Federica Vian Quantum Field Theories with Symmetries 9 9 9 in the 1 y a Wilsonian Exact Renormalization Group M 9 1 1 v 2 Ph.D. Thesis 4 1 5 0 9 9 / h t - p e h : v i X r a Universit`a degli Studi di Parma PARMA - 1998 2 Contents Introduction 3 1 Wilson Renormalization Group 9 1.1 Wilson effective action . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.2 The RG flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2 The massless scalar case 17 2.1 The RG flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.1.1 Physical couplings and boundary conditions . . . . . . . . . . . . 20 2.1.2 Loop expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.1.3 One-loop vertices . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2 Perturbative renormalizability . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3 Infrared behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3 The Quantum Action Principle 33 3.1 The Quantum Action Principle . . . . . . . . . . . . . . . . . . . . . . . 34 3.2 Effective Slavnov-Taylor identities . . . . . . . . . . . . . . . . . . . . . . 35 3.3 Perturbative solution of ∆ = 0 . . . . . . . . . . . . . . . . . . . . . . 38 Γ 3.3.1 Solution of ∆ = 0 at Λ = Λ . . . . . . . . . . . . . . . . . . . 40 Γ 0 4 The breaking of dilatation invariance: the Callan-Symanzik equation 43 4.1 Dilatation invariance in the RG . . . . . . . . . . . . . . . . . . . . . . . 45 4.2 The one-loop beta function . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.3 Computation of γ(1) and σ(1)(Λ ) via ∆ˆ(1)(Λ ) . . . . . . . . . . . . . 49 2 0 2 0 5 SU(N) Yang-Mills theory 51 5.1 RG flow for SU(N) Yang-Mills theory . . . . . . . . . . . . . . . . . . . 52 5.1.1 Relevant parameters . . . . . . . . . . . . . . . . . . . . . . . . . 54 2 CONTENTS 5.1.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . 56 5.2 Effective ST identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 5.3 Solution of the fine-tuning at Λ = Λ . . . . . . . . . . . . . . . . . . . . 57 0 5.3.1 Solution of the fine-tuning at the first loop . . . . . . . . . . . . . 60 5.3.2 Vertices of ∆ˆ with more than two fields . . . . . . . . . . . . . . 64 Γ 5.4 Comparison with the fine-tuning at Λ = 0 . . . . . . . . . . . . . . . . . 67 6 Chiral gauge theories and anomalies 69 6.1 Renormalization group flow and effective action . . . . . . . . . . . . . . 70 6.1.1 Boundary conditions: physical parameters and symmetry . . . . . 72 6.2 Solution of ∆ = 0 at Λ = Λ . . . . . . . . . . . . . . . . . . . . . . . 73 Γ 0 (1) 6.2.1 Explicit solution of ∆ (Λ ) = 0 . . . . . . . . . . . . . . . . . 74 Γ 0 6.3 The ABJ anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 7 The Wess-Zumino model 81 7.1 The RG flow for the Wess-Zumino model . . . . . . . . . . . . . . . . . . 82 7.1.1 Evolution equation . . . . . . . . . . . . . . . . . . . . . . . . . . 83 7.2 Relevant couplings and boundary conditions . . . . . . . . . . . . . . . . 85 7.2.1 Loop expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 8 Supersymmetric Gauge Theories and Gauge Anomalies 89 8.1 N = 1 Super Yang-Mills . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 8.1.1 Matter fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 8.1.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . 92 8.2 Effective ST identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 8.2.1 Perturbative solution of ∆ = 0 . . . . . . . . . . . . . . . . . . . 96 Γ 8.3 Gauge anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 A Supersymmetric conventions 103 Bibliography 106 Introduction The main goal of renormalization from a traditional point of view is to determine when and how the cancellation of ultraviolet (UV) divergences in quantum field theory occurs. Such a cancellation is essential if a theory has to yield quantitative physical predictions. What is not obvious is how the quantum fluctuations associated with extremely short distances (i.e. very high momenta) can be so harmless to affect a theory only through the values of a few of its parameters, typically the bare masses and coupling constants or the counterterms in renormalized perturbation theory. Anotherpictureforrenormalizationcanbeconceived, andtheideaisduetoK.Wilson [1]. Hesuggestedthatalloftheparametersofarenormalizablefieldtheorycanbethought of as scale dependent objects and such a scale dependence is described by differential equations, the so-called renormalization group (RG) equations or flow equations. The RG method is based on the functional integral approach to field theory and the origin of the ultraviolet (UV) divergences is perused by isolating in the functional integral the short-distance degrees of freedom of the field. Actually in the generating functional Z[J] the basic integration variables are the Fourier components φ(k) of the field, namely Z[J] is expressed by Z[J] = φei [L+Jφ] = dφ(k) ei [L+Jφ]. D Z R k Z ! R Y In order to cure the ultraviolet divergences, a sharp UV cutoff M is imposed. This means we integrate only over the fields φ(k) with k M and set φ(k) = 0 for k M so that | | ≤ | | ≥ in therealspace thefields aredefined ona lattice ofspacing 2π/M. According toWilson, the fundamental fields are replaced by their averages over a certain space-time volume (blockspin transformations) and thus defined on a coarser lattice. By this averaging process small scale fluctuations which correspond to high frequencies are eliminated. Rather than in the discrete Wilson RG, we are interested in the continuous Wilsonian RG[2]-[4]. Theidea behindit isverysimilar tothatofthediscrete case: inthegenerating 4 CONTENTS functional (partition function in the Euclidean) we do not integrate over all momenta in one go, but we first integrate out modes between a cutoff scale Λ (UV) and a very 0 much lower energy scale Λ. What is left over —integration between Λ and zero— may still be expressed as a generating functional but the bare action is replaced by a very complicated effective action, S , containing an infinite series of non-local terms. This eff is naturally interpreted as the high frequency modes of the fields generating effective couplings for the low-energy modes. However, we expect the behaviour at small scales to be controlled only by a finite number of parameters, i.e. the “relevant parameters” (with non-negative mass dimension). Deviations from locality will be of order Λ/Λ . Thus, 0 when the relevant parameters have been fixed at low energies, the dependence on Λ will 0 be given by powers of Λ/Λ at any order in perturbation theory. Moreover, by requiring 0 the physical Green functions to be independent of the cutoff Λ, it follows the functional S obeys an evolution equation. Hence, the evolution equation with a suitable set eff of boundary conditions —which encode both renormalizability and the renormalization conditions— can be thought of as an alternative definition of a theory. From what we have seen so far we should be driven to view the RG formulation as a natural setting for the analysis of effective theories [5]. Effective theories arevery popular nowadays: Chiral Perturbation Theory [6], Heavy Quark Effective Theory [7], low energy N = 2 Super Yang Mills [8] are just a few examples. Even though the dream of modern physics is to achieve a simple understanding of all the observed phenomena in terms of some fundamental dynamics (unification), assuming a theory of everything appeared at some point, the description of nature at all physical scales would have little to do with a quantitative analysis at the most elementary level. Therefore, in order to study a particular physical system in a huge surrounding world, the key issue is to identify and pick up the most appropriate variables. Usually, a physics problem involves widely separated energy scales. The basic idea is to identify those parameters which are very large (small) compared with the energy scale of the system and to set them to infinity (zero). A sensible description of the system would obviously consider the corrections induced by the neglected energy scales as small perturbations. Effective field theories are the appropriate theoretical tool to examine low-energy physics, where low is referred to some energy scale M (Λ in the 0 RG).Only the relevant degrees of freedom, i.e. the states with k << M are kept, whereas the heavier excitations with k >> M are integrated out from the action. The by- product of such integration is a bunch of non-renormalizable interactions among the light states, which can be expanded in powers of E/M, E being the energy. Thus the CONTENTS 5 information on the heavier degrees of freedom is stored in the couplings of the low-energy Lagrangian. Althoughaneffectivefieldtheorycontainsaninfinitenumber ofinteractions, renormalizability can be trusted since, at a given order in the expansion in E/M (Λ/Λ 0 in the RG), the low-energy theory is specified by a finite number of couplings. We naturally expects the effective theory keeps track of the symmetries of the funda- mental theory. Global symmetries, such as Lorentz invariance, isotopic spin invariance and so on, are automatically maintained in the RG method. It is certainly not so for gauge symmetries. In fact the division of momenta into large or small (according to some scale Λ) —which is fundamental in the RG approach— is not preserved by gauge transformations, since in the momentum space the field is mapped into a convolution with the element of the gauge group. We are forced to conclude that the symmetry of the fundamental theory is lost at the effective level. Nevertheless, a remainder of the original invariance survives in the form of an effective symmetry which constraints the flow of all the couplings of the theory at the scale Λ. Unfortunately, the task of solving the relations among the couplings coming from those constraints is impossible to carry out, due to non-linearity. Therefore one is left with two options: either work in non- perturbative field theory by means of an analytic approximation or solve equations in the perturbative regime. In the former case we have to face the unpleasant aspect that there is no known truncation consistent with gauge invariance and the best one can do is to give a numerical estimate of the symmetry breaking term by using effective Ward identities [9]. In this thesis we will choose the latter option and the implementation of symmetries in perturbation theory will be extensively treated. Eventhoughthetopicsofthethesiswillbediscussedattheperturbativelevel,wemust recall for completeness that the RG formulation is in principle non perturbative. Clearly analytic approximation methods must be employed in non-perturbative quantum field theory where there are no small parameters to expand in. In this direction much progress has been made. Let us just mention the applications to chiral symmetry breaking, phase transitions, finite temperature, large N limit and to many other sectors. For a review see [10]. Two major problems affect non-perturbative RG. Of the first of these, that any known truncation violates gauge invariance, we have already said. The second problem is the lack of a recipe to evaluate errors in a certain approximation scheme. We now present the outline of the thesis. In the first chapter we will introduce the Wilsonian Exact Renormalization Group for a general theory (i.e. containing both bosons and fermions, scalars and vectors). The procedure of integrating out the modes with frequency above Λ2 and below Λ2 will be performed multiplying the quadratic part 0 6 CONTENTS of the classical action by a cutoff function which is one between Λ2 and Λ2 and rapidly 0 vanishesoutsidethisinterval. Wewillthenderive theRGflowbyrequiring thegenerating functional of the theory to be independent of the infrared cutoff Λ. As an example of how the RG method works, in chapter 2 we will apply it to the masslessscalartheory. Wewillseeindetailshowaniterativesolutionoftheflowequation, together with a set af suitable boundary conditions, provides the usual loop expansion. Furthermore we will explicitly compute the one-loop two-point and four-point vertices. For this theory we will also prove perturbative renormalizability, i.e. the existence of the Λ limit, and infrared finiteness, that is the vertex functions at non-exceptional 0 → ∞ momenta are finite order by order in perturbation theory. The third chapter will be devoted to establishing the Quantum Action Principle (QAP) in the RG and we will show that the Slavnov-Taylor (ST) identities, which com- pletely characterize the classical theory, can be directly formulated for the cutoff effective action at any Λ. Afterwards we will use these effective identities to fix the couplings in the bare action. In the fourth chapter the QAP will be exploited to analyse the breaking of dilata- tion invariance occurring in the scalar theory in the RG approach. An analogue of the Callan-Symanzik equation will be derived for the cutoff effective action and from the effective Ward identities of dilatation the one-loop beta function for such a theory will be reproduced. In the fifth chapter we will address SU(N) Yang-Mills theory. After deriving the evolution equation, we will treat the key issue of boundary conditions which, in this case, have also to ensure restoration of symmetry for the physical theory when the cutoffs are removed (in the limits Λ 0 and Λ ). We will then use the effective ST identities 0 → → ∞ to derive some of the bare couplings (fine-tuning) at the first loop. The next step will be the extension of the RG method to chiral gauge theory. This will be the subject of chapter 6. Since in the RG formulation the space-time dimension is four, there is no ambiguity in the definition of the matrix γ and in the regularized action left and right 5 fermions will not be coupled. Therefore the solution of the fine-tuning procedure we be simpler than in the standard case (global chiral symmetry is preserved) and will be explicitly performed. We will then show how the chiral anomaly can be obtained in the RG. Having gone through non-supersymmetric theories, chapter 7 will be dedicated to extending the RG formulation to supersymmetric theories. Regularization will be imple- mented in such a way that supersymmetry is preserved. Actually, it suffices to write the CONTENTS 7 classical action in terms of superfields and multiply the propagators by the same cutoff function. In components this corresponds to use the same cutoff for all fields. We will start with the Wess-Zumino model to set up the formalism and then, in chapter 8, we will approach supersymmetric gauge theories. We will solve the fine-tuning equation at the first loop and show how the gauge anomaly can be derived. Finally, the appendix contains the supersymmetric conventions. 8 CONTENTS