5. renormalization questions will.5. Thus the theory is automatically ultraviolet finite from the start. we assume that we shall still want to formulate the theory in terms of an action of the form ˙ S = d4 x Ä(φ. since it is hard to see how they could be numerically simulated. as we shall discuss in section 16. As long as the lattice spacing is much smaller than the physical size of the hadrons one is studying.6. (16. Lorentz invariance is sacrificed in such an approach and replaced by some form of hypercubic symmetry: we must hope that for small enough a this will not matter. 16. there is now a corresponding maximum momentum = π/a (see following equation (16. allowing physical insights and numerical techniques to pass from one subject to the other.4 (see also appendix O). This at once seems to rule out any formalism based on non-commuting operators.2. The sum over paths approach does not involve quantum operators. in a way that has been very beneficial to both. the same would be true of ordinary quantum mechanics.1) Copyright 2004 IOP Publishing Ltd . The discrete formulation of quantum field theory should be suitable for numerical computation. a formulation does exist which avoids operators: Feynman’s sum over paths approach. of course. In particular. This method is the essential starting point for the lattice approach to quantum field theory and it will be briefly introduced in section 16. The chapter ends with some sample results obtained from lattice simulations of QCD. the lattice version of the theory should be a good approximation. Indeed. one quite natural non-perturbative way of regulating ultraviolet divergences is to approximate continuous spacetime by a discrete lattice of points. Since no two points can ever be closer than a . but fermions still have to be accommodated somehow. enter when we consider the a dependence of our parameters. without presupposing the existence of any perturbative expansion. Postponing until section 16. We shall discuss how a simple field theory is ‘discretized’ in the next section: the following one will show how a gauge theory is discretized.6)) in the lattice version of the theory. Fortunately.4. ∇φ. as we shall see in section 16.2 Discretization We start by considering a simple field theory involving a bosonic field φ.theory? We need to be able to regulate the divergences in a way which does not rely on their appearance in the Feynman graphs of perturbation theory. Such a lattice will introduce a minimum distance— namely the lattice spacing ‘a ’ between neighbouring points.2. the physics of renormalization and of the RGE is considerably illuminated from a lattice/statistical mechanics perspective. φ). As Wilson (1974. It turns out that this formulation enables direct contact to be made between quantum field theory and statistical mechanics.4 the question of exactly how we shall use it.

by formulating the theory in ‘Euclidean’ space. In that case. It follows that the maximum allowed magnitude of the momentum is then π/a. In this notation. instead of Minkowskian.8) Copyright 2004 IOP Publishing Ltd .6) As in all our previous work. which will also be discretized. i d3 x dτ ÄE ≡ iSE . N1 − 1. .3) A typical free bosonic action is then SE (φ) = 1 2 d3 x dτ [(∂τ φ)2 + (∇φ)2 + m 2 φ 2 ].1) becomes S → −i ≡ i d3 x dτ Ä φ. . . by introducing t = −iτ : further motivation for doing this will be provided in section 16.4) We now represent all of spacetime by a finite-volume ‘hypercube’. the action (16. from the start.It seems plausible that it might be advantageous to treat space and time as symmetrically as possible. . . (16. ∂φ ∂τ (16.2) (16.7) which has the inverse ˜ φ(ν1 ) = a N1 1 2 e−i2πν1 n1 /N1 φ(n 1 ) n1 (16. integrals and differentials are replaced by the finite difference expressions dx → a n1 1 ∂φ → [φ(n 1 + 1) − φ(n 1 )] ∂x a (16. N1 /2 (we take N1 to be even). 1. Then the allowed kvalues may be taken to be kν1 = 2πν1 /L with ν1 = −N1 /2 + 1. we may have N1 lattice points along the x-axis. For example. indicating that a −1 is (as anticipated) playing the role of our earlier momentum cut-off . we can alternatively consider a formulation in momentum space. a2 (16. . . . so that a field φ(x) is replaced by the N1 numbers φ(n 1 a) with n 1 = 0. 0.4.5) so that a typical integral (in one dimension) becomes dx ∂φ ∂x 2 →a n1 1 [φ(n 1 + 1) − φ(n 1 )]2 . ∇φ. . It is convenient to impose periodic boundary conditions such that φ(x) = φ(x + L). We write L = N1 a for the length of the cube side. We then write φ(n 1 ) = ν1 1 (N1 a) 1 2 ˜ ei2πν1 n1 /N1 φ(ν1 ) (16. .

ν2 . (16. (16.9) is a discrete version of the δ -function relation given in (E. 2. Following the same procedure for fermion field leads. 3.4) is 1 2 ˜ φ(kν1 ) kν1 4 sin2 (kν1 a/2) ˜ + m 2 φ(−kν1 ). The manipulations we have been going through will be recognized by readers familiar with the theory of lattice vibrations and phonons. which implies that the propagator in the discrete case is proportional to 4 sin2 (kν1 a/2) + m2 a2 (16.16) Copyright 2004 IOP Publishing Ltd .9) Equation (16. a one-dimensional version of the free action (16.4) then becomes (problem 16. a2 (16.25) of volume 1.1) 1 N1 N1 −1 n 1 =0 ei2π n 1 (ν1 −ν2 )/ N1 = δν1 .2. They satisfy M by γ E usual Minkowski ones γµ 1. The two expressions do coincide in the continuum limit a → 0. First note that the Euclidean Dirac matrices γµ are related to the M .10) while 1 2 dx ∂φ dx 2 → = 2 a2 1 2a 2 ˜ φ(ν1 ) sin2 ν1 πν1 N1 ˜ φ(−ν1 ) ˜ φ(−kν1 ). The Euclidean Dirac Lagrangian is then E ¯ ψ(x)[γµ ∂µ + m]ψ(x).13) would be replaced by 1 2 dk ˜ ˜ φ(k)[k 2 + m 2 ]φ(−k) 2π −1 (16. 2 (16. γ E } = 2δ {γµ ν µν for µ = 1. γ E ≡ −iγ M ≡ γ M . (16.2) 1 2 dx m 2 φ(x)2 → 1 m 2 2 ˜ ˜ φ(ν1 )φ(−ν1 ) ν1 (16. to E difficulties.2.13) In the continuum case.3 ≡ −iγ1. which should be written now in Hermitean form E E ¯ ¯ ¯ m ψ(x)ψ(x) + 1 {ψ(x)γµ ∂µ ψ(x) − (∂µ ψ(x))γµ ψ(x)}. A one-dimensional version of the mass term in (16. however.12) ˜ φ(kν1 )4 sin2 kν1 k ν1 a 2 Thus. 4.3 4 4 0 E .14) as usual.15) rather than to [k 2 + m 2 ]−1 (remember we are in one-dimensional Euclidean space).since (problem 16.11) (16.

17) ¯ ¯ ψ(n 1 + 1) − ψ(n 1 ) E γ1 ψ(n 1 ) a ¯ m ψ(n 1 )ψ(n 1 ) + 1 E ¯ [ψ(n 1 )γ1 ψ(n 1 + 1) 2a 2 =a n1 E ¯ − ψ(n 1 + 1)γ1 ψ(n 1 )] . Thus two modes survive as a → 0. an alternative finite a → 0 limit is found at kν1 → π/a (consider expanding a −1 sin [(π/a − δ)a ] for small δ ). effectively banishing the ‘doubled’ state to a very high mass. Banks et al 1976). 2 which changes our inverse propagator to E iγ1 1 sin(kν1 a) + m + (1 − cos(kν1 a)). a a (16.4. Susskind 1977. Further discussion of this aspect of lattice fermions is contained in Montvay and M¨ nster (1994). Various solutions to this problem have been proposed (Wilson 1975.21) By considering the expansion of the cosine near kν1 ≈ 0.20) sin(kν1 a) a But here is the (first) problem with fermions: in addition to the correct continuum limit (a → 0) found at kν1 → 0. It is a consequence of the fact that the Dirac Lagrangian is linear in the derivatives. a phenomenon known as the ‘fermion doubling problem’ (actually in four dimensions there are 16 such corners of the hypercube). u A second problem concerning fermions is the more obvious one already alluded to: how are we to represent such entirely non-classical objects. Wilson (1974).3) ¯ E sin(k ν1 a) ˜ ˜ + m ψ(−kν1 ) ψ(kν1 ) iγ1 a sin(kν1 a) a (16. Thus the propagator itself is m2 + sin2 (kν1 a) . for example.18) In momentum space this becomes (problem 16. which in particular obey the exclusion principle? We shall return to this question in section 16. a2 (16. (16. for kν1 ≈ π/a it gives a large term of order 1/a . suggested adding ¯ a term of the form 1 ψ(n 1 )[ψ(n 1 + 1) + ψ(n 1 − 1) − 2ψ(n 1 )] to the Lagrangian.19) kν1 E and the inverse propagator is [iγ1 E m − iγ1 + m]. Copyright 2004 IOP Publishing Ltd .The corresponding ‘one-dimensional’ discretized action is then a n1 ¯ m ψ(n 1 )ψ(n 1 ) + − n1 1 2 E ¯ ψ(n 1 )γ1 n1 ψ(n 1 + 1) − ψ(n 1 ) a (16. However. it can be seen that the second term disappears in the continuum limit.

Our discussion in chapter 13 should have prepared us for this: we are trying to compare two ‘vectors’ (here. It is very instructive to see how the same ideas arise naturally in the lattice case. In the usual (continuum) case. We need to parallel transport one field to the same point as the other. x) dx . Consider the quantity Ç(x. 1 φ † (n 1 ) [φ(n 1 + 1) − φ(n 1 )a] (16. y) (16. QED.3 Gauge invariance on the lattice Having explored the discretization of derivatives.24) dθ (x) . (16. y) transforms by expieθ(y) φ(y) = Ç(x.3. We illustrate the idea in the simple case of the Abelian U(1) theory. for example. we have φ(x) → eieθ(x)φ(x). y) = φ † (x) exp Ç(x. the geometrical significance of which (in terms of parallel transport) was discussed in section 13. (16. gauge invariant. for example.22) will not be gauge invariant either. We do not expect (16. Under a gauge transformation for the continuous case.27) Copyright 2004 IOP Publishing Ltd . Ç(x. Consider. we replaced φ † ∂µ φ by φ † (∂µ + ie Aµ )φ .25) Under the gauge transformation (16. A(x) → A(x) + then φ † (x)φ( y) transforms by φ † (x)φ( y) → e−ie[θ(x)−θ(y)]φ † (x)φ( y) (16. a charged boson field φ(x). so we ask: what is the discrete analogue of this? The term φ † (x) ∂∂x φ(x) becomes. x + dx) − Ç(x.23) and is clearly not invariant.26) and it is. y) → φ † (x)e−ieθ(x) exp{ie x y x ie y A dx φ(y). before they can be properly compared. and by considering the gauge-invariant quantity Adx +ie[θ(x)−θ(y)]} dx→0 lim Ç(x. The familiar ‘covariant derivative’ rule can be recovered by letting y = x + dx for infinitesimal dx.2 and 13.23).22) by itself to be gauge invariant and it is easy to check that it is not. with charge e. it is now time to think about gauge invariance. as we have seen.62 ) shows us how to do this.16. To construct a gauge-invariant current. dx (16.22) a in one dimension. fields) at two different points. The solution (13. The essential reason is that this operator involves the fields at two different points—and so the term φ † (n 1 )φ(n 1 + 1) in (16. therefore. we saw in chapter 13 how this was implemented by replacing ordinary derivatives by covariant derivatives. when the ‘coordinate axes’ are changing as we move about.

27). n x .4) the result φ † (x) d − ie A φ(x) dx (16. What about the Maxwell action for the U(1) gauge field? This does not exist in only one dimension (∂µ Aν − ∂ν Aµ cannot be formed) so let us move into two.28) (16. n y+1 ) × U (n x+1 . n 1 ) may be drawn as in figure 16. Again. Consider the product U£ of link variables around a square path (figure 16. Under a gauge transformation. n y . ‘e A’ in (16. n x . for example. In the discrete case. n y )U (n x+1 .33) . we merely keep the finite version of (16.3) guides us to the answer. n y ) − θ (n x . n 2 ) from (16. n y+1 )U (n x . the link U (n x+1 .30) A dx (16. as in the continuum form of the covariant derivative.1. n 1 + 1)φ(n 1 + 1) where the link variable U is defined by U (n 1 . n y )]} Copyright 2004 IOP Publishing Ltd (16. n y+1 .Figure 16.31) is replaced by gt a Aa (n 1 ) where the t ’s are the appropriate matrices. n 1 ) in one dimension.29) ≡ φ † (x) Dx φ(x) with the usual definition of the covariant derivative.31) → exp[−ie A(n 1)a] in the small a limit. n x+1 . n y ). that U£ is gauge invariant. transforms by a factor (cf equation (16. (16. Evaluating (16. In the non-Abelian SU(2) or SU(3) case. Thus gauge-invariant discretized derivatives of charged fields can be constructed. first. n y+1 . n y ). and replace φ † (n 1 )φ(n 1 + 1) in (16. Link variable U (n 2 . n x . A link variable U (n 2 .22) by the gauge-invariant quantity φ † (n 1 )U (n 1 . n 1 + 1) = exp ie n1 a (n 1 +1)a (16. our discussion of the geometrical significance of Fµν as a curvature (see section 13. Note that the order of the arguments is significant: U (n 2 . The generalization to more dimensions is straightforward.2) of side a (reading from the right): U£ = U (n x . which is why the link carries an arrow. n 1 ) = U −1 (n 1 . n y .31). one finds (problem 16.1.26)) exp{ie[θ (n x+1 .25). n x+1 .32) It is straightforward to verify. n y .

Adding the exponentials again. n y + 1) − A x (n x . For small ‘a’ we may expand the exponential in (16.2.Figure 16.5). the exponentials contain no matrices and the accumulated phase factors cancel out. n y + 1) + iea A y (n x + 1. Next. let us see how to recover the Maxwell action.36) using the derivative definition of (16. and similarly for the three other links in U£ . verifying the gauge invariance.34) (16.36). n y ) − A y (n x . leading to £ where (1 − Re U£ ) → 1 2 £ e2 a 4 (Fx y )2 (16. To relate this to the continuum limit we must note that we sum over each such plaquette with only one definite orientation. n y ) = exp −iea 2 a A y (n x + 1. We also take the real part to remove the imaginary terms. n y ) + iea A x (n x .35) (16. n y ) +iea 2 a ∂ Ay ∂ Ax − = exp +iea 2 ∂x ∂y (16. A simple plaquette in two dimensions.37) Fx y = ∂ Ay ∂ Ax − ∂x ∂y as usual. so that the sum over plaquettes Copyright 2004 IOP Publishing Ltd . In this Abelian case. n y )} A x (n x . we can write U£ ≡ exp{−iea A y (n x . n y ) − iea A x (n x .

41) where q(ti ) = q i . as should be no surprise to the reader by now.38) (Note that in two dimensions ‘e’ has dimensions of mass. described by a Lagrangian L depending on one degree of freedom q(t). as required.4 Representation of quantum amplitudes So we have a naturally gauge-invariant ‘classical’ field theory defined on a lattice. we have not yet considered how we are going to turn this classical lattice theory into a quantum one.40) where the trace is over the SU(3) matrices.) However. In this section we give a brief introduction to this formalism. therefore. In section 5.2. We shall now provide some justification for this assertion. with a suitable continuum limit. is proportional to (with h = 1) ¯ tf exp i all paths q(t ) ti L(q(t).n 2 1 2 4 Fx y → e2 a 2 d x d y. (16.37) is Sg = 6 g2 £ Tr 1 − 1 Re U£ 3 (16. to pass from a state in which q = q i at t = ti to a state in which q = q f at time t = tf . an essential element in the lattice approach to quantum field theory.2. Thus £ (1 − Re U£ ) → 1 4 2 e2 a 4 Fx y n 1 . the analogue of the left-hand side of (16. ‘e A’ is replaced by ‘gt · A’. q(t)) dt ˙ (16.39) for the lattice action. more subtle than the naive replacements (16. the a → 0 limit of the quantum theory is. This is precisely what is provided by Feynman’s sum over paths formulation of quantum mechanics (Feynman and Hibbs 1965) and of quantum field theory and it is. we stated that in this approach the amplitude for a quantum system.) In four dimensions similar manipulations lead to the form SE = 1 e2 £ (1 − Re U£ ) → 1 4 2 d3 x dτ Fµν (16. (Actually. Copyright 2004 IOP Publishing Ltd . In the non-Abelian case.7. for SU(3).is equivalent to half of the entire sum. and q(tf ) = q f . 16. The fact that the calculations are mostly going to have to be done numerically seems at once to require a formulation that avoids non-commuting operators.5) because of renormalization issues. as we shall see in section 16. as noted earlier.

ˆ (16. with an associated complete set of eigenvectors |q such that q|q = q|q . Consider now the quantity ˆ H qtff |qtii H (16. Using (16.41).42)–(16. we can write to q f H qtff |qtii H = q f |e−i H (tf −ti ) |q i . Inserting a complete set of momentum eigenstates.49) 1 Remember that q| p is the q-space wavefunction of a state with definite momentum p and is.42) (16. (16. We shall demonstrate the connection explicitly for the special case of a free particle.47) can be represented as (16. we can evaluate (16. state vectors and observables are related in the Schr¨ dinger and Heisenberg pictures (see appendix I o of volume 1). Now define the Heisenberg picture state |qt H by |qt H = ei H(t −t0 ) |q . ˆ (16. (16.48) 2m For this case.We begin by recalling how.47) We want to understand how (16. the amplitude for the system described by H to go from q i at ti f at t . Let q be the canonical coordinate operator in the Schr¨ dinger ˆ o picture. a plane wave. we obtain1 q f |e−i H(tf −ti ) |q i = ˆ ∞ −∞ q f | p p|e−i H (tf−ti ) |q i d p ∞ −∞ ∞ −∞ ˆ 1 = 2π = 1 2π ei pq e−i p f 2 (t f −ti )/2m e−i pq d p i exp −i p2 (tf − ti ) − p(q f − q i ) 2m d p.44) the result qH (t)|qt ˆ H = q|qt H (16.46) ˆ which is. we are using the normalization of equation (E.45) which shows that |qt H is the (Heisenberg picture) state which at time t is an eigenstate of qH (t) with eigenvalue q. for which p2 ˆ ˆ H= . ˆ The corresponding Heisenberg operator qH (t)is defined by ˆ qH (t) = ei H (t −t0) qe−i H (t −t0 ) ˆ ˆ ˆ ˆ (16.47) directly as follows. in ordinary quantum mechanics.43) ˆ where H is the Hamiltonian and t0 is the (arbitrary) time at which the two pictures coincide. therefore.25) in volume 1. indeed.44) We then easily obtain from (16. Copyright 2004 IOP Publishing Ltd .44).

we complete the square via the steps p2 (tf − ti ) − p(q f − q i ) = 2m = = where p = p− tf − ti 2m tf − ti 2m tf − ti 2m p2 −   2mp(q f − q i ) tf − ti 2 m(q f − q i ) p−  tf − ti p − 2  m 2 (q f − q i )2  − (tf − ti )2  (16.55) We must now understand how the result (16.52) As it stands.51) We then shift the integration variable in (16. being rapidly oscillatory for large p . 2m −∞ (16.To evaluate the integral.2. it is at this point that the motivation for passing to ‘Euclidean’ spacetime arises. 2(τf − τi ) (16.56) in the free-particle case.52) is not well defined. We interpret the τ integral in terms of a discretization procedure. (16. We split the interval τf − τi Copyright 2004 IOP Publishing Ltd .55) can be represented in the form (16. Using the result ∞ −∞ dξ e−bξ = 2 π b (16.54) we finally obtain q |e f ˆ − H (τf −τi ) m |q = 2π(τf − τi ) i 1 2 exp − m(q f − q i )2 .41).53) and the integral is a simple convergent Gaussian. If we make the replacement t → −iτ . the integral in (16.52) becomes q f |e−i H (tf−ti ) |q i = ˆ m(q f − q i )2 1 exp i 2π 2(tf − ti ) d p exp − q f |e− H (τf −τi ) |q i = ˆ m(q f − q i )2 1 exp − 2π 2(τf − τi ) ∞ −∞ d p exp − (τf − τi ) p 2 2m (16. In Euclidean space.49) to p and obtain i(tf − ti ) p 2 . However.41) is exp paths − τf τi 1 m 2 dq dτ 2 dτ (16. (16. tf − ti ∞ (16.50) m(q f − q i )2 2(tf − ti ) m(q f − q i ) . similar to that introduced in section 16.

We follow the method of Feynman and Hibbs (1965. q N−1 at τ1 . .. is now interpreted as a multiple integral over all the intermediate positions q 1 . as shown in figure 16. q N−1 which paths can pass through at ‘times’ τ1 .58) are all of Gaussian form and since the integral of a Gaussian is again a Gaussian (cf the manipulations leading from (16. . .3. . . . 2 A( ) A( ) A( ) (16. Copyright 2004 IOP Publishing Ltd . depending on . . into N segments each of size . section 3..Figure 16. but without the ‘i’ in the exponents).3.57) and the ‘sum over paths’.. in going from q 0 ≡ q i at τi to q N ≡ q f at τf . τ2 . . . .52). . via the intermediate positions q 1 . . . τ2 . The τ -integral in (16. The integrals in (16. τ N−1 . exp − m j =1 dq N−1 (q j − q j −1)2 dq 1 dq 2 . . ..56) becomes the sum N m j =1 (q j − q j −1)2 2 (16. τ N−1 : 1 A( ) N . q 2 .49) to (16. .1). A ‘path’ from q 0 ≡ q i at τi to q N ≡ q f at τf .58) where A( ) is a normalizing factor. which is to be determined. we may perform all the integrations analytically. q 2 .

the pattern is now clear: after n − 1 steps we shall have an exponential factor exp[−m(q n − q i )2 /(2n )]. and all that remains is to choose A( ) in (16.60) Now the procedure may be repeated for the q 2 integral I2 ≡ exp − m 2 m (q − q i )2 − (q 3 − q 2 )2 dq 2 4 2 (16.66) exp − m 2 (q − q i )2 . (16.55) in (16.56) are concerned. The required A( ) is 2π (16. so that the result (16. 4 For the second (q 2 ) integration.59) This can be evaluated by completing the square.58) so as to produce the same normalization as (16.63) remembering that q N ≡ q f and that τf − τi = N . after N − 1 steps we shall have a factor exp[−m(q f − q i )2 /2(τf − τi )] (16.5) I = 2 4π 3m 1 2 exp −m 3 (q − q i )2 .5) I1 = π m 1 2 exp −m 2 (q − q i )2 .Consider the integral over q 1 : I1 ≡ exp − m [(q 2 − q 1 )2 + (q 1 − q i )2 ] dq 1 .60) becomes m I1 = 2π [ A( )]2 1 = m 2π2 π m 1 2 1 2 exp − m 2 (q − q i )2 4 (16.67) . Hence. For the first (q 1 ) integration. 2 (16. So we have recovered the correct exponential factor of (16.64) (16. 4 (16.55). 6 (16. shifting the integration variable and using (16.65) A( ) = m as we now verify.58) contains two factors of A−1 ( ). the accumulated constant factor is m 1 A( ) 2π2 Copyright 2004 IOP Publishing Ltd 1 2 4π 3m 1 2 = m 2π3 1 2 .54) to obtain (problem 16.61) which yields (problem 16. the formula (16.55).62) As far as the exponential factors in (16.

(16. chapter 5) and in Peskin and Schroeder (1995. τi ).71) where T is the usual time-ordering operator. established for the free-particle case. tb ) = dq a dq b q a q b q f |e− H (τf −τb ) |q b × q b |e− H (τb −τa ) |q a q a |e− H (τa −τi ) |q i . say.70) shows that if we take the limits τi → −∞.43) and (16.Proceeding in this way. our amplitude will represent the process in which the system at begins in its ground state | at τi → −∞.56) (in the sense → 0) is. with q = q i . then the state of lowest energy E 0 (the ground state) provides the dominant contribution. the accumulated constant is m 2π N 1 2 m = 2π(τf − τi ) 1 2 (16.58). ˆ ˆ ˆ ˆ ˆ ˆ (16.72) becomes G fi (ta . for tb > ta . Inserting a complete set of states and Euclideanizing. tb ) ≡ qtff |T {qH (ta )qH (tb )}|qtii (16. (16.55) and (16.55) with. The conventional notation for the path-integral amplitude is q f |e− H (τf −τi ) |q i = ˆ q(τ )e − τf τi L dτ (16. τf → ∞.55). The equivalence of (16. chapter 9). τf → ∞. Using (16.70) Equation (16. More general cases are discussed in Feynman and Hibbs (1965. with q = q How do we represent propagators in this formalism? Consider the expression (somewhat analogous to a field theory propagator) ˆ ˆ G fi (ta . N1 − 1 variables in the interval (τa . as ˆ ˆ G fi (ta . tb ) = q f |e−i H (tf−tb ) qe−i H (tb −ta ) qe−i H (ta −ti ) |q i .69) is interpreted in the sense of (16. one can convince oneself that after N − 1 steps. We now proceed to discuss further aspects of the path-integral formulation. τa ) Copyright 2004 IOP Publishing Ltd .44). Thus.73) Now.71) can be written.72) (16. and ends in | f.69) where the right-hand side of (16. (16. ˆ Consider the (Euclideanized) amplitude q f |e− H (τf −τi ) |q i and insert a complete ˆ set of energy eigenstates |n such that H |n = E n |n : q f |e− H (τf −τi ) |q i = n ˆ q f |n n|q i e−En (τf −τi ) . therefore.68) as in (16. each of the three matrix elements has a discretized representation of the form (16. N2 − 1 in (τb . in this limit.

74). the surviving part of (16.78) with τ → ∞.79) Copyright 2004 IOP Publishing Ltd . q(τn ) ≡ q b . If we now identify q(τa ) ≡ q a .73). Just as in (16. τa ) = φ b that a four-dimensional discretization of Euclidean spacetime is implied.77) is φ f (x)| e−E τ ˆ ˆ |T {φH (x a )φH (x b )}| e−E τ |φ i (x) (16. τb ). Then the order of τa and τb inside the interval (τi . ∂τ φ) d4 xE (16. Two of these factors can be associated with the dq a dq b integration in (16.75) is equal to ˆ ˆ φ f (x)|e− Hτf T {φH (x a )φH (x b )}e H τi |φ i (x) . as discussed after (16.74). in this limit.77) Taking the limits τi → −∞. τf ) = φ f (x). or q(τa ) and q(τb ) in (16.58). ˆ ˆ τf τi ÄE (φ.71) of the time-ordered product. In this way. Suppose alternatively that τa > τb .77) but without the additional fields φ(x a ) and φ(x b ).74) In obtaining (16. a (x) and φ(x.73).7). Each such representation carries one ‘surplus’ factor of [ A( )]−1 . τi ) = φ(x. 4 E d xE] Ä (16.70). which in this case is the (interacting) vacuum state | . making an overall factor of [ A( )]−3 . say. τf ) is simply reversed but since q a and q b in (16.75) (16. are ordinary (commuting) numbers. ∇φ.71)-(16. τ ) = φ b (x). we have to understand φ(x.and N3 − 1 in (τf . The generalizations of these results to the field theory case are intuitively clear.74) is unaltered and it does actually represent the matrix element (16. it follows that (16. so that we have a total of N1 + N2 + N3 − 1 properly normalized integrations and one ‘surplus’ factor [ A( )]−1 as in (16. we expect the analogue of (16.75). (16. the fields being Fourier-analysed by four-dimensional generalizations of expressions such as (16.76) and the boundary conditions are given by φ(x.73) is simply q(τ )q(τa )q(τb )e − τf τi L dτ .4)) φ(x) φ(x a )φ(x b ) exp − where d4 x E = d3 x dτ φ i (x). In (16. we obtain the formula for the field theory propagator in fourdimensional Euclidean space: ˆ ˆ |T {φH (x a )φH (x b )}| = lim φφ(x a )φ(x b ) exp[− φ exp[− τ −τ τ →∞ Ä τ 4 −τ E d x E ] . (16. we took the case τb > τa . (16. Thus. in the case of a single scalar field φ(x).74) to be (cf (16. the formula (16.74). τf → ∞ will project out the configuration of lowest energy. The exponential and overlap factors can be removed by dividing by the same quantity as (16. For example.

1) is an element of U(1). Such ‘classical’ anti-commuting variables are called Grassmann variables. since the gauge-field propagator will not exist unless we ‘fix the gauge’ by imposing some constraint. the exponential in (16. In the case of the unitary groups. in which the integral is carried out over all gauge fields Aµ (x) (in the U(1) case. Peskin and Schroeder 1995) and one finds exactly the same ‘Feynman rules’ as in the canonical (operator) approach. and are briefly described in appendix O. we can easily imagine a formula similar to (16. hence.79) for the gauge-field propagator. Perturbation theory can be developed in this formalism also. In the case of gauge theories. in the continuum limit. indeed. with a finite number of lattice points. E E E E for example λφ 4 . for vevs of gauge-invariant quantities. For our purposes. as mentioned in section 13.79) may be expressed as vevs of products of free fields. In the discrete case. this was the route followed by Faddeev and Popov (1967) in first obtaining the Feynman rules for non-Abelian gauge theories. where Ä0 describes a free scalar field and Äint is an interaction. The integration measure for the link variables can be chosen so as to be gauge invariant and. There remains one more conceptual problem to be addressed in this approach—namely. where the ‘Hermitean matrix’ can be parametrized in some convenient way—for example. for example). Suppose ÄE = Ä0 + Äint . the formalism provides well-defined expressions. must nevertheless anti-commute with each other.80) and both numerator and denominator of (16. the variables in the parametrization of U vary over some bounded domain (they are essentially ‘angle-type’ variables. as in the simple U(1) case).35)) ∼ exp (i Hermitean matrix). Compact techniques exist for analysing this formulation of perturbation theory (Ryder 1996. Then.Vacuum expectation values (vevs) of time-ordered products of more fields will simply have more factors of φ on both sides. n 1 + 1) of (16. In all these cases. Further details are contained in Ryder (1996) and in Peskin and Schroeder (1995). though not quantum field operators. the integral over the link variables is well-defined without gauge-fixing. the appropriate integration variables are the link variables U(li ) where li is the i th link. assuming λ is small. But we already know from chapter 7 (or from chapter 13 in the non-Abelian case) that. They are elements of the relevant gauge group—for example U(n 1 . Such constraints can be imposed on the corresponding path integral and. how are we to deal with fermions? It seems that we must introduce new variables which. independently of perturbation theory. and so.3.5.31) for SU(2).79) can be expressed as exp − d4 x E (Ä0 + Äint ) = E E exp − d 4 x E Ä0 E 1−λ d4 x E φ 4 + · · · (16. the important point is that the fermion Lagrangian is bilinear in Copyright 2004 IOP Publishing Ltd . such as the Lorentz gauge condition. provided the action is gauge invariant. such elements typically have the form (cf (12. as in (12. we shall not be able to construct a well-defined perturbation theory in this way.3.

Copyright 2004 IOP Publishing Ltd . the expectation value of any operator Ç(U ) is just Ç(U ) = U Ç(U ) exp[− Sg (U )] U exp[− Sg (U )] (16. moreover. the partition function Z defined by Z= configurations exp − H kB T (16. computers are just about reaching the stage of being able to calculate such a vast determinant numerically. The problem is that M is a very large matrix indeed. H is the classical Hamiltonian 2 See appendix O. The connection is made via the fundamental quantity of equilibrium statistical mechanics. We introduce this connection now. by way of a preliminary to the discussion of renormalization in the following section. but hitherto most calculations have been done in the quenched approximation.81) / where M is a matrix representing the Dirac operator i D − m q in its discretized and Euclideanized form. 16. This means that in a typical fermionic amplitude of the form (cf the denominator of (16.79)) Zψ = ¯ ψ ψ exp[− Sψ ] (16. the determinant of this matrix. the fermionic action having the form Sψ = ¯ d4 x E ψ M(U )ψ (16. In the quenched approximation.82) one has essentially an integral of Gaussian type (albeit with Grassmann variables). This is equivalent to the neglect of closed fermion loops in a Feynman graph approach.5 Connection with statistical mechanics Not the least advantage of the path integral formulation of quantum field theory (especially in its lattice form) is that it enables a highly suggestive connection to be set up between quantum field theory and statistical mechanics.2 The result is simply det[ M(U )]. At the time of writing.40). setting the determinant equal to a constant independent of the link variables U .83) where Sg (U ) is the gauge action (16.84) which is simply the ‘sum over states’ (or configurations) of the relevant degrees of freedom. if we want anything like a reasonably sized lattice (say 20 lattice spacings in each of the four dimensions).the (Grassmann) fermion fields ψ . which can actually be performed analytically. the gauge field degrees of freedom must be itemized (via the link variables U ) at each site. with the Boltzmann weighting factor.

86) φ exp(−SE ) (16.89) G(x) ∝ |x| ξ(T ) Copyright 2004 IOP Publishing Ltd . not the four of our Euclideanized spacetime.40).40) 6/g 2 would play the rˆ le of 1/kB T .84). such changes of ‘phase’ can occur. the gauge coupling is seen to be analogous to an inverse temperature. In an Ising model (see the following section). which for a spin system may be defined via the spin–spin correlation function G(x) = s(x)s(0) = all s(x) s(x)s(0)e−H/kB T (16.evaluated for each configuration. the effective ‘Hamiltonian’ is not quite the same—though we may note that Wilson (1971b) argued that in the case of a φ 4 interaction the parameters can be chosen so as to make the values φ = ±1 the most heavily weighted in SE .87) is inserted into (16. For large |x|. in (16. Nevertheless.79). the system variables taking the values ±1. The factor exp(−SE ) is then a sensible statistical weight for the fluctuations in φ. Consider. (16.39) or (16. indeed.84). as coupling constants (or other parameters) vary—and. One insight we may draw from this connection is that. in four ‘spatial’ dimensions. of course. of course. One is led to wonder whether something o like transitions between different ‘phases’ exist. the Hamiltonian has the form H = −J n sn sn+1 (16.88) where we are once more reverting to a continuous x variable. deal in three spatial dimensions. and Z φ may be interpreted as the partition function for a system described by the field degree of freedom φ but. for comparison. by comparison with (16. In statistical mechanics. the denominator in (16. Statistical mechanics does. A second point is somewhat related to this.85) in the case of a single scalar field with mass m and self-interaction λφ 4 . it is remarkable that quantum field theory in three spatial dimensions appears to have such a close relationship to equilibrium statistical mechanics in four spatial dimensions. When (16. The Euclideanized Lagrangian density ÄE is like an energy density: it is bounded from below and increases when the field has large magnitude or has large gradients in τ or x. The parallel becomes perhaps even stronger when we discretize spacetime. an important quantity is the correlation length ξ .6). in the case of pure gauge actions (16. this takes the form −|x| 1 exp . we arrive at something very reminiscent of the φ(n 1 )φ(n 1 + 1) term in (16. For example. Naturally.87) where J is a constant and the sum is over lattice sites n. namely Zφ = where SE = d4 x E ÄE = d4 x E { 1 (∂τ φ)2 + 1 (∇φ)2 + 1 m 2 φ 2 + λφ 4 } 2 2 2 (16.

it is clear that (16. precisely the way in which. This shows rather clearly why it is not going to be easy: correlations over all scales will have to be included.79) for example. at the critical point ξ(Tc ) ∼ L . thanks to Wilson’s work. it provides access to a more intuitive way of understanding renormalization theory. the methods of chapter 15 were rather formal and the reader may well feel the need of a more physical picture of what is going on. which are controlled (in a renormalizable theory) by the procedure of renormalization. as was indeed always our assumption. so the regime ξ a is equivalent to m . Another important advantage of this formalism is. where L is the size of the system. the idea suggests Copyright 2004 IOP Publishing Ltd . as in (16. In fact. fluctuations over short distances of order −1 generally lead to divergences in the limit → ∞. Thus studying a quantum field theory this way is analogous to studying a fourdimensional statistical system near a critical point.79). using the ‘lattice + path integral’ approach. all other quantities become finite. Nevertheless. 16. so as to illuminate the formal treatment of the previous chapter. Comparing (16. as we have seen.The Fourier transform of this (in the continuum limit) is ˜ G( k2 ) ∝ ( k 2 + ξ −2 (T ))−1 (16. In the quantum field theory.90) then shows that ξ −1 (T ) is playing the role of a mass term m . to affect a renormalizable theory only through the values of some of its parameters and. In the ‘lattice + path integral’ approach to quantum field theory. Such a picture was provided by Wilson (1971a) (see also Wilson and Kogut 1974).3. even as → ∞. in fact.6. However. correlations exist over very large scales ξ compared to the inter-atomic spacing a .1 Introduction In the continuum formulation which we have used elsewhere in this book. in fact. therefore. this is by no means all there is to renormalization theory: we have seen the power of ‘renormalization group’ ideas in making testable predictions for QCD. we may regard a −1 as playing a role analogous to a momentum cut-off .88) with (16. it should be possible to examine specifically how the ‘short distance’ or ‘high momentum’ degrees of freedom affect the result.90) as we learned in section 2. From this point of view. if these parameters are taken from experiment. the degrees of freedom involved are the values of the field(s) at each lattice site. Now. we are naturally led to the consideration of renormalization in the lattice formulation. Such divergent fluctuations turn out. indeed.6 Renormalization and the renormalization group on the lattice 16. At this point. is quite surprising.88) is proportional to the propagator (or Green function) for the field s( x): (16. as indicated earlier. The aim of this section is to give a brief introduction to Wilson’s ideas. Quantum amplitudes are formed by integrating suitable quantities over all values of these degrees of freedom. near a critical point for a statistical system. This latter assertion is not easy to prove and.

Thus. we shall be considering the system as described by parameters (including masses and couplings) appropriate to that scale. Having done this once.87) and partition function N−1 Z= {sn } exp K n=0 sn sn+1 (16. which is necessarily very large in the case ξ a. . In (16. or separated by only a lattice site or two in coordinate space. .91) all the sn variables take the values ±1 and the ‘sum over {sn }’ means that all possible configurations of the N variables s0 .4 shows a portion of the one-dimensional lattice with the spins on the sites. itself that we might be able to perform explicitly the integration (or summation) over those degrees of freedom located near the cut-off in momentum space. in general. We now consider some simple examples which we hope will illustrate these points. And third. to introduce new kinds of effective interactions into the theory. . Figure 16. we can imagine doing it again—and indeed iterating the process. 16.6.4. s N−1 are to be included. The spin sn is located at the lattice site na and we shall (implicitly) be assuming the periodic boundary condition sn = s N+n . we may perhaps anticipate that the result of this ‘integrating out’ will be not only to render the parameters of the theory scale-dependent but also. at any given stage of the integration. which is of course strongly reminiscent of RGE ideas.91) where K = J/(kB T ) > 0.Figure 16. s1 . for this portion we Copyright 2004 IOP Publishing Ltd . Thus. the process of ‘integrating out’ short-distance degrees of freedom will obviously reduce the number of effective degrees of freedom. First. . it must be a step in the right direction. Second. If we can do this. as previously envisaged. until eventually we arrive at some kind of ‘effective theory’ describing physics in terms of ‘long-distance’ degrees of freedom. A portion of the one-dimensional lattice of spins in the Ising model. There are several aspects of such a programme which invite comment. each site being separated by the lattice constant a.2 The one-dimensional Ising model Consider first a simple one-dimensional Ising model with Hamiltonian (16. s2 . this sketch of the ‘integrating out’ procedure suggests that. the result may be compared with the theory as originally formulated to see how this ‘integration over short-distance degrees of freedom’ affects the physical predictions of the theory.

s3 .s3 .s4 (16. expanded as a power series: exp(K s0 s1 ) = 1 + K s0 s1 + K3 K2 + (s s1 ) + · · · 2! 3! 0 (16. The new (primed) spin variables are situated twice as far apart as the original (unprimed) ones. these sums are easy to do.96) . Consider the quantity exp(K s0 s1 ).s2 . s N−1 .Figure 16. providing a more ‘coarse grained’ description of the system.5. s1 and s2 . we shall end up (for this portion) with an expression involving the ‘effective’ spin variables s0 . s N−1 . Summing over s1 and s3 corresponds to ‘integrating out’ two short-distance degrees of freedom as discussed earlier. It follows that exp(K s0 s1 ) = cosh K (1 + s0 s1 tanh K ) and similarly exp(K s1 s1 ) = cosh K (1 + s1 s1 tanh K ). are evaluating exp[K (s N−1 s0 + s0 s1 + s1 s2 + s2 s3 + s3 s4 )] .s0 .95) (16.s2 (16. therefore.93). situated twice as far apart as the original ones and. There are many ways we could choose to describe the sn but here we shall only consider a very simple one (Kadanoff 1977) in which each sn is simply identified with the sn at the corresponding site (see figure 16.s1 . then.4. A ‘coarsening’ transformation applied to the lattice portion shown in figure 16. In fact.s1 . with lattice spacing 2a and corresponding new spin variables sn .94) where we have used (s0 s1 )2 = 1.s1 .92) becomes exp[K (s N−1 s0 + s0 s1 + s1 s1 + s1 s3 + s3 s2 )]. (16.5). Copyright 2004 IOP Publishing Ltd (16.93) If we can now perform the sums over s1 and s3 in (16.s0 . For the portion of the lattice under consideration.92) Now suppose we want to describe the system in terms of a ‘coarser’ lattice.

103) tanh K (n) = (tanh K )n .100). and thence to K . (16.104) .99) Exactly the same steps can be followed through for the sum on s3 in (16. Of particular interest is a point (or points) K ∗ such that tanh K ∗ = tanh2 K ∗ . Copyright 2004 IOP Publishing Ltd (16.101) (16. (16. In fact. . by tanh K = (tanh K )2 = (tanh K )4 .98) contains a new ‘nearest-neighbour’ interaction. the effective lattice constant is 2n a and the effective coupling is given by (16.Thus the sum over s1 is cosh2 K (1 + s0 s1 tanh K + s1 s1 tanh K + s0 s1 tanh2 K ). (16. The successive values K . just like the original one in (16. apart from the accumulated spin-independent part. we can write (16. the new partition function.98) in the standard form exp[g1 (K ) + K s0 s1 ] and then use (16. s1 =±1 (16. It is clear that we could apply the same procedure to the new Hamiltonian.93)—and indeed for all the sums over the ‘integrated out’ spins. the lattice spacing has doubled and the coupling K has been renormalized to K .100) (16.98) Remarkably.91) but with an altered coupling (and a different spinindependent piece). defined on a lattice of size 2a.100) is an example of a renormalization transformation: the number of degrees of freedom has been halved. . Equation (16. the terms linear in s1 vanish after summing and the s1 sum becomes just 2 cosh2 K (1 + s0 s1 tanh2 K ).95) to set tanh K = tanh2 K and identify g1 (K ) = ln 2 cosh2 K cosh K .102) This is equivalent to iterating the renormalization transformation. of the coupling under these iterations can be regarded as a ‘flow’ in the (one-dimensional) space of K -values: a renormalization flow. K . introducing a coupling K which is related to K . has the same form as the old one but with a new coupling K related to the old one K by (16. after n iterations. . (16. s0 s1 . The upshot is that.97) Clearly.

which goes to infinity in the thermodynamic limit. This is called a fixed point of the renormalization tranformation.87). 2 −2K e . At such a point in K -space.108) 2 or. changing the scale by a factor of 2 (or 2n for that matter) will make no difference. (16.103) as in figure 16.107) confirms that there is no finite temperature T at which ξ → ∞ but as T → 0 we do have ξ → ∞ and the system is ordered. ‘Renormalization flow’: the arrows show the direction of flow of the coupling K as the lattice constant is increased. the only fixed points are K ∗ = ∞ and K ∗ = 0.90). Thus there is no critical point at a non-zero T ∗ and. For large K . we may calculate ξ(T ) exactly (problem 16. we see that K = K ∗ when the temperature is ‘tuned’ to the value T = T ∗ = J/(kB K ∗ ). hence.6) and find that ξ(T ) = −a .107) Equation (16.104) as Kn 1 − 2e−2 K 1 2 (16.6. which means that the system must be. The starred values are fixed points.105) (16. However. In the present case. Remembering that K = J/(kB T ).107) we find a 2K ξ(T ) e for T → 0 (16. corresponding to the onset of ordering. Note that the flow is away from the quasi-fixed point at K ∗ = ∞(T = 0) and towards the (non-interacting) fixed point at K ∗ = 0. so that this is ‘almost’ a fixed point.6. no transition to an ordered phase.89) and (16. ln tanh K (a) (16. At a critical point. we may use tanh K to write (16. in terms of the equivalent mass parameter m(T ) = ξ −1 (T ) introduced after (16. ξ ∼ L (the system size).Figure 16. More precisely.90).106) K (n) = K − ln n which shows that changes only very slowly (logarithmically) under iterations when in the vicinity of a very large value of K . ordered. Such a T ∗ would be the temperature of a critical point for the thermodynamics of the system. in some sense. In the present model. we may describe the behaviour as T → 0 as ‘quasi-critical’. Another way of looking for a critical (or fixed) point would be to calculate the correlation length ξ(T ) introduced in (16. We may represent the flow of K under the renormalization transformation (16. from (16. with the Hamiltonian (16.109) m(T ) a Copyright 2004 IOP Publishing Ltd .

A simple such example given by Kadanoff (1977) is the transformation K = 1 (2K )2 2 for a doubling of the effective lattice size.3 Further developments and some connections with particle physics A renormalization transformation which has a fixed point at a finite (neither zero nor infinite) value of the coupling is clearly of greater interest. As we iterate away from Copyright 2004 IOP Publishing Ltd .103) on ξ . In particular. n (16. under successive iterations.110) ξn (T ) = − ln tanh K (n) From (16. (16. since this will correspond to a critical point at a finite temperature. K ∗ = ∞ and the finite point K ∗ = 1 . 2 The striking feature of this flow is that the motion is always away from the finite fixed point. The renormalization transformation (16. In this sense.111) confirms what we might have expected: the correlation length is measured in units of the only available length unit.7.6.113) involves fermions in one dimension but the details are irrelevant to our purpose here. ξ → 0 as n → ∞.113) has three fixed points: K ∗ = 0. The renormalization flow is shown in figure 16. and when this increases to na after n iterations. 16. Let us denote by ξn (T ) the effective correlation length after n iterations.7. This may be understood by recalling that at the fixed point (which is a critical point for the statistical system) the correlation length ξ must be infinite (as L → ∞). It is interesting to consider the effect of the renormalization transformation (16. a. The expression (16.103). ξ must decrease by n −1 so as to maintain the same physical distance.113) (16. or K (n) = 1 (2K )n 2 (16. so that a . The model leading to (16. The renormalization flow for the transformation (16.111) Equation (16. it then easily follows that ξn (T ) = 1 ξ(T ).Figure 16. T = 0 is a kind of ‘asymptotically reachable’ fixed point.113).109) becomes very small at small T but never vanishes at a finite value of T .112) for n such iterations.

analogous to the RGE’s for α and αs considered in chapter 15. ξ must decrease according to (16. In this case.117) which is sketched in figure 16. the flow being away 2 from it as f increases.7 and the related discussion. say by a factor of f . however. as given in the previous chapter. (16.113) then becomes K ( f a) = 1 (2 K (a)) f . ξ is given by Kadanoff (1977) as ξ= a |ln 2 K | (16. The β -function of (16. The zero of β is indeed at the fixed (critical) point K = 1 and this is an infrared unstable fixed point. 2 Differentiating (16.114) which indeed goes to infinity at K = 1 . this point. In the present model. we need to consider a continuous change of scale. First.111). For this model. Note.116) a renormalization group equation. describing the ‘running’ of K ( f a) with the scale f . 2 Let us now begin to think about how all this may relate to the treatment of the renormalization group in particle physics. figure 15. that in the Copyright 2004 IOP Publishing Ltd . we find that f d K ( f a) = K ( f a) ln [2 K ( f a)] . in particular. the transformation (16.115) We may reasonably call (16. df (16.8. and therefore we must leave the fixed (or critical) point.117)—the arrows indicate increasing f .115) with respect to f . The foregoing is exactly analogous to the discussion in section 15.Figure 16.8. the β -function is β(K ) = K ln(2 K ).4—see.116) (16.

and in the continuum limit. not momentum space. Enormous progress was made in the theory of critical phenomena when the powerful methods of quantum field theory were applied to calculate critical exponents (see. say. it is clear that scaling a by f is the same as scaling k by f −1 = t . Thus. the mass parameter (cf (16. at a critical point. of course.4—it is no longer true in an interacting theory. Generally. Peskin and Schroeder (1995. η is a critical exponent. Hamiltonians will consist of several terms and the behaviour of all their coefficients will need to be considered under a renormalization transformation. k (16. and Binney et al (1992)). one of a number of such quantities characterizing the critical behaviour of a system.118) generally becomes modified to 1 ˜ (16. characterized by their behaviour in the vicinity of a fixed point: Copyright 2004 IOP Publishing Ltd . we see that η/2 is precisely the anomalous dimension of the field s(x). The general analysis of renormalization flow in multi-dimensional coupling space was given by Wegner (1972). We have emphasized that. at a critical point. for example. η will depend on the coupling constant η( K ): at a non-trivial fixed point. G(x) ∝ 1 |x|1+η (16. In the interacting case. and accounts for the fact that K = 1 is an infrared 2 ∗ unstable fixed point in figure 16. while αs is an infrared stable fixed point in figure 15.7(a). the correlation length ξ → ∞ or. Allowing for the change in sign. This will produce a change in sign in d K /dt relative to d K /d f . the coefficients show one of three types of behaviour under renormalization transformations such that a → f a. but—as we learned for the fermion propagator in section 15. chapter 13). equivalently. the Fourier transform of the spin–spin correlation function should behave as 1 ˜ G(k2 ) ∝ 2 .90)) m = ξ −1 → 0.118) This is indeed the k2 -dependence of the propagator of a free.119) with (15.present case we are considering rescalings in position space.65). In general. In this case. equivalently. Comparing (16. In the statistical mechanics case. the spin–spin correlation function exhibits scaling under the transformation x = f x but it is not free-field scaling. Since momenta are measured in units of a −1 .7(b). we have only considered simple models with just one ‘coupling constant’. In our discussion so far. massless scalar particle.119) G(k2 ) ∝ η 2 )1− 2 (k or. η will be evaluated at the fixed point value K ∗ . η(K ∗ ).8.4—we have an example of scaling with anomalous dimension. so that diagrams of renormalization flow were onedimensional.120) in three spatial dimensions. In simple terms. figure 16. so—just as in section 15.8 is quite analogous to figure 15. (16.

(16. (16. This explains the property of universality observed in the physics of phase transitions. (i) relevant. A simple mathematical analogue may illustrate the point. respectively. therefore. We write Zxy = ∞ −∞ ∞ dx exp[−(x 2 + λx 4 )] 2 4 ∞ −∞ dy exp[−(y 2 + λx 2 y 2 )] 1 2 π = dx exp[−(x + λx )] 1 + λx 2 −∞ . (16. so that the system moves away from the fixed point (as in the single-coupling examples considered earlier). (ii) irrelevant and (iii) marginal couplings—the terminology is also frequently applied to the operators in the Hamiltonians themselves.122) may be expanded as Zxy π2 1 ∞ −∞ dx exp[−(x 2 + λx 4 )][1 − 1 λx 2 + 3 λ2 x 4 − 2 8 5 3 6 16 λ x + · · ·].122) Assuming that λ is small and may be treated perturbatively. so the system moves towards the fixed point. in fact. the action being similar to that of two scalar fields with quartic couplings. be independent of the number and type of all irrelevant couplings and will be determined by the relatively few (in general) marginal and relevant couplings. whereby many apparently quite different physical systems are described (in the vicinity of their critical points) by the same critical exponents. all systems which flow close to the fixed point will display the same critical exponents determined by the dynamics of these few couplings. (ii) the difference decreases as f increases. Additional terms in the Hamiltonian are. and (iii) there is no change in the value of the coupling as f changes. Suppose we want to ‘integrate out’ the variable y.121) which may be regarded as the ‘partition function’ for a system with two variables x and y.(i) the difference from the fixed point value grows as f increases. The corresponding coefficients are called. Thus. Consider the expression Zxy = ∞ −∞ ∞ dx −∞ dy exp[−(x 2 + y 2 + λx 4 + λx 2 y 2 )] (16. The intuitive meaning of ‘irrelevant’ is clear enough: the system will head towards a fixed point as f → ∞ whatever the initial values of the irrelevant couplings. The critical behaviour of the system will.124) . generally introduced following a renormalization transformation.123) The terms in the series expansion may be regarded as arising from an expansion of the exponential exp[− 1 λx 2 − 1 λ2 x 4 + 2 4 Copyright 2004 IOP Publishing Ltd 3 3 6 16 λ x − · · ·] (16.

and (c) the generation of a 4 ‘new interaction’ of sixth order in the degree of freedom x. except the ‘kinetic’ piece. summations over momenta appearing in loops do not. so that their effect becomes negligible as we iterate to scales much larger than a. The additional interactions compensate for this change.8 ( in four dimensions φ has mass dimension unity. At first sight. Thus.125) 2 4 The result (16. Thus. that we have integrated out. after one iteration a → f a. A simple way to do this is to rescale distances by x (16. We will be left with a functional integral of the form (16. this is the ‘free-field’ fixed point. so that A must have mass dimension −2). we may expect that renormalization transformations associated with a → f a. we require the kinetic action to be unchanged: d4 x E (∂µ φ)2 = d4 x E (∂µ φ )2 Copyright 2004 IOP Publishing Ltd . in (16. we must rewrite it so that it has the same general form as the original Z φ of (16.85) but with φ(x) restricted to |x| > f a. Imagine. so that to this order. (16. if we start with a typical ‘φ 4 ’ scalar theory as given by (16. It is.125) may be interpreted as (a) a renormalization of the ‘mass’ term x 2 so that its coefficient changes from 1 to 1 + 1 λ. therefore.85). In the quantum field case.86). We may think of this new interaction (in more realistic quantum field cases) as supplying the effective interaction between the ‘x-fields’ that was previously mediated by the ‘y-fields’. In order λ3 . and iterations thereof. We now define the fixed point of the renormalization transformation to be that in which all the terms in the action are zero. (16. We consider continuous changes of scale characterized by a factor f . but an interaction such as Aφ 6 is not renormalizable according to the criterion given in section 11. and we discuss a theory with only a single scalar field φ for simplicity.126) x = f so that the functional integral is now over φ(x ) with |x | > a.123) may be written as Zxy π2 1 ∞ −∞ 3 dx exp{−[x 2(1 + 1 λ)+ x 4 (λ− 1 λ2 )+ 16 λ3 x 6 +· · ·]}. but only up to the lower cut-off π/ f a. this may seem disturbing: after all.85). those components of φ(x) with a < |x| < f a. as in (16. however. will in general lead to an effective theory involving all possible couplings allowed by whatever symmetries are assumed to be relevant. and with additional interaction terms in the action. essential to remember that in this ‘Wilsonian’ approach to renormalization.85). the original theory (in four dimensions) is a renormalizable one. we shall now see how the coefficients of non-renormalizable interactions correspond precisely to irrelevant couplings in Wilson’s approach. run up to the original cut-off value π/a. In order to interpret the result in Wilson’s terms. (b) a renormalization of the 2 x 4 term so that the coupling λ changes to λ − 1 λ2 . we shall expect to generate all possible couplings involving φ and its derivatives.

in this picture. which is a rapid growth with f . The mass dimension of φ is unity.127) from which it follows that φ = f φ. we have to include higher-order loop corrections. Finally. Thus.129) showing that m 2 = m 2 f 2 and the ‘coupling’ m 2 is relevant. But such an interaction is precisely a non-renormalizable one (in four dimensions).g. the non-renormalizable terms disappear and we are left with an effective renormalizable theory.8. in Wilson’s sense. for instance). We saw that such terms were suppressed at low energies by factors of E/ .128) shows that the ‘new’ A is related to the old one by A = A/ f 2 and in particular that. since it grows with f 2 . the mass term m 2 φ 2 behaves differently: m2 d4 x E φ 2 = m 2 f 2 d4 x E φ 2 (16. 6 (16. As we have seen.128) (16. Consider now a term of the form Aφ 6 : A d4 x E φ 6 = A f2 d4 x E φ . the λφ 4 interaction transforms as λ d4 x E φ 4 = λ d4 x E φ 4 (16.8. might be the Planck mass). Such a term has positive mass dimension and corresponds to a ‘superrenormalizable’ interaction. The reader may verify the generality of this result for any interaction with p powers of φ and q derivatives of φ. To find whether such couplings increase or decrease with f . according to the criterion of section 11. tending to zero as we reach large scales. The coupling is marginal. where is the cut-off scale beyond which the theory is supposed to fail on physical grounds (e. The foregoing analysis in terms of the suppression of non-renormalizable interactions by powers of f −1 parallels precisely the similar one in section 11. One further word should be said about terms such as ‘m 2 φ 2 ’ (which arise in the Higgs sector of the Standard Model. then the ‘initial’ value of m will have to be very finely ‘tuned’ in order to end up with a mass of order 1 TeV. down to much lower energy scales.130) and so λ = λ. The result is that as we renormalize. m 2 scales by m 2 = m 2 f 2 . However. which may correspond (though not necessarily) to a renormalizable interaction. If we imagine starting at a very high scale. This is the field theory analogue of ‘universality’. it seems unnatural to have scalar particles with Copyright 2004 IOP Publishing Ltd . and hence that of A must be −2 so that the action is dimensionless: couplings with negative mass dimensions correspond to non-renormalizable interactions. as f increases. A decreases and is therefore an irrelevant coupling. such as 1015 TeV and flowing down to 1 TeV.= 1 4 d x E (∂µ φ )2 f2 (16.

so that L ∼ 0. we need a R ∼ 1/m L = Na (16. The lattice must be large enough so that the spatial dimension R of the object we wish to describe—the Compton wavelength of a quark or the size of a hadron—fits comfortably inside it. computing demands are formidable. as a result of renormalization effects. so as to ensure that M remains finite.masses much less than the physical cut-off scale. as → ∞. we shall now see that its value actually controls the physical size of the lattice spacing. we shall also have to change gL suitably. (16.131) so that N must be large. We have strayed considerably from the initial purpose of this chapter— but we hope that the reader will agree that the extra insight gained into the physical meaning of renormalization has made the detour worthwhile. of course. the bare parameters of the theory depend on the momentum cut-off in such a way that. with a brief survey of some of the results obtained numerically. essentially. must be related to the only quantity with mass dimension. in approaching the continuum limit a → 0.4).2. Actual calculations are done by evaluating quantities such as (16.79) by ‘Monte Carlo’ methods (similar to the method which can be employed to evaluate multi-dimensional integrals). Quenched calculations now use four-dimensional cubes with N = 64. We also need R a or else the granularity of the lattice resolution will become apparent.1. finite values are obtained for the corresponding physical parameters (see the last paragraph of section 10. otherwise the result will be subject to ‘finite-size effects’ as the hypercube side length L is varied. quite analogous to saying that. for example). Despite the fact that gL is dimensionless. ideally. where gL is the bare lattice gauge coupling (we assume for simplicity that the quarks are massless). as indicated earlier.7 Numerical calculations Even in the quenched approximation (section 16. in a renormalizable theory. Ignoring any statistical inaccuracy. Thus. the extent to which the lattice ‘a’ can really be taken to be very small is severely limited by the computational resources available—that is.1. for example. by a relation of the form 1 M = f (gL ). a −1 . by the number of mesh points N. however. the masses m which could be simulated would be limited (from (16.132) a Thus. the results will depend on the parameters gL and N. Further. In practice.64 fm. We shall return to this problem in section 22.10. The computed mass of a hadron M. we expect the mass m (which is of order R −1 ) to be very much less than a −1 . We now finally return to lattice QCD. a.01 fm.131)) by Copyright 2004 IOP Publishing Ltd . This is. If we were to think of an a of order 0. 16. say.

136) where the sign is the opposite of (15. which is a severe restriction against light quarks or hadrons.140)). as usual.1 fm with L ∼ 6. which would allow the pion to be reached but note that. in practice. it is clear that the lattice cut-off is. in this case. A more reasonable value might be a ∼ 0. a (16. However. and lead to an interesting prediction for M (equations (16. and it is defined by L ≡ lim gL →0 2π 1 exp − 2 a bgL . really read as M= 1 f (gL (a)) .117)). a dimensionful constant of integration (cf (15.m 300 MeV. we find that 0= so that a 1 1 d f dgL(a) dM = − 2 f (gL (a)) + da a dgL da a dgL(a) da df = f (gL (a)) . calculable in perturbation theory.135) Meanwhile. The integration of (16. At all events. we know that the behaviour of gL(a) at small scales is. of course.136) requires.139)–(16.133) As a → 0. in fact. L can be regarded as a lattice equivalent of the continuum QCD . which seems low and certainly rules out simulations with very ¯ massive quarks. as we expect from asymptotic freedom. (16. in view of the left-hand inequality in (16.4 fm. thanks to the asymptotic freedom of QCD.131). the equivalent momentum cut-off is h /a ∼ 6 GeV. (16.137) Equation (16. be included.133). Copyright 2004 IOP Publishing Ltd .134) (16.132) should. M should be finite and independent of a.138) Equation (16. Equation (16.54)): 2 gL (a) 1 = 4π b ln(1/a 2 2) L . therefore. a long way from the ‘in principle’ → ∞ situation. dgL (16.137) shows that gL(a) tends logarithmically to zero as a → 0. Differentiating (16. up to a constant. Higher-order corrections can.138) may also be read as showing that the lattice spacing a must go exponentially to zero as gL tends to zero. the scale dependence of gL is given (to one-loop order) by a b 3 dgL(a) = g (a) da 4π L (16.48) since a ∼ µ−1 is the relevant scale parameter here (compare the comments after equation (16. This will allow us to determine the form of f (gL ).

should vary as a function of gL . The horiztonal lines are the experimental Copyright 2004 IOP Publishing Ltd . τf → +∞ in (16. we show in figure 16.74 MeV in this case. How are masses calculated on the lattice? The principle is very similar to the way in which the ground state was selected out as τi → −∞. in (16.141). (16. The ρ and π masses are missing since they were used to fix. the term with the minimum value of E n .In a similar way. and open circles correspond to fitting the φ mass.140) Equation (16.139) (16. two lattice results are shown: filled circles correspond to fixing the strange quark mass by fitting the K mass. to the same universal scale factor L. the result being expressed in terms of a certain number of lattice units a −1 at a certain value of gL . Figure 16. section 5.140) can be tested in actual calculations. The behaviour predicted by (16. Consider a correlation function for a scalar field. namely E n = Mφ . A quantity such as the ρ meson mass is calculated (via a correlation function of the form (16. For each hadron. d masses.140) is remarkable.135) using (16.10 a precision calculation by the CP-PACS collaboration (Aoki et al 2000) of the spectrum of light hadrons in quenched QCD. τ )φ(0)| | |φ(0)|n |2 e− En τ . one can then go on to make predictions for the other low-lying hadrons. as it implies that all calculated masses must be proportional. in the continuum limit a → 0. but we must remember that the renormalization schemes involved in the two cases are quite different. Having fixed the physical state of a by the ρ mass. a −1 can be converted to GeV. (16.132). Then the calculation is repeated for a different gL value and the new a −1 (GeV) extracted. A plot of ln[a −1 (GeV)] 2 versus 1/gL should then give a straight line with slope 2π/b and intercept ln L . By comparison with the known ρ mass. for simplicity: C(τ ) = = n |φ(x = 0. The value of L has been adjusted to fit the numerical data and has the value L = 1.6). the lattice spacing (as just noted) and the u. M = constant × = constant × 2π 1 exp − 2 a bgL L. The form (16. will survive: Mφ can be measured from a fit to the exponential fall-off as a function of τ.136) gives.139) and (16. integrating (16. As one example.141) As τ → ∞.1. we may expect QCD ≈ 50 L (Montvay and Munster 1994.70). from which it appears that the calculations are indeed being performed close to the continuum limit. expressed in lattice units a −1 . respectively.9 shows such a plot. In fact. taken from Ellis et al (1996). This may seem alarmingly far from the kind of value expected for QCD .139) is known as asymptotic scaling: it predicts how any physical mass.

The ρ and π masses are absent because they are used to fix the lattice spacing and the u. taken from Ellis et al (1996). d and s quarks. calculated in the quenched approximation (Aoki et al 2000): Filled circles are the results calculated by fixing the s quark mass to give the correct mass for the K meson. The horizontal lines are the experimental values. containing u. Copyright 2004 IOP Publishing Ltd . ln(a −1 in GeV) plotted against 1/gL . as adapted from Allton (1995).9.10. d masses.2 Figure 16. The mass spectrum of light mesons and baryons. Figure 16. open circles are the results of fitting the s quark mass to the φ mass.

This means that the ‘theory’ does not allow a consistent definition to be given of the s quark mass. UKQCD Collaboration) q using two degenerate flavours of dynamical (i. we show in figure 16. Although the overall picture is quite impressive. The data are well described by c 3 Comparison with matched data in the quenched approximation revealed very little difference.Figure 16.142) to the Cornell (Eichten et al 1980) or Richardson (1979) phenomenological potentials gives r0 0. r=r0 (16. at about the 10% level. The broken curve is the functional form (16. masses. As a second example of a precision result.11.e.145). The inconsistency must be attributed to the quenched approximation. unquenched) quarks3 on a 163 × 32 lattice. between the pairs of results.11 a lattice calculation of the static q¯ potential (Allton et al 2002. one dimensionful quantity has to be fixed in order to set the scale. The static QCD potential expressed in units of r0 (Allton et al 2002.65.49 fm. UKQCD Collaboration). As usual. defined by 2 r0 dV dr = 1. Copyright 2004 IOP Publishing Ltd . this has been done via the scale parameter r0 of Sommer (1994). This calculation was the first to establish quantitatively that the error to be expected of the quenched approximation is of order 10%—an encouragingly small value. implying that quenched calculations of other phenomenologically important quantities will be reliable to within that sort of error also. in this case. conveniently in the range ¯ which is well determined by c¯ and bb data. In the present case. there is clear evidence of a disagreement.142) Applying (16.

since—if the potential maintained this form—it would cost an infinite amount of energy to separate a quark and an anti-quark. Nevertheless. There is no evidence for q string breaking in figure 16.the expression V (r ) = V0 + σ r − where. Luscher 1981).2 that there is good evidence to believe that the hadron spectrum predicted by QCD should show signs of a symmetry (namely.3.143) is—up to a constant—exactly the functional form mentioned in √ chapter 2.143) and where V0 has been chosen such that V (r0 ) = 0.146) 3 where µ is some energy scale. in which A has the ‘universal’ ¨ ¨ value π/12 0. Besides. the form (16. equation (2. But we also noted that the most obvious signs of such a symmetry—parity doublets in the hadronic spectrum—are conspicuously absent. The parameter A is found to have a value of about 0. enough energy will be stored in the ‘string’ to create a q¯ pair from the vacuum: q the string then breaks and the two q¯ pairs form mesons. A would be given by one-gluon exchange as A = 4 αs (µ) (16. The quantity σ (there called b ) is referred to as the ‘string tension’ and has a value of about 465 MeV in the present calculations.3. we propose to include the topic at the present stage. In lowest-order perturbation theory and in the continuum limit. This would give αs 0. Thus.65 − A) r0 r −1 . chiral symmetry) which would be exact if the u and d quarks were massless. The resolution of this puzzle lies in the concept of ‘spontaneous symmetry breaking’.26.22.143) becomes r0 V (r ) = (1.142).11 but we must note that the largest distance probed is only about 1. in accordance with (16. −1 − A r0 r (16. Copyright 2004 IOP Publishing Ltd . Our third and last example of lattice QCD calculations concerns chiral symmetry breaking. a reasonable value for µ 3 GeV.145) is predicted by the ‘universal bosonic string model’ (Luscher et al 1980. and which should survive as an approximate symmetry to the extent that m u and m d are small on hadronic scales. Interestingly. σ = (1. seeing the concept in action here will provide good motivation for the detailed study in part 7. We learned in section 12.22). which forms the subject of the next part of this book (chapters 17–19).3 fm. since it is one on which significant progress has recently been made within the lattice approach.65 − A) 2 r0 (16. The existence of the linearly rising term with σ > 0 is a signal for confinement. Phenomenological models suggest a value of around 440 MeV (Eichten et al 1980).145) Equation (16. (16.144) A r (16. But at some point.

147) by the Ginsparg–Wilson (1982) relation / / / / γ5 D + Dγ5 = a Dγ5 D. All these approaches are being numerically implemented with very promising results. chiral symmetry simply could not be realized at non-zero lattice spacing: this is the content of the Nielsen–Ninomiya theorem (Nielsen and Ninomiya 1981a. while avoiding the fermion doubling problem. b. 1994. break chiral symmetry explicitly.149)–(16.150) The symmetry under (16. But the modification to the derivative made in (16. the ‘Wilson’ (1974a) fermions mentioned earlier. As we shall see in chapter 17. or a suitable generalization of it.c). Admittedly. 1995). with infinitesimal variations proportional to / δψ = γ5 (1 − 1 a D)ψ 2 δψ = ψ(1 − 1 / 2 a D)γ5 . supposing we could somehow ignore the right-hand inequality in (16. not so easy—but that problem has now been solved. This can easily be seen by noting (see (12. therefore. In particular. rather than having to be extracted only in the continuum limit.131)— in such a way as not to violate chiral symmetry from the start. which reduces to (16. however. Unfortunately.155) for example) that the crucial property required for chiral symmetry to hold is / / γ5 D + Dγ5 = 0 (16. at finite lattice spacing. (16. it is difficult to construct a lattice theory with fermions—even massless ones.21) contains a piece with no γ -matrix.132). provides a lattice theory with all the fundamental symmetry properties of continuum chiral gauge theories (Hasenfratz et al 1998). which will not satisfy (16.147) / where D is the SU(3)c -covariant Dirac derivative. For example. (16. In the last few years a way has been found to formulate chiral gauge theories satisfactorily on the lattice at finite a . subject to quite mild assumptions. ‘classically perfect fermions’ (Hasenfratz and Niedermayer 1994).‘Spontaneous symmetry breaking’ is an essentially non-perturbative phenomenon and its possible occurrence in QCD is. indeed in three different ways: Kaplan’s ‘domain wall’ fermions (Kaplan 1992). in view of the comments following (16. b. so that the effects attributable to spontaneous symmetry breaking can be studied at finite a . Indeed. terms responsible for the breaking will vanish in the continuum limit but it would be much better to start with an action that preserved chiral symmetry.147) as a → 0. Finding an operator which satisfies (16.150).148) is. a particularly suitable problem for investigation by lattice calculations.147). and ‘overlap fermions’ (Narayanan and Neuberger 1993a. this would open up the possibility of being able to tackle light quarks and hadrons. for a long time it was thought that.148) This relation implies (Luscher 1998) that the associated action has an exact ¨ symmetry. a dramatic signal of the spontaneous breaking of a global symmetry is the appearance of a massless (Goldstone) boson: in the Copyright 2004 IOP Publishing Ltd . The key is to replace the condition (16.149) (16.

is something that can be π explicitly checked in a numerical simulation: the square of the pion mass should be proportional to the quark mass (taking m u = m d ).2—it is reasonable to suppose that this is because chiral symmetry is broken explicitly by (small) quark masses m u . m d in the QCD Lagrangian. ˆ ¯ˆ ¯ˆ ˆ in the physical vacuum. a simple model of chiral symmetry breaking implies the relation (Gasser and Leutwyler 1982) m2 = − π (m u + m d ) 2 2 fπ ¯ˆ ¯ˆ ˆ |u u + d d| ˆ (16.155) µ ∂B Copyright 2004 IOP Publishing Ltd . The crucial point here is that ˆ this vev remains non-zero even as the quark mass tends to zero—that is. this is taken to be the pion.3.79)) it is clear that the vev of ˆ ˆ ψ ψ can be written as ˆ ˆ |ψ ψ| = ∂ (ln Z QCD ) ∂m (16.11)). in the Nambu vacuum (see equation (18. The average value of s is then given by kB T ∂ ln Z s = . not massless and—as discussed in section 18. Now consider an analogous problem in statistical mechanics. such a non-zero vev can arise. (16.54)) f π 93 MeV and |u u + d d| is the expectation value. of course. we shall show in equation (18.1 we shall learn how. in particular.66) that m 2 is proportional to (m u + m d ).152) where Sg is the action for the gauge fields with link variables U and only one fermion field ψ of mass m is treated. The partition function is Zs = exp −(Hs − µs B)/kB T (16. Indeed. A simple analogy may help us to see how such a non-zero vev may arise. Here. therefore.151) ¯ˆ ¯ˆ ˆ where (cf (18. The physical pion is. The existence of such a nonzero vev for a field operator is a fundamental feature of spontaneous symmetry breaking.153) and we are specifically thinking of the limit as m → 0. Then (cf (16. as the ‘explicit’ chiral symmetry breaking is ‘turned off ’. Let us note here. in which the degrees of freedom are spins which can interact with one another via a Hamiltonian Hs and (via their associated magnetic moment µ) with an external magnetic field B.154) where s is the component of spin along the field B. as we shall see in the following chapter. Consider the quantity Z QCD = U ψ ψ exp − Sg − / ψ(i D − m)ψ d4 x E (16. of course. of the operator u u + d d .present (chiral) case. taken after the infinite volume limit a → 0 (see the penultimate paragraph of section 17. In section 18. imply that the vev is zero. that the conventional definition of the vacuum | that we have used hitherto would.1). Indeed.

The limitations of this analogy will be discussed in section 17. It is clear that the expectation m 2 ∝ m π K is generally well borne out. of the pion mass 2 squared (in units of 1/ r0 where r0 = 0. normalized at the reference point m = m ref (Giusti 2002).13 on figure 16. More sophisticated calculations predict corrections to (16. In particular. The existence of such a non-zero vacuum value. done in the quenched approximation.3. Let us now see whether the prediction (16. Note that the physical pion mass is at (m π r0 )2 0.1 but for the moment it does provide a useful physical picture in ˆ ˆ which ‘ |ψ ψ| ’ can be regarded as an ‘internal field’ arising spontaneously in the ground state | . Figure 16. Such a non-zero s occurs in a ferromagnet below its transition temperature.151).151) is confirmed by numerical calculations in lattice QCD.−2 Figure 16. The pion mass squared in units of r0 . surviving in the symmetry limit m → 0.12 shows a recent compilation (Giusti 2002) of calculations.12: in physical units the lowest pion masses reached at present lie somewhat below 200 MeV and Copyright 2004 IOP Publishing Ltd . versus the quark mass m . normalized √ at the reference point m = m ref such that m π (m = m ref ) = 2m K with m K = 2 495 MeV (note that 2m 2 r0 3. Sharpe 1992). including terms of the form m ln m (Bernard and Golterman 1992. An even better analogy is to the ‘condensates’ relevant to superfluids and superconductors. even when the ‘external field’ m is reduced to zero. we may regard a non-zero value of s as arising ‘spontaneously’ if it survives (in the thermodynamic limit V → ∞) even as B → 0.5 fm) versus quark mass m .12. ¯ˆ ¯ ˆ ˆ which we shall study in chapter 17: the vev |u u + d d| is called the ‘chiral ˆ condensate’. is fundamental to the concept of spontaneous symmetry breaking. where it is related to the ‘internal field’.16). which may be seen at the lowest m values (Draper et al 2002).

a highly mature field. .9). some 30 years after its initial inception by Wilson (1974). Problems 16.19).1 Verify equation (16. 3. .2 Verify equation (16.the lightest quark mass is about 15 MeV (Draper et al 2002). We have been able to give only a brief introduction into what is now. we can confidently say that. QCD is established as the correct theory of the strong interactions of quarks. Show that the partition function Z N is given by N−1 ZN = 2 n=1 (2 cosh K n ) where K n = Jn /kB T . Hence.18) is (16. 2.4 Use (16. .5 Verify (16. 16. . Z N ∂ K p ∂ K p+1 . calculate the entropy for the particular case in which all the Jn ’s are equal to J and N 1 and discuss the behaviour of the entropy in the limits T → ∞ and T → 0. . N take the values sn = ±1 and the energy of each spin configuration is N−1 E =− n=1 Jn sn sn+1 where all the constants Jn are positive. Lattice QCD is now a major part of particle physics. both in the short-distance (perturbative) regime and in the long-distance (non-perturbative) regime.62). to the numerical algorithms and to the treatment of fermions—to name a few of the issues. . Show that the average value s p s p+1 of the product s p s p+1 is given by s p s p+1 = Show further that s p s p+ j = 1 ∂ j ZN . spin variables sn at sites labelled by n = 1. Let ‘ p’ denote a particular site such that 1 p N.25) in (16.28). 16.10). 16. The actual physical values can be expected to be reached quite soon. 16.60) and (16.6 In a modified one-dimensional Ising model.27) to verify (16. beyond reasonable doubt. ZN ∂Kp Copyright 2004 IOP Publishing Ltd . A great deal of effort has gone into ingenious and subtle improvements to the lattice action. From the perspective of this chapter and the previous one.3 Show that the momentum space version of (16. ∂ K p+ j 1 ∂ ZN . 16.

in the case J1 = J2 = · · · = JN = J . Copyright 2004 IOP Publishing Ltd . show that. considering the T → ∞ and T → 0 limits explicitly. s p s p+ j = e− j a/ξ where ξ = −a/[ln(tanh K )] and K = J/kB T . Discuss the physical meaning of ξ .Hence.


By contrast. |B :1 ˆ |A = φ † |0 A in mind that |0 is the full (interacting) vacuum. such as parity doublets in particular. ˆ |B = φ † |0 B (17.17 SPONTANEOUSLY BROKEN GLOBAL SYMMETRY Previous chapters have introduced the non-Abelian symmetries SU(2) and SU(3) in both global and local forms. these symmetries are also applied. To understand this. The physical consequences of spontaneous symmetry breaking turn out to be rather different in the global and local cases. 17. rather than | Copyright 2004 IOP Publishing Ltd .2? (ii) How can weak interactions be described by a local non-Abelian gauge theory when we know the mediating gauge field quanta are not massless? The answers to these questions each involve the same fundamental idea. of the global chiral symmetry introduced in section 12. in which the generators of the symmetry are Hermitian field operators and the states are created by operators acting on the vacuum. Remarkably enough. consider two states |A . spontaneously broken local symmetry will be discussed in chapter 19 and applied in chapter 22. the essentials for a theoretical understanding of the phenomenon are contained in the simpler global case. and we have seen how they may be applied to describe such typical physical phenomena as particle multiplets and massless gauge fields. However. however. The application to spontaneously broken chiral symmetry will be treated in chapter 18.3. Consider the following two questions: (i) Why are there no signs in the baryonic spectrum. what could go wrong with the argument for symmetry multiplets that we gave in chapter 12. in response to question (i). which we consider in this chapter.3. we must use the field theory formulation of section 12.1) . which is a crucial component of the Standard Model and perhaps also of theories which go beyond it.1 Introduction We begin by considering. in two cases where the physical phenomena appear to be very different. This is the idea that a symmetry can be ‘spontaneously broken’ or ‘hidden’. but it must be borne 1 We now revert to the ordinary notation |0 for the vacuum state. in the Standard Model. the symmetries considered hitherto may be termed ‘manifest symmetries’. Thus.

7) ˆ ˆˆ ˆ ˆˆ ˆ ˆ E B |B = H |B = H φ † |0 = H ( Q φ † − φ † Q)|0 .3) (17.2) (17. Thus φ † is ‘rotated’ into φ † A B ˆ by U . we see that EA = EB if (17. a multiplet structure will emerge provided that the vacuum is left invariant under the symmetry transformation.100)) A B ˆ ˆ ˆ [ Q.3) H Q φ † |0 = Q H φ † |0 A A ˆ ˆ ˆ = Q H |A = E A Q|A † ˆˆ ˆ ˆ ˆ = E A Q φ A |0 = E A (φ † + φ † Q)|0 B A if (17.6) (17. comparing (17. We want to see what assumptions are necessary to prove that EA = EB We have Now if ˆ ˆ where H |A = E A |A and H |B = E B |B .ˆ ˆ where φ † and φ † are related to each other by (cf (12.2) (17. such that ˆ ˆ [ Q.10) Thus.8) ˆ ˆ Remembering that U = exp(iα Q). (17. (17.8) with (17. we see that (17.7) holds. (17.7) holds. which is to say when ˆ Q|0 = 0. φ † ] = φ † A B ˆ for some generator Q of a symmetry group. The ‘spontaneously broken symmetry’ situation arises in the contrary case—that is. B A A ˆ Q|0 = 0 we can rewrite the right-hand side of (17. (17.7) is equivalent to ˆ |0 ≡ U |0 = |0 . and the operators will create states related by the symmetry transformation.11) In this case.9) using (17.2) is equivalent to ˆ ˆ ˆ ˆ ˆ U φ † U −1 ≈ φ † + i φ † A A B (17.6). H ] = 0.5) (17. = E A |B whence.6) as ˆ ˆˆ ˆ ˆˆ using (17. the argument for the existence of symmetry multiplets breaks down and although the Hamiltonian or Lagrangian may exhibit a non-Abelian Copyright 2004 IOP Publishing Ltd .4) ˆ ˆ ˆ ˆ for an infinitesimal transformation U ≈ 1 + i Q. (17. when the vacuum is not invariant under the symmetry.

H ] = 0). we obtain ˆ ˆ ˆ ˆ ˆ ˆ 0| j 0(x) Q|0 = 0|ei P·x j 0 (0)e−i P·x Q|0 ˆ ˆ ˆ ˆ = 0|ei P·x j 0 (0) Qe−i P·x |0 where the second line follows from ˆ ˆ [ P µ . though intuitively helpful. so that ˆ Q|0 = 0. theorem about the quantum field vacuum. from Schr¨ dinger picture operators at t = 0 to o Heisenberg operators at t = 0. implemented by the unitary operator2 U (x) = ˆ ˆ exp i P · x (where P µ is the 4-momentum operator). consider the vacuum matrix element 0| j 0 (x) Q|0 . and it seems tempting to infer that.symmetry. but there is another way of thinking about it which is somewhat less abstract though also less rigorous. in any case. The ˆ basic condition is Q|0 = 0. The associated ‘charge’ is the Hermitian operator Q = j 0 d3 x that ∂µ j ˙ ˆ and Q = 0. We have hitherto assumed that the transformations of such a U(1) group are representable in the space of physical states by unitary operations ˆ ˆ ˆ U (λ) = exp iλ Q for arbitrary λ. ˆ ˆ To prove this result.12) 2 If this seems unfamiliar. application of Q Thus we have the physically suggestive idea of ‘degenerate vacua’ (they must be ˆ degenerate since [ Q. We shall see in a moment why this notion. U |0 = |0 ). Copyright 2004 IOP Publishing Ltd . such ˆ ˆ ˆµ = 0. not zero. suggested before. ˆ From translation invariance. so we begin our discussion with a somewhat formal. Fabri and Picasso (1966) showed that there are actually two possibilities: ˆ ˆ (a) Q|0 = 0 and |0 is an eigenstate of Q with eigenvalue 0. the ˆ to the vacuum gives. |0 . so that |0 is ˆ (i. or ˆ invariant under U ˆ (b) Q|0 does not exist in the space (its norm is infinite). but another possible vacuum. this will not be manifested in the form of multiplets of mass-degenerate particles. Q] = 0 (17. is not rigorous. but nonetheless fundamental. It would seem.2 The Fabri–Picasso theorem ˆ Suppose that a given Lagrangian Ä is invariant under some one-parameter ˆ continuous global internal symmetry with a conserved Noether current j µ . The statement (b) is technically more correct than the more intuitive statements ˆ ˆ ‘ Q|0 = 0’ or ‘U |0 = |0 ’. that the properties of the vacuum are all important. in this case. with the vacuum invariant under U . it may be regarded as the four-dimensional generalization of the transformation (I.7) in appendix I of volume 1. 17.e. The preceding italicized sentence does correctly define what is meant by a spontaneously broken symmetry in field theory.

in a sense. be reversed. we introduce the BCS ground state for a superconductor in a way which builds on the Bogoliubov model of a superfluid. We are then prepared for the application. involving just one complex scalar field. there are crucial insights to be gained by considering the analogous phenomenon in condensed matter physics. the argument can also. unless Q|0 = 0.14) (17.13) which states that the matrix element we started from is. After a brief look at the ferromagnet. We might now simply proceed to the chiral symmetry application. In particular. however. and so ˆ ˆ ˆ ˆ 0| j 0 (x) Q|0 = 0| j 0(0) Q|0 (17. either ˆ 0 = 0 or Q|0 has infinite norm. we shall introduce the ‘Goldstone model’ which is the simplest example of a spontaneously broken global U(1) symmetry. we shall describe the Bogoliubov model for the ground state of a superfluid. independent of x . ˆ Now consider the norm of Q|0 : ˆ ˆ 0| Q Q|0 = = ˆ ˆ d3 x 0| j 0(x) Q|0 ˆ ˆ d3 x 0| j 0(0) Q|0 (17. We shall see that the excitations away from the ground state are massless modes and we shall learn. in chapter 18. Next. The generalization of this to the non-Abelian case will draw us in the direction of the Higgs sector of the Standard Model.ˆ ˆ since Q is an internal symmetry. Returning to condensed matter systems. that the concept of spontaneous symmetry breaking is so important to particle physics that a more extended discussion is amply justified. to spontaneous chiral symmetry Copyright 2004 IOP Publishing Ltd . and the symmetry is unitarily ˆ ˆ implementable by operators U = exp(iλ Q).17) ˆ ˆ then in fact ∂µ j µ = 0. Q is independent of t . The foregoing can be easily generalized to ˆ Q| ˆ non-Abelian symmetry operators Ti .15) ˆ which must diverge in the infinite volume limit. in fact. that such modes are an inevitable result of spontaneously breaking a global symmetry. Thus. via Goldstone’s theorem. But the vacuum is an eigenstate of P µ with eigenvalue zero.16) is the spatial integral of the µ = 0 component of a 4-vector (but not assumed to be conserved) and if it annihilates the vacuum ˆ Q(t)|0 = 0 (17. Coleman (1986) proved that if an operator ˆ Q(t) = ˆ d3 x j 0(x) (17. Remarkably enough. We believe. which provides an important physical example of a spontaneously broken global Abelian U(1) symmetry.

3 Spontaneously broken symmetry in condensed matter physics 17. with the symmetry unitarily implemented. indeed. in case (a) the charge annihilates a ground state. Then one ground state will be related to another by |0. in fact. In case (a) of the theorem. in the case of fields (extending presumably throughout all space). though. existed. However. a . one involving interacting fields) may well have unsuspected properties—which may. An essential aid to understanding hidden symmetry in quantum field theory is provided by Nambu’s (1960) remarkable insight that the vacuum state of a quantum field theory is analogous to the ground state of an interacting many-body system. in the infinite volume limit. Sometimes we may meet systems in which apparently more than one state has the same lowest energy eigenvalue. In simple models one can verify that these alternative ground states are all orthogonal to each other. which is often described as involving a ‘degenerate ground state’—a term we shall now elucidate. In chapter 19 we shall see how a different aspect of superconductivity provides a model for the answer to question (ii).18)— ˆ since Q|0 does not exist—and we do have the possibility of many degenerate ground states. . and the true ground state will turn out to be a unique linear superposition of them. However.1 The ferromagnet We have seen that everything depends on the properties of the vacuum state. b ˆ (17. and so all of them are really identical. in practice. It is the state of lowest energy—the equilibrium state. none of these states will be the true ground state: tunnelling will take place between the various degenerate states. a = eiλ Q |0. But we can postulate (even if we cannot yet prove) properties of the quantum field theory vacuum |0 which are analogous to those of the ground states of many physically interesting many-body systems—such as superfluids and superconductors. following Nambu’s profound analogy with one aspect of superconductivity. Now the ground state of a complicated system (for example. the Fabri–Picasso theorem shows that there is an alternative possibility. the only possibility for systems of finite spatial extent. . built on these alternative ground states. suppose that several ground states |0.breaking (question (i) of the first paragraph of this chapter). that the ground state of any system described by a Hamiltonian is non-degenerate. be very hard to predict from the Hamiltonian. given the kinetic and potential energies as specified in the Hamiltonian. b . Now it is generally the case. a state which is not the true ground state may have an extremely long lifetime.3. This is. we cannot write (17. Copyright 2004 IOP Publishing Ltd . in fact. And each member of every ‘tower’ of excited states. For. the ground state is unique. . however. in quantum mechanics. Yet.18) for some λ. 17. |0. to name two with which we shall be principally concerned. In case (b).

the eigenstates of H S are labelled by the eigenvalues of ˆ ˆ total squared spin. The spontaneous symmetry breaking which is the true model for particle physics is that in which a many-body ground state is not an eigenstate (trivial or otherwise) of the symmetry operators of the Hamiltonian: rather it is a superposition of such eigenstates. It is clear that the ground state is not invariant under the spin-rotation symmetry of H S . each one built on one of the possible orthogonal ground states. At first sight. that is. −N/2) of Sz for this value of S. they cannot be related by unitary transformations of the form ˆ eiλ Q . As usual with angular momentum in quantum mechanics. say the state with eigenvalue Sz = N/2. consider two ground states differing by a spin rotation. This Hamiltonian is invariant under spatial rotations. Nevertheless.. say. and ‘the ground state is degenerate’. a familiar example of these ideas seems to be that of a ferromagnet below its Curie temperature TC . However. there are some useful insights to be gained from the ferromagnet. the spins are all aligned along the 3-axis. cos α). say of Sz = i Siz . and this state is (2 · N/2 + 1) = (N + 1)-fold degenerate. . and spins at different sites are assumed to commute. The quantum mechanical ground state of H S is an eigenstate with total spin quantum number S = N/2. so that the spins are fully aligned. But any single tower must constitute a complete space of states. It follows that states in different towers belong to different complete spaces of states. this ground state is degenerate. it has to be emphasized that this ferromagnetic ground state does. as would be the case for an invariant ground state. sin α. in fact. Consider an ‘ideal Heisenberg ferromagnet’ with N atoms each of spin. In the first. N/2 − 1. since it only depends on the dot product of the spin operators. to different—and inequivalent—‘worlds’. 2 1 1 0 (N products) N (17.1 . First. So two important features of what we have so far learned to expect of a spontaneously broken symmetry are present—namely. We shall explore this for the superfluid and the superconductor in due course.. and of one component of spin. We are free to choose any one of these degenerate states as ‘the’ ground state. .19) Copyright 2004 IOP Publishing Ltd . . . Thus the first ground state is ˆ χ0 = 1 0 1 0 . ‘the ground state is not invariant under the symmetry of the Hamiltonian’. In particular. 2 ˆ ˆ described by a Hamiltonian of Heisenberg exchange form H S = −J Si · S j . which would require the eigenvalues S = Sz = also orthogonal to all the members of other towers. ˆ allowing for all the possible eigenvalues (N/2. Furthermore. Such ˆ ˆ ˆ rotations are implemented by unitary operators exp(i S · α) where S = i Si . respect the symmetry of H S in the sense that it belongs to an irreducible representation of the symmetry group: the unusual feature is that it is not the ‘trivial’ (singlet) representation. and in the second along the axis n = (0. where i and j label the atomic sites.

they behave like massless particles when quantized—and this is another feature we should expect when a symmetry is spontaneously broken. Thus any two such ‘rotated ground states’ are. We have been assuming that one particular ground state (e. Suppose for simplicity that the magnet is one-dimensional (but the spins have all three components). 1 cos α/2 i sin α/2 .while the second is (cf (4.23) 2 ( Si− S j + + S j − Si+ )χn = i= j i =n ˆ Thus the state χn is not an eigenstate of H S .24) χq = √ ˜ N n are eigenstates. In this respect. The states (17. However. the only non-zero contributions from the first (bracketed) term in (17. leading to the result 1 ˆ ˆ ˆ ˆ χi . N (17. indeed. such that 1 0 ˆ = (17. where Si+ = Si x + i Siy . orthogonal in the infinite volume limit. so as to restore it to ‘up’. the one with Sz = N/2) has been somehow Copyright 2004 IOP Publishing Ltd .g.. Here q is one of the discretized wavenumbers produced by appropriate boundary conditions. Consider the state ˆ ˆ ˆ ˆ ˆ χn = Sn− χ0 where Sn− is the spin-lowering operator Sn− = ( Snx − i Sny ) at site n . as is usual in one-dimensional ‘chain’ problems (see section 16.22) ˆ ˆ ˆ (remembering that spins on different sites commute).24) represent spin waves and they have the important feature that for low q (long wavelength) their frequency ω tends to zero with q (actually ω ∝ q 2 ).74)) χ0 (α) = cos α/2 i sin α/2 . (17. which goes to zero as N → ∞.22) come from terms in which ˆ ˆ either Si+ or S j + act on the ‘down’ spin at n . The ˆ action of H S on χn can be found by writing ˆ ˆ Si · S j = i= j i= j 1 ˆ ˆ 2 ( Si− S j + ˆ ˆ ˆ + S j − Si+ ) + Siz Sˆj z (17.20) is (cos α/2) N . The ˆ ˆ ‘partner’ operator Si− (or S j − ) then simply lowers the spin at i (or j ). therefore.20) The scalar product of (17. We may also enquire about the excited states built on one such ground state. The ferromagnet gives us one more useful insight.2). a little more work shows that the superpostitions 1 eiqna χn (17.21) Sn− 0 n 1 n ˆ so Sn− χ0 differs from the ground state χ0 by having the spin at site n flipped.19) and (17.. ˆ Since all Si+ operators give zero on a spin ‘up’ state. ˆ say the one with Sz eigenvalue N/2.

both at finite temperature. it strongly suggests that what is really characteristic about the phenomenon is that it entails ‘spontaneous ordering’. The ferromagnet. This term will. Consider now the two limits B → 0 and N → ∞. the N + 1 different Sz eigenstates become degenerate.26) 3 It is worth pausing to reflect on the idea that ordering is associated with symmetry breaking. section 55 or Parry 1973.3. if N → ∞ at finite B = 0. and has Sz = N/2. and we have an ensemble in which each enters with an equal weight: there is. provides an easily pictured system exhibiting many of the features associated with spontaneous symmetry breaking. When B → 0 at finite N. even as N → ∞ (but only after B → 0). indeed. chapter XIV of Landau and Lifshitz (1980)).25) ˆ ˆ where φ † (x) creates a boson of mass m at position x. representing the effect of an applied field B directed along the z-axis. for example. therefore. which is assumed to be weak (see. However. chapter 1). the single state with Sz = N/2 will be selected out as the unique ground state and this asymmetric situation will persist even in the limit B → 0. The field which develops a non-zero equilibrium value below TC is called an ‘order parameter’. ensure that the ground state is unique. TC . which is aligned with the external B and survives even as B → 0. This concept forms the basis of Landau’s theory of second-order phase transitions (see. thus ‘spontaneously’ breaking the symmetry. for example. most importantly. no loss of symmetry. Copyright 2004 IOP Publishing Ltd .27) (17.3 Generally such ordering occurs below some characteristic ‘critical temperature’. 17. This H describes identical bosons of mass m interacting via a potential v. therefore.2 The Bogoliubov superfluid Consider the non-relativistic Hamiltonian (in the Schr¨ dinger picture) o 1 ˆ H= 2m + 1 2 ˆ ˆ d3 x ∇ φ † · ∇ φ ˆ ˆ ˆ ˆ d3 x d3 y v(|x − y|)φ † (x)φ † ( y)φ( y)φ(x) (17.‘chosen’. We note at once ˆ that H is invariant under the global U(1) symmetry ˆ ˆ ˆ φ(x) → φ (x) = e−iα φ(x) the generator being the conserved number operator ˆ N= ˆ ˆ φ † φ d3 x (17. Schiff 1968. We now turn to an example much more closely analogous to the particle physics applications: the superfluid. In a (classical) mean-field theory approximation we suppose that an ‘internal field’ is ‘spontaneously generated’. But what does the choosing? The answer to this is clear enough in ˆ the (perfectly realistic) case in which the Hamiltonian H S is supplemented by a ˆ term −gµB i Siz .

We impose periodic boundary conditions at the cube faces and the free particle energies are k = k 2 /2m . Inserting (17. below the superfluid transition temperature TS . H ] = 0. (17. Naturally. we expect from the statistical mechanics of Bose–Einstein condensation (Landau and Lifshitz 1980.k . with all ˆ ˆ† other commutators vanishing. it is reasonable to hope that most of the particles remain in the condensate.33) When a weak repulsive v is included. only relatively few being excited to states with 4 This is non-relativistic physics. We shall work always at zero temperature. in the limit v → 0. Our ultimate concern will be with the way this symmetry is ‘spontaneously broken’ in the superfluid ground state. which has k = 0.1) to ˆ H= k a k |0 ˆ† ˆ† k ak ak + 1 2 v(k 1 − k1 )ak1 ak2 ak ak ¯ ˆ† ˆ† ˆ ˆ 2 1 ( k1 + k2 − k1 − k 2 ) (17. the ground state will be proportional to |N. symmetry the physics will not involve any (hidden) multiplet structure. (17. section 62) that in the limit as v → 0 the ground state has all the particles ‘condensed’ into the lowest energy state. k2 is scattered (conserving momentum) to a pair in states k 1 . rather than a non-Abelian. so there is no anti-particle part.25) leads (problem 17.29) where the sum is over all momenta k1 . Copyright 2004 IOP Publishing Ltd . ak ] = δk. k 1 . We expand φ(x) as a superposition of solutions of the v = 0 problem. k2 subject to the conservation law imposed by the function: ( k) = 1 =0 if k = 0 if k = 0.29) is easily visualized as in figure 17.31) The interaction term in (17.ˆ ˆ which obeys [ N . But the nature of the ‘symmetry breaking ground state’ in this U(1) case (and in the BCS model of section 17. k2 .28) 2 k where ak |0 = 0.7) will serve as a physical model for non-Abelian cases also. which are plane waves4 quantized in a large cube of volume : 1 ˆ φ(x) = 1 ak eik· x ˆ (17.28) into (17. since this is an Abelian. k2 via the Fourier transform of v : v(k) = ¯ v( r)eik· r d3 r.30) (17. 0 = (a0 ) N |0 . ˆ† (17.1. A pair of particles in states k1 . ˆ We begin by re-writing H in terms of mode creation and annihilation ˆ operators in the usual way. ˆ is a one-particle state and [ak .32) Now. Thus.

N0 has been replaced by N).29). conceptually. |ground B .35) is of a completely different kind.29).34) the two terms on the left-hand side are each of order N0 / and hence finite. This amounts to saying that in the commutator a0 ˆ 1/2 a0 ˆ† 1/2 − a† ˆ 1/2 a0 ˆ 1/2 = 1 (17. from that which would be associated with a (hypothetical) ‘explicit’ number violating term in the original Copyright 2004 IOP Publishing Ltd . Let N0 be the number of particles with k = 0 where.2) to the following approximate form for H : ˆ ˆ H ≈ HB ≡ k ak ak E k + ˆ† ˆ + 1 2 N k 1 N2 v(0) ¯ 2 (17.35). in this limit. N0 ≈ N . we may effectively replace both a0 and a0 in the second term in ˆ 1/2 (17.1.29) by the number N0 . Bogoliubov (1947) argued ˆ† that. cannot be expected to be an eigenstate of the number operator. the number of a’s must equal the number of a † ’s in every term. as compared with H of ˆ B does not conserve the U(1) (number) symmetry (17.Figure 17.26) while (17. the ground state of ˆ ˆ ˆ HB . ˆ The most immediately striking feature of (17. Thus. by assumption.36) primed summations do not include k = 0 and terms which tend to zero as → ∞ have been dropped (thus. it is important to be clear that the number-non-conserving aspect of (17.35) v(k)[a k a−k + ak a−k ] ¯ ˆ† ˆ† ˆ ˆ where Ek = k + N v(k) ¯ (17. Replacing a0 and a0 by N0 leads ˆ ˆ† ˆ (problem 17. is that H ˆ H does: it is easy to see that for (17.26) to be a good symmetry. However. k = 0. while 1/2 their difference may be neglected as → 0. We now consider the limit N (and N0 ) → ∞ and → ∞ such that the density ρ = N/ (and ρ0 = N0 / ) stays constant. The interaction term in (17.

From this point of view. We now assert that HB can be written in the form ˆ HB = k ωk αk αk + β ˆ† ˆ (17. The limit is crucial here: it enables us to picture the condensate N0 as providing an infinite reservoir of particles.38) where f k and gk are real functions of k = |k|. αk ] = f k2 − gk ˆ ˆ† (17. N0 ≈ N and N0 / remains finite as → ∞. We choose f k and gk such that f k2 − gk = 1.26) either. A convenient choice is ˆ f k = cosh θk . ˆ What is |ground B ? Remarkably. the operators αk will ˆ ˆ ˆ† ˆ† ˆ† turn out to be precisely creation operators for quasiparticles which exchange particle number with the ground state. the addition of a term of the form ‘a † a a’. cosh θl al† + sinh θl a−l ] = ωl (cosh θl al† + sinh θl a−l ) Copyright 2004 IOP Publishing Ltd (17. We note first that it implies that.40) slightly indirectly. We must again at once draw attention to the fact that this transformation does not respect the symmetry (17. In arriving ˆ ˆˆ at (17. In fact. gk = sinh θk .42) . ˆ† The commutator of αk and αk is easily evaluated: ˆ 2 [αk . of course. as just anticipated. for one particular mode operator αl† . we have effectively replaced (17.38) is then said to be ‘canonical’. ˆ ˆ so that the a’s and the α’s have the same (bosonic) commutation relations and ˆ ˆ the transformation (17. The ultimate test. We verify (17. with which excitations away from the ground state can exchange particle number. and that αk acts as the creation operator for the quasiparticle of energy ωk .40) for certain constants ωk and β. of course. is whether such a state is a good approximation to the true ground state for a large but finite system. αl† ] = ωl αl† .35).37) where ρ0 = N0 / . Substituting for αl† from (17.40) implies.41) (17.28) by 1/2 ˆ φB (x) = ρ0 + 1 1/2 k=0 ak eik·x ˆ (17. since ak → e−iα ak while a−k → e+iα a−k . HB can be exactly diagonalized by means of the Bogoliubov quasiparticle operators (for k = 0) ˆ ˆ† αk = f k ak + gk a−k ˆ αk = f k ak + gk a−k ˆ† ˆ† ˆ† (17. Equation (17. that the ˆ ˆ† eigenvalues of HB are β + k (n + 1/2)ωk . ˆ ˆ ˆ ˆ [ HB .38).39) 2 while two α’s or two α † ’s commute.Hamiltonian—for example. a number-non-conserving ground state may appear more reasonable. we require ˆ ˆ ˆ ˆ ˆ ˆ [ HB.

46) where ρ = N/ . ¯ For consistency. ωl will tend to zero as |l| → 0 and ¯ we will have massless ‘phonon-like’ modes. but we can give ˆ ˆ† a formal expression for it. therefore. However. as the state with no non-zero-momentum quasiparticles in it. in fact. indicating that in the case of such a long-range force the massless mode associated with spontaneous symmetry breaking acquires a mass.48) .e. This is a complicated state in terms of the original ak and ak operators. As long as v(l) is less singular than l −2 as |l| → 0. then ¯ ωl → constant as |l| → 0 and the spectrum would not be that of a massless excitation.46) interpolates between phonon-like behaviour at small |l| and particle-like behaviour at large |l|. we require El2 − ωl2 − N 2 2 (v(l))2 = 0 ¯ (17. it is clear that it is defined as the state |ground B such that αk |ground ˆ B =0 for all k = 0 (17. ˆ (17.43) or (17.43) (17. (17. Since the α’s and a † ’s are related by a ˆ ˆ ˆ canonical transformation. In particular. giving the frequency as a function of the momentum (or wavenumber).which must hold as an identity in the a ’s and a †’s. ωl behaves essentially like ¯ l 2 /2m and the spectrum returns to the ‘particle-like’ one of massive bosons. it is an example of a ‘dispersion relation’.45) or (recalling the definitions of El and l ) ωl = l2 2m l2 + 2ρ v(l) ¯ 2m 1/2 (17. let us now concentrate on the ground state in this model. of course.44). there must exist a unitary operator UB such that ˆ ˆ ˆ −1 α k = UB a k U B ˆ Copyright 2004 IOP Publishing Ltd ˆ −1 ˆ ˆ a k = U B α k UB . Thus.44) v(l) cosh θl − (ωl + El ) sinh θl = 0. for large |l|. we note that if.47) i. Indeed.35) ˆ ˆ ˆ for HB . From (17.46) is an important result. if v(l) ∼ e2 /l 2 . the speed ¯ of sound will be (ρ v(0)/ m)1/2 . This will be the topic of chapter 19. Having discussed the spectrum of quasiparticle excitations. Furthermore. Colombic (the Fourier ¯ transform of e2 /| x|). if v(0) = 0. one finds that (ωl − El ) cosh θl + N N v(l) sinh θl = 0 ¯ (17. Equation (17.3). and some patient work with the commutation relations (problem 17. Using the expression (17. The value of tanh θl is then determined via either (17. then ωl ∼ |e|(ρ/ m)1/2 for small |l|. which ¯ is just the ‘plasma frequency’.40). as follows. Such a v is. v(l) ∼ 1/l 2 .

we have 1/2 ˆ (17. The important point is the following.e. in the non-interacting ground state (17. ˆ Consider now the expectation value of φ(x) in any state of definite particle ˆ number—that is. This generalizes (in the form of products of such ˆ operators) to the full HB case but we shall not need the detailed result: an analogous result for the BCS ground state is discussed more fully in section 17. This is just the situation alluded to in the paragraph before equation (17.19).51) ˆ ground|φB (x)|ground B = ρ0 .7. using the inverse of (17. Hence it follows that ˆ α k UB | 0 = 0 ˆ ˆ (17. 0 of (17.33).4. we abstract from the Bogoliubov model the crucial feature that the field acquires a non-zero expectation value in the true ground state. UB is evaluated for ˆ an HB consisting of a single k -mode only. we find the similar result: B (17. 0|φB (x)|N. but in (17. In problem 17. Furthermore. ˆ it is not an eigenstate of the symmetry operator N . (17. it is easy to see that this must vanish (there will always be a spare annihilation operator). ˆ |ground B = UB |0 does not have a fixed number of particles in it: that is to say. It is clear from expanding the exponentials ˆ that UB creates a state in which the number of a-quanta (i. as we have just seen.52) to the complete φ(x) and the true ground state |ground .38) ak = cosh θk αk − sinh θk α−k ˆ ˆ ˆ† together with (17.53) that is. in the infinite volume limit. as anticipated in the comment following (17. the original bosons) is not fixed.50) or (17. in an eigenstate of the symmetry operator N .Now we know that ak |0 = 0. If it were. Recalling the association Copyright 2004 IOP Publishing Ltd . → ∞ with fixed N/ .53)—we must ˆ again emphasize—|ground is not an eigenstate of the symmetry operator N . in which case the operator effecting ˆ the transformation analogous to (17. unlike the simple non-interacting ground state |N.52) ˆ The question is now how to generalize (17. 1/2 (17. in the limit N.47). However.36).49) ˆ ˆ and we can identify |ground B with UB |0 .50) N. Condition (17.53) would vanish. (17. We make the assumption that ˆ ground|φ(x)|ground = 0. Thus. in our discussion of the ferromagnet. ˆ this is not true of φB (x): for example.33).48) is U1 = exp[θ (a a − a †a † )/2] where ˆˆ ˆ ˆ θ replaces θk in this case.53) has the form of an ‘ordering’ condition: it is analogous to the non-zero value of the total spin in the ferromagnetic case. We are now at the heart of spontaneous symmetry breaking in field theory. 0 = ρ0 .

α B such that B ˆ ground. There are infinitely many such ground states since α is a continuous parameter. None of the states |ground.57) the original choice (17. then the Uα factors iα N · e−iα N and would cancel out. nevertheless. Whether these Copyright 2004 IOP Publishing Ltd . Instead.37) and (17. we are deciding ˆ (arbitrarily) to pick the ground state such that B ground|φB |ground B is aligned in the ‘real’ direction. knowing that |ground B is not an eigenstate of N . α is an eigenstate for N : instead. ˆ −1 we can regard Uα |ground B as an ‘alternative ground state’ |ground.52) corresponding to α = 0. such that ˆ the expectation value of φB has a definite phase. In choosing say α = 0. ˆ (17. α|φB |ground.55) may be written as B ˆ ˆ ˆ −1 ground|Uα φB Uα |ground B ˆ = e−iα B ground|φB |ground B .‘quantum vacuum ↔ many body ground state’ we expect that the occurrence of a non-zero vacuum expectation value (vev) for an operator transforming nontrivially under a symmetry operator will be the key requirement for spontaneous symmetry breaking in field theory. leaving a in (17. Hence (17.4 Goldstone’s theorem We return to quantum field theory proper and show. they are certain coherent superpositions of states with different eigenvalues N.55) ˆ ˆ ˆ −1 ˆ ˆ ˆ Now φB = Uα φB Uα where Uα = exp(iα N ).56) ˆ ˆ If |ground B were an eigenstate of N with eigenvalue N. Before leaving the superfluid. (17. which is only rigorous for a finite system but is. Salam and Weinberg (1962)) how in case (b) of the Fabri– Picasso theorem massless particles will necessarily be present. a similar situation obtains for the true ground ˆ state.54) But in that case our condition (17. we should be at liberty to replace (17. thus ‘spontaneously’ breaking the symmetry. (17. Since the original H has a U(1) symmetry under which φ transforms ˆ ˆ to φ = exp(−iα)φ. 17. In the next section we show how this requirement necessitates one (or more) massless modes. say. No physical consequence follows from choosing one rather than another but we do have to choose one.52) becomes B ˆ ground|φB |ground B ˆ = e−iα B ground|φB |ground B . α B ˆ = e−iα B ground|φB |ground B (17. we examine (17. very ˆ ˆ suggestive.52) in another way. via Goldstone’s theorem (1961). following Goldstone (1961) (see also Goldstone. however. By hypothesis.56) would become just e ˆ contradiction.37) by 1/2 ˆ φB = e−iα ρ0 + 1 1/2 k=0 ak e−iα eik·x .

64) ˆ d3 x j0(x) + d3 x div ˆ j(x) = 0 (17.53). (17. (17. (17.62) If the surface integral vanishes in (17.58). which is independent of time and is the space integral of the µ = 0 component of a conserved Noether current: ˆ Q= ˆ j0(x) d3 x.64) involves local operators separated by a very large space-like interval and. we can write (17. is ˆ not annihilated by Q. call not invariant under Q. φ(y)]|0 . in fact. ˆ Suppose. we have.59) will give some other field. If Q were to annihilate |0 . that we have a Lagrangian Ä with a continuous symmetry ˆ generated by a charge Q. as usual. ∂ ∂x0 whence ∂ ∂x0 ˆ ˆ d3 x 0|[ j0(x). φ(0)]|0 = − = − ˆ d3 x 0|[div ˆ j(x).59) as ˆ 0 = 0|φ (0)|0 = 0| ˆ ˆ d3 x j0(x). the vanishing of (17. and gauge fields are absent: the local symmetry case is treated in chapter 19. (17. From (17. In this chapter we are concerned solely with global symmetries. ∂µ j µ = 0. ˆ Suppose φ(y) is some field operator which is not invariant under the continuous symmetry in question and consider the vacuum expectation value ˆ ˆ 0|[ Q. the hallmark of the hidden symmetry situation is the existence of ˆ some field (here φ (0)) with non-vanishing vacuum expectation value. just as in (17.63) (17. then. by assumption. (17.60) (17. φ(0)]|0 .59) Just as in equation (17. Since φ is ˆ the commutator in (17. φ(0) |0 . this would ˆ clearly vanish: we investigate the consequences of its not vanishing. therefore.e. on whether the theory also contains gauge fields. however.64) would seem to be Copyright 2004 IOP Publishing Ltd .particles will actually be observable depends. and we may set y = 0. ˆ independent of y. i. The commutator in (17.64).13).58) We consider the case in which the vacuum of this theory is not invariant. φ(0)]|0 ˆ dS · 0|[ ˆ j(x). ˆ it φ (0).61) will be independent of x 0 . thus. translation invariance implies that this vev is.61) ˆ Since.

But it has been assumed to be not zero. Copyright 2004 IOP Publishing Ltd . Performing the spatial integral on the right-hand side. and the reader is referred to Guralnik et al (1968) and Bernstein (1974). Inserting a complete set of states in (17.67) But this expression is independent of x 0 . we obtain 0= = d3 x n ˆ ˆ ˆ ˆ { 0| j0(x)|n n|φ(0)|0 − 0|φ(0)|n n| j0 (x)|0 } (17. with pn the 4-momentum eigenvalue of the state |n . in the theory. Although in a sense less ‘dynamical’ than the Bogoliubov superfluid (or the BCS superconductor. of course.67). by choosing a parameter in the Lagrangian appropriately. ˆ hence. The superfluid provided us with an explicit model exhibiting the crucial ˆ non-zero expectation value ground|φ|ground = 0. Massive states |n will produce explicit x 0 -dependent factors e±iMn x0 ( pn0 → Mn as the δ-function constrains pn = 0). in a theory with a continuous symmetry. A detailed analysis of exactly how this changes the argument being presented here will take us too far afield at this point. Thus. (17.unproblematic.61) on x 0 leads to the necessity for a massless particle in the spectrum. The only possibility is states whose energy pn0 goes to zero as their 3-momentum does (from δ 3 ( pn )). appears to lead inevitably to the necessity of having a massless particle.65) d3 x n ˆ ˆ ˆ ˆ { 0| j0(0)|n n|φ(0)|0 e−i pn ·x − 0|φ(0)|n n| j0(0)|0 ei pn ·x } (17. to be discussed shortly). some state or states must exist among |n such that 0| j0|n = 0 and yet (17.67) is independent of x 0 . the existence of a non-vanishing vacuum expectation value for a field. We shall treat the ‘spontaneously broken’ gauge theory case in chapter 19. Let us now see how the independence of (17.66) using translation invariance. in which the now expected massless mode emerged dynamically. we find (omitting the irrelevant (2π)3 ) that 0= n ˆ ˆ ˆ ˆ δ 3 ( pn )[ 0| j0(0)|n n|φ(0)|0 ei pn0 x0 − 0|φ(0)|n n| j0 (0)|0 e−i pn0 x0 ]. (17. Equally. Hence. We now discuss a simpler relativistic model. but in less formal terms.67) vanishes identically. This is the Goldstone (1961) result.61). if we take |n = |0 . in which the symmetry breaking is brought about more ‘by hand’—that is. massless: they are called generically Goldstone modes. and such states contribute zero to (17. Such states are. or particles. the matrix elements of j0 between |0 and such a massive state must vanish. Indeed so it is—with the exception of the case in which the symmetry is local and gauge fields are present. this Goldstone model does provide a very simple example of the phenomenon of spontaneous symmetry breaking in field theory.

we expand φ about the classical minimum at φ = 0. with 1 ˆ ˆ ˆ φ = √ (φ1 − iφ2 ) 2 described by the Lagrangian ˆ ˆ ˆ ˆ ˆ ÄG = (∂µ φ † )(∂ µ φ) − V (φ). ˆ ˆ ˆ ˆ ˆ ˆ V = VS ≡ 1 λ(φ † φ)2 + µ2 φ † φ 4 with µ2 . V. λ > 0. ˆ Clearly ÄG is invariant under the global U(1) symmetry ˆ ˆ ˆ φ → φ = e−iα φ.69) 1 ˆ ˆ ˆ φ † = √ (φ1 + iφ2 ) 2 (17. In general. non-perturbative. We shall see how this symmetry may be ‘spontaneously broken’. In the quantum theory. For λ = 0. To implement this. It is clear that the absolute minimum of the classical Hamiltonian ÀG is reached for ˙ (i) φ = constant. leading to the usual quantized modes.70) but without the hats and with λ and µ2 both positive. the vacuum of the quantum field theory.71) (17. For V = VS as in (17. The Hamiltonian density is then ˙ ˙ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ÀG = φ † φ + ∇φ † · ∇φ + V (φ). a complex scalar field φ as in section 7.73) (17. the Lagrangian is simply ˆ ˆ ˆ ˆ ˆ Äfree = ∂µ φ † ∂ µ φ − µ2 φ † φ (17.74) Copyright 2004 IOP Publishing Ltd .68) We begin by considering the ‘normal’ case in which the potential has the form (17. We know that everything depends on the nature of the ground state of this field system—that is. the minimum of VS is clearly at φ = 0 and is unique.72) ˆ the generator being Nφ of (7.5 Spontaneously broken global U(1) symmetry: the Goldstone model ˆ We consider. writing as usual ˆ φ= d3 k √ [a(k)e−ik·x + b† (k)eik·x ] ˆ (2π)3 2ω (17. and (ii) φ = φ0 . following Goldstone (1961). where φ0 is the minimum of the classical version of the potential.23). (17. it is a difficult.17. which reduces the φ and ∇φ terms to zero.1. we expect to treat small oscillations of the field about this minimum as approximately ˆ harmonic. problem to find the ground state (or a good approximation to it—witness the superfluid) but we can make some progress by first considering the theory classically.70) where the plane waves are solutions of the ‘free’ (λ = 0) problem.

alternatively.Figure 17.73). 4 (17.80) Copyright 2004 IOP Publishing Ltd . Thus. This time. The classical potential VSB of (17.79) where v= 2|µ| .70) ‘by hand’. consisting of two degrees of freedom. so that the classical potential is now the ‘symmetry breaking’ one: V = VSB ≡ 1 λ(φ † φ)2 − µ2 φ † φ.76) is consistent with the fact that the classical minimum energy configuration has φ = 0.2. when 2 2 φ1 + φ2 = 2µ2 λ (17. on this interpretation.77).2.76) It seems reasonable to interpret quantum field average values as corresponding to classical field values and. ω = (k2 + µ2 )1/2 and the vacuum is defined by ˆ a(k)|0 = b(k)|0 = 0 ˆ and so clearly ˆ 0|φ|0 = 0. each with the same mass µ (see section 7. which represents a complex scalar field. (17. although the origin φ1 = φ2 = 0 is a stationary point. Consider now the case in which the classical minimum is not at φ = 0.77) This is sketched versus φ1 and φ2 in figure 17. in (17. The minimum of VSB occurs when (φ † φ) = or.1). (17.75) (17.78) 4µ2 ≡ v2 λ (17. it is an (unstable) maximum rather than a minimum. This can be achieved by altering the sign of µ2 in (17. λ1/2 (17.

This is the case for T greater than the ferromagnetic transition temperature TC . therefore. but the system must choose one particular direction (e.76). or some other variable.83) is lost. with a minimum at | M| = 0. (17.2 strongly suggests that if we want a sensible quantum interpretation of a theory with the potential (17. It is clear that this is exactly the situation met in the superfluid (but ‘B’ here will stand for ‘broken symmetry’) and is. It may also be the case in particle physics that parameters such as µ2 change sign as a function of T .81). be |0 B such that B 0|φ|0 B does not vanish. The minimum condition (17. In quantum field theory. Before proceeding further. which is the vacuum. If we maintain the idea that the vacuum expectation value of the quantum field should equal the ground-state value of the classical field.g. if one assumes that µ2 (T ) changes sign at TC . becoming negative for T < TC .82) φ(x) = (ρ(x)/ 2) exp(iθ (x)/v) where. However.2.77). we briefly outline a condensed matter analogue of (17. Any direction of M is possible (only | M| is specified). This symmetry has been broken ‘spontaneously’—though this is still only a classical analogue.1) and when it does so the rotational invariance exhibited by F of (17. therefore. and it suggests that we should think of the ‘symmetric’ and ‘broken symmetry’ situations as different phases of the same system. in contrast to (17. Figure 17. thereby effectively precipitating a phase change. the v is inserted so that θ has the same dimension (mass) as ρ and φ . via the influence of a very weak external field. as discussed in section 17. at any value of θ . about which stable Copyright 2004 IOP Publishing Ltd . the condition for the existence of massless (Goldstone) modes.83) valid for weak and slowly varying magnetization.3. If the parameter µ2 is positive. the model is essentially the Landau mean field theory of ferromagnetism. the minimum being at | M| = 0.The condition (17. we had better expand the fields about a point on the circle of minima. moreover.79) can also be written as √ |φ| = v/ 2.77) and (17.81) which may help in understanding the change in sign of the parameter µ2 . represents a possible classical ground state—and it is clear that they are (infinitely) degenerate. the vacuum in this ˆ µ2 < 0 case must. it is helpful to introduce the ‘polar’ variables ρ(x) and θ (x) via √ (17. for convenience. represents the circle ρ = v : any point on this circle. then F will now resemble a vertical section of figure 17. Let us see how they emerge in this model. Nevertheless. Consider the free energy F of a ferromagnet as a function of the magnetization M at temperature T and make an expansion of the form F ≈ F0 (T ) + µ2 (T ) M 2 + λ 4 M + ··· 4 (17.81) To have a clearer picture. particles are thought of as excitations from a ground state. it is clear that F has a simple ‘bowl’ shape as a function of | M|.

27) and (7. rather than about the obviously unstable point φ = 0. We might well guess that ‘radial’ oscillations in ρ would correspond to a conventional massive field ˆ ˆ (having a parabolic restoring potential). then. α B v = e−iα √ 2 −iα ˆ = e B 0|φ|0 (17.89) .ˆ oscillations are likely.88) as in (17. Equation (17.37) and (17. from (17. it has to be noted that the triggering mechanism for the symmetry breaking (µ2 → −µ2 ) has to be put in by hand. θ = 0 in the classical case. we set 1 ˆ ˆ ˆ φ(x) = √ (v + h(x)) exp(−iθ (x)/v) 2 ˆ ˆ ˆ ˆ ˆ ˆ ÄG = 1 ∂µ h∂ µ h − µ2 h 2 + 1 ∂µ θ∂ µ θ + µ4 /λ + · · · 2 2 (17.85).52). say θ = −α. is essentially phenomenological. we may perfectly well choose a vacuum corresponding to a classical ground state with non-zero θ . merely reflects the fact that the minimum value of VSB is −µ4 /λ. which does not affect equations of motion. and with the occurrence of massless modes.87) B (17.86) may be compared with (17. But we know (see (7. contains much of the essence of spontaneous symmetry breaking in field theory: a non-zero vacuum value of a field which is not an invariant under the symmetry group.) The ansatz (17.84) and the non-zero vev (17. α|φ|0. in short. zero mass bosons and massive excitations in a direction in field space which is ‘orthogonal’ to the degenerate ground states. in the superfluid case.28)) that ˆ ˆ ˆ −1 ˆ ˆ e−iα φ = φ = Uα φ Uα Copyright 2004 IOP Publishing Ltd (17. As in the case of the superfluid. This implies.5) that ÄG (with V = VSB of (17.85) shows that the particle spectrum in the ‘spontaneously broken’ case is dramatically different from that in the normal case: instead of two degrees of freedom with the same mass µ. Then B ˆ 0. respectively.85) ˆ ˆ the dots representing interaction terms which are cubic and quartic in θ . while ‘angle’ oscillations in θ —which pass through all the degenerate vacua—have no restoring force and are massless. We expect the vacuum |0 B to be annihilated by the mode operators ah and aθ for these fields. However. Goldstone’s model.57). in contrast to the—admittedly approximate but more ‘dynamical’—Bogoliubov approach. (The constant term in (17. however.86) B 0|φ|0 B = v/ 2 which is consistent with our interpretation of the vacuum expectation value (vev) as the classical minimum. Let us pick the point ρ = v .84) ˆ ˆ that √ ˆ (17. The Goldstone model.84) ˆ ˆ ˆ and find (problem 17.77) with hats on) becomes (17. Accordingly. one (the θ -mode) is massless and the other (the √ h-mode) has a mass of 2µ. h.

94) asserts that. jφ becomes ˆ ˆ ˆ ˆµ ˆ ˆ (17. this time. which.72). α B ˆ ˆ ˆ −1 = B 0|Uα φ Uα |0 B (17. Note also that (17. (17. when the symmetry is spontaneously ˆ ˆ† broken. with x = 0. In words. forms part of the Higgs sector of the Standard Model. We are now ready to generalize the Abelian U(1) model to the (global) nonAbelian case.95) Copyright 2004 IOP Publishing Ltd .90) ˆ 0.3 but of bosons: ˆ φ= ˆ φ+ ˆ φ0 ≡ 1 ˆ √ (φ1 2 1 ˆ √ (φ3 2 ˆ + iφ2 ) ˆ + iφ4 ) (17. in the ‘broken symmetry’ case. ˆµ after (17.94). It is interesting to find out what happens to the symmetry current corresponding to the invariance (17.6 Spontaneously broken global non-Abelian symmetry We can illustrate the essential features by considering a particular example. p = −i p µ v e−i p·x µ (17. This is easily seen by writing the usual normal mode ˆ ˆ expansion for θ and h . as usual. with an amplitude which is proportional to the symmetry breaking vacuum expectation value v . not valid in the infinite volume limit where Nφ fails to exist).23) which we write again here in slightly different notation: ˆµ ˆ ˆ ˆ ˆ jφ = i(φ † ∂ µ φ − (∂ µ φ)† φ) (17. is precisely of the type that was shown to be non-zero in the proof of the Goldstone theorem.92) µ ˆ ˆ ˆ normal ordering being understood. α|φ|0. (17. ˆ The term involving just the single field θ is very remarkable: it tells us that there is a non-zero matrix element of the form B ˆ 0| jφ (x)|θ. α B (this argument ˆ is. and using the standard bosonic commutation relations for aθ (k). We consider an SU(2) doublet but. aθ (k ).67).93) jφ = v∂ µ θ + 2h∂ µ θ + h 2 ∂ µ θ /v. in fact. The matrix element (17. not of fermions as in section 12. 17. as is required for the massless θ . the symmetry current connects the vacuum to a state with one Goldstone quantum. with momentum p µ .where So (17.94) is consistent with ∂µ jφ = 0 only if p 2 = 0. This current is given in (7. Written in terms of the h and θ of (17.84). p stands for the state with one θ -quantum (Goldstone boson).91) ˆ −1 and we may interpret Uα |0 B as the ‘alternative vacuum’ |0.94) where |θ.88) becomes B ˆ ˆ Uα = eiα Nφ .

97) is invariant under global SU(2) transformations ˆ ˆ ˆ (17. As before.98) as a condition on the vev of φ † φ. except that in that case it is a local symmetry.ˆ where the complex scalar field φ + destroys positively charged particles and ˆ creates negatively charged ones and the complex scalar field φ 0 destroys neutral particles and creates neutral anti-particles. we need to define ‘ 0|φ|0 ’ and expand about it. with µ2 positive in both cases. Thus. each with the same mass µ. for the ‘normal’ sign of µ2 . so that the full symmetry is SU(2)×U(1). This U(1) symmetry leads to a conserved quantum number which we call y. Let us see what happens in the broken symmetry case. (17. α2 .98) ˆ ˆ As in the U(1) case. ˆ ˆ 0|φ † φ|0 = v 2 /2 (17. In the present case. however. via Q = e(t3 + y/2) (17. K0 ). the free (λ = 0) part would describe a ˆ ˆ ‘−µ complex doublet. the Lagrangian we shall use has an additional U(1) symmetry. we note that (17.84). As we shall see in a moment.101) where α is to be distinguished from α ≡ (α1 . Before proceeding.97) which has the ‘spontaneous symmetry breaking’ choice of sign for the parameter ˆ ˆ µ2 . That is.97) with µ2 > 0.99) where now |0 is the symmetry-breaking ground state.100) φ → φ = exp(−iα · τ /2)φ but also under a separate global U(1) transformation ˆ ˆ ˆ φ → φ = exp(−iα)φ (17.69) and (17. Plainly. The Lagrangian we choose is a simple generalization of (17. in which ‘+µ2 φ † φ’ is replaced by 2 φ † φ’. the situation is more Copyright 2004 IOP Publishing Ltd . α3 ). as in (17. in order to get a sensible particle spectrum we must expand the ˆ ˆ fields φ not about φ = 0 but about a point satisfying the stable ground state ˆ (vacuum) condition (17. which is the symmetry of the electroweak sector of the Standard Model.98). We associate the physical charge Q with the eigenvalue ˆ t3 of the SU(2) generator t3 .96) so that y(φ + ) = 1 = y(φ 0 ). For the Lagrangian (17. the minimum of the classical potential is at the point (φ † φ)min = 2µ2 /λ ≡ v 2 /2. The full symmetry is then referred to as SU(2)×U(1). and with y. φ + and φ 0 can be thought of as analogous to the hadronic iso-doublet (K+ . and the subscript ‘B’ is omitted.77): ˆ Ä ˆ ˆ ˆ ˆ = (∂µ φ † )(∂ µ φ) + µ2 φ † φ − λ † 2 ˆ ˆ (φ φ) 4 (17. with four degrees of freedom. we interpret (17.

103) In considering what symmetries are respected or broken by (17. Copyright 2004 IOP Publishing Ltd 0 √ v/ 2 = 0 0 (17. irrespective of where that points.106) We say that ‘the vacuum is invariant under (17. which would therefore not be ‘spontaneously broken’ after all. it is easiest to look at infinitesimal transformations.85) might be.98) becomes ˆ2 ˆ2 ˆ2 ˆ2 0|φ1 + φ2 + φ3 + φ4 |0 = v 2 . φ. Suppose we were considering just SU(2) and the field ‘φ’ was ˆ ˆ an SU(2)-triplet.103) since (1 + τ3 ) so that 0|φ|0 = 0|φ + δφ|0 . (17.103). Just this happens (by design) in the present case. an arbitrary 0|φ|0 can be brought to the form ˆ 0|φ|0 = 0 √ v/ 2 . ˆ Suppose.104) (which is a combination of (17.105) . not four.95).complicated than (17. (17.100)) is still a symmetry of (17. we would then get fewer of these bosons than expected.104)’ and when we look at the spectrum of oscillations about that vacuum we expect to find only three massless bosons.95) contains four real fields as indicated in (17. by using the freedom of global SU(2)×U(1) ˆ phase changes. This would effectively mean that this particular choice of the vacuum state respected that subset of symmetries.102) holds. however. In the present case. and it is not at first obvious what an appropriate generalization of (17.84) since the complex doublet (17.84) and (17. Furthermore. then. but this form is invariant under rotations about the n-axis. Since each broken symmetry is associated with a massless Goldstone boson. that we could choose the 0|φi |0 so as to break this SU(2)×U(1) symmetry completely: we would then expect four massless fields. and (17. It is then clear that the particular transformation ˆ ˆ δ φ = −i (1 + τ3 )φ (17. in this more complicated (non-Abelian) situation a qualitatively new feature can arise: it may happen that the chosen condition ˆ 0|φi |0 = 0 is invariant under some subset of the allowed symmetry transformations.102) ˆ It is evident that we have a lot of freedom in choosing the 0|φi |0 so that (17. An analogy may make ˆ this point clearer.101) and the ‘third component’ of (17. Then we could always write 0|φ|0 = vn where n is a unit vector. (17. Actually. it is not possible to make such a choice.

in the ‘spontaneously broken’ case. which is easily done by replacing the ordinary derivatives in (17. we find that no mass term is generated for the θ fields. made local). no such doublet structure is seen: instead.Oscillations about (17. but the physical spectrum is completely different.107).103) are conveniently parametrized by ˆ ˆ φ = exp(−i(θ(x) · τ /2)v) 1 √ (v 2 0 ˆ + H (x)) (17. Again. there is one massive scalar field and three massless scalar fields.6). we work at zero temperature.7 The BCS superconducting ground state We shall not attempt to provide a self-contained treatment of the Bardeen– Cooper–Schrieffer (1957)—or BCS—theory.97) (see problem 17.97) by suitable covariant ones. Let us now note carefully that whereas in the ‘normal symmetry’ case with the opposite sign for the µ2 term in (17.108) 2 √ just as in (17. in the following chapter we shall see how Nambu (1960) interpreted a chiral symmetry breaking fermionic mass term as an analogous ‘gap’. the SU(2)×U(1) symmetry will be ‘gauged’ (i. as a preliminary to our discussion of chiral symmetry breaking in the following chapter. We may summarize this (anticipated) result by saying. with the choice (17. The number of degrees of freedom is the same in each case.85). leaving that topic for chapter 19.e. those gauge fields corresponding to symmetries that are broken by the ˆ choice of 0|φ|0 acquire a mass. rather. 17. then. while the H field piece is ˆ ˆ ˆ = 1 ∂µ H ∂ µ H − µ2 H 2 + interactions (17. In the application of this to the electroweak sector of the Standard Model. Inserting (17. while those that correspond to symmetries that ˆ are respected by 0|φ|0 do not. showing that m H = 2µ. We emphasize at the outset that we shall here not treat electromagnetic interactions in the superconducting state. the free-particle spectrum consisted of a degenerate doublet of four degrees of freedom all with the same mass µ.107) into (17.84). Copyright 2004 IOP Publishing Ltd ˆ ÄH . we wish simply to focus on one aspect of the theory.97). The existence of such a gap is a fundamental ingredient of the theory of superconductivity. will be to end up with three massive gauge fields (those mediating the weak interactions) and one massless gauge field (the photon). Exactly how this happens will be the subject of chapter 19. namely the occurrence of an energy gap separating the ground state from the lowest excited levels of the energy spectrum. that when a spontaneously broken non-Abelian symmetry is gauged. We end this chapter by considering a second important example of spontaneous symmetry breaking in condensed matter physics. We shall see in chapter 19 that the result.107) which is to be compared with (17.

in the non-interacting (V = 0) part. Cooper (1956) was the first to observe that the Fermi ‘sea’ was unstable with respect to the formation of bound pairs. At this point we recall the characteristic feature of a weakly interacting gas of electrons at zero temperature: thanks to the Exclusion Principle.k c† c−k c−k ck ˆk ˆ† ˆ ˆ (17. This instability modifies the Fermi sea in a fundamental way: a sort of ‘condensate’ of pairs is created around the Fermi energy and we need a many-body formalism to handle the situation. in the presence of an attractive interaction. mediated by phonon exchange. the c’s are fermionic ˆ operators obeying the usual anti-commutation relations and the vacuum is such that ck |0 = 0. we thus arrive at the BCS reduced Hamiltonian ˆ HBCS = k ˆ† ˆ k ck ck −V k. we shall be dealing with electrons— which are fermions—rather than the bosons of a superfluid. we shall see that a similar kind of ‘condensation’ occurs in the superconductor too. Note that. Naturally. With this by way of motivation. such a phenomenon can only occur for bosons. It turns out (see. an essential element in the BCS theory is the identification of a mechanism whereby pairs of electrons become correlated. Thus. But the positively charged ions do provide sources of attraction for the electrons. and the ˆ single-particle energies k are measured relative to E F .2. so their total momentum is zero. It can also be argued that the effective attraction will be greater when the spins are anti-parallel but the spin will not be indicated explicitly in what follows: ‘k’ will stand for ‘k with spin up’ and ‘−k’ for ‘−k with spin down’. where ωD is the Debye frequency associated with lattice vibrations. the electrons populate singleparticle energy levels up to some maximum energy E F (the Fermi energy). In the present case.109) which is the starting point of our discussion. Kittel (1987) chapter 8) that electron–electron scattering.Our discussion will deliberately have some similarity to that of section 17. we shall consider pairs of equal and opposite momentum k. Now the Coulomb interaction between a pair of electrons is repulsive and it remains so despite the screening that occurs in a solid. which then become bound to a state with a total energy less than 2E F . The sum is over states lying near E F . c† ck is the number ˆk ˆ operator for the electrons. whose value is fixed by the electron density. which because of the Pauli principle has eigenvalues 0 Copyright 2004 IOP Publishing Ltd . In (17. The constant V (with the minus sign in front) represents a simplified form of the effective electron–electron attraction. What this means is that the energy of the system can be lowered by exciting a pair of electrons above E F .109). For simplicity. Nevertheless. the behaviour of which may have some similarity to that of bosons. leads to an effective attraction between two electrons whose energies k lie in a thin band E F − ωD < k < E F + ωD around E F . and may be used as intermediaries (via ‘electron–phonon interactions’) to promote an effective attraction between electrons in certain circumstances.3. of course. as before. for example.

111) only make sense in the context of a number-non-conserving ground state. (17.38). just as for the superfluid. analogous to (17. in terms of the ˆ ˆ mode operators β and β † : ˆ HBCS = k ˆ† ˆ ˆ† ˆ ωk (βk βk + β−k β−k ) + γ (17.110). completely analogous to (7. We seek a diagonalization of (17. This last condition implies (problem 17. It is easy to check (problem 17. the transformations (17. since they do not respect the ˆ symmetry (17. We follow Bogoliubov (1958). we shall see that the BCS ground state does not respect the symmetry. of course.109).115) then implies that the ωk are the energies of states created by ˆ† ˆ† the quasiparticle operators βk and β−k .50) and sums the singleparticle energies k for each occupied level. (17. which is equivalent to ψ ( x) = e−iα ψ(x) for the electron field operator ˆ at x . depend only on k = |k| and are chosen so as to preserve anti-commutation relations for the β ’s.40). Although HBCS of (17. However.110) ˆ ˆ for all k .116) . Bogoliubov et al (1959) (see also Valatin (1958)) and make a canonical transformation on the operators ck .7) 2 u 2 + vk = 1 k (17.41). We set ˆ βk = u k ck − vk c−k ˆ† ˆ ˆ ˆk β−k = u k c−k + vk c† † βk = u k c† − vk c−k ˆk ˆ † β−k = u k c−k + vk ck ˆ† ˆ (17.or 1: this term is.112) so that we may conveniently set u k = cos θk vk = sin θk . βl† ] = ωl βl† as in (17. Thus fermion number is conserved by HBCS .8) that the form (17.113) Just as in the superfluid case.111) where u k and vk are real. the ground state being defined by ˆ βk |ground Copyright 2004 IOP Publishing Ltd BCS ˆ = β−k |ground BCS = 0. ˆ We immediately note that HBCS is invariant under the global U(1) transformation ck → ck = e−iα ck ˆ ˆ ˆ (17. c−k similar to ˆ ˆ† the one employed for the superfluid problem in (17. as motivated by the ‘pair condensate’ picture. despite the fact that the operators obey anticommution relations.114) for certain constants ωk and γ .114) implies ˆ ˆ ˆ (17. Equation (17. we shall shortly make a crucial number-non-conserving approximation.109) is number-conserving.115) [ HBCS .

in fact. cos θl cl† − sin θl c−l ] = ωl (cos θl cl† − sin θl c−l ) ˆ ˆ ˆ ˆ (17. Since these operators carry fermion number ±2. we may consider the excitation of two particles above a BCS state with N − 1 pairs.119) −V cos θl k c† c−k + (ωl + l ) sin θl = 0. ˆk ˆ † It is at this point that we make the crucial ‘condensate’ assumption: we ˆ ˆ ˆ† ˆ† replace the operator expressions k c−k c k and k c k c−k by their average values. Such an excited state is represented by βk β−k |ground BCS . we therefore require ˆ [ HBCS . Evaluating (17. which are assumed to be non-zero in the ground state. Recalling that l is measured relative to E F .9) (ωl − l ) cos θl − V sin θl k c−k ck = 0 ˆ ˆ (17.111). we see that it implies that all excited states are separated from E F by a finite amount. It is this quantity which is usually called the † † energy gap.125) Copyright 2004 IOP Publishing Ltd . it is clear that this assumption is only valid if the ground state does not.122) (17.120) . We accordingly make the replacements V k c−k ck → V ˆ ˆ c† c−k → V ˆ k ˆ† k BCS ground| k c−k ck |ground ˆ ˆ c† c−k |ground ˆ k ˆ† k BCS ≡ ∗ (17.122) and taking now to be real and equal to | |. (17. We shall need the expressions for cos θl and sin θl which may be obtained as follows.117) which must hold as an identity in the cl ’s and cl† ’s. In interpreting (17. Squaring (17.118) and (17.121) V BCS ground| BCS ≡ In that case.ˆ Substituting for βl† in (17. we must however be careful to reckon energies for an excited state as relative to a BCS state having the same number of pairs. one ˆ ˆ obtains (problem 17.124) ωl sin θl = − sin θl + cos θl + | |2 ]1/2 .124) is the fundamental result at this stage.123) (17. have a definitive number of particles—just as in the superfluid case. we obtain | |2 (cos2 θl − sin2 θl ) = 2 l | | cos θl sin θl (17.124).119) become ωl cos θl = which are consistent if ωl = ±[ 2 l l cos θl + l sin θl ∗ (17. if we consider experimental probes which do not inject or remove electrons. equations (17. considering a component of |ground BCS with N pairs.115) from (17. Thus. Equation (17. namely | |.117).118) (17. The minimum energy for this to be possible is 2| |.

| |=V BCS (17. The condition (17. so that when electromagnetic interactions are included in the superconducting state. not observed: instead.120) and (17. Remarkably. An explicit formula for can be found by using the definition (17. Replacing the sum by Copyright 2004 IOP Publishing Ltd .111): ˆ ˆ ˆ† ck = cos θk βk + sin θk β−k .130) The sum in (17. in this case. However.129).which leads to tan 2θl = | |/ and then cos θl = 1 l 1+ 2 ωl 1/2 l (17.127) All our experience to date indicates that the choice ‘ of phase for the ground-state value: V BCS = real’ amounts to a choice = | |.126) sin θl = 1 l 1− 2 ωl 1/2 .110).130) is only over the small band E F − ωD < k < E F + ωD over which the effective electron–electron attraction operates. the by now anticipated form for a spontaneously broken U(1) symmetry. this is the physics of the Meissner effect in a superconductor. the massless photon field will enter. other phases for are equally possible.120). rendering it massive. ground| k c−k ck |ground ˆ BCS (17. using (17. of course.128) has.110) to become a local function of x. we have to allow the α in (17. together with the expression for ck found by inverting (17. As we shall see. (17.128) By making use of the U(1) symmetry (17.129) ground| k ˆ ˆ† (cos θk β−k + sin θk βk ) BCS BCS ˆ ˆ† × (cos θk βk + sin θk β−k )|ground =V =V k BCS ground| k ˆ ˆ† cos θk sin θk β−k β−k |ground . At the same time. that degree of freedom is incorporated into the gauge field. we shall learn in chapter 19 that the expected massless (Goldstone) mode is. we may now recall that the electrons are charged. and that of the ‘Higgs mechanism’ in the Standard Model. ˆ This gives. and we must therefore expect the occurrence of a massless mode. | | 2[ 2 k + | |2 ]1/2 (17.

and compare problem 17. This is the celebrated BCS solution for the gap parameter | |. Equation (17.133) ˆ The operator UBCS is (Blatt 1964. and so no perturbative treatment starting from a normal ground state could reach this result. The explicit form of the ground state in this model can be found by a method similar to the one indicated in section 17.132) is in reasonably good agreement with experiment.4) ˆ exp[θk (c† c−k − ck c−k )]. and may be refined. Perhaps the most significant thing to note about it.138) .135) ˆ = UBCS |0 (17. (17.48)) ˆ ˆ UBCS ck UBCS = βk ˆ ˆ† ˆ† ˆ ˆ† ˆ † UBCS c−k UBCS = β−k . When the exponential in UBCS is expanded out and applied to the vacuum state |0 .131) yields | |= ωD ≈ 2ωD e−1/ V NF sinh(1/ V NF ) (17. ˆ ˆk ˆ† ˆ ˆ We have ˆk ˆ† ˆ ˆ ˆ ˆ ˆk ˆ† sk = −c† c−k ck c−k − ck c−k c† c−k ˆ2 Copyright 2004 IOP Publishing Ltd (17. great simplifications occur.4. we have ˆ ˆ† ˆ ˆ UBCS βk UBCS |0 = 0 showing that we may identify |ground BCS (17.136) ˆ via the condition (17.137) (17. we obtain the gap equation 1= 1 V · NF 2 ωD −ωD d [ 2 + | |2 ] 2 (17. section V. The estimate (17. for our purpose. since ck |0 = 0.2 for the superfluid. Since ˆ the transformation from the c’s to the β ’s is canonical. there must exist a unitary ˆ operator which effects it via (compare ( integral. Yosida 1958. ˆk ˆ † ˆ ˆ (17. is that the expression for | | is not an analytic function of the dimensionless interaction parameter V NF (it cannot be expanded as a power series in this quantity).132) for V NF 1.3.134) UBCS = k Then. Consider the operator sk = c† c−k − ck c−k .131) 1 = V NF sinh−1 (ωD /| |) where NF is the density of states at the Fermi level.116).

Copyright 2004 IOP Publishing Ltd .1 Verify (17.141) where a0 is the Bohr radius.44).140) = k (cos θk + sin θk c† c−k )|0 . 17. ˆk ˆ † As for the superfluid. If we compare E b to the Coulomb repulsion at a distance ξ . is of order 10−3 . We should emphasize that this is only the barest outline of a simple version of BCS theory. and as many as 106 pairs may have their centres of mass within one coherence length of each other. one finds (see. which to calculate one needs to evaluate the constant γ in (17. It follows that ˆ2 ˆ exp(θk sk )|0 = 1 + θk s k − ˆ 2 θ3 θk − k sk . Hence. Nevertheless.2 Verify (17. 17. the pairs are not really bound.3 Derive (17. from which many subtleties have been omitted. Problems 17. we find that E b /(α/ξ ) ∼ a0 /ξ (17. One can also calculate the approximate spatial extension of a pair. Consider.114). 17.141). θ are real parameters. Enz 1992) that E b ≈ 3 2 /E F . which is denoted by the coherence length ξ and is of order vF /π where kF = mvF is the Fermi momentum.35). .so that sk |0 = −|0 . in conventional superconductors. To a good approximation.43) and (17. for example. a † ] = 1 and λ. ˆ ˆ ˆ (a) Show that Uλ is unitary. only correlated.139) (17. the simple theory presented here contains the essential features which underly all attempts to understand the dynamical occurrence of spontaneous symmetry breaking in fermionic systems. . Numerical values show that the right-hand side of (17.29). for example.140) represents a coherent superposition of correlated pairs. with no restraint on the particle number. (17. |0 ˆ 2 3 ˆ = (cos θk + sin θk sk )|0 ˆk ˆ † = (cos θk + sin θk c† c−k )|0 and hence |ground BCS (17. the binding energy E b of a pair. We now proceed to an important application in particle physics.4 Let ˆ Uλ = exp[ 1 λθ (a 2 − a †2] ˆ ˆ 2 where [a.

and that d2 Iˆλ ˆ = θ 2 Iλ .77) ˆ ˆ and show that the result for the constant term and the quadratic terms in h and θ .85). the terms quadratic in the ˆ ˆ ˆ ˆ fields H and θ reveal that θ is a massless field.7 Verify that the β’s of (17. ˆ 17.5 Insert the ansatz (17.118) and (17.(b) Let Show that ˆ ˆ ˆ −1 ˆ I λ = Uλ a Uλ and ˆ d Iλ ˆ = θ Jλ dλ ˆ ˆ ˆ ˆ −1 Jλ = Uλ a †Uλ . 17. 2 ˆ ˆ ˆ ˆ 17.69) with V = VSB of (17.48)) that and where ˆ ˆ ˆ −1 ˆ ˆ ˆ U1 a †U1 = sinh θ a + cosh θ a † ≡ α † ˆ ˆ ˆ ˆ U1 ≡ Uλ=1 = exp[ 1 θ (a 2 − a †2)].9 Derive (17.8 Verify (17.119).112) holds.115).84) for φ into ÄG of (17. Copyright 2004 IOP Publishing Ltd .38) and (17.107) is inserted in (17.6 Verify that when (17. is as given in (17.97). show that and thus finally (compare (17.111) satisfy the required anti-commutation relations if (17. 17. while the quanta of the H field √ have mass 2µ. 17. dλ2 ˆ ˆ ˆ Iλ = cosh(λθ ) a + sinh(λθ ) a † ˆ ˆ ˆ −1 U1 a U1 = cosh θ a + sinh θ a † ≡ α ˆ ˆ ˆ (c) Hence.

an immediate physical consequence should be the appearance of massless (Goldstone) bosons. it would perhaps hardly be sufficient grounds for accepting the hypothesis. no such massless pseudoscalar particles are observed. If this is so. one for every symmetry not respected by the vacuum. that this chiral symmetry is spontaneously broken as a dynamical effect—presumably. 1 ˆ( ) ˜ (18. Indeed. m d → 0. as discussed in section 18. as we saw.e.1) T 2 |d = |u +5 we now interpret the state |u (which is degenerate with |d ) as |d + ‘π + ’ where ˜ ‘π + ’ is a massless particle of positive charge but a pseudoscalar (0− ) rather than a scalar (0+ ) since. but it is natural to hope that when the small up and down quark masses are included. the real pions (π + . We can now resolve this puzzle by making the hypothesis (section 18. the Goldstone state(s)] in the form of the pion’.166)) happen to be the very same currents which enter into strangeness-conserving semileptonic weak interactions (such as Copyright 2004 IOP Publishing Ltd . which differ only in q¯ spin alignment from the pseudoscalar (0− ) octet.1) first articulated by Nambu (1960) and Nambu and Jona-Lasinio (1961a). In the same way. one of the simplest consequences of such a symmetry should be the existence of nucleon parity doublets. which are not observed. As Nambu and Jonaq Lasinio (1961a) stated. This is indeed how they do appear.2. as we showed.109) and (12. the vector and axial vector 1 1 ˆ ( )µ and T ( 2 )µ of the u–d strong interaction SU(2) ˆ symmetry currents T 2 5 symmetries (see (12. as a property of the QCD interactions. returning to (12.3. If this was the only observable consequence of spontaneously breaking chiral symmetry.18 CHIRAL SYMMETRY BREAKING In section 12.169) which we repeat here for convenience. interacting via the colour gauge fields of QCD. ‘It is perhaps not a coincidence that there exists such an entity [i. should exhibit signs of the non-Abelian chiral symmetry SU(2)f 5 . ˜ 1 1 ˆ(2) ˆ (2) ‘π − ’ and ‘π 0 ’ will be associated with T−5 and T3 5 .2 we arrived at a puzzle: there seemed good reason to think that a world consisting of u and d quarks and their anti-particles. π 0 ) will emerge as ‘anomalously light’. from today’s perspective. First. particularly with respect to the octet of vector (1− ) mesons. rather than strictly massless. But. π − . But there are two circumstances which greatly increase the phenomenological implications of the idea. |u has opposite parity to |u . which was exact in the massless limit m u . Of course.

5. indeed. which provide a very successful description of the low-energy dynamics of the hadronic degrees of freedom. If vanishes. that chiral symmetry breaking is an important part of the Standard Model.5).1 The Nambu analogy We recall from section 12. It would take us too far from our main focus on gauge theories to pursue these interesting avenues in detail. c−l . Thus.3) and the Adler–Weisberger (in Adler 1965) relation.123) which describe the elementary excitations in a superconductor (in the case is real) and which we repeat here for convenience: ωl cos θl = l cos θl + l sin θl + sin θl cos θl . The result of this spontaneous breaking of the U(1) symmetry is the creation of the gap (or 2 for a number-conserving excitation) and the appearance of a massless mode.4) (18. some ¯ ¯ remarkable connections between weak. and their interactions with other hadrons such as nucleons. Consider Copyright 2004 IOP Publishing Ltd .n → pe− νe and π − → µ− νµ ).3.2) (18. in terms of which the Dirac equation reads Eφ = σ · pφ + mχ Eχ = − σ · pχ + mφ. and to encourage further study of a subject which may at first sight appear somewhat peripheral to the Standard Model as conventionally understood.4) have been developed. cos θl and sin θl are. (18. (17. respectively. Second.4) and (18. as we shall see in chapter 20. it turns out that the dynamics of the Goldstone modes.97) for the Dirac matrices.and strong-interaction parameters can be established. 18. such as the Goldberger–Treiman (1958) relation (see section 18.6) ˆ The superposition in βl combines operators which transform differently under the U(1) (number) symmetry. we shall introduce the subject of chiral anomalies in section 18.5) ωl sin θl = − In (18. in this chapter. are strongly constrained by the underlying chiral symmetry of QCD.126) implies that ˆ ˆ† θl = 0 and we revert to the symmetry-respecting operators cl . surprisingly detailed effective theories (see section 18.3) and equations (17. (18. But we hope to convince the reader.122) and (17. the components of the ˆ† electron destruction operator cl and the electron creation operator c−l in the ˆ ˆ quasiparticle operator βl (see (17. Finally.3) Nambu (1960) and Nambu and Jona-Lasinio (1961a) pointed out a remarkable analogy between (18.2 that for ‘almost massless’ fermions it is natural to use the representation (4.111)): ˆ ˆ ˆ† βl = cos θl cl − sin θl c−l .2) and (18. (18.

Since the vacuum carries neither linear nor angular momentum. Here φ and χ are the components of definite chirality in the Dirac spinor ω (compare (12. the mass of a Dirac particle arises from the (presumed) spontaneous breaking of a chiral (or γ5 ) symmetry. Copyright 2004 IOP Publishing Ltd (18. the particle symmetry we want to interpret this way is SU(2)f 5 not U(1)5. According to Nambu’s analogy. the vacuum for a massive Dirac particle is to be pictured as a condensate of correlated pairs of massive fermions. First. and this will be accompanied by a massless boson.148)) that they have opposite chirality.3). since the vacuum does not violate fermion number conservation. In short.7) (18.140) for the BCS ground state.2) and (18. However. we must again note that the BCS electrons are charged. This means (recalling the discussion after (12. the members of a pair must have equal and opposite spin: they therefore have the same helicity.2) and (18. so that in the real superconducting case we are dealing with a spontaneously broken local U(1) symmetry. As usual. Second. analogous to (17. By contrast. then Eφ+ = | p|φ+ + mχ+ Eχ+ = − | p|χ+ + mφ+ . one has to be a fermion and the other an anti-fermion.1.1.3). The type of fermion–anti-fermion in the Nambu ‘condensate’. separate ones for the chirality eigenstates φR ≡ 0 χ Nambu therefore made the following analogy: superconducting gap parameter ↔ Dirac mass m quasiparticle excitation ↔ massive Dirac particle U(1) number symmetry ↔ U(1)5 chirality symmetry Goldstone mode ↔ massless boson. which is itself not a chirality eigenstate when m = 0.150)). the Dirac equation for ω decouples into two φ 0 and φL ≡ . We may easily write down an expression for the Nambu vacuum. Before proceeding. now (18. so the appropriate generalization (Nambu and Jona-Lasinio 1961b) has to be understood.Figure 18. the SU(2)f 5 chiral symmetry is not gauged.8) . therefore. we should note that there are features of the analogy on both sides which need qualification. Thus. Consider solutions φ+ and χ+ of positive helicity in (18. a typical pair in the Nambu vacuum is as shown in figure 18. not a global one. the quantum field theory vacuum is analogous to the many-body ground state. When m vanishes.

Thus. Hence cs ds acquires a factor e2iβs . the nucleon mass could not be dynamically predicted.111). with ‘m’ the (spontaneously generated) nucleon mass. the fermion under discussion was taken to be the nucleon.s ˆ† (cos θ p − sin θ p cs ( p)ds (− p))|0 ˆ† m=0 (18. Thus.127)) cos θ p = sin θ p = | p| 1 1+ 2 E | p| 1 1− 2 E 1/2 (18. since under ψ = e−iβγ5 ψ we find that ¯ † γ 0 ψ → ψ † eiβγ5 γ 0 e−iβγ5 ψ = ψe−2iβγ5 ψ. the Goldberger–Treiman (1958) relation was derived and a number of other applications were suggested. under a chiral transformation ψ = e−iβγ5 ψ. Thus. unlike the analogous gap parameter in BCS theory. Subsequently.11) carries ±2 units of chirality. As we have seen in volume 1. The fermion–fermion interaction—necessarily invariant under chiral transformations—was taken to be of the four-fermion type. This work was generalized to the SU(2)f 5 case by Nambu and Jona-Lasinio (1961b).10) where E = (m 2 + p2 )1/2 . in analogy with (17. we can read off the mixing coefficients cos θ p and sin θ p as (cf (17. who showed that if the chiral symmetry was broken explicitly by the introduction of a small nucleon mass (∼ 5 MeV). for a given s.Comparing (18. In addition. Nevertheless.11) ˆ† where cs ’s and ds ’s are the operators in massless Dirac fields. In the original conception by Nambu and co-workers. Depending on ˆ† the sign of the helicity s.131) could be formulated. somewhat analogous to the Fermi energy E F .5). this is actually a non-renormalizable theory but a physical cut-off was employed.8) with (18. while ˆ† ds (− p) operators accompany v-spinors for which the γ5 eigenvalue equals minus ˆ ˆ ˆ the helicity. Nambu with other collaborators 1 A different phase convention is used for d † (− p) as compared to that for c† in (17. a massless bound state automatically appeared in the ¯ channel (Nambu ff and Jona-Lasinio 1961a). each pair in (18. a gap equation similar to (17. The Nambu vacuum is then given by1 |0 N = p.120). ˆs ˆ−k Copyright 2004 IOP Publishing Ltd . a ˆ ˆ ˆ ˆ ˆ ˆ ψ ¯ ˆ ˆ Dirac mass is associated with a non-zero value for N 0|ψ ψ|0 N .9) 1/2 (18. cs ( p) operators go with u-spinors for which the γ5 eigenvalue equals the helicity. cs → e−iβs cs ˆ ˆ † † ˆ† ˆs → eiβs ds . A ¯ ˆ ˆ ˆ ˆ Dirac mass term ψ ψ is of just this kind. and it was possible to show that when it had a non-trivial solution. We ˆ ˆ may check this by noting that in the mode expansion of the Dirac field ψ. then the Goldstone pions would have their observed non-zero (but small) mass.4) and (18. Thus ˆ† and d ˆ the Nambu vacuum does not have a definite chirality and operators carrying non-zero chirality can have non-vanishing vacuum expectation values (vevs).7) and (18.

Weisberger 1965) which involves two soft pions.1 Two flavour QCD and SU(2)f L×SU(2)f R Let us begin with the massless case.g) degree of freedom of the u and d quarks. Let us therefore now turn to two-flavour QCD. In this case. which conserve chirality as noted in section 12.18) Copyright 2004 IOP Publishing Ltd . low momentum) pion could be calculated.15) (18.17) (18. of course. for various processes.13) and the λ matrices act on the colour (r.16) (18. (18. QCD became widely accepted as that theory in around 1973. the ‘fermions in question’ are quarks and the interactions between them are gluon exchanges.(Nambu and Lurie 1962.b. ˆ ˆ D µ = ∂ µ + igs λ/2 · Aµ (18. This work was all done in the absence of an agreed theory of the strong interactions (the N–J-L theory was an illustrative working model of dynamicallygenerated spontaneous symmetry breaking. while a small explicit quark mass term in the Lagrangian is held to be responsible for the non-zero pion mass. The bulk of the quark masses inside bound states forming hadrons is then interpreted as being spontaneously generated.14) (18.12) ˆ where u and d now stand for the field operators.3.2. for which the fermionic part of the Lagrangian is ¯ ˆˆ ¯ ˆ ˆ ˆ ˆ / Äq = u i D u + d i D d ˆ / (18. Nambu and Schrauner 1962) showed how the amplitudes for the emission of a single ‘soft’ (nearly massless. but not a complete theory of strong interactions). 18. In addition to the local SU(3)c symmetry.1. These developments culminated in the Adler– Weisberger relation (Adler 1965. this Lagrangian is invariant under (i) U(1)f ‘quark number’ transformations ˆ q → e−iα q ˆ (ii) SU(2)f ‘flavour isospin’ transformations q → exp(−iα · τ /2) q ˆ ˆ (iii) U(1)f 5 ‘axial quark number’ transformations ˆ q → e−iβγ5 q ˆ (iv) SU(2)f 5 ‘axial flavour isospin’ transformations ˆ q → exp(−iβ · τ /2γ5 ) q ˆ where q= ˆ u ˆ ˆ d .

3.21) i = 1. such as must be present if the symmetry is spontaneously broken. however.22) ˆ 2 previously denoted by Ti. there are not any near-massless isoscalar 0− mesons.101)) ˆ Qi = 1 q† ˆ τi 3 qd x ˆ 2 (18.20) The corresponding ‘charges’ are (compare (12. The symmetry currents associated with (iv) are those already given in (12.5 = qγ µ γ5 q ˆ ˆ 2 Similarly. Q i. Finally. it is interesting to discover ˆ ˆ the algebra of the generators. the currents associated with (ii) are ˆ µ ¯ τi ˆ ˆ ji = qγ µ q 2 i = 1. The η meson is an isoscalar 0− meson. (18. 2. (18. explicitly broken by the electromagnetic interactions (and by the difference in the masses m u and m d . so it need not concern us further. which includes the s quark (see. Weinberg 1996.5).5 d3 x = q † γ5 ˆ τi 3 qd x ˆ 2 (18.23) ˆ( ) previously denoted by Ti 2 .19) but there is still a missing lig√ isoscalar 0− meson. Weinberg (1996.5 and (compare (12. for example. As with all symmetries. for example. symmetry (iii) becomes extended to u → e−iβγ5 u ˆ ˆ ˆ ˆ d → e−iβγ5 d ˆ s → e−iβγ5 s ˆ (18. Symmetry (ii) is the standard isospin symmetry of chapter 12.166). This is the well-known ‘U(1) problem’: it was resolved by ’t Hooft (1976a.5 in this case. either. 3. but no such particle exists. Copyright 2004 IOP Publishing Ltd . It can be shown that its mass ht must be less than or equal to 3 m π (Weinberg 1975). but we give them again here in a slightly different notation which will be similar to the one used for weak interactions: τi ¯ ˆµ ji. section 23. symmetry (iv) is the one with which we are presently concerned. In that case. it can be understood as one of the Goldstone bosons associated with the spontaneous breaking of the larger group SU(3)f 5.Symmetry (i) is unbroken and its associated ‘charge’ operator (the quark number operator) commutes with all other symmetry operators. 2. 1986). In fact.10). by showing that the inclusion of instanton configurations (Belavin et al 1975) in path integrals leads to violations of symmetry (iii)—see. when included).167)) ˆ Q i. which are the six charges Q i .5 ≡ (1) ˆ0 ji. section 19. Symmetry (iii) does not correspond to any known conservation law. but with a mass of 547 MeV it is considerably heavier than the pion.

1) ˆ ˆ ˆ [ Qi .30) ˆ ˆ ˆ Q i.26) ˆ Relation (18.L ] = 0.21) can be written as ˆµ ˆµ ˆµ ji = ji.R .L therefore behave like two commuting (independent) angular momentum operators.24) has been seen before in (12.Patient work with the anti-commutation relations for the operators in q(x) and ˆ q † (x) gives the results (problem 18. qL = ((1 + γ5 )/2)q.2) ˆ ˆ [ Q i.3. the symmetry group of the combined symmetries (ii) and (iv) is called SU(2)f L × SU(2)f R .L where τi ¯ ˆ ˆ ˆ ji.L = q L γ µ qL . (18.2.L .33) (18. 2 (18.L introduced in section 12.5 ) 2 (18.L ] = i ˆ i j k Q k. For this reason. Q j. But ((1 ± γ5 )/2) are just the projection operators PR. Q i. which project out the chiral parts of any fermion 2 2 ˆ field.5 .R + ji.R = qR ˆ† τi qR d3 x ˆ 2 ˆ Q i.29) (18.26) to something much simpler.28) (18.R ] = i ˆ ˆ [ Q i.R . Referring to (18.R . Q j ] = i ˆ ˆ [ Q i .L can also be written as ˆ Q i. it is easy to see that PR = PR and PL = PL .L = 1 ( Q i − Q i. Q j. In a similar way. ˆ ˆ The operators Q i.34) Copyright 2004 IOP Publishing Ltd .103) and simply states that the Q i ’s obey an SU(2) algebra.23). Q j. we see that ˆ Q i.25) (18.32) where qR = ((1 − γ5 )/2)q. Q j.L (18.5 ] = i ˆ ˆ [ Q i.L .R = 1 ( Q i + Q i. Defining ˆ ˆ ˆ Q i. Furthermore.27) has a simple interpretation.R and ˆ Q i.L = qL ˆ† τi qL d3 x ˆ 2 (18. The decoupling effected by (18.5 = ji.L τi ¯ ˆµ ˆ ˆ ji.R = q† ˆ 1 + γ5 2 τi 3 qd x ˆ 2 (18.27) ˆ ˆ [ Q i.R = q R γ µ qR 2 ˆµ ˆµ ˆµ ji.5 ˆ i j k Qk . Q j.R ˆ i j k Q k.5 ] = i ˆ i j k Qk ˆ i j k Q k. each obeying the algebra of SU(2).R − ji.24)–(18.31) ˆ and similarly for Q i. so that Q i.20) and (18.5 ) 2 we find (problem 18. A simple trick reduces the rather complicated algebra of (18. the currents ˆ ˆ ˆ (18.22) and (18.24) (18.

ˆ + η and δ = − η.36) by γ5 and adding the result to (18.5). which is why it is called chiral symmetry.36) (18. Q j. ˆ ˆ ˆ Multiplying (18.35) ˆ which is plainly invariant under (18. There is as yet no formal proof that this SU(2)L ×SU(2)R chiral symmetry is spontaneously broken in QCD. Under an infinitesimal SU(2) isospin transformation.132)). Even granted that chiral symmetry is spontaneously broken in massless twoflavour QCD.37) and (18.R .41) (18.12) is chirally ˆ invariant: problem 18. since D is flavour-blind.35).2 which is why [ Q i. Numerical solutions of QCD on a lattice (see chapter 16) do provide strong evidence that quarks acquire significant dynamical (SU(2)f 5breaking) mass. This is.38). ˆ ˆ (18. of course. ˆ ˆ 2 2 Now and 2 We may set γ = (18. how do we know that it breaks in such a way as to leave the isospin (‘R+L’) symmetry unbroken? A plausible answer can be given if we restore the quark mass terms via ¯ˆ ¯ˆ ¯ˆ ¯ ˆ ˆ ˆ ˆ Äm = m u u u + m d d d = 1 (m u + m d )q q + 1 (m u − m d )qτ3 q. (18. though it can be argued that the larger symmetry SU(3)L ×SU(3)R —appropriate to three massless flavours—must be spontaneously broken (see Weinberg 1996.38) ˆ ˆ Hence qR and qL transform quite independently.L ] = 0.Thus the SU(2)L and SU(2)R refer to the two chiral components of the fermion fields. we find that qR = (1 − i( + η) · τ /2)qR ˆ ˆ and similarly qL = (1 − i( − η) · τ /2)qL .3 verifies that Äq can be written as ¯ ˆ ¯ ˆ ˆ ˆ Äq = q R i DqR + q L i DqL ˆ / ˆ / (18.42) Copyright 2004 IOP Publishing Ltd . ˆ ˆ This formalism allows us to see immediately why (18. an issue that cannot be settled within perturbation theory (compare the comments after (17.39) (18.37) (18.40) ¯ˆ ¯ ˆ ¯ ˆ q q = q L qR + q R qL ˆ ˆ ˆ ¯ ¯ ¯ ˆ ˆ ˆ ˆ qτ3 q = q L τ3 qR + q R τ3 qL . q transforms by ˆ q → q = (1 − i · τ /2)q ˆ ˆ ˆ while under an axial isospin transformation q → q = (1 − iη · τ /2γ5 )q. section 22.

The ¯ ˆ qτ3 q term explicitly breaks the third component of isospin (resembling an ˆ electromagnetic effect) but its magnitude may be expected to be smaller than that ¯ˆ of the q q term.46) is consistent with ˆµ ∂µ ji. we find that ¯ ˆ ¯ ¯ q L qR → q L qR = q L (1 + i( − η) · τ /2)(1 − i( + η) · τ /2)qR (18. We therefore set (cf (17. being proportional to the difference of the masses.5 = 0 if p2 = 0.46) where f π is a constant with dimensions of mass. From sections 17.Including these extra terms is somewhat analogous to switching on an external field in the ferromagnetic problem.86). This suggests that the vacuum will ‘align’ in such a way as to preserve isospin but break axial isospin.44) and (18.5 (suitably generalized) we know that this current has non-zero matrix elements between the vacuum and a ‘Goldstone’ state which. in the σ -model ˆµ of section 18.2 Pion decay and the Goldberger–Treiman relation We now discuss some of the rather surprising phenomenological implications of spontaneously broken chiral symmetry—specifically. we find that f π = −vσ . ∂µ ji.5 is given by (18. from (18. if the pion is massless. in which ji. i. p = i p µ f π e−i p·x δi j µ and (18.40) is invariant Equations (18. which we expect to be related to a symmetry-breaking vev.43) ˆ ˆ ˆ ˆ ¯ qR − iη · q τ qR ¯ ˆ = qL ˆ ˆ ˆL (18.81) and the field σ develops a vev given in (18.38).45) q R qL → q R qL + iη · q R τ qL . rather than ˆ their sum. 18.3. in our case. which determines a preferred direction for the symmetry breaking.41) and (18. is the pion. It is clear that neither of (18.42) preserves SU(2)L ×SU(2)R since they treat the L and R parts differently.e.37) and (18. Indeed. ˆ ¯ q in (18. We start by ignoring any ‘explicit’ quark masses. Note that (18. Its value can be determined from the rate for the decay π + → µ+ νµ by the following reasoning. the spontaneous breaking of the axial isospin symmetry.45) confirm that the term q ˆ ˆ under the isospin part of SU(2)L ×SU(2)R (since is not involved) but not invariant under the axial isospin transformations parametrized by η .94)) ˆ 0| ji.5 (x)|π j .4 and 17.5 = 0. In chapter 20 we shall learn that the effective weak Hamiltonian density for this low-energy strangeness-non-changing semileptonic transition is ¯ ˆ ˆ ˆ ÀW (x) = √F cos θC ψ d (x)γ µ (1 − γ5 )ψu (x) G 2 Copyright 2004 IOP Publishing Ltd . The precise relation between f π and a vev depends on the dynamical theory (or model) being considered: for example.44) ¯ ˆ ¯ ˆ ¯ ˆ ˆ ˆ (18. ˆµ so that the axial isospin current is conserved. We treat f π as a phenomenological parameter.

Further.46). by the following argument. the invariant amplitude for the process. Using p = p1 + p2 . as in (18. taking the neutrino to p be massless. But since the initial and final parities are different. we note that ¯ ˆ ˆµ ˆµ ˆµ ˆµ ˆ ψ d (x)γ µ (1 − γ5 )ψu (x) = j1 (x) − i j2 (x) − j1. p2 | ˆ d4 x ÀW (x)|π + . ¯ ˆ ˆ ˆ ˆ × [ψνe (x)γµ (1 − γ5 )ψe (x) + ψ νµ (x)γµ (1 − γ5 )ψµ (x)] (18. νµ . (18. p = − √ 0| j1. On the other hand. the lowestorder contribution to the S-matrix is − i µ+ . For the pionic one.3 However. p1 . therefore. we find that 2 i ¯ ˆ ˆ ˆµ ˆ 0|ψ d (x)γ µ (1 − γ5 )ψu (x)|π + .52) by m µ . precisely for ˆµ this reason the axial currents ji.5(x) µ (18. p2 |ψ νµ (x)γµ (1 − γ5 )ψµ (x)|0 2 ¯ ˆ ˆ × 0|ψ d (x)γ µ (1 − γ5 )ψu (x)|π + . Å.5|π1 + iπ2 (18. Thus.20) and (18.5 − i j2. νµ .5 do have a non-zero matrix element.47) where G F is the Fermi constant and θC is the Cabibbo angle.21).50) 2 √ (18. 1 Noting that |π + = √ |π1 + iπ2 . Helicities of massless leptons in π + → µ+ νµ due to the ‘V − A’ interaction. it would have to be an axial 4-vector. the only 4-vector available is the pion’s momentum pµ which is an ordinary (not an axial) 4-vector. From Lorentz invariance such a matrix element has to be a 4-vector. p GF ¯ ˆ ˆ = −i √ cos θC d4 x µ+ . we may replace / in (18.5(x) + i j2. the currents ji can have no matrix elements between the vacuum (which is a 0+ state) and the π (which is 0− ). 3 See page 284. p1 .51) = 2 p µ fπ e−i p·x so that (18.48) becomes i(2π)4 δ 4 ( p1 + p2 − p)[−G F cos θC u ν ( p2 )γµ (1 − γ5 )v( p1 ) pµ f π ].Figure 18.48) The leptonic matrix element gives u ν ( p2 )γµ (1−γ5 )vµ ( p1 )ei( p1+ p2 )·x in the usual ¯ way. p .52) The quantity in brackets is. ¯ (18.49) ˆ from (18. Copyright 2004 IOP Publishing Ltd .2.

4 shows that the rate computed from (18. the amplitude is proportional to m µ . p | ji. µ This is a very important conclusion. so that the leptons carry off a net spin angular momentum. This is why. we comment on the physics of (18.5 .2. Problem 18. at least. Now the muon. would correspond to positive helicity for the µ+ (compare the discussion in section 12. if the muon was massless. while grossly contradicting the naive expectation that the rate with the larger energy release should dominate. this enables the value fπ 93 MeV (18.3.4. ˆµ ˆµ Since (with massless quarks) ji. Following an analysis similar to that in section 8. ˆµ Consider now another matrix element of ji.99))— we find that τi 5 5 q 0 = u( p )[ / γ5 F1 (q 2 ) + q 2 γ5 F3 (q 2 )] u( p) ¯ 2 τi 5 5 = u( p )[( / − / )γ5 F1 (q 2 ) + q 2 γ5 F3 (q 2 )] u( p) ¯ p p 2 5 2 2 5 2 τi = u( p )[−2Mγ5 F1 (q ) + q γ5 F3 (q )] u( p) ¯ (18. the amplitude vanishes for massless muons and neutrinos.52). The rate is therefore proportional to m 2 . it selects the negative helicity neutrino state. is one of the main indications for the ‘vector–axial vector’. The (1 − γ5 ) factor acting on a v spinor selects out the γ5 = −1 eigenvalue which. in fact. Hence.52).56) 2 Copyright 2004 IOP Publishing Ltd . because it implies that the rate to muons is ∼ (m µ / m e )2 ∼ (400)2 times greater than the rate to electrons—a result which agrees with experiment. this time between nucleon states.2). we write ˆ N. or ‘V − A’.5 (0)|N. This. taking the (1 − γ5 ) through the γ 0 γ µ factor to act on u † .52) is π→µν = 2 G 2 m 2 f π (m 2 − m 2 )2 π µ F µ 4π m 3 π cos2 θC .Before proceeding. in an amount proportional to m µ . M is the nucleon mass and q = p − p . structure of (18.47). as we have just remarked after (18. since the pion spin is zero.53) Neglecting radiative corrections. p 5 = u( p ) γ µ γ5 F1 (q 2 ) + ¯ µ iσ µν τi 5 5 qν γ5 F2 (q 2 ) + q µ γ5 F3 (q 2 ) u( p) 2M 2 (18.5 (0) = 0 (cf (8. (18. But this is forbidden.55) are understood to be written in flavour and Dirac space. as we shall see in more detail in section 20. Likewise.8 for the matrix elements of the electromagnetic current operator between nucleon states.5 is conserved—that is qµ ji.54) to be extracted. the ν configuration is as shown in figure 18.55) where the Fi5 ’s are certain form factors. is not massless and some ‘wrong’ helicity is present in its wavefunction. The spinors in (18. Hence.

3.46). therefore. first given by Nambu (1960). 5 5 that either M = 0 (which is certainly not so) or F1 (0) = 0. But F1 (0) can be measured in β decay and is found to be approximately equal to 1. the contribution of figure 18.61) Copyright 2004 IOP Publishing Ltd .3. Such a contribution can only arise from the propagator of a massless particle—which. (18. using / γ5 = −γ5 / and the Dirac equations for u( p).5 ‘saves’ the conservation of the current.58) igπNNπi N γ5 τi N which is SU(2)f -invariant and parity-conserving since the pion field is a ¯ pseudoscalar and so is N γ5 N. Putting these pieces together. Hence. the form p p ¯ 5 5 factors F1 and F3 must satisfy 5 5 2M F1 (q 2 ) = q 2 F3 (q 2 ). This elegant physical argument.57) with (18. u( p ).60) from this contribution. and the (massless) propagator is i/q 2 . The pion-current matrix element is given by (18.5 Figure 18. sheds a revealing new light on the phenomenon of spontaneous symmetry breaking: the existence of the ˆµ massless particle coupled to the symmetry current ji. Combining (18. Here.57) appears to predict. of course.57) Now the matrix element (18. One pion intermediate state contribution to F3 . is the pion.3 to the current matrix element is ¯ 2gπNNu( p )γ5 and so 5 F3 (q 2 ) = i τi u( p) 2 − iq µ f π 2 q 1 2gπNN f π q2 gπNN fπ M (18.26: 5 it is conventionally called gA . the conventional Lagrangian is ˆ ˆ ˆ ¯ (18. we deduce 5 gA ≡ lim F1 (q 2 ) = q 2 →0 (18. The only possible conclusion is that F3 must contain a part proportional to 1/q 2 . The process is pictured in figure 18. 5 We calculate the pion contribution to F3 as follows.59) (18. For the π–N vertex.60).55) enters into neutron β-decay (as does the ˆµ matrix element of ji (0)). q 2 0 and (18.

67) (18.the well-known Goldberger–Treiman (G–T) (1958) relation. the pion mass is not zero and neither are the ‘explicit’ quark masses m u . Now consider the matrix element ¯ ˆ ˆ ˆ ∂µ 0| ji.60). we see that m 2 is proportional to the sum of quark masses and π π tends to zero as they do.64) (18. 5 5 −2M F1 (q 2 ) + q 2 F3 (q 2 ) = i τi u( p) 2 m 2 fπ 2 q − m2 π π (18. ˆµ This means that the one-nucleon matrix element of ∂µ ji.55) it is given by τi 5 5 iu( p )[−2Mγ5 F1 (q 2 ) + q 2 γ5 F3 (q 2 )] u( p).46) tells us that ∂µ ji. With m u and m d reinstated.66) Since p2 = m 2 . We can repeat the argument leading to the G–T relation but retaining ˆµ m 2 = 0.68) 2gπNNm 2 fπ π . Equation (18.5 /(m 2 f π ) behaves like a properly π π normalized pion field. ¯ 2 Hence.26 and f π = 93 MeV. m d . Then {m.65) (18. the equations of motion for the quark fields are ˆ ˆ ¯ ˆ ¯ / i D q = mq ˆ − i Dµ qγ µ = q m ˆ ˆ (18.5 = iq m. at least when operating on a near mass-shell pion state.69) Also.62) where m= µ mu 0 0 md . 2 µ (18. q2 − m2 π (18. we now have 5 F3 (q 2 ) = 1 2gπNN f π .59)) ¯ 2gπNNu( p )γ5 while from (18. gA = 1. In the real world. which is only 5% below the experimental value of this effective pion–nucleon coupling constant. ˆ ˆ ∂µ ji. 2 Let us take the case i = 1. in place of (18. Taking M = 939 MeV.7.5 and find (problem 18.5) that τi ¯ ˆµ γ5 q.5 is (cf (18. (18. q2 − m2 π (18.70) Copyright 2004 IOP Publishing Ltd . τ1 } = (m u + m d )τ1 . for example.63) ˆ We can re-calculate ∂µ ji. one obtains gπ NN ≈ 12.5 (0)|π1 ( p) = − p2 f π = 1 (m u + m d ) 0|i qτ1 γ5 q|π1 .

Equations (18.21) and the second is equivalent to (12. π an isotriplet ˆ and σ an isoscalar.69) and (18. What are the global symmetries of (18. rather than quarks. We refer the interested reader to Georgi (1984).73) (18. The Lagrangian is (with λ > 0) ˆ ˆ Äσ ¯ /ˆ ¯ ¯ ˆˆ ¯ ˆ ˆ ˆ ˆ ˆ = ψi∂ ψ + igπNNψτ γ5 ψ · π + gπNN ψ ψ σ + 1 ∂µ π · ∂ µ π ˆ 2 ˆ ˆ ˆ ˆ ˆ ˆ + 1 ∂µ σ ∂ µ σ − 1 µ2 (σ 2 + π 2 ) − 1 λ(σ 2 + π 2 )2 . ˆ ˆ ˆ ˆ ji 2 (18.72)? We can at once infer that it is ˆ invariant under global isospin transformations if ψ is an isodoublet. The associated symmetry current is ¯ ˆµ (σ ) = 1 ψγ µ τi ψ + (π × ∂ µ π)i . As we will see. as we shall see) and a massless pseudoscalar isotriplet π (the pions). it effectively embodies many of the preceding results.3 The linear and nonlinear σ -models ˆ The linear σ -model involves a massless fermion isodoublet ψ (which will be identified with the nucleon—its mass being generated spontaneously.72) We have seen all the different parts of this before: the massless fermions. It will be a somewhat more complicated model than that of section 17. We now turn to another example of a phenomenological model exhibiting spontaneously broken axial isospin symmetry. since it contains no rapidly π varying pion pole contribution. and so we recover the G–T relation again. Amplitudes involving two ‘Goldstone’ pions can be calculated by an extension of these techniques.74) (18. π (18.76) is as in (18. an analogous σ –ψ coupling (here with the same coupling constant).71) 5 F1 (q 2 ) varies only slowly from q 2 = 0 to q 2 = m 2 .4. 2 2 4 (18. The corresponding charges ˆ (σ ) Qi = ˆ0 (σ ) d3 x ji (18. though similar to it in that the spontaneous breaking is put in ‘by hand’ via a suitable potential.75) under an infinitesimal SU(2)f transformation. 18. transforming by ˆ ˆ ˆ ˆ ψ → ψ = (1 − i · τ /2)ψ ˆ π →π =π + ×π ˆ ˆ ˆ σ →σ =σ ˆ ˆ (cf (12. and π and σ fields with a symmetrical mass parameter µ2 and a quartic potential.124). this time realized in terms of hadronic (nucleon and pion) degrees of freedom. the π –ψ coupling as in (18.58). There is also a scalar ˆ field σ which is an isoscalar.76) The first term of (18.77) Copyright 2004 IOP Publishing Ltd .64)) (18.70) are consistent for q 2 = m 2 if π 5 F1 (q 2 = m 2 ) = gπ NN fπ / M.

5 ’s.79) (18. for which of course no γ5 matrix can enter? Just as the algebra of SU(2) is the same as that of the generators of 3D rotations (section 12. 24 and 34 directions.1).78) is the same as (18.80) where η is a second set of three infinitesimal parameters. of σ 2 + π 2 . Remarkably. But the SO(4) symmetry of (18. therefore. we have i a model of interacting fields with a global SU(2)L×SU(2)R symmetry. where now ¯ ˆµ ˆ ˆ ˆ ˆ ji. which is the same number of generators as the ˆ (σ ) ˆ (σ ) three Q i ’s together with the three Q i. if we regard (σ . However. therefore. which are analogous to spacetime (velocity) transformations in special relativity. namely ˆ ˆ ˆ ψ → ψ = (1 − iη · τ /2 γ5 )ψ π → π = π + ησ ˆ ˆ ˆ ˆ σ →σ =σ −η·π ˆ ˆ (18.5 would not be conserved if we added an explicit mass ˆ for ψ . once again.5 = 1 ψγ µ γ5 τi ψ + (σ ∂ µ πi − πi ∂ µ σ ).82) ˆ and Q (σ ) satisfy relations of the form (18.5 ) d3 x (18.36) and is. It can ˆµ easily be verified that ji.are constants of the motion and obey the SU(2) algebra (18.24)–(18. The transformations of SO(4) ˆ ˆ ˆ preserve the (length)2 of 4-vectors—in this case. so the algebra of SU(2)×SU(2) turns out to be the same as that of the generators of rotations in a four-dimensional Euclidean space—here. 2.5 ˆ0 (σ ji. at least in the meson sector.e. 3 and 4.72) is also invariant under a further set of transformations (see problem 18.5 . This makes six ‘rotations’ in all.1. Transformation (18. leading to an SU(2) algebra.24). section M. π ) as being the four components of a ‘4-vector’. 13 and 23 which would be ‘spatial’ rotations. If we label the four directions as 1. But how can we understand this SU(2)×SU(2) structure for the meson fields. ˆ 2 (18. Thus. (18. pions and nucleons on the other). we know that the ‘L’ and ‘R’ refer to the components of different chirality. the group of 4D rotations) is the same as SU(2)×SU(2). again an axial isospin transformation on ˆµ ˆ the doublet ψ . ˆ As far as the fermion field ψ is concerned.72) is apparent. the first term is as in (18. this by itself by no means proves that the algebra of the generators of SO(4) (the special orthogonal group in 4D—i. Note that all these algebraic results hold independently of the specific realization of the operators in terms of the fields in the model under consideration (quarks on the one hand. we can imagine rotations in the planes 12.81) Again.78) (18. This suggests we call the second set of currents ji.26).3). of course. But there are also rotations in the 14. Just this Copyright 2004 IOP Publishing Ltd .20) and there is a new ‘mesonic’ piece. an ‘internal’ space involving the field components (see appendix M. Of course.4.6). the charges ˆ Q (σ ) = i.

Finally. We now consider the symmetry ˆ ˆ breaking case µ2 < 0. in fact.83) where the ‘new’ µ2 in (18.R = 1 [ Q i + Q i.74) and (18. Thus far we have supposed that the parameter µ2 in (18.7) that Äσ becomes ˆ Äσ ¯ / ¯ ¯ ˆˆ ˆ ˆ ˆ ˆ ˆ ˆ = ψ(i∂ + gπNNvσ )ψ + igπNN ψτ γ5 ψ · π + gπNN ψ ψ σ µ µ 2 2 1 1 + 2 ∂µ π · ∂ π + 2 ∂µ σ ∂ σ + µ σ − λvσ σ (σ 2 + π 2 ) ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ − 1 λ(σ 2 + π 2 )2 + constant. π → x . We choose 0|π |0 = 0 ˆ 0|σ |0 = vσ .83) is positive. 1 (18.5 2 (σ ) (σ ) (σ ) ˆ ˆ ˆ Q i. we interpret (18.78). ˆ ˆ 4 (18. (to first order in η ) σ 2 + π 2 = σ 2 + π 2 as ˆ ˆ ˆ required. i.L i i.84) 2 as 0|π 2 + σ 2 |0 = vσ . while (18.5 ] as before. ˆ (18. Then the ˆ ˆ ˆ associated ‘L’ and ‘R’ generators can be found by writing Q (σ ) = 1 ( Q (σ ) − Q (σ ) ).79) and (18.2 to the surface of a sphere in 4D.97). note that (18.72).combination is visible in the potential terms of (18. indeed. the ground state having 0|σ |0 = 0|π|0 = 0. as in (17. The symmetry is then unbroken.72) now becomes V (π.87) has. we find (problem 18. These relations provide us with the ‘L’ and 2 ‘R’ transformation law for the meson fields in this model. This potential has a minimum when 2 π 2 + σ 2 = vσ (18. So (18. while for σ we ˆ ˆ ˆ need to set ˆ (18.102)—a point to which we shall eventually return in chapter 22. which we may reasonably call SU(2)f 5 in view of (18. The classical potential in (18.88) The spontaneous breaking typified by (18. Introducing ˆ ˆ (18.79) and (18. σ ) = − 1 µ2 (σ 2 + π 2 ) + 1 λ(σ 2 + π 2 )2 2 4 (18. exactly analogous to (17.87) σ = vσ + σ ˆ and then expand σ in terms of creation and annihilation operators.77) and (17. it is also.86) ˆ Then π will have a standard expansion in terms of a ’s and a † ’s.72) is positive. therefore.85) with vσ = (µ2 /λ) 2 . representing a normal mass parameter.84) (18.84) generalizes the circular minimum of figure 17.80) record their transformations under the other SU(2).80) that.79) and (18.87) into (18. As in these previous cases.72). and we can easily verify from (18. induced the following results: Copyright 2004 IOP Publishing Ltd .75) tell us how π and σ transform under ˆ ˆ ˆ ˆ ordinary SU(2)f . and t 2 − x 2 invariant ˆ ˆ rather than σ 2 + π 2 . and we need to choose one particular ground state before ˆ ˆ we can get a proper particle interpretation. with σ → t .80) are analogous to an infinitesimal velocity transformation in relativity.

We can ask to what extent the rather simple Lagrangian (18. ∂µ ji. Previously this was of course zero (from (18.2. In addition. The π–N interaction itself leads.90) (at least to tree level in the interactions of Äσ + cσ ) as follows.46). there is no real dynamical explanation of the symmetry breakdown here. ˆ We may therefore regard Äσ as some kind of ‘effective Lagrangian’ embodying the SU(2)L×SU(2)R symmetry of QCD.86) respects it.(i) the fermion has acquired a mass −gπ N N vσ . In the QCD case. proportional to the symmetry breaking parameter vσ .86).85)). broken spontaneously in just the same way as we assumed for QCD.5 is proportional to the pion field. there is a pion mass term 2 ˆ − 1 (−µ2 + λ2 vσ )π 2 .72) (with the µ2 term as in (18.84) ˆ gets shifted to a new point such that 0|σ |0 —which we still call vσ —satisfies ˆ 2 vσ (−µ2 + λvσ ) = c. of course.8) that now ∂µ ji.90) ˆ ˆ We can check (18. while leaving SU(2)f intact. of course. and we recover a G–T type relation (at tree level) with gA = 1. but SU(2)f 5 is spontaneously broken. One quickly verifies (problem 18. Copyright 2004 IOP Publishing Ltd . precisely to the ‘one pion exchange’ potential between two nucleons. With the explicit symmetry-breaking addition cσ .72). as postulated by Yukawa. it is analogous to an ‘alignment ˆ ˆµ field’).5 is no longer zero but is given by ˆµ ∂µ ji. √ (iii) the σ field has a mass 2µ.83)) describes other features of low-energy interactions among pions and nucleons. proportional to vσ . we should have c = −m 2 f π . The empirical consequences—massless pions.91) we can identify m 2 as c/vσ . In the same way. (iv) there are new trilinear couplings between π and σ . but now 2 from (18. In that case. just as in the previous section.72) is preserved. ˆ (18. we saw that a quark mass term broke the full SU(2)×SU(2) symmetry explicitly. from item (i) we learn that the nucleon mass is gπNN f π . corresponding to oscillations in the ‘radial’ ˆ direction in figure 17. for consistency with (18. In fact. a massive nucleon—are the same but.90) if π our symmetry breaking parameter vσ is identified with − f π . π (18. being the three Goldstone modes ‘perpendicular’ ˆ to the one selected out by the symmetry-breaking condition (18. the addition of ˆ a term +cσ to Äσ will have the same effect (again. (18. (ii) the π ’s remain massless.91) Note that vσ returns to (µ2 /λ)1/2 for c → 0.5 = −cπi . and ˆ (v) the SU(2)f symmetry of (18. since the vacuum choice (18. the minimum (18.89) ˆµ Thus. This model can be easily modified to include a finite mass for the pions. just a ‘by hand’ choice of the sign of µ2 in (18. This will be consistent with (18.

this π –N theory is renormalizable. although in view of the magnitude of gπ NN.94) via its expansion ˆ Änlσ m = 1 2 (π · ∂ π)(π · ∂ π)(1 − π 2 / f π )−1 ˆ ˆ ˆ ˆ ˆ 2 2 fπ 1 = (π · ∂µ π )(π · ∂ µ π ) + O(π 6 ).95) (18. This implies that λ also becomes very large.2. Nevertheless. However. The potential terms disappear altogether from (18. It opens up a large field of ‘effective Lagrangians’ for low-energy hadronic physics (Donoghue et al 1992). ˆ ˆ ˆ ˆ ˆ 2 2 fπ (18.but now incorporating the conservation of isospin.93) (18. However. we note that the effective interaction 2 1 2 (π · ∂π)2 is not renormalizable. First. provided one worked at energies below the scale set by the dimensional coupling.92) is rather remarkable.72) is that there is no plausible candidate. The result was first given by Weinberg (1966). We interpret the denominator in (18. We can actually get rid of σ by supposing that its mass √ 2µ is very large indeed but in such a way that the ratio vσ = (µ2 /λ)1/2 remains finite. for which it makes a quite specific prediction. if we are lucky). The main phenomenological problem with (18. since it has a coefficient f π with dimension (mass)−2 . rising as the square of the 4 scaling of µ. At any rate. Furthermore.96) The first term in (18. it is striking that symmetry conditions have determined the low-energy dynamics of the Goldstone modes. the discussion in section 11. which 4 λ becomes very large and negative. Thus in terms of a picture such as figure 17. This is a general feature and clearly a most important one. among the observed mesons with masses below 1 GeV. even in loops. since f π is known. perturbation theory is of little value.92) The result of imposing (18. the potential has a very deep and narrow minimum and (assuming a semi-classical picture) the fields are effectively constrained to lie on the ‘chiral circle’ 2 ˆ σ 2 + π 2 = fπ .94) (π · ∂µ π )(π · ∂ µ π) ˆ ˆ ˆ ˆ 2 ( fπ − π 2) ˆ .96) invites a number of comments. Let us consider now just the meson sector. Here that scale is f π = 93 MeV (or perhaps this multiplied by numbers such as 2π. and is consistent with experiment (see Donoghue et al (1992) for a review). using a different technique. for the ‘σ ’ (0+ ) meson. ˆ (18. Relation (18. Copyright 2004 IOP Publishing Ltd . the value of the potential at the minimum is − 1 µ .96) describes a pion–pion scattering process of the form π +π → π +π.8 showed that such a Lagrangian could still be useful.72) and we are left with pions interacting via the interaction ˆ Änlσ m = 1 ∂µ 2 = 1 2 2 fπ − π 2 ∂ µ ˆ 2 fπ − π 2 ˆ (18. we expect the theory to work only for energies not too far from threshold.

happen and when it does the afflicted current (or its divergence) is said to be ‘anomalous’ or to contain an ‘anomaly’. Donoghue et al (1992)). A combination of the low-energy Goldstone mode dynamics and the (at present) phenomenological ρ-meson contribution provides a good representation of nonstrange mesonic dynamics below about 1 GeV. there is a corresponding conserved current.3. We demonstrated this result in some special cases. approximate and spontaneously broken—there is one result on which we have relied and never queried. However. Expanding this in powers of q 2 / m 2 . t = ( p1 − p1 )2 or u = ( p1 − p2)2 . where p1 + p2 = p1 + p2 expresses 4-momentum conservation. 18. J P = 1− π − π resonance called the ρ (see section 9. Despite the low-energy success of (18. In particular. with a mass of 770 MeV. we may expect an anomaly when we introduce electromagnetism into our chiral models (such as the linear σ -model). as discussed in section 12. A proper calculation shows that indeed such σ terms exactly cancel the 1 λ(π 2 )2 ones (see. so that the corresponding Feynman amplitudes for π( p1 ) + π( p2 ) → π( p1 ) + π( p2) contain two powers of the momenta. On the π other hand.88). derivatives are absent from the meson sector of the ‘spontaneously broken’ model (18. General analysis shows that anomalies occur in renormalizable theories of fermions coupled to both vector and axial vector currents. The importance of ρ -meson exchange in hadronic dynamics was first stressed by Sakurai (1960) in his ‘vector meson dominance’ theory. The situation is no better with the ‘linear σ -model’ of (18. for example. it is interesting that the effective π –π interaction involves two derivatives.Second. which from Lorentz invariance must appear in the form of s = ( p1 + p2 )2 . ˆ 4 leaving the leading contribution proportional to q 2 (where q is any of the possible momentum transfers).96) (which is called the ‘nonlinear σ -model’ in this context). in fact. suggesting a contradiction.4 Chiral anomalies In all our discussions of symmetries so far—unbroken. In particular. They therefore vanish (for m 2 = 0) at the points s = 0.88).5). we did not investigate whether quantum corrections might violate the classical conservation law.1. but we have now to point out that while it is undoubtedly valid at the level of the classical Lagrangian and field equations. In particular. This can. since then Copyright 2004 IOP Publishing Ltd . there is no sign of the T = 1. there will also be contributions from σ exchange of the form A/(q 2 − m 2 ) where m 2 = 2µ2 and A is proportional σ σ 2 to λ2 vσ . the interaction 1 ˆ2 2 4 λ(π ) would seem to lead to a constant (non-vanishing) contribution as the momenta went to zero. We refer to Noether’s theorem. It is phenomenologically rather successful but it has not yet been possible to derive it directly from QCD. we find a leading term − A/ m 2 ∼ σ σ 2 2 −λ2 vσ / m 2 ∼ −λ2 vσ /µ2 ∼ −λ. it fails to account for prominent phenomena as the energy scale rises into the 500–1000 MeV region. This states that for every continuous symmetry of a Lagrangian. therefore. t = 0 or u = 0.

In that case. One way of understanding how anomalies arise is through consideration of the renormalization process. in some cases. this will invariably entail some regularization of divergent integrals. We shall not attempt an extended discussion of this technical subject. both axial vector and vector symmetry currents will be involved. to indicate how they arise in one simple model. while in others they must be eliminated. Anomalies arise when both axial and vector symmetry currents are present. We begin by writing the matrix element of j3. chapter 19). The amplitude considered in (18. which is in general necessary once we get beyond the classical (‘tree level’) approximation.5 (x) should then be conserved. Jackiw 1972. Adler and Bardeen 1969). This is an example of a chiral anomaly. to be discussed shortly. We consider the classic case of π 0 → γ γ . since it is not possible to find a regularization scheme which preserves both vector and axial vector current conservation (Adler 1970. using a different approach. they are to be welcomed. The axial isospin ˆµ current ji. as first pointed out by Veltman (1967) and ˆµ Sutherland (1967).21). But the specific example of the O(e2 ) photon self-energy studied in section 11. but we shall see that this implies that the amplitude for π 0 → γ γ must vanish. A full exposition of the anomaly in the axial vector current in spinor electrodynamics was first given by Adler (1969) and the occurrence of the anomaly in the σ -model (with electromagnetic interactions) was pointed out by Bell and Jackiw (1969). A more modern non-perturbative perspective is provided by Peskin and Schroeder (1995. a typical illustration of which is provided by the calculation of the rate for π 0 → γ γ . As we saw in volume 1. But we do want to alert the reader to the existence of these anomalies. The need for particular care in the calculation of π 0 → γ γ was first recognized by Schwinger (1951).1 Figure 18.97) and the one-pion intermediate state contribution to it. and to explain why.5(x) between Copyright 2004 IOP Publishing Ltd .3 showed that a simple cutoff form of regularization already violated the current conservation (or gauge invariance) condition (11.4. in the context of spontaneously broken global chiral symmetry with massless quarks and pions. it was possible to find alternative regularizations which respected electromagnetic current conservation and were satisfactory.

k 2 . due to the anti-symmetry of the symbol). it is helpful to consider a detailed calculation performed within a specific model. so the Lagrangian Äσ of (18. The contribution of the process shown in figure 18.97) As in figure 18.97) then shows that this implies that qµ ŵνλ = 0 (18. one contribution to ŵνλ has the form (constant/q 2 ) due to the massless π 0 propagator. shown in figure 18.the vacuum and a 2γ state. γ . therefore.5 ( x)|0 ∗ ∗ 1ν (k 1 ) 2λ (k 2 ) = (2π)4 δ 4 (k1 + k2 − q) ŵνλ (k1 .4. Note that this automatically incorporates electromagnetic gauge invariance (the amplitude vanishes when the polarization vector of either photon is replaced by its 4-momentum.72) so as to include only one fermion of charge e (the proton) and two mesons. by Steinberger (1949). It is also effectively the model used long before. as d4 x e−iq·x γ .46). shown in figure 18. k1 . there are two graphs to consider.4 to ŵνλ is then iq µ f π i iA q2 νλαβ k1α k2β (18.72) is an obvious choice. will not have the 1/q 2 factor in (18. give a vanishing contribution to qµ ŵνλ at q 2 = 0 which is the on-shell point for the (massless) pion.99) which with (18. and hence that π 0 → γ γ is forbidden.5(x). the σ and π3 . of course. once again. the axial current connects the pion state to the vacuum. It is important to realize that all other contributions to ŵνλ . In the present case.98) and will. 2 | j3. true that m 2 = 0. For our purposes. apart from the π 0 one shown in figure 18.3. ˆµ 1 . A partial integration in (18. In such a situation. when chiral symmetry is spontaneously broken. This was the model considered by Bell and Jackiw (1969) and also by Adler (1969). which should be zero. Consequently. Now consider replacing ˆµ ˆµ j3. To order α . in the first calculation of the π 0 → γ γ rate. (18. in momentum space. as described by the matrix element (18. when enlarged to include electromagnetism in the usual gaugeinvariant way. It turns out that the fermion loop integral is actually convergent: details of its evaluation may be found in Copyright 2004 IOP Publishing Ltd .98) implies that A = 0. This is because.5(a) and (b). k2 ). It is. there is a problem for the hypothesis of spontaneously broken (approximate) chiral symmetry. But estimates (Adler 1969) of the π consequent corrections suggest that the predicted rate of π 0 → γ γ for real π 0 ’s is far too small.5(x) in (18.97) by ∂µ j3. and it is symmetrical under interchange of the photon labels.98) ∗ ∗ where we have parametrized the π 0 → γ γ amplitude as A νλαβ 1ν (k1 ) 2λ (k2 ) k1α k2β . we want a model which exemplifies ˆ spontaneously broken chiral symmetry. it will be sufficient to simplify (18.4.

But a careful analysis of the ‘triangle’ graph contributions to the matrix element ŵνλ of (18. various possibilities now exist. section 11.2).101) is in very good agreement with experiment. 2 64π 3 fπ (18. at the same time as electromagnetic gauge invariance. no such regularization can be found. Figure 18. reveals that the fault lies in assuming that a regularization exists such ˆµ that for these amplitudes the conservation equation qµ γ γ | j3.97). The two O(α) graphs contributing to π 0 → γ γ decay in the simplified version of (18. one finds a non-zero result ˆµ for qµ γ γ | j3. shown in figure 18. Copyright 2004 IOP Publishing Ltd . In principle.101) (18. the result is −1 (using g/ m = f π ) e2 A= (18.5(0)|0 (again the details are given in Itzykson and Zuber (1980)).100) to give (π 0 → 2γ ) = α2 m 3 π . When the amplitudes of figure 18. In fact. Problem 18.72).6. O(α) contributions to the matrix element in (18.Figure 18.100) 4π 2 f π where A is the π 0 → γ γ amplitude introduced earlier. In the limit q 2 → 0.5.97).5. Itzykson and Zuber (1980.6.5(0)|0 = 0 can be maintained.9 evaluates the π 0 → γ γ rate using (18.6 are calculated using an (electromagnetic) gauge-invariant procedure.

103) which is indeed consistent with (18. and is already evident from (18. we have an anomaly in a current associated with a local symmetry.102) predicts a definite value for A. It might be thought that the result (18. In the foregoing discussion. Colour supplies just this degeneracy.48).10) iqµ ŵνλ = e2 4π 2 ανβλ µ k1α k2µ (18. possible between quark and lepton ‘triangles’ in the weak interaction Copyright 2004 IOP Publishing Ltd . this gives 1/3. For the u and d quarks of chiral SU(2)×SU(2).99) is no longer valid. As we shall see in chapter 20. The triangles considered earlier actually used a fermion with integer charge (the proton). If.102) means that (18. and thus the π 0 → γ γ rate is important evidence for such a degree of freedom.97) and (18.102) produces (problem 18. Taking the vacuum → 2γ matrix element of (18. axial currents are indeed present in weak interactions and they are coupled to the W± . Equation (18. after suitably interchanging the labels on the symbol.5(x) is not zero after all.164) to eliminate unwanted gauge degrees of freedom and ensure unitarity of the S -matrix. So anomalies in currents coupled to gauge fields cannot be tolerated. the previous numerical value for A is multiplied by the factor τ3 Q 2 for each contributing quark.ˆ This implies that ∂µ j3. Thus cancellations are. But Adler and Bardeen (1969) showed that ˆµ such ‘triangle’ loops give the only anomalous contributions to the ji. Equation (18. we will have a serious problem. of a symmetry of the classical Lagrangian. however. so that A need no longer vanish: indeed.5 − γ − γ vertex. at the quantum level. all anomalies must somehow cancel away.5(x) = − (18.102) is true to all orders in α .100). In this case. so that (18. the axial isospin current was associated with a global symmetry: only the electromagnetic currents (in the case of π 0 → γ γ ) were associated with a local (gauged) symmetry and they remained conserved (anomaly free). which carry fractional charge.102) is a typical example of ‘an anomaly’—the violation. That this is possible rests essentially on the observation that the anomaly (18.21) or (13. in principle. the calculation producing the specific value e2 ανβλ ˆ ˆ ˆµ ∂µ j3. The whole rather elaborate construction of a quantum gauge field theory relies on current conservation equations such as (11.102) is only valid to order α (though the O(α 2 ) correction would presumably be very small).102) is independent of the mass of the circulating fermion. identical in their electromagnetic and SU(2)×SU(2) properties. Z0 gauge fields.100). if this theory is to be satisfactory at the quantum level.102) Fαν Fβλ 32π 2 ˆ where the F ’s are the usual electromagnetic field strength tensors. (18. Hence. We clearly should use quarks. Consequently agreement with experiment is lost unless there exist three replicas of each quark. so we need to see if it is consistent with (18.

18. the Lagrangian (18.26).24)–(18. 18.72) is invariant under the transformations (18.2 Verify (18. Copyright 2004 IOP Publishing Ltd .101). 18. the divergence of the ˆ ˆµ ˆ axial vector current is given by ∂µ ji. Problems 18.72).9 Verify (18. Remarkably Show that the rate for π + → µ+ νµ . Weinberg 1996. 18. section 22.28)–(18.88).3 Show that Äq of (18.5).12) can be written as (18. Bouchiat et al (1972) were the first to point out that each generation of quarks and leptons will be separately anomaly free if the fractionally charged quarks come in three colours.78)–(18. section 20.52).39).30). is given by (18. complete cancellation of all anomalies does occur in the GSW theory (see Peskin and Schroeder 1995.72) becomes (18. 18. 18. calculated from the lowest-order matrix element (18. 18.64). Anomaly cancellation is a powerful constraint on possible theories (’t Hooft 1980.5 (x) = −cπi .2).80).6 Show that (18. 18.103).7 Show that after making the ‘shift’ (18. 18.8 Show that when a term cσ is added to Äσ of (18.5 Verify (18.1 Verify (18. and calculate the π 0 lifetime in seconds.10 Verify (18.87).53).

as mentioned in section 2. . The reasons include: (a) the goal of unification (e. so that their mediating quanta must be massive. (19. as pointed out in section 13.19 SPONTANEOUSLY BROKEN LOCAL SYMMETRY In earlier parts of this book we have briefly indicated why we might want to search for a gauge theory of the weak interactions.51. if Aµ were to represent a massive field. we also know from section 2.3.5 that weak interactions are short-ranged. which are reminiscent of a similar situation in QED (see comment (ii) in section 3. is invariant under the gauge transformation Aµ → A µ = Aµ − ∂ µ χ.3) is manifestly not invariant under (19.3).g. Nevertheless.1 after equation (13. and we will review again in the following section. £ + M 2 (compare sections 4.6) and which are particularly characteristic of a non-Abelian gauge theory. and (b) certain ‘universality’ phenomena (to be discussed more fully in chapter 20). The detailed application to the electroweak theory will be made in chapter 22. Equation (19. local) symmetry. as discussed in section 3. which is by ‘spontaneously breaking’ the gauge (i. as we pointed out for the photon in section 11.21)) ν £ Aν − ∂ ν (∂µ Aµ) = jem (19.44). since a simple gauge boson mass violates gauge invariance. 19. the relevant wave equation would be ν (19.1 and 5.4 and for non-Abelian gauge quanta in section 13.6 and also section 11. However. we have simply replaced the massless ‘Klein–Gordon’ operator £ by the corresponding massive one.e.3) (£ + M 2 )Aν − ∂ ν (∂µ Aµ ) = jem .1) which.5. with the U(1) gauge theory QED). The same conclusion follows Copyright 2004 IOP Publishing Ltd However. there is a way of giving gauge field quanta a mass. The electromagnetic potential satisfies the Maxwell equation (cf (3. this seems to rule out the possibility of a gauge theory of weak interactions.2) To get this. At first sight.2) and it is precisely the mass term M 2 Aν that breaks the gauge invariance.1 Massive and massless vector particles Let us begin by noting an elementary (classical) argument for why a gauge field quantum cannot have mass. This is the topic of the present chapter.

one adds a mass term + 1 M 2 Aµ Aµ to the Lagrangian of (7. Apart from the ‘i ’. which we will call X-particles. Fermion–fermion scattering via exchange of two X bosons. Now consider the loop integral in figure 19. which shows some kind of fermion– fermion scattering (we need not be more specific).1. in a Lagrangian treatment: to obtain (19. thus indicating already that a massless vector particle seems to be a very different kind of thing from a massive one (we cannot just take the massless limit of the latter). but without the current on the right-hand side (so as to describe a free field): ˜ [(−k 2 + M 2 )g νµ + k ν k µ ] X µ (k) = 0 (19. Consider figure 19.5) k2 − M 2 A proper field-theoretic derivation would yield this result multiplied by an overall factor ‘i’ as usual and would also include the ‘i ’ via k 2 − M 2 → k 2 − M 2 + i .2 for photons.4).1 shows that.4 and 13. as we now discuss. without it being a gauge theory.1. We remark immediately that (19.3) for the corresponding X ν field. Similar reasoning holds for the non-Abelian case too. Such a theory is certainly possible but it will not be renormalizable.87). To calculate this amplitude. which can be found by following the ‘heuristic’ route outlined in section 7.1) and this clearly violates invariance under (19. the propagator should be proportional to the inverse of the quantity in the square brackets in (19. Perhaps. we need the propagator for the X-particle. Problem 19.5) gives nonsense in the limit M → 0.5. proceeding in fourth-order perturbation theory via the exchange of two massive vector bosons.63) (see also 2 sections 11.2). At each vertex we will have a coupling constant g. (19. this inverse does exist and is given by −g µν + k µ k ν / M 2 . W± for example.1.4) which should be compared with (7.3. then. associated with an interaction Lagrangian having the general Copyright 2004 IOP Publishing Ltd . unlike the (massless) photon case. we must settle for a theory involving massive charged vector bosons.3) as the corresponding Euler–Lagrange equation. We consider the momentum–space version of (19.Figure 19.

|k|): then the x. indeed. Now. as we shall now explain. They come from the longitudinal polarization state of the massive X-particle.6) for high k-values (we are not troubling to get all the indices right. so that the divergence becomes ∼ d4 k 11 // kk (19.¯ ˆ ˆ ˆ form g ψγµ ψ X µ (a γµ γ5 coupling could also be present but will not affect the argument). Now the first two bracketed terms in (19. (19.6) behave like a constant at large k. (19. 0.8) e . indeed. the polarization vectors ν ν −ik·x (19. as we warned the reader in section 11.8. this is a case where it does not. consider the leading divergent behaviour of figure 19. this may not guarantee renormalizability and.10) with kν leads to M 2k · = 0 which implies (for M 2 = 0!) k · = 0.12) Equation (19. To get an idea of why this might be so. in the rest frame of the Copyright 2004 IOP Publishing Ltd .1.89) for example.and y-directions are ‘transverse’ while the z-direction is ‘longitudinal’. 0.11) (19.5).6) come from? The answer is simple and important.10) Taking the ‘dot’ product of (19. as we expect for a spin1 particle. so that the leading divergence is effectively ∼ d4 k k µk ν k2 kρ kσ k2 11 // kk (19.12) is a covariant condition.9) satisfy the condition ν (−k 2 + M 2 ) + k ν kµ µ = 0. Let us take k µ = (k 0 . This will be associated with the k µ k ν terms in the numerator of (19. we are omitting the spinors altogether and we are looking only at the large-k part of the propagators). this ‘ g ’ is dimensionless but. Just as in QED. The freeparticle wave equation is (£ + M 2 )X ν − ∂ ν (∂µ X µ ) = 0 and plane-wave solutions have the form Xν = Hence. which has the effect of ensuring that there are just three independent polarization vectors. Where have these dangerous powers of k in the numerator of (19.7) which is quadratically divergent and. exactly what we would get in a ‘four-fermion’ theory—see (11. This strongly suggests that the theory is non-renormalizable. (19.

±1 Consider now the propagator for a spin. ¯ (19. (19.14) (19. 1).20) Equation (7. it is straightforward to verify the result (problem 19. The ’s are ‘orthonormalized’ so that (cf (7.17) as before. λ = 0) = M −1 (| k|.18) Note that k · µ (k. but the longitudinal polarization vector becomes (problem 19. λ = ±1) = µ (krest .19): in particular. we can Lorentz transform µ (krest . (19.2) µ (k. 0. the dangerous factor k µ k ν /M 2 can be traced to the spin sum (19. 0). s) = (k + m)αβ . − m2 + i (19.22) arises from the spin sum (19. (19.20) arises from the spin sum / u α (k.12) reduces to three independent ’s as µ 0 = 0 and we may choose (19. λ) as required. (19. For the result in a general frame. Copyright 2004 IOP Publishing Ltd . ±i. λ = ±1) (19. 0. k 0 ).22) and the numerator in (19.21) s In just the same way.X. s)u β (k. 0) along the z -axis. λ) = (0.19) λ=0.16) (19. λ) ν∗ (k.61) shows that the factor in the numerator of (19. 0. λ = 0) = 0 as required. | k|).15) These states have definite spin projection (λ = ±1. 0. 0. Thus. the massive spin-1 propagator is given by i[−g µν + k µ k ν /M 2 ] k2 − M 2 + i (19. given in (7. such that krest = ( M. From (19. in a frame such that k µ = (k 0 .17) and (19. 0. we find that µ (k. 0) (λ = 0) = (0. For example.60): 2 k2 / i(k + m) .1 particle.19). (λ)) with (λ = ±1) = ∓ 2−1/2 (1.83)) (λ)∗ · (λ ) = δλλ .13) (krest .3) µ (k.18). λ) = −g µν + k µ k ν /M 2 . 0.

making the theory renormalizable? The answer is yes.3. We shall not give further details here (see also section 20.75) for the propagator in such a theory.5)). Anderson (1963) pointed out that several situations in solid state physics could be interpreted in terms of an effectively massive electromagnetic field. This is the natural generalization of the spontaneous global symmetry breaking considered in chapter 17. Soon afterwards. This gives another way of seeing in what way a massless vector particle is really very different from a massive one: the former has only two (spin) degrees of freedom. instead of the three we might have expected for a vector (spin-1) particle—and as do indeed exist for massive vector particles. in such a way that the dangerous k µ k ν term can be ‘gauged away’. by choice of the gauge parameter ξ (namely by taking ξ = 1). Is there.23) This contains somewhat similar factors of k µ k ν (admittedly divided by k 2 rather than M 2 ) but they are gauge-dependent and can. and discussed its possible relevance to contemporary attempts (Sakurai 1960) to interpret the recently discovered vector mesons (ρ. ω. as section 7. φ. Does this matter? In section 11. the crucial formula is ( large values of k the longitudinal state µ (k. then. and this is the origin of the numerator factors k µ k ν / M 2 in (19. . free photons exist in only two polarization states (electromagnetic waves are purely transverse).22). as we have seen. The first serious challenge to the then widely held view that electromagnetic gauge invariance requires the photon to be massless was made by Schwinger (1962). such ‘gauging’—essentially the freedom to make gauge transformations— seems to be possible only in a massless vector theory. so as (perhaps) to render the theory renormalizable? Consider the photon propagator of chapter 7 repeated here: i[−g µν + (1 − ξ )k µ k ν / k 2 ] . any way of getting rid of the offending ‘k µ k ν ’ terms in the Xpropagator. . via the idea of spontaneous breaking of gauge symmetry. be ‘gauged away’ entirely. k2 + i (19.1 showed.22). it is clear that Anderson had his doubts about the hadronic application. λ = 0) is proportional to k µ .3) but merely state that theories with massive charged vector bosons are indeed non-renormalizable. . These considerations therefore suggest the following line of thought: is it possible somehow to create a theory involving massive vector bosons. By way of advance notice. as we have seen. A closely related point is that. while the latter has three. which is to be compared with (19. by letting M → 0 in (19. He outlined a general framework for treating the phenomenon of the acquisition of mass by a gauge boson. precisely Copyright 2004 IOP Publishing Ltd .) as the gauge quanta associated with a local extension of hadronic flavour symmetry. From his discussion. But. and it is not at all clear how to ‘lose’ the offending longitudinal state smoothly (certainly not.8 we explained why it is thought that the relevant theories at presently accessible energy scales should be renormalizable theories. in fact.

chapter 8). the physics is well illustrated by the non-relativistic Abelian (i. Further. a situation similar to that in the Bogoliubov superfluid. The Ginzburg–Landau field ψ is commonly referred to as the ‘macroscopic wave function’. More recently. as noted after (17. We refer interested readers to the book by Schrieffer (1964. section 17. It is certainly possible to complete the BCS theory in this way.128). we gave a brief introduction to some aspects of the BCS theory of superconductivity. certainly leads to identifiable multiplets.g. It was Weinberg (1967) and Salam (1968) who made the correct application of these ideas to the generation of mass for the gauge quanta associated with the weak force. In particular. as he remarked. we omitted completely all electromagnetic couplings of the electrons in the ‘microscopic’ Hamiltonian. There is. This terminology originates from the recognition that in the BCS ground state a macroscopic number of Cooper pairs have ‘condensed’ into the state of lowest energy. well before the appearance of BCS theory. that the multiplet structure ordinarily associated with a non-Abelian symmetry would be lost.e. an analogous ‘effective theory’ for time-dependent phenomena (at zero temperature) has been derived from a BCS-type model (Aitchison et al 1995). however. But we know that flavour symmetry. This work all relates to static phenomena. a quantity which was a measure of the ‘degree of ordering’ of a system below some transition temperature. nothing specifically relativistic about the basic mechanism involved. 19. as originally proposed by Ginzburg and Landau (1950). Our presentation is influenced by that of Anderson (1963). in certain domains of temperature and magnetic field. so as to include within the same formalism a treatment of electromagnetic effects (e. as we saw in chapter 17. all pairs having the same total momentum Copyright 2004 IOP Publishing Ltd . we shall follow a less ‘microscopic’ and somewhat more ‘phenomenological’ approach.7. In the case of superconductivity. Gorkov (1959) and others showed how the Ginzburg–Landau description could be derived from BCS theory.1 we introduced the concept of an ‘order parameter’. nor need we start with the non-Abelian case. In section 17. we shall follow a more qualitative approach.3. actually closer (at least formally) to our eventual application in particle physics. even if admittedly not exact. the Meissner effect) in a superconductor. Instead. the order parameter (in this sense) is taken to be a complex scalar field ψ. electromagnetic) case—which is nothing but the physics of superconductivity. In fact. For the moment.2 The generation of ‘photon mass’ in a superconductor: the Meissner effect In chapter 17. which are at least approximately degenerate in mass. Subsequently. This has the consequence. which has a long history in theoretical studies of superconductivity and is. gauge bosons can only acquire a mass if the symmetry is spontaneously broken. in some ways. this state is highly coherent. We were concerned mainly with the nature of the BCS ground state and with the non-perturbative origin of the energy gap for elementary excitations.because.

ψ has no knowledge of the coordinates of individual electron pairs. as an ‘empirical’ order parameter. in particular.(namely zero.24).28) We now replace |ψ|2 in (19. The Ginzburg–Landau description proceeds by considering the quantummechanical electromagnetic current associated with ψ. in particular. Copyright 2004 IOP Publishing Ltd . namely j em = −2e ∗ [ψ (∇ + 2ie A)ψ − {(∇ + 2ie A)ψ}∗ ψ].26) by n s /2 in accordance with (19. (19.25) Note that we have supplied an overall factor of −2e to turn the Schr¨ dinger o ‘number density’ current into the appropriate electromagnetic current. The quantities n c and n s will depend on temperature T .121). referring essentially to paired electrons.24) and take the curl of the resulting equation to obtain ∇ × j em = − e2 n s me B.40).25) as j em = − 2e2 me A+ 1 ∇φ |ψ|2 2e (19. These considerations suggest that a successful phenomenology can be built by invoking the idea of a macroscopic wavefunction ψ. Nevertheless. the charge associated with ψ is taken to be −2e and the mass is 2m e . it may be thought of as in some way related to the ground-state ‘pair’ expectation value introduced in (17. This current was considered in section 3.4 and is given by the gauge-invariant form of (A.26) where ψ = eiφ |ψ|. ψ is varying primarily through its phase degree of freedom φ. Note that ψ is a ‘bosonic’ quantity.7).27) (19. which can be written in this case as A → A + ∇χ φ → φ − 2eχ.29) Equation (19. tending to zero as T approaches the superconducting transition temperature Tc from below. Perhaps the single most important property of ψ is that it is assumed to be normalized to the total density of Cooper pairs n c via the relation |ψ|2 = n c = n s /2 (19. The precise connection between ψ and the microscopic theory is indirect.26) is invariant under the gauge transformation (3. describing the condensate.24) where n s is the density of superconducting electrons. we can rewrite (19. in the presence of a static external electromagnetic field described by a vector potential A. (19. 4m e i (19.140)).29) is known as the London equation (London 1950) and is one of the fundamental phenomenological relations in superconductivity. rather than its modulus |ψ|. consistently with (19. We easily verify that (19. in the case of (17. Assuming now that.

we find that λ= me n s e2 1/2 ≈ 10−8 m (19.The significance of (19. we find that e2 n s B. by comparison with the left-hand side of (19. say. screening currents are set up within the superconductor. (19. This expresses the fact that. First. The physical origin of the screening is provided by Lenz’s law: when a magnetic field is applied to a system of charged particles. With n s ∼ 4 × 1028 m−3 (roughly one conduction electron per atom).32) Next. (19. consider for simplicity one-dimensional variation 1 d2 B = 2B 2 dx λ in the half-plane x ≥ 0.29) holds.33) have the form B(x) = B 0 exp −(x/λ) (19.33) the exponentially growing solution being rejected as unphysical.31) is a very characteristic one encountered in a number of contexts in condensed matter physics. penetrates only a distance of order λ into the region x ≥ 0.29) emerges when we combine it with the (static) Maxwell equation ∇ × B = j em .31) ∇2 B = me The variation of magnetic field described by (19. Then the solutions of (19. The range parameter λ is called the screening length. The field.30) Taking the curl of (19. therefore. this is the cause of atomic diamagnetism. Here the effect is occurring on a macroscopic scale (as mediated by the ‘macroscopic wavefunction’ ψ) and leads to the Meissner effect—the exclusion of flux from the interior of a superconductor. the magnetic field will be ‘screened out’ from penetrating further into the medium.34) (19. λ2 (19. On the atomic scale. over distances of order λ from the exterior boundary of the material. we note that the quantity (e2 n s /m e ) must—in our units—have the dimensions of (length)−2.35) Copyright 2004 IOP Publishing Ltd . induced EMF’s are set up which accelerate the particles and the magnetic effect of the resulting currents tends to cancel (or screen) the applied field. Let us write e2 n s me = 1 . In this case.31). in a medium such that (19.30) and using ∇ × (∇ × B) = ∇(∇ · B) − ∇ 2 B and ∇ · B = 0. These exactly cancel—perfectly screen—the applied flux density in the interior.

Copyright 2004 IOP Publishing Ltd . in the rest frame of the X-particle in which (see equation (19. it suggests rather strongly that a second field.38) ∇2 X = M 2 X which is exactly the same in form as (19. we could take the curl of (19.31) in terms of an ‘effective nonzero photon mass’. which is due to Higgs (1964) and is essentially the local version of the U(1) Goldstone model of section 17.31) via B = ∇ × A). n s → 0 and λ becomes arbitrarily large. In particular. Now consider the static version of (19. Taking the divergence via ∂ν leads to M 2 ∂ν X ν = 0 (cf (19. so that flux is no longer screened. is an essential element in the story (here. It is quite simple to interpret equation (19. how is all the foregoing related to what we learned in chapter 17 about spontaneous symmetry breaking? Where is the Goldstone mode? Is it really all gauge invariant? And what about Lorentz invariance? Can we provide a Lagrangian description of the phenomenon? The answers to these questions are mostly contained in the model to which we now turn. a mass has the ¯ dimension of an inverse length and so we naturally expect to be able to interpret λ−1 as an equivalent mass (for the photon. The connection is made precise by making the association M2 = e2 n s me = 1 . Equation (19.37).36) which simply expresses the fact that each component of X ν has mass M . many questions remain unanswered so far. Consider the equation (19.which is the correct order of magnitude for the thickness of the surface layer within which screening currents flow.39) Equation (19.8) for a free massive vector field.8) can be written as (£ + M 2 ) X ν = 0 (19.37) (19. As T → Tc . For example.39) shows very directly another way of understanding the ‘screening length ↔ photon mass’ connection: in our units h = c = 1. in this case). in addition to the electromagnetic one.13)) the ν = 0 component vanishes. it is the ψ field).11)) and so (19.31) (if X were the electromagnetic field A. and over which the applied field falls to zero. λ2 (19. This treatment conveys much of the essential physics behind the phenomenon of ‘photon mass generation’ in a superconductor.38) to obtain (19. Nevertheless. This provides a partial answer to the puzzle about the discontinuous change in the number of spin degrees of freedom in going from a massless to a massive gauge field: actually. some other field has to be supplied.37) reduces to (19.5.

40).42) Before proceeding any further. This current can be obtained just as in (7.84). extended so as to be locally.41) can be compensated by an appropriate transformation (19. we need only replace the ∂ ’s by D ’s according to the rule (7. we note at once that this model contains four ˆ field degrees of freedom—two in the complex scalar Higgs field φ and two in ˆ µ.44) .40) (19.5 that the form of the potential terms in (19. This is the first unexpected result in the local case. which only appears after making a ‘shift to the classical minimum’.42) on Aµ .69) and (17. Somehow the system must keep track of the fact that we started with four degrees of freedom. However.3 Spontaneously broken local U(1) symmetry: the Abelian Higgs model ˆ This model is just ÄG of (17. 4 4 (19.41) when accompanied by the gauge transformation on the potentials 1 ˆ ˆ ˆ ˆ Aµ (x) → A µ (x) = Aµ (x) + ∂ µ α(x).40) (specifically the µ2 one) does not lend itself to a natural particle interpretation. namely ˆ ˆ ˆ ˆ φ(x) → φ (x) = e−iα(x)φ(x) (19. namely ν ˆ ˆ £ Aν − ∂ ν (∂µ Aµ) = jˆem (19. To see what is going on.120) and add the Maxwell piece. q (19. the massless gauge field A We learned in section 17. But there is a remarkable difference between the global and local cases. the phase of φ is completely arbitrary. Copyright 2004 IOP Publishing Ltd (19. a degree of freedom (the Goldstone mode) cannot simply disappear. rather than merely globally. In the ˆ present (local) case. Due originally to Higgs (1964). This must mean that there is somehow no longer any physical manifestation of the massless mode. it provides a deservedly famous and beautifully simple model for investigating what happens when a gauge symmetry is spontaneously broken. We may also be reminded of our desire to ‘gauge away’ the longitudinal polarization states for a ‘massive gauge’ boson: we shall return to this later.84) can be ‘gauged away’ ˆ was precisely the Goldstone field. let us study the field equation ˆ for Aν . ˆ To make (17.77).72). in the global altogether.137) and is given by ˆν ˆ ˆ ˆ ˆ ˆ ˆ ˆ jem = iq(φ † ∂ ν φ − (∂ ν φ † )φ) − 2q 2 Aν φ † φ. U(1) invariant. the field θ in (17. since any change in α ˆ ˆ of (19.40) is invariant under the local version of (17. as in (17.43) ˆν where jem is the electromagnetic current contained in (19. if we choose! But θ case. This produces ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ÄH = [(∂ µ + iq Aµ)φ]† [(∂µ + iq Aµ)φ] − 1 Fµν F µν − 1 λ(φ † φ)2 + µ2 (φ † φ).19. Thus.69) locally U(1) invariant. ˆ ˆ leaving ÄH the same as before.

48) shows how the ˆ phase degree of freedom of the Higgs field φ has been incorporated into a new ˆ ν . as in (17.84)) 1 ˆ ˆ ˆ φ(x) = √ (v + h(x)) exp(−iθ (x)/v) (19.26). we obtain ˆ ˆ £ Aν − ∂ ν ∂µ Aµ = −v2 q 2 ∂νθ ˆ Aν − vq .81). First. The foregoing analysis shows us two things. equation (19. (19. So we can certainly regard the whole expression ( Aν − ∂ ν θ /vq) as a perfectly ˆ acceptable gauge field.47) following. see Higgs (1966). has ‘three’ spin degrees of gauge field A freedom.46) is indeed a relativistic analogue of (19. where they ˆν would represent interactions) and placing this jem in (19.45) 2 √ 1 into (19.48) Then. for arbitrary ˆ ˆ α. in the relativistic context.48) will contribute something Copyright 2004 IOP Publishing Ltd .40) where v/ 2 = 21/2 |µ|/λ 2 is the position of the minimum of the classical potential as a function of |φ|. Retaining just the linear terms in (19. the current (19.4) ˆν jem = −v 2 q 2 ˆ ∂νθ ˆ Aν − vq + terms quadratic and cubic in the fields.47) is ˆ invariant under (19. with mass M = vq! This fundamental observation was first made. in that it provides a ‘screening’ (mass generation) effect on the gauge field. interpreting ψ as essentially playing the role of ‘φ ’ 2 the role of v 2 . which is massive and. Higgs (1964) and Guralnik et al (1964).8) for a free massive vector field.47) ˆ Now a gauge transformation on Aν has the form shown in (19.50) But (19. Aν and θ . we can go further. If we imagine plane-wave solutions for ˆ ˆ ˆ ˆ A ν . (19.50) is nothing but the equation (19. for a full account.46) (the others and |ψ| would appear on the right-hand side of equation (19. by Englert and Brout (1964). we see that the ∂ ν θ /vq part of (19.49) (19. since we know (or can easily verify) that the left-hand side of (19. Let us define ˆ ∂νθ ˆ ˆ . In fact.46) The linear part of the right-hand side of (19. Second.26). the resulting equation for A ν is ˆ ˆ ˆ £ A ν − ∂ ν ∂µ A µ = −v2 q 2 A ν or ˆ ˆ (£ + v 2 q 2 ) A ν − ∂ ν ∂µ A µ = 0.We now insert the field parametrization (cf (17. A ν = Aν − vq (19.43). We obtain (problem 19. therefore.42).46) is directly analogous to the nonˆ relativistic current (19.42). (19.

3.54) ˆ we see that the phase of φ in (19. In superconductivity. It follows that in order to interpret the relation (19. as Bardeen (1957) first recognized. When this aspect of the dynamics of the electrons is included. to the condition ˆ ∂µ A µ = 0 (19.48) to make the massive vector ˆ field A ν .50).proportional to k ν / M to the polarization vector of A ν (recall M = v q ).51). we know from Lorentz covariance that all the components (transverse and longitudinal) of the vector field must have the same mass and this has. after equation (17.46). of course.52) ˆ ˆ ˆ In going from the situation described by Aµ and θ to one described by A µ alone via (19. The Goldstone field disappears as a massless degree of freedom. We may therefore say that the massless gauge field ˆ ˆ Aν has ‘swallowed’ the Goldstone field θ via (19. the longitudinal part of A couples strongly to longitudinal excitations of the electrons: primarily. But the transverse and longitudinal degrees of freedom respond differently in the non-relativistic (superconductor) case. This is a high-frequency mode and is essentially the one discussed in section 17. the choice of gauge which takes the Copyright 2004 IOP Publishing Ltd . regard Aµ as being in a gauge specified by 1 ˆ 1 ˆ ˆ £θ = M £θ. in this choice ˆ of gauge.51) ˆ on A µ .45). via its gradient. We return to equations (19. and reappears. In this (relativistic) case. a fully gauge-invariant description of the electromagnetic properties of superconductors.48).42)) ˆ α(x) = −θ (x)/v. but then the ˆ µ . ˆ ˆ Recalling then the form of the associated local phase change on φ(x). emerged automatically from our covariant treatment. to be consistent with (19.50) leads. chapter 8). to the collective density fluctuation mode of the electron system—that is.45) has been reduced to zero.48) as a gauge ˆ ˆ transformation on Aν we must.48)–(19. satisfying the massive equation (19. as the longitudinal part of the massive vector field. In this way the four degrees of freedom are all now safely accounted for: three are ˆ ˆ in the massive vector field A ν and one is in the real scalar field h (to which we shall return shortly).50) (ignoring vector field we must use is A other interactions). ∂µ Aµ = vq (19. Taking the divergence of (19. Thus it is indeed possible to ‘gauge θ away’ in (19. But this is exactly the (large k ) behaviour of the longitudinal polarization vector of a massive vector particle. is obtained (Schreiffer 1964. within the BCS theory. ˆ ˆ ˆ ˆ φ(x) → φ (x) = e−iα(x)φ(x) (19. to plasma oscillations. as we have seen.2. There. we have evidently chosen a gauge function (cf (19.53) (19.

That the propagator is gauge dependent should. In the next section we shall discuss a subtlety in the argument which applies in the case of real superconductors.macroscopic wavefunction to be real (i. furthermore.50). even though this must be an unphysical degree of freedom. are renormalizable (’t Hooft 1971b).5 that the equation of motion.55). and the second √ line is the Lagrangian of a scalar particle of mass 2µ. and hence the propagator for the vector particle depend on the choice of gauge.1 would be completely unresolved. The fact that this ‘Higgs mechanism’ leads to a massive vector field can be ˆ seen very economically by working in the particular gauge for which φ is real and inserting the parametrization (cf (19. on reflection.e. The necessity for the existence of one or more massive scalar particles (‘Higgs bosons’) when a gauge symmetry is spontaneously broken in this way was first pointed out by Higgs (1964).5). After all. if the massive vector boson generated in this way were simply described by the wave equation (19. Feynman graphs involving quanta corresponding to the degree of freedom associated with the phase field ˆ θ will have to be included for a consistent theory.55) ˆ into the Lagrangian ÄH .45)) 1 ˆ ˆ φ = √ (v + h) 2 (19. the Maxwell part with the addition of a mass term (note that the sign of the mass term is correct for the spatial (physical) degrees of freedom). as follows from the fact that a gauge can be chosen in which this field vanishes. ˆ ˆ We may now ask: what happens if we start with a certain phase θ for φ but ˆ ˆ do not make use of the gauge freedom in Aν to reduce θ to zero? We shall see in section 19. ˆ However. Copyright 2004 IOP Publishing Ltd .e. The latter is the mass ˆ of excitations of the Higgs field h away from its vacuum value (compare the global U(1) case discussed in section 17. before discussing other gauges than the one in which φ is given by (19.56) is exactly the Lagrangian for a spin-1 field of mass v q — i. we first explore another interesting aspect of superconductivity.56) The first line of (19. Retaining only the terms quadratic in the fields. 2 (19. in which the vector particles acquire mass through the spontaneous symmetry breaking mechanism. come as a relief. and which leads to the phenomenon of flux quantization. φ = 0 in (19.50). one finds (problem 19. a different choice of gauge from that which renders φ real has precisely the effect of ameliorating the bad high-energy behaviour associated with (19.5) that ˆ Äquad = H ˆ ˆ ˆ ˆ ˆ ˆ − 1 (∂µ Aν − ∂ν Aµ )(∂ µ Aν − ∂ ν Aµ ) + 1 q 2 v 2 Aµ Aµ 4 2 ˆ ˆ ˆ + 1 ∂µ h∂ µ h − µ2 h 2 . all the troubles with massive vector particles outlined in section 19. ˆ As we shall see. This is ultimately the reason for the following wonderful fact: massive vector theories.26)) is called the ‘London gauge’.

2. the material is superconducting and the field dies off over the characteristic penetration length λ. Our discussion of superconductivity so far has dealt.4 Flux quantization in a superconductor Though a slight diversion. had been drilled in a block of type-I material. in the form of thin filaments of flux. while we have a number of relevant results assembled. with only one class of superconductors. a more precise criterion is that type-II behaviour holds if ξ < 21/2 λ: both ξ and λ are. Apart from its intrinsic interest. Outside the core of the filaments. when an external magnetic field of less than a certain critical value is applied.Figure 19. There is a quite separate class—type-II superconductors—which allow partial entry of the external field. 19. Around each filament of magnetic flux. thereby producing a non-simply connected geometry.7. It is as if numerous thin cylinders. of course. called type I: these remain superconducting throughout the bulk of the material (exhibiting a complete Meissner effect). the phenomenon may also provide a useful physical model for the ‘confining’ property of QCD. Thus. temperature-dependent. The behaviour of |ψ| and B in the vicinity of a flux filament is shown in figure 19. According to the Ginzburg–Landau (GL) theory. where ξ is the ‘coherence length’ of section 17. Within each filament the field is high and the material is not superconducting.7. it is convenient to include a discussion of flux quantization at this point. in fact. In real superconductors. whereas for simple Copyright 2004 IOP Publishing Ltd . there circulates a vortex of screening current: the filaments are often called vortex lines. each enclosing flux. Magnetic field B and modulus of the macroscopic (pair) wavefunction |ψ| in the neighbourhood of a flux filament. For type-II behaviour to be possible. screening currents are associated with the macroscopic pair wavefunction (field) ψ . |ψ| must vanish at the centre of a flux filament and rise to the constant value appropriate to the superconducting state over a distance ξ < λ.2. already discussed in section 16.

An important result is that the flux through a vortex line is quantized. ones with holes in them. It is interesting to consider now a situation—so far entirely hypothetical—in which a magnetic monopole is placed in a superconductor.57) as before. The phenomenon is seen in non-simply connected specimens of type-I superconductors (i. and in the flux filaments of type-II materials: in the latter case each filament carries a single flux quantum 0 . If the wavefunction ψ is singlevalued. as usual).e.61) where 0 = 2π/q is the flux equation (or 2π h /q in ordinary units). Rearranging.type-I superconductivity. Far enough away from the vortex. but in (19. the monopole strength gm had to satisfy the ‘Dirac quantization condition’ qgm = n/2 Copyright 2004 IOP Publishing Ltd (19.60) where [φ] is the change in phase around . for consistency with quantum mechanics. It ¯ is not entirely self-evident why ψ should be single-valued. we obtain the result that the flux through any surface spanning is quantized: = B·dS = 2πn =n q 0 (19.58) as in (19.59) around any closed loop superconductor. |ψ| is simply set equal to a constant. the screening currents j em will have dropped to zero. we have ∇φ m . the change in phase [φ] for any closed path can only be zero or an integer multiple of 2π. Transforming the left-hand side of (19. which encloses a flux (or vortex) line. Solutions of the coupled GL equations for A and ψ can be obtained which exhibit this behaviour. and hence A·ds = 1 q ∇φ · d s = 1 [φ] q (19.62) .58) we are leaving the charge parameter q undetermined for the moment: the mass parameter m will be unimportant. such as a ring).26).60) by Stoke’s theorem. but experiments do indeed demonstrate the phenomenon of flux quantization.59) A = − 2 2 j em + q |ψ| q in the type-II Let us integrate equation (19. To see this. Dirac showed (1931) that. in the type-II case |ψ| has the variation shown in this figure. we write ψ = eiφ |ψ| (19. (19. in units of 0 with |q| = 2e (which may be interpreted as the charge on a Cooper pair. The expression for the electromagnetic current is j em = − q2 m A− ∇φ q |ψ|2 (19.

62) that the flux 4πgm out of any closed surface surrounding the monopole is quantized in units of 0 . the Maxwell equations have such a symmetry when monopoles are present. other schemes are also possible. as was first pointed out by Nambu (1974). whose symmetry is spontaneously broken to an electromagnetic U(1)em gauge group. solutions of the magnetic monopole type do occur in the case of non-Abelian gauge field theories. As the distance between the pair—the length of the filament—increases. The group structure of SU(3) is also quite different from that of U(1) models. It follows from (19. Indeed. This is the basic model which. Whether in fact the real QCD vacuum (ground state) is formed as some such coherent plasma of monopoles. The configuration is stable because there is no way for the flux to leak away: it is a conserved quantized quantity. Just this circumstance can arise in a grand unified theory which contains SU(3)c and a residual U(1)em . these monopole solutions provide an illuminating way of thinking about charge quantization: as Dirac (1931) pointed out. in one way or another. with confinement of electric charges and flux. so does the energy of the filament. that charge is quantized. provides a model of a meson. As so often stressed. For the eventual application to QCD. The monopole–antimonopole pair in a type-II superconducting vacuum. Now magnetic monopoles do not occur naturally as solutions of QED: they would have to be introduced by hand. one will want (presumably) particles carrying non-zero values of the colour quantum numbers to be confined. from his quantization condition (19. These E filaments would then terminate on electric charges.143)).62). and we do not want to be restricted just to static solutions (as in the GL theory. therefore. A vacuum which confined filaments of E rather than B may be formed as a coherent state of condensed magnetic monopoles (Mandelstam 1976. however. rather than of magnetic charge. Hence. joined by a quantized magnetic flux filament. at a rate proportional to its length. underlies many theoretical attempts to understand confinement. Incidentally. This is exactly the kind of linearly rising potential energy required by hadron spectroscopy (see equation (16. since the flux cannot spread out in directions transverse to the filament.e. the difficulty lies in the non-perturbative nature of the confinement problem. they carry an SU(3)c index). is a subject of continuing research. a flux filament in a superconductor can originate from. or be terminated by.where q is any electronic charge and n is an integer. The essential feature of the superconducting ground state was that it involved the coherent state formed by condensation of electrically charged bosonic fermion pairs. a Dirac monopole (with the appropriate sign of gm ). When these ideas are applied to QCD. here used as an analogue). We imagine. E and B must be understood as the appropriate colour fields (i. Remarkably enough. interchanging the roles of magnetism and electricity in all of the foregoing. ’t Hooft 1976b). the existence of just one monopole implies. Copyright 2004 IOP Publishing Ltd . These quantum numbers are the analogues of electric charge in the U(1) case.

counted either ˆ ˆ ˆ as two for the original massless Aν plus one each for θ and h .68) .43) and (19. by the decomposition ˆ ˆ ˆ φ = 2−1/2 [v + χ1 (x) + iχ2 (x)] (19. as was done in section 19. there seem ˆ ˆ to be three degrees of freedom for the massive field Aν and one for each of χ1 and χ2 . in section 7. The operator on the left-hand side can be simply inverted. Actually. If we take this new equation at face value.64) This was the gauge that enabled us to transform away the phase degree of freedom and reduce the equation of motion for the electromagnetic field to that of a massive vector boson. Copyright 2004 IOP Publishing Ltd (19.3. this certainly cannot be right. At first sight this just looks like the equation of motion of ˆ an ordinary massive vector field Aν coupled to a rather complicated current.67). the consequence of the unwanted degree of freedom is quite subtle. However.44)) ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ £ Aν − ∂ ν (∂µ Aµ) = iq[φ †∂ ν φ − (∂ ν φ)† φ] − 2q 2 Aν φ † φ leads to the equation of motion ˆ ˆ (£ + M 2 ) Aν − ∂ ν (∂µ Aµ ) = − M∂ ν χ2 + q(χ2 ∂ ν χ1 − χ1 ∂ ν χ2 ) ˆ ˆ ˆ ˆ ˆ 2 ˆν 2 2 − q A (χ + 2v χ1 + χ ) ˆ ˆ ˆ (19. In terms of (19. or as three for the ˆ ˆ massive A ν and one for h .2. as we can see by a count of the degrees of freedom. we know perfectly well that we can make use ˆ ˆ of the freedom gauge choice to set χ2 to zero.67) 1 2 (19.19. but it is basic to all gauge theories and already appeared in the photon case. The difficulty arises when we try to calculate the ˆ propagator for Aν from equation (19. Substituting this form for φ into ˆ ˆ ˆ the master equation for Aν (obtained from (19. making five in all. we now choose to parametrize ˆ φ .1. say. quite generally. In the previous gauge we had four degrees of freedom. reducing φ to a real quantity ˆ and eliminating a spurious degree of freedom: we have then returned to the form (19.63) we do not choose the gauge (cf (19.5 ’t Hooft’s gauges We must now at last grasp the nettle and consider what happens if.55).66) with M = qv .65) ˆ where the vacuum values of χ1 and χ2 are zero. in the parametrization ˆ ˆ ˆ φ = |φ| exp(iθ (x)/v) (19. to yield (apparently) the standard massive vector boson propagator i(−g µν + k µ k ν /M 2 )/(k 2 − M 2 ).52)) ˆ ˆ ∂µ Aµ = £θ /M.67). (19. Instead of using the modulus and phase as the two independent ˆ degrees of freedom for the complex Higgs field φ .

of course.71) is precisely (with a lowered index) the one we needed for the photon propagator in (7.6) that ˆ (19.5 and perform the sum. therefore. ˆ In figure 19. However.69) confirms that χ1 is a massive field with mass 2µ (like the h in ˆ (19. while χ2 is massless. the momentum 4-vector is the only quantity that can couple to the vector index of ˆ the vector boson. i/ k 2 . ˆ Figure 19.12) and (11. Inserting the expressions for the propagators and vector–scalar coupling and keeping track of the indices. it does not exist.ˆ ˆ Figure 19. we are interested in pursuing the case χ2 = 0.70) ˆ ÄH = 1 ∂µ χ1 ∂ µ χ1 + 1 ∂µ χ2 ∂ µ χ2 − µ2 χ1 + cubic and quartic terms. This is the term − M∂ ν χ2 .69) √ ˆ Equation (19. there is also ˆ a term involving only one field.3. ˆ However. This can ˆ ˆ be easily found by substituting (19.7) −g µλ + k µ k λ /M 2 ν (gλ − k ν kλ /k 2 )−1 (19.4.4 the only unknown factor is the propagator for χ2 . Clearly.4 as in figure 19.65) into ÄH and examining the part which is quadratic in the χ ’s: we find (problem 19. (19. The existence of this coupling shows that the propagators of Aν and χ2 are necessarily mixed: the complete vector propagator must be calculated ˆ by summing the infinite series shown diagrammatically in figure 19. The χ2 propagator is. In momentum space this corresponds to a coupling strength −ik ν M and an associated vertex as shown in figure 19.4.88)—and.56)). we can formally sum the series by generalizing the well-known result ((cf 10.3. Series for the full A ν propagator. But the inverse required in (19. This complication is. Now that ˆ ˆ all the elements of the diagrams are known.71) i k2 − M 2 for the full propagator. we finally arrive at the result (problem 19. At last the fact that we are dealing with a gauge theory has caught up with us! Copyright 2004 IOP Publishing Ltd . completely eliminated by the gauge choice χ2 = 0. A ν –χ2 coupling. which tells us that ˆ Aν actually couples directly to the scalar field χ2 via the gradient coupling (− M∂ ν ). the current on the right-hand side is rather peculiar: instead of having ˆ only terms corresponding to Aν coupling to two or three particles. as we saw there. ˆ ˆ ˆ ˆ ˆ2 2 2 Diagrammatically. we rewrite the propagator of figure 19.26)) (1 − x)−1 = 1 + x + x 2 + x 3 + · · · . for a scalar particle.

75) This propagator is very remarkable. For any finite ξ .72) into (19. As we saw in section 7. Formal summation of the series in figure 19. A full treatment (’t Hooft 1971b) shows that this pole is always cancelled by an exactly similar pole in the propagator for the χ2 ˆ 1 We shall not enter here into the full details of quantization in such a gauge: we shall effectively treat (19. Inserting (19.2. Note. so that Higgs particles were not present and the physical limit was obtained only as ξ → 0. (19. the high-energy behaviour of the gauge boson propagator is actually ∼ 1/k 2 . 2 A vector boson propagator of similar form was first introduced by Lee and Yang (1962) but their discussion was not within the framework of a spontaneously broken theory.76) is seen to correspond to the limit ξ → ∞ and. This condition is manifestly covariant and. however. k2 − ξ M 2 (19. A clever way to do this in the present (spontaneously broken) case was suggested by ’t Hooft (1971b). it effectively reduces the degrees of freedom by one. the high-energy disease outlined in section 19.73) 2 ˆν 2 2 − q A (χ1 + 2v χ1 + χ2 ).72) as a classical field relation.75) at k 2 = ξ M 2 : this is surely unphysical since it depends on the arbitrary parameter ξ .Figure 19. Copyright 2004 IOP Publishing Ltd . that there seems to be another pole in the propagator (19.72) where ξ is an arbitrary gauge parameter1 (not to be confused with the superconducting coherence length).1 appears to threaten renormalizability (in fact.5. it can be shown that there is a consistent set of Feynman rules for this gauge and the theory is renormalizable thanks to many cancellations of divergences).2 The standard massive vector boson propagator i(−g µν + k µ k ν /M 2 )(k 2 − M 2 )−1 (19.67). which is as good as the renormalizable theory of QED (in Lorentz gauge).8) and yields the general form for the gauge boson propagator i −g µν + (1 − ξ )k µ k ν (k 2 − M 2 )−1 . moreover. His proposal was to set ˆ ∂µ Aµ = Mξ χ2 ˆ (19. however. we obtain ˆ ˆ (£ + M 2 ) Aν − ∂ ν (∂µ Aµ )(1 − 1/ξ ) = q(χ2 ∂ ν χ1 − χ1 ∂ ν χ2 ) ˆ ˆ ˆ ˆ (19. in this gauge. to obtain a well-defined gauge field propagator we need to fix the gauge.74) ˆ ˆ ˆ The operator appearing on the left-hand side now does have an inverse (see problem 19.3.4.

This is. ˆ As noted there. just the local version of the model studied in section 17. The general non-Abelian case was treated by Kibble (1967). and also global U(1) transformations (17. These finite-ξ gauges are called R gauges (since they are ‘manifestly renormalizable’) and typically involve unphysical Higgs fields such as χ2 .77) ˆ so that the SU(2) covariant derivative acting on φ is as given in (13. which is the one relevant to the electroweak theory. and one U(1) gauge ˆ ˆ field B µ (x).81) . of course. in the local version.75) for the gauge boson propagator the limit M → 0 may be safely taken (compare the remarks about this limit for the ‘naive’ massive vector boson propagator in section 19. it is easiest to use the U-gauge Feynman rules: the technical difficulties with this gauge choice only enter in loop calculations. 3). therefore. exactly as in equation (7. Thus.79) 4 4 4 (19.6 Spontaneously broken local SU(2)×U(1) symmetry We shall limit our discussion of the spontaneous breaking of a local non-Abelian symmetry to the particular case needed for the electroweak part of the Standard Model.101).1).23).field itself.97) is invariant under global SU(2) transformations of the form (17. This yields the massless vector boson (photon) propagator in a general ξ -gauge.1). which we write as ig B µ /2. We recall that the scalar field φ is an SU(2)-doublet ˆ φ= ˆ φ+ ˆ φ0 (19. Note that in our master formula (19.119) or (19. 2. ˆ ÄG where ˆµ ˆ ˆ ˆ ˆ D µ φ = (∂ µ + igτ · W /2 + ig B µ /2)φ µν µ ν µ ˆ ν ν ˆ µ ˆ ˆ ˆ F = ∂ W − ∂ W − gW × W Copyright 2004 IOP Publishing Ltd 1 2 (19.100).10). in fact. we shall have to introduce three SU(2) gauge ˆµ fields (as in section 13.80) (19. The ˆ infinite-ξ gauge is known as the U gauge (U for unitary) since only physical particles appear in this gauge.6. the Lagrangian Ä of (17. for which the R-gauge choice is easier. For tree-diagram calculations. the later convenience. We now proceed with the generalization of these ideas to the non-Abelian SU(2) case. which we call Wi (x) (i = 1. namely ˆµ ˆ D µ = ∂ µ + igτ · W /2.78) being for 1ˆ ˆ ˆ ˆ ˆ ˆ µν 1 ˆ ˆ ˆ ˆ λ ˆ ˆ = ( Dµ φ)† ( D µ φ)+µ2 φ † φ − (φ † φ)2 − F µν · F − G µν G µν (19. The Lagrangian is. 19. ˆ To this must be added the U(1) piece.

45)). we should be left with three massive boson gauge bosons (which will be the W± and Z0 ) and one massless gauge boson. which we have already considered in the global case in section 17.55).87) 1 ˆ √ (v + H (x)) 2 Copyright 2004 IOP Publishing Ltd . t3 anticipating that it is the weak. As pointed out there.86) φ = exp(−iθ (x) · τ /2v) √ 1 ˆ (v + H (x)) 2 (compare (19.83) implies ˆ ( 1 + t3 2 ) 0|φ|0 = 0 2 and hence ˆ ˆ 0|φ|0 → ( 0|φ|0 ) = exp[iα( 1 + t3 2 (1/2) (1/2) (1) (19. As in (17. isospin—see chapter 21). This means that we do not want to give a vacuum value to a charged field (as is done in the BCS ground state). The choice suggested by Weinberg (1967)) was ˆ 0|φ|0 = 0 √ v/ 2 (19.107). as usual.and ˆ ˆ ˆ G µν = ∂ µ B ν − ∂ ν B µ . However. Put differently. and it is simplest to examine the particle spectrum in this (unitary) gauge. we certainly do not want a ‘superconducting’ massive photon to emerge from the theory in this case. (19.83) implies that the vacuum remains invariant under the combined transformation of ‘U(1) + third component of SU(2) isospin’—that is. (19. we do want it to behave as a ‘weak’ superconductor. The essential point for the electroweak application is that. we can reduce the phase fields θ to zero by an appropriate gauge transformation. (19. However. generating mass for W± and Z0 .107) but just as in ˆ (19.84) ˆ ˆ )] 0|φ|0 = 0|φ|0 (19. as the physical vacuum is not an electromagnetic superconductor.83) in order to see the physical particle spectrum.83) √ √ where v/ 2 = 2µ/λ1/2 . where.85) = τ3 /2 (we are using lower case t for isospin now. we parametrize these conveniently as 0 ˆ ˆ (19. in contrast to (17. this time. We may reasonably guess that the massless boson will be associated with a symmetry that is unbroken by the vacuum expectation value.82) We must now decide how to choose the non-zero vacuum expectation value that breaks this symmetry. the photon. rather than hadronic. Substituting 0 ˆ φ= (19.6. after symmetry breaking. We now need to consider oscillations about (19.

91) (19.94) (19.95) . we have Copyright 2004 IOP Publishing Ltd (19.88) tells us that we have a scalar field of mass 2µ (the Higgs ˆ ˆ boson. again). kinetic energies or mass terms).90) together with the orthogonal combination ˆ ˆµ ˆ Aµ = sin θW W3 + cos θW B µ . together with four φ-fields. But they can easily be unmixed by noting that the last term in (19.93) 4 8 4 where Thus MZ = 1 v(g 2 + g 2 )1/2 = MW / cos θW 2 and MA = 0. (19.e.88) √ The first line of (19. which is eight degrees of ˆ freedom in all.79)—three massless W ’s and one massless B. (19. This suggests introducing the normalized linear combination ˆ ˆµ ˆ Z µ = cos θW W3 − sin θW B µ where cos θW = g/(g 2 + g 2 )1/2 sin θW = g /(g 2 + g 2 )1/2 (19.96) Counting degrees of freedom as in the local U(1) case.into (19. W3 ) acquire a mass (cf (19. we originally had 12 ˆ ˆ in (19.9) that ˆ ˆ ˆ ˆ Äfree = 1 ∂µ H ∂ µ H − µ2 H 2 G 2 µ µ µ µ ˆ ˆ ˆν ˆ ˆ ˆ − 1 (∂µ W1ν − ∂ν W1µ )(∂ µ W1 − ∂ ν W1 ) + 1 g 2 v 2 W1µ W1 4 8 ˆ ˆ ˆν ˆ ˆ ˆ − 1 (∂µ W2ν − ∂ν W2µ )(∂ µ W2 − ∂ ν W2 ) + 1 g 2 v 2 W2µ W2 4 8 ˆ ˆ ˆν ˆµ ˆ ˆ − 1 (∂µ W3ν − ∂ν W3µ )(∂ µ W3 − ∂ ν W3 ) − 1 G µν G µν 4 4 ˆ ˆµ ˆ ˆ + 1 v 2 (g W3µ − g Bµ )(g W − g B µ ). The next two lines tell us that the components W1 and W2 of the ˆ 1 .92) ˆ ˆ ˆ Fµν = ∂µ Aν − ∂ν Aµ . we find (problem 19.79) and retaining only terms which are second order in the fields (i. W2 .89) ˆ ˆ The last two lines show us that the fields W3 and B are mixed.88) involves only the combination ˆµ ˆ g W3 − g B µ . (19. which evidently acquires a mass.88) become ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ − 1 (∂µ Z ν − ∂ν Z µ )(∂µ Z ν − ∂ ν Z µ ) + 1 v 2 (g 2 + g 2 ) Z µ Z µ − 1 Fµν F µν (19.56) in the U(1) case) ˆ ˆ triplet ( W M1 = M2 = gv/2 ≡ MW . After symmetry breaking. 8 3 (19. We then find that the last two lines of (19.

as well as those of the physical Higgs scalar σ . and this will.88) corresponds to a choice of gauge.65) and (17. Of course. which we do not wish to be non-zero.5). Copyright 2004 IOP Publishing Ltd .2 In this case the gauge boson propagators are all of the form (19. (19.3 and 19.3. as in the Abelian case. the W and Z particles have propagators of the form (19. The complete Feynman rules of the electroweak theory are given in appendix B of Cheng and Li (1984). discarding the W1 . be advantageous when doing loop calculations involving renormalization. We would then return to a general parametrization such as (cf (19. We then make the identification e = g sin θW (19.95)) ˆ φ= 0 √ v/ 2 1 +√ 2 ˆ ˆ φ2 − iφ1 ˆ3 σ − iφ ˆ (19.75) and ξ dependent.22).5. W2 pieces: ˆ Dµ φ = ∂µ + ig sin θW + ig cos θW 1 + τ3 2 ˆ Aµ 1 + τ3 2 ˆ ˆ Z µ φ.ˆ ˆ ˆ three massive vector fields W1 . as observed in (19.97). W2 and Z with nine degrees of freedom. and ˆ µ does not acquire a mass when 0|φ|0 = 0 (gauge fields coupled ˆ this is why A ˆ 0 do not become massive). one ˆ ˆ massless vector field A with two. and the A field with the massless photon. In such gauges. ˆ The identification of Aµ with the photon field is made clearer if we look at ˆ ˆ ˆ ˆ ˆ the form of Dµ φ written in terms of Aµ and Z µ . There is always the possibility of using other gauges.87). In the gauge (19.84).100) i=1. in general. and one massive scalar H .99) and add ’t Hooft gauge-fixing terms − 1 2ξ ˆµ ˆ ˆ ˆ ˆ (∂µ Wi + ξ MW φi )2 + (∂µ Z µ + ξ MZ φ3 )2 + (∂µ Aµ )2 . the Feynman rules will have to involve graphs ˆ corresponding to exchange of quanta of the ‘unphysical’ fields φi . Although certainly to unbroken symmetries of 0|φ| ˆ ˆ not unique.98) in order to get the right ‘electromagnetic Dµ ’ in (19. There will also have to be suitable ˆ ghost interactions in the non-Abelian sector as discussed in section 13. We emphasize once more that the particular form of (19. (19. We are interpreting the zero eigenvalue of (1 + τ3) as the electromagnetic charge of the vacuum. this choice of φ and 0|φ|0 is undoubtedly very economical and natural.97) τ3 − sin2 θW 2 ˆ Now the operator (1+τ3) acting on 0|φ|0 gives zero. the ˆ ˆ physical application will be to identify the W and Z fields with those physical ˆ particles. namely the unitary one (cf the discussions in sections 19.

and meeting these constraints seems to require rather elaborate kinds of models. spontaneously. the leptons are not coupled to QCD. quite simple to construct a theory ˆ in which the Higgs fields φ appear as bound. or composite. in the last part of the book. the W± and Z0 gauge bosons can be given masses.1. As we will see in chapter 22. The dynamical generation of fermion masses would. of course. We now turn. In the latter case. and which will be explained in chapter 22. Thus for both quarks and leptons a chiral-symmetry-breaking mass seems unavoidable. and even as far as the quarks are concerned we saw that some small Lagrangian mass was required (to give a finite mass to the pion. behind the Higgs sector. Copyright 2004 IOP Publishing Ltd .3. by supposing that the potential in (19. not via an explicit mass term in the Lagrangian. attributing the masses of fermions to their couplings with the Higgs field. be closely analogous to the generation of the energy gap in the BCS theory. It is. But generating masses for the gauge bosons is not the only job that the Higgs sector does in the Standard Model: it also generates masses for all the fermions. One may. for example). in a way quite analogous to that in which the nucleon acquired a mass in the linear σ -model of section 18. Peskin (1997)). One problem is that such theories are already tightly constrained by the precision electroweak experiments (see chapter 22). We have seen how. technicolour theories do offer the prospect of a new strongly interacting sector. as we saw in section 18. So we may ask: is it possible to find a dynamical theory which generates both masses for the vector bosons and for the fermions? Such theories are generically known as ‘technicolour models’ (Weinberg 1979. we know that it can be derived from ‘microscopic’ BCS theory and this naturally leads to the question whether there could be a similar underlying ‘dynamical’ theory. wonder whether this ‘mechanism’ is after all purely phenomenological. in fact. which could be probed in the next generation of colliders. somewhat akin to the GL theory of a superconductor. But such ideas take us beyond the scope of the present volume. in fact. However. to weak interactions and the electroweak theory. for example. Susskind 1979) and they have been intensively studied (see.The model introduced here is actually the ‘Higgs sector’ of the Standard Model but without any couplings to fermions. the gauge symmetry of the weak interactions is a chiral one which requires that there should be no explicit fermion masses in the Lagrangian. of course. Within the Standard Model.79) has the symmetry-breaking sign of the parameter µ2 . this must—in its turn—be interpreted as arising ‘spontaneously’ also. But. in fact. To preserve the weak gauge symmetry. one proceeds along what seems a more phenomenological route. break chiral symmetry for the quarks. We saw in chapter 18 how it is likely that the strong QCD interactions do. that is. This seems to be an ingenious and even elegant ‘mechanism’ for arriving at a renormalizable theory of massive vector bosons. states of heavy fermions.

ˆ 19.5 Insert (19.18).65) into ÄH of (19.75) without the ‘i’ (cf problem 19.56) for the quadratic terms. 19.71).69).2 Verify (19.55) into ÄH of (19. 19.4 Verify (19.Problems 19. Copyright 2004 IOP Publishing Ltd .3 Verify (19.1). −gµρ + kµ kρ /M 2 k2 − M 2 ν = gρ .40) and derive (19. 19.8 Write the left-hand side of (19. and ˜ ˆ show that the inverse of the factor multiplying Aµ is (19.1 Show that [(−k 2 + M 2 ) g νµ + k ν k µ ] 19.6 Insert (19.7 Derive (19.88). 19. ˆ 19.19).74) in momentum space (as in (19.46). 19.40) and derive the quadratic terms of (19.9 Verify (19.4)).


which have the spin 1 and obey the exclusion principle and 2 which further differ from light quanta in that they do not travel with the velocity of light. Mr Debye. it is best not to think about it. Technischen Hochschule Gloriastr. . since one would certainly have seen the neutrons by now if they existed. As the bearer of these lines. The mass of the neutrons should be of the same order of magnitude as the electron mass and in any event not larger than 0. 4 December 1930 u Dear Radioactive Ladies and Gentlemen. like the new taxes.—Unfortunately I cannot appear in T¨ bingen personally. Copyright 2004 IOP Publishing Ltd . I admit that on a first look my way out might seem to be quite unlikely.01 proton masses. and also Mr Back. the possibility that there could exist in the nuclei electrically neutral particles. how because of the ‘wrong’ statistics of the N and 6 Li nuclei and the continuous β-spectrum. who recently told me in Brussels: ‘Oh. that I wish to call neutrons. But nothing ventured nothing gained.— So. . to whom I graciously ask you to listen. W Pauli Quoted from Winter (2000). .—With my best regards to you. and the seriousness of the matter with the continuous β-spectrum is illustrated by a quotation of my honoured predecessor in office. since I am indispensable here in Z¨ rich because of a ball u u on the night of 6/7 December. I have hit upon a desperate remedy to save the ‘exchange theorem’ of statistics and the law of conservation of energy. Z¨ rich u Z¨ rich. will explain to you in more detail. . a neutron is emitted in addition to the electron such that the sum of the energies of the neutron and electron is constant.’ Therefore one should earnestly discuss each way of salvation. examine and judge it. pp 4–5.20 INTRODUCTION TO THE PHENOMENOLOGY OF WEAK INTERACTIONS Public letter to the group of the Radioactives at the district society meeting in T¨ bingen: u Physikalisches Institut der Eidg.—The continuous β-spectrum would then become understandable by the assumption that in β-decay. dear Radioactives. . Namely. your humble servant.

Pauli (in his ‘Dear Radioactive Ladies and Gentlemen’ letter) had suggested that the continuous e− spectrum in β-decay could be understood by supposing that. for many purposes. In the new Copyright 2004 IOP Publishing Ltd . not a half-odd integer. But such an approach. after which it was rapidly accepted that nuclei consisted of protons and (Chadwick’s) neutrons. though economical. would assume considerable knowledge of weak interaction phenomenology on the reader’s part. namely the terms involving just the gauge and Higgs fields. Fermi then developed his theory of β-decay. the decaying nucleus also emitted a light. Taking 14 N for definiteness. spin. at the Solvay Conference in 1933. as will be explained in section 21. electrically neutral particle. Pauli restated his hypothesis (Pauli 1934). namely electrons and protons. the couplings of these fields to the quarks and leptons. a very useful approximation to the full theory at energies well below the masses of the W± (∼80 GeV) and Z0 (∼90 GeV). This implies a half-odd integer value for the total nuclear spin. one requires 14 protons and seven electrons for the known charge of seven. of course. and proceed at once to discuss the empirical consequences. so that—if the usual ‘spin-statistics’ connection were to hold—the spin of the nitrogen nucleus should be an integer.2. In 1930. the β-spectrum problem remained and.1 . Much of what we shall discuss is still. with little motivation. b) for nuclear β-decay. as we shall see. in addition to the e− . not Fermi–Dirac statistics. This second part of Pauli’s hypothesis was quite soon overtaken by the discovery of the (real) neutron by Chadwick (1932). We could at this point simply write these couplings down. 20. Pauli regarded his hypothetical particle as a constituent of the nucleus. building on the ‘neutrino hypothesis’ of Pauli. But data from molecular spectra indicated that the nitrogen nuclei obeyed Bose–Einstein. for example). Stimulated by the discussions at the Solvay meeting. using now the name ‘neutrino’ which had meanwhile been suggested by Fermi. Besides. the problem was as follows. following a ‘semi-historical’ route (for fuller historical treatments we refer the reader to Marshak et al (1969) or to Winter (2000). in the neutrino sector especially the ‘historical’ data need to be carefully interpreted in order to understand the focus of much ongoing research.At the end of the previous chapter we arrived at an important part of the Lagrangian of the Standard Model. In this first 2 version of the proposal. Assuming that the nucleus was somehow composed of the only particles (other than the photon) known in 1930. This had the attraction of solving not only the problem with the continuous e− spectrum but a second problem as well—what he called the ‘wrong’ statistics of the 14 N and 6 Li nuclei.1 Fermi’s ‘current–current’ theory of nuclear β-decay and its generalizations The first quantum field theory of a weak interaction process was proposed by Fermi (1934a. which he called the ‘neutron’. We prefer to keep this book as self-contained as possible and so in the present chapter we shall provide an introduction to this phenomenology. However. The full electroweak Lagrangian also includes.

it was found Copyright 2004 IOP Publishing Ltd . Soon after Fermi’s theory was presented. Fermi relied heavily on the analogy with electromagnetism.2) The discovery of positron β-decay soon followed and then electron capture: these processes were easily accommodated by adding to (20. implying that the e− ν pair had also to appear in a similar (4vector) ‘current’ form. Four-fermion interaction for neutron β-decay. the ‘weak ¯ current’ had the form u p γ µ u n . Instead. picture of the nucleus.1. however. Thus a ‘current–current’ amplitude was proposed. however. with the emission of an e− ν pair. the electron–neutrino pair had somehow to be created and emitted in the transition process of the nuclear decay. it became clear that the observed selection rules in some nuclear transitions could not be accounted for by the forms (20. The ¯ lepton pair was also charged. Indeed. (20.2) its Hermitian conjugate ¯ ¯ ˆ ˆ ˆ ˆ Aψ n (x)γ µ ψp (x)ψ ν (x)γ µ ψe (x) (20. Correspondingly.2) and (20. in ‘allowed’ transitions (where the orbital angular momentum carried by the leptons is zero). of the form Au p γ µ u n u e− γµ u ν ¯ ¯ (20.Figure 20. Fermi conjectured that the nucleons participated via a weak interaction analogue of the electromagnetic transition currents frequently encountered in volume 1 for QED. The basic process was assumed to be the transition neutron→proton. The n and p were then regarded as ‘elementary’ and without structure (point-like). The whole interaction then had to be Lorentz invariant.1) where A was constant. neither the electron nor the neutrino were to be thought of as nuclear constituents.3). rather than having the ‘charge conserving’ form of u p γ µ u p for instance. The sum of (20. the process was described field theoretically in terms of the local interaction density ¯ ¯ ˆ ˆ ˆ ˆ Aψ p (x)γ µ ψn (x)ψ e (x)γµ ψν (x). in which the charge of the nucleon changed.1. when used to calculate transition probabilities in first-order perturbation theory. In this case. as shown in figure 20. Specifically. much as a photon is created and emitted in nuclear γ -decay.2) and (20. obviously. like the emission of a photon in QED. the whole process took place at a single spacetime point.3) gave a good account of many observed characteristics of β-decay.3) taking A to be real. Further.

2) and (20. the combination ¯ ¯ ˆ ˆ ˆ ˆ ψ p (x)ψn (x)ψ e (x)ψν (x) could occur. Hence some other interaction(s) must be present.2). and this assumption had been built into the structure of the proposed β-decay interactions. Note that the interaction must always be Lorentz invariant.3) imply that the nucleon spins cannot ‘flip’. it had always been assumed that any physical interaction had to be such that parity was conserved. Other combinations are also possible.6. as we shall discuss shortly. With the discovery of the muon (section 1.5) allows J = 1.4) The non-relativistic limit of (20. and also ¯ ¯ ˆ ˆ ˆ ˆ ψ p (x)σµν ψn (x)ψ e σ µν ψν (x) where σµν = i (γµ γν − γν γµ ).4) gives J = 0. following the analysis of Lee and Yang.8) Note that we have deliberately called all the neutrinos just ‘ν’. in part.1) and the pion (section 2. Hitherto. for others—of comparable strength—a change of nuclear spin by one unit ( J = 1) occurred. more weak decays became experimentally accessible. Gamow and Teller (1936) introduced the general four-fermion interaction. µ− ) might have the same form (‘universality’) but the data were incomplete and.1) that. in nuclear decays the energy release is very small (∼ few MeV) compared to the mass of a nucleon. The breakthrough came in 1956. but (20. (ν.that. such as (20. Now. Copyright 2004 IOP Publishing Ltd . For example. without any particle/anti-particle indication or lepton flavour label: we shall have more to say on these matters in section 20. but not their derivatives. (20.6) (20. for example µ decay (20.7) µ− → e− + ν + ν and π decay π − → e− + ν. in this limit. There were hopes that the couplings of the pairs (p. the interactions (20.2).5). when Lee and Yang (1956) suggested that parity was not conserved in all weak decays. It is then easy to see (problem 20. parity violation was indeed found to be a strikingly evident feature of weak interactions. Once it was looked for properly. apparently contradictory. e− ) and (ν. (20.n). constructed from bilinear combinations of the nucleon pair and of the lepton pair. Thus began a long period of difficult experimentation to establish the correct form of the β -decay interaction.4) or (20.3. while for many transitions the nuclear spin did not change ( J = 0). 2 (20.5) (20. and so the non-relativistic limit is an excellent approximation for the nucleon spinors.

20. such as spin.14) (20. A polar vector is one which transforms in the same way as the coordinate x under the parity operator P P : x → −x. the experiment of Wu et al (1957) established. By extension. specifically nuclear β-decay.11) Thus a polar vector V is defined by the behaviour P : V → −V (20.10) where v. Consider now how the distribution (20. which is in the direction opposite to the applied magnetic field. (20. 60 Co decays to 60 Ni ( J = 4). p and E are. any angular momentum. a J = 1 transition. we must now distinguish between a scalar such as U · V (the dot product of two polar vectors) and a pseudoscalar such as U · (V × W) (the triple scalar product of three polar vectors): P : U · V → +U · V P : U · (V × W) → −U · (V × W). In forming scalar products therefore. is also an axial vector. that parity was violated in a weak interaction.12). Why does this indicate parity violation? To see this. momentum and energy.2 Parity violation in weak interactions In 1957. for the first time.11) and (20.12) under parity. The experiment involved a sample of 60 Co ( J = 5) cooled to 0.13) under parity. and the results were consistent with a distribution of the form I (θ ) = 1 − J · p/E = 1 − v cos θ (20. At this temperature most of the nuclear spins are aligned by the magnetic field and so there is a net polarization J . (20. The relative intensities of electrons emitted along and against the magnetic field direction were measured. Examples are the velocity v. The vector product of two such vectors defines the behaviour of an axial vector P : U × V → (−U) × (−V ) = U × V (20. The degree of 60 Co alignment was measured from observations of the angular distribution of γ -rays from 60 Ni. an axial vector does not reverse sign under P: the most common example is angular momentum l = x × p. In contrast to (20.9) (20. we must first recall the definition of vectors (‘polar vectors’) and pseudovectors (‘axial vectors’).10) would be described in a paritytransformed coordinate system. momentum p and electromagnetic current j em . and θ is the angle of emission of the electron with respect to J .15) The scalar remains the same under P but the pseudoscalar changes sign.01 K in a solenoid. respectively. J → J and Copyright 2004 IOP Publishing Ltd . the electron speed. Applying the rules just stated.

The two are inequivalent. This is an operational consequence of ‘parity violation’. t) for observers using the transformed coordinate system such that ‘their’ Dirac equation has exactly the same form in their system as (20.4 of volume 1) if we can find a wavefunction ψP (x .17): i ∂ψP (x . t). (20.21) Comparing (20. is the appearance of the pseudoscalar quantity J · p in (20.19) Now we know that ∇ = −∇.17) will be covariant under (20.22) . Hence. in fact. it follows that we may consistently ‘translate’ between ψ and ψP using the relation ψ(x.17). in contrast to all the other coordinate system equivalences which we have previously studied (e. since x = −x. t) + βmψ(x.19) becomes i (20. employing only vector currents. The Fermi theory.p → − p so that. needs a modification to account for this result. (20. ∂t ∂ψP (x . be using.17) under the coordinate transformation (20. t)] + βm[βψ(x .16) The difference between (20. we find that i∂ [βψP (x . under three-dimensional rotations and Lorentz transformations). evidently. the distribution would have the form IP (θ ) = 1 + v cos θ. The crucial point in this example. t) + βmψP (x . by performing the measurement. t)] = −iα · ∇[βψP (x . alongside the obviously scalar quantity ‘1’. t) = −iα · ∇ψ(x. 20. t) + βmψP (x .18) (see appendix D of volume 1 and also section 4.9). (20. t)]. we need to consider the behaviour of the Dirac spinors and fermion fields under P. t) ∂t P : x → x = −x. t) Copyright 2004 IOP Publishing Ltd (20.16) and (20.g. ∂t (20. t).10) implies that. we can determine which of the two coordinate systems we must. as described by the new system. t) = −iα · ∇ ψP (x .18) Equation (20. t → t.20) Multiplying this equation from the left by β and using βα = −αβ. ∂t (20.21) and (20.3 Parity transformation of Dirac wavefunctions and field operators We consider the behaviour of the free-particle Dirac equation i ∂ψ(x. To see how this may be done. t) = βψP (−x. t) = iα · ∇ψP (x .

ψP (x. The possibility of such a phase factor did not arise in the case of Lorentz transformations. t). We have. t)ββγ β ψ2 (−x. as the electric charge density. t) = ψ1P (x. t)γ ψ(−x.25) which can be conveniently summarized by the simple statement that the threemomentum p as seen in the parity transformed system is minus that in the original one. t) = N −σ · p E+m φ σ·p E+m φ φ exp(−iEt + i p · x).11 of Bjorken and Drell (1965).23) In fact.27) showing that the µ = 0 component is a scalar under P: this is to be expected. t)βγ ψ2P (x.26) ˆ An explicit form for P is given in section 15. for µ = 0. is also a scalar. t) ˆ† ˆ = ψ1 (−x. since we can insist that for infinitesimal ones the transformed ψ and the original ψ differ only infinitesimally (not by a finite phase factor). But the parity transformation cannot be built up out of infinitesimal steps—the coordinate system is either reflected or it is not. (20. we could include an arbitrary phase factor ηP on the right-hand side of (20.24) φ exp(−iEt − i p · x) (20.23). t) ≡ Pψ(x. t) 1 (20.46): ψ(x. t)ψ2P (x. ˆ2 ˆ1 ψ ¯ ˆ† ˆ ˆ ˆ ψ 1P (x. (20.or. t). we have ¯ ˆ ˆ ˆ† ˆ ψ 1P (x.23): such a phase leaves the normalization of ψ and all bilinears of the ¯ form ψ (gamma matrix) ψ unaltered. for example. t) ˆ† ˆ = ψ1 (−x. t)ψ (−x. equivalently. (4. In the same way we can introduce the idea of a unitary quantum field operator ˆ ˆ P which transforms Dirac field operators ψ(x. t)γ 0 ψ2P (x. t) 1 2 (20. ˆ Consider now the behaviour under P of a 4-vector current of the form ¯ (x)γ µ ψ (x). t)P−1 = β ψ(−x. (20. t)γ ψ2P (x. Note that σ does not change sign. For the spatial parts. consider the free-particle solutions in the standard form (4. t) = βψ(−x. As an example of (20.28) Copyright 2004 IOP Publishing Ltd .40). t) = N Then ψP (x. t) ¯ ˆ ˆ = ψ (−x. t) = ψ1P (x. t)β · β ψ2 (−x. t) = βψ(−x. t) ¯ ˆ ˆ = − ψ (−x. as expected. t) according to ˆ ˆ ˆ ˆ ˆ ψP (x. Parity is said to be a discrete transformation in contrast to the continuous Lorentz transformations (and rotations). ρ.

32) ¯ ˆ ˆ Consider now the quantity ψ 1 (x. t)βγ5 ψ2P ( x. Copyright 2004 IOP Publishing Ltd (20.30) and γ0 = β . t). With the usual choice of the Dirac matrices used in chapter 4.33) ¯ ˆ ˆ using βγ5 = −γ5 β and β 2 = 1. t)γ5 ψ2P ( x. (20. we easily find that γ5 = 0 1 1 0 .35) . namely β= 1 0 0 −1 α= 0 σ σ 0 (20. Finally.3. like the current density j . A0 → − A0 and A → A. We call this kind of 4-vector ¯ an axial 4-vector.34) ˆ ˆ ˆ ˆ Then. we must. We recall the definition (20. A). for an ordinary 4-vector ˆ ˆ ˆ V µ = (V 0 . V ). this is done via the γ5 matrix already introduced in section 12. Similarly. For Dirac particles. under parity. t) = ψ1P ( x. t)γ ψ (− x. t)γ µ γ5 ψ2 ( x. where we are suppressing possible spacetime arguments x. Thus. Thus the spatial parts transform as a polar vector. while the spatial part is an axial vector. so as to create pseudoscalars well as scalars. γ = βα . this combination of ψ 1 and ψ2 is a pseudoscalar.29) γ5 = iγ 0 γ 1 γ 2 γ 3 .2) that the µ = 0 component of this is a pseudoscalar.using βγ = −γ β . t) ˆ† ˆ = ψ1 (− x. t)γ5 ψ2 ( x. consider the combination ¯ ˆ ˆ ψ 1 ( x.31) (20. t) ¯ ˆ ˆ = − ψ (− x. for short. have both axial and polar vectors.2. t)ββγ5 β ψ2 (− x. To accommodate parity violation. t) 1 5 2 (20. (20. the usual ‘ψγ µ ψ’ one being just a vector. t). γ5 can easily be shown to anti-commute with the other four γ matrices: {γ5 . Under the parity transformation. The reader can easily check (problem 20. this becomes ¯ ˆ ˆ† ˆ ˆ ψ 1P ( x. however. t. and most importantly. γµ } = 0. ˆ Let us write the components of an axial 4-vector Aµ as ˆ ˆ ˆ Aµ = ( A0 .

Consider. these ¯ authors proposed the form u e− γµ (1 − γ5 )u ν . For the part involving the nucleons. therefore. as we have seen. namely the V − A (‘V minus A’) structure. From our present perspective. for example. essentially the simple ‘V − A’ one. in place of the leptonic combination u e− γµ u ν . it is advantageous to work with a different representation of the Dirac matrices. In this case. in fact.2. once again. Copyright 2004 IOP Publishing Ltd .1 This V − A structure for quarks and ¯ leptons is fundamental to the Standard Model. having the form u p γµ (1 − r γ5 )u n ¯ where r had the empirical value ∼ 1.10 that a slight modification is necessary. certain combinations of V-type and A-type currents. Indeed. u u γµ (1 − γ5 )u d . We work in a representation in which γ5 is 1 We shall see in section 20. to within some experimental error) that the currents participating in Fermi’s current–current interaction are. which is that the (1 − γ5 ) factor can be thought of as acting either on the u spinor or on the u spinor. This proves to be the key to unlocking the structure of the weak interaction. a term ¯ u e− γµ (1 − γ5 )u ν . the proposal was slightly more complicated. while Aµ Aµ and Vµ V µ both transform as scalars. the hadronic transition is actually occurring at the quark level. the remarkable fact is that the appropriate current to use is. it was eventually established (always. of course. We must now at once draw the reader’s attention to a rather remarkable feature of this V − A structure. (20. Sudarshan and Marshak (1958) and then Feynman and Gell-Mann (1958) and Sakurai (1958) proposed a specific form for the current–current interaction. of course. 20. so that rather than a transition n → p we now think in terms of a d → u one. We have ¯ u e− γµ (1 − γ5 )u ν = u †− βγµ (1 − γ5 )u ν ¯ e = u †− (1 − γ5 )βγµ u ν e = [(1 − γ5 )u e− ]† βγµ u ν = [(1 − γ5 )u e− ] γµ u ν .37) To understand the significance of this.36) ˆ ˆ ˆ ˆ transforms as a pseudoscalar under parity. that one way to introduce a Lorentz invariant ˆ ˆ pseudoscalar interaction is to form the ‘dot product’ of a V µ and an Aµ type object. It follows that the Lorentz invariant product ˆ ˆ ˆ ˆ ˆ ˆ Aµ V µ = A0 V 0 − A · V (20. For example.4 V − A theory: chirality and helicity Quite soon after the discovery of parity violation. after many years of careful experiments. We learn.ˆ ˆ ˆ ˆ the components transform by V 0 → V 0 . for both nucleons and leptons. being the difference (with equal ¯ weight) of a V-type and an A-type current. V → − V under parity. and many false trails.

43) which formalizes (20.42) for any u. We may introduce the projection operators PR .chosen to be diagonal. so that writing p u= φ χ (20.31) by a unitary transformation.37) and emphasizes the fact that only the chiral L components of the u spinors enter into weak interactions. PL of section 12. In the terminology of section 12. namely −σ 0 (20. Then u 1 γµ ¯ 1 − γ5 2 2 u 2 = u 1 γµ PL u 2 = u 1 γµ PL u 2 ¯ ¯ = u 1 γµ PL u 2L = u 1 PR γµ u 2L ¯ ¯ = u † PL βγµ u 2L = u 1L γµ u 2L ¯ 1 (20.38) which is related to our ‘usual’ choice (20. positive energy spinors are defined as solutions of ( / − m)u = 0. The α and β of (20.3. To see the physical consequences of this.44) Copyright 2004 IOP Publishing Ltd . ‘(1 − γ5 )u ’ has definite chirality.15: readers who have not worked through that problem are advised to do so now. meaning that it belongs to the eigenvalue −1 of γ5 . First of all. it is clear that any combination ‘(1 − γ5 )u ’ is an eigenstate of γ5 with eigenvalue −1: γ5 = 1 0 0 −1 α= 0 −σ β= 0 1 1 0 γ = 0 σ γ5 (1 − γ5 )u = (γ5 − 1)u = −(1 − γ5 )u (20. We may also suggest a backward glance at section 12.3.40) PR PL = PL PR = 0 u L ≡ PL u u R ≡ PR PR + PL = 1 (20.3. a remarkably simple statement.2 and chapter 17. As usual. PL ≡ satisfying 2 PR = PR 2 PL = PL 1 − γ5 2 PR ≡ 1 + γ5 2 (20.38) were introduced earlier in equation (4.2. which we shall now derive explicitly.39) σ 0 2 using γ5 = 1. namely L (‘left-handed’).41) and define (20.2. we need the forms of the Dirac spinors in this new representation. for convenience.97) and problem 4.

hence.51) φ χ = 0 χ .46) σ · pφ+ = | p|φ+ while the eigenstate φ− with λ = −1 satisfies (20.51) and (20. u( p. λ = +1) = √ .50) Equations (20.46) with a minus on the righthand side.52) are very important. only the negative helicity u-spinor will enter. (20.52) √ 0 E − | p|φ+ (20. χ is to take them to be helicity eigenstates (see section 4.47) The normalization N is fixed as usual by requiring uu = 2m . using m E 2 − p2 ≈ E − | p| = √ 2E E + | p| Copyright 2004 IOP Publishing Ltd for m E.53) . λ = +1) can be written as u( p. from which it ¯ follows (problem 20. (20.38). λ = +1) = N φ+ ( E−| p|) φ+ m . the projector PL ‘kills’ the top two components: PL In particular. λ = +1) = and PL u( p. the spinor u( p. For example. (20.45) A convenient choice of two-component spinors φ .51) implies that in the limit of zero mass m (and. equation (20.49) Now we have agreed that only the chiral ‘L’ components of all u-spinors enter into weak interactions. But from the explicit form of γ5 given in (20. (20. in the Standard Model. we see that when acting on any spinor u. In particular. More quantitatively. PL u( p. Thus. Thus. (20. E → | p|). finally.3) that N = ( E + | p|)1/2 . λ = −1) = √ √ E − | p|φ− E + | p|φ− . we have √ E + | p|φ+ u( p.48) E − | p|φ+ Similarly. (20. λ = −1) = √ 0 E + | p|φ− . the eigenstate φ+ with positive helicity λ = +1 satisfies (20.3). (20.we obtain ( E − σ · p)φ = mχ ( E + σ · p)χ = mφ.

since E is clearly a scalar while σ · p is a pseudoscalar (note that when m = 0 we can infer 2 The proportionality of the negative helicity amplitude to the mass of the anti-fermion is. of course.2. which reduces via (20. Copyright 2004 IOP Publishing Ltd . of course. remember the ‘small ¯ subtlety’ to do with the labelling of v -spinors.4) the results √ − E − | p|χ− √ v( p. each of the equations in (20. (20. this would mean that the positive helicity components of all anti-fermions dominate in weak interactions. however.5. similarly. in the massless limit.38) for the Dirac matrices.56) violates parity.45) (in the m = 0 limit) to the p two independent two-component ‘Weyl’ equations. it was realized that the result could be elegantly explained by the assumption that the neutrinos were strictly massless particles (Landau 1957. In Pauli’s original letter. They satisfy ( / + m)v = 0 p and the normalization vv = −2m . Recalling the ‘hole theory’ interpretation of section 4. We must. (20. Salam 1957).5. Lee and Yang 1957. were derived in the convenient representation (20. This can be shown by using general helicity projection operators. Immediately after the discovery of parity violation.we can say that positive helicity components of all fermions are suppressed in matrix elements by factors of order m/E. the action of PL removes the top two components. discussed in section 4. helicity is Lorentz invariant. In this case. Bearing in mind that the helicity operator σ · p/| p| is a pseudoscalar. stated in italics. λ = +1) actually satisfy σ · pχ− = −| p|χ− and. the χ+ ’s in v( p. since the direction of p cannot be reversed by a velocity transformation with v < c. u and v spinors satisfy the same equation / (u or v) = 0.55) Once again. only the λ = +1 state survives.3: the two-component spinors χ in v( p. We then find (problem 20. negative helicity components being suppressed2 by factors of order m/ E . Eφ0 = σ · p φ0 Eχ0 = −σ · p χ0 . he suggested that the mass of the neutrino might be of the same order as the electron mass.3. we see that φ0 has positive helicity and χ0 negative helicity. λ = +1) = (20. they actually hold independently of any choice of representation. exactly as noted for π + → µ+ νµ decay in section 18. Furthermore. precisely related to the parity violation built into the V − A structure. this ‘unequal’ treatment for λ = +1 and λ = −1 components is. In this strictly massless case.56) Remembering that E = | p| for a massless particle. leaving the result that. We should emphasize that although these two results. A similar analysis may be done for the v -spinors.54) E + | p|χ− and v( p. λ = −1) = √ E √ + | p|χ+ − E − | p|χ+ . λ = −1) satisfy σ · pχ+ = | p|χ+ .

31) or (20.58). so as to produce the equation ∗ (iγ µ∗ ∂µ + eγ µ∗ Aµ + m)ψC = 0 (20. We now turn to the question of whether there is a distinction to be made between neutrinos and anti-neutrinos. for example. if we can find a C0 such that −1 C0 γ µ∗ C0 = −γ µ (20. we first introduce another discrete symmetry operation.60) from the left by a matrix C0 (to be determined). 20. Consider the Dirac equation for a particle of charge −e (e > 0) in a field Aµ : / / (i∂ + e A − m)ψ = 0.62) ∗ we may identify C0 ψC with ψ (up to an ever-possible phase factor). (20. that of charge conjugation C.45) that. (20. we obtain −1 ∗ [C0 γ µ∗ C0 (i∂µ + e Aµ ) + m]C0 ψC = 0. all γ -matrices are real except γ 2 which is pure imaginary. there is a sort of ‘covariance’ involved here under the transformation e → −e: we can relate ψC to ψ in such a way that (20. involving the K-capture reaction e− +152 Eu → ν +152 Sm∗ (20.59) Remarkably.59).59) follows from (20. the i being inserted for convenience so as to make C0 real. Take the complex conjugate of (20.61) Thus. Multiplying (20.38). a result taken as confirming the ‘two-component’ neutrino theory and the V − A theory. A possible choice for C0 is then iγ 2 .58) The equation satisfied by a particle of the same mass and opposite charge +e is / / (i∂ − e A − m)ψC = 0. As we have seen. Thus the (massless) neutrino could be ‘blamed’ for the parity violation. the massless limit of the (four-component) V − A theory leads to the same conclusion.from (20. (20. In either of our two representations (20.56) and with the form of β in (20. They found that the helicity of the emitted neutrino was (within errors) 100% negative.60) assuming Aµ to be real. which is assumed to be non-singular. which is consistent with (20.5 Charge conjugation for fermion wavefunctions and field operators We begin by following rather similar steps to those in section 20.3 for parity.38)). Copyright 2004 IOP Publishing Ltd . in this representation.57) as described by Perkins (2000). either positive or negative. Which helicity is actually chosen by Nature was determined in a classic experiment by Goldhaber et al (1958). neutrinos have one definite helicity. As a preliminary. φ ↔ χ under P. In this model.

70) and so the transformed wavefunction ψC is precisely the wavefunction of a positive 4-momentum solution with positive helicity (cf (20. 2 σ2 = 1. we find that iγ 2 v ∗ ( p. we see that (20. λ = −1). we find for this case that √ E + | p| φ+ 2 ∗ −i p·x ψC = iγ v ( p. to a particle ↔ anti-particle transformation. λ = −1) = u( p. Then. Thus.65) (20.54). This transformation is the nearest we can get. σ1 σ3 = −σ3 σ1 .The wavefunctions ψ and ψC are then related by ∗ ψ = iγ 2 ψC or ψC = iγ 2 ψ ∗ (20. λ = +1)ei p·x √ − E − | p|χ− √ = E + | p|χ− ei p·x (20. defining iσ2 χ+ as φ− .68) Multiplying by σ2 from the left and. we obtain (inserting a −i freely) ∗ ∗ σ · p(−iσ2 χ− ) = | p|(−iσ2 χ− ) (20. in a wavefunction theory.63) where we have used (iγ 2 )2 = I.48)). say φ+ .67) is ∗ ∗ (σ1 p1 − σ2 p2 + σ3 p3 )χ− = −| p|χ− .64) using (20. ψC = iγ 2 ψ ∗ = ∗ iσ2 0 E + | p| χ− Now the χ− here satisfies σ · pχ− = −| p|χ− ∗ ∗ σ ∗ · pχ− = −| p|χ− (20. In a similar ∗ way. √ ∗ 0 −iσ2 − E − | p| χ− √ e−i p·x .69) ∗ showing that (−iσ2 χ− ) is a spinor with positive helicity.67) and. For the latter. taking ψ to be a negative 4momentum solution of the free-particle Dirac equation with a helicity λ = +1 v -spinor: ψ = v( p. recalling that σ1 and σ3 are real while σ2 is pure imaginary. our transformation takes negative 4-momentum solutions with helicity λ into positive 4-momentum solutions with helicity λ. in the representation (20. we want an operator Copyright 2004 IOP Publishing Ltd . Thus.66).3.38). we find that (20. in this case. λ = +1)e e−i p·x = √ E − | p| φ+ = u( p.66) as agreed in section 20. Taking the complex conjugate of (20. using σ1 σ2 = −σ2 σ1 . λ = +1)e−i p·x (20. (20. Consider the application of this result.

59). but it does not. λ).1. since we have seen that the V− A interaction treats a positive helicity particle very differently from a negative ˆ helicity anti-particle. the k and λ values are not altered. Let us consider the effect of the transformation (20. λ)eik·x ] ˆ (20. which follows by inverting (20. ˆ ˆ Returning to (20.71) ˆ (Note that we write explicitly †T for ∗ as ψ will contain annihilation and creation operators for which ∗ is undefined). There. the required operator is C. (20. We shall discuss CP further in section 22. However.70).73) ˆ† λ ˆ ˆ from which it is clear that the field ψC (x) is just the same as ψ(x) but with cλ (k) ˆ ˆ† ˆ ˆ† replaced by dλ (k) (and cλ (k) by dλ (k))—that is.which simply changes a particle with a certain 4-momentum and helicity into the corresponding anti-particle with the same 4-momentum and helicity. We have introduced the idea of particle–anti-particle conjugation within the context of electromagnetic interaction. we see that.56) automatically incorporates CP invariance. Thus the CP-conjugate of an e− with λ = −1 is an e+ with λ = +1.72) λ where E = m 2 + k2 . it is clear that the helicity operator itself is odd under P. In particular. λ)e−ik·x ] (20. and so the V − A interaction does preserve the combined symmetry of CP. This can ˆ only be done in a quantum field formalism. as required.71) on a standard normal mode expansion of a Dirac field: ˆ ψ(x) = 1 d3 k √ 3 (2π) 2E ˆ† [cλ (k)u(k. we find that ˆ ψC (x) = 1 d3 k √ 3 (2π) 2E ˆ [cλ (k)v(k. − emitted in µ− -decay is In perhaps more physical terms. while one is precisely transformed into the other under C. particle operators replaced by ˆ anti-particle ones. λ)eik·x + dλ (k)u(k. λ)e−ik·x + dλ (k)v(k. But we must now ask whether it is also a good symmetry in weak interactions. This is just what we need to describe fermions which carry a conserved quantum number (such as their Copyright 2004 IOP Publishing Ltd . where it is indeed a good symmetry. Using iγ 2 u ∗ (k. and the similar relation for λ = −1. Particle–anti-particle symmetry would predict that an e+ emitted in the C-conjugate process should also have helicity λ = −1. The answer to this question must be an immediate negative. λ) = v(k.7. whether we use ψ or ψC .5) that the ‘two-component’ theory of (20. just as the C-conjugate field should be.58) and (20. we know that the e predominately in the λ = −1 helicity state. four distinct kinds of ‘modes’ are involved: there are particles with either sign of helicity and anti-particles with either sign of helicity. It may easily be verified (problem 20. with the property ˆ ˆˆ ˆ ˆ ˆ ψC ≡ Cψ C−1 = C0 ψ †T = iγ 2 ψ †T .

an almost undetectable ‘contamination’ of negative helicity component in the ‘¯ ’.74) ¯ should be observed. in this book. by which—according to some convention—‘particle’ can be distinguished from ‘anti-particle’. 20. this is. adopting a traditional approach in which the data are interpreted in terms of conserved quantum numbers carried by neutrinos. that the ‘neutrinos’ actually used are those which accompany electrons in β − -decay. ‘¯ ’. The e− in β-decay has predominately negative helicity and its accompanying ‘ν ’ has predominately ¯ positive helicity. If. Indeed. Suppose. since we have only considered charged fermions. Thus the property of the V − A interaction. Specifically. together with the very small value of the neutrino mass. a result which could clearly be interpreted as confirming the ‘conserved electron number hypothesis’. the argon being 17 18 detected through its radioactive decay. the cross-section is extremely small but by using a large enough target volume this might perhaps be compensated.electromagnetic charge).1 of volume 1. the inverse reaction ν + n → e− + p should also exist. we gave a brief discussion of leptonic quantum numbers (‘lepton flavours’). on the other hand.3. of course. the reaction (20. in principle. well established.74). We must now examine the matter more closely.3. Thus far. carrying ν a conserved lepton number.75) . which serve to distinguish neutrinos from anti-neutrinos. then the reaction ‘ν ’ +37 Cl → e− +37 Ar ¯ 17 18 (20. and the neutrino mass is less than 1 eV.74). Then. in the light of what we have learned about the helicity properties of the V − A interaction. Consider the e− capture reaction e− + p → ν + n. independently of any considerations about ‘lepton number’. as we shall now discuss. Davis found no evidence for reaction (20.74) should not be observed. The fraction of the other helicity present is of the order m/E. Davis (1955)—following a suggestion made by Pontecorvo (1948)—argued as follows.1) these are to be regarded as anti-neutrinos. If (as was supposed in section 1. But in the case of the neutral neutrinos the situation is not so clear. Hence. at the expected level of cross-section. However. this has been the case. however. the positive helicity ‘ν ’ from β − -decay will (predominately) produce a positive helicity lepton. which ¯ must be the e+ not the e− . conspire effectively to forbid (20. In 1955. therefore. the ‘helicity-allowed’ reaction ‘ ν ’ + p → e+ + n ¯ Copyright 2004 IOP Publishing Ltd (20.6 Lepton number In section 1. ν Now the property of the V − A interaction is that it conserves helicity in the zero mass limit (in which chirality is the same as helicity). the reaction ν +37 Cl → e− +37 Ar was proposed. which was. Of course. another interpretation is possible. where E ∼ few Mev. the ‘ν’ in the capture process and the ‘ν ’ in β-decay are not distinguished by the weak interaction.

all four ‘modes’—e−(λ = +1). ψM (x). in contrast to the Dirac variety.74). even though only two generally dominate in weak interactions (e− (λ = −1) and e+ (λ = +1)). e+ (λ = −1)— are experimentally accessible via electromagnetic interactions.76) which is the Majorana condition. which have all four possible modes present (two helicities. of course. will have a mode expansion of ˆ† the form (20. which is. ‘neutrinos are their own anti-particles’. if indeed they exist. the carrier of lepton number +1). can be interpreted in terms of ‘¯ ’ carrying a ν lepton number of −1. If. As we have seen. Such ‘C-self-conjugate’ fermions are called Majorana fermions. The situation may therefore be summarized as follows. in fact. Such ˆ† a field will then clearly obey the relation ˆ ˆ ˆ ˆ CψM C−1 = ψM (20. the question is: are these ‘neutrinos’ distinguished only by their helicity. In the case of e− and e+ . no quantum number—other than the helicity—exists which distinguishes the neutrino states. in fact. so small that the ‘wrong helicity’ supression factors will make it very difficult to see the presence of the possible states νe (λ = +1). to a very good approximation. the massless limit is approached smoothly and it seems highly likely that neutrino masses are. In short. In their case. seem to interact only weakly. we may if we wish say that the participating states are (in association with e− or e+ ) νe ¯ (λ = +1) and νe (λ = −1). The distinction between the ‘Dirac’ and ‘Majorana’ neutrino possibilities becomes an essentially ‘metaphysical’ one in the limit of strictly massless neutrinos. rather than of a particle and its anti-particle. Unfortunately.72) but the operator cλ will appear instead of the operator dλ . e− (λ = −1). or is there an additional distinguishing characteristic (‘electron number’)? In the latter case. then we would have to say that the C-conjugate of a neutrino state is a neutrino. e+ (λ = +1). A neutrino would be a fermionic state somewhat like a photon. also its own anti-particle. in contrast. ¯ Copyright 2004 IOP Publishing Ltd . Reaction (20. since they both conserve chirality which is the same as helicity in the massless limit. where ‘ν ’ is the helicity −1 state (or. not an anti-neutrino—that is. two particle/anti-particle). equal to that of the e+ . We would have just the two states νe (λ = −1) and νe (λ = +1) ¯ ¯ and no way of creating νe (λ = +1) or νe (λ = −1). It was also established that only ‘ν ’ produced e− via (20. the helicity rules do the job required just as well as the lepton number rules. Neutrinos. of course. νe (λ = −1). nor by any weak (or electromagnetic) interaction. The ‘− ’ label then becomes superfluous.75) too. But we may also regard these two states as simply two different helicity states of one particle. The quantum theory of free Majorana fermions is described in appendix P.was observed by Reines and Cowan (1956) (see also Cowan et al (1956)). ˆ The field operator for a Majorana fermion. since then (as we have seen) a given helicity state cannot be flipped by going to a suitably moving Lorentz frame. we should expect the ‘other’ two states νe (λ = −1) and νe (λ = +1) to exist as well as the ¯ ones known from weak interactions. on the other interpretation.

but in the second vertex the V − A interaction will ‘want’ it to have λ = −1. Again this can be interpreted either in terms of helicity conservation or in terms of conservation of a leptonic quantum number L µ . no clear evidence for this process has been obtained.7. Thus. In the same way. there is bound to be one ‘m/ E ’ suppression factor. We shall assume the analogous properties are true for the ν ’s accompanying τ leptons. A are nuclei.77) are observed to produce only µ+ ’s when they interact with matter. exactly as in Davis’ original argument. Double β -decay without emission of a neutrino.78) to proceed. ‘ν ’ particles accompanying the µ− ’s in π − decay ¯ ¯ π − → µ− + ‘ ν ’ (20. in this type of reaction. in which states of one such flavour can acquire a component of another. via the diagram shown in figure 20. for the simple case of νe ↔ νµ mixing (for example).2) have combined to demonstrate convincingly that ‘neutrino oscillations’ do occur. for a flavour quantum number distinguishing neutrinos which interact in association with one kind of charged lepton from those which interact in association with a different charged lepton. the review by Kayser in Hagiwara et al (2002) and references therein) concerns ‘neutrinoless double β -decay’. which is the process A → A + e− + e− . As yet. ¯ On the other hand.2.2.Figure 20. The ν emitted along with the e− at the first vertex will be predominately λ = +1. Thus there is evidence.1 the experiment of Danby et al (1962) found no evidence for it.3. For the moment we simply state that. It would create too big a detour to continue with the details of this interesting physics at this point: we shall return to it in section 22. the probability that an initially pure νµ state becomes a νe state in Copyright 2004 IOP Publishing Ltd . like the outgoing e− .2. where A. then in principle it can initiate a second weak interaction. helicity arguments alone would allow the reaction ‘ ν ’ + p → e+ + n ¯ (20. If the neutrino emitted in the first β -decay carries no electron-type conserved quantum number. not µ− ’s. However. but as we saw in section 1. as it propagates. a test for Majorana-type neutrinos. whichever vertex we choose to make ‘easy’. for example. There is also a complicated nucleus physics overlap factor. One much-discussed experimental test case (see.7. a number of observations (see section 22.

82) Copyright 2004 IOP Publishing Ltd . Klein 1948. e− ).n) by the corresponding quarks (u. It is also the case that none of the known (2003) neutrino experiments is sensitive to the difference between the Dirac and Majorana ‘option’.vacuo is proportional to sin2 ( L/ L 0 ) (20.8 × E(GeV)/ m 2 (eV2 ) (20.7) L 0 (in km) 0. Lee et al 1949. and a certain form of ‘universality’ does hold. 1981.6).b).2). τ − ) as well as two other quark pairs (c. Fermi’s original vector-like current ψ e γ µ ψν becomes modified to a total leptonic charged current ˆµ ˆµ ˆµ ˆµ jCC (leptons) = jwk(e) + jwk(µ) + jwk(τ ) where.81) (20. for example.4. If m 2 ∼ 10−4 eV2 and E ∼ 1 GeV.n). so ν ν that L must be of the order of the radius of the earth before (20.79) where L is the distance from the νµ production point. µ− ). as already indicated in section 20. we shall postpone the discussion of the quark currents until section 20. using the V − A modification to Fermi’s theory. It is to these pairs that the ‘V − A’ structure applies.7 The universal current–current theory for weak interactions of leptons After the breakthroughs of parity violation and V − A theory. section 9. In this case. Just such oscillations have been observed via atmospheric or solar neutrinos.80) where E is the neutrino beam energy and m 2 is the squared mass difference |m 2µ − m 2e |. section P.d). it is fair to say that the Standard Model treats neutrinos as Dirac particles. and the ‘oscillation length’ L 0 is given by (Perkins 2000. The upshot of all this is that. From our modern standpoint. (20. we shall see why. (νe . if neutrinos are Majorana particles. ¯ ˆµ ˆ ˆ jwk(e) = ν e γ µ (1 − γ5 )e.10. 20. this list has to be changed by the replacement of (p. Puppi 1948. (νµ . see also appendix P. Because of certain complications which arise. Mohapatra and Senjanovic 1980. and that is what we shall generally assume in the rest of this part of the book. the way their mass must appear would suggest an origin in ‘physics beyond the Standard Model’. but the experiment of Danby et al (1962) was obviously unable to see them. while the ‘Majorana neutrino’ hypothesis is interesting and still viable.79) is appreciable. the earlier hopes (Pontecorvo 1947. Yanagida 1979.s) and (t. we see that L 0 ∼ 8000 km. and by the inclusion of the third lepton pair (ντ . as we now describe. and one for which some appealing theoretical arguments can be made (Gell-Mann et al 1979. In due course (section 22. Tiomno and Wheeler 1949) were revived of a universal weak interaction among the pairs of particles (p. concentrating here on the ¯ ˆ ˆ leptonic currents.

83). Renton 1990.4) using ¯ ¯ ˆ ˆ ˆ 2ν eL γ µ eL = ν e γ µ (1 − γ5 )e ˆ (20. (20.87) (see.8 at the end of the previous volume: G F has dimensions (mass)−2 . say. the term GF ¯ ˆ¯ ˆ ν ˆ √ ν µ γ µ (1 − γ5 )µ eγµ (1 − γ5 )ˆ e 2 describes µ− decay: ¯ µ− → νµ + e− + νe (20.9 that there are also electrically neutral weak currents. is denoted by e(x) rather than ψe (x) ˆ and the ‘x ’ argument is suppressed. in contrast to an ¯ ˆ electromagnetic current operator such as eγ µ e which is electrically neutral.90) .83) 2 Note that ¯ ¯ (ν e γ µ (1 − γ5 )e)† = eγ µ (1 − γ5 )ˆ e ˆ ˆ ˆ ν (20.84) and similarly for the other bilinears. The ‘charged’ current terminology refers to ˆµ the fact that these weak current operators jwk carry net charge. The currents can also be written in terms of the chiral components of the fields (recall section 20.166 × 10−5 GeV−2 .In (20. For example. The interaction Hamiltonian density accounting for all leptonic weak interactions is then taken to be G µ † ˆ lep ÀCC = √F jˆCC(leptons) jˆCCµ (leptons).85) for example.2) and it is found to be GF 1.88) The value of G F can be determined from the rate for process (20.4. section 6. describe many physical processes. (20.86) (20. when it is multiplied out.89) This is a convenient moment to note that the theory is not renormalizable according to the criteria discussed in section 11. for the first time. and the same ‘strength parameter’ G F / 2 multiplies all of the products in (20. a useful shorthand whereby ˆ the field operator for the electron field. (20.87) as well as all the reactions related by ‘crossing’ particles from one side to the other. The terms in (20. We shall return to this aspect of Fermitype V − A theory in section 21. There are also what we might call ‘diagonal’ terms in which the same lepton ˆ† ˆµ pair is taken from jwk and jwkµ . for example. We ˆ shall see in section 20. for example GF ¯ √ ν e γ µ (1 − γ5 )e eγµ (1 − γ5 )ˆ e ˆ¯ ˆ ν ˆ 2 Copyright 2004 IOP Publishing Ltd (20. ‘Universality’ is manifest in the fact that all the lepton pairs h√ ave the same form of the V − A coupling.82) we are now adopting. for example νµ + e− → µ− + νe .83).1.

83)). generating a whole new wave of experimental activity.91) was measured by Reines et al (1976) after many years of effort. ¯ (20. The amplitude. (20. In 1973 such neutral current processes were discovered (Hasert et al 1973). However.93) We shall be interested in energies much greater than any of the lepton masses and so we shall work in the massless limit: this is mainly for ease of calculation—the full expressions for non-zero masses can be obtained with more effort. ¯ ¯ (20. So (20. If the lepton pairs ¯ ¯ are arranged so as to have no net lepton number (for example e− νe . while those with zero charge participate in neutral current processes.92) is a purely ‘neutral current’ process.88). Today we know that charged current processes are mediated by the W± bosons and the neutral current ones by the Z0 . in fact. is Å = −i(G F / √ 2)u(µ)γµ (1 − γ5 )u(νµ )u(νe )γ µ (1 − γ5 )u(e). in contrast to all the others (such as β-decay or µ-decay) we have discussed so far.9. the leptons can be grouped either as (νe e− ) which is charged.91) The cross-section for (20. there is no way of pairing the leptons in (20.91). in the Fermi-like V − A current theory.129) for 2 → 2 scattering we have.which describes reactions such as ¯ νe + e− → νe + e− . to first order in ÀCC (see (20. or as (νe νe ) ¯ or (e+ e− ) which is neutral. which are called ‘charged current’ processes. it is time to calculate something.9.91).92) For reasons which will become clearer in section 20. ¯ ¯ (20. In the case of ¯ (20. been predicted in the first version of the Standard Model. 1 dσ = |Å|2 (20.94) d 64π 2 s Copyright 2004 IOP Publishing Ltd . µ+ νµ .92) is called a ‘neutral current’ process. neglecting all masses.92) so as to cancel the lepton number and have non-zero charge. in principle.8 Calculation of the cross-section for νµ + e− → µ− + νe After so much qualitative discussion. νµ νµ etc) then pairs with non-zero charge occur in charged current processes. sometimes called inverse muon decay. which is a pure ‘charged current’ process. We choose the process (20. due to Glashow (1961). It is interesting that some seemingly rather similar processes are forbidden ˆ lep to occur. while some ‘neutral current’ contribution could be present in (20. From the general formula (6. 20. for example νµ + e− → νµ + e− . We shall discuss the neutral current couplings in section 20. Their existence had.

let us look at it in some detail.101) b c d (20.where |Å|2 is the appropriate spin-averaged matrix element squared.96) G2 F 2 where the νµ → µ− tensor Nµν is given by / / Nµν = Tr[k γµ (1 − γ5 )k γν (1 − γ5 )] without a 1/(2s + 1) factor. we must average over initial electron states for unpolarized electrons and sum over the final muon polarization states. as in (8. in fact. We define |Å|2 = |Å|2 = G2 F 2 Nµν E µν (20. Similarly. and the e− → νe tensor is E µν = 1 2 (20. sum over both helicity states of both neutrinos since the (1 − γ5 ) factors guarantee that right-handed neutrinos contribute nothing to the cross-section.184) for example. 2 Since this calculation involves a couple of new features. the calculation then reduces to that of a product of traces: 1 / / Tr[k γµ (1 − γ5 )k γν (1 − γ5 )] Tr[ / γ µ (1 − γ5 ) / γ ν (1 − γ5 )] p p 2 (20.7. As for the eµ scattering example in section 8. we can. In the case of neutrino–electron scattering. However.95) all lepton masses being neglected. For the neutrinos there is no averaging over initial neutrino helicities.102) .97) Tr[ / γ µ (1 − γ5 ) / γ ν (1 − γ5 )] p p (20. for convenience of calculation. since only left-handed (massless) neutrinos participate in the weak interaction.100) The first trace is the same as in our calculation of eµ scattering (cf (8.99) (1 − γ5 )2 = 2(1 − γ5 ) the tensor Nµν may be written as / / Nµν = 2 Tr[k γµ (1 − γ5 )k γν ] / / / / = 2 Tr(k γµ k γν ) − 2 Tr(γ5 k γν k γµ ). there is no sum over final neutrino helicities. (20. The second trace must be evaluated using the result Tr(γ5 a b/d ) = 4i / /c / Copyright 2004 IOP Publishing Ltd αβγ δ a α β γ δ (20.98) including a factor of 1 for spin averaging.185)): / / Tr(k γµ k γν ) = 4[kµ kν + kν kµ + (q 2 /2)gµν ]. By commuting the (1 − γ5) factor through two γ matrices ( / γ ν ) and p using the result that (20.

Note that this definition has the consequence that 0123 = +1 = −1. (20. We can now evaluate Nµν .188)).104) We will also need to contract two tensors. 2.37) in appendix J of volume 1).104) and the 2! from the fact that the two indices are contracted.110) as for the electromagnetic tensor L µν (cf (8.(see equation (J. By looking at the possible combinations. i j k ilm = δ j l δkm − δkl δ j m . it should be easy to convince yourself of the result i j k ilm = δ jl δkl δ jm δkm (20. in the approximation of neglecting all lepton masses. one can show (see problem 20. (20.108) For the electron tensor E µν we have a similar result (divided by 2): E µν = 4[( pµ pν + p ν pµ + (q 2 /2)g µν ) − i µνγ δ pγ pδ ]. Hence.111) .e. (20. q µ Nµν = q ν Nµν = 0 (20. 1. The totally anti-symmetric tensor αβγ δ is just the generalization of i j k to four dimensions and is defined by  for 0123 and all even permutations of 0.106) For the four-dimensional tensor. 3 αβγ δ =  0 otherwise.107) where the minus sign arises from (20. (20. 1. after some rearrangement of indices. the result for the νµ → µ− tensor: Nµν = 8[(kµ kν + kν kµ + (q 2 /2)gµν ) − i µναβ k α k β ].109) Now. we may replace p = p+q Copyright 2004 IOP Publishing Ltd (20.103) but 0123 (20. We obtain.105) i.7) that µναβ µνγ δ = −2! δα δ δα γ δβ δ δβ γ (20. Its appearance here is a direct consequence of parity violation. 2. 3  +1 −1 for 1023 and all odd permutations of 0.

and drop all terms involving q in the contraction with Nµν . In the anti-symmetric term, however, we have
µνγ δ

pγ ( pδ + qδ ) =

µνγ δ

pγ qδ
µνγ δ .

(20.112) Thus, we (20.113)

since the term with pδ vanishes because of the anti-symmetry of arrive at µν E eff = 8 p µ pν + 2q 2 g µν − 4i µνγ δ pγ qδ .

We must now calculate the ‘ N · E ’ contraction in (20.96). Since we are neglecting all masses, it is easiest to perform the calculation in invariant form before specializing to the ‘laboratory’ frame. The usual Mandelstam variables are (neglecting all masses) s = 2k · p u = − 2k · p t = − 2k · k = q 2 satisfying s + t + u = 0. The result of performing the contraction Nµν E µν = Nµν E eff

(20.114) (20.115) (20.116)


(20.118) tensors (see

may be found using the result (20.107) for the contraction of two problem 20.7): the answer for νµ e− → µ− νe is Nµν E µν = 16(s 2 + u 2 ) + 16(s 2 − u 2 )


where the first term arises from the symmetric part of Nµν , similar to L µν , and the second term from the anti-symmetric part involving µναβ . We have also used t = q 2 = −(s + u) (20.120)

valid in the approximation in which we are working. Thus, for νµ e− → µ− νe we have (20.121) Nµν E µν = +32s 2 and with 1 dσ = d 64π 2 s we finally obtain the result G2 s dσ = F2 . d 4π
Copyright 2004 IOP Publishing Ltd

G2 F 2

Nµν E µν



G2 s F . (20.124) π Since t = −2 p2 (1 − cos θ ), where p is the CM momentum and θ the CM scattering angle, (20.123) can alternatively be written in invariant form as (problem 20.8) G2 dσ = F. (20.125) dt π All other purely leptonic processes may be calculated in an analogous fashion (see Bailin (1982) and Renton (1990) for further examples). When we discuss deep inelastic neutrino scattering in section 20.11, we shall be interested in neutrino ‘laboratory’ cross-sections, as in the electron scattering case of chapter 9. A simple calculation gives s 2m e E (neglecting squares of lepton masses by comparison with m e E ), where E is the ‘laboratory’ energy of a neutrino incident, in this example, on a stationary electron. It follows that the total ‘laboratory’ cross-section in this Fermi-like current–current model rises linearly with E. We shall return to the implications of this in section 20.11. The process (20.88) was measured by Bergsma et al (CHARM collaboration) (1983) using the CERN wide-band beam ( E ν ∼ 20 GeV). The ratio of the observed number of events to that expected for pure V − A was quoted as 0.98 ± 0.12. σ = 20.9 Leptonic weak neutral currents ¯ The first observations of the weak neutral current process νµ e− → νµ e− were ¯ reported by Hasert et al (1973), in a pioneer experiment using the heavy-liquid bubble chamber Gargamelle at CERN, irradiated with a νµ beam. As in the ¯ case of the charged currents, much detailed experimental work was necessary to determine the precise form of the neutral current couplings. They are, of course, predicted by the Glashow–Salam–Weinberg (GSW) theory, as we shall explain in chapter 22. For the moment, we continue with the current–current approach, parametrizing the currents in a convenient way. There are two types of ‘neutral current’ couplings: those involving neutrinos ¯ ¯ ˆ ˆ of the form ν l . . . νl ; and those involving the charged leptons of the form lˆ . . . l. ˆ We shall assume the following form for these currents (with one eye on the GSW theory to come): (1) neutrino neutral current ¯ ˆ gN cνl ν l γ µ (2) charged lepton neutral current ¯ l ˆ gNlγ µ cL (1 − γ5 ) l (1 + γ5 ) ˆ + cR l 2 2 l = e, µ, τ. (20.127) 1 − γ5 2 νl ˆ l = e, µ, τ ; (20.126)

The total cross-section is then

Copyright 2004 IOP Publishing Ltd

This is, of course, by no means the most general possible parametrization. The neutrino coupling is retained as pure ‘V − A’, while the coupling in the charged lepton sector is now a combination of ‘V − A’ and ‘V+A’ with certain coefficients l l cL and cR . We may also write the coupling in terms of ‘V’ and ‘A’ coefficients l l l l l l defined by cV = cL + cR , cA = cL − cR . An overall factor gN determines the strength of the neutral currents as compared to the charged ones: the c’s determine the relative amplitudes of the various neutral current processes. As we shall see, an essential feature of the GSW theory is its prediction of weak neutral current processes, with couplings determined in terms of one parameter of the theory called ‘θW ’, the ‘weak mixing angle’ (Glashow 1961, Weinberg 1967). The GSW predictions for the parameter gN and the c’s is (see equations (22.37)–(22.40)) gN = g/ cos θW cνl =
1 2 l cL = − 1 + a 2 l cR = a


for l = e, µ, τ , where a = sin2 θW and g is the SU(2) gauge coupling. Note that a strong form of ‘universality’ is involved here too: the coefficients are independent of the ‘flavour’ e, µ or τ , for both neutrinos and charged leptons. The following reactions are available for experimental measurement (in addition to the charged current process (20.88) already discussed): νµ e− → νµ e− νe e− → νe e− νµ e− → νµ e− ¯ ¯ − νe e → νe e− ¯ ¯ (NC) (NC + CC) (20.129) (20.130)

where ‘NC’ means neutral current and ‘CC’ charged current. Formulas for these cross-sections are given in section 22.4. The experiments are discussed and reviewed in Commins and Bucksbaum (1983), Renton (1990) and, most recently, by Winter (2000). All observations are in excellent agreement with the GSW predictions, with θW determined as sin2 θW 0.23. The reader must note, however, that modern precision measurements are sensitive to higher-order (loop) corrections, which must now be included in comparing the full GSW theory with experiment (see section 22.8). The simultaneous fit of data from all four reactions in terms of the single parameter θW provides already strong confirmation of the theory—and indeed such confirmation was already emerging in the late 1970s and early 1980s, before the actual discovery of the W± and Z0 bosons. It is also interesting to note that the presence of vector (V) interactions in the neutral current processes may suggest the possibility of some kind of link with electromagnetic interactions which are, of course, also ‘neutral’ (in this sense) and vector-like. In the GSW theory, this linkage is provided essentially through the parameter θW , as we shall see. 20.10 Quark weak currents The original version of V − A theory was framed in terms of a nucleonic current ¯ ˆ ˆ of the form ψ p γ µ (1 − r γ5 )ψn . With the acceptance of quark substructure it was
Copyright 2004 IOP Publishing Ltd

natural to re-interpret such a hadronic transition by a charged current of the form ¯ ˆ uγ µ (1 − γ5 )d, very similar to the charged lepton currents: indeed, here was a ˆ further example of ‘universality’, this time between quarks and leptons. Detailed comparison with experiment showed, however, that such d → n transitions were very slightly weaker than the analogous leptonic ones: this could be established ¯ ¯ ¯ by comparing the rates for n → pe− νe and µ → νµ e− νe . But for quarks (or their hadronic composites), there is a further complication, which is the very familiar phenomenon of flavour change in weak hadronic processes (recall the discussion in section 1.3.2). The first step towards the modern theory of quark currents was taken by Cabibbo (1963). He postulated that the strength of the hadronic weak interaction was shared between the S = 0 and S = 1 transitions (where S is the strangeness quantum number), the latter being relatively suppressed as compared to the former. According to Cabibbo’s hypothesis, phrased in terms of quarks, the total weak charged current for u, d and s quarks is ¯ (1 − γ5 ) d + sin θC uγ µ (1 − γ5 ) s . ¯ ˆµ ˆ jCab (u, d, s) = cos θC uγ µ ˆ ˆ ˆ 2 2 We can now postulate a total weak charged current ˆµ ˆµ ˆµ jCC (total) = jCC (leptons) + jCab (u, d, s) ˆµ where jCC (leptons) is given by (20.81), and then generalize (20.83) to
µ † ˆ tot ÀCC = √F jˆCC(total) jˆCCµ (total).






The effective interaction (20.133) describes a great many processes. The purely leptonic ones discussed previously are, of course, present in ˆ ˆµ jCC (leptons) jCCµ (leptons). But there are also now all the semi-leptonic processes such as the S = 0 (strangeness conserving) one ¯ d → u + e− + νe and the S = 1 (strangeness changing) one ¯ s → u + e− + νe . (20.135) (20.134)

The notion that the ‘total current’ should be the sum of a hadronic and a leptonic part is already familiar from electromagnetism—see, for example, equation (8.90). The transition (20.135), for example, is the underlying process in semileptonic decays such as − → n + e− + νe ¯ (20.136) and K− → π 0 + e− + νe ¯ (20.137)

Copyright 2004 IOP Publishing Ltd

Figure 20.3. Strangeness-changing semi-leptonic weak decays.

as indicated in figure 20.3. The ‘s’ quark is assigned S = −1 and charge − 1 e. The s → u transition 3 is then referred to as one with ‘ S = Q’, meaning that the change in the quark (or hadronic) strangeness is equal to the change in the quark (or hadronic) charge: both the strangeness and the charge increase by one unit. Prior to the advent of the quark model and the Cabibbo hypothesis, it had been established empirically that all known strangeness-changing semileptonic decays satisfied the rules | S| = 1 and S = Q. The u–s current in (20.131) satisfies these rules automatically. Note, for example, that the process apparently similar to (20.136), + → n + e+ + ν , is forbidden in the lowest order (it requires a double quark e transition from suu to udd). All known data on such decays can be fit with a value 0.23 for the ‘Cabibbo angle’ θC (not to be confused with θW ). This sin θC relatively small angle is, therefore, a measure of the suppression of | S| = 1 processes relative to S = 0 ones. The Cabibbo current can be written in a more compact form by introducing the ‘mixed’ field ˆ ˆ ˆ (20.138) d ≡ cos θC d + sin θC s . ¯ (1 − γ5 ) d . ˆµ ˆ ˆ (20.139) jCab(u, d, s) = uγ µ 2 In 1970 Glashow, Iliopoulos and Maiani (GIM) (1970) drew attention to a theoretical problem with the interaction (20.133) if used in second order. Now it is, of course, the case that this interaction is not renormalizable, as noted previously for the purely leptonic one (20.83), since G F has dimensions of an inverse mass squared. As we saw in section 11.7, this means that oneloop diagrams will typically diverge quadratically, so that the contribution of such a second-order process will be of order (G F .G F 2 ) where is a cut-off, compared to the first-order amplitude G F . Recalling from (20.89) that G F ∼ 10−5 GeV−2 , we see that for ∼ 10 GeV such a correction could be significant if accurate enough data existed. GIM pointed out, in particular, that some secondorder processes could be found which violated the (hitherto) well-established phenomenological selection rules, such as the | S| = 1 and S = Q rules already discussed. For example, there could be S = 2 amplitudes contributing to the KL − KS mass difference (see Renton 1990, section 9.1.6, for example), as
Copyright 2004 IOP Publishing Ltd


well as contributions to unobserved decay modes such as K+ → π + + ν + ν ¯ (20.140)

which has a neutral lepton pair in association with a strangeness change for the hadron. In fact, experiment placed very tight limits on the non-existence of (20.140)—and still does: the present limit on the branching is less than about 10−9 %. This seemed to imply a surprisingly low value for the cut-off, say ∼ 3 GeV (Mohapatra et al 1968). Partly in order to address this problem and partly as a revival of an earlier lepton-quark symmetry proposal (Bjorken and Glashow 1964), GIM introduced a fourth quark, now called c (the charm quark) with charge 2 e. Note that in 1970 3 the τ -lepton had not been discovered, so only two lepton family pairs (νe , e), (νµ , µ) were known: this fourth quark, therefore, did restore the balance, via the twoquark family pairs (u,d), (c,s). In particular, a second quark current could now be hypothesized, involving the (c,s) pair. GIM postulated that the c-quark was coupled to the ‘orthogonal’ d–s combination (cf (20.138)) ˆ s = − sin θC d + cos θC s . ˆ ˆ The complete four-quark charged current is then ¯ (1 − γ5 ) d + cγ µ (1 − γ5 ) s . ¯ ˆ ˆµ ˆ jGIM(u, d, c, s) = uγ µ ˆ ˆ 2 2 (20.142) (20.141)

The form (20.142) had already been suggested by Bjorken and Glashow (1964). The new feature of GIM was the observation that, assuming an exact SU(4)f symmetry for the four quarks (in particular, equal masses), all second-order contributions which could have violated the | S| = 1, S = Q selection rules now vanished. Further, to the extent that the (unknown) mass of the charm quark functioned as an effective cut-off , due to breaking of the SU(4)f symmetry, they estimated m c to lie in the range 3–4 GeV, from the observed KL − KS mass difference. GIM went on to speculate that the non-renormalizability could be overcome if the weak interactions were described by an SU(2) Yang–Mills gauge theory, involving a triplet (W+ , W− , W0 ) of gauge bosons. In this case, it is natural to introduce the idea of (weak) ‘isospin’, in terms of which the pairs (νe , e), (νµ , µ), (u,d ), (c, s ) are all t = 1 doublets with t3 = ± 1 . Charge-changing currents then 2 2 involve the ‘raising’ matrix 1 1 τ+ ≡ (τ1 + iτ2 ) = 2 2 0 1 0 0 (20.143)

and charge-lowering ones the matrix τ− /2 = (τ1 − iτ2 )/2. The full symmetry must also involve the matrix τ3 /2, given by the commutator [τ+ /2, τ− /2] = τ3 . Whereas τ+ and τ− would (in this model) be associated with transitions mediated
Copyright 2004 IOP Publishing Ltd

by W± , transitions involving τ3 would be mediated by W0 and would correspond to ‘neutral current’ transitions for quarks. We now know that things are slightly more complicated than this: the correct symmetry is the SU(2) × U(1) of Glashow (1961), also invoked by GIM. Skipping, therefore, some historical steps, we parametrize the weak quark neutral current as (cf (20.127) for the leptonic analogue) gN
q=u,c,d ,s q (1 − γ5 ) q (1 + γ5 ) ¯ + cR q ˆ qγ µ cL ˆ 2 2 q


for the four flavours so far in play. In the GSW theory, the cL ’s are predicted to be
u, cL c = d, cL s 1 2

− 2a 3
1 2

u, cR c = − 2 a 3 1 3a d, cR s

(20.145) (20.146)

= −



1 3a

where a = sin2 θW as before and gN = g/ cos θW . One feature of (20.144) is worth nothing. Consider the terms ¯ ¯ ˆ ˆ d {. . .}d + s {. . .}ˆ . ˆ s (20.147)

It is simple to verify that, whereas either part of (20.147) alone contains a ¯ ¯ ˆ ˆ strangeness-changing neutral combination such as d{. . .}ˆ or s {. . .}d , such s ˆ combinations vanish in the sum, leaving the result diagonal in quark flavour. Thus, there are no first-order neutral flavour-changing currents in this model, a result which will be extended to three flavours in sections 22.3 and 22.7.1. In 1974, Gaillard and Lee (1974) performed a full one-loop calculation of the KL − KS mass difference in the GSW model as extended by GIM to quarks and using the renormalization techniques recently developed by ’t Hooft (1971b). They were able to predict m c ∼ 1.5 GeV for the charm quark mass, a result spectacularly confirmed by the subsequent discovery of the c¯ states in c charmonium, and of charmed mesons and baryons of the appropriate mass. 20.11 Deep inelastic neutrino scattering We now present another illustrative calculation within the framework of the ‘current–current’ model, this time involving neutrinos and quarks. We shall calculate cross-sections for deep inelastic neutrino scattering from nucleons, using the parton model introduced (for electromagnetic interactions) in chapter 9. In particular, we shall consider the processes νµ + N → µ− + X νµ + N → µ+ + X ¯ (20.148) (20.149)

which, of course, involve the charged currents for both leptons and quarks. Studies of these reactions at Fermilab and CERN in the 1970s and 1980s played
Copyright 2004 IOP Publishing Ltd

Figure 20.4. Inelastic neutrino scattering from a nucleon.

a crucial part in establishing the quark structure of the nucleon, in particular the quark distribution functions. The general process is illustrated in figure 20.4. By now we are becoming accustomed to the idea that such processes are, in fact, mediated by the W+ , but we shall assume that the momentum transfers are such that the W-propagator is effectively constant (see the discussion in section 21.2). The effective lepton– quark interaction will then take the form ¯ ˆ eff ˆ ¯ Àνq = √F µγµ (1 − γ5 )ˆµ [uγ µ (1 − γ5 )d + cγ µ (1 − γ5 )ˆ] ν ¯ ˆ ˆ s ˆ 2 G (20.150)

leading to expressions for the parton-level subprocess amplitudes which are exactly similar to that in (20.93) for νµ + e− → µ− + νe . Note that we are considering only the four flavours u, d, c, s to be ‘active’, and we have set θC ≈ 0. As in (20.96), the νµ cross-section will have the general form dσ (ν) ∝ Nµν W(ν) (q, p)


where Nµν is the neutrino tensor of (20.108). The form of the weak hadron tensor µν W(ν) is deduced from Lorentz invariance. In the approximation of neglecting lepton masses, we can ignore any dependence on the 4-vector q since q µ Nµν = q ν Nµν = 0.


Just as Nµν contains the pseudotensor µναβ , so too will W(ν) since parity is not conserved. In a manner similar to equation (9.10) for the case of electron scattering, we define neutrino structure functions by
(ν) W(ν) = (−g µν )W1 + µν

1 µ ν (ν) i p p W2 − 2 M 2M 2

µνγ δ

(ν) pγ qδ W3 .


In general, the structure functions depend on two variables, say Q 2 and ν, where Q 2 = −(k − k )2 and ν = p · q/M; but in the Bjorken limit approximate scaling is observed, as in the electron case: Q2 → ∞ ν→∞
Copyright 2004 IOP Publishing Ltd

x = Q 2 /2Mν fixed


(ν) (ν) ν W2 ( Q 2 , ν) → F2 (x) (ν) M W1 ( Q 2 , ν) (ν) ν W3 ( Q 2 , ν)

(20.155) (20.156) (20.157)

→ →

(ν) F1 (x) (ν) F3 (x)

where, as with (9.21) and (9.22), the physics lies in the assertion that the F ’s are finite. This scaling can again be interpreted in terms of point-like scattering from partons—which we shall take to have quark quantum numbers. In the ‘laboratory’ frame (in which the nucleon is at rest) the cross-section in terms of W1 , W2 and W3 may be derived in the usual way from (cf equation (9.11)) dσ (ν) = GF √ 2
2 3 1 µν d k 4π M Nµν W(ν) . 4k · p 2k (2π)3


In terms of ‘laboratory’ variables, one obtains (problem 20.10) G2 k d2 σ (ν) k +k (ν) (ν) (ν) W2 cos2 (θ/2) + W1 2 sin2 (θ/2) + sin2 (θ/2)W3 . = F 2π k M d Q 2 dν (20.159) For an incoming anti-neutrino beam, the W3 term changes sign. In neutrino scattering it is common to use the variables x, ν and the ‘inelasticity’ y where y = p · q/ p · k. (20.160) In the ‘laboratory’ frame, ν = E − E (the energy transfer to the nucleon) and y = ν/ E . The cross-section can be written in the form (see problem 20.10)
2 2 G2 d2 σ (ν) (ν) 1 + (1 − y) (ν) 1 − (1 − y) = F s F2 + x F3 dx d y 2π 2 2


in terms of the Bjorken scaling functions, and we have assumed the relation 2x F1

= F2



appropriate for spin- 1 constituents. 2 We now turn to the parton-level subprocesses. Their cross-sections can be straightforwardly calculated in the same way as for νµ e− scattering in section 20.8. We obtain (problem 20.11) νq, ν q : ¯¯ ν q, νq : ¯ ¯ G2 Q2 d2 σ = F sxδ x − dxdy π 2Mν G2 d2 σ Q2 = F sx(1 − y)2 δ x − dxdy π 2Mν . (20.163) (20.164)

Copyright 2004 IOP Publishing Ltd

Figure 20.5. Suppression of νµ q → µ− q for y = 1: (a) initial-state helicities; (b) ¯ ¯ final-state helicities at y = 1.

The factor (1 − y)2 in the ν q, νq cases means that the reaction is forbidden at ¯ ¯ y = 1 (backwards in the CM frame). This follows from the V − A nature of the current, and angular momentum conservation, as a simple helicity argument shows. Consider, for example, the case ν q shown in figure 20.5, with the ¯ helicities marked as shown. In our current–current interaction there are no gradient coupling terms and, therefore, no momenta in the momentum-space matrix element. This means that no orbital angular momentum is available to account for the reversal of net helicity in the initial and final states in figure 20.5. The lack of orbital angular momentum can also be inferred physically from the ‘point-like’ nature of the current–current coupling. For the νq or ν q cases, the ¯¯ initial and final helicities add to zero and backward scattering is allowed. The contributing processes are νd → l − u ¯ νu → l −d ¯ νd → l+u ¯¯ ¯ + νu → l d ¯ (20.165) (20.166)

the first pair having the cross-section (20.163), the second (20.164). Following the same steps as in the electron scattering case (sections 9.2 and 9.3), we obtain
νn ¯ F2 = F2¯ = 2x[d(x) + u(x)] νp F3 νn F2 νn F3 νn = F3¯ = 2[d(x) − u(x)] ¯ νp ¯ F2 νp ¯ F3 νp

(20.167) (20.168) (20.169) (20.170)

= =

¯ = 2x[u(x) + d(x)] ¯ = 2[u(x) − d(x)].

Inserting (20.167) and (20.168) into (20.161), for example, we find that d2 σ (νp) = 2σ0 x[d(x) + (1 − y)2 u(x)] ¯ dxdy where σ0 = G2 M E G2 s F = F 2π π 1.5 × 10−42(E/GeV) m2 (20.172) (20.171)

is the basic ‘point-like’ total cross-section (compare (20.124)). Similarly, one finds that ¯ d2 σ (ν p) ¯ = 2σ0 x[(1 − y)2 u(x) + d(x)]. (20.173) dx dy
Copyright 2004 IOP Publishing Ltd

Figure 20.6. Charged-current inelasticity (y) distribution as measured by CDHS (from Winter 2000, p 443).

The corresponding results for νn and ν n are given by interchanging u(x) and d(x) ¯ ¯ and u(x) and d(x). ¯ The target nuclei usually have approximately equal numbers of protons and neutrons and it is appropriate to average the ‘n’ and ‘p’ results to obtain an ¯ ‘isoscalar’ cross-section σ (νN) or σ (ν N) : d2 σ (νN) = σ0 x[q(x) + (1 − y)2 q(x)] ¯ dx dy ¯ d2 σ (ν N) = σ0 x[(1 − y)2 q(x) + q(x)] ¯ dx dy (20.174) (20.175)

¯ where q(x) = u(x) + d(x) and q(x) = u(x) + d(x). ¯ ¯ Many simple and striking predictions now follow from these quark parton results. For example, by integrating (20.174) and (20.175) over x, we can write dσ (νN) ¯ = σ0 [Q + (1 − y 2 ) Q] dy ¯ dσ (νN) ¯ = σ0 [(1 − y)2 Q + Q] dy (20.176) (20.177)

where Q = xq(x) dx is the fraction of the nucleon’s momentum carried ¯ by quarks and similarly for Q. These two distributions in y (‘inelasticity
Copyright 2004 IOP Publishing Ltd

Figure 20.7. Low-energy ν and ν cross-sections (from Winter 2000, p 427). ¯

distributions’), therefore, give a direct measure of the quark and anti-quark composition of the nucleon. Figure 20.6 shows the inelasticity distributions as reported by the CDHS collaboration (de Groot et al 1979), from which the authors extracted the ratio ¯ ¯ Q/(Q + Q) = 0.15 ± 0.03 (20.178) after applying radiative corrections. An even more precise value can be obtained ¯ by looking at the region near y = 1 for ν N which is dominated by Q, the small Q ¯ contribution (∝ (1 − y)2 ) being subtracted out using νN data at the same y. This method yields ¯ ¯ Q/(Q + Q) = 0.15 ± 0.01. (20.179) Integrating (20.176) and (20.177) over y gives ¯ σ (νN) = σ0 (Q + 1 Q) 3 (ν N) ¯ 1 ¯ σ = σ0 ( Q + Q)

(20.180) (20.181) (20.182) (20.183)

and hence while

¯ ¯ Q + Q = 3(σ (νN) + σ (ν N) )/4σ0

1 ¯ ¯ Q/(Q + Q) = 2

3r − 1 1+r

¯ where r = σ (ν N) /σ (νN). From total cross-section measurements and including c and s contributions, the CHARM collaboration (Allaby et al 1988) reported

¯ Q + Q = 0.492 ± 0.006(stat) ± 0.019(syst)
Copyright 2004 IOP Publishing Ltd


¯ ¯ Q/( Q + Q) = 0.154 ± 0.005(stat) ± 0.011(syst).


The second figure is in good agreement with (20.179) and the first shows that only about 50% of the nucleon momentum is carried by charged partons, the rest being carried by the gluons, which do not have weak or electromagnetic interactions. Equations (20.180) and (20.181), together with (20.172), predict that the ¯ total cross-sections σ (ν N) and σ (ν N) rise linearly with the energy E . Figure 20.7 shows how this (parton model) prediction received spectacular confirmation as early as 1975 (Perkins 1975), soon after the model’s success in deep inelastic ¯ scattering. In fact, both σ (ν N)/ E and σ (ν N) / E are found to be independent of E up to E ∼ 200 GeV. Detailed comparison between the data at high energies and the earlier data ¯ of figure 20.7 at E ν up to 15 GeV reveals that the Q fraction is increasing with energy. This is in accordance with the expectation of QCD corrections to the ¯ parton model (section 15.7): the Q distribution is large at small x and scaling violations embodied in the evolution of the parton distributions predict a rise at small x as the energy scale increases. Returning now to (20.167)–(20.170), the two sum rules of (9.65) and (9.66) can be combined to give 3=
0 1

¯ dx [u(x) + d(x) − u(x) − d(x)] ¯
1 νn dx (F3 + F3 ) νp

(20.186) (20.187) (20.188)

= ≡

1 2

0 1 0

νN dx F3

which is the Gross–Llewellyn Smith sum rule (1969), expressing the fact that the number of valence quarks per nucleon is three. The CDHS collaboration (de Groot et al 1979), quoted IGLLS ≡
0 1 νN dx F3 = 3.2 ± 0.5.


In perturbative QCD, there are corrections expressible as a power series in αs , so that the parton model result is only reached as Q 2 → ∞:
2 IGLLS (Q 2 ) = 3[1 + d1 αs /π + d2 αs /π 2 + · · ·]


where d1 = −1 (Altarelli et al 1978a, b), d2 = −55/12 + Nf /3 (Gorishny and Larin 1986) where Nf is the number of active flavours. The CCFR collaboration (Shaevitz et al 1995) has measured IGLLS in anti-neutrino–nucleon scattering at Q 2 ∼ 3 GeV2 . It obtained IGLLS ( Q 2 = 3 GeV2 ) = 2.50 ± 0.02 ± 0.08
Copyright 2004 IOP Publishing Ltd


Figure 20. x 2 (20.106)) (ν) d αs (Q 2 ) (x F3 (x.002 ± 0. taken from Shaevitz et al (1995). Q 2 )) = 2 π d ln Q 1 x Pqq (z)x F3 x . 3 in agreement with the O(αs ) calculation of Larin and Vermaseren (1991) using QCD = 250 ± 50 MeV.194) Copyright 2004 IOP Publishing Ltd .003.111 ± 0.192) Figure 20. z (20. To leading order.8. This 2 fit yields a value of αs at Q 2 = MZ given by 2 αs (MZ ) = 0. CCFR neutrino–iron structure functions x F3 (Shaevitz et al 1995).193) νp The Adler sum rule (Adler 1963) involves the functions F2 and F2 : IA = 0 1 dx ν p ¯ νp (F − F2 ). the x F3 evolution is given by (cf (15. νp ¯ (20. Q2 z dz . MS The predicted Q 2 evolution of x F3 is particularly simple since it is not coupled to the gluon distribution. shows a comparison of the CCFR data with the next-to-leading order calculation of Duke and Owens (1984). The full line is the next-to-leading order (one-loop) QCD prediction and the dotted line is an extrapolation to regions outside the kinematic cuts for the fit.8.

since this is a gauge symmetry current.170). the neutrino and charged lepton structure functions should be approximately in the ratio 18/5. which from the data are concentrated at small x . however.9 shows the neutrino results on F2 and x F3 together with those from several µN experiments scaled by the factor 18/5. chapter 4).12 Non-leptonic weak interactions The effective weak Hamiltonian of (20. ν ν From (20. chapter 5). 20. as we shall see in the following chapter).58) and (9.199) 2 Copyright 2004 IOP Publishing Ltd . and this gives further confirmation of the quark parton picture.170) allow the F2 functions for electron (muon) and neutrino scattering to be simply related.197) while (20. Relations (20. we see that the differences F2 − x F3 involve the anti-quark (sea) contribution. The agreement is reasonably good.167) and (20.167)–(20. We have mentioned QCD corrections to the simple parton model at several points. depends precisely on separating the non-isoscalar contribution ( IA vanishes for the isoscalar average ‘N’).133) (as modified by GIM) clearly contains the term G µ † ˆq ÀCC (x) = √F jˆGIM(x) jˆµGIM(x) (20.198) Assuming that the non-strange contributions dominate. as we already inferred in section 9.170). and Winter (2000. the right-hand side of IA is just 1 2 0 ¯ d x (u(x) + d(x) − d(x) − u(x)) ¯ (20.02 ± 0. which is the reciprocal of the mean squared charge of the u and d quarks in the nucleon.169) give νp ν ν ¯ F2 N ≡ 1 ( F2 + F2 n ) = x(u + d + u + d).195) which represents four times the average of I3 (isospin) of the target.196) in agreement with the expected value 2. we have eN en F2 = 1 ( F2 + F2 ) = 2 ep 5 18 x(u ¯ + u + d + d) + 1 x(s + s ) + · · · ¯ ¯ 9 (20. This sum rule follows from the conservation of the charged weak current (as will be true in the Standard Model.3. 1985): IA = 2.40 (20.167)–(20. The only published result is that of the BEBC collaboration (Allasia et al 1984.61). ¯ 2 (20. can be employed for the case of neutrino scattering also.167)–(20. For further access to this area we refer to Ellis et al (1996. in the context of deep inelastic charged lepton scattering. From (9.In the simple model of (20. Figure 20. Its measurement. Clearly the full machinery introduced in chapter 15. which is 1 2 for the proton.

these QCD interactions cannot be treated Copyright 2004 IOP Publishing Ltd . Comparison of neutrino results on F2 (x) and x F3 (x) with those from muon production properly rescaled by the factor 18/5. in which no lepton fields are present (just as there are no quarks in (20. p 455).10 can occur. This interaction is responsible. rather schematic since there are strong QCD interactions (not shown) which are responsible for binding the three-quark systems into baryons and the q¯ system into a meson. Unlike the case q of deep inelastic lepton scattering.9. of course. the process shown in figure 20. This figure is. For example. at the quark level.83)). which represents the non-leptonic decay 0 → pπ − . Figure 20. Effective four-fermion non-leptonic weak transition at the quark level.11.10. for a Q 2 ranging between 10 and 1000 GeV2 (from Winter 2000. which undergo no weak interaction.Figure 20. for transitions involving four-quark (or anti-quark) fields at a point. By ‘adding on’ another two quark lines u and d. we arrive at figure 20.

We shall see how CP violation arises in the Standard ¯ ¯ K Model in section 22. In the case of transitions involving at least one heavy quark Q. one has to proceed in a phenomenological way.10.1. with the perturbatively.47). 1990) noticed that a considerable simplification occurs in the linit m Q → ∞.11.4 Verify (20. where perturbation theory fails. 20.56) are invariant under CP. Copyright 2004 IOP Publishing Ltd . parametrizing the decay amplitudes in terms of appropriate form factors (which are analogous to the electromagnetic form factors introduced in chapter 8). we refer to Leader and Predazzi (1996). Isgur and Wise (1989. 20. Problems 20. The non-leptonic sector is the scene of some very interesting physics. since the distance scales involved are typically those of the hadron sizes (∼ 1 fm). For an introduction to the Isgur–Wise theory. For example.Figure 20. t)γ ψ2 (x.55). though substantial progress is being made via lattice QCD. For a convenient review of this large and important area. In general. evidently. t) is an axial vector.1 Show that in the non-relativistic limit (| p| M) the matrix element u p γ µ u n ¯ of (20.7.54) and (20. in the case of semi-leptonic decays. t) is a pseudoscalar under P.2 Show that ψ1 (x. 0 using the process of figure 20. t)γ 0 ψ2 (x. we refer to Donoghue et al (1992). 20. and that ¯ ˆ ˆ ψ(x.3 Verify the normalization N = (E + | p|)1/2 in (20. ¯ ˆ ˆ ˆ 20. and CP violation in the 0 −K0 and B0 −B0 systems. Similar difficulties also arise. This means that non-leptonic weak interactions among hadrons are difficult to analyse quantititively. ¯ ¯ however. Non-leptonic weak decay of addition of two ‘spectator’ quarks.1) vanishes if p and n have different spin states. such as K0 − K0 and B0 − B0 oscillations. one universal function (the ‘Isgur– Wise form factor’) is sufficient to describe a large number of hadronic form factors introduced for semi-leptonic transitions between two heavy pseudoscalar (0− ) or vector (1− ) mesons.5 Verify that equations (20.

(b) from the anti-commutation relations of the other γ matrices. i j k lmn Prove the results i j k imn = δ jm δkm δ jn δkn ijk ijn = 2δkn ijk ijk = 3! (c) Extend these results to the case of the four-dimensional (Lorentz) tensor µναβ (remember that a minus sign will appear as a result of 0123 = +1 but 0123 = −1). γµ } = 0 and hence that (1 + γ5 )γ0 = γ0 (1 − γ5 ) and (1 + γ5 )γ0 γµ = γ0 γµ (1 + γ5 ).7 (a) Consider the two-dimensional anti-symmetric tensor 12 ij defined by = +1. 21 = −1 11 = 22 = 0. in two dimensions. convince yourself of the result i j kl = +1(δik δ j l − δil δ j k ). = 2).6 The matrix γ5 is defined by γ5 = iγ 0 γ 1 γ 2 γ 3 . it can be shown that the product of two three-dimensional anti-symmetric tensors has the form = δil δ jl δkl δim δ jm δkm δin δ jn δkn .20. Hence prove that i j il = δ jl and i δii ij ij =2 (remember. properties: 2 (a) γ5 = 1 and hence Prove the following (1 + γ5 )(1 − γ5 ) = 0. 20. Copyright 2004 IOP Publishing Ltd . By explicitly enumerating all the possibilities (if necessary). show that {γ5 . (b) By similar reasoning to that in part (a) of this question.

kν ) where the phase space factor ‘dLips’ is defined in (16.8 leading to the result dσ/dt = G 2 /π F (neglecting all lepton masses). √ Å = −i(G F / 2)u(µ)γµ (1 − γ5 )u(νµ )g µν u(νe )γν (1 − γ5 )u(e) ¯ ¯ verify the intermediate results given in section 20. Discuss the e+ /µ+ ratio in the light of your result.8 Starting from the amplitude for the process νµ + e− → µ− + νe given by the current–current theory of weak interactions. ke . Repeat the calculation using the amplitude √ Å = (G F / 2) fπ pµ u(ν)γµ (gV + gA γ5 )v(e) ¯ and retaining a finite neutrino mass. 20. Show that the ratio of π + → e+ ν and π + → µ+ ν rates is given by (π + → e+ ν) = (π + → µ+ ν) Me Mµ 2 2 2 Mπ − Me 2 2 Mπ − Mµ 2 .111).11.9 The invariant amplitude for π + → e+ ν decay may be written as √ Å = (G F / 2) fπ pµ u(ν)γµ (1 − γ5 )v(e) ¯ where pµ is the 4-momentum of the pion and the neutrino is taken to be massless. Evaluate the decay rate in the rest frame of the pion using the decay rate formula 2 = (1/2 Mπ )|Å|2 dLips( Mπ . Hence.20.10 (a) Verify that the inclusive inelastic neutrino–proton scattering differential cross-section has the form G2 k d2 σ (ν) (ν) (ν) W2 cos2 (θ/2) + W1 2 sin2 (θ/2) = F 2πk dQ 2 dν (k + k ) 2 (ν) + sin (θ/2)W3 M in the notation of section 20. F 20. Copyright 2004 IOP Publishing Ltd . show that the local total cross-section for this process rises linearly with s : σ = G 2 s/π.

dx d y 2π xy k sin2 (θ/2) = . namely G2 dσ (ν q) = F . dt π (a) Show that the cross-section for scattering by anti-quarks νµ q has the form ¯ G2 dσ (ν q) = F (1 − y)2 . .11 The differential cross-section for νµ q scattering by charged currents has the same form (neglecting masses) as the νµ e− → µ− νe result of problem 20.(b) Using the Bjorken scaling behaviour ν W2 (ν) → F2 (ν) M W1 (ν) → F1 (ν) ν W3 (ν) → F3 (ν) rewrite this expression in terms of the scaling functions.8.11: G2 d2 σ (νq) = F sxδ(x − Q 2 /2Mν) dx dy π and G2 d2 σ (ν q) = F sx(1 − y 2 )δ(x − Q 2 /2Mν) ¯ dx dy π (c) Use the parton model prediction G2 d2 = F sx[q(x) + q(x)(1 − y)2 ] ¯ dx dy π Copyright 2004 IOP Publishing Ltd (where M is the nucleon mass). ¯ dt π (b) Hence prove the results quoted in section 20. M 2 (c) Insert the Callan–Gross relation 2 x F1 (ν) Remember that = F2 (ν) to derive the result quoted in section 20. 20.11: G 2 (ν) d2 σ (ν) = F s F2 dx d y 2π x F3 1 − (1 − y)2 1 + (1 − y)2 + (ν) 2 2 F2 (ν) . neglect all massses and show that G2 d2 σ (ν) (ν) (ν) (ν) = F s[ F2 (1 − y) + F1 x y 2 + F3 (1 − y/2) yx]. In terms of the variables x and y .

to show that and F2 (ν) = 2x[q(x) + q(x)] ¯ = q(x) − q(x) ¯ . q(x) + q(x) ¯ (ν) x F3 (x) (ν) F2 (x) Copyright 2004 IOP Publishing Ltd .

that one might well wonder why it should not be regarded as a fully-fledged theory. not only as between different types of lepton but also (within the context of Cabibbo– GIM ‘mixing’) between the quarks and the leptons. One good reason for wanting to do this would be in order to carry out calculations beyond the lowest order. which is essentially all we have used it for so far (with the significant exception of the GIM argument). now recognized as one of the pillars of the Standard Model. further discussions of these old issues concerning the current–current model might seem irrelevant. An understanding of them provides valuable motivation for the GSW theory—and some idea of what is ‘at stake’ in regard to experiments relating to those parts of it (notably the Higgs sector) which have still not been experimentally established. and the generalization of this property in the non-Abelian theories of chapter 13. We saw that this gives a remarkably accurate account of a wide range of data—so much so. This is the remarkable ‘universality’ structure noted in the previous chapter. it is worth emphasizing once again a more positive motivation for a gauge theory of weak interactions. Before reviewing the difficulties. these difficulties do raise several important points of principle. was formulated long before such precision measurements existed. with certain in-principle difficulties associated with the current–current model. A gauge theory would provide a natural framework for such universal couplings. under the impetus of quite compelling theoretical arguments. These had to do. in fact.21 DIFFICULTIES WITH THE CURRENT–CURRENT AND ‘NAIVE’ INTERMEDIATE VECTOR BOSON MODELS In the preceding chapter we developed the ‘V − A current–current’ phenomenology of weak interactions. Since we now believe that the GSW theory is the correct description of electroweak interactions up to currently tested energies. Salam and Weinberg. But the electroweak theory of Glashow. however. mainly. This recalls very strongly the ‘universality’ property of QED. However. if viewed as a ‘theory’. Such higher-order calculations are indeed required by the precision attained in modern high-energy experiments. Copyright 2004 IOP Publishing Ltd .

It is a consequence of unitarity or flux conservation (see. Consider the process (figure 21. up to a numerical factor. in the previous chapter.6) (21.5) (21. which is the square of the total energy in the CM frame.1 Violation of unitarity in the current–current model We have seen several examples. Merzbacher 1998. for example. Å = G F E 2 (1 + cos θ ) (21. the cross-section in each partial wave is bounded by σ J ≤ 4π(2 J + 1)/k 2 Copyright 2004 IOP Publishing Ltd (21.7) . treated to lowest order in perturbation theory.1). chapter 13) that the partial wave amplitude may be written in terms of a phase shift δ J : f J = eiδ J sin δ J so that | f J | ≤ 1. For spinless particles the total cross-section may be written as a sum of partial wave cross-sections σ = 4π k2 (2 J + 1)| f J |2 J (21. Since the troubles we shall find occur at high energies. ¯ ¯ 21.1) in the current–current model. we can simplify the expressions by neglecting the lepton masses without altering the conclusions. F Consider now a partial wave analysis of this process.2) where E is the CM energy and θ is the CM scattering angle of the e− with respect to the direction of the incident µ− . In this limit the invariant amplitude is (problem 21. regarding it as fundamental interaction.1. A similar process was discussed in chapter 20. Current-current amplitude for νµ + µ− → νe + e− .1) νµ + µ− → νe + e− ¯ ¯ (21.4) where f J is the partial wave amplitude for angular momentum J and k is the CM momentum. This leads to the following behaviour of the cross-section: (21. in which cross-sections were predicted to rise indefinitely as a function of the invariant variable s .3) σ ∼ G 2 E 2. Thus.Figure 21. We begin by showing why this is ultimately an unacceptable behaviour.

2). At energies well above lepton masses. which ultimately have to do with the dimensionality of the coupling constant in Copyright 2004 IOP Publishing Ltd . at some energy E. To fill in all the details. G F is found to have the value (Hagiwara et al 2002) : GF 1. the discussion referred to a hypothetical ‘four-fermion’ interaction without the V −A structure and it was concerned with renormalization rather than unitarity. In that case. the square of the CM energy E. so that σ ∼ G 2 E 2 . This is the essence of the ‘unitarity disease’ of the current–current model. F At this point the reader may recall a very similar-sounding argument made in section 11. By contrast. It must also be relativistically invariant. since the amplitude (equation (21.10).8) Moreover.8. (21. The net effect of the spin structure is to produce the (1 + cos θ ) factor in equation (21.9) The unitarity bound (21. We shall avoid those details and instead sketch the conclusions of such an analysis. (21. Given this fact. Therefore. The gamma-matrix structure is irrelevant to these issues.1 particles interacting via a V − A interaction we have seen that helicity is 2 conserved.166 39(1) × 10−5 GeV−2 (21. the remaining amplitude can be regarded as that appropriate to a J = 0. the unitarity bound will be violated by this lowest-order (Born approximation) expression for σ . which is the state required by the V − A interaction.11) and has the dimensions of [M]−2 . however. spinless process (since there are no other factors of cos θ ) so that J f eff=0 ∼ G F E 2 . This embodies the fact that the initial state with Jz = −1 ( Jz quantized along the µ− direction in the CM system) is forbidden by angular momentum conservation to go to a Jz = +1 state at θ = π. we can arrive at the form for the cross-section for νµ µ− → νe e− at high energy without calculation. which led to the same estimate of the ‘dangerous’ energy scale (21. For massless spin. (21. the only invariant quantity available to restore the correct dimensions to σ is s. and is not due to more and more partial waves contributing to the sum in σ .2)) only involves (cos θ )0 and (cos θ )1 contributions.which falls as the CM energy rises. The ¯ ¯ cross-section has dimensions of [L]2 = [M]−2 but must involve G 2 which has F dimension [M]−4 .3) we have a cross-section that rises with CM energy: σ ∼ E 2. in (21. it is clear that this rise in σ is associated with only a few partial waves.6) is therefore violated for CM energies E ≥ GF −1/2 ∼ 300 GeV. From calculated decay rates. involves a careful treatment of the appropriate partial wave analysis for the case when all particles carry spin.10) This difficulty with the current–current theory can be directly related to the fact that the Fermi coupling constant G F is not dimensionless. Extracting this angular-momentum-conserving kinematic factor.

This strongly suggests that we should take Fermi’s analogy further and look for a weak interaction analogue of (21. Pressing the analogy with QED further will help us to see how one might arise. the Fermi current–current interaction has the ‘four-point’ structure ¯ ˆ ¯ ˆ ˆ ˆ ‘G F (ψ ψ) · (ψ ψ)’ while QED has the ‘three-point’ (Yukawa) structure ¯ ˆ ˆ ˆ ‘eψ ψ A’. we are led to search for a theory of weak interactions that involves a dimensionless coupling constant.15) Lorentz invariant.14) (21. was found to be (see problem 8. If we accept this clue from QED.88)) σ = 4πα 2 /3E 2 (21. and ask: what happens in QED? We consider an apparently quite similar process. in fact. whereas in QED they interact only indirectly via the mediation of the electromagnetic field. But. In fact. once again. unitarity and renormalizability are actually rather closely related. Fermi’s current–current model was. M Since the weak currents are. One-photon annihilation graph for e+ e− → µ+ µ− . (21. analogous to G F .2. Dimensional analysis shows. that [g] = 0.2). as we said.Figure 21. both cases. In this case the coupling constant α. vector-like. so that a factor E 2 is required in the denominator to give σ ∼ [L]2 . of course. we must assume that the ˆ W fields are also vectors (spin-1) so as to make (21. in Fermi’s case. neglecting lepton masses.19 and equation (9.15) ˆ where W is a bosonic field. namely e+ e− → µ+ µ− in lowest order (figure 21.13) Dimensional analysis easily shows. Faced with this unitarity difficulty. having the form ¯ ˆ ˆ ˆ ‘g ψ ψ W ’ (21. is dimensionless. as we shall see. that [G F ] = M −2 while [e] = M 0 . More formally.14). In chapter 8 the total cross-section for this process.12) which obediently falls with energy as required by unitarity. like electromagnetic Copyright 2004 IOP Publishing Ltd . motivated by the vector currents of QED. the currents interact directly with each other. we appeal to the most successful theory we have. And because the weak interactions are plainly not long-range.

Figure 21. we need a neutral vector boson as well. In addition to all these. quite analogous to figure 21.16) and the propagator (ignoring the i ) is i 2 (−g µν + k µ k ν /MW ) 2 k 2 − MW . Matrix elements will have the general form (up to constant factors) g 2 j1 µ (−gµν 2 + kµ kν /MW ) 2 k 2 − MW ν j2 (21.1). one further difference with electromagnetism. ψ p γµ (1 − r γ5 )ψn destroys negative charge or creates positive charge). Because both charge-raising and charge-lowering currents exist. The reaction (21.17) Let us first see how the IVB model relates to the current–current one. so as to make the total effective interaction chargeˆ conserving.2. Because we also have weak neutral currents. For the moment.1. charged and uncharged.6. It follows that the W fields must then be charged. there is the familiar massless neutral vector boson.2 The IVB model As discussed in section 19. we are going to follow a more scenic route and accept (as Glashow did in 1961) that we are dealing with ordinary ‘unsophisticated’ massive vector particles. There is.g. which is ¯ ˆ ˆ that the currents in β -decay. One-W− annihilation graph for νµ + µ− → νe + e− .18) Copyright 2004 IOP Publishing Ltd . for example. ψ e γ µ (1 − γ5 )ψνe creates negative charge or destroys positive charge). The ‘companion’ current ¯ ˆ ˆ carries the opposite charge (e. is 2 (£ + MW )W µ − ∂ µ ∂ ν Wν = 0 (21. carry charge (e.3.15) are neutral. the W± and Z0 can be understood as gauge quanta. Z0 . described by the field W µ . the ‘intermediate vector boson’ (IVB). and to suppose that weak interactions are mediated by the exchange of IVB’s. the photon.3. so that expressions of the form (21. for example. So we are led to postulate the existence of a massive weak analogue of the photon. as required. the classical wave equation for a massive vector particle. 21. we need both W+ and W− . ¯ ¯ ones.g. of course. (21. the mass of the W quanta cannot be zero. however. Despite the fact that they are not massless. thanks to the symmetry-breaking mechanism explained in section 19. is then conceived as proceeding via the Feynman diagram shown in figure 21.

and ‘g ’ is playing a role analogous to ‘e’(compare (21. (21.Figure 21. µ (21. we 2 certainly have k 2 << MW so that all k -dependence in the W-propagator (21.3 Violation of unitarity bounds in the IVB model As the section heading indicates.19).4. For typical β -decay energies (and knowing that MW ∼ 80 GeV!).21) which is indeed quite close to the true value. out of so much apparent dissimilarity between the weak and electromagnetic interactions. 21. with the important qualitative connection 2 G F ∼ g 2 / MW . in a sense. and we shall take up that issue again in section 21. Maybe the intrinsic strengths g and e are roughly equal: g ∼ e. be rescued.3 but also for the IVB version of β -decay shown in figure 21.14) and (21. via (21.19) This is a fundamental relation—the exact version of it in the GSW theory is given in equation (22.18) can be ignored.1. One-W− exchange process in β -decay.46). ν where j1 and j2 are certain weak-current matrix elements. why the weak interactions are weak! They are so (i. Such a structure will appear not only for processes such as that shown in figure 21.20) This would then lead. Indeed. to an order-of-magnitude estimate of MW : MW ∼ e/ G F 1/2 ∼ 90 GeV (21. and we arrive back at Fermi’s current–current amplitude.15)).4. This simple idea is essentially correct: the precise relation between g and e in the GSW theory will be given in (22. matters will turn out to be fundamentally no better in the IVB model. after all.4. perhaps some simple similarity can. We now investigate whether the IVB model can do any better with unitarity than the current–current model. The analysis will bear a close similarity to the discussion of the renormalizability of the model in section 19. but the demonstration is instructive. GF is ‘small’) principally because MW is so large.e. We begin Copyright 2004 IOP Publishing Ltd .29) of chapter 22. It shows us why the Fermi constant has dimension [mass]−2 and even.

for example. Does the IVB modification ensure that the unitarity bounds are not violated for any Born (i. ¯ p Using the Dirac equation for the spinors in the forms u( p)( / − m) = 0 and ¯ p ( / + m)v( p) = 0 (see problem 4.27) proceeding via the graph shown in figure 21. together with {γ µ .2.26) m e u(e)(1 − γ5 )v(νe ) − m ν u(e)(1 + γ5 )v(νe ). viewed as proceeding as in figure 21.24) (21.25) ¯ p¯ = u(e) / e (1 − γ5 )v(νe ) + u(e) / ν (1 − γ5 )v(νe ).5.12) and hence we conclude that.22) For convenience. the factors of two have been chosen to be those that would actually appear in the GSW theory in unitary gauge. in a dynamical sense. Using 4-momentum conservation.1. The unitarity-violating processes turn out to be those involving external W particles (rather than internal ones. thus the k µ k ν factors have disappeared. (21. At first sight. we might conclude that the high-energy behaviour of (21. as we can see as follows. for the fall with energy of the QED cross-section (21. the IVB modification of the four-fermion model does avoid the violation of unitarity in lowest order. 2 (21. We may compare (21. they turn out to be considering the process (21. the amplitude for which we take to be (1 − γ5 ) u(µ). Consider.2.11).22) is going to be considerably ‘worse’ than that of (21. neglecting the lepton masses by comparison with MW . the effect of the 2 IVB is simply to replace the photon propagator −g µν / k 2 by −g µν /(k 2 − MW ). as in figure 21. which is i v(νµ )γν ¯ ie2 v(e)γµ u(e) ¯ −g µν k2 u(µ)γν v(µ) ¯ (21.3).25) p becomes ¯ ¯ (21. Indeed.1).23) in view of the presence of the k µ k ν factors in the numerator of (21.22) with the amplitude for figure 21.23) g2 (1 − γ5 ) −g µν + k µ k ν / M 2 u(e)γµ ¯ v(νe ) 2 2 k2 − M 2 where k is the 4-momentum of the photon. γ5 } = 0. However. µ ¯ k µ u(e)γµ (1 − γ5 )v(νe ) = ( pe + pν )u(e)γµ (1 − γ5 )v(νe ) ¯ ¯ µ (21. It was this photon propagator which was really responsible. The fact that this is experimentally a somewhat esoteric reaction is irrelevant for the subsequent argument: the Copyright 2004 IOP Publishing Ltd . tree-graph) process? The answer is no. the process νµ + νµ → W+ + W− ¯ (21. A similar result holds for the k ν factor. at least for this process. as will be explained in section 22.3.22).e.

which is. we need the result ∗ µ (k. λ = 0) = MW kµ ˆ (−1. as usual. The amplitude for this process is proportional to Åλ1λ2 = g 2 × / ( / 1 − k 1 + m µ) ν p γ (1 − γ5 )u( p1 ) ( p 1 − k 1 )2 − m 2 µ −∗ +∗ µ ¯ µ (k 2 . λ) = −gµν + kµ kν /MW (21. carried solely in the longitudinal polarization vector µ (k. λ) ν (k. From (19. ¯ proposed theory. will necessarily generate the amplitude shown in figure 21. Copyright 2004 IOP Publishing Ltd . we must form |Å|2 and sum over the three states of polarization for each of the W’s. We may write this as (k. and which will dominate real production of W’s at high energy. λ2 ) ν (k 1 .18) we see that in a frame in which k µ = (k 0 . λ2 ) is that associated − with 4-momentum k and polarization state λ .28) −∗ where the ± are the polarization vectors of the W’s: µ (k2 .17) and (19. and with the outgoing W 2 2 +∗ similarly for ν . 0. λ1 )v( p2 )γ (1 − γ5 ) (21. 0. in which regime it is clear that the kµ kν /MW term in (21. be in the high-energy 2 behaviour of the cross-section.30) which at high energy tends to k µ /MW . it is clear that it is the longitudinal polarization states which are responsible for the k µ k ν parts of the polarization sum (12.5. λ = 0). Our interest will. λ = ±1) involve no momentum dependence.21). To calculate the total cross-section. |k|). µ− -exchange graph for νµ + νµ → W+ + W− . and since this amplitude violates unitarity. represented by the IVB modification of the four-fermion model. It is therefore worth looking a little more closely at this term.Figure 21.29) λ=0.19). in fact.5. the theory is unacceptable.±1 already given in (19.29) will dominate the gµν term. Thus. To do this. the transverse polarization vectors µ (k. k) + 0 MW (k + |k|) (21.

In fact.2)). (21. which are / ‘compensated’ by the factor k 2 . Retaining only the leading powers of energy. precisely as in the four-fermion model (equation (21.36) −2 We see that the longitudinal ’s have brought in the factors MW . Recalling (21. In the case of the latter. The production of longitudinal polarized W’s is actually a pure J = 1 process and the J = 1 partial wave amplitude is f 1 = G F E 2 /6π.34) The unitarity bound | f 1 | ≤ 1 is therefore violated for E ≥ (6π/ G F )1/2 ∼ 103 GeV (cf (21. one obtains (Gastmans 1975) E 2 sin2 θ dσ = G2 (21.33) F d 8π 2 and. we argued that the root of the disease lay in the fact that G F was not dimensionless.28)) and setting µ = kµ /M for the longitudinal polarization vectors. v( p2 )k 2 (1 − γ5 ) 1 ¯ (21. and neglected neutrino masses in forming spins |Å00 |2 . just as before.35) 2 ( p 1 − k 1 )2 MW 2 Using the Dirac equation / 1 u( p1) = 0 and p1 = 0. in this respect. we see that we are involved with an effective amplitude / p g2 / − k1 / / k 1 (1 − γ5 )u( p1 ).Concentrating therefore on the production of longitudinal W’s.10)).19).32) spins where E is the CM energy and θ the CM scattering angle. hence.λ2 for ν ν → W+ W− (equation (21. a total cross-section which rises with energy as E 2 . commuted the (1 − γ5 ) factors through. and we have to conclude that the IVB model is. and it is this latter factor which causes the rise Copyright 2004 IOP Publishing Ltd . putting all the factors in. we are led to examine the quantity g4 4 MW ( p1 − k 1 )4 / / / p / p / / p Tr[k 2 (1 − γ5 )( / 1 − k 1 )k 1 / 1 k 1 ( / 1 − k 1 )k 2 / 2 ] p (21. this can be reduced to p − g2 2 MW / v( p2)k 2 (1 − γ5 )u( p1 ). no more fit to be called a theory than was the four-fermion model.2) that 4 4 |Å00 |2 ∼ (g 4 / MW )( p1 · k2 )( p2 · k2 ) = (g 4 / MW ) E 4 (1 − cos2 θ ) (21. we find (see problem 21. Other unitarity-violating processes can easily be invented. ¯ (21. yet somehow this was not a good enough cure after all: perhaps (it is indeed so) ‘dimensionlessness’ is necessary but not sufficient (see the following section). Why is this? Returning ¯ to Åλ1 .31) where we have neglected m µ . we see that the (unsquared) amplitude must behave essentially as G F E 2 .

Once more we turn in our distress to the trusty guide.λ2 = e2 ∗ ∗ µ ¯ µ (k 2 . This is a non-trivial point. The fact that there are two rather than one will turn out to be significant. the ‘dangerous’ term kµ kν /MW arises entirely from the longitudinal polarization vectors.38). with energy. The amplitude for this diagram is the same as before except that ¯ (1 − γ5 )/2 is replaced by 1.Figure 21. 2 since this diverges.37) ( p 1 − k 1 )2 − m 2 e In the cross-section we would need to sum over the photon polarization states. For the moment we just concentrate on figure 21. However. λ) 2 = −gµν + kµ kν /MW (21. In QED there are two graphs contributing in lowest order (the ‘crossed’ versions of figure 8. λ2 ) ν (k 1 . (21. These are shown in figures 21. in its turn. The longitudinal polarization states have effectively reintroduced a dimensional coupling constant g/ MW . therefore.38) for photons (this time involving only the two transverse polarization states). the longitudinal state of polarization is absent altogether! We might well suspect.6.3. And this circumstance can. Indeed. The analogous electromagnetic process to consider is e+ e− → γ γ .3 that.14).38) λ and so we would need the analogue of this result for massless photons. for real photons.6(a) which is directly analogous to the diagram for ν ν → W+ W− . For the massive spin-1 particles.6. that since it was the longitudinal W’s that caused the ‘bad’ high-energy behaviour of the IVB model.6(a) and (b). and we learnt in section 7. the ‘good’ high-energy behaviour of QED might have its origin in the absence of such states for photons. the right-hand Copyright 2004 IOP Publishing Ltd . Clearly the answer is not to take the MW → 0 limit of (21.1 ) to the gauge invariance property of QED. we used ∗ µ (k. in section 8. (b) crossed graph. λ) ν (k. g/21/2 by e and the vectors now refer to photons: Åλ1 . λ1 )v( p2 )γ / p / 1 − k1 + me ν γ u( p1 ). Lowest-order amplitudes for e+ e− → γ γ : (a) direct graph. be traced (cf section 7. as we shall see in section 21.4.3 we saw that in the analogue of (21. QED.

in which the W’s are gauge quanta.1 but we shall now be somewhat more specific. if the local symmetry is spontaneously broken (chapter 19). It must. We gave an indication of why this is so in section 19. ‘Dimensionlessness’ alone is not enough. for example. We might. We have arrived here at an important theoretical indication that what we really need is a gauge theory of the weak interactions. we have already seen how this ‘peculiarity’ can indeed arise.4 The problem of non-renormalizability in weak interactions The preceding line of argument about unitarity violations is open to the following objection. 21. ‘weak interactions become strong at energies of order 1 TeV’.7 and 21. just as in QED. be a peculiar kind of gauge theory. Consider.side could be taken to be just −gµν . is simply that perturbation theory must fail in theories of the type considered at some sufficiently high energy. also had dimension (mass)−2 . relating the discussion to the previous one about unitarity. The objection is perfectly valid. one needed to supply the values of an ever-increasing number of parameters from experiment and the theory lost predictive power. Does the IVB model fare any better? In this case. We gave an elementary introduction to renormalization in chapters 10 and 11 of volume 1. in fact. Its coupling constant. since normally gauge invariance requires the gauge field quanta to be massless. However. But before proceeding to implement that idea in the GSW theory. provided that the Ward identity (8. the latter for the IVB-mediated process νµ νµ → νµ νµ . In particular. rather than unitarity. The ‘non-renormalizable’ problem was essentially that. as one approached the ‘dangerous’ energy scale (21. arguments based on lowestorder perturbation theory are irrelevant. What it shows. the difficulties that arise when one tries to do higher-order calculations in the case of a four-fermion interaction with the same form (apart from the V − A structure) as the current–current model.8. it turns out: the IVB model is not renormalizable either. It seems plausible ¯ ¯ Copyright 2004 IOP Publishing Ltd . and we shall take account of it by linking high-energy behaviour to the problem of renormalizability. The essential reason is that the effective expansion parameter for perturbation theory 1/2 1/2 is E G F . we discussed in some detail. It is an argument conducted entirely within the framework of perturbation theory. the coupling constant is dimensionless. Since E G F becomes large at high energy.8—the former for the QED process e+ e− → e+ e− via an intermediate 2γ state. just note in passing that yet another way of stating the results of the previous two sections is to say that. the fourth-order processes shown in figure 21. in section 11. we discuss one further disease (related to the unitarity one) possessed by both current–current and IVB models—that of nonrenormalizability.10).165) held. which we called G F . for both the current–current and IVB theories. a condition directly following from gauge invariance. however. however.

So it is plain that.6(a)and (b). This is exactly the comparison we were making ¯ in the previous section.1—namely. already discussed in the previous section.7. via figure 21. the amplitudes must be formed by somehow ‘sticking together’ two of the lower-order graphs shown in figures 21. therefore. 2 the ‘bad’ behaviour of the k µ k ν / MW factors in the W-propagators.4.7 that there are two graphs corresponding to figure 21. Let us see how QED—a renormalizable theory— manages to avoid this problem. at large k . Copyright 2004 IOP Publishing Ltd . mimicking for figures 21. on the other. it seems clear from inspection of figure 21. on the one hand. a non-renormalizable divergence. Figure 21. the blame lies with the longitudinal polarization states for the W’s. Indeed.8.1 Accepting this for the moment. namely figures 21. and knowing that ultraviolet divergences of loop graphs originate in the high-k behaviour of their integrands. and this turns out to produce.5. once again. Consider.5.5 and 21. we are led to compare the high-energy behaviour of the process e+ e− → γ γ . due to the k dependence of the ¯ longitudinal polarization vectors. ¯ ¯ from the diagrams that. and νµ νµ → W+ W− .6(a) and (b) the calculation we did for figure 21. We would obtain the leading high-energy behaviour by replacing the 1 The reader may here usefully recall the discussion of unitarity for one-loop graphs in section 13.3. for the reason indicated in section 19. rather than unitarity. O(e4 ) contributions to e+ e− → e+ e− . In this case. O(g 4 ) contribution to νµ νµ → νµ νµ . we saw in section 20.3 that the high-energy behaviour of the amplitude ν ν → W+ W− (figure 21. in each case.5) grew as E 2 .Figure 21.8. but now we have arrived at it from considerations of renormalizability.6.

Four-point e+ e− vertex.8.5.10. as in the Fermi (current–current) case. would not be renormalizable.7. Thus. This divergence entails the introduction of a new vertex. Thus. If it were. the theory without this vertex is non-renormalizable—and if we include it. since we knew in advance that it is always possible to choose polarization vectors for real photons such that they are purely transverse. the calculation is highly relevant to the question of renormalizing figure 21. Our presentation hitherto has emphasized the fact that.8—and ¯ this time there is no ‘crossed’ amplitude to cancel it. provided that both contributions in figure 21.5) that when this replacement is made for each photon the complete amplitude for the sum of figures 21.10.7(a) and (b)). include their longitudinal ( µ ∝ kµ ) states as well as the transverse ones (see section 13. Nevertheless. The photons in this process are not real external particles but are instead virtual. This is exactly what does happen in the case of figure 21. an infinite counter term proportional to a four-point vertex e+ e− → e+ e− (figure 21. the sum of these two contributions is not divergent. Four-point ν ν vertex. of course. one way to ‘fix’ the IVB theory might be to hypothesize a new physical process to be added to Copyright 2004 IOP Publishing Ltd . In physical terms.9) would have to be introduced and the original QED theory.6(a) and (b) (or figures 21. so that no physical process can depend on a part of µ proportional to kµ . not present in the original IVB theory. the bad highenergy behaviour is rendered harmless by a cancellation between contributions from figures 21.3 then suggests that these longitudinal states are harmless. we are landed with a four-field point-like vertex which is non-renormalizable.Figure 21. Figure 21.9. The bad high-energy behaviour of ν ν → W+ W− translates into a divergence of figure 21. internal ones. This has the consequence that we should.6 are included.3 for something similar in the case of unitarity for one-loop diagrams). in general. which of course lacks such a fundamental interaction. figure 21. ¯ photon polarization vectors by the corresponding momenta and it can be checked (problem 21. this result was expected. The calculation of problem 21. in QED.6(a) and (b) vanishes. Indeed.

38)). the same as in the renormalizable QED case. Find the explicit form of u for the case ˆ p = (sin θ. ˆ Copyright 2004 IOP Publishing Ltd . the potentially harmful longitudinal polarization states are.1. p = p| p|. Our IVB propagator corresponds to the ξ → ∞ limit and with this choice of ξ . We have therefore arrived once more. a little hindsight.1 (a) Using the representation for α.figure 21. the general expression for the gauge boson propagator in such a theory (in a ’t Hooft gauge) is i −g µν + (1 − ξ )k µ k ν 2 k 2 − ξ MW 2 (k 2 − MW + i ) (21. indeed. renormalizable.2) that the vanishing of amplitudes when an µ is replaced by the corresponding kµ is due to gauge invariance: in other words. at the idea that we need a gauge theory of massive vector bosons. harmless in a gauge-invariant theory. ’t Hooft’s proof that they are (’t Hooft 1971b) triggered an explosion of theoretical work.4 (equation (20. There we saw that. we are in possession of a more powerful principle. massless particles are described by spinors of the form u = E 1/2 φ+ φ− (normalized to u † u = 2E) ˆ where σ · pφ± = ±φ± . in fact. after a somewhat more leisurely discussion than that of section 19. But for any finite ξ (for example ξ = 1) the high-energy behaviour of the propagator is actually ∼ 1/ k 2 . as it became clear that. We now have all the pieces in place. so that the offending k µ k ν part of the propagator can be ‘gauged away’ as in the photon case. In QED. it would be possible to make higher-order calculations for weak interaction processes using consistent renormalization procedures of the kind that had worked so well for QED. Problems 21. we have already seen (section 8. This is precisely what is provided by the ‘spontaneously broken’ gauge theory concept. 0. for the first time. β and γ5 introduced in section 20. The search for such high-energy cancellation mechanisms can.5 in such a way that a cancellation occurred at high energies.6. in fact. be pushed to a successful conclusion. arguably. This strongly suggests that such theories—in particular non-Abelian ones—are. cos θ ). However. all the troubles we have been discussing appear to be present.39) where ξ is a gauge parameter. given sufficient ingenuity and. and can proceed to introduce the GSW theory based on the local gauge symmetry of SU(2) × U(1). taking the U(1) case for simplicity. as developed in chapter 19.

) 21. L)γ µ (1 − γ5 )v(¯ e . cos θ ). in the ¯ ¯ limit in which all masses are neglected. as explained in section 20.3 Check that when the polarization vectors of each photon in figures 21. R) ¯ ν ¯ ν where we have explicitly indicated the appropriate helicities R or L (note that.6(a) and (b) is replaced by the corresponding photon momentum. Verify that the amplitude is proportional to G F E 2 (1 + cos θ ). L)u(e− . discussed in section 21. The amplitude is proportional to G F v(¯ µ . so that only the µ = x. 0. (Hint: evaluate the ‘easy’ part v(¯ µ )γµ (1 − γ5 )u(µ− ) first.(b) Consider the process νµ + µ− → νe + e− . let the initial µ− momentum be (0. E) and the final e− momentum be E(sin θ. In the CM frame. Copyright 2004 IOP Publishing Ltd . This will show that the components µ = 0. 0.2 Verify equation (21. the sum of these two amplitudes vanishes. (1 − γ5 )/2 is the projection operator for a right-handed anti-neutrino). 21. y components of the dot product need to be calculated.4.1.32). z ¯ ν vanish. R)γµ (1 − γ5 )u(µ− .

the vacuum is not invariant and we must expect that the degenerate multiplet structure will then. Moreover. This is precisely the situation in the electroweak theory. and later extended by other authors. a prominent symmetry of the Schrodinger equation.e.1 Weak isospin and hypercharge: the SU(2) × U(1) group of the electroweak interactions: quantum number assignments and W and Z masses An important clue to the symmetry group involved in the weak interactions is provided by considering the transitions induced by these interactions. by studying electromagnetic transitions. the relations it requires between otherwise unrelated masses and couplings—are accessible to experiment. However. in general. but in section 17. as we shall see. the relevant weak gauge group? Several possibilities were suggested. despite the fact that members of a multiplet of a global symmetry which is spontaneously broken Copyright 2004 IOP Publishing Ltd . there ¨ is one very important difference between the ‘weak multiplets’ we shall be considering and those associated with symmetries which are not spontaneously broken. When the symmetry is spontaneously broken. disappear completely. This is somewhat analogous to discovering the multiplet structure of atomic levels and hence the representations of the rotation group. 22. introducing elements of the data at illustrative points.1 we learned that that result only holds provided the vacuum is left invariant under the symmetry transformation. produces a theory which is in remarkable agreement with currently known data. subsequently treated as a spontaneously broken gauge symmetry by Weinberg (1967) and by Salam (1968). We saw in chapter 12 how an unbroken non-Abelian symmetry leads to multiplets of states which are degenerate in mass.22 THE GLASHOW–SALAM–WEINBERG GAUGE THEORY OF ELECTROWEAK INTERACTIONS Given the preceding motivations for considering a gauge theory of weak interactions. We shall not give a critical review of all the experimental evidence but instead proceed directly to an outline of the GSW theory. but it is now very well established that the one originally proposed by Glashow (1961). essential consequences of the weak symmetry group—specifically. the remaining question is this: what is the relevant symmetry group of local phase transformations. Nevertheless. i.

the two gauge fields associated with transitions between doublet members will have charge ±1. the subscript ‘L’ refers to the fact that only the left–handed chiral components of the fields enter as a consequence of the V − A structure. transforming as a t = 1 multiplet (a triplet) under the group. we saw in section 20.4. the prime indicates that these fields are quantum–mechanical superposition of the fields ˆ ˆ ˆ d. as anticipated for a spontaneously broken symmetry. This is a generalization to 3×3 mixing of the 2×2 GIM mixing introduced in section 20. Because (as with the ordinary SU(2)f of hadronic isospin) the members of a weak isodoublet differ by one unit of charge. We use the symbols t. In (22. (22. no longer have even approximately the same mass. a ‘weak SU(2) group’ called ‘weak isospin’. (22.will. under an SU(2)L transformation νe ˆ νe ˆ = exp(−iα · τ /2) . For the moment. We emphasize once more that this weak isospin is distinct from the hadronic isospin of chapter 12. to show that the weak isospin assignments and corresponding transformation properties apply only to these left–handed parts: for example. the weak isospin group is referred to as SU(2)L. s . which is part of SU(3)f . Making this SU(2)L into a local phase invariance (following the logic of chapter 13) will entail the introduction of three gauge fields.10.7.3) e− L ˆ e− L ˆ Note that.2).3) that the associated gauge quanta still mediate interactions between members of a given symmetry multiplet.1. The quanta of these fields will. the concept of a multiplet is still useful. νµ ↔ µ etc. in general. in both cases.1) ˆ t ˆ b L ˆ where eL = 1 (1 − γ5 )e etc. The simplest possibility is therefore to suppose that. just as in the manifest local non-Abelian symmetry example of QCD. as we saw in chapter 20. and for the quark fields ˆ 2 t= 1 2 t3 = +1/2 t3 = −1/2 u ˆ ˆ d c ˆ s ˆ .2 and 22.7. the leptonic transitions associated with the weak charged currents are. be the now Copyright 2004 IOP Publishing Ltd . c which are classified by their strong interaction quantum numbers. we shall find (in sections 22. This suggests that these pairs should be regarded as doublets under some group.10 how weak transitions involving charged quarks suggested a similar doublet structure for them also. t3 for the quantum numbers of weak isospin and make the following specific assignments for the leptonic fields: t= 1 2 t3 = +1/2 t3 = −1/2 νe ˆ e− ˆ νµ ˆ µ− ˆ ντ ˆ τ− ˆ L L (22. For this reason. νe ↔ e. is involved.2) L L L As discussed in section 20. of course. these doublets all involve pairs of particles which are not mass degenerate. we ignore the corresponding mixing in the neutrino sectors but return to it in section 22.2. This is because when the symmetry is made a local one. Further. and it will be discussed further in section 22. Now.

e− ) etc then have y = −1. The new Abelian U(1) group is associated with a weak analogue of hypercharge—‘weak hypercharge’— just as SU(2)L was associated with ‘weak isospin’. the W’s must somehow acquire mass. so that W masses were put in ‘by hand’. and a very economical and appealing idea would be to associate this neutral vector particle with the photon— thereby unifying the weak and electromagnetic interactions. and associated with the weak isospin raising and lowering operators t± . (νe . Glashow proposed that the Gell–Mann-Nishijima relation for charges should also hold for these weak analogues. The Copyright 2004 IOP Publishing Ltd . we reproduce here the main results of section 19. that the W mass must arise via the spontaneous breakdown of a non-Abelian gauge symmetry and. Indeed. This is exactly the piece of mathematics we went through in section 19. we now know (see chapter 20) that there exist neutral current weak interactions.6. Of course. Furthermore. this does constrain the W mass. Schwinger arranged this by introducing appropriate couplings of the vector bosons to additional scalar and pseudoscalar fields. These couplings were arbitrary and no prediction of the W masses could be made.familiar W± bosons mediating the charged current transitions. but. which between them will be responsible for the weak neutral current processes and for electromagnetism. similar ideas were also advanced by Salam and Ward (1964).6. while the photon remains massless. and an obvious suggestion is to have these currents mediated by the neutral member W0 of the SU(2)L gauge field triplet.4. the neutral currents are not pure V − A. giving eQ = e(t3 + y/2) (22. We now believe. the lepton doublets. the neutral currents will have the same pure left– handed V − A structure as the charged currents. which we now appropriate as an important part of the Standard Model. Clearly. therefore. A key contribution was made by Glashow (1961). while the quark doublets (u. Now. has been lost. as we saw in chapter 20. A model of this kind was originally suggested by Schwinger (1957). the attractive feature of including the photon. in addition to those of the charged currents. What about the third gauge boson of the triplet? This will be electrically neutral. as we saw in section 19. however. everything falls marvellously into place: the charged vector bosons appear as before but there are now two neutral vector bosons. Glashow suggested enlarging the Schwinger–Bludman SU(2) schemes by inclusion of an additional U(1) gauge group. resulting in an ‘SU(2)L × U(1)’ group structure.4) for the electric charge Q (in units of e) of the t3 member of a weak isomultiplet. Apart from the question of the W mass in Schwinger’s model. 3 when this group is gauged. For convenience.6. and thus unifying weak and electromagnetic interactions. again pre-Higgs. In this model. as emphasized in section 21. We must also include these in our emerging gauge theory. assigned a weak hypercharge y. Such a scheme was indeed proposed by Bludman (1958). dC ) etc have y = + 1 .

10) (22.82) for the gauge field B µ .7) in (22.15) (22.ˆ Higgs field φ is an SU(2) doublet ˆ φ= ˆ φ+ ˆ φ0 (22.8) 4 4 4 ˆµ ˆ where F µν is the SU(2) field strength tensor (19. The Lagrangian for the sector consisting of the gauge fields and the Higgs fields is ÄG 1ˆ ˆ ˆ ˆ ˆ ˆ µν 1 ˆ ˆ ˆ ˆ λ ˆ ˆ = ( Dµ φ)† ( D µ φ) + µ2 φ † φ − (φ † φ)2 − F µν · F − G µν G µν (22.11) ˆ ˆν − ∂ν W2µ )(∂ µ W2 ˆ ˆ ˆ − 1 (∂µ Z ν − ∂ν Z µ )(∂ µ Z ν − ∂ Z ) + 4 ˆ ˆ − 1 Fµν F µν 4 where ˆ ˆ (g 2 + g 2 ) Z µ Z µ (22.16) and with ˆ F µν = ∂ µ Aν − ∂ ν Aµ cos θW = g/(g 2 + g 2 )1/2 Copyright 2004 IOP Publishing Ltd sin θW = g /(g 2 + g 2 )1/2 .5) with an assumed vacuum expectation value (in unitary gauge) given by ˆ 0|φ|0 = 0 √ v/ 2 .80).9) ˆ ˆ ˆ ˆ Äfree = 1 ∂µ H ∂ µ H − µ2 H 2 G 2 ˆ ˆ ˆν − 1 (∂µ W1ν − ∂ν W1µ )(∂ µ W1 4 − 1 ˆ 4 (∂µ W2ν (22.17) .12) (22.7) ˆ where H is the (physical) Higgs field.8)) the quadratic parts of (22.81) for the gauge fields W .e. the insertion of (22. After symmetry breaking (i.9) ˆµ − ∂ ν W1 ) + ˆµ − ∂ ν W2 ) + v2 ν ˆµ 8 1 2 2 ˆ ˆµ 8 g v W1µ W1 1 2 2 ˆ ˆµ 8 g v W2µ W2 (22.8) can be written in unitary gauge as (see problem 19.14) (22.13) ˆµ ˆ ˆ Z µ = cos θW W3 − sin θW B µ ˆ ˆµ ˆ Aµ = sin θW W3 + cos θW B µ (22. and D µ φ is given by (19.6) Fluctuations about this value are parametrized in this gauge by ˆ φ= 1 √ (v 2 0 ˆ + H) (22. (22. ˆ ˆ ˆ G µν is the U(1) field strength tensor (19. (22.

13) shows that the Aµ field describes a massless particle (to be identified with the photon). We have included ‘R’ components for the neutrinos in the table. the ‘R’ components do interact via ˆ the U(1) field B µ : it is this that allows electromagnetism to emerge free of parityviolating γ5 terms. can be read off from (22.19) where g is the SU(2)L gauge coupling constant.Table 22. νµR . t νeL .8).9) tells us that the mass of the Higgs boson is √ √ (22. fair to say that in the original Standard Model the neutrinos were taken to be massless. at present. tL uR .1. we arrive at the assignments shown in table 22. no evidence for any weak interactions coupling to the right-handed field components and it is therefore natural—and a basic assumption of the electroweak theory—that all ‘R’ components are singlets under the weak isospin group. sL . There is.4)). tR dL . ντ R eL . Weak isospin and hypercharge assignments. Crucially. bR φ+ φ0 1/2 0 1/2 0 1/2 0 1/2 0 1/2 1/2 t3 1/2 0 −1/2 0 1/2 0 −1/2 0 1/2 −1/2 y −1 0 −1 −2 1/3 4/3 1/3 −2/3 1 1 Q 0 0 −1 −1 2/3 2/3 −1/3 −1/3 1 0 Feynman rules for the vector boson propagators (in unitary gauge) and couplings. It is. bL dR .18) m H = 2µ = 2λv √ where v/ 2 is the (tree-level) Higgs vacuum value.9)–(22.11) show that the charged W’s have a mass MW = gv/2 (22. ντ L νeR . as we shall see. Still unaccounted for are the right–handed chiral components of the fermion fields. (22. τR uL . however.1. With the help of the weak charge formula (equation (22. cR . νµL . We have seen in chapter 20 that it is Copyright 2004 IOP Publishing Ltd . µR .10) and (22. with no neutrino mixing.12) gives the mass of the Z0 as (22. Equations (22. µL .13) give the tree-level masses of the Higgs boson and the gauge bosons: (22. cL . and are given in appendix Q. and for the Higgs couplings. however.20) MZ = MW / cos θW and (22. (22. τL eR . sR .

24) L ˆ and lˆ f R are the singlets lˆeR = eR etc. Thus the amplitude Copyright 2004 IOP Publishing Ltd . 22. Thus.7.τ ¯ / lˆ f L i Dlˆ f L + f =e.2 The leptonic currents (massless neutrinos): relation to current–current model ˆµ ˆ We write the SU(2)L ×U(1) covariant derivative in terms of the fields W and B µ of section 19.26) is also the amplitude for the µ − νµ − W and τ − ντ − W vertices. The ‘τ · W/2’ terms can be written as ˆµ ˆµ ˆµ ˆµ (W − iW ) (W + iW ) 1 ˆµ τ · W /2 = √ τ+ 1 √ 2 + τ− 1 √ 2 2 2 2 + τ3 µ W 2 3 (22.τ ¯ / lˆ f R i Dlˆ f R (22.15)).µ.23) picks out the process e− → νe + W− for example. In this case. there are no ‘R’ components for neutrinos. We shall mention their masses again in section 22.µ. with the result that the corresponding vertex is given by ig (1 − γ5 ) −√ γ µ . but for the moment we proceed in the ‘massless neutrinos’ approximation.2. except when discussing experimental situations specifically sensitive to neutrino oscillations.6 as ˆ ˆ ˆµ D µ = ∂ µ + igτ · W /2 + ig y B µ /2 and as ˆ D µ = ∂ µ + ig y B µ /2 on ‘L’ SU(2) doublets (22.21) on ‘R’ SU(2) singlets.23) where the lˆ f L are the left-handed doublets lˆ f L = νf ˆ fˆ− (22.26) 2 2 The ‘universality’ of the single coupling constant ‘g’ ensures that (22.25) where τ± = (τ1 ± iτ2 )/2 are the usual raising and lowering operators for the doublets. (22. (22. the ‘ f = e’ contribution to the first term in (22. Consider. first.for many purposes an excellent approximation to treat the neutrinos as massless. The correct normalization for √ ˆ ˆµ ˆµ the charged fields is that W µ ≡ (W1 − iW2 )/ 2 destroys the W+ or creates the ˆ W− (cf (7. the charged leptonic currents.22) The leptonic couplings to the gauge fields therefore arise from the ‘gaugecovariantized’ free leptonic Lagrangian: ˆ Älept = f =e.

29) This is an important equation giving the precise version. in the GSW theory.Figure 22.8) by able √ v/ 2 = 2µ/λ1/2 (cf (17. leading to the amplitude − ig 2 u(µ)γµ (1 − γ5 )u(νµ )u(νe )γ µ (1 − γ5 )u(e) ¯ ¯ 2 8MW (22.93). This comparison gives g2 GF √ = . µ of (22.19) and (22. for the νµ + e− → µ− + νe process considered in section 20. (22.1. we may quote v/ 2 (the vacuum value of the Higgs field): √ v/ 2 174 GeV.98)).28) which may be compared with the form we used in the current–current theory.1. we can deduce √ G F / 2 = 1/(2v 2 ) (22.30) so that from the known value (21. there follows the value of v: v 246 GeV.32) This parameter sets the scale of electroweak symmetry breaking. Putting together (22. but as yet no theory is √ to predict its value. 2 For k 2 MW we can replace the W-propagator by the constant value 2 g µν /MW . 2 8MW 2 (22.8 is 2 i[−g µν + k µ k ν /MW ] g ig (1 − γ5 ) (1 − γ5 ) u(νµ ) u(e) − i √ u(νe )γν − √ u(µ)γµ ¯ ¯ 2 − M2 2 2 k 2 2 W (22.29).19) introduced earlier. W-exchange process in νµ + e− → µ− + νe . (22. equation (20. It is related to the parameters λ.11) of G F . of the qualitative relation (21. Copyright 2004 IOP Publishing Ltd .31) √ Alternatively.27) corresponding to the Feynman graph of figure 22.

we shall have both L and R couplings from (22. e− . τ − .µ.τ 1 − γ5 2 l + cR 1 + γ5 2 lˆ (22.38) (22. up to the usual factors of g ’s and √ 2’s. .33) has the form µ† ˆµ ˆ ˆ† − jCC (leptons)Wµ − jCC (leptons)Wµ (22. written as ˆ ˆ − jNC (leptons) Z µ . Turning now to the leptonic weak neutral current. (22. For the Q = 0 neutrinos with t3 = + 1 . . νµ . and we can write ˆµ jNC (charged leptons) = where l l cL = t3 − sin2 θW Q l = − 1 + sin2 θW 2 l cR g cos θW ¯ l ˆ lγ µ cL l=e.35) ˆµ ˆ Referring to (22.39) (22.36). the charge-changing part of (22. µ (22. 2 ˆµ jNC(neutrinos) = g 2 cos θW (1 − γ5 ) ¯ νl ˆ νl γ µ ˆ 2 (22.37) l where now l = e. The Feynman rules for the lepton–W couplings (appendix Q) can be read off from (22.In general. this will appear via the couplings to the Z0 . For the other (negatively charged) leptons. we find (problem 22.34) ˆµ where the leptonic weak charged current jCC (leptons) is precisely thatused in the current–current model (equation (20. 2 2 Copyright 2004 IOP Publishing Ltd . µ.40) = − sin θW Q l = sin θW .36) where the sum is over the six lepton fields νe .14) for the linear combination of W3 and B µ which represents µ . .1) that ˆ Z ˆµ jNC (leptons) = g cos θW ¯ l ˆ ψ l γ µ t3 l 1 − γ5 2 ˆ − sin2 θW Q l ψl (22.23) can be written as g ¯ (1 − γ5 ) (1 − γ5 ) (1 − γ5 ) e + ν µγ µ ˆ ¯ ˆ µ + ντ γ µ ˆ ¯ ˆ τ Wµ ˆ ˆ ˆ − √ νeγ µ 2 2 2 2 + Hermitian conjugate ( 2 2 .81)). Thus the dynamical symmetry currents of the SU(2)L gauge theory are exactly the ‘phenomenological’ currents of the earlier current–current model. .3 3 ) √ ˆµ ˆµ ˆ where W µ = (W1 − iW2 )/ 2. τ .33).

the Z0 coupling is not pure ‘V − A’.µ.2) reduces to − ig 2 l l u(e)γµ [cL (1 − γ5 ) + cR (1 + γ5 )]u(e) ¯ 2 4 cos2 θW MZ l l × u(µ)γ µ [cL (1 − γ5 ) + cR (1 + γ5 )]u(µ).126)–(20.45) Copyright 2004 IOP Publishing Ltd .20)) in the absence of loop corrections.2. where l gV = − 1 + 2 sin2 θW 2 l gA = − 1 . These relations (22.43) It is customary to define the parameter 2 2 ρ = MW /(MZ cos2 θW ) (22. the amplitude for e− µ− → e− µ− via Z0 exchange (figure 22. ¯ (22.25.38).43) and (22. Z0 -mediated processes reduce to the current–current form at low k 2 . the Feynman rules for lepton–Z couplings (appendix Q) are contained in (22. Z0 -exchange process in e− µ− → e− µ− . As in the case of W-mediated charge-changing processes.28) (i.τ ¯ ˆ lγ µlˆ (22.128).37)– (22. As noted earlier.40) are exactly the ones given earlier.37) and (22.41) (22. by looking for ˆ the piece that couples to Aµ .Figure 22. ¯ ‘neutral current process’/‘charged current process’) is then 2ρ. in particular. The ratio of factors in front of the u .44) which is unity at tree level (see (22. Again. We find that ˆµ jemag = −g sin θW l=e. the couplings are independent of ‘l ’ and hence exhibit lepton universality. in (20. 2 g 2 cos θW ¯ l l ˆ lγ µ (gV − gA γ5 )lˆ l (22. . u expressions in (22. We may also check the electromagnetic current in the theory. For example. The alternative notation ˆµ jNC(charged leptons) = is often used. .e.42) l Note that gV vanishes for sin2 θW = 0.

m s and m b and the V -matrix is the Cabibbo–Kobayashi–Maskawa matrix (Cabibbo 1963. Kobayashi and Maskawa 1973) which generalizes the 2 × 2 GIM mixing introduced in section 20.which allows us to identify the electromagnetic charge e as e = g sin θW (22. Thus.1).49) (22.45).7. Thus jNC(quarks) will be given by the same expression as (20.146).98) of chapter 19.131) to three generations and supplies the factor g/ 2. except that the t3 = − 1 2 components of the L-doublets have to be understood as the flavour-mixed (weakly interacting) states      ˆ ˆ d d Vud Vus Vub  s  =  Vcd Vcs Vcb   s  . We shall discuss this matrix further in section 22.48) √ which generalizes (20.49) as ˆµ jNC(quarks) = g 2 cos θW q q ¯ qγ µ (gV − gA γ5 )q ˆ ˆ q (22. we can alternatively write (22.144). as is of course required.50) (22.51) = − sin θW Q q .1. s and b are the strongly interacting fields with masses m d . ˆ ˆ (22.10. 22. The neutral currents are diagonal in flavour if the matrix V is unitary (see ˆµ also section 22. As for the charged leptons.3 The quark currents The charge-changing quark currents which are coupled to the W± fields have a form very similar to that of the charged leptonic currents. the charge-changing weak quark current is g ¯ (1 − γ5 ) ˆ ¯ ¯ (1 − γ5 ) s + tˆγ µ (1 − γ5 ) b ˆµ ˆ ˆ ˆ ˆ jCC (quarks) = √ uγ µ d + cγ µ 2 2 2 2 (22. as for the leptons.47) ˆ ˆ Vtd Vts Vtb b L b L ˆ ˆ ˆ where d.46) as already suggested in (19.144)–(20. Note that all the γ5 ’s cancel from (22.7. 2 These expressions are exactly as given in (20. except that now the sum will be over all six quark flavours. The neutral weak quark current is thus ˆµ jNC (quarks) = where cL = t3 − sin2 θW Q q q cR q q g cos θW q q (1 − γ5 ) q (1 + γ5 ) ¯ + cR q ˆ qγ µ cL ˆ 2 2 (22.52) Copyright 2004 IOP Publishing Ltd .

9 that. each with two leptons and two quarks.2) and indicates a remarkable connection. p 384): Nc ( Q u + Q d ) + Q e = 0 (22.73 GeV.1) and (22.20) we predict MZ = MW / cos θW 88. we note one important feature of the Standard Model currents in general.1) and (22. between the facts that quarks occur in three colours and have charges which are 1/3 fractions. ‘d’ and ‘e’ type fields in each family.56) Similarly.29) and (22. The width for W− → e− + νe can be calculated using the vertex (22.28 GeV sin θW 77.2) together ‘vertically’.54) = q t3 . While they can be tolerated in global (non-gauged) currents—and indeed we saw that an anomaly in the strong isospin current is responsible for π 0 → γ γ decay—they must cancel in the symmetry currents of a gauge theory or else renormalizability is destroyed.55) is true for the families in (22.55) where Nc is the number of colours and Q u . (22. which are all determined in terms of only a few basic parameters.23. the then known data were consistent with a single value of θW given (using a modern value) by sin2 θW 0. Q d and Q e are the charges (in units of e) of the ‘u’.4 Simple (tree-level) predictions We noted in section 20. The condition that anomalies cancel in the gauged currents of the Standard Model is a remarkably simple one (Ryder 1996.4. 22.53) (22. (22. we may then predict the value of MW : MW = πα √ 2G F 1/2 1 sin θW 37.where gV = t3 − 2 sin2 θW Q q q gA q q (22. Clearly (22.2) (W− → e− νe ) = ¯ 1 g2 GF M 3 MW = 1/2 W 12 4π 6π 2 205 MeV (22. the leptons and quarks are grouped in three families.57) These predictions of the theory (at lowest order) indicate the power of the underlying symmetry to tie together many apparently unrelated quantitities. ¯ with the result (problem 22.46). Before proceeding to discuss some simple phenomenological consequences. Using (22. using (22. Reading (22.26).58 GeV. before the discovery of the W and Z particles. The theoretical motivation for such family grouping is that anomalies are cancelled within each complete family.58) Copyright 2004 IOP Publishing Ltd . at some deep level. The Standard Model provides no explanation of this connection. Anomalies were discussed in section 18.

37) and (22.Figure 22. the GIM mechanism ensuring that all flavour¯ changing terms cancel. producing an estimated total width of approximately 2. The width for Z0 → ν ν can be found from (22. The total W width for all these channels will therefore be about 1. and MW by MZ .85 GeV.38). 2 we find that ¯ (22. q Cross-sections for lepton–lepton scattering proceeding via Z0 exchange can 2 be calculated (for k 2 MZ ) using the currents (22.60). while the branching ratio for W → eν is B(eν) = (W → eν)/ (total) 11%.62) (Z0 → l l) 76.58) by replacing g/21/2 by ¯ g/2 cos θW . Neglecting GIM-type ¯ ¯ flavour mixing among the two energetically allowed quark channels ud and cs.60) l l using (22. to the three (invisible) neutrino channels 20.59) In making these estimates we have neglected all fermion masses. Charged lepton pairs couple with both cL and cR terms. ¯ ¯ their widths would also be the same.61) ν ν The values cL = 1 .49).1. The total width to u¯ . cR = 0 in (22. giving (Z0 → ν ν ) = ¯ 1 g 2 MZ GF M 3 = 1/2 Z 24 4π cos2 θW 2 12π 152 MeV (22.61) reproduce (22. τ − ντ are the same.) The branching ratio to charged leptons is approximately 3.4%. The widths to µ− νµ . (QCD corrections will increase these estimates by a factor of order 1. c¯ . (22.3%. leading (with neglect of lepton masses) to ¯ (Z0 → l l) = l l |cL |2 + |cR |2 6 g 2 MZ .56).5 MeV.57). using (22. Quark pairs couple as in (22.3. With sin2 θW 0.23.5% and to hadrons (via hadronization of the q¯ channels) about 69. dd. apart from a factor of three for the different colour channels. 4π cos2 θW (22. Neutrino–electron graphs involving Z0 exchange.22 GeV. s¯ and bb channels (allowing u ¯ c s three for colour and neglecting masses) is then 1538 MeV. and the Copyright 2004 IOP Publishing Ltd .

A third lepton–lepton process is experimentally available. The cross-section for (22. Examples are νµ e− → νµ e− and ¯ νµ e− → νµ e− ¯ (22.4. and possibly an interference term.65) by interchanging cL and cR .8.11. as shown in figure 22.64) is found from (22. cR → sin2 θW .67) 1 In the kinematics. Copyright 2004 IOP Publishing Ltd .4.65) where E is the energy of the incident neutrino in the ‘laboratory’ system and y = ( E − E )/ E as before. ∗ ∗ ¯ νe e− → νe e− .1 Equation (22.125).Figure 22. lepton masses have been neglected wherever possible. The interference ¯ same (1 − y) term is negligible for E m e .66) turns out to be given by an expression of the form (22.125) by noting that t = −2m e E y: the |cL |2 l term agrees with the pure V − A result (20. where E is the energy of the outgoing neutrino in the ‘laboratory’ system. while the |cR |2 term involves the 2 factor discussed for ν q scattering in section 20.66) In this case there is a single W intermediate state graph to consider as well as the Z0 one.64) as shown in figure 22. (22. One-W annihilation graph in νe e− → νe e− .65) may be compared with the νµ e− → µ− νe l (charged current) cross-section of (20. The cross-section for (22.65) but with the replacements l cL → 1 2 l + sin2 θW .3.63) (22. l l as was the charged current. Since the neutral current for the electron is not pure V−A. we expect to see terms involving both |cL |2 and |cR |2 .63) is found to be (’t Hooft 1971c) l l l l l l dσ/d y = (2G 2 Em e /π)[|cL |2 + |cR |2 (1 − y)2 − 1 (cR cL + cL cR ) ym e / E] F 2 (22. The cross-section for the anti-neutrino process l l (22. ¯ ¯ method of section 20. ¯ (22.

72) where NF is the number scattered into the forward hemisphere 0 ≤ cos θ ≤ 1 and NB that into the backward hemisphere −1 ≤ cos θ ≤ 0. As mentioned in section 20. the process is mediated by both γ .69) e e e B = 4gA gA Re χ(s) + 8gA gV gA gV |χ(s)|2 f f f f f f (22.19. LEP and SLC. is exactly as calculated in problem 8. the differential cross-section for the scattering of unpolarized e− and e+ is given by πα 2 dσ = [(1 + cos2 θ )A + cos θ B] d cos θ 2s (22.68) one Copyright 2004 IOP Publishing Ltd . Integrating (22. which is therefore the pure single photon contribution.9. (a) One-γ and (b) one-Z0 annihilation graphs in e+ e− → f f¯.71) χ(s) = s/[4 sin θW cos θW (s − 2 2 2 MZ +i Z MZ )]. Note that the term surviving when all the g’s are set to zero. In modern notation. The presence of the cos θ term leads to the forward–backward asymmetry noted in that problem. The forward–backward asymmetry AFB may be defined as AFB ≡ (NF − NB )/(NF + NB ) (22.5. We emphasize once more that all these cross-sections are determined in terms of the Fermi constant G F and only one further parameter.5. Renton (1990) and Winter (2000).68) where θ is the CM scattering angle of the final-state lepton.Figure 22. Calculations of the cross-section were made some 30 years ago (for example Budny (1975)). sin2 θW . Particularly precise determinations of the Standard Model parameters may be made at the e+ e− colliders.70) (22. at energies where the lepton masses may be neglected in the final answers.and Z0 annihilation as shown in figure 22. In lowest order. experimental fits to these predictions are reviewed by Commins and Bucksbaum (1983). Consider the reaction e+ e− → f f¯ where f is µ or τ . s = ( pe− + pe+ )2 and e e e A = 1 + 2gV gV Re χ(s) + [(gA )2 + (gV )2 ][(gA )2 + (gV )2 ]|χ(s)|2 (22.

(22.73) becomes (neglecting the photon contribution) AFB (Z0 peak) = e e 3gA gV gA gV e e {[(gA )2 + gV )2 ][(gA )2 + (gV )2 ]} f f f f . whereas a true parity-violating effect would involve terms odd (linear) l in gA .063. direct evidence for parity violation in l e+ e− → µ+ µ− . 2 Many such measurements have been made ‘on the Z peak’. in any comparison with modern high precision data.5(a). (22.78) is usually denoted by A f : A f = 2gV gA /[(gV )2 + (gA )2 ]. LR f f f (22. i. In that case. through interference between 1γ and 2γ annihilation processes (which have different charge conjugation parity).and final-state bremsstrahlung corrections to figure 22. as well as between initial. QED alone produces a small positive AFB . Longitudinally polarized electrons were inelastically scattered Copyright 2004 IOP Publishing Ltd . However.78) The quantity on the right-hand side of (22.easily finds that AFB = 3 B/8 A.77) (22. for example.73) l For sin2 θW = 0. so they are very 2 small for sin θW √ 23.42) that the gV ’s vanish. term in (22.and right-handed incident electrons: ALR ≡ (σL − σR )/(σL + σR ) for which the tree-level prediction is e e e e ALR = 2gV gA /[(gV)2 + (gA )2 ]. The effect is therefore controlled essentially by the first 0.79) The asymmetry AFB is not. in fact. At s = 29 GeV.e.71). (22. However. (22. the asymmetry is AFB −0.25 we noted after (22.70) that it is even under gA → l −gA . f f f f (22.74) Another important asymmetry observable is that involving the difference of the cross-sections for left. since we see from (22.76) A similar combination of the g’s for the final-state leptons can be measured by forming the ‘L–R F–B’ asymmetry AFB = [(σLF − σLB ) − (σRF − σRB )]/(σR + σL ) LR for which the tree-level prediction is 2 AFB = 2gV gA /[(gV )2 + (gA )2 ].70). at s = MZ in the parametrization (22. all one-loop radiative effects must clearly be considered.75) (22.69) and (22. electroweak-induced parity violation effects in an apparently electromagnetic process were observed in a remarkable experiment by Prescott et al (1978). Indeed.

x 2 ) ˆ 2 s σ = 4π 2 α(1/4 sin2 θW )δ(ˆ − MW ) ˆ (22.80) for massless partons. The total cross-section Copyright 2004 IOP Publishing Ltd . ¯ ¯ Estimates of W and Z0 production in pp collisions may be obtained (see. so that the ¯ √ valence q’s in the proton and q’s in the anti-proton will dominate (at s = 1. We now turn to some of the main experimental evidence. O(αs ) calculations are presented in Hamberg et al (1991) and by van der Neerven and Zijlstra (1992). as usual.5 The discovery of the W± and Z0 at the CERN pp collider ¯ 22.83) = π2 1/2 2 G F MW δ(x 1 x 2 s − 2 MW ). for example.44). An asymmetry between the results.1 Production cross-sections for W and Z in pp colliders The possibility of producing the predicted W± and Z0 particles was the principal motivation for transforming the CERN SPS into a p¯ collider using the stochastic p cooling technique (Rubbia et al 1977.1.5–2 at these energies.8 ¯ TeV. and they should be evaluated at Q 2 = 2 2 MW .4. With s 1/2 = MW ∼ 80 GeV and s 1/2 = 630 GeV ˆ for the pp collider energy. was observed—a clear signal for parity violation. which are often accounted for 2 by a multiplicative factor ‘K ’.81) must.13. As in (14. and the 3 subprocess cross-section σ for q¯ → W± + X is ˆ q σ (p¯ → W± + X) = p 1 3 1 1 dx 1 dx 2 σ (x 1 . This was the first demonstration of parity-violating effects in an ‘electromagnetic’ process: the corresponding value of sin2 θW is in agreement with that determined from ν data.from deuterium and the flux of scattered electrons was measured for incident electrons of definite helicity. we see that the x’s are typically ∼0. QCD corrections to (22. we denote by s the subprocess invariant ˆ s = (x 1 p1 + x 2 p2 )2 = x 1 x 2 s ˆ (22.5. Staff of the CERN pp project 1981). There will be further (O(αs )) corrections. which is of order 1.12) and for two-jet cross-sections in section 14. ¯ 22. depending on the helicities. Quigg 1977) from the parton model.04 and the sea quarks will be expected to contribute).81) where the 1 is the same colour factor as in the Drell–Yan process. The parton model cross-section p¯ → W± + anything is p then (setting Vud = 1 and all other Vi j = 0) ¯ ¯ u(x 1 )d(x 2 ) + d(x 1 )u(x 2 ) u(x 1 )d(x 2 ) + d(x 1 )u(x 2 ) ¯ ¯ 0 0 (22.4 with γ replaced by W or Z0 . in a way analogous to that used for the Drell–Yan process in section 9.3. beginning with the discoveries of the W± and Z0 1983. be included. see also Ellis et al (1996) section 9. appropriate to the Fermilab Tevatron. as shown in figure 22.82) (22.6 (cf figure 9. Leading logarithms will make the distributions Q 2 -dependent. x 0.

5.84) (22.7.3. The rates could. Thus. √ for production of W+ and W− at s = 630 MeV is then of order 6. despite the unfavourable branching ratios. be increased by using the q¯ modes of W and Z0 . the leptonic modes provide the better signatures. which has a cross-section some 103 higher than (22. and (22. being very hard to distinguish from conventional two-jet events produced via the mechanism discussed in section 14. hence (22. as discussed further in section 22. But the detection of these is very difficult. and the calorimetric jet energy resolution capable of resolving such an effect is hard to achieve.1. Preferred direction of leptons in W+ decay.07 nb (22. while a 0 gives about 2 nb. W and Z0 would appear as slight shoulders on the edge of a very steeply falling invariant mass distribution. at Copyright 2004 IOP Publishing Ltd . Parton model amplitude for W± or Z0 production in pp collisions. ¯ Figure 22.84) p represents ∼ 10−8 of the total cross-section.85) √ s = 630 MeV.84).7 nb 0. The total cross-section for p¯ is about 70 mb at these energies.Figure 22. q which have bigger branching ratios.85) is 10 times smaller. Multiplying these by the branching similar calculation for the Z ratios gives σ (p¯ → W + X → eν X) p σ (p¯ → Z + X → e e X) p 0 + − 0.6.5 nb. similar to that shown in figure 9.13.3. of course.

Arnison et al 1983a.11 with reference to ν q and ν q scattering). UA2.2 Charge asymmetry in W± decay At energies such that the simple valence quark picture of (22. but there is a suggestion that the tree-level prediction is on the low side. the best signatures for W and Z production in p¯ collisions are provided by the leptonic modes p p¯ → W± X → e± ν X p p¯ → Z0 X → e+ e− X. The actual distribution ¯ ∗ ∗ has the form ∼ (1 + cos θe )2 . Clearly the agreement with (22.2(syst.87) Reaction (22.4(stat) ± 3. in W− → e− νe .89) MZ = 92. ¯ 22.) GeV.57) is good. Very clean events can be isolated by imposing a modest transverse energy cut—the e+ e− pairs required come from the decay of a massive relatively slowly moving Z0 . p (22. where θe is the angle between the e− and the p (for ¯ ¯ W− → e− νe ) or the e+ and the p (for W+ → e+ νe ).5 ± 1. Similarly. Banner et al 1983). an isolated and approximately backto-back.3(stat. e+ e− pair with invariant mass peaked around 90 GeV (cf (22.9 shows (Geer 1986) the invariant mass distribution for a later sample of 14 UA1 events in which both electrons have wellmeasured energies. the systematic error reflects the uncertainty in the absolute calibration of the calorimeter energy scale.) ± 1. Figure 22.5. a right-handed e+ and a left-handed νe are emitted. we see that angular momentum conservation allows e+ production parallel to the direction of the anti-proton but forbids it parallel to the direction of the proton. Bagnaia et al 1983) will be discussed first.5(syst. Arnison et al 1983b.86) has the larger cross-section. was (22. the W+ is created in the annihilation of a left-handed u quark from the proton ¯ and a right-handed d quark from the p (neglecting fermion masses). by a factor of 10 (cf (22. The signature for (22.87) is simpler and so the Z0 discovery (UA1.86) (22.) GeV. together with the Breit–Wigner resonance curve appropriate to MZ = 93 GeV/c2 .1.81) is valid.3 Discovery of the W± and Z0 at the pp collider and their properties As already indicated in section 22. Figure 22. the e− is emitted preferentially parallel ¯ to the proton (these considerations are exactly similar to those mentioned in section 20. UA2.57)).85)) and was observed first (UA1. based on 13 well-measured pairs. of course. The UA1 result for the Z0 mass was MZ = 93. In the ¯ W+ → e+ νe decay. However.84) and (22. with experimental mass resolution folded in. the kinematics of (22.0 ± 1.22.8 shows the transverse energy distribution of a candidate Z0 event from the first UA2 sample. In both cases. Referring to figure 22.5. Copyright 2004 IOP Publishing Ltd .7.88) The corresponding UA2 result (DiLella 1986). (22.5.87) is.

Figure 22. we have Z (2.Figure 22. Bagnaia et al 1983) in the θ and φ plane. The cell transverse energy distribution for a Z0 → e+ e− event (UA2. Invariant mass distribution for 14 well-measured Z0 → e+ e− decay (UA1. beyond νe .8. th Indeed.19 GeV. νµ and ντ ) which contribute to the width through the process Z0 → Copyright 2004 IOP Publishing Ltd .57) to a value MZ 91.16 Nν ) GeV (22. for any fermion family additional to the three known ones. loop corrections adjust (22.4. Geer 1986).90) from section 22. in excellent agreement with the current experimental value (Hagiwara et al 2002). If we assume that.9. where θ and φ are the polar and azimuth angles relative to the beam axis. where Nν is the number of additional light neutrinos (i.e. The total Z0 width Z is an interesting quantity. only the neutrinos are significantly less massive than MZ /2.5 + 0.

Thus. The advent of LEP provided precision checks on (22. Thus (22. which established Nν = 3. Banner et al 1983). Further. so that ¯ (22. and cannot measure the ν’s. we show data from DELPHI (Abreu et al 1990.35 30 25 20 15 10 5 0 88 89 90 91 92 s (GeV) 93 DELPHI 94 95 96 Figure 22. it is clear that—as in the case of Z0 → e+ e− decay—slow moving massive W’s will emit isolated electrons with high transverse energy. three and four massless neutrino species respectively.90): at the cost of departing from the historical development.) ν ν .e. The cross-section for e+ e− → hadrons around the Z0 mass (DELPHI collaboration.10. In this case an invariant mass plot is impossible. (From Abe. similar in magnitude to that of the e− but opposite in azimuth) were the signatures used for the early event samples (UA1.e. which can be measured by calorimetry and which should balance the transverse energy of the electrons.90) could not be used directly. electrons of high E T accompanied by balancing high missing E T (i. We turn now to the W± . 1991. such electrons should be produced in association with large missing transverse energy (corresponding to the ν’s). Abreu et al 1990). Arnison et al 1983a. The dotted. since we are looking for the eν (µν) mode. full and dashed lines are the predictions of the Standard Model assuming two. UA2.90) can be used as an important measure of such neutrinos (i. ¯ generations) if Z can be determined accurately enough. The determination of the mass of the W is not quite so straightforward as that of the Z. However. The mass resolution of the pp experiments was of the same order as the total expected Z0 width. Abe 1991) in figure 22. since we cannot construct directly an invariant mass plot for the Copyright 2004 IOP Publishing Ltd .10.

92) where φ is the azimuthal separation between peT and pνT.1 (stat. the magnitude of the transverse momentum of the e− .Figure 22. In fact.) ± 1. the transverse 2 2 momentum distribution is given by dσ d cos θ dσ dσ = = d peT d cos θ d peT d cos θ 2 peT MW 1 2 2 M − peT 4 W −1/2 (22. as may be seen 2 from the following argument.94) .93) MW = 81.91). and the most probable value was found by a maximum likelihood fit. We have | pe | = 1 MW and | peT | = 1 MW sin θ ≡ peT . Consider the decay of a W at rest (figure 22.0 (stat. which is peT ≈ 1 MW .11.) ± 2. eν pair: only the missing transverse momentum (or energy) can be attributed to the ν.91) and the last (Jacobian) factor in (22.1 MW = 83. the distribution of events in peT . Thus.) GeV (22. The quoted results were UA1 (Geer 1986) : UA2 (DiLella 1986) : Copyright 2004 IOP Publishing Ltd 1. Early determinations of MW used (22. A Monte Carlo simulation was used to generate MT distributions for different values of MW . This peaking will be smeared by the width and transverse motion of 2 the W. should show a pronounced peaking towards the maximum kinematically allowed value.8 (syst. defined by 2 MT = (E eT + E νT )2 − ( peT + pνT )2 2 peT pνT (1 − cos φ) (22. since some unidentified longitudinal momentum will always be lost down the beam pipe. but sensitivity to the transverse momentum of the W can be much reduced (Barger et al 1983) by considering instead the distribution in ‘transverse mass’.2 ± 1.3 (syst.11).91) produces a strong peaking towards peT = 1 MW .5 ±1.) GeV (22. Kinematics of W → eν decay.

l. the agreement between the experiments and of both with (22.56) is good. We show in figure 22. W → eν transverse mass distribution measured by the CDF collaboration (Abe et al 1995a).20) to obtain sin2 θW via 2 2 sin2 θW = 1 − MW /MZ . one finds that sin2 θW 0. the systematic errors again reflecting uncertainty in the absolute energy scale of the calorimeters. (22.022 (stat.97) Radiative corrections have.12 a more to MW modern determination of MW by the CDF collaboration (Abe et al 1995a). to be applied but one renormalization scheme (see section 22.212 ± 0. The two experiments also quoted (Geer 1986.).96) The weighted average of UA1 and UA2 yielded sin2 θW = 0. DiLella 1986) UA1 : UA2 : W W < 6.56) 80.Figure 22.39 GeV (Hagiwara et al 2002). in general.0 GeV 90% c. Loop corrections adjust (22.95) Once again. (22.222. Using this scheme and quoted values of MW and MZ (Hagiwara et al 2002).96) to a definition of the renormalized sin2 θW to all orders in perturbation theory. Copyright 2004 IOP Publishing Ltd . (22.5 GeV < 7.8) promotes (22. The W and Z mass values may be used together with (22. the predictions again being on the low side.12.

22.2). measured with respect to a direction parallel (anti-parallel) ∗ to the p(p) beam. in the rest frame of the W. for a total of 75 events.6.5.13. Finally. we may say that the early discovery experiments provided remarkably convincing confirmation of the principal expectations of the GSW theory. The expected form (1 + cos θe )2 is followed very closely. Let us recall first that the standard way to introduce the interactions of gauge fields with matter fields (e.g. The W decay angular distribution of the emission angle θe of the positron (electron) with respect to the anti-proton (proton) beam direction. immediately appears to create a fundamental problem as far as the masses of these particles are concerned: we mentioned this briefly at the end of section 19. We now consider some further aspects of the theory. as outlined in the preceding sections. ¯ In summary. figure 22. fermions) is via the covariant derivative Copyright 2004 IOP Publishing Ltd .6 The fermion mass problem The fact that the SU(2)L gauge group acts only on the L components of the fermion fields. background subtracted and acceptance corrected. θe is the e+ (e− ) angle in the W rest frame.13 shows (Arnison et al 1986) the angular distribution of ∗ the charged lepton in W → eν decay (see section 22.∗ Figure 22.

We studied such a reaction in section 21. But suppose. once again.3). for example λ = +1 for the fermion f.3) to check that the ordinary ‘kinetic’ part of a free Dirac fermion does not mix the L and R components of the field: ¯ / ¯/ˆ ¯ / ˆ ˆ ˆ ψ ∂ ψ = ψ R ∂ ψR + ψ L ∂ ψL . mass terms for Dirac fermions will explicitly break SU(2)L . following Appelquist and Chanowitz (1987). contemplate ‘gauging’ the L and the R components / / differently. The same is also true for Majorana fermions (see appendix P) which might describe the neutrinos. therefore. Now it is a simple exercise for SU(2) fields (compare problem 18.98) acting on t = 1/2 doublets. Z0 and γ with the usual SU(2)L × U(1) 2 2 couplings.100) (see equation (18.3 in the context of unitarity violations (in lowest-order perturbation theory) for the IVB model. Appelquist and Chanowitz (1987) considered first the case in which ‘ f ’ is a lepton with t = 1 . in the case of QCD (cf (18.99) but this was because QCD conserves parity and must. They found that in the ‘right’ helicity channels for the leptons (λ = +1 for f¯.4.replacement Wµ ∂ µ → D µ ≡ ∂ µ + igτ · W µ /2 (22. a fermion–anti-fermion annihilation process of the form f f¯ → W+ W− .14. Of course. Consider. in principle. we examine channels involving the ‘wrong’ helicity component.100) is not invariant. a Dirac mass term has the form ¯ ˆ ¯ ˆ ˆ ˆ −m(ψ L ψR + ψ R ψL ) (22. to violations of unitarity and then of renormalizability. Such cancellations are a feature of gauge theories.5 was cancelled by that of the diagram shown in figure 22. The sum of the amplitudes tends to a constant as s (or E 2 ) → ∞. It is easy to see that if only ψL is subject to a transformation of the form (22.41) for example) and it precisely couples the L and R ˆ components.99). Then it is found that the cancellation no longer occurs and we shall ultimately have a ‘nonrenormalizable’ problem on our hands. t3 = − 1 coupling to W’s. and represent one aspect of the renormalizability of the theory. Weak interactions are parity-violating and the SU(2)L covariant derivative acts only in the second term of (22. (22. λ = −1 for f ) the bad highenergy behaviour associated with a fermion-exchange diagram of the form of figure 21. Thus. treat L and R components the same. as we indicated at the end of section 21. However. for example.99) Thus we can. in the sense that it will lead. (22. but having an explicit (Dirac) mass m f . then (22. all over again. This kind of explicit breaking of the gauge symmetry cannot be tolerated.39)) the replacement ∂ → D was made equally in each term on the right-hand side of (22.101) 0 0 where the subscript indicates the λ = 0 (longitudinal) polarization state of the W± . Copyright 2004 IOP Publishing Ltd .

However.Figure 22. at the Lagrangian ˆ level. then this will generate a fermion mass. But now insert just the vacuum value (22. Thus. (22.103) L ˆ the Higgs doublet φ and the R-component of the electron field: ¯ ˆ ˆ ˆˆ Äe = −ge(l¯ˆeL φ eR + eR φ †lˆeL ).104) In each term of (22.105) Copyright 2004 IOP Publishing Ltd . for example. in a theory with spontaneous symmetry breaking. E t ∼ 18 TeV.104): we find the result ˆ Äe (vac) = −ge √ Yuk v ¯ ¯ ˆ ˆ (e L eR + e R eL ) ˆ ˆ 2 (22. for m t ∼ 175 GeV. Attending to all the factors in the partial wave expansion gives the result that the unitarity bound will be saturated at E = E f (TeV) ∼ π/m f (TeV). by such a term. which multiplies the SU(2)L scalar R-component. ˆ Yuk (22. One-Z0 and one-γ annihilation contribution to fλ=−1 f¯λ=1 → W+ W− .4). which here we can take to be m f / s. the two SU(2)L doublets are ‘dotted together’ so as to form an SU(2)L scalar. This would constitute a serious flaw in the theory.6) of φ into (22. even though the breakdown occurs at energies beyond those currently reachable.104). Thus. The linear σ -model of (18. Consider the electron.14. and let us hypothesize such a coupling between the electron-type SU(2) doublet lˆeL = νe ˆ e− ˆ (22. 0 0 An estimate of the energy at which this will happen can be made by recalling that the ‘wrong’ helicity state participates only by virtu√of a factor (m f /energy) e (recall section 20.102) The estimate obtained by Appelquist and Chanowitz differs only by a factor of √ 2. (22. The typical bad highenergy behaviour for an amplitude Å was Å ∼ G F s. there is a way of giving fermion masses without introducing an explicit mass term in the Lagrangian.72) shows how. and the symmetry is preserved. which we expect to be modified here to √ √ Å ∼ G F sm f / s ∼ G F m f s.104) is SU(2)L-invariant. if a fermion has a ‘Yukawa’-type coupling to a scalar field which acquires a vev.

One-H annihilation graph.107) ¯e H .7).15. allowing us to make the identification √ m e = ge v/ 2.1.104) will generate a coupling between the electron and the Higgs fields of the form √ ¯ˆ ˆ ¯ˆ ˆ −ge ee H / 2 = − (m e /v)ee H ˆ ˆ (22.102)—and by this time the reader will not be surprised to be told that this is indeed the case.110) Copyright 2004 IOP Publishing Ltd . must be included.108) = − (gm e /2 MW )e ˆ ˆ Such a coupling. But we learned in u section 12. At first sight it might seem that this stratagem will only work for the t3 = − 1 2 ˆ components of doublets.14.109) d d then the charge conjugate states iτ2 u∗ d∗ ˆ Thus if. (22.100). in addition to the f -exchange graph analogous to figure 21. if present for the process f f¯ → W+ W− considered earlier. in our case.106) When oscillations about the vacuum value are considered via the replacement (22. the term (22. ˆ (22. will 0 0 mean that.3 that if a pair of states forming an SU(2) doublet transform d by u u = e−iα·τ /2 (22.15.5 and the annihilation graph of figure 22.Figure 22. which is exactly a (Dirac) mass of the form (22. The presence of the fermion mass in the coupling to H suggests that this graph might be just what is required to cancel the ‘bad’ high-energy behaviour found in (22. φ is the SU(2) doublet ˆ φ= transform in exactly the same way. because of the form of 0|φ|0 . 1 ˆ ˆ ˆ √ (φ1 − iφ2 ) ≡ φ + 2 1 ˆ ˆ ˆ √ (φ3 − iφ4 ) ≡ φ 0 2 (22. a further graph shown in figure 22.

This does appear to indicate that we are dealing here with a ‘phenomenological model’.1). quarks as well as leptons. But one must admit to a certain uneasiness concerning the enormous difference in magnitudes represented by the couplings gνe . however. Suppose. . We will look more closely at the quarks in the next section.113) √ which is precisely a (Dirac) mass for the neutrino. we could consider economically dropping it altogether. making do with just the νeL component. transforming in just the same way as φ . . . which we may take to be νeL and νeR . it is certainly possible to construct a Dirac-type mass term of the form (22. . With these. In this case.110) and ¯ (22.111). If m νe < 1 eV. while gt ∼ 1! Besides. which is that they could be Majorana (not Dirac) fermions. to acquire a mass by the same ‘mechanism’.113). namely ¯ ˆ ˆ ¯ ˆ† −gνe (lˆeL φC νeR + ν eR φClˆeL ) ˆ (22. K− ) isospin doublets in SU(3)f ).6.then the charge conjugate field ˆ ˆ φC ≡ iτ2 φ ∗ = 1 √ (φ3 + iφ4 ) 2 1 − √ (φ1 + iφ2 ) 2 ≡ ¯ ˆ φ0 ˆ −φ − (22.114) Now we know that the charge-conjugate field transforms under Lorentz transformations in the same way as the original field (see appendix P) and so we can use (ˆ eL )C to form a Lorentz invariant ν (ˆ eL )C νeL ν Copyright 2004 IOP Publishing Ltd (22.112) ˆ then yields √ ¯ ˆ ¯ ˆ −(gνe v/ 2)(ν eL νeR + ν eR νeL ) ˆ ˆ (22. if we set gνe v/ 2 = m νe . But since. then gνe < 10−11. after all. the νeR component has zero quantum members both for SU(2)L W-interactions and for U(1) B-interactions (see table 22. then. whereas the use of the Higgs field ‘mechanism’ in the W–Z sector is quite economical. rather than a ‘theory’. once more.115) . As far as the neutrinos are concerned.112) assuming now the existence of the field νeR . ν ˆ †T (22. that we keep only the field νeL .111) ˆ is also an SU(2) doublet. ge .71)) ¯T ˆ (ˆ eL )C = iγ2 γ0 ν eL = iγ2 νL . only two possibilities exist for neutrinos. .111) may be thought of as analogous to the (K+ . Note that the vacuum value (22. and their anti-particles) which exist for massive Dirac particles. rather than the four degrees of freedom (νeL . gt. there is another possibility. as indicated in section 20. With the help of φC we can write down another SU(2)-invariant coupling in the νe –e sector. ((22. (22. νeR . . K0 ) and (K. in the present case it seems rather unsatisfactory simply to postulate a different ‘ g ’ for each fermion–Higgs interaction. Its charge-conjugate is defined ˆ by (see (20.6).6) will now appear in the ˆ upper component of (22. In the Higgs vacuum (22. It is clearly possible to go on like this and arrange for all the fermions.

116) is not invariant under a global U(1) phase transformation νeL → e−iα νeL ˆ ˆ (22. so that (22. using only the Standard Model Higgs φ. we have † (ˆ eL )C νeL = νeL . which is actually conserved by all the Standard Model interactions. as a low-energy limit of a theory defined at a higher mass scale. (22.1) operator to form a singlet. ν ˆ ˆT ν (22. is ˆT − geM ¯ ˆ† ˆ (lˆeL φC )T iσ2 (φClˆeL ). The typical form of such ˆ operator we need. This means that we cannot ˆ form an SU(2)L invariant with it. however. we could generate suitable ‘effective’ operators via loop corrections.117) in our representation (20. as the current– current model is the low energy limit of the GSW one. However. t3 ) quantum numbers of the term are (1. much as we generated an effective operator representing an anomalous magnetic moment interaction in QED (cf section 11.121) Copyright 2004 IOP Publishing Ltd .which has mass dimension M 3 . since two νeL ˆ operators appear rather than a νe and a νe (which would lead to L e conservation). Thus.120) which would correspond to lepton number (if accompanied by a similar transformation for the electron fields).119) would vanish for classical (commuting) fields.7).50)) and (22. in order to generate a term νeL iσ2 νeL .1). Thus—as. Hence. we may write a ‘Majorana mass term’ in the form ¯ ν − 1 m M [(ˆ eL )C νeL + ν eL (ˆ eL )C ] ν ˆ ˆ (22. since 1 the latter has t = 2 and cannot combine with the (1.117) is effectively (ˆ eL )C νeL = νeL (iσ2 )ˆ eL . perhaps. Now iγ2 γ0 = −iσ2 0 0 iσ2 . such an effective operator could not be generated in perturbation theory.38). the operator would have to violate lepton number conservation. in fact. followed by symmetry breaking.116) 2 where the 1 2 is conventional. It is at once apparent that the mass term (22. M (22. we cannot make a ‘tree-level’ Majorana mass by the mechanism of Yukawa coupling to the Higgs field. only its two lower components are present (cf ˆ (20. But whatever it is. we already knew— Majorana neutrinos do not carry a lepton number.119) which is that. There is a further interesting aspect to (22. Thus. ˆ ˆ† the (t. Written out in more detail.119) Note that iσ2 is an anti-Hermitian 2 × 2 matrix. It could arise. − iγ2 γ0 νeL = νeL iγ2 γ0 νeL ν ˆ ˆT ˆ ˆT ˆ (22.118) But since νeL is an L-chiral field.

7. u L1 = u L . These considerations are tending to take us ‘beyond the Standard Model’. for example. in the spirit of the discussion in section 11. that the operator ‘(lφ)(φl)’ in (22.121) will generate the required Majorana mass term. 2 and 3 refer to the family ˆ ˆ ˆ number. In fact. M (22. dL1 = dL . However. Suppose. which is why we introduced the factor M −1 in the coupling: it is indeed a non-renormalizable effective interaction. most importantly. The labels 1. for the sake of argument. the d ones to the t3 = − 1 2 2 components.7 Three-family mixing 22. so we shall not pursue them at any greater length.104) and (22. with no mixing at all. 22. the only renormalizable Yukawa coupling is of the form ¯ ˆ ˆ ˆ ‘ψ ψ φ’. section P.Note. an effective coupling of ‘natural’ size geM ∼ 1 emerges from this argument. Apart from the symmetry.7. which has a dimensionless coupling (as in the ge and gνe of (22. We may interpret M as the mass scale at which ‘new physics’ enters.1 Quark flavour mixing We introduce three doublets of left-handed fields: qL1 = ˆ u L1 ˆ ˆ dL1 qL2 = ˆ u L2 ˆ ˆ dL2 qL3 = ˆ u L3 ˆ ˆ dL3 (22. Further discussion of neutrino masses is contained in appendix P. if we drop this we might as well abandon the whole motivation for the ‘gauge’ concept.124) which transform in the now familiar way under SU(2) L × U(1).123) and the corresponding six singlets u R1 ˆ ˆ dR1 u R2 ˆ ˆ dR2 u R3 ˆ ˆ dR3 (22. with m M ∼ geM v2 ∼ geM 10−2 eV. after all. the only other theoretical requirement is renormalizability—for. we must now generalize the discussion to the three-family case. We ˆ have to consider what is the most general SU(2)L× U(1)-invariant interaction between the Higgs field (assuming we can still get by with only one) and these various fields. (22. After symmetry breaking.112)). there is no a priori requirement for it to be ‘diagonal’ in Copyright 2004 IOP Publishing Ltd . and to their ‘R’ partners. Instead. This implies (as in the discussion ¯ ˆ ˆ ˆ ˆ of the Higgs potential V ) that we cannot have terms like (ψ ψ φ)2 appearing— −4 and would be non– which would have a coupling with dimensions (mass) renormalizable.122) Thus.121) has mass dimension five. etc. then. this was M ∼ 1016 GeV (a scale typical of Grand Unified Theories). The u-fields ˆ ˆ correspond to the t3 = + 1 components of SU(2)L .2. if indeed the mass of the νe is of order 10−2 eV. just like the current–current one.

125) and a sum on the family indices i and j (from 1 to 3) in (22. After symmetry breaking.  ˆ dL1 ˆ ˆ dL =  dL2  ˆ dL3  etc (22. an interaction of the form (summing on repeated indices) ¯ ˆ ˆ ¯ ˆˆ ˆ ˆ ˆ Äψφ = ai j q Li φC u R j + bi j q Li φ dR j + h. 2 So.130) (22. we can still manage with only the one Higgs field. therefore. It couples to each fermion with a strength proportional to the mass of that fermion. divided by MW .c. u ˆ ˆ Lα=t ≡ tˆL .129) (22.132) Rather remarkably.127) becomes ˆ Äq H = − 1+ ˆ H v ¯ˆ ¯ˆ ˆ [m u u u + · · · + m b bb]. Feinberg et al 1959) that m u and m d can both be made Hermitean. where ‘h. γ5 -free and diagonal by making four separate unitary transformations on the ‘family triplets’:  u L1 ˆ ˆ u L =  u L2  ˆ u L3 ˆ  via (u) ˆ u Lα = (UL )αi u Li ˆ (d) ˆ ˆ dLα = (UL )αi dLi (u) u Rα = (UR )αi u Ri ˆ ˆ (d) ˆ ˆ dRα = (UR )αi dRi . we find (problem 22.’ stands for ‘Hermitian conjugate’. using the gauge (22.104) and (22.3) that ˆ Äfφ = − 1+ ˆ H v ¯ ¯ ˆ ˆ [u Li m uj u R j + d Li m dj dR j + h. The allowed generalization of (22.112) is. ‘α’ is the index of the ‘mass diagonal’ basis and ‘i’ is that of the ‘weak interaction’ basis.2 Then (22.127) where the ‘mass matrices’ are v m uj = − √ ai j i 2 v m dj = − √ bi j . ˆ Copyright 2004 IOP Publishing Ltd . It can be shown (Weinberg 1973. dLα=s ≡ sL .c.7). ˆ (22.] ˆ i ˆ i (22.126) (22. i 2 (22. the m u and m d matrices could involve a ‘γ5 ’ part as well as a ‘1’ part in Dirac space. for example.131) In this notation.c. etc.128) Although we have not indicated it.125) is assumed. qLi = ˆ u Li ˆ ˆ dLi (22.the weak interaction family index i.

ˆ 2 g ¯ (u) (d)† ˆ ˆ = − √ u Lα [(UL )αj (UL ) jβ ]γ µ dLβ Wµ + h. though it is unitary. Kobayashi and Maskawa 1973). ˆ 2 Vαβ ≡ [UL UL (u) (d)† (22.Ri ˆ (cf (22.Rα .B = i(u L j . Indeed.c.134) then has the form g ˆ ¯ ¯ ¯ ˆ ˆ ˆ − √ Wµ [u L γ µ dL + cL γ µ sL + t L γ µ bL ] + h.4.Rα and dL.137) It is important to know how many independent parameters the CKM matrix contains.Now consider the SU(2)L ×U(1) gauge-invariant interaction part of the ˆ Lagrangian.Ri and dL. (22. The interaction (22.133) where a sum on j is understood. The charged current processes.135) is not diagonal.22)). d L j )γ µ (∂µ + igτ · W µ /2 + ig y Bµ/2) ˆ uL j ˆ ˆ dL j ¯ ¯ ˆ ˆ ˆ ˆ ˆ + iu R j γ µ (∂µ + ig y Bµ /2)u R j + id R j γ µ (∂µ + ig y Bµ /2)dR j ˆ (22. − √ (u L j . This now has to be rewritten in terms of the ˆ mass-eigenstate fields u L. ˆ Problem 22.c.133) and this spoils the argument used in problem 22.139) Copyright 2004 IOP Publishing Ltd .130) and (22. that is. ˆ ˆ bL (22. ˆ ˆ ˆ 2 where    ˆ dL Vud  s  =  Vcd ˆL ˆ Vtd bL Vus Vcs Vts  Vub Vcb   Vtb  ˆ dL sL  .133) is diagonal in the mass basis. Written out in terms of the ‘weak interaction’ fields u L. and ∗ ∗ ∗ Vub Vud + Vcb Vcd + Vtb Vtd = 0. The unitarity relation V † V = I provides nine conditions. d L j )γµ τ+ Wµ ˆ ˆ dL j 2 g ¯ ˆ ˆ = − √ u L j γ µ dL j Wµ + h.c.136) (22. however.131) are unitary.138) (22. A general 3×3 complex matrix has 2×32 =18 real parameters. namely |Vud |2 + |Vcd |2 + |Vtd |2 = 1 together with two other similar diagonal equations.4 shows that the neutral current part of (22. provided the U matrices of (22.21) and (22. it is ¯ ¯ ˆ ˆ ˆ ˆ Ä f. This is the well known CKM matrix (Cabibbo 1963.134) where the matrix ]αβ (22.c.25) we find that the charged current piece is ˆ ÄCC = g ¯ uL j ˆ ¯ ˆ ˆ + h. using (22. involve the non-diagonal matrices τ1 and τ2 in (22.W. the neutral current interactions do not change the flavour of the physical (mass eigenstate) quarks.

132) real). Thus.044) (0.9990 − 0.044)  .219 − 0.141) we see that the elements which are least well determined are Vtd and Vub . The standard parametrization of V (Hagiwara et al 2002) is   s12 c13 s13 e−iδ13 c12 c13 V =  −s12 c23 − c12 s23 s13 eiδ13 c12 c23 − s12 s23 s13 eiδ13 s23 c13  iδ13 iδ13 s12 s23 − c12 c23 s13 e −c12 s23 − s12 c23 s13 e c23 c13 (22.140) where ci j = cos θi j and si j = sin θi j . it is helpful to note the 90% confidence limits on the elements of the matrix V as quoted by Hagiwara et al (2002).004 − 0. (22. (22. taking into account the unitarity constraints:   (0.9732 − 0. with Vud ≈ 1 ≈ Vtb . If all the elements of the matrix were real.226) (0. there are really only five removable phases. From (22. The remaining six parameters in the unitary V are.038 − 0.9741 − 0. parametrized by three real numbers. moreover. This leads to a parametrization due to Wolfenstein (1983). Vcd Vcb has been chosen to lie along Copyright 2004 IOP Publishing Ltd . to relate these two elements by ∗ Vub + Vtd ∗ s12 Vcb .and five other similar ‘off–diagonal’ equations.219 − 0.143) Aλ2 −λ 1 − λ2 /2 3 (1 − ρ − iη) 2 Aλ −Aλ 1 retaining terms up to O(λ3 ). we write Vcb 3 (ρ − iη) with A 1 and |ρ − iη| < 1.140) is often satisfactory. However.135). it would be orthogonal ( V T V = I ) and could be thought of as a ‘rotation’. This would seem to give enough freedom to redefine away six phases from V as defined in (22. phases. s23 = Aλ2 and Vub = s13 e−iδ13 = Identifying s12 = λ.0048)  (0. without loss of generality.144) This relation may be represented as a triangle in the complex plane as shown in ∗ figure 22.9993) From this it follows that the mixing angles are small and. therefore.014) (0. (22. Before proceeding further. satisfy a definite hierarchy θ23 θ13 .141) (0. We may use the unitarity relation (22.16 where.226) (0. The phase of each of the quark fields is arbitrary and may be freely changed (though qL and qR must change in the same ˆ ˆ way to keep the mass terms in (22. we can see that V is left invariant if we change all the quarks by the same phase.9748) (0.037 − 0.142) 1 θ12 Thus a small-angle approximation to the angles θi j in (22.139). This gives Aλ   λ Aλ3 (ρ − iη) 1 − λ2 /2  V  (22. and V may therefore be parametrized in terms of three real ‘rotation angles’ and one phase.9756) (0.0025 − 0.

t) ˆ ψ(x. t). for CP to be conserved.5) that (for example) † ¯ ˆˆ ¯ ˆ ˆ ˆ ˆˆ (CP)u L Vud γ µ dL W µ (CP)−1 = d L Vud γ µ u L W µ ˆ ˆ ˆ (22. t)P−1 = γ0 ψ(−x.Figure 22. t). the 2 × 2 mixing matrix (following the same counting of parameters as before) can have only one real parameter.150) where the fact that the fields are evaluated at x will not matter since we integrate over x in the action. The existence of the phase δ13 represents a fundamental difference from the simple 2 × 2 mixing of the original Cabibbo–GIM type considered in section 20. The triangle can also be rescaled so that its base is of ¯ unit length: the apex is then at ρ = ρ(1 − λ2 /2). The action of P and C on fermionic fields was discussed in sections 20.144). This corresponds to setting all angles θi 3 to zero in (22.149) Careful application of these relations then shows (problem 22.143). But the corresponding term from the ‘h.c. the horizontal. the Cabibbo (GIM) angle θ .16.136) is † ¯ ∗ ˆ d L Vud γ µ u L W µ . η = η(1 − λ2 /2) in terms of the ¯ parameters in (22. (22. The ‘unitarity triangle’ representing equation (22.140).145) Vtd = |Vtd |e−iβ and from (22. The reason that the appearance of the phase δ13 in (22.146) (22. ˆ ˆ ˆ C Hence. the entries in the V -matrix must be real. t). ˆˆ ˆˆ ˆ ˆ† CP W µ (x. t)P−1 = iγ 2 ψ †T (x.140) is so significant is that it implies ˆ ˆ (if δ13 = 0) that CP is violated. With three families (though not with only two) the possibility exists of having a Copyright 2004 IOP Publishing Ltd .140) γ = δ13 . The angles β and γ are defined by Vub = |Vub |e−iγ (22. This is known as the ‘unitarity triangle’. ˆ ˆ (22. In that case.’ part in (22.151) It follows that.148) (22. t)(CP)−1 = W µ (−x.147) (22. We also have ˆˆ ˆ ˆˆ ˆ CPψ(x. t)(CP)−1 = iγ 2 γ 0 ψ †T (−x.3 and 20.10.5: ˆ ˆ ˆ ˆ Pψ(x.

2 Neutrino flavour mixing An analysis similar to the previous one can be carried out in the leptonic sector. since its discovery in 1964 (Christenson et al 1964). however. For a full introduction to the phenomenology of CP violation we refer the reader to Leader and Predazzi (1996. In fact. We would also have lepton mass ¯ ¯ ˆ terms (ν Li m νj νR j + h. namely the MNS matrix (Maki et al 1962. It is.153). ˆ ˆ ˆ (22.) and (eLi m ej eR j + h. are all proportional to the quantity 2 s12 s13 s23 c12 c13 c23 sin δ13 (22. and to Branco et al (1999) and Bigi and Sanda (2000). As an indication. which is a major motivation for the construction of B factories.) in analogy with (22. We would then have leptonic flavour mixing in charged current processes.153) when averaged.143)). as was stressed by Kobayashi and Maskawa (1973). or differences of rates. It can be shown (Jarlskog 1985a.127).7. see also Pontecorvo 1967). According to the summary in the current Review of Particle Properties (Hagiwara et al 2002) all processes can be quantitatively understood by one value of the CKM phase δ13 (= γ ) = 59◦ ± 13◦. b) that CP-violating amplitudes. 22.143) (they involve third-generation mixing and are O(λ3 )). In the Standard Model. a very small effect. The value of β = 24◦ ± 4◦ as determined from the overall fit is consistent with (22. as seen from the magnitudes of Vub and Vtd in (22. present experimental results from Babar (Aubert et al 2002) and Belle (Abe et al 2002) yield sin 2β = 0. or if—improbably—its eigenvalues neutrino mass matrix m ˆ were all equal.08 (22.155) so as to reduce the charged current term to family-diagonal form.154) where V (l) is the leptonic analogue of the CKM matrix. via interactions of the form (l) ¯ ν Lα Vαβ γ µ eLβ Wµ + h. But it is now Copyright 2004 IOP Publishing Ltd .c. Larger effects are expected in the B0 system.c. CP violation in the decays of neutral K mesons is an intensively studied phenomenon. in that case.152) which is just twice the area of the unitarity triangle. we would be free to redefine νL and νR by ˆ ˆ ˆ ν L = V (l)† νL ˜ ˆ ˆ ν R = V (l)† νR ˜ (22. chapters 18 and 19).78 ± 0.c. it is interpreted as arising from the phases in the Vub and Vtd elements (see (22. If the ˆ i ˆ i ˆ ν was identically zero. It is these˜states that we would identify with the physical neutrino states.CP-violating phase parameter δ13 .

Bahcall (1989). Bilenky (2000) and Kayser in Hagiwara et al (2002). Kayser et al (1989).158) from measurements at SLD. experimentally. The asymmetry parameter Ae (see (22. indicating that neutrinos do have (different) masses.00216 (22. We saw following equation (22. Such remarkable precision is a triumph of machine design and experimental art—and it is the reason why we need a renormalizable electroweak theory. this has an interesting consequence for the parametrization of the matrix V (l) .159) (22. quantummechanical oscillations between neutrinos of different (leptonic) flavour can be observed to occur over macroscopic distances. 1) (22.clear.156) do not affect neutrino oscillations but do affect the rate for neutrinoless double β-decay (see section 20.79)) is (Abe et al 2000) Ae = 0.0021 GeV (22.6). Thus. Because the neutrino mass differences are (apparently) so small. eiα2 /2 . including higherorder corrections.6). The reader will recall that it is an open question whether neutrinos are Dirac or Majorana particles (sections 20. that neutrino flavour mixing does take place.15138 ± 0. two of the phases which could be removed from the 3 × 3 quark mixing matrix V .061 GeV. These are just three examples from the table of 35 observables listed in the review of the electroweak model by Erler and Langacker in Hagiwara et al (2002).1876 ± 0.451 ± 0. Boehm and Vogel (1987). the matrix V (l) can be parametrized as V (l) (Majorana) = V (CKM type) × diag(eiα1 /2 . as they carry no lepton number.156) and we have three CP-violating phases. D0: Abbott et al (2000). cannot be removed from the leptonic analogue matrix V (l) if the neutrinos are Majorana particles.119) that global phase transformations (ordinarily corresponding to a number conservation law) cannot be freely made on Majorana fields. The subject is extensively covered in a number of books and reviews. 22. is quoted by Erler and Langacker as χ 2 /degree of freedom Copyright 2004 IOP Publishing Ltd . In the Majorana case. We note that the extra CP-violating phases in (22. UA2: Alitti et al (1992)) is MW = 80.6 and 22.157) from the Z lineshape at LEP1 (Tournefier 2001). for example Mohapatra and Pal (1991). If they are of Majorana type.8 Higher-order corrections The Z0 mass is presently (2002) determined to be MZ = 91. The W mass (CDF: Affolder et al (2001). The overall fit to the data.

In the electroweak case. θ23 ν ν ν MNS angles θ12 . can be separated off. especially insofar as they have phenomenological implications. some loop corrections are proportional to the square of the top mass (see (22. the parameters in the Lagrangian are gauge couplings g. By contrast.2σ deviation in the hadronic charge asymmetry (3/4) Ae Ab .3/38. θ13 . Some information about MH can therefore be gained by seeing how the fits vary with MH . and the Higgs mass. This reasonably strong numerical consistency lends impressive support to the belief that we are indeed dealing with a renormalizable spontaneously broken gauge theory.163) (22. so the fits are not very sensitive to MH . α2 ?). we shall see in equation (22. we obtain cut-off independent results from loop corrections in a renormalizable theory by taking the values of certain parameters—those appearing in the original Lagrangian—from experiment. In fact. the tree-level vacuum value v).10. Instead. θ13 . (22. After all this (and earlier) emphasis on the renormalizability of the electroweak theory. leaving g. µ2 Higgs–fermion Yukawa couplings g f phase δ13 CKM angles θ12 . Second.174) that the dependence on MH is only logarithmic—it acts rather like a cut-off. requiring quite a bit more formal machinery which would be outside the intended scope of this book (suitable references include Altarelli et al (1989). But the fact is that this is a very complicated and technical story. Actually. The probability of a larger χ 2 is 14%: one of the major discrepancies is a 3. g and one combination of λ and µ2 (for instance. another is a 2.9. and the introduction to one-loop calculations in QED at the end of volume 1. g Higgs potential parameters λ. have had to be introduced.161) (22. gµ − 2. θ23 ν phase δ13 (+α1 . the one remaining unobserved element in the theory—the Higgs boson—has a mass MH which is largely unconstrained by theory (see section 22.2) and it is therefore a parameter in the fits. and consequently very tight bounds could be placed on m t via its virtual presence (in loops) before its real presence was confirmed. First. the reader perhaps now has a right to expect an exposition of loop corrections in the electroweak theory.160) (22. in more than one way. because no extra parameters. one can turn this around.173)). These three parameters are usually replaced by the equivalent and more Copyright 2004 IOP Publishing Ltd .= 47.162) (22. as we shall discuss shortly and in section 22. very careful analysis of small discrepancies between precision data and electroweak predictions may indicate the presence of ‘new physics’. especially the pedagogical account by Consoli et al (1989). and the equally approachable lectures by Hollik (1991)).164) The fermion masses and mixings.5σ deviation in the muon anomalous magnetic moment. according to a welldefined procedure (‘renormalization scheme’). As we have seen. we want to touch on just a few of the simpler and more important aspects of one-loop corrections. not in the original Lagrangian.

165)–(22. Hollik 1990. (22. Another scheme is the modified minimal subtraction (MS) scheme (section 15.167) MZ = 1 v g2 + g 2 2 1 GF = √ (22. L3: Acciari et al 2000).167) (together with MH and the fermion masses and mixings). at tree level α = g 2 g /(g 2 + g )4π 2 2 GF (22. VENUS: Okada et al 1998. for example.168) 2v 2 but these relations become modified in higher order. for example. In this scheme.convenient set α (Mohr and Taylor 2000) (22. and Copyright 2004 IOP Publishing Ltd . Kennedy et al 1989. We shall continue here with the scheme defined by (22.169) is promoted into a definition of the renormalized sin2 θW to all orders in 2 perturbation theory.170). for reviews see Langacker 1995). ˆ2 This is the scheme used by Erler and Langacker in Hagiwara et al (2002). the differences are very small).171) where r includes the radiative corrections relating α. OPAL: Abbiendi et al 2000. one can derive radiatively-corrected values for physical quantities in terms of the set (22. 1984. g and v . of course.170) The radiatively-corrected value for MW is then 2 MW √ (πα/ 2G F ) = 2 sW (1 − r ) (22. G F .165) (Marciano and Sirlin 1988. it being then denoted by sW : 2 2 2 sW = 1 − MW / MZ . related to g . But a renormalization scheme has to be specified. Kennedy and Lynn 1989. MW and MZ . α( MZ ). These are. The renormalized parameters will ‘run’ in the way described in chapters 15 and 16: the running of α . van Ritbergen and Stuart 1999) (22. at any finite order (though. We cannot go into detail about all the contributions to r but we do want to highlight two features of the result—which are surprising. After renormalization. has been observed directly (TOPAZ: Levine et al 1997. the tree-level formula 2 2 sin2 θW = 1 − MW / MZ (22. Attention is then focused on sZ ≡ sin2 θW (MZ ). in practice.166) MZ (Tournefier 2001). important phenomenologically.5) ˆ which introduces the quantity sin2 θW (µ) ≡ g 2 (µ)/[g 2 (µ) + g 2 (µ)] where the ˆ ˆ ˆ couplings g and g are defined in the MS scheme and µ is chosen to be MZ for ˆ ˆ ˆ most electroweak processes. Bardin et al 1989. One conceptually simple scheme is the ‘on-shell’ one (Sirlin 1980.

42). electroweak radiative corrections in e+ e− → Z0 → f f¯ can be included by replacing the fermionic f f couplings gV and gA (see (22.172) In (22. this can make a relatively big difference. Clearly.3).173) is quite different.H ≈ 2 2 3 6 16π MW As the notation suggests.related to an interesting symmetry. appearing as O(|q 2 |/m 2 ). In such a situation f f (which is the usual one). As regards (22. we saw that the contribution of heavy fermions ‘(|q 2| m 2 )’ was suppressed.53) and (22. t. As explained there. (22. Beenakker and Hollik 1988).5.175) (22. ρ is then a radiative correction to ρ. but (22.5 but this time in weak boson propagators.174) contain surprising features. It turns out that.172).173)) and κ f ≈ 1 + 2 2 [sW /(1−sW )] ρ. and a contribution from the Higgs boson which is (for m H MW ) √ 2 2G F MW 11 m2 5 H − ln . The corrections have the form (in the on-shell scheme) ρ f ≈ 1 + ρ (of equation (22. there is an additional contribution ¯ coming from the presence of the virtual top quark in vertex corrections to Z → bb (Akhundov et al 1986. (22. ρ is a leading contribution to the parameter ρ introduced in (22.173) and (22.0664(2) (see section 11. of the kind discussed for ¯ [2] in γ section 11. ρ is given by (Veltman 1977) ρ= 3G F (m 2 − m 2 ) t √ b 8π 2 2 (22.38) for example.174) ( r )rem.176) √ (f) ρ f t3 together with corrections to the Z0 propagator. it is associated with top–bottom quark loops in vacuum polarization amplitudes. (22. for f = b. For the b-quark. to good approximation. This is why some precision measurements are surprisingly sensitive to the value of Copyright 2004 IOP Publishing Ltd . Hollik 1991) that the leading terms in r have the form r= r0 − 2 (1 − sW ) 2 sW ρ + ( r )rem . with a large value of m t . It turns out (Consoli et al 1989. the heavy particles are said to ‘decouple’. the fermion masses being in the numerator. referring to equation (11. But the correction (22. it measures the strength of neutral current processes relative to charged current ones.44). The running of α in r0 is expected. r0 = 1 − α/α(MZ ) is due to the running of α.173) while the ‘remainder’ ( r )rem contains a significant term proportional to ln(m t /m Z ). In the QED case. and has the value r0 = 0.173).54)) by gV = ¯ and f √ (f) 2 ρ f (t3 − 2Q f κ f sW ) gA = ¯ f (22.

let us go back to the Higgs gauge field ˆ Lagrangian ÄG of (22. Problem 22. where ρ = 1.110). The nearness of the measured ρ parameter to 1 is. in fact. the sensitivity is only logarithmic. r would then have been very sensitive to m H but. Consider.174). At tree level.173)).¯ Figure 22. To see the significance of this. But we now have to explain why ρ in (22. As we have seen. Since g f ∼ m f /v (see f (22.17. the contribution from the longitudinal polarization components of the W’s. f f An important message is that particles which acquire their mass spontaneously do not ‘decouple’.173) as follows. Let us first consider the situation at tree level. t–b vacuum polarization contribution.6 explores a simple model with a Higgs field in the triplet representation.106)) we arrive at an estimate ∼ m 2 /4πv 2 ∼ G F m 2 /4π as in (22.173).173) would vanish if m 2 = m 2 — t b and why only ln m 2 appears in (22. as regards the dependence on m H . these components are nothing but three of the four Higgs components which the W± and Z0 ‘swallowed’ to become massive.17. With the doublet Higgs of the form (22. Both these facts are related to a H symmetry of the assumed minimal Higgs sector which we have not yet discussed. it is a striking fact that the Higgs potential only involves the Copyright 2004 IOP Publishing Ltd . in particular. good support for the hypothesis that there are only doublet Higgs fields. we might well have expected it to involve m 2 in the H numerator if we considered the typical divergence of a scalar particle in a loop (we shall return to this after discussing (22. Second. The shift ρ is associated with vector boson vacuum polarization contributions. say)— or indeed by any number of doublets. But the couplings of these ‘swallowed’ Higgs fields to fermions are determined by just the same Higgs–fermion Yukawa couplings as we introduced to generate the fermion masses via spontaneous symmetry breaking. m t in the range near (as we now know) the physical value. It may be shown (Ross and Veltman 1975) that ρ = 1 is a natural consequence of having the symmetry broken by an SU(2)L doublet Higgs field (rather than a triplet. We can understand the appearance of the fermion masses (squared) in the numerator of (22. it is simplest to think of ρ in connection with the mass ratio (22.44). for example the one shown in figure 22.8) which produced the masses. in fact. Hence we expect the fermion loops to contribute (to these longitudinal W states) something of order g 2 /4π. where g f is the Yukawa coupling.

The condition g = 0 corresponds (cf (22. It is clear from (22.177) as does the vacuum condition (17.111) becomes ˆ φC = √ 2 (σ + iπ3 )/ √ ˆ ˆ ˆ −(π2 − iπ1 )/ 2 ˆ . we now write √ ˆ (π2 + iπ1 )/√ 2 ˆ ˆ (22. ˆ ˆ to be of any interest. 1) v (τ · W µ )(τ · W ) 2 4 0 1 = 1 2 ˆ ˆµ MW W µ · W 2 (22.178) φ= (σ − iπ3 )/ 2 ˆ ˆ ˆ while the φC field of (22.6) corresponds to ˆ ˆ σ = v. MW = MZ .181) that ˆ the three components W µ are treated on a precisely equal footing by the Higgs Copyright 2004 IOP Publishing Ltd .182) with MW = gv/2 as usual. and suppose to begin with that g = 0. this symmetry has to be present in the (Dµ φ)† (D µ φ) term as well.highly symmetrical combination of fields ˆ2 ˆ2 ˆ2 ˆ2 φ1 + φ2 + φ 3 + φ 4 (22.110).181). as expected if cos θW = 1.181) using τi τ j = δi j + i i j k τk .17)) ˆ ˆ to θW = 0. Now the vacuum choice (22. π = 0.180) ˆ Consider now the covariant SU(2)L × U(1) derivative acting on φ. (22.182) states that in the limit g → 0. This suggests that there may be some extra symmetry in (22.8). we will ˆ ˆ get just 1 g2 2 ˆ ˆµ (0. as in (22.179) We then find that these can be written as 1 ˆ φ = √ (σ + iτ · π ) ˆ ˆ 2 0 1 1 ˆ φC = √ (σ + iτ · π ) ˆ ˆ 2 1 0 . But of course. so that when we form (Dµ φ)† (D µ φ) from (22. and thus to W3µ = Z µ . Then 1 ˆ ˆ Dµ φ = √ (∂µ + igτ · W µ /2)(σ + iτ · π ) ˆ ˆ 2 g 1 ˆ ˆ ˆ ˆ = √ ∂µ σ + iτ · ∂µ π + i σ τ · W µ 2 2 g 0 ˆ ˆ ˆ − [π · W µ + iτ · W µ × π] ˆ 1 2 0 1 (22.102). (22. and so (22. The nature of this symmetry is best brought out by introducing a change of ˆ ˆ notation for Higgs doublet φ + and φ 0 : instead of (22.8) which is special to the doublet structure.

since they do not mix π and σ .178).14) and (22. there are additional terms in (22. In like term.185) which single out the ‘τ3 ’ component. i. t Consider a typical mass term.185) − sin2 θW Z µ (σ + iτ · π) 1 cos θW 2 ˆ We see from (22. from the term W µ σ 2 .15) to write W3µ and Bµ in ˆ ˆ terms of the physical fields Z µ and Aµ as in (19. of the form discussed in section 22. all three W fields have the same mass in this g → 0 limit. ˆ ˆ It is straightforward to calculate ( Dµ φ)† ( D µ φ) from (22. We can now answer this question.184) π →π + ×π ˆ ˆ ˆ  σ →σ ˆ ˆ ˆ2ˆ This is why. ρ = 1. However.183) This expression now reveals what the symmetry is: (22. yet involves the quark masses and the question of why 2 m 2 − bb appears in the numerator in (22. as in (22.184). and this relation is preserved under ‘rotations’ of the form (22. ˆ ˆ If we now reinstate g . there is a ‘τ · W ’ˆ ˆ 3 ’ part of it is now Z / cos θW .e. (22.173). and indeed the notation suggests that W µ and π should perhaps be ˆ regarded as some kind of new triplets. ˆ ˆ 4 4 (22.7) ˆ ˆ ( Dµ φ)† D µ φ = 1 g 1 µ (∂µ σ )2 + (∂µ π )2 − ∂µ σ π · W ˆ ˆ ˆˆ ˆ 2 2 2 g g µ ˆµ ˆ ˆ ˆ + σ ∂µ π · W + ∂µ π · (π × W ) ˆ ˆ 2 2 g2 ˆ 2 2 g2 2 ˆ ˆ + W µ (σ + π 2 ) + π 2 W µ .ˆ field (22. and use (22.181) becomes ˆ τ1 ˆ τ2 ˆ τ3 Z µ 1 1 + τ3 ˆ √ ∂µ + ig W1µ + ig W2µ + ig + ig sin θW Aµ 2 2 2 cos θW 2 2 ig 1 + τ3 ˆ 0 ˆ ˆ .181). except that the ‘W the vacuum σ = v.7. First. and of the relations (22.97). but we expect that they will cause O(g 2 ) corrections to ρ = 1 at the one-loop level. (22. however. These terms vanish as g → 0 and do not contribute at tree level.185) that g = 0 has two effects.15) which embody the requirement of a massless photon.183) is invariant under ˆ ˆ global SU(2) transformations under which W µ and π are vectors—that is (cf (3. Hence this mass relation (and ˆ ˆ ρ = 1) is a consequence of the global SU(2) symmetry of the interactions and the vacuum under (22. and therefore break this global SU(2). for a quark Copyright 2004 IOP Publishing Ltd .184). π = 0 which simply means that the mass of the Z is ˆ ˆ MZ = MW / cos θW .9))  ˆ ˆ ˆ Wµ → Wµ + × Wµ  .181): one finds (problem 22.1.14) and (22. None of this. (22.

190) . we also make ˆ the transformation of π given in (22.189) From (22. This is why a t–b loop in a W vacuum polarization correction can produce the ‘non-decoupled’ contribution (22. (22. The resulting complete transformation.doublet of the i th family ¯ ¯ ˆ ˆ ¯ ¯ ˆˆ ˆ Äm = −g+ (u Li d Li )φC u Ri − g−(u Li d Li )φ dRi . it is precisely an ordinary ‘isospin’ transformation of the type ui ˆ ˆ di Copyright 2004 IOP Publishing Ltd → (1 − i · τ /2) ui ˆ ˆ di .178) and (22. the transformation on the L components is just the same as a standard SU(2)L transformation.188) but only under transformations with 1 = 2 = 0. ˆ dRi (22. however.188).173) to ρ. which was responsible for ρ = 1 at the tree level. (22. which grows as m 2 − m 2 t b and produces quite detectable shifts from the tree-level predictions.188).188) if we wish.184).189) we see that if. they are totally decoupled in the gauge dynamics. Returning to (22. given the accuracy of the data.187). 3 = 0. will not be invariant under (22. As far as the R components are concerned. the first term of (22.187) becomes (to first order in ) − (g+ + g− ) ¯ ¯ ˆ ˆ ˆ √ (u Li d Li )[σ + iτ · (π + π × )] ˆ ˆ 2 2 u Ri ˆ ˆ dRi .186) are proportional to the masses of the quark doublet—this symmetry is explicitly broken by the quark ¯ mass difference. dLi )T and (u Ri .179). can be extended also to the quark sector. dRi )T : ˆ ˆ u Li ˆ ˆ dLi → (1 − i · τ /2) u Li ˆ ˆ dLi u Ri ˆ ˆ dRi → (1 − i · τ /2) u Ri ˆ . which does the same to both the L and R components. and we are free to make the transformation (22.188) Under (22. is a nonchiral one—in fact. but—because the g± in (22. so the gauge interactions of the quarks obey this symmetry also. (22. ˆ ˆ ˆ ˆ Using (22. We conclude that the global SU(2) symmetry of (22. then this first term in Äm will be invariant under these combined transformations. The second term in (22. this can be written as ˆ Äm = −g+ ¯ ¯ g− ¯ ¯ 0 u Ri ˆ √ (u Li d Li )(σ + iτ · π ) ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ − √ (u Li d Li )(σ + iτ · π ) ˆ 0 dRi 2 2 (g+ + g− ) ¯ ¯ u Ri ˆ = − ˆ ˆ √ (u Li d Li )(σ + iτ · π ) ˆ ˆ ˆ dRi 2 2 (g+ − g− ) ¯ ¯ u Ri ˆ .186) Consider now a simultaneous (infinitesimal) global SU(2) transformation on ˆ ˆ the two doublets (u Li .184). at the same time as (22.188).187) − ˆ ˆ (u Li d Li )(σ + iτ · π )τ3 ˆ ˆ √ ˆ dRi 2 2 (22. except that it is global.

1979) that one-loop radiative corrections to electroweak observables grow at most like ln m 2 for large m H . of course. that the sensitivity of the data to this important parameter of the Standard Model is only logarithmic. Chanowitz et al 1978.1. Even without a Higgs contribution however. ˆ ˆ Copyright 2004 IOP Publishing Ltd .174) and it means. and hence one might expect to see some radiative correction becoming very large ( O(m 2 )) H as one tried to ‘banish’ H from theory by sending m H → ∞ (m H would be acting ˆ ˆ like a cut-off).Z the O(m 2 ) pieces are approximately O(m 2 ) for m 2 H H H H cancel out from all observable quantities.18. effect seen at the close of play at LEP gave m H ∼ 115 GeV (LEP Higgs (2001)). This is just H what we have in (22. Without the Higgs particle H the theory is non-renormalizable.10. as usual. This loop contributes to the Higgs self-energy. the expected O(m 4 ) behaviour becomes O(m 2 ) H H instead (van der Bij and Veltman 1984. For further details. the simplest loop we meet is that shown in figure 22. one suspects that the large m 2 effects will not H be so dramatic after all. that it diverges as the square of the cut-off. In fact. it turns out that the electroweak theory is renormalizable at the one-loop level if the fermion masses are zero (Veltman 1968.3 leaving only ln m 2 terms. appears (relative to the one-loop contributions) with an additional factor of 3 Apart from the φ † φ coefficient! See section 22. The reason is that in such a (φ † φ)2 theory.d. for m 2 MW. see the review of searches for Higgs bosons by Igo-Kemenes in Hagiwara et al (2002). calculation shows (Veltman 1977. 1970). At the two-loop level. We will return to this particular detail in section 22. The reader will recognize that the mathematics here is exactly the same as that in section 18.3 involving the SU(2) of isospin in the σ -model. Fits to data typically give m H in the region of 100 GeV at the minimum of the χ 2 curve but the error (which is not simple to interpret) is of the order of 50 GeV.Z .8) from experiment. Direct searches now rule out a Higgs mass less than about 110 GeV . Thus. and will be renormalized by taking the value of the coefficient ˆ ˆ of φ † φ in (22. unfortunately.10. While there are finite corrections which H 2 2 MW.18 and it is easy to see by counting powers. while the ∼ 2 s. One-boson self-energy graph in (φ † φ)2 . The isospin SU(2) is frequently called ‘custodial SU(2)’ since it ‘protects’ ρ = 1.ˆ ˆ Figure 22. van der Bij 1984)—and. What about the absence of m 2 corrections? Here the position is rather more H subtle. This analysis of the symmetry of the Higgs (or a more general symmetry breaking sector) was first given by Sikivie et al (1980).

In the late 1980s and early 1990s. Abachi et al 1995b) at the CDF and D0 detectors at FNAL’s Tevatron. It was shown by Einhorn and Wudka (1989) that this screening is also a consequence of the (approximate) isospin SU(2) symmetry we have just discussed in connection with (22. b) announced 12 events consistent with the hypothesis of production of a t¯ pair and.191) the central figure and first error being based on m H = 300 GeV. on this t hypothesis. and for some years subsequently. was discovered by Veltman (1977) and called a ‘screening’ phenomenon by him: for large m H (which also means.8 TeV. the mass was found to be m t = 174 ± 10 ±13 GeV 12 Copyright 2004 IOP Publishing Ltd (22.9 The top quark Having drawn attention to the relative sensitivity of radiative connections to loops containing virtual top quarks. between experimentalists searching for the real top and theorists fitting ever more precise data to get tighter and tighter limits on m t from its virtual effects. 22. as we have seen. 1995b. Phenomenologically. b. Langacker and Erler gave the result of a fit to all electroweak data as m t = 169 ±16 ±17 GeV. the data were not precise enough to be sensitive to virtual t effects. the second (+) error assuming m H = 1000 GeV and the second (−) error assuming m H = 60 GeV. LEP at CERN and SLC at Stanford began to produce new and highly accurate data which did allow increasingly precise predictions to be made for the top quark mass. in p–¯ collisions at E CM = 1. p The W and Z particles were. 18 20 (22. At about the same time.193) . it is worth devoting a little space to a ‘backward glance’ at the year immediately prior to the discovery of the t-quark (Abe et al 1994a.192) without any assumption for m H . In fact. the experimental error in m t was just about the same as the theoretical one (and—of course—the central values were consistent). a kind of race began. in their May 1994 review of the electroweak theory (contained in Moutanet et al 1994. as we have seen. in the limit of large m H . A month or so earlier. the situation was very different. the upshot is that it is unfortunately very difficult to get a good handle on the value of m H from fits to the precision data. the CDF collaboration (Abe et al 1994a.O(α). discovered in 1983 and at that time. Thus. This relative insensitivity of the radiative corrections to m H . With the top quark. by the time of the actual experimental discovery of the top quark. large λ).173). Thus. m t . Ellis et al (1994) gave the extraordinarily precise value m t = 162 ± 9 GeV (22. p 1304ff). we have an effectively strongly interacting theory whose principal effects are screened off from observables at lower energy.

1 GeV (Hagiwara et al 2002) as compared to the value predicted by fits to the electroweak data of 175. Unlike the mass of the other quarks. Now we 2 have seen that the virtual process is supressed by ∼ 1/MW if the energy release (as in the case of b-decay) is well below MW . bound by QCD. while the electroweak interactions are undoubtedly well described by the SU(2)L × U(1) theory. this fast decay of the t quark means that there will be no detectable t¯ ‘toponium’. the hadronization is associated with the b quark. however.10. spontaneously broken. the b quark itself decays by the usual virtual W processes. 22. All the particles in this hypothetical world are. they can only travel some 10−16 m before decaying. In the real world. non-Abelian chiral gauge theory. which has a more typical weak lifetime (∼ 1.10 The Higgs sector 22. (22. m t is greater than MW and this means that it can decay to b + W via real W emission: t → W+ + b. the strong interactions which should eventually ‘hadronize’ them t will not play a role until they are ∼ 1 fm apart. The 2002 experimental value for m t is 174. massless.5×10−12 s). But if they are travelling close to the speed of light. all the interactions being determined by just two Lagrangian parameters g and g in a ‘universal’ way.1 Introduction The Lagrangian for an unbroken SU(2)L × U(1) gauge theory of vector bosons and fermions is rather simple and elegant. made use of the subtle intricacies of a renormalizable. In fact (problem 22.3 ± 5. Instead. Thus t’s tend to decay before they experience the confining QCD interactions. Abachi et al 1995b). One feature of the ‘real’ top events is particularly noteworthy.3 ± 44 GeV. and its discovery is of the utmost importance. Consider now the production of a t¯ pair in the collision between two partons. t With the t quark now safely real. the one remaining missing particle in the Standard Model is the Higgs boson.8) the top quark lifetime from (22. But the real process (22. It is surely remarkable how the quantum fluctuations of a yet-to-be detected new particle could pin down its mass so precisely.194) In contrast. It is fitting that we should end this final chapter with a brief review of Higgs physics. By February 1995 both groups had amassed more data and the discovery was announced (Abe et al 1995b. It seems hard to deny that Nature has. neither the mediating gauge quanta (apart from the photon) nor the Copyright 2004 IOP Publishing Ltd . As the t t and ¯ separate.194) suffers no such suppression and proceeds very much faster. indeed.194) is estimated to be ∼ 4 × 10−25 s! This is quite similar to the lifetime of the W+ itself. This represents an extraordinary triumph for both theory and experiment.and this was followed by nine similar events from D0 (Abachi et al 1995a). By the same token. via W+ → e+ νe for example.

In that case ‘electron pairing’ was a crucial ingredient. The whole phenomenology depends on the renormalized coefficient having a negative value. They must acquire mass in some way that does not break the gauge symmetry of the Lagrangian. Such considerations may indicate that the ‘Higgs Sector’ of the Standard Model is on a somewhat different footing from the rest of it—a commonly held view. This means that the O( 2 ) one-loop ˆ ˆ correction must be cancelled by the ‘bare’ mass term 1 m 2 φ † φ so as to achieve 2 H. In chapter 19 we discussed how such a breaking of a gauge symmetry does happen. none has yet emerged as both theoretically compelling and phenomenologically viable. following the original ideas of Higgs and others (Higgs 1964. Guralnik et al 1964. rather than (as in BCS theory) obtaining the fermion mass gaps dynamically. Higgs 1966) that a suitable scalar (‘Higgs’) field exists. In section 22. of course. the Yukawalike fermion couplings are not gauge interactions either. Perhaps it should be regarded as more a ‘phenomenology’ than a ‘theory’. as is commonly assumed (say 1016 GeV). much as the current–current model was. It has to be admitted that this part of the Standard Model appears to be the least satisfactory. Englert and Brout 1964. triggering the spontaneous breaking of the symmetry. we may mention a point which has long worried many theorists. This cancellation between m 2 and 2 will a negative coefficient of order −v H. all these are renormalizable couplings—but this basically means that their values are not calculable and have all to be taken from experiment. This term would ordinarily. In this connection. indicates that some additional fields must be present in order to give mass to the originally massless gauge bosons. in a superconductor.18 gives a quadratically divergent ( O( 2 )) and positive ˆ ˆ contribution to the φ † φ term in the Lagrangian.8 we noted that figure 22. indeed. dynamically.fermions are massless. While the coupling of the Higgs field to the gauge fields is determined by the gauge symmetry. Furthermore. In particle physics. the matter is much more delicate. True. while a lot of effort has gone into examining various analogous ‘dynamical symmetry breaking’ theories. or else the renormalizability of the theory is destroyed and its remarkable empirical success (at a level which includes loop corrections) would be some kind of freak accident. and they are both unconstrained and uncomfortably different in orders of magnitude. But in the Higgs case. it is simply assumed. a simple count of the number of degrees of freedom in a massive vector field. they too are put in ‘by hand’ via Yukawa-like couplings to the Higgs field. as opposed to a massless one.0 2 . in the Standard Model. And so. at one-loop order. the Higgs self-coupling is not a gauge interaction and is unrelated to anything else in the theory.0 have to be very precise indeed if —the scale of ‘new physics’—is very high. However. Likewise. with a potential which breaks the symmetry spontaneously. The reader may wonder why attention should now be drawn to this particular piece of renormalization: aren’t all divergences handled this way? In a sense Copyright 2004 IOP Publishing Ltd . be just the mass term of the scalar field.

and benefit from the protection enjoyed by the fermions. Whatever may lie ‘beyond’ it. A second possibility is supersymmetry (Peskin 1997). chapter 11) and the review by Igo-Kemenes in Hagiwara et al (2002)). rather like the BCS-pairing idea referred to earlier. Similarly. gauge invariance for the vector bosons prohibits any O( 2 ) connections in perturbation theory.they are. in which there is a ‘protective’ symmetry operating. upon which we remarked in section 11. The elucidation of the mechanism of gauge symmetry breaking is undoubtedly of the greatest importance to particle physics: quite apart from the SU(2)L × U(1) theory. we Copyright 2004 IOP Publishing Ltd . if we take the apparently reasonable point of view that the Standard Model will ultimately fail at some scale where new physics enters. Thus the difficulty is: why is the empirical parameter v ‘shielded’ from the presumed high scale of new physics? This ‘problem’ is often referred to as the ‘hierarchy problem’. The discovery of this Higgs boson has. always been a vital goal in particle physics. Various possibilities have been suggested for the kind of physics that might enter at energies of a few TeV. There is a good reason for this in the case of the electron mass. Before turning to experiment. a few TeV. which ‘protects’ the coefficient of φ † φ in this way. the problem would go away if the scale were as low as. which one would really like to understand far better. √ physical scalar field necessarily remains. But there is no symmetry.3 (for a full account see Dawson et al (1990) and for more compact ones see Ellis et al (1996. however. the same scale at which the Standard Model ceases to be a perturbatively calculable theory. Of course. A third possibility is that of large extra dimensions (Antoniadis 2002). to which we must now return. These undoubtedly fascinating ideas obviously take us well beyond our proper subject. The most significant prediction of the Higgs mechanism in SU(2)L ×U(1)— and one originally pointed out by Peter Higgs himself (1964)—is that even after the gauge bosons have swallowed three of the scalar fields to √ acquire mass. say.2. for which there is nothing like such a dramatic ‘fine-tuning’ problem. therefore.10. as we shall briefly discuss in section 22. It is hard to understand what can be stopping it from being of order 2 . As we shall see in the next section this happens to be. The other mass corrections have all been logarithmic. the Lagrangian of the Higgs sector of the Standard Model leads to many perfectly definite predictions which may be confronted with experiment. not accidentally. within the ˆ ˆ Standard Model. Chiral symmetry forces self-energy corrections for fermions to be proportional to their mass and. but the fact is that this is the first case we have had in which we have to cancel a quadratic divergence. We stress again that we are dealing here with an absolutely crucial symmetry-breaking term. hence. with mass m H = 2µ = one √ λv/ 2. very many of the proposed theories which go ‘beyond the Standard Model’ face a similar ‘mass problem’ and generally appeal to some variant of the ‘Higgs mechanism’ to deal with it. to contain only logarithms of the cut-off. For example. ‘technicolour’ models (Peskin 1997) regard the Higgs field as a composite of some new heavy fermions. since scalar fields can be put alongside fermions in supermultiplets.

For the ˆ ˆ (φ † φ)2 interaction of (22. as m H increases.196) Note that this is exponentially sensitive to the ‘low-energy’ coupling constant λ(v)—and that E ∗ decreases rapidly as λ(v) increases. independent. this gives m H < 160 GeV.197) For ∼ 1016 GeV. is unfortunately undetermined in the Standard Model. which determines m H given the known value of v . This is an oversimplified argument for various reasons. The precise critical values are sensitive to the value of m t . though the essential point is correct. The conclusion is that for 130 GeV < m H < 190 GeV the perturbative regime could extend to ∼1016 GeV. As we saw. An important omission is the contribution of the top quark to the running of λ( E).want to mention some theoretical considerations concerning m H by way of orientation.195) Like QED. 3λ(v) (22.18)) m H is essentially proportional to λ1/2 (v). But (see (22. some quite strong theoretical arguments suggest that m H cannot be arbitrarily large. in a non-gauge theory with massive vector bosons. We have previously considered violations of unitarity by the lowestorder diagrams for certain processes (see chapter 21 and section 22. However. In fact. Like all coupling constants in a renormalizable theory.8). but that for m H < 130 GeV the theory would be non-perturbative at a much lower scale. (22.6). Then we require 4π 2 mH < v 3 ln( /v) 1/2 . m H < 750 GeV.2 Theoretical considerations concerning mH The coupling constant λ. a one-loop calculation of the β -function leads to λ( E) = λ(v) 1− 3λ(v) 8π 2 ln( E/v) . More refined versions place both approximate upper. λ must ‘run’. argument which suggests that m H cannot be too large. 22. which carry Copyright 2004 IOP Publishing Ltd . such violations are associated with the longitudinal polarization states of the bosons. (22. non-perturbative behaviour sets in increasingly early. Hambye and Riesselmann 1997). the theory becomes non-perturbative at the scale E ∗ such that E ∗ ∼ v exp 8π 2 . bounds on m H (Cabibbo et al 1979. Hence.10. this theory is not asymptotically free: the coupling increases with the scale E. There is another. and lower. then the bound on m H is weaker. Isidori et al 2001. However. if the non-perturbative regime sets in at 1 TeV. Suppose we say that we should like perturbative behaviour to be maintained up to an energy scale .

in fact. strong cancellations in the high-energy behaviour occur between different lowestorder diagrams. unexpected result can be understood as follows. together with the W–W self-interaction.199)). Rather. chapter 8)). this puts a bound on m H . ˆ The longitudinal components of the W’s arise from the ‘∂ µ φ’ parts in (22. as predicted by the Standard Model. √ (22.factors proportional to the 4–momentum k µ (see (21. This behaviour is characteristic of gauge theories (Llewellyn Smith 1973. Thus. however. Cornwall et al 1974) and is related to their renormalizability. curiously reminiscent of the original situation with the four-fermion current– current interaction itself (compare (21. at first sight. for which the Feynman amplitude is just proportional to λ. it can do so if λ itself is too big—and since λ ∝ m 2 .30)). One process of this sort which we have yet to consider.10) with (22. Ellis et al (1996. the Higgs coupling constant (see. Copyright 2004 IOP Publishing Ltd .9)) and so. in order of magnitude.8) (compare equation (19. Z and Higgs particles. Now. it states that if m H gets bigger than a certain value. b) gives 1/2 mH < ≈ 1 TeV. or else some electroweak interactions will become effectively strong with new physical consequences. (22. the scattering of longitudinal W’s is effectively the scattering of the three Goldstone bosons in the complex Higgs doublet. In terms of standard quantities. We now consider some simple aspects of Higgs production and decay processes at collider energies. In a gauge theory. A considerable number of diagrams (seven in all) contribute to this process. (22. we expect unitarity to imply λ < 1. It is. is that in which two longitudinally polarized W’s scatter from each other. This. These bosons have self-interactions arising ˆ ˆ from the λ(φ † φ)2 Higgs potential. perturbation theory will fail or ‘new physics’ will enter. this line of reasoning seems to imply that the Higgs boson will either be found at a mass well below 1 TeV. for example. When all these are added up the high-energy behaviour of the total amplitude turns out to be proportional to λ. This ‘no lose’ situation provided powerful motivation for the construction of the LHC. At all events. A constant H amplitude is pure J = 0 (compare (21. although such a constant term obviously cannot violate unitarity as the energy increases (as has happened in the other cases).198) λ = m2 GF/ 2 H and so we expect mH < GF √ 8 2π 3G F −1/2 .200) Like the preceding argument.48) in the U(1) case). which produce k µ factors.199) A more refined analysis (Lee et al 1977a. this one does not say that m H must be less than some fixed number. Perhaps this is a clue that we need to replace the Higgs phenomenology. in leading order: exchange of γ .

the cross-section is about 50 pb for m H 100 GeV and 1 pb for m H 700 GeV. detection of H depends Copyright 2004 IOP Publishing Ltd . Higgs boson production process by ‘gluon fusion’. which are probed at larger √ collider energy s. the cross-section for this process will rise with energy.20.20. Stange et al 1994a. This already excludes many possibilities in both production and decay.4 GeV (95% c. Higgs boson production process by ‘vector boson fusion’.l.2 pb for m H 200 GeV. the production cross-section ranges from about 100 GeV to 0. 22. Figure 22. roughly ten times smaller. p The next largest cross-section. At √ the Tevatron with s = 2 TeV.201) m H ≥ 114.19. At an LHC energy of 1 pb for m H √ s = 14 TeV. b).). However.19 (Georgi et al 1978. is for ‘vector boson fusion’ via the diagram of figure 22. At both the Tevatron and the LHC.3 Higgs phenomenology Our discussion is based on the existing lower bound on m H established at LEP (LEP 2003): (22. Since the gluon probability distribution rises rapidly at small x values. Glashow et al 1978. The cross-section is the same for pp and for p¯ colliders. the dominant production mechanism is expected to be ‘gluon fusion’ via an intermediate top quark loop as shown in figure 22.Figure 22.10.

23 (taken from Ellis et al 1996) shows the cross-sections for the various production processes as a function of m H . A fourth possibility is ‘associated production with top quarks’ as shown in figure 22. t crucially on being able to separate the signal from the large backgrounds expected in many of the Higgs decay channels (to be discussed shortly). For such ¯ m H values. (H → f f¯) is less than 10 MeV.21. of which bb has ¯ is easily calculated to lowest the largest branching ratio.22. Higgs boson production in association with a t¯ pair.pT leptons from ¯ the decay b → cl − ν . decays to fermion–anti-fermion pairs dominate. A p¯ collider gives a somewhat larger p cross-section for this process than a pp collider. The situation changes significantly when m H becomes greater than twice the Copyright 2004 IOP Publishing Ltd .202) where the colour factor C is three for quarks and one for leptons. for example.Figure 22. The final state ZH → µ+ µ− bb provides a clean signature through tagging b-jets using the high. Figure 22. the mechanism of associated production of a Higgs boson with a vector boson. Higgs boson production in association with W or Z.9) (H → f f¯) = C GFm2 mH f √ 4π 2 1− 4m 2 f m2 H 3/2 (22. have to be detected via its decays. shown in figure 22. The width of H → f f order and is (problem 22. of course.22. Figure 22. could also be important since the leptonic decays of the W or Z can be exploited for triggering. For this reason.21. The Higgs boson will. For ¯ m H < 140 GeV.

203) and the width for H → ZZ is the same with MW → MZ and a factor of 1 to 2 allow for the two identical bosons in the final state.Figure 22. The tree-level width for H → W+ W− is (problem 22.24. p 399). Higgs boson production cross-sections in pp collisions at the LHC (figure from Ellis et al 1996. which is just another facet of the ‘strong interaction’ regime discussed earlier. Higgs boson decay via quark triangle. Copyright 2004 IOP Publishing Ltd . These widths rise rapidly with m H .9) (H → W+ W− ) = G Fm3 √H 8π 2 1− 2 4MW m2 H 1− 2 4MW m2 H + 12 4 MW m4 H (22. vector boson masses. reaching ∼ 1 GeV when m H ∼ 200 GeV and ∼ 100 GeV for m H ∼ 500 GeV. Figure 22.23. It is apparent that for m H any larger than this (say m H ∼ 1 TeV) the width of the state will become comparable to its mass.

127). The vector bosons themselves decay to leptons and quark jets.2 Using the vertex (22. 22. derive (22.133) is diagonal in the ‘mass’ basis. Figure 22. The rarer purely leptonic final states such as l +l − l + l − are likely to prove the best hope for discovery. Problems 22.5 Verify (22. Copyright 2004 IOP Publishing Ltd . 22. The ATLAS and CMS detectors have been optimized for Higgs boson searches and should be well able to discover a Higgs boson with m H in the range 100 GeV to 1 TeV.1 By identifying the part of (22.150).36). 22. The LHC is scheduled to start physics runs in the year 2007. will allow this crucial energy regime to be explored with high precision. p 393).4 Verify that the neutral current part of (22. Another rare but characteristic decay is H → γ γ via intermediate W and quark triangle loops (figure 22.24).26).56).7) into (22.3 Insert (22. verify (22.125) to derive (22.Figure 22. shows the complete set of phenomenologically relevant Higgs branching ratios for a ‘light’ Higgs boson.35). Future machines. taken from Ellis et al (1996). But final states containing hadronic jets will have to contend with large backgrounds at hadron collides. such as a high-energy e+ e− or µ+ µ− collider.23) which has the form (22. Branching ratios of the Higgs boson (from Ellis et al 1996.25. with the latter having the larger branching ratios.25. 22.

Copyright 2004 IOP Publishing Ltd .9 Using the Higgs couplings given in appendix Q.7 Use (22. t2 ] = it3 ).203). with the same sin θW as in (22. Show that the photon and Z fields are still given by (22.17). 22. The non-vanishing component has t3 = −1. but that now √ MZ = 2MW / cos θW .6 Suppose that the Higgs field is a triplet of SU(2)L rather than a doublet.15).14) and (22. Since we want the charge of the vacuum ˆ to be zero and we have Q = t3 + y/2.8 Calculate the lifetime of the top quark to decay via t → W+ + b. What is the value of the parameter ρ in this model? 22.183). So the covariant ˆ is derivative on φ ˆ ˆ ˆ (∂µ + ig t · W µ + ig Bµ )φ where   t1 =  0 1 √ 2 1 √ 2 0 1 √ 2      t2 =  0 i √ 2 −i √ 2 0 −i √ 2    0 1 √ 2 0 i √ 2 0 0 0 0 and t3 is as before (it is easy to check that these three matrices do satisfy the required SU(2) commutation relations [t1 . we must assign y(φ) = 2. using   1 0 0 0  t3 =  0 0 0 0 −1 in the ‘angular-momentum-like’ basis. and suppose that its vacuum value is   1 0 0 ˆ 0|φ|0 =  0 0 0  0 0 −1 in the gauge in which it is real. 22. verify (22.22.181) to verify (22.202) and (22.

the group is Abelian. ‘e’ being the first (unit) matrix and the inverse being the usual matrix inverse.1) Note that. the product ab ∈ (the symbol ‘∈’ means ‘belongs to’ or ‘is a member of’). The reader may check that each of (i)–(iv) is satisfied.APPENDIX M GROUP THEORY M. b so as to form their ordered ‘product’ ab. with the law of combination being ordinary multiplication. b. If the set does not have a finite number of elements it is an infinite group. i. such that aa −1 = a −1 a = e. it happens to be so for these Copyright 2004 IOP Publishing Ltd . (iv) For all a ∈ . ae = ea = a. . such that the following four conditions hold: (i) For every a. −1. . iii) contains a unique identity element. ab = ba: if it is commutative (ab = ba). i.2) (M. c. the order of the group being equal to the number of elements in the set. the set of four numbers (1. e. As a simple example. b ∈ . . if not. a −1 . −i ) form a finite Abelian group of order 4.4) with the combination law being matrix multiplication. in general. such that for all a ∈ . (ii) The law of combination is associative. the law of combination is not commutative. (ab)c = a(bc). it is non-Abelian.e. with e taken to be the number 1 and the inverse being the algebraic reciprocal.1 Definition and simple examples A group is a set of elements (a. Any finite set of elements satisfying the conditions (i)–(iv) forms a finite group.e.) with a law for combining any two elements a.3) (M. A second group of order 4 is provided by the matrices 1 0 0 1 0 −1 1 0 −1 0 0 −1 0 −1 1 0 (M. i. Although matrix multiplication is not commutative in general. there is a unique inverse element. (M.

c ). . −i) (in that order) do. α2 . i. In particular. groups whose elements are labelled by a number of continuously variable real parameters α1 . α2 . . we can ‘move out’ from that neighbourhood to other nearby elements. M. The analyticity condition implies that if we are given the form of the group elements in the neighbourhood of any one element. the way these four matrices multiply together is (as the reader can verify) exactly the same as the way the four numbers (1. For a continuous group. −1. if this is the case. a. c ↔ c . . . The matrices of SU(3) are specified by the values of eight real parameters. Further. the correspondence between the elements of the two groups is ‘one to one’: that is. . α1 . we are concerned with various kinds of ‘coordinate transformations’ (not necessarily spacetime ones but including also ‘internal’ transformations such as those of SU(3)). Two groups with the same multiplication structure and with a oneto-one correspondence between their elements are said to be isomorphic.6) Copyright 2004 IOP Publishing Ltd . rotations in three dimensions form a group whose elements are specified by three real parameters (e. . M. By convention. if we label the two sets of group elements by (e. The simplest group element to consider is the identity. a . three for pure velocity transformations). . . b ↔ b . using the mathematical procedure known as ‘analytic continuation’ (essentially. which we shall now denote by I . c) and (e . .3 Generators of Lie groups Consider (following Lichtenberg 1970. . the group is a Lie group. Lorentz transformations also form a group. they are homomorphic. For example. In fact. by repeating the process. A more restrictive condition is that γ should be an analytic function of α and β. condition (i) takes the form g(α)g(β) = g(γ (α. Lie proved that the properties of the elements of a Lie group which can be reached continuously from the identity I are determined from elements lying in the neighbourhood of I . . parametrizations are arranged in such a way that g(0) is the identity element of the group.2 Lie groups We are interested in continuous groups—that is. αr ) (M. chapter 5) a group of transformations defined by x i = f i (x 1 . α2 .particular matrices. If they have the same multiplication structure but the correspondence is not one-to-one.5) where the parameters γ are continuous functions of the parameters α and β. b . . . β)) (M. αr ) ≡ g(α). x 2 . we should be able to reach all group elements which are ‘continuously connected’ to the original element.g. αr : g(α1 . . x N . using a power series expansion). b. this time with six real parameters (three for 3D rotations. two for defining the axis of the rotation and one for the angle of rotation about that axis). . we have the correspondences e ↔ e . . a ↔ a .

1). X µ ] = cλµ X ν (M. . By convention. 0). Finite transformations are obtained by ‘exponentiating’ the quantity in braces in (M. so there are as many generators as there are parameters labelling the group elements. We have N F → F + dF = F + i=1 N ∂F dxi ∂ xi r ν=1 =F+ i=1 r ∂ fi ∂F dαν ∂αν ∂ xi (M. the i) in the definition of X .8). . . The commutation relations (M.10) is a generator of infinitesimal transformations. 2.1 Note that in (M. N ) are the ‘coordinates’ on which the transformations act. for example those of SO(3) (see section M. and the α ’s are the (real) parameters of the transformations.11) ˆ where we have written = α · X. 0). so x i = f i ( x.7) dxi = ν=1 ∂ fi dαν ∂αν (M.where the x i ’s (i = 1.30)): ˆ ˆ U (α) = exp {−iα · X} r ˆ ν=1 αν X ν (M.4. 1 Clearly there is lot of ‘convention’ (the sign.12) ν where the constants cλµ are complex numbers called the structure constants of the group. An important theorem states that the commutator of any two generators of a Lie group is a linear combination of the generators: ν ˆ ˆ ˆ [ X λ .10) ν runs from 1 to r .9) ≡ 1− ν=1 ˆ dαν i X ν F where ˆ Xν ≡ i N i=1 ∂ fi ∂ ∂αν ∂ x i (M. Copyright 2004 IOP Publishing Ltd .9) (compare (12. A transformation in the neighbourhood of the identity is then given by r (M. α = 0 is the identity transformation.8) where the {dαν } are infinitesimal parameters and the partial derivative is understood to be evaluated at the point ( x. It is chosen for convenient ˆν consistency with familiar generators. Consider now the change in a function F( x) under the infinitesimal transformation (M. .12) are called the algebra of the group. a sum over ν from 1 to r is understood on the right-hand side.

10) are then ∂ ∂ ˆ − ix 2 X 1 = ix 3 ∂ x2 ∂ x3 ∂ ∂ ˆ − ix 3 X 2 = ix 1 ∂ x3 ∂ x1      ∂ ∂   ˆ 3 = ix 2  − ix 1 X ∂ x1 ∂ x2 Copyright 2004 IOP Publishing Ltd (M. we have etc.18) dx 3 = − 1 x 2 + 2 x 1 . Thus in (M. i. x T x = x T x. which is not continuously connected to the identity.8).19)).4.e.13) where R is a real 3 × 3 matrix such that the length of x is preserved.20) ∂ f1 = −x 3 ∂α2 ∂ f1 = x2 ∂α3          The generators (M.64)) or dx 1 = − 2 x 3 + 3 x 2 ∂ f1 =0 ∂α1 dx 2 = − 3 x 1 + 1 x 3 1. Expanding this out to first order in δ R gives δ R T = −δ R (M.1 SO(3) and three-dimensional rotations Rotations in three dimensions are defined by x = Rx (M. Those with det R = 1 are ‘proper rotations’ and they form the elements of the group SO(3): the S pecial O rthogonal group in 3 dimensions. We may parametrize δ R as   − 2 0 3 δR =  − 3 0 (M. An R close to the identity matrix I can be written as R = I + δ R where (I + δ R)T (I + δ R) = I. This implies that R T R = I .19) dα3 ≡ 3.15) (M. (M.4 Examples M. It follows that (M.16) so that δ R is an anti-symmetric 3 × 3 matrix (compare (12.M.17) 1  − 1 0 2 and an infinitesimal rotation is then given by x =x− ×x (compare (12.14) 1 = det(R T R) = det R T det R = (det R)2 and so det R = ±1. (M. Those R’s with det R = −1 include a parity transformation (x = −x). so that R is an orthogonal matrix. (M. identifying dα1 ≡ dα2 ≡ 2.21) .

23) The action of finite rotations. parametrized by α = (α1 .24) ˆ The operators U (α) form a group which is isomorphic to SO(3).27)) q1 q2 so that dq1 = i 1 3 2 q1 + + q2 2 2 2 −i 3 i 1 2 q2 + − q1 . (M.30) (M.23).4.26) Then (with dα1 ≡ 1 etc) ∂ f1 iq1 = ∂α3 2 ∂ f2 iq2 =− ∂α3 2 (M. The two groups are Copyright 2004 IOP Publishing Ltd . α2 . = =− .2 SU(2) We write the infinitesimal SU(2) transformation (acting on a general complex two-component column vector) as (cf (12.31) It is an interesting exercise to check that the commutation relations of the ˆ ˆ X i ’s are exactly the same as those of the X i ’s in (M. The structure constants of SO(3) are i i j k . ∂α1 2 ∂α2 2 and (from (M. α3 ). on functions F is given by ˆ ˆ U (α) = exp{−iα · X}. ∂α1 2 ∂α2 2 ∂ f2 iq1 ∂ f 2 q1 .10)) 1 ∂ ∂ ˆ q2 X1 = − + q1 2 ∂q1 ∂q2 i ∂ ∂ ˆ q2 − q1 X2 = 2 ∂q1 ∂q2 1 ∂ ∂ ˆ −q1 . M.27) (M.25) (M.23). from (M. (M.29) (M.which are easily recognized as the quantum-mechanical angular momentum operators ˆ X = x × −i∇ (M.22) which satisfy the SO(3) algebra ˆ ˆ [Xi . + q2 X3 = 2 ∂q1 ∂q2 (M.28) iq2 ∂ f1 q2 ∂ f1 . = = . X j ] = i ˆ ijk Xk. dq2 = 2 2 2 i = (1 + i · τ /2) q1 q2 (M.

(18. We choose to parametrize δS in such a way that the Euclidean 4-vector (x. where I is the 4 × 4 unit matrix.7. and they are in fact isomorphic in the vicinity of their respective identity elements.35) . Y3 ] = 0.4.37) (M. Y2 ≡ X 5 .36) ˆ ˆ ˆ ˆ ˆ ˆ Relabelling these last three generators as Y1 ≡ X 4 . however.80)) x = x − × x − ηx 4 x4 = x4 + η · x (M. (18. Y1 ] = [ X 2 . η2 . with the condition det S = +1. (M.74). x 3 ) and η = (η1 .therefore said to have the same algebra with the same structure constants.33) ˆ ˆ and similarly for X 2 and X 3 as in (M. as we discuss in section M. The six generators are (with dα1 ≡ 1 etc) ∂ ∂ ˆ − ix 2 X 1 = ix 3 ∂ x2 ∂ x3 (M. so that SO(3) is a subgroup of SO(4). which have six real parameters. Y3 ≡ X 6 . They are not the same for ‘large’ transformations. namely by 4 × 4 real anti-symmetric matrices δS. η3 ). together with (defining dα4 = η1 etc) ∂ ∂ ˆ + x1 X 4 = i −x 4 ∂ x1 ∂ x4 ∂ ∂ ˆ + x2 X 5 = i −x 4 ∂ x2 ∂ x4 ∂ ∂ ˆ + x3 X 6 = i −x 4 ∂ x3 ∂ x4 (M.79) and (18. x 2 . Y j ] = i ˆ ˆ [Yi . Y2 ] = [ X 3 .38) (M.75). X j ] = i ˆ ˆ [Xi . Infinitesimal SO(4) transformations are characterized by the 4D analogue of those for SO(3). we find the following algebra: ˆ ˆ [Xi .3 SO(4): The special orthogonal group in four dimensions This is the group whose elements are 4 × 4 matrices S such that S T S = I . (M. The Euclidean (length)2 2 2 2 2 x 1 + x 2 + x 3 + x 4 is left invariant under SO(4) transformations. Note that the first three components transform by (M. x 4 ) is transformed to (cf (18.18) when η = 0. Y j ] = i together with ˆ ijk Xk ˆ i j k Yk ˆ ijk Xk (M. M.34) (M.40) Copyright 2004 IOP Publishing Ltd .39) ˆ ˆ ˆ ˆ ˆ ˆ [ X 1 .32) where x = (x 1 .21).

43) (M.37)–(M. K j ] = i ˆ ijk Xk ˆ i j k Kk ˆ ijk Xk. X j ] = i ˆ ˆ [Xi . The six ˆ ˆ ˆ generators are then X 1 . They therefore behave like two independent angular momentum operators.51) (M. together with ∂ ∂ ˆ K1 = − i x 1 0 + x 0 1 ∂x ∂x ∂ ∂ ˆ K2 = − i x 2 0 + x 0 2 ∂x ∂x ∂ ∂ ˆ K3 = − i x 3 0 + x 0 3 ∂x ∂x The corresponding algebra is ˆ ˆ [Xi . From (M.4.41) (M. N j ] = i ˆ i j k Mk ˆ i j k Nk (M.50) (M.40) constitute the algebra of SO(4).52) (M.46).48) .45) we see that. (M.46) where η is now the infinitesimal velocity parameter (the reader may check that (x 0 )2 − x 2 is indeed left invariant by (M.43)–(M.21). M. K j ] = − i Copyright 2004 IOP Publishing Ltd . (M. the six generators have separated into two sets of three. We may think of infinitesimal Lorentz transformations as corresponding physically to ordinary infinitesimal 3D rotations. (M.49) ˆ ˆ [Ki .47) (M.44) (M.43)–(M. X 3 as in (M. to first order in and η). together with infinitesimal pure velocity transformations (‘boosts’). each set obeying the algebra of SO(3) (or of SU(2)) and commuting with the other set. The basic 4-vector then transforms by x0 x = = x0 − η · x x − × x − ηx 0 (M. This algebra may be simplified by introducing the linear combinations ˆ ˆ ˆ Mi = 1 ( X i + Yi ) 2 ˆ ˆ ˆ Ni = 1 ( X i − Yi ) 2 which satisfy ˆ ˆ [ Mi . X 2 .4 The Lorentz group In this case the quadratic form left invariant by the transformation is the Minkowskian one (x 0 )2 − x 2 (see appendix D of volume 1). The algebra (M.37) confirms that the three generators controlling infinitesimal transformations among the first three components x obey the angular momentum commutation relations.42) ˆ ˆ [ Mi .(M. N j ] = 0. M j ] = i ˆ ˆ [ Ni .45) is referred to as SU(2)×SU(2). in this form.45) (M.

X 3 of (M.55) 1 0 0 i 0 0 0 1 0  1    √ 0 0 0 0 0 3  0 1 √ 0 . . η = (η1 . (M.53) 2 q3 q3 where there are now eight of these η’s. η2 . .5 SU(3) A general infinitesimal SU(3) transformation may be written as (cf (12.56) λ7 =  0 0 −i  λ8 =   3 2 0 i 0 0 0 −√ 3 ˆ In this parametrization the first three of the eight generators G r (r = 1. (M.30). The structure constants are i f abc .72))     q1 q1 1  q2  = 1 + i η · λ  q2  (M.39).52) as compared with (M.4.54) 0 0 0 0 0 0 0 0 0       0 0 1 0 0 −i 0 0 0 λ4 =  0 0 0  λ5 =  0 0 0  λ6 =  0 0 1  (M. . 2. 8) ˆ ˆ ˆ are the same as X 1 .57) where a. . M. and the non-vanishing f ’s are as follows: f 123 = 1 f 345 = 1/2 f 147 = 1/2 f 156 = −1/2 f 246 = 1/2 f257 = 1/2 √ √ f 458 = 3/2 f 678 = 3/2. η8 ) and the λ-matrices are the Gell-Mann matrices       0 1 0 0 −i 0 1 0 0 λ1 =  1 0 0  λ2 =  i 0 0  λ3 =  0 −1 0 (M. .71) and (12. (M. chapter 2). i ∂ ∂ ˆ q3 − q1 G5 = 2 ∂q1 ∂q3 The SU(3) algebra is found to be ˆ ˆ ˆ [G a . . .Note the minus sign on the right-hand side of (M. G b ] = i f abc G c (M.10).59) (M.58) i ∂ ∂ ˆ q3 G7 = − q2 2 ∂q2 ∂q3 . .60) f 367 = −1/2 Note that the f ’s are anti-symmetric in all pairs of indices (Carruthers 1966. for example.29)–(M. X 2 . Copyright 2004 IOP Publishing Ltd . b and c each run from 1 to 8. The others may be constructed as usual from (M.

the quantum number J specifies each different representation. For example. The reason it does so ˆ2 is because (as should be familiar) the corresponding operator J commutes with every generator: ˆ2 ˆ (M. Clearly. together with the physical requirement ˆ that the Ji ’s (and the matrices representing them) must be Hermitian. 1. an identical matrix representation may be obtained for them: these matrices were denoted by Ti(T ) in section 12.61). J j ] = i ˆ i j k Jk (M.61). J2 .. X 2 . 3/2. These generators satisfy certain commutation relations. J3 ) all with the same commutation relations as (M. X r (without ˆ hats) which satisfy the same commutation relations as the X ν ’s—that is. since the generators of SU(2) have the same algebra as (M. . where the entire theory may be developed from the commutation relations (with h = 1) ¯ ˆ ˆ [ Ji .M.61) ˆ for the angular momentum operators Ji . it is also possible to find sets of matrices X 1 . . .63) [ J . Ji ] = 0. . Such an operator is called a Casimir operator and by a lemma due to Schur (Hammermesh 1962. pp 100–1) it must be a multiple of the unit operator. the matrices Ji are of dimension (2 J + 1) × (2 J + 1). For any given Lie algebra. M J ≤ J . It is important to note that J (or T ) can take an infinite sequence of values J = 0. corresponding physically to various ‘spin’ magnitudes.2.5 Matrix representations of generators and of Lie groups ˆ ˆ ˆ We have shown how the generators X 1 . analogous to |J M . equivalently. understood to be acting on functions of the ‘coordinates’ to which the transformations of the group refer. Thus there are infinitely many sets of three matrices (J ) (J ) (J ) (J1 . In general. X 2 . of the generators. .1. The numerical value it has differs for each different representation and may. more than one such operator is needed to characterize a ˆ2 representation completely. they have the same algebra. the Lie algebra of the group. . . In this case the matrices are of the form (in quantum-mechanical notation) Ji( J ) MJ MJ ˆ ≡ J M J | Ji | J M J (M. the problem amounts to finding a correct labelling of the base states. Since M J and M J each run over the 2 J + 1 values defined by (J ) − J ≤ M J . etc). ‘ J = 1/2’. section 27). X r of a Lie group can be constructed as differential operators. . 1/2. The idea is familiar from the study of angular momentum in quantum mechanics (Schiff 1968. Such matrices are said to form a (matrix) representation of the Lie algebra or. In physical terms.62) ˆ2 ˆ where | J M J is an eigenstate of J and of J3 with eigenvalues J ( J + 1) and M J respectively. A similar method for obtaining matrix representations of Lie algebras may be followed in other cases. therefore. . the two operators M and Copyright 2004 IOP Publishing Ltd . In the latter case. . in SO(4). be used to characterize a representation (namely as ‘ J = 0’. .

1. N = 0. since G 3 and ˆ 8 commute. so is SU(3). 3∗ (anti-triplet). . the labelling of the matrix elements of the generators is the same as it would be for two independent particles. SU(3) and the Lorentz group.66) (M. a unique labelling. 8 (octet) and 10 ˆ (decuplet) is. we may write such a matrix as r 1+i ν=1 ( ) ν Xν (M.65) (M. In a suitable parametrization. it is more common to denote them (as we have in the text) by their dimensionality. λb } = 4 δab I + 2dabc λc 3 (M. In the case of groups whose elements are themselves matrices. This is also equal to the number of independent mutually commuting generators (though this is by no means obvious). 3∗ ) = 4/3 ˆ C2 (8) = 3 ˆ C2 (10) = 6.3 and 12. where M. a further labelling is then required to characterize the states ˆ within a given representation (the analogue of the eigenvalue of J3 for angular momentum). The values of C2 in these representations are ˆ C2 (1) = 0 ˆ C2 (3. In practice. 1/2. For given M. which for the cases 1 (singlet). For SO(4) these further labels may be taken to be the eigenvalues ˆ ˆ ˆ ˆ of M3 and N3 : for SU(3) they are the eigenvalues of G 3 and G 8 —i. with two commuting generators M3 and N3 . SO(4) is a rank ˆ ˆ ˆ two group. SO(4). in fact. Two Casimir operators are therefore required to characterize the G representations of SU(3).1). such as SO(3). those corresponding to the third component of isospin and hypercharge. . (M. in the flavour case (see figures 12. . 3 (triplet). N the matrices are of dimension [(2M + 1) × (2N + 1)] × [(2M + 1) + (2N + 1)].64) and are symmetric in all pairs of indices (they are tabulated in Carruthers 1966. The number of Casimir operators required to characterize a representation is called the rank of the group (or the algebra).68) Copyright 2004 IOP Publishing Ltd . for the few SU(3) representations that are actually required. .e. one of spin M and the other of spin N. SU(2). Thus. table 2.4).67) Having characterized a given representation by the eigenvalues of the Casimir operator(s). one particular representation of the generators may always be obtained by considering the general form of a matrix in the group which is infinitesimally close to the unit element.ˆ2 N commute with all the generators and take values M(M + 1) and N(N + 1) respectively. Thus. which may be taken to be the ‘quadratic’ one ˆ ˆ1 ˆ2 ˆ8 C2 ≡ G 2 + G 2 + · · · + G 2 together with a ‘cubic’ one ˆ ˆ ˆ ˆ C3 ≡ dabc G a G b G c where the coefficients dabc are defined by the relation {λa .

ν. in section 12. Thus. . r ) are infinitesimal parameters. and the fact that the generators are independent. these matrices are of dimensionality r × r . It is not a coincidence that the coefficients on the right-hand side of (M. where r is the number of generators. . (M. X r ) ) are matrices representing the generators of the (matrix) group . [ X λ . X ν ]] + [ X µ .( ( ( where ( 1 . . . and each of µ. For SO(3). .1.70) are (minus) the SO(3) structure constants.12).73) Copyright 2004 IOP Publishing Ltd .48): Ti (1) jk = −i ijk . in angular momentum terms. λ runs from 1 to r . satisfying the SU(2) algebra. [ X ν . One representation of a Lie algebra is (R) always provided by a set of matrices {X ν } whose elements are defined by Xλ (R) µν ν = −cλµ (M. β (M.71) where the c’s are the structure constants of (M. [ X µ . X µ ]] = 0. . we obtain α α β α β cµν cλα + cνλ cµα + cλµ cνα = 0. These particular two representations are called the fundamental representations of the SU(2) and SU(3) algebras. where we found (SU(2)) ’s were just τ /2. This is exactly the same procedure we followed for SU(2) in section 12.26) that the three X ν (SU(3)) Similarly. respectively. 2 . (M. respectively.17))   0 0 0 (SO(3)) X1 =  0 0 −i  0 i 0   0 0 i (SO(3)) =  0 0 0  X2 −i 0 0   0 −i 0 (SO(3)) (M. from (12. they (SO(3)) ’s are the representations of lowest dimensionality.70) The matrices τi /2 and Ti(1) correspond to the values J = 1/2. .2 we saw that the eight SU(3) X ν ’s were just λ/2. the three X ν are (from (M. X λ ]] + [ X ν . J = 1.12) to evaluate the commutators. and ( X 1 ) .69) =  i 0 0  X3 0 0 0 which are the same as the 3 × 3 matrices Ti(1) of (12.72) Using (M. satisfying the SU(3) algebra.1. . X 2 ) . That this prescription works is due to the fact that the generators satisfy the Jacobi identity ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ [ X λ .

exponentiating the generators X ( ) just recreates the general matrices of the group. as required.46). Consider the infinitesimal Lorentz transformation (M.74) and (12. α2 . (M. for example the Dirac equation. precisely the (νβ) matrix element of (R) α (R) [X λ . . .80) Copyright 2004 IOP Publishing Ltd . In the case of the groups whose elements are matrices. that is. The 4 × 4 matrix corresponding to this may be written in the form 1 + i · X (LG) − iη · K (LG) where X1 (LG) (M. .71).31) and (12. (M. β)). a corresponding matrix representation of the group elements can be obtained by exponentiation. of course. The representation in which the generators are represented by (minus) the structure constants.81) for SU(3)). M.49) for SU(2).76). so we may call this the ‘self-representation’: the one in which the group elements are represented by themselves. . and (12. say. in the sense of (M.12). αr ) (see (12.76) where α = (α1 . if the group elements obey g(α)g(β) = g(γ (α.74) is. In the more general case (M. is called the regular or adjoint representation. β)) then D (P) (α)D (P) (β) = D (P) (γ (α.6 The Lorentz group Consideration of matrix representations of the Lorentz group provides insight into the equations of relativistic quantum mechanics.75) (R) showing that the X µ ’s satisfy the group algebra (M.78) It is a rather remarkable fact that there are certain. via D (P) (α) = exp{iα · X (P) } (M. Having obtained any particular matrix representation X (P) of the generators of a group . the crucial property of the matrices D (P) (α) is that they obey the same group combination law as the elements of the group they are representing.The reader may fill in the steps leading from here to the desired result: Xλ (R) να (R) Xµ αβ (R) − Xµ να Xλ (R) αβ α (R) = cλµ X α νβ .74) (M. X µ ] = cλµ X α (R) (M.79) 0  0 =  0 0  0 0 0 0  0 0 0 0   etc 0 −i  i 0 (M. 10 × 10 matrices which multiply together in exactly the same way as the rotation matrices of SO(3).77) (M.

69) but with an extra border of 0’s) and  0 −i 0  −i 0 0 (LG) K1 =   0 0 0 0 0 0  0 0 −i  0 0 0 (LG) K2 =   −i 0 0 0 0 0  0 0 0  0 0 0 (LG) K3 =   0 0 0 −i 0 0 In (M. the limiting velocity c is not included (the γ -factor goes to infinity). Q j ] = 0 Copyright 2004 IOP Publishing Ltd i j k Pk i j k Qk 1 2 (X (M.84) − iK ). (M.81) the matrices are component vector  0 x  x1   x2 x3   . Q j ] = i [Pi . Let us introduce the combinations P ≡ 1 (X + iK ) 2 Q≡ Then the algebra becomes [Pi .83) (M.50)–(M.52).52) as expected.86) (M. P j ] = i [Q i .87) . 0  0 (M. A theorem states that only the generators of compact Lie groups can be represented by finite-dimensional Hermitian matrices.50)–(M.81) understood to be acting on the four- (M. and to the corresponding matrices of SU(2) and SU(3). In a general representation of the Lorentz group.(as in (M. are not Hermitian. For the Lorentz group. in contrast to X i (SO(3)) or X i .82) (LG) (LG) It is straightforward to check that the matrices X i and K i satisfy the algebra (M. and so the group is non-compact. K i will obey the algebra (M. (LG) (LG) An important point to note is that the matrices K i .   0 0   0  0  0 0   0  0  −i 0  . Here ‘compact’ means that the domain of variation of all the parameters is bounded (none exceeds a given positive number p in absolute magnitude) and closed (the limit of every convergent sequence of points in the set also lies in the set). the generators X i .80) and (M.85) (M.

The ‘−η · σ /2’ part shows how such a spinor transforms under a pure (infinitesimal) velocity transformation. (M. So we can appropriate standard angular momentum theory to set up the representations of the algebra of the P’s and Q’s—namely. and similarly for Q3. 1/2. and the same is. Exponentiating this part gives a transformation law φ = exp(− 1 v n · σ )φ 2 ˆ Copyright 2004 IOP Publishing Ltd (M. We label this representation by the values of 2 P ( 1 ) and Q (0) (these are the eigenvalues of the two Casimir operators). for P = 0. so that X = iK (M.0) = − σ .88) and (M. . this becomes 1 + i · σ /2 − η · σ /2. first met in section 4.4: it exponentiates to give exp(iα · σ /2) for finite rotations.95) .43)–(M. Then 2 using (M. which therefore transform under an infinitesimal Lorentz transformation by φ = (1 + i · σ /2 − η · σ /2)φ. (M.81) that the matrices iK (LG) are Hermitian.88) in this representation.90) 1 i (M. in fact.1 . they behave just like two independent (mutually commuting) angular momenta. 0) representation of the Lorentz group’.89) we find that X ( 2 .0) = 1 σ 2 and 1 (M.91) K ( 2 ..89) the familiar matrices for spin. Consider the particular case where the eigenvalue of Q 2 is zero (Q = 0) and the value of P is 1/2. We can see from (M. and act on two-component spinors. The eigenvalues of P 2 are of the form P(P +1).45). true in a general finitedimensional representation.92) (M. 2 Now recall that the general infinitesimal Lorentz transformation has the form 1 + i · X − iη · K .which are apparently the same as (M.94) We say that φ ‘transforms as the (1/2. . In the present case. The ‘1 + i · σ /2’ part is the familiar (infinitesimal) rotation matrix for spinors. and similarly for Q 2 : the eigenvalues of P3 are M P where −P ≤ M P ≤ P. while the second condition tells us that P = 1 (X + iK ) = 1 σ 2 2 (M. . The first condition implies that the Q’s are identically zero.93) These matrices are of dimension 2 × 2.

1 wave equations we have introduced in 2 the text—that is.for a finite boost characterized by a speed v and a direction n. (M.46) takes (E.98) We see that φ and χ behave the same under rotations but ‘oppositely’ under boosts. φ will transform by φ = exp(iα · σ /2)φ. How does all this relate to the spin. (M.99) and inserting Vη Vη gives −1 −1 [Vη (E − σ · p)Vη ]Vη φ = 0.101).103) (M. (M. the previous steps yield X (0. a second two-dimensional representation.99) for a massless spin.150) and its u L analogue (see also (20. (M. 2 ) = 1 (M. 2 ) = 1 σ 2 1 (M. under a three-dimensional rotation. the Dirac equation of chapter 4 and the Weyl equation (4. (M. 2 So the corresponding two-component spinor χ transforms by K (0. 2 The work of section 4. 1/2).106) Copyright 2004 IOP Publishing Ltd .97) χ = (1 + i · σ /2 + η · σ/2)χ. 0) object.104) φ = (1 − η · σ /2)φ exactly as in (M. which is characterized by the labelling P = 0.103) to be consistent.101) p = p − η E.100) So let us consider boosts. demonstrating that this φ is (as the notation has assumed) a (1/2. (M. however. (M.96) as before.1 particle with energy E and momentum p such that E = | p|. Q = 1/2.4 guarantees that.99) becomes (E − σ · p )φ = 0.99).102) We shall verify that for (M. φ must transform as (M. (M. In this case.94).105) −1 −1 Then applying Vη to (M. An infinitesimal velocity transformation (see M.56))? Consider the Weyl equation (E − σ · p)φ = 0 (M. In this primed frame. but i σ. p) to E = E −η· p (M. which we denote by (0. Let Vη = (1 − η · σ /2). ˆ There is.102) and (M.

via φ ψ= (M.152).110) as discussed in problem 4. (M. (M.1 and M.15 and section 20. In a similar way.4.101) and (M.15 will recognize these manipulations. Furthermore.2) that the algebras of these two groups are identical. 1/2) object.151) and (4. So the groups are isomorphic in the vicinity of their respective identity elements. as we shall now discuss.4. it might appear that the groups are fully isomorphic. 0) object.111) M. Since exponentiating these infinitesimal matrix transformations produces matrices representing group elements corresponding to finite transformations in both cases. one can verify that the Weyl spinor χ satisfying ( E + σ · p)χ = 0 (M.112) (M.107) Hence (M. The φ and χ spinors are projected out of ψ by the chirality operators PR.109) χ and describes a massive spin. in the representation such that γ5 = 1 0 0 −1 .4. The four-component Dirac wavefunction ψ is put together from one φ and one χ .4.The part in square brackets is.104).102) as compared with (4. to first order in η. But actually they are not.103) follows with the identification (M.L = 1 (1 ± γ5 ) 2 as discussed in section 20. showing that also under boosts φ transforms as a (1/2. matrix representations of one algebra automatically provide representations of the other. The reader who has worked through problem 4. with a sign change for the infinitesimal velocity parameters in (M. Copyright 2004 IOP Publishing Ltd .108) transforms as a (0.1 particle according to the equations 2 Eφ = σ · pφ + mχ Eχ = −σ · pχ + mφ (M.7 The relation between SU(2) and SO(3) We have seen (sections M. (1 + η · σ /2)( E − σ · p)(1 + η · σ /2) = ( E − η · p) − σ · ( p − Eη) = E −σ · p.

a = cos θ/2 + in z sin θ/2 b = (n y + in x ) sin θ/2 (M. Now let us consider the correspondence—or mapping—between the matrices of SO(3) and SU(2): we want to see if it is one-to-one. and one x y z can convince oneself that the full range of a and b is covered if θ/2 lies between 0 and π (in particular.1. α2 . of a sphere of radius 2π .35) we proposed.113) 0 0 1 On the other hand. The notation strongly suggests that the matrix V (n. where θ is an angle and n is a unit vector.117) Thus if we agree to limit n to directions in the upper hemisphere of figure M. to write α = nθ . z (M. by taking advantage of the fact that R(n. since (as the ˆ ˆ reader may verify) exp(iθ τ · n/2) = cos θ/2 + iτ · n sin θ/2 ˆ ˆ (M. ˆ   cos θ sin θ 0 R(ˆ . the space of the α’s is the interior. θ ). that is.116) with n 2 + n 2 + n 2 = 1. as shown in figure M. we can say that the whole sphere represents the parameter space of SU(2). the same parameters n and θ specify a rotation. ˆ we may allow the range of θ to extend to 2π. θ ) ≡ exp(iθ n · τ/2) of SU(2) corresponds ˆ ˆ to the matrix R(n. Then.32) we wrote V as exp(iα · τ /2). However. but now θ (rather than θ/2) runs from 0 to π. and below (12. What about the parameter space of SO(3)? In this case. π + θ ) = R(−n. since the corresponding region of a. the choice of which we are now going to examine in more detail than previously. which has not been constrained in any way). It ˆ follows that the parameters α satisfy α2 ≤ 4π 2 . ˆ ˆ (M. ˆ where n is the axis of the rotation and θ is the angle of rotation. in (M. For example. it is not necessary to extend the range of θ/2 so as to include the interval π to 2π. but that of SO(3) is provided by the upper half only. further.1.114) where |a|2 +|b|2 = 1 from the unit determinant condition. ˆ Copyright 2004 IOP Publishing Ltd . A general 3D rotation is described by the SO(3) matrix R(n. b can be covered by changing the orientation of n. Clearly the condition |a|2 + |b|2 = 1 is satisfied. and surface. ˆ for 3D rotations.We begin by re-considering the parameters used to characterize elements of SO(3) and SU(2). which certainly involves three real parameters α1 .114). θ ) =  − sin θ cos θ 0  . In (12. α3 .115) it follows that this latter parametrization corresponds to writing. It therefore depends on three real parameters. we can write the general SU(2) matrix V in the form V = a −b ∗ b a∗ (M. θ ) of SO(3) but the way it actually works has a subtlety. θ ).

So we have D(a) = D(a)D(e). θ ) x · τ V † (n. not an isomorphism). a homomorphism. from (M. θ = 2π).118) establishes a precise mapping between the elements of SU(2) and those of SO(3) but it is not one-to-one (i.1. and hence so will the associated SO(3) matrix R(n. using (M. It is. the upper (stippled) hemisphere that of SO(3). is the identity matrix. the SU(2) matrices cannot be said to represent rotations.119) 0 −1 but the corresponding rotation matrix. θ ) ˆ ˆ (M. The parameter spaces of SO(3) and SU(2): the whole sphere is the parameter space of SU(2). be a familiar manipulation). but can an SU(2) matrix be said to provide a valid ˆ representation of SO(3)? Consider the case V (n = z . However. ae = a by the property of e. θ ). We form the quantity x · τ .e. Let D(a). θ ).113). Thus. θ ). since plainly V can always be replaced by −V and x will be unchanged. and assert that x · τ = V (n. and hence D(e) = I .113): the general case follows with more labour (but the z general infinitesimal case should. therefore. From (M.118) where x = R(n.a = nq Figure M. by now. (M.115) ˆ ˆ this is equal to −1 0 (M. ˆ Next. Then D(ae) = D(a)D(e) by the fundamental property (M. since (M. We can easily verify (M. in the Copyright 2004 IOP Publishing Ltd . e of . D(e) represent the elements a.78) of representation matrices. our theorem is violated.118) for the special case ˆ R(ˆ . we prove a little theorem to the effect that the identity element e of a group must be represented by the unit matrix of the representation: D(e) = I . V (n. θ )x. Now let us return to the correspondence between SU(2) and SO(3). Hence.119) is plainly not the identity matrix of SU(2). θ ) ˆ corresponds to R(n.

SU(2) matrices provide a representation of SO(3). spin.118)—just as. of course. In fact. θ ).1 particles certainly do exist. in general. V and −V correspond to the same SO(3) element R(n. θ = 4π). If we ‘factor out’ this sign. chapter 8). two and only two distinct elements of SU(2). A more mathematically precise way of saying this is given in Jones (1990.120) correspond to the identity element of SO(3) in the correspondence (M. namely 1 0 0 1 and −1 0 0 −1 (M. so Nature appears to 2 make use of these ‘not quite’ representations! The SU(2) identity element is. the groups are isomorphic. Copyright 2004 IOP Publishing Ltd . up to a sign. The failure to be a true representation is localized simply to a sign: we may indeed say that. Nevertheless. as we ˆ saw. V (n = z .strict sense. confirming that the rotational properties of a spinor ˆ ˆ are quite other than those of a classical object.

Then the nature of the divergence as d → 4 can be exposed (much as we did with the cut-off procedure in section 10. We consider (N. The idea behind dimensional regularization (’t Hooft and Veltman 1972) is to treat d as a variable parameter.16)) and shifting the origin of the loop momentum to complete the square (cf (10.1) or to a similar integral with factors of k (such as kµ kν ) in the numerator.42) and (11. We write Id = 1 (n − 1)! ∂ ∂ n−1 1 dd k . We begin (as discussed in connection with (15.40) and (11. so that (N.APPENDIX N DIMENSIONAL REGULARIZATION After combining propagator denominators of the form ( p2 − m 2 + i )−1 by Feynman parameters (cf (10. Then the 2 Minkowskian square k 2 becomes −(k e )2 − k2 ≡ −kE .75)) by ‘Euclideanizing’ the integral by setting k 0 = ik e with k e real. the case of physical interest is d = 4. unlike the simple cut-off regularization we used in chapters 10 and 11.3) and dealt with by a suitable renormalization scheme.3) 2 (n − 1)! ∂ (2π)d (kE + ) the ‘i ’ may be understood as included in . of course. and n is commonly 2 (e. n) ≡ 1 dd k (2π)d [k 2 − + i ]n (N. so that now −i 1 ∂ n−1 dd k E Id = (N.g.4) Copyright 2004 IOP Publishing Ltd . For our purposes.2) The d dimensions are understood as one time-like dimension k 0 and d − 1 space-like dimensions.1) converges and can be evaluated explicitly as a function of d (and.16)).1) first. Power-counting shows that (N. The crucial advantage of dimensional regularization is that it preserves gauge invariance. the other variables. and dd k becomes idd kE . d [k 2 − (2π) +i ] (N.1) diverges for d ≥ 2n. in one-loop self-energies). all one-loop Feynman integrals may be reduced to evaluating an integral of the form Id ( . The integral is evaluated by 2 introducing the following way of writing (kE + )−1 : 2 (kE + )−1 = 0 ∞ dβe−β(kE + 2 ) (N. including n). taking values smaller than 2n.

. (N.10) z where γ is the Euler–Mascheroni constant having the value γ ≈ 0.5) The interchange of the orders of the β and kE integrations is permissible since Id is convergent.11) (N.9) gives an explicit definition of Id which can be used for any value of d..5772.76) for the case n = 2 (remembering that Id was defined in Minkowski space). it is convenient to write (N.8) Since (n) = (n − 1)!. in fact. . −1. Equation (N. The behaviour near z = 0 is given by 1 (z) = − γ + O(z) (N. −2. (N.6) dβe−β β −d/2 (N.which leads to Id = −i (n − 1)! ∂ ∂ n−1 0 ∞ dβ dd kE −β(k 2 + ) E e . The last integral can be written in terms of Euler’s integral for the gamma function (z) defined by (z) = 0 ∞ x z−1 e−x dx.7) dβe−β β n−(d/2)−1. Id = = −i 1 (n − 1)! (4π)d/2 −i (−1)n−1 (n − 1)! (4π)d/2 e−β (2π)d ∂ ∂ π β n−1 d/2 . (z) has isolated poles (see apppendix F of volume 1) at z = 0. (2π)d (N.9) Id = i (4π)d/2 (n) This formula agrees with (15. Using z (z) = (z + 1) we find the behaviour near z = −1: (−1 + t) = −1 (t) 1−t 1 + 1 − γ + O(t) = − t (N. . The kE integrals are.12) Copyright 2004 IOP Publishing Ltd . As a function of z. not necessarily an integer.8) entirely in terms of functions as (−1)n (n − d/2) (d/2)−n . a series of Gaussians:    d dk 2 dk j −βk j  d E −β(k 2 + ) E e e = e−β   (2π)d (2π) j =1 = Hence. (N.

13) Consider now the case n = 2. For example. 2) = i 2 − γ + ln 4π − ln (4π)2 + O( ) . for small . Setting d = 4 − . (N.10).17) From Lorentz covariance this must be proportional to the only second-rank tensor available. So A= 1 d 1 = d dd k k2 d (k 2 − (2π) + i )n 1 dd k + (2π)d (k 2 − + i )n−1 1 dd k (2π)d (k 2 − + i )n Copyright 2004 IOP Publishing Ltd .79)) that Id ( .7) that a gauge coupling which is dimensionless in d = 4 dimensions will acquire mass dimension µ /2 in d = 4 − dimensions. the divergent behaviour is given by (2 − d/2) = 2 − γ + O( ) (N. (N. 2 z 2 (N.17) with gµν . (2π)d [k 2 − + i ]n (N.15) When − /2 and (4π)−2+ /2 are expanded in powers of .16) to ln( /µ2 ). 2) = i (4π)2− /2 − /2 2 − γ + O( ) . Finally. we find (see (15.9) will have a pole at d = 4. in the MS scheme one subtracts the pole and the ‘−γ + ln 4π’ piece.78). When expanded in powers of . as stated in (15. Renormalization schemes will subtract the explicit pole pieces (which diverge as → 0). the terms linear in will produce terms independent of when multiplied by the −1 in the bracket of (N.15). A vacuum polarization loop with two powers of the coupling will then contain a factor µ . n) ≡ µν dd k k µk ν . namely g µν : µν Id = Ag µν . (N. Using x ≈ 1 + ln x + O( 2 ). using g µν gµν = d in d dimensions. this will convert the ln in (N. 2) is then given by Id ( . consider the integral Id ( . but may also include in the subtraction certain finite terms as well.similarly near z = −2: (−2 + t) = 1 1 3 + − γ + O(t) .16) Another source of -dependence arises from the fact (see problem 15.18) The constant ‘ A’ can be determined by contracting both sides of (N.14) from (N. Id ( . for which (n − d/2) in (N.

19) Using these results. one can show straightforwardly (problem 15.18)—i. the piece in braces—vanishes. starting from a gauge-invariant formulation of the theory. the renormalization programme can be carried out while retaining manifest gauge invariance.= i(−1)n (4π)d/2 (d/2)−n+1 d − (n − 1 − d/2) + (n − 1) (n − d/2) (n) i(−1)n (d/2)−n+1 (n − 1 − d/2) {−n + (n − d/2)} = (4π)d/2 d (n) i(−1)n−1 (d/2)−n+1 1 (n − 1 − d/2) . Copyright 2004 IOP Publishing Ltd .e.5) that the gauge-non-invariant part of (11. = (4π)d/2 2 (n) (N. With the technique of dimensional regularization.

∂θ1 then necessarily (O. as with ordinary numbers and functions. So let us begin with differentiation.71)). 2 2 θ1 = θ2 = 0. and ∂ (θ1 θ2 ) = θ2 . We define ∂(aθ ) =a (O. f (θ ). For our application. integration would be some kind of inverse of differentiation.5) ∂θ2 Consider now a function of one such variable.2) Grassmann numbers can be added and subtracted in the ordinary way and muliplied by ordinary numbers. But something new is needed to represent. (O. the necessary mathematics was developed by Grassmann in 1855 and applied to quantum amplitudes by Berezin (1966). It is natural to think that. it seems that to represent amplitudes involving fermionic operators by path integrals we must think in terms of ‘classical’ anti-commuting variables. An expansion of f in powers of θ terminates after only two terms because of the property (O. θ2 satisfy the fundamental relation θ1 θ2 + θ2 θ1 = 0 and. (O. Thus. Any two Grassmann numbers θ1 . since the fermionic operators anti-commute.6) Copyright 2004 IOP Publishing Ltd .APPENDIX O GRASSMANN VARIABLES In the path integral representation of quantum amplitudes (chapter 16) the fields are regarded as classical functions. the essential thing we need to be able to do with Grassmann numbers is to integrate over them. (O. for example.3) ∂θ where a is any ordinary number.4) ∂ (θ1 θ2 ) = −θ1 . the time-ordered product of two fermionic operators: there must be a sign difference between the two orderings. Matrix elements of time-ordered products of bosonic operators could be satisfactorily represented (see the discussion following (16.1) (O.2): f (θ ) = a + bθ. of course. Fortunately.

One further property fixes its value: we require the result to be invariant under translations of θ by θ → θ + η.4. the innermost integral is done first. which are complex-valued. Thus we may define 1 ψ = √ (θ1 + iθ2 ) 2 Copyright 2004 IOP Publishing Ltd 1 ψ ∗ = √ (θ1 − iθ2 ) 2 (O.10) This has changed the constant (independent of θ ) term but left the linear term unchanged. we define dθ (a + bθ ) = b (O.So ∂ f (θ ) =b ∂θ (O. then the next. (O.12) that is.9) but also Such an integral should be linear in f . it must be a linear function of a and b . and so on. We only need to consider integrals over the complete range of θ .11) which means that integration is. we could deduce 0 = A−1 ( A2 ) = ( A−1 A) A = A for all A). So we require dθ (a + bθ ) = dθ ([a + bη] + bθ ). in some sense. of the form dθ f (θ ) = dθ (a + bθ ). This property is crucial to manipulations made in the path integral formalism.7) ∂2 f =0 (O. for instance in ‘completing the square’ manipulations similar to those in section 16. Thus. where η is a Grassmann number. the same as differentiation! When we integrate over products of different θ ’s. we need to specify a convention about the order in which the integrals are to be performed. but with Grassmann numbers. Hence.13) . (O. we must approach Grassmann integration other than via an inverse of differentiation. We adopt the convention dθ1 dθ2 θ2 θ1 = 1 (O. we need to introduce complex Grassmann numbers. which is conventionally taken to be simply b. Thus. which are built out of real and imaginary parts in the usual way (this would not be necessary for Majorana fermions).8) ∂θ 2 for any such f . The only linear function of a and b which behaves like this is a multiple of b. Since our application will be to Dirac fields. thus. the operator ∂/∂θ has no inverse (think of the matrix analogue A2 = 0: if A−1 existed.

in the Grassman case. These are the terms ∗ ∗ ∗ ∗ 1 2 [M11 M22 (ψ1 ψ1 ψ2 ψ2 + ψ2 ψ2 ψ1 ψ1 ) ∗ ∗ ∗ ∗ + M12 M21 (ψ1 ψ2 ψ2 ψ1 + ψ2 ψ1 ψ1 ψ2 )]. (O. we find that it becomes dψ ∗ dψ ψψ ∗ (1 + bψψ ∗ ) = dψ ∗ dψ ψψ ∗ = 1 (O.22) Copyright 2004 IOP Publishing Ltd .18) and the insertion has effectively produced a factor b −1 .17) The important point here is that. whose entries are ordinary numbers.and then −idψdψ ∗ = dθ1 dθ2 .14) is consistent under complex conjugation. rather than a negative. which is essentially all we need in the path integral formalism. (O. However. in the expansion of the exponential.21) and M is a 2 × 2 matrix. b appears with a positive. ψ1 . The only terms which survive the integration are those which. (O. We are now ready to evaluate some Gaussian integrals over Grassmann variables. This effect of an insertion is the same in the ‘ordinary variables’ case: dx dy(x 2 + y 2 )/2 e−b(x 2 +y 2 )/2 = 2π/b 2. if we insert a factor ψψ ∗ into the integrand in (O. ∗ ∗ contain each of ψ1 . (O. (O.16). power on the right-hand side.20) where ψ= ψ1 ψ2 (O.15) Then (O.16) Note that the analogous integral with ordinary variables is dx dy e−b(x 2 +y 2 )/2 = 2π/b. ψ2 and ψ2 exactly once.19) Now consider a Gaussian integral involving two different Grassmann variables: ∗T ∗ ∗ dψ1 dψ1 dψ2 dψ2 e−ψ Mψ (O.14) It is convenient to define complex conjugation to include reversing the order of quantities: (ψχ)∗ = χ ∗ ψ ∗ . (O. We begin with dψ ∗ dψ e−bψ ∗ψ = = dψ ∗ dψ(1 − bψ ∗ ψ) dψ ∗ dψ(1 + bψψ ∗ ) = b.

ψ3 . this produces ∗ ∗ ( M11 M22 − M12 M21 )(ψ2 ψ2 ψ1 ψ1 ). or take on trust. (O. . Let H be diagonalized by the unitary transformation z1 z2 with UU † = I .29) can be evaulated using (O. (O.30) Copyright 2004 IOP Publishing Ltd . consider now a two-variable (both complex) analogue of (O. with positive eigenvalues b1 and b2 . and the result is proportional to (b1 b2 )−1 .28) = 1. This result is sufficient to establish the assertion made in section 16. ψ N . b2 > 0.24) with an analogous result for two ordinary complex numbers z 1 . Thus—compare (O. In this case we consider the integral ∗ ∗ dz 1 dz 1 dz 2 dz 2 e−z ∗Hz (O. ψ2 . The integral (O.26) dz 1 dz 2 = det U dz 1 dz 2 ∗ ∗ dz 1 dz 1∗ dz 2 dz 2∗ = dz 1 dz 1 dz 2 dz 2 (O. . setting z 1 = (x 1 + √ √ iy1 )/ 2. we need to ∗ ∗ re-order the terms into the form ψ2 ψ2 ψ1 ψ1 .18): ∗ ∗ ∗ dψ1 dψ1 dψ2 dψ2 ψ1 ψ2 e−ψ ∗T Mψ .25) where z is a two-component column matrix with elements z 1 and z 2 . Returning to integrals of the form (O.4 concerning the integral (16.17)—Gaussian integrals over complex Grassmann variables are proportional to the determinant of the matrix in the exponent. We take the matrix H to be Hermitian. according to the convention (O. when written in ‘discretized’ form. Next.24) The reader may show. while those over ordinary complex variables are proportional to the inverse of the determinant. .22) conveniently. z 2 = (x 2 + iy2 )/ 2. when diagonalized. the obvious generalization to N independent complex Grassmann variables ψ1 . We may contrast (O.29) the integrals converging provided b1 .27) (O.20).82). z 2 .20) is therefore just ∗ ∗ dψ1 dψ dψ2 dψ2 e−ψ ∗T Mψ = det M.16) and (O.23) and the integral (O. .25) then becomes dz 1 dz 1∗ e−b1 z1 z1 ∗ dz 2 dz 2∗ e−b2 z2 z2 ∗ (O. which is the inverse of the determinant of the matrix H . (O.12).To integrate (O. (O.17). Then and so since | det U |2 =U z1 z2 (O.

31) With this result we can make plausible the fermionic analogue of (16. Copyright 2004 IOP Publishing Ltd . (O. ¯ |T ψ(x 1 )ψ(x 2 ) | = 1 The reader may interpret this as a finite-dimensional determinant. namely ¯ ¯ ¯ / ψ ψ ψ(x 1 )ψ(x 2 ) exp[− d4 x E ψ(i∂ − m)ψ] 4 x ψ(i∂ − m)ψ] ¯ ψ ψ exp[− d E ¯ / (O.∗ This time.32) ¯ note that ψ and ψ ∗ are unitarily equivalent. p the familiar Dirac propagator. but this is just ( / − m)−1 in momentum space. only the term ψ1 ψ2 in the expansion of the exponential will survive the integration and the result is just −M12 . while the numerator is this same determinant multiplied by the / inverse of the operator (i∂ − m). after discretization. the reader should be convinced that the general result is dψi∗ dψi ψk ψl∗ e−ψ i ∗T Mψ = (M −1 )kl det M.79). The denominator of this expression / is 1 det(i∂ − m). By exploring a similar integral (still ∗ with the term ψ1 ψ2 ) in the case of three complex Grassmann variables.

Specifically. up/down’ description with four degrees of freedom. in terms of the way two-component spinors transform under Lorentz transformations. For the Dirac field. whereas a neutral scalar field has only one. spin. at first sight. in that case. which appears to lead inevitably to the ‘particle/anti-particle.6— that is. a charged scalar field (or one carrying some other conserved quantum number) has two field degrees of freedom. for example. But. for neutral fermions such as neutrinos.1) could be understood as being built from two different kinds of spinor. corresponding to particles and anti-particles with spin up or down (or helicity +1 or −1.1 Spin. one way of approaching the ‘number of degrees of freedom’ issue is via the discussion of the Dirac equation in appendix M.1 wave equations 2 Within the framework of quantum-mechanical wave equations (rather than that of quantum fields).1 particles.2. The two degrees of freedom correspond to the physically distinct states of particle and anti-particle. 2 P. section M. It is not so clear.2). and so have only two degrees of freedom corresponding just to the two possible spin states. Such particles (which clearly cannot be electrically charged) would then be fermionic analogues of the π 0 . 0) representation of the Lorentz group.1. which is its own anti-particle. via φ = (1 + i · σ /2 − η · σ /2)φ Copyright 2004 IOP Publishing Ltd (P. there is an additional doubling of the number of degrees of freedom to four in all. We therefore begin by re-considering relativistic wave equations for spin. etc). introduced in section 7.2) . where there is room for such a possibility in the conventional presentation of the Dirac equation (see section 4. One kind (φ) transforms according to the (1/2.APPENDIX P MAJORANA FERMIONS In this appendix we aim to give an elementary introduction to ‘Majorana’ fermions—that is. As we saw in section 7. we saw there that the 4-component Dirac spinor ψ= φ χ (P.1 particles with no conserved quantum number allowing 2 an observable distinction to be made between the ‘particle’ and the ‘anti-particle’. the possibility exists that they might be their own anti-particles.

4) by φc . and similarly with R ↔ L.5).9) that (−i∂t + iσ · ∇)[(i∂t + iσ · ∇)φ] = m(−i∂t + iσ · ∇)(iσ2 φ ∗ ) = − m 2 φ. which transforms like a ‘χ ’ rather than a ‘φ ’? Then we could. constructed from the components of φ (and.4) and (P. consider1 φc ≡ iσ2 φ ∗ .7): we find that (−i∂t + iσ · ∇)(iσ2 φ ∗ ) = −mφ.4) (P. multiply from the left by iσ2 .8) and (P. It follows from (P. we may say that iσ2 ψR transforms like ψL . (P. the discussion of the SU(2) representations 2 and 2∗ in the paragraph containing equations (12. shows that ‘iσ2 φ ∗ so as to make iσ2 real). at this point. Equation (P.8) Take the complex conjugate of this equation.4. Copyright 2004 IOP Publishing Ltd . Then φc = (iσ2 φ ∗ ) = iσ2 (1 − i · σ /2 − η · σ /2)∗ φ ∗ = iσ2 (1 + i · σ ∗ /2 − η · σ ∗ /2)φ ∗ = (1 − i · σ /2 + η · σ /2)(iσ2 φ ∗ ) (P. In ‘R–L’ language.7) ∗ ∗ ∗ using σ1 = σ1 . φ is ψR and χ is ψL ). It is indeed possible to find such a φc : namely.for infinitesimal rotations and boosts. for example.9) (P. σ2 = −σ2 . (P. We may ask the question: is it possible to find a spinor φc .5) In (P.7) ∗ ’ does transform like a ‘χ ’ (the ‘i’ is inserted for convenience. replace χ on the right-hand side of (P. (In terms of the ‘R–L’ classification of section 20. (P. σ3 = σ3 and σi σ j + σ j σi = 2δi j .10) 1 The reader may usefully bear in mind. 1/2) representation via χ = (1 + i · σ /2 + η · σ /2)χ. as in (P. without additional degrees of freedom).3) The Dirac equation itself is then (i∂t + iσ · ∇)φ = mχ (i∂t − iσ · ∇)χ = mφ. the two-component wave equation (P.6) (i∂t + iσ · ∇)φ = m(iσ2 φ ∗ ). and the two sides would transform consistently but involve only the degees of fredom in φ . hence. and commute the σ2 through the bracket on the left-hand side. since both sides must transform consistently under Lorentz transformations.53)–(12. the ‘i∂t ± iσ · ∇ ’ factors on the left-hand sides are just what is required to change the transformation character of a φ into that of a χ or vice versa. then.57) in chapter 12. The other kind (χ ) transforms as a (0. (P. Consider.

(P.1) in this case too.15) ψM = iσ2 φ ∗ Consider now the charge conjugation operation C0 of section 20.2 However. may be said to describe a self-conjugate particle.27) on the Dirac wavefunction ψ .11) showing that φ satisfies the KG equation. hence.17) φ 0 ψM.1 particle. Comparison of (P.18) 2 The true ‘anti-particle’ is defined via the operator CPT.5. φ is a valid relativistic wavefunction for a massive spin. 2 Plane-wave solutions can be found.5) reveals that it is consistent to use a fourcomponent wavefunction of the type (P. since this is believed to be conserved by all ˆˆˆ Copyright 2004 IOP Publishing Ltd .14) ∂t ρ + ∇ · j = 0. identifying χ with iσ2 φ ∗ . in this sense. (P.8). We find that a∗ b∗ .12) via the wave equation (P. We find that ∗ ψM C = iγ2 ψM = 0 iσ2 −iσ2 0 = ψM .13) We can also check that ρ = φ † φ and j = φ † σ φ satisfy the continuity equation (cf (4.9) and (P.R = interactions. (∂t2 − ∇ 2 + m 2 )φ = 0 (P. Hence. φ∗ iσ2 φ (P. These results were first given by Case (1957). and writing φ .L = 0 iσ2 φ ∗ (P.55)): (P.16) = φ iσ2 φ ∗ Hence ψM satisfies the Majorana condition ψM C = ψM and.and. having the form φ = e−i p·x where c d will be related to c d = a b a b + ei p·x c d (P. in terms of the ‘R–L’ classification ψM. assuming that it is the ‘right’ operation in the present case as well. precisely the condition we imposed in (4. even though it has only two components. E +σ · p . − iσ2 m (P.

24) (P.L ψM.L ψM. λ = +1) + ei p·x (iγ2 u ∗ ( p. as usual.R }. when we try to describe this theory in terms of a Lagrangian.23) (P.46). which can also be written.23) vanishes identically. Let us therefore now consider the quantum field case. however.119) but resolved there by the fact that φ was replaced by a two-component fermionic (anti-commuting) field. as ¯ ¯ m{ψM. If φ is an ordinary c-number spinor of the form φ= a b (P. and we seem unable to form the required Lagrangian.21).9) from the Lagrangian ¯ / ÄM = ψM (i∂ − m)ψM .L + ψM.we obtain ψM.L C = ψM. we find that the four-component wavefunction has the form     a a∗ E+σ · p .R .15).47) and the normalization (20.20) a Let us take = φ+ .13) into (P.22) = im{φ † σ2 φ ∗ − φ T σ2 φ}.R ψM. The same point was noticed with respect to (22. λ = +1)) (P.48).  + ei p·x  ψM = e−i p·x  ∗   E−σ · p   a a iσ2 m b b∗ (P. (P.16) holds.20) can be written as ψM (λ = +1) = e−i p·x u( p. since σ2 is anti-symmetric.R C = ψM. it is evident that the Majorana condition (P. Then ψM of b (P.19) Inserting (P. In the form (P. Copyright 2004 IOP Publishing Ltd . such that σ · pφ+ = | p|φ+ as in (20.25) then each term in (P. we would expect to derive (P.12) and (P. But consider the mass term: ¯ m ψM ψM = m(φ † φ T (−iσ2 )) 0 1 1 0 φ iσ2 φ ∗ (P. − iσ2 m     b b∗ .21) using (20. A problem arises. Mimicking what we learned in the Dirac case.8) and (P.

Thus.5.29) (P.6. iγ2 u ∗ ( p.21)): ˆ ψM (x) = d3 k 1 √ (2π)3 2E cλ (k)u(k.28) λ Let us write (with a slight abuse of notation) u= as usual.26) λ ˆ† clearly showing the four degres of freedom—namely via the operators cλ and dλ .27) T . Consider. iγ2 v ∗ ∼ u.27) by cλ (compare (P. similarly. the expansion of the real scalar field in (5. the charge-conjugate field transforms in the same way as the original field. iγ2 v ∗ ( p. ˆ which respectively destroy particles with λ = ±1 and create anti-particles with λ = ±1. in ψD C . as claimed in section 22. Then iγ2 u ∗ = −iσ2 χ ∗ iσ2 φ ∗ φ χ ∼ φ χ ∼u (P. (P. a ˆ ˆ and b† .32) λ Copyright 2004 IOP Publishing Ltd .16) of a complex scalar field.30) where ‘∼’ means ‘transforms in the same way as’—a result which follows from the work of section P. as required.116) has only one type of operator.70) shows more: that. λ)eik·x ˆ (P. λ = +1) = v( p. the field ˆ ˆ ˆ† ˆ† obtained by replacing dλ in (P. λ)e−ik·x + dλ (k)v(k. λ = +1) is actually equal to u( p. λ))eik·x ˆ ˆ† (P. ˆ Apart from the spinor factors. (P.31) These results show that.1. with our conventions. a and its Hermitian conjugate a †. The charge-conjugate field is ˆ ψD C (x) = d3 k 1 √ (2π)3 2 E ˆ cλ (iγ2 u ∗ )eik·x + dλ (iγ2 v ∗ )e−ik·x . (20. Similarly. λ = +1). In fact. (P. with the property (20. In contrast.P. therefore. λ = +1) ˆ and.2 Majorana quantum fields ˆ The field-theoretic charge-conjugation operator C was introduced in section 20.71): ¯ ˆ ˆˆ ˆ ˆ ˆ ψC ≡ Cψ C−1 = iγ2 ψ †T = iγ2 γ0 ψ The standard normal mode expansion of a Dirac field is ˆ ψD (x) = d3 k 1 √ (2π)3 2 E ˆ† cλ (k)u(k. λ)e−ik·x + cλ (k)(iγ2 u ∗ (k. which also has two distinct kinds of mode operator. ˆ† (P. the operators ˆ c and d are interchanged.27) is analogous to the expansion (7.

32) destroy and create just one kind of particle. But so are other types of mass term. which coresponds to a CP-conserving theory in this sector. D (P.35) (P.39) are approximately given by m1 ≈ mR Copyright 2004 IOP Publishing Ltd m 2 ≈ −m 2 /m R .37) (P. by finding the matrix which diagonalizes M.L + ψ M.L C ˆ ψM.34) These may all be combined into the most general Lorentz-invariant mass term (for one neutrino flavour) ¯ ˆ ˆ ˆ ˆ − 1 NL C M NL + N L M NL C 2 where ˆ NL = ˆ ψM. for example.39) We shall assume for simplicity that the parameters m L .R C ψ ˆ NL C = ˆ ψM. m R and m D are all real.R + ψM.L ψM.38) and the matrix M is given by M= mL mD mD mR .L + ψM.R and also two ‘Majorana’ mass terms. Reverting to the L–R notation.116).R C ψM. The mass term considered at the end of the previous section is now clearly possible in this quantum field formalism.L C ψM.R C . one of the form ˆ ˆ ˆ ˆ − 1 m L ψM.33) ˆ ˆ† and the Majorana condition is satisfied for the field ψM . One case that has m R .40) .L ˆ M.L C 2 as in (22. we may have a ‘Dirac’ mass of the form ¯ ¯ ˆ ˆ ˆ ˆ −m D ψ M.R ψM. 2 (P. (P. Then it is easy to see that ˆ ˆ ˆ ˆ CψM C−1 = ψM (P. which has two possible spin states. taking the fermionic operators to anticommute as usual. The theory of (one-flavour) neutrino mixing may now be developed. and also ˆ ˆ ˆ ˆ − 1 m R ψM.R ψM. A clear discussion is given in Bilenky (2000).where we have used v = iγ2 u ∗ . and hence obtaining the fields in the ‘mass’ basis as linear combinations of those in the ‘weak interaction (L–R)’ basis.R (P. Then the attracted considerable attention is that in which m L = 0 and m D eigenvalues of (P.L ψM. The operators cˆλ and cλ in (P.36) (P.

one eigenvalue is (by assumption) very large compared to m D and one is very much smaller.40) is very much lighter than that—as.122). The reader may consult Bilenky (2000). The condition m L = 0 ensures that the leptonnumber-violating term (P. or Kayser et al (1989). This is the well-known ‘see-saw’ mechanism of GellMann et al (1979). 1981). appears to be true for the neutrinos. mixing. for the devlopment of the theory of neutrino masses. in fact. Copyright 2004 IOP Publishing Ltd . and oscillations.36) is characterized by a large mass scale m R . If. for example. which would then imply that m 2 of (P. Yanagida (1979) and Mohapatra and Senjanovic (1980. we recover an estimate for m 2 which is similar to that in (22. It may be natural to assume that m D is a ‘typical’ quark or lepton mass term. The experimental status is reviewed by Kayser in Hagiwara et al (2002). Thus. m R ∼ 1016 Gev.where the apparently troubling minus sign in m 2 can be absorbed into the mixing parameters. of course.

the factor ∗ µ (k .1. the spinor 2 u( p .5) Copyright 2004 IOP Publishing Ltd .APPENDIX Q FEYNMAN RULES FOR TREE GRAPHS IN QCD AND THE ELECTROWEAK THEORY Q. (Q.4) Q. external gluons also have a ‘colour polarization’ vector a c (c = 1.3) and for gluons leaving the graph. .1 particles leaving the graph.1 External particles Quarks The SU(3) colour degree of freedom is not written explicitly: the spinors have 3 (colour) × 4 (Dirac) components. =i 2 p − m2 p / −m (Q. For each fermion or anti-fermion line entering the graph include the spinor u( p.2 Propagators Quark = p i / +m . Gluons Besides the spin-1 polarization vector. .1.1 QCD Q. 2. 8) specifying the particular colour state involved. λ)a c or v( p . s ) ¯ as for QED. . For each gluon line entering the graph include the factor µ (k. . s) (Q.2) (Q. s) or v( p.1) and for spin. s ) ¯ (Q. λ )a c∗ .

Q. τ ). νµ .7) Here a and b run over the eight colour indices 1. k2 (Q.6) in which no unphysical particles appear. b. µ.1.3).5 and 19.Gluon i kµkν −g µν + (1 − ξ ) 2 δ ab (Q. . . ντ ). These U-gauge rules are given here for the leptons l = (e. Q.2 The electroweak theory For tree-graph calculations. 2. t. and for the t3 = −1/2 CKM2 mixed quarks denoted by f where f ≡ d.5. The mixing matrix Vf f is discussed in section 22. . Calculations are usually performed in Lorentz or Feynman gauge with ξ = 1 and gluon propagator equal to = =i (−g µν )δ ab .3 Vertices − gs f abc [gµν (k1 − k2 )λ + gνλ(k2 − k3 )µ + gλµ (k3 − k1 )ν ] 2 − igs [ f abe f cd e (gµλ gνρ − gµρ gνλ ) + f ad e f bce (gµν gλρ − gµλ gνρ ) + f ace f dbe (gµρ gνλ − gµν gλρ )] It is important to remember that the rules given here are only adequate for tree-diagram calculations in QCD (see section 13. s.7. .1. Copyright 2004 IOP Publishing Ltd . c. for the t3 = + 1 quarks denoted by f. 8.6) k2 k for a general ξ gauge. it is convenient to use the U-gauge Feynman rules (sections 19. νl = (νe . where f ≡ u.

the spinor 2 u( p .13) p i / +m .9) (Q.1 particles leaving the graph. Q.2 Propagators Leptons and quarks = Vector bosons (U gauge) = i (−gµν + kµ kν /m 2 ) V 2 − MV (Q. s ). λ ). see section 22. s) (Q.11) Q.8) and for spin.1 External particles Leptons and quarks For each fermion or anti-fermion line entering the graph include the spinor u( p. (Q. Higgs particle = i − m2 H (Q.14) p2 Copyright 2004 IOP Publishing Ltd .10) and for vector bosons leaving the graph.12) k2 where ‘V’ stands for either ‘W’ (the W-boson) or ‘Z’ (the Z0 ). the factor ∗ µ (k . s) or v( p. =i 2 p − m2 p / −m (Q.2.7.2.Note that for simplicity we do not include neutrino flavour mixing.2. ¯ (Q. s ) ¯ Vector bosons For each vector boson line entering the graph include the factor µ (k. λ) or v( p .

16) .15) (Q. Copyright 2004 IOP Publishing Ltd f f f (Q.2.3 Vertices Charged-current weak interactions Leptons g 1 − γ5 − i √ γµ 2 2 Quarks g 1 − γ5 Vf f − i √ γµ 2 2 Neutral-current weak interactions (no neutrino mixing) Fermions −ig f 1 − γ5 f 1 + γ5 + cR γ µ cL cos θW 2 2 where cL = t3 − sin2 θW Q f cR = − sin2 θW Q f and f stands for any fermion.Q.

Vector boson couplings (a) Trilinear couplings: γ W+ W− vertex ie[gνλ(k1 − k2 )µ + gλµ (k2 − kγ )ν + gµν (kγ − k1 )λ ] Z0 W+ W− vertex ig cos θW [gνλ (k1 − k2 )µ + gλµ (k2 − k3 )ν + gµν (k3 − k1 )λ ] (b) Quadrilinear couplings: − ie2 (2gαβ gµν − gαµ gβν − gαν gβµ ) − ieg cos θW (2gαβ gµν − gαµ gβν − gαν gβµ ) Copyright 2004 IOP Publishing Ltd .

− ig 2 cos2 θW (2gαβ gµν − gαµ gβν − gαν gβµ ) ig 2 (2gµα gνβ − gµβ gαν − gµν gαβ ) Higgs couplings (a) Trilinear couplings: HW+ W− vertex ig MW gνλ HZ0 Z0 vertex ig MZ gνλ cos θW Copyright 2004 IOP Publishing Ltd .

Fermion–Yukawa couplings (fermion mass m f ) −i Trilinear self-coupling g mf 2 MW −i (b) Quadrilinear couplings: HHW+ W− vertex 3m 2 g H 2MW ig 2 gµν 2 HHZZ vertex ig 2 gµν 2 cos2 θW Copyright 2004 IOP Publishing Ltd .

Quadrilinear self-coupling − i3m 2 g 2 H 2 4MW Copyright 2004 IOP Publishing Ltd .

74 2626 Abe K 1991 Proc. Phys. Lett. C 13 199 ——1983 Z. Rev. 177 2426 ——1970 Lectures on Elementary Particles and Quantum Field Theory (Proceedings of the Brandeis Summer Institute) vol 1. Phys. ed S Deser et al (Boston. Phys. Phys. Lett. Phys. 14 1051 ——1969 Phys. C 28 321 Allton C R et al 1995 Nucl. Phys. D 52 4784 ——1995b Phys. 74 2422 ——1995b Phys. Phys. D 66 032007 Abramovicz H et al 1982a Z. 38 403 Allasia D et al 1984 Phys. Rev. C 17 283 Abreu P et al 1990 Phys. Phys. C: Solid State Phys. Rev. 74 2632 Abbiendi G et al 2000 Eur. B 51 6531 Akhundov A A et al 1986 Nucl. Phys. D 65 054502 Alper B et al 1973 Phys. Lett. Lett. Rev. B 242 536 Acciari M et al 2000 Phys. B 476 40 Adler S L 1963 Phys. on High Energy Physics ed K K Phua and Y Yamaguchi (Singapore: World Scientific) p 33 Abe K et al 2000 Phys. Lett. B 135 231 ——1985 Z. B 157 461 Copyright 2004 IOP Publishing Ltd . Rev. B 235 389 Alitti S et al 1992 Phys. B 146 544(E) ——1979 Nucl. Lett. C 13 553 Abbott B et al 2000 Phys. Lett. 25th Int. Rev. B 126 298 Altarelli G et al 1978a Nucl. Lett. Rev. Rev. C 12 289 ——1982b Z. Conf. Phys. Rev. Phys. Rev. B 437 641 ——2002 UKQCD collaboration Phys. D 64 052001 Aitchison I J R et al 1995 Phys. J. Rev. 143 1144 ——1965 Phys. Lett. 73 225 ——1995a Phys. 182 1517 Affolder T et al 2001 Phys. Lett. Lett. Rev. Rev. Lett. 84 5945 ——2002 Belle collaboration Phys. D 50 2966 ——1994b Phys. B 44 521 Altarelli G 1982 Phys. 84 222 Abe F et al 1994a CDF collaboration Phys. Phys. B 276 1 Akrawy M Z et al 1990 OPAL collaboration Phys. 81 1 Altarelli G and Parisi G 1977 Nucl. B 143 521 ——1978b Nucl.REFERENCES Abachi S et al 1995a D0 collaboration Phys. B 276 354 Allaby J et al 1988 J. Lett. Rev. MA: MIT) Adler S L and Bardeen W A 1969 Phys. Rev. Rev. Rep.

Rev. 59 2405 Arnison G et al 1983a Phys. C 30 341 Appelquist T and Chanowitz M S 1987 Phys. B 122 103 ——1983b Phys. 52 54 Bludman S A 1958 Nuovo Cimento 9 443 Boas M L 1983 Mathematical Methods in the Physical Sciences (New York: Wiley) Copyright 2004 IOP Publishing Ltd . Mod. D 46 853 Bernstein J 1974 Rev. Lett. Lett. B 295 159 Anderson P W 1963 Phys. Rev. Rev. B 103 124 Belavin A A et al 1975 Phys. B 166 484 Aubert B et al 2002 BaBar collaboration Phys. Rev. Lett. D 13 1043 Banner M et al 1982 Phys. 130 439 Antoniadis I 2002 2001 European School of High Energy Physics ed N Ellis and J MarchRussell CERN 2002-002 (Geneva) pp 301ff Aoki S et al 2000 CP-PACS collaboration Phys.——1989 Z. Phys. B 118 203 ——1983 Phys. Lett. Lett. Rev. Phys. Bjorken J D and Kogut J B 1971 Phys. Lett. B 136 294 ——1985 Phys. B 122 465 Berends F A et al 1981 Phys. 11 255 Blatt J M 1964 Theory of Superconductivity (New York: Academic) Bloch F and Nordsieck A 1937 Phys. Rev. B 223 485 Berezin F A 1966 The Method of Second Quantisation (New York: Academic) Berman S M. 46 7 Bigi I I and Sanda A I 2000 CP Violation (Cambridge: Cambridge University Press) Bilenky S M 2000 in 1999 European School of High Energy Physics ed A Olchevski CERN 2000-007 (Geneva) Binney J J et al 1992 The Modern Theory of Critical Phenomena (Oxford: Clarendon) Bjorken J D 1973 Phys. Phys. Lett. at LEP-1 CERN 89-08 (Geneva) Amaudruz P et al 1992 NMC collaboration Phys. Cooper L N and Schrieffer J R 1957 Phys. Rev. D 4 3388 Bernard C W and Golterman M F 1992 Phys. Lett. 43 830 Bardeen J 1957 Nuovo Cimento 5 1766 Bardeen J. Phys. C 21 99 Beenakker W and Hollik W 1988 Z. C 40 569 Bergsma F et al 1983 Phys. Lett. Rev. Lett. B 129 130 ——1984 Phys. Lett. B 158 494 ——1986 Phys. B 126 398 ——1984 Phys. B 144 283 Bahcall J 1989 Neutrino Astrophysics (Cambridge: Cambridge University Press) Bailin D 1982 Weak Interactions (Bristol: Adam Hilger) Banks T et al 1976 Phys. D 8 4098 Bjorken J D and Drell S D 1965 Relativistic Quantum Fields (New York: McGraw-Hill) Bjorken J D and Glashow S L 1964 Phys. B 59 85 Bell J S and Jackiw R 1969 Nuovo Cimento A 60 47 Benvenuti A C et al 1989 BCDMS collaboration Phys. B 122 476 Barber D P et al 1979 Phys. Phys. Lett. Lett. D 18 3998 Bardin D Yu et al 1989 Z. Rev. Lett. Phys. 89 201802 Bagnaia P et al 1983 Phys. Lett. Lett. Rev. Lett. C 44 493 Barger V et al 1983 Z. Lett. Rev. Lett. 108 1175 Bardeen W A et al 1978 Phys. Rev. 84 238 Appel J A et al 1986 Z.

37 383 Collins P D B and Martin A D 1984 Hadron Interactions (Bristol: Adam Hilger) Combridge B L et al 1977 Phys. Phys. Rev. Sci. 44 560 Chadwick J 1932 Proc. Phys. Lett. Rev. B 46 471 Cabibbo N 1963 Phys. Lett. Phys. Lett. Lett. Math. USSR 11 23 ——1958 Nuovo Cimento 7 794 Bogoliubov N N et al 1959 A New Method in the Theory of Superconductivity (New York: Consultants Bureau. XVI Int. Conf.Boehm F and Vogel P 1987 Physics of Massive Neutrinos (Cambridge: Cambridge University Press) Bogoliubov N N 1947 J. B 239 429 Commins E D and Bucksbaum P H 1983 Weak Interactions of Quarks and Leptons (Cambridge: Cambridge University Press) Consoli M et al 1989 Z. Rev. B 86 243 Budny R 1975 Phys. 97 766 Dawson S et al 1990 The Higgs Hunter’s Guide (Reading. Rev. Phys. B 85 277 Chetyrkin K G and Kuhn J H 1993 Phys. D 2 1541 Carruthers P A 1966 Introduction to Unitary Symmetry (New York: Wiley) Case K M 1957 Phys. Phys. Lett. on High Energy Physics (Chicago. A 136 692 Chanowitz M et al 1978 Phys. Rev.) Bosetti P C et al 1978 Nucl. Lett. Rev. 13 138 Coleman S 1985 Aspects of Symmetry (Cambridge: Cambridge University Press) ——1966 J. Lett. Inc. 44 1068 ——1965 High Energy Physics ed C de Witt and M Jacob (New York: Gordon and Breach) Danby G et al 1962 Phys. D 12 187 Cheng T-P and Li L-F 1984 Gauge Theory of Elementary Particle Physics (Oxford: Clarendon) Chetyrkin K G et al 1979 Phys. 104 1189 Cornwall J M et al 1974 Phys. 107 307 Caswell W E 1974 Phys. Lett. 9 36 Davis R 1955 Phys. 33 244 Celmaster W and Gonsalves R J 1980 Phys. D 10 1145 Cowan C L et al 1956 Science 124 103 Dalitz R H 1953 Phil. B 55 227 B¨ sser F W et al 1972 Proc. B 70 234 Combridge B L and Maxwell C J 1984 Nucl. Mag. Lett. B 38 519 Branco G C et al CP Violation (Oxford: Oxford University Press) Brandelik R et al 1979 Phys. Part. Nucl. Soc. Rev. 10 531 Cabibbo N et al 1979 Nucl. IL) vol 3 u (Batavia: FNAL) ——1973 Phys. Lett. Phys. Rev. B 153 402 Chen M-S and Zerwas P 1975 Phys. B 308 127 Christenson J H et al 1964 Phys. 7 787 Coleman S and Gross D J 1973 Phys. Rev. Rev. Rev. Lett. Lett. 31 851 Collins J C and Soper D E 1987 Annu. Lett. MA: Addison-Wesley) Copyright 2004 IOP Publishing Ltd . Lett. Rev. B 142 1 Bouchiat C C et al 1972 Phys. B 158 295 Callan C G 1970 Phys. B 78 285 ——1979 Nucl. R. Rev. at LEP-I ed G Altarelli et al CERN 89-08 (Geneva) Cooper L N 1956 Phys. Phys.

especially footnote 9 Fermi E 1934a Nuovo Cimento 11 1 ——1934b Z. 49 895 Gasser J and Leutwyler H 1982 Phys. Lett. Yale Theoretical Advanced Study Institute (Singapore: World Scientific) Gell-Mann M 1961 California Institute of Technology Report CTSL-20 (reprinted in GellMann and Ne’eman 1964) Gell-Mann M et al 1979 Supergravity ed D Freedman and P van Nieuwenhuizen (Amsterdam: North-Holland) p 315 Gell-Mann M and Low F E 1954 Phys. 26 697 ——1977 in Weak and Electromagnetic Interactions at High Energies ed R Balian and C H Llewellyn Smith (Amsterdam: North-Holland) p 121 Feynman R P and Gell-Mann M 1958 Phys. 109 193 Feynman R P and Hibbs A R 1965 Quantum Mechanics and Path Intergrals (New York: McGraw-Hill) Gaillard M K and Lee B W 1974 Phys. 88 161 Feynman R P 1963 Acta Phys. Landshoff P V. C 1 143 DiLella L 1985 Annu. Rev. Symp. Stirling W J and Webber B R 1996 QCD and Collider Physics (Cambridge: Cambridge University Press) Ellis S D. Lett. 16 408 Faddeev L D and Popov V N 1967 Phys. Lett. Nucl. B 333 118 Ellis R K. Polon. Rev. 95 1300 Copyright 2004 IOP Publishing Ltd . 13 321 Enz C P 1992 A Course on Many-Body Theory Applied to Solid-State Physics (World Scientific Lecture Notes in Physics 11) (Singapore: World Scientific) ——2002 No Time to be Brief (Oxford: Oxford University Press) Fabri E and Picasso L E 1966 Phys. A 133 60 Dokshitzer Yu L 1977 Sov. 1975 ed M Levy et al (New York: Plenum) pp 109ff Geer S 1986 High Energy Physics 1985. Europhysics Conf. Lett. Rev. Phys. on High Energy Physics. D 10 897 Gamow G and Teller E 1936 Phys. Lett. Rev. July 1985 ed L Nitti and G Preparata (Bari: Laterza) pp 761ff Dine M and Sapirstein J 1979 Phys. Olive D I and Polkinghorne J C 1966 The Analytic S-Matrix (Cambridge: Cambridge University Press) Eichten E et al 1980 Phys. 69 3615 Englert F and Brout R 1964 Phys. Rev. Cargese. D 21 203 Einhorn M B and Wudka J 1989 Phys. Rev. Lett. Rev. 43 668 Dirac P A M 1931 Proc. Lett. D 30 49 Eden R J. D 39 2758 Ellis J et al 1976 Nucl. Soc. Sci. R. Rev. B 111 253 ——1994 Phys.–JETP 46 641 Donoghue J Groot J G H et al 1979 Z. 35 107 ——1986 Proc. on Lattice Field Theory (LATTICE 2002) Boston. Italy. Phys. 3 527. Golowich E and Holstein B R 1992 Dynamics of the Standard Model (Cambridge: Cambridge University Press) Draper T et al 2002 Talk Presented at 20th Int. Phys. B 25 29 Feinberg G et al 1959 Phys. Phys. 87 77 Gastmans R 1975 Weak and Electromagnetic Interactions at High Energies. Rev. Kunszt Z and Soper D E 1992 Phys. Rev. Proc. Rep. Bari. Rev. Int. MA. June 2002 hep-lat/0208045 Duke D W and Owens J F 1984 Phys. Rev. Rev.

Lett. 20 1064 Giusti L 2002 plenary talk at 20th Int. Teor. Lett. Rev. 38 165 ——1991 1989 CERN-JINR School of Physics CERN 91-07 (Geneva) p 50ff Hughes R J 1980 Phys. D 25 25 Ginzburg V I and Landau L D 1950 Zh. Rev. MA hep-lat/0211009 Glashow S L 1961 Nucl. B 232 113 ——1990 Phys. D 55 7255 Hammermesh M 1962 Group Theory and its Applications to Physical Problems (Reading. 13 508 ——1966 Phys. 13 585 ——1968 Advances in Particle Physics vol 2. D 66 010001 Halzen F and Martin A D 1984 Quarks and Leptons (New York: Wiley) Hamberg R et al 1991 Nucl. Phys. J. B 259 144 Gorkov L P 1959 Zh. Phys. 109 1015 Goldstone J 1961 Nuovo Cimento 19 154 Goldstone J. Symp. Lett. Rev. Rev. Rev. B 139 1066 Hasenfratz P et al 1998 Phys. D 18 1724 Glashow S L. MA: Addison-Wesley) Han M Y and Nambu Y 1965 Phys. Lett. Lett. Fiz. B 359 343 Hambye T and Reisselmann K 1997 Phys. B 186 376 Isgur N and Wise M B 1989 Phys. Lett. Rev. Eksp. Rev. Rev. Phys. Lett. Rev. D 9 416 Ginsparg P and Wilson K G 1982 Phys. CA: Benjamin/Cummings) Georgi H et al 1978 Phys. Rev. Phys. Rev. Salam A and Weinberg S 1962 Phys. Rev. 22 579 Glashow S L et al 1978 Phys. Rev. Rev. B 237 527 Isidori G et al 2001 Nucl. Phys. Phys. B 414 785 Hasert F J et al 1973 Phys. 13 598 Gribov V N and Lipatov L N 1972 Sov. Lett. B 609 387 Itzykson C and Zuber J-B 1980 Quantum Field Theory (New York: McGraw-Hill) Copyright 2004 IOP Publishing Ltd . 145 1156 Hollik W 1990 Fortsch. Fiz. 30 1343 ——1974 Phys. D 21 102 Greenberg O W 1964 Phys. Lett.Gell-Mann M and Ne’eman 1964 The Eightfold Way (New York: Benjamin) Georgi H 1984 weak Interactions and Modern Particle Theory (Menlo Park. D 2 1285 Goldberger M L and Treiman S B 1958 Phys. ed R Cool and R E Marshak (New York: Interscience) pp 567ff Hagiwara K et al 2002 Phys. Phys. Rev. Phys. Teor. Rev. D 9 980 Guralnik G S et al 1964 Phys. 15 438 Gross D J and Llewellyn Smith C H 1969 Nucl. Rev. Rev. B 97 246 ——1981 Nucl. Phys. B 46 138 Heisenberg W 1932 Z. Iliopoulos J and Maiani L 1970 Phys. B 14 337 Gross D J and Wilczek F 1973 Phys. Eksp. B 427 125 Hasenfratz P and Niedermayer F 1994 Nucl. 127 965 Gorishny S G et al 1991 Phys. Nucl. Lett. Lett. 40 692 Georgi H and Politzer H D 1974 Phys. 95 1300 Goldhaber M et al 1958 Phys. 77 1 Higgs P W 1964 Phys. 36 1918 Gottschalk T and Sivers D 1980 Phys. on Lattice Field Theory (LATTICE 2002) Boston.

Lett. 55 1039 ——1985b Z. NJ: Princeton University Press) pp 97–254 Jacob M and Landshoff P V 1978 Phys. B 180 317 u ——1998 Phys. DELPHI. B 428 342 L¨ scher M et al 1980 Nucl. Phys. Lett. 128 885 LEP 2003 (The LEP working Group for Higgs Searches. Vid. Rev. 3 650 Kittel C 1987 Quantum Theory of Solids second revised printing (New York: Wiley) Klein O 1948 Nature 161 897 Kobayashi M and Maskawa K 1973 Prog. B 565 61 LEP Higgs 2001 (LEP Higgs Working Group) CERN-EP/2001-055 (Geneva) Levine I et al 1997 Phys. 3rd edn (Oxford: Pergamon) Langacker P (ed) 1995 Precision Tests of the Standard Electroweak Model (Singapore: World Scientific) Larin S A and Vermaseren J A M 1991 Phys. Theor. Lett. D 16 1519 Lee T D and Nauenberg M 1964 Phys. Suppl. Lett. 105 1671 ——1962 Phys. Dan. Phys. Mod. Rev. Phys. Phys. B 173 365 u Copyright 2004 IOP Publishing Ltd . Rev. Rev. Phys. Rev. Rev.-Fys. 3 127 Landau L D and Lifshitz E M 1980 Statistical Mechanics part 1. R Jackiw and D J Gross (Princeton. ALEPH. B 75 531 Jones H F 1990 Groups. Representations and Physics (Bristol: IOP Publishing) Kadanoff L P 1977 Rev. Rev. B 288 342 Kayser B et al 1989 The Physics of Massive Neutrinos (Singapore: World Scientific) Kennedy D C et al 1989 Nucl. Lett. B 259 345 ——1993 Phys. Selsk. 75 9905 Lee T D and Yang C N 1956 Phys. B 321 83 Kennedy D C and Lynn B W 1989 Nucl. Rosenbluth R and Yang C N 1949 Phys. B 133 1549 Lee T D. 35 1 Leader E and Predazzi E 1996 An Introduction to Gauge Theories and Modern Particle Physics vol 2 (Cambridge: Cambridge University Press) Lee B W et al 1977a Phys. B 303 334 Lautrup B 1967 Kon. Math. B 322 1 Kibble T W B 1967 Phys. 104 254 ——1957 Phys. Phys. Phys. L3 and OPAL collaborations) Phys. Med. Phys.Jackiw R 1972 Lectures in Current Algebra and its Applications ed S B Treiman. Phys. Rep. Lett. 66 1 Kunszt Z and Pietarinen E 1980 Nucl. Lett. Lett. Rev. Rev. C 48 285 Jarlskog C 1985a Phys. B 164 45 Landau L D 1957 Nucl. C 29 491 Jones D R T 1974 Nucl. Rev. 38 883 ——1977b Phys. Rev. D 17 2800 Kinoshita T 1962 J. Lett. Phys. 78 424 Lichtenberg D B 1970 Unitary Symmetry and Elementary Particles (New York: Academic) Llewellyn Smith C H 1973 Phys. 49 652 Kugo T and Ojima I 1979 Prog. Phys. Mat. Macroscopic theory of Superconductivity (New York: Wiley) L¨ scher M 1981 Nucl. B 46 233 London F 1950 Superfluids Vol I. 155 1554 Kim K J and Schilcher K 1978 Phys. Theor. Phys. 49 267 Kaplan D B Phys.

Rev. 30 1346 Copyright 2004 IOP Publishing Ltd . 28 247 Mandelstam S 1976 Phys. Rev. B 193 173 ——1981c Nucl. Lett. D 50 6734 ——2002 Eur. J. Rev. Lett. 125 1429 Nambu Y and Schrauner E 1962 Phys. 71 3251 ——1994 Nucl. Lett. Phys. Rev. Phys. Ges. Brussels 1933 (Paris: Gautier-Villars). B 185 20 ——1981b Nucl. Lett. 49 1171 Nielsen H B and Ninomaya M 1981a Nucl. Lett. Rev. B 195 541 Noether E 1918 Nachr. Phys. Rev. Lett. Solvay. MA: Addison-Wesley) Politzer H D 1973 Phys. B 302 62 ——1993b Phys. Rev. Rev. C 23 73 Merzbacher E 1998 Quantum Mechanics 3rd edn (New York: Wiley) Mohapatra R N et al 1968 Phys. 20 1081 Mohapatra R N and Pal P B 1991 Massive Neutrinos in Physics and Astrophysics (Singapore: World Scientific) Mohapatra R N and Senjanovic G 1980 Phys. 46 393 Perkins D H in Proc. Phys. Phys. Rev. Septi` me Conseil Phys. 72 351 Montanet L et al 1994 Phys. G¨ ttingen 171 o Okada S et al 1998 Phys. Rev. 124 246 Nambu Y and Lurie D 1962 Phys. 26 222 Nielsen N K 1981 Am. Rev. Lett. Rev. 4 380 ——1974 Phys. D 10 4262 Nambu Y and Jona-Lasinio G 1961a Phys. Phys. Rev. 122 345 ——1961b Phys. Prog. D 23 165 Mohr P J and Taylor B N 2000 Rev. Phys. Int. 81 2428 Pais A 2000 The Genius of Science (Oxford: Oxford University Press) Parry W E 1973 The Many Body Problem (Oxford: Clarendon) Pauli W 1934 Rapp. Symp.Maki Z et al 1962 Prog. Mod. Phys. Phys. e reprinted in Winter (2000) pp 7. J. CA p 571 ——2000 Introduction to High Energy Physics 4th edn (Cambridge: Cambridge University Press) Peskin M E 1997 in 1996 European School of High-Energy Physics ed N Ellis and M Neubert CERN 97-03 (Geneva) pp 49-142 Peskin M E and Schroeder D V 1995 an Introduction to Quantum Field Theory (Reading. C 23 245 Mandl F 1992 Quantum Mechanics (New York: Wiley) Marciano W J and Sirlin A 1988 Phys. Lett. B 412 574 ——1995 Nucl. 128 862 Narayanan R and Neuberger H 1993a Phys. 8 Pennington M R 1983 Rep. B 443 305 Ne’eman Y 1961 Nucl. 44 912 ——1981 Phys. on Lepton and Photon Interactions at High Energies. Rev. Rep. Theor. Phys. Stanford. 61 1815 Marshak R E et al 1969 Theory of Weak Interactions in Particle Physics (New York: Wiley) Martin A D et al 1994 Phys. D 50 1173 Montvay I and M¨ nster G 1994 Quantum Fields on a Lattice (Cambridge: Cambridge u University Press) Nambu Y 1960 Phys. Phys. Wiss. Rev.

–JETP 26 989) Prescott C Y et al 1978 Phys. B 411 839 Staff of the CERN pp project 1991 Phys. 82 664 ——1957 Ann. transl. Rev. Lett. Phys. D 19 2619 Sutherland D G 1967 Nucl.. Sov. and Math. Acta 26 499 Sudarshan E C G and Marshak R E 1958 Phys. Rev. Rev. 125 397 Shaevitz et al 1995 CCFR collaboration Nucl. Phys. Rev. Phys. Suppl. Fiz. NY 2 407 ——1962 Phys. Rev. Phys. Phys. Phys. Int. 38 188 Sharpe S R 1992 Phys. 10 153 (Engl. B 82 272 Ross D A and Veltman M 1975 Nucl. Rev. B Proc. Phys. Lett. D 50 4491 Steinberger J 1949 Phys. 10 99) Sikivie P et al 1980 Nucl.Pontecorvo B 1947 Phys. 1976 (Braunschweig: Vieweg) p 683 Ryder L H 1996 Quantum Field Theory 2nd edn (Cambridge: Cambridge University Press) Sakurai J J 1958 Nuovo Cimento 7 649 ——1960 Ann. Rev. B 2 433 Symanzik K 1970 Commun. Mat. D 29 89 Sommer R 1994 Nucl. Mod. Lett. 76 1180 Sterman G and Weinberg S 1977 Phys. Rev. B 95 135 Rubbia C et al 1977 Proc. Rev. 66 560 Schiff L I 1968 Quantum Mechanics 3rd edn (New York: McGraw-Hill) Schrieffer J R 1964 Theory of Superconductivity (New York: Benjamin) Schutz B F 1988 A First Course in General Relativity (Cambridge: Cambridge University Press) Schwinger J 1951 Phys. B 107 306 ¯ Stange A et al 1994a Phys.. Phys. Rev. D 49 1354 ——1994b Phys. 13 168 Samuel M A and Surguladze L R 1991 Phys. D 16 3031 ——1979 Phys. Rev. Lett. Rev. Phys. 109 1860 Susskind L 1977 Phys. Theor. 39 1436 Stueckelberg E C G and Peterman A 1953 Helv. D 22 971 ——1984 Phys. B 77 347 Puppi G 1948 Nuovo Cimento 5 505 Quigg C 1977 Rev. Aachen. 18 227 Copyright 2004 IOP Publishing Ltd . Math. Lett. Lett. Rev. Theor. D 46 3146 Shaw R 1995 The problem of particle types and other contributions to the theory of elementary particles PhD Thesis University of Cambridge Slavnov A A 1972 Teor. Rev. Rev. B 173 189 Sirlin A 1980 Phys. transl. Phys. 53 1717 (Engl. 72 246 ——1948 Chalk River Laboratory Report PD-205 ——1967 Zh. 37 315 Renton P 1990 Electroweak Interactions (Cambridge: Cambridge University Press) Richardson J L 1979 Phys. Neutrino Conf. 49 297 Reines F and Cowan C 1956 Nature 178 446 Reines F. Phys. Eksp. Gurr H and Sobel H 1976 Phys. Phys. NY 11 1 Salam A 1957 Nuovo Cimento 5 299 ——1968 Elementary Particle Physics ed N Svartholm (Stockholm: Almqvist and Wiksells) Salam A and Ward J C 1964 Phys. Phys. Lett..

B 248 141 van der Bij J J and Veltman M 1984 Nucl. Rev. Lett. B 231 205 van der Neerven W L and Zijlstra E B 1992 Nucl.Tarasov O V et al 1980 Phys. especially footnote 8 ——1975 Phys. Phys. Phys. B 37 195 ——1976a Phys. D 8 605. B 4 3184 ——1974 Phys. 14 1047 Williams E J 1934 Phys. Rep 12C 75 Winter K 2000 Neutrino Physics 2nd edn (Cambridge: Cambridge University Press) Copyright 2004 IOP Publishing Ltd . B 382 11 van Ritbergen T and Stuart R G 1999 Phys. B 4 3174 ——1971b Phys. Phys. 179 1499 ——1971a Phys. Phys. Rev. Rev. D 11 3583 ——1979 Phys. 82 488 Veltman M 1967 Proc. Rev. Rev. Rev. B 21 288 ——1977 Acta Phys. B 33 173 ——1971b Nucl. Lett. 45 729 Wilson K G 1969 Phys. B 8 475 von Weisz¨ cker C F 1934 Z. 88 612 a Weerts H 1994 D0 collaboration Proc. Cargese Summer Institute 1979 ed G ’t Hooft et al (New York: Plenum) ——1986 Phys. Rev. Phys. 21 153 Tournefier E 2001 LEP collaborations Proc. European Physical Society Int. B 33 436 ——1976 Gauge Theories of Weak Interactions (Cambridge: Cambridge University Press) ’t Hooft G 1971a Nucl. Phys. Polon. Conf. Rev. Phys. Rev. Lett. Lett. R. D 10 2445 ——1975 New Phenomena in Subnuclear Physics. B 7 637 ——1970 Nucl. B 35 167 ——1971c Phys. Soc. ed A Zichichi (Bologna: Editrice Composition) p 1225 ——1980 Recent Developments in Gauge Theories. Proc. B 44 189 Tiomno J and Wheeler J A 1949 Rev. 36th Rencontre de Moriond: Electroweak Interactions and Unified Theories (Les Arcs) Valatin J G 1958 Nuovo Cimento 7 843 van der Bij J J 1984 Nucl. School on Subnuclear Physics ‘Ettore Majorana’ ed A Zichichi (New York: Plenum) Wilson K G and Kogut J 1974 Phys. Rev. 142 357 ’t Hooft G and Veltman M 1972 Nucl. 19 1264 ——1973 Phys. B 93 429 Tavkhelidze A 1965 Seminar on High Energy Physics and Elementary Particles (Vienna: IAEA) p 763 Taylor J C 1971 Nucl. A 301 107 ——1968 Nucl. Rev. Rev. Phys. Phys. 1975 Int. Proc. Phys. B 5 4529 Weinberg S 1966 Phys. Mod. Rev. Rep. Lett. Rev. D 14 3432 ——1976b High Energy Physics. D 19 1277 ——1996 The Quantum Theory of Fields Vol II Modern Applications (Cambridge: Cambridge University Press) Weisberger W 1965 Phys. Lett. Phys. 31st Rencontre de Moriond: QCD and High-energy Hadronic Interactions (Les Arcs) FermiLab-Conf-96-132-E Wegner F 1972 Phys. 17 616 ——1967 Phys.

Rev. 111 1255 Copyright 2004 IOP Publishing Ltd . Lett.Wolfenstein L 1983 Phys. Rev. Rev. Rev. 51 1945 Wu C S et al 1957 Phys. 96 191 Yosida K 1958 Phys. 105 1413 Yanagida T 1979 Proc. Workshop on Unified Theory and Baryon Number in the Universe ed O Sawada and A Sugamoto (Tsukuba: KEK) Yang C N and Mills R L 1954 Phys.