MATHEMATICAL INTRODUCTION TO QUANTUM ELECTRODYNAMICS
JAN DEREZINSKI´
Slides of the minicourse at IHP 21-25.06.2010
Contents
3 0 Introduction 5
1 General scattering theory 29
2 Axioms of quantum field theory 65
3 Neutral scalar bosons 89
4 Massive photons 137 5 Massless photons 175
6 Linearly perturbed quadratic Hamiltonians 209
7 Charged scalar bosons 221
8 Dirac fermions 267 Chapter 0
Introduction 5 Quantum Electrodynamics (QED) is an extremely successful theory. It depends on very few adjustable parameters – the fine structure constant and the masses of charged particles – and it can be used to compute with great precision many physical quantities, explaining a wide range of phenomena. It has a rich and complex formal structure, even though it can be developed from a very small number of postulates without any arbitrariness, The agreement of the QED computations with the experimental data is astounding. Its most spectacular successes are the computation of the anomalous magnetic moment of the electron and of the energy levels of light atoms (the Lamb shift). Its range of applicability is however much wider: one can argue that a large part of modern physics can be derived from QED. QED has serious problems. It has been noticed already by F. Dyson that the perturbation expansion computed in QED is probably divergent. Even worse: as noted by L. D. Landau, if we try to sum up this series we encounter a singularity, the so-called Landau pole. The common wisdom says that QED does not exist as a consistent satisfactory nonperturbative theory. QED is an example of a quantum field theory (QFT). It is a relatively sophisticated example of a QFT because of the gauge invariance. It is natural to ask whether QED, or more generally, other QFT’s, fit into the framework of usual quantum mechanics and of some common-sense postulates: whether they can be described on a Hilbert space, with a bounded from below Hamiltonian, satisfy the Poincar´e covariance, and the Einstein causality. In the literature two basic sets of such postulates can be found: the Wightman axioms and the Haag-Kastler axioms. They are viewed by many physicists as an irrelevant pedantry of narrow-minded mathematicians. In my opinion, this is a wrong assessment – in reality, most of these postulates express basic intuitions coming from quantum mechanics. One can view these axioms not as an unshakable foundation, but as a set guiding principles. Unfortunately, so far the only known nontrivial models of Wightman or Haag-Kastler axioms in 1+3 dimensions are free theories. Interacting quantum field theories, such as eg. the λ:φ4: theory, satisfy these axioms at best in the sense of formal power series. There is probably no hope that a model of interacting QFT will be rigorously and non-perturbatively constructed in a near future. In particular, it is commonly believed that it is impossible to construct the λ:φ4: theory satisfying the Wightman axioms (because of the Landau pole). The situation with QED is, to my understanding, even worse, since it does not even satisfy Wightman axioms on the formal level, because of the gauge invariance. According to the Wightman axioms, the full informations about a QFT can be encoded in the so-called Wightman functions. Actually, in usual physics textbooks, QFT is studied following a somewhat different philosophy. According to them, the objective of QFT is computation of various quantities related to the scattering operator (of course, as formal, probably divergent power series). These quantities are easier to describe in massive theories. In theories containing massless particles, the situation is messed-up by the so-called the infra-red problem. Basic quantities computed in QFT
massive QFT QFT possessing massless particles
N-point time-ordered N-point time-ordered Green’s functions Green’s functions (with a physical normalization); (without a physical normalization);
scattering amplitudes; ————-
scattering cross-sections; inclusive scattering cross-sections;
(complex) energy levels; (complex) energy levels. In standard modern textbooks, N-point time-ordered Green’s functions are treated as the ba- sic objects of a given QFT. Other quatities, such as the energy levels and scattering amplitudes and cross-sections can be in principle derived from Green’s functions. Scattering amplitudes (that is matrix elements of the scattering operator) do not, however, exist in the massless case (even as formal power series). Physically measurable are inclusive cross-sections and energy levels. In QED (both massive and massless), there is a problem with Green’s functions, because they depend on the gauge (both in the massless and massive case). Neither the (inclusive) scattering amplitudes nor energy levels depend on the gauge. One can name 4 major issues in QED that need a special attention:
1. the ultra-violet problem;
2. the infra-red problem;
3. the gauge-invariance;
4. bound states of many particle systems.
The ultra-violet problem is common to all kinds of interacting relativistic quantum field theories. It is the main reason why we seem to be limited to a perturbative treatment and why the renormalization is necessary. Historically, it posed the biggest obstacle in the development of QED, and was mastered only in the late 40’s, with major contributions of Schwinger, Feynman, Tomonaga and Dyson. The infra-red problem is common to all QFT’s possessing massless particles. In QED it is due to the presence of massless photons and is absent in the massive QED. It appears already in non-relativistic quantum mechanics – in scattering theory with Coulomb forces. These forces are long-range, which makes the usual definition of scattering operators impossible. Its another manifestation is the appearance of inequivalent representations of canonical commutation re- lations when we consider scattering of photons against a classical current that has a different direction in the past and in the future. Thus, even in these toy non-relativistic situations the usual scattering operator is ill-defined. Therefore, it is not surprising that (much bigger) problems are present in the full relativistic case. One can cope with the infra-red problem by approximating massless QED with the massive one and restricting computations only to inclusive cross-sections justified by an imperfect resolution of the measuring device. The expression gauge invariance has in the context of QED several meanings.
1. The most common, discussed in classical electrodynamics, is the fact that if we add to a 4-potential solving the Maxwell equation a total derivative, then it still solves the Maxwell equations. Of course, this no longer holds for the Proca equations – the massive gener- alization of the Maxwell equations. Therefore, it is often stressed that gauge-invariance implies that the photons are massless.
2. There exists another meaning of the gauge-invariance: we can multiply charged fields by a space-time dependent phase factor and compensate it by changing the external potentials. 1. and 2. go together in the full Lagrangian of QED, which is invariant with respect to these two gauge transformations applied simultaneously. 3. In QED one often uses the term “gauge invariance” in yet another meaning: we have the freedom of the choice of the (free) photon propagator. This meaning applies both to the massive and massless QED. Some of these propagators are distinguished. In particular, the Feynman propagator plays a special role in the renormalization in the ultraviolet regime.
There are several distinct approaches to treat the problem of the gauge invariance QED, all of them equivalent what concerns the physical quantities. These approaches are quite different in the massive and massless case. Nevertheless, all physical quantities depend continuously on the mass, down to zero. Bound states energies are defined as the positions of singularities of Green’s functions. They usually have a nonzero imaginary part. The understanding of bound states seems to be one of the major problems of QFT. It is in fact dificult to study bound states by purely perturbative methods, since the basic quantities that we compute involve elementary particles as asymptotic states. The same problem would appear if we wanted to compute the scattering matrix of an N-body Schr¨odingeroperator in a purely perturbative way. QED is a quantized theory that describes charged particles interacting with photons. Charged particles are either fermions described by the Dirac equation or bosons described by the Klein- Gordon equation. From the point of view of applications, the case of charged fermions is much more important, nevertheless, from the theoretical point of view, charged bosons can be considered on the same footing as charged fermions.
Photons are massless vector bosons, described by the Maxwell equation. The zero mass of photons leads to the infra-red divergencies, which are difficult from the theoretical point of view. The usual way to cope with them is to assume that photons are massive, described by the Proca equation. Then we can treat the (physical) massless QED as the limit of the (non- physical) massive one. Therefore, it is useful to discuss in parallel the massless and massive QED. In full QED both photons and charged particles are quantized. It is however useful to consider the formalism, where only one kind of particles is quantized. In fact, the following three theories can be viewed as the main pillars of the full QED:
1. charged scalar bosons interacting with an external potential;
2. Dirac fermions interacting with an external potential;
3. (massless or massive) photons interacting with an external current.
All the above three theories are (or at least can be) well understood in the non-perturbative sense. They have a well-dynamics on a Hilbert space, or at least in an appropriate C∗-algebra. They can be used to compute quite a number of physically relevant quantities and to explain tricky theoretical questions. One can distinguish three, increasingly complex, basic varieties of the above theories: (i) theory without external potentials, resp. currents; (ii) theory with external potentials, resp. currents in the vacuum representation; (iii) theory with external potentials, resp. currents in the interaction (Furry) representation. Only the first theory is Poincar´einvariant. All of them are in some sense very simple-minded. Their Hamiltonians are quadratic, and hence their equations of motions are linear. After solving the classical equations, we can say everything about their quantizations. Full QED is much more difficult. It involves quantizing simultaneously charged particles and photons. One way to formulate full QED goes as follows. First we formulate the theory of charged particles in an external potential and the theory of photons in an external current as perturbative theories for many-point Green’s functions and scattering amplitudes. An elegant and convenient way to do this a representations of terms in the perturbation expansions as diagrams. Rules to evaluate them are often called the Feynman rules. Then we put the Feynman rules for charged particles and photons together obtaining the Feynman rules for the full QED. These rules (in a somewhat different language) were known already in the 30’s. If one applies them to the so-called tree diagram, one obtains physically meaningful and well defined answers (such as the Klein-Nishina formula, Bhabha formula, etc.). When one tries to evaluate diagrams involving loops, one usually encounters divergencies. Only at the end of 40’s physicists found methods that enabled them to make sense of these diagrams. As a result, the quantities of interest can be expressed in terms of formal power series. These power series, though probably divergent, involve a very small coupling constant. Therefore, for practical purposes, if we take a few first terms, they seem to behave as convergent series. To my experience, in most textbooks, QED is not presented as a consistent theory. Essen- tially all commonly available textbooks on QED present it as a collection of “cookbook recipes”, somewhat chaotic and not quite honest. The reader is supposed to excuse the inconsistencies and use his or her “physical intuition” to cope with its problems. To my understanding, in reality, properly interpreted, QED is a rigorous theory with unam- biguous rules. There are several methods of renormalizing Feynman diagrams. All of them, to my understanding, are equivalent and lead to a uniquely defined formal power series for various quantities of interest depending on very few parameters. The parameters of QED do not appear explicitly in the inital Feynman rules. They enter the theory in the so-called normalization conditions, usually imposed on the values of certain Green’s functions at some chosen points in the momentum space. There is a considerable freedom in choosing these conditions. For instance, we can demand that the 2-point Green’s functions have an appropriate singularity at the value of the physical mass (the on-shell normalization condition). However, other conditions can often be more convenient. The value of the electric charge (and hence of the fine structure constant) can be fixed by demanding that the appropriate Ward(-Takahashi) identity holds. One can distinguish at least 3 basic varieties of the full QED: (i) with no external potentials; (ii) with weak external potentials; (iii) in the interaction (Furry) picture. QED with no external potentials is fully Poincar´ecovariant. It can be viewed as the basic theory. The normalization conditions for the masses and the charge are formulated within its framework. QED with external potentials can in principle be derived as an effective approximation of the QED with no external potentials. We need to assume that some infinitely heavy charged particles generate a potential, which can be treated classically (in addition to quantized photons already present in the theory). One possibility is to keep the vacuum representation and expand the external potential in the perturbation series. Another possibility is to use the interaction (Furry) picture, where we assume that the external potential (or at least its major part) is time-invariant and its effect can be included in the propagator of charged particles. (Of course, there can be an additional time-dependent external potential treated perturbatively). QED in an external potential is very important from the point of view of practical compu- tations. The definition of the anomalous magnetic moment involves QED in a weak external potential. A natural setup for the Lamb shift is the QED in the Furry picture. To sum up, one can view QED as a family of closely related theories. Beside the full QED, this family contains the theories of one kind of particles interacting with external potentials or currents. These partial theories can be understood non-perturbatively. Even though they are in some sense “trivial”, they are actually quite interesting, both physically and mathe- matically. A number of nontrivial and often pathological problems can be partly understood within their setup, such as the infra-red problem, the infinite renormalization of the charge, non-implementability of the dynamics, gauge invariance. The full QED is much less trivial. It seems to be necessarily perturbative.
Chapter 1
General scattering theory 29
Heisenberg picture Let H be a Hilbert space with a self-adjoint operator (called the Hamiltonian) H. Let B be an operator and t ∈ R. The operator B in the Heisenberg picture at time t is defined as
itH −itH BH(t) := e Be . Wightman functions Let Φ be a normalized vector satisfying HΦ = EΦ.
Given operators BN ,...,B1 and times tn, . . . , t1 (not necessarily ordered), the corresponding Wightman function is defined as
W (BN , tN ,...,B1, t1)
:= (Φ|BN,H(tN ) ··· B1,H(t1)Φ) i(−tN +tN−1)(H−E) i(−t2+t1)(H−E) = Φ|BN e ··· e B1Φ . Dyson’s time ordered product
Let tn . . . , t1 ∈ R be pairwise distinct. We define Dyson’s time-ordered product of Bn(tn),...,
Bi(t1) by
T(Bn(tn) ··· B1(t1)) := Bin (tin ) ··· Bi1 (ti1 ),
where i1, . . . , in is a permutation such that tin ≥ · · · ≥ ti1 . Green’s functions
The (time-ordered) Green’s functions are defined for pairwise distinct tN , . . . , t1 as
G (BN , tN ,...,B1, t1)
:= (Φ|T(BN,H(tN ) ··· B1,H(t1)) Φ) .
Clearly, we can pass from the Wightman to Green’s functions:
G(BN , tN ,...,B1, t1) X = θ tσ(N) − tσ(N−1) ··· θ tσ(2) − tσ(1) W (Bσ(N), tσ(N),...,Bσ(1), tσ(1)).
σ∈SN Analytic continuation of Wightman functions Assume in addition, that Φ is a ground state of H, that is H ≥ E. Then it is easy to see that Wightman functions can be extended to a function defined for Im(−zN + zN−1) ≥ 0,
Im(−z2 + z1) ≥ 0:
W (BN , zN ,...,B1, z1) i(−zN +zN−1)(H−E) i(−z2+z1)(H−E) = Φ|BN e ··· e B1Φ . The Wightman function is holomorphic in the interior of its domain and continuous up to the boundary. For any real tN > ··· > t1 we have
G(BN , tN ,...,B1, t1) = W (BN , tN ,...,B1, t1). (1.1)
This allows us in principle to determine the Wightman functions from Green’s functions. This follows from the Edge of the Wedge Theorem, which says that in certain situations a holomor- phic function is determined by its values on the boundary of its domain. Note that the holomorphic extension of thw Wightman function at its “euclidean points” is often called the Schwinger function:
S(BN , sN ,...,B1, s1) := W (BN , isN ,...,B1, is1). Standard Møller and scattering operators
Assume now that H0 and V are self-adjoint operators and H := H0 + λV . The standard Møller and scattering operators (if they exist) equal
S± := s− lim eitH e−itH0 , t→±∞ S := S+∗S−.
± Theorem 1.1 1. If the standard Møller operators exist, they are isometric and satisfy S H0 = HS±.
2. If the standard scattering operator exists, it satisfies H0S = SH0. Problem with eigenvalues It is easy to show the following fact:
Theorem 1.2 If the standard Møller operators exist and H0Ψ = EΨ, then HΨ = EΨ.
This is the reason why in practice the standard formalism of scattering theory is usually applied in a suituations where the unperturbed Hamiltonian H0 has only absolutely continuous spectrum.
In quantum field theory, even when one can define (even formally) H0 and H, typically both have ground states, and these ground states are different. Thus standard scattering theory is not applicable. Instead one can sometimes try other approaches to scattering theory. Time-ordered exponentials
Consider a family of operators [t−, t+] 3 t 7→ B(t). We define the time-ordered exponential
∞ Z t+ X Z Z Texp B(t)dt := ··· B(tn) ··· B(t1)dtn ··· dt1 t− n=0 t+≥tn≥···≥t1≥t− ∞ X Z t+ Z t+ 1 = ··· T(B(t ) ··· B(t )) dt ··· dt , t ≥ t ; n! n 1 n 1 + 1 n=0 t− t− −1 Z t+ Z t− Texp B(t)dt := Texp B(t)dt t− t+ ∞ Z Z X n = ··· (−1) B(t1) ··· B(tn)dt1 ··· dtn, t+ ≤ t−. n=0 t+≤t1≤···≤tn≤t− For shortness, let us write
Z t+ U(t+, t−) := Texp B(t)dt . t− Note that d U(t+, t−) = B(t+)U(t+, t−), dt+ d U(t+, t−) = U(t+, t−)B(t−), dt− U(t, t) = 1l,
U(t2, t1)U(t1, t0) = U(t2, t0).
(t −t )B If B(t) = B, then U(t+, t−) = e + − . Time dependent Hamiltonian Let R 3 t 7→ H(t) be a family of self-adjoint operators. H(t) is called the Hamiltonian at time t. We introduce the dynamics (evolution)
Z t+ U(t+, t−) := Texp −i H(s)ds . t−
The dynamics U(t+, t−) is unitary. Heisenberg picture for time-dependent Hamiltonians If R 3 t 7→ A(t) is an operator-valued function, the operator A(t) in the Heisenberg picture is defined as
AH(t) := U(0, t)A(t)U(t, 0).
Note that if A(t) does not depend on t, then AH(t) is the solution of d A (t) = i [H (t),A (t)] , dt H H H AH(0) = A. Interaction picture
Let H0 be a self-adjoint operator. Let R 3 t 7→ V (t) be a family of self-adjoint operators. Set H(t) := H0 + λV (t). Let R 3 t 7→ A(t) be an operator-valued function. The operator A(t) in the interaction picture is defined as
itH0 −itH0 AI(t) := e A(t)e .
The evolution in the interaction picture is defined as
it+H0 −it−H0 UInt(t+, t−) := e U(t+, t−)e .
The interaction Hamiltonian is defined as
HInt := VI(t).
Note that
Z t+ UInt(t+, t−) = Texp −iλ HInt(t)dt . t− Scattering operator for time-dependent perturbations
Quite often, the dynamics in the interaction picture has a limit t− → −∞ or t+ → ∞. This is in particular the case if V (t) decays in time sufficiently fast.
Definition 1.3 The Møller or wave operators, resp. the scattering operator (if they exist) are defined as
± S := s− lim UInt(0, t), t→±∞ S := S+∗S−. Theorem 1.4 1. If S± exist, then they are isometric.
2. If both S+ and S− exist, then so does S and
S = w− lim UInt(t+, t−). t+,−t−→∞
3. If RanS+ = RanS−, then S is unitary.
± Note that all the operators UInt(t+, t−), resp. S can be viewed as special cases of the scattering operator, if we multiply V (t) by 1l[t−,t+](t), resp. 1l[0,+∞[(±t). Wick’s time-ordered product For further applications, it is convenient to modify the definition of Dyson’s time-order product. Assume that I is a unitary involution on H. We call an operator B even, resp. odd, if B = ±IBI. Let t 7→ Bk(t),...,B1(t) be time dependent operators of pure parity.
Definition 1.5 Let tn . . . , t1 ∈ [t+, t−] be pairwise distinct. We define Wick’s time-ordered product of Bn(tn),..., Bi(t1) by
T(Bn(tn) ··· B1(t1)) := sgnaσBσn (tσn ) ··· Bσ1 (tσ1 ),
where σ1, . . . , σn is a permutation such that tσn ≥ · · · ≥ tσ1 and sgnaσ is the sign of the permutation restricted to the odd elements among Bi. Note that Dyson’s time-ordered product is a special case of Wick’s time-ordered product for I = 1l. Therefore, we will keep the same symbol for Dyson’s and Wick’s time-ordered product. Note also that if B(t) is even for all t, then all the formulas of (1) remain true for Wick’s time-ordered product. Green’s functions
Assume that H0 and V (t) are even. Let Φ0 be a fixed vector with H0Φ0 = 0. Suppose that there exist ± Φ := lim UInt(0, t)Φ0 = lim U(0, t)Φ0. (1.2) t→±∞ t→±∞ The Green’s function for the dynamics U is defined as
G(Ak, tk,...,A1, t1) + − := Φ |T Ak,H(tk) ··· A1,H(t1) Φ . (1.3)
Let G0(... ) denote Green’s functions for the free dynamics. We can express interacting Green’s functions by the free ones:
G(Ak, tk,...,A1, t1) ∞ X (−iλ)n Z ∞ Z ∞ = ds ··· ds G (V, s , ··· , V, s ,A , t , ··· ,A , t ) . n! n 1 0 n 1 k k 1 1 n=0 −∞ −∞ In fact, it suffices to assume that tk ≥ · · · ≥ t1. Then the left hand side of (1.3) equals −it+H0 lim lim U(0, t+)e Φ0|U(0, tk)AkU(tk, 0) ··· t+→∞ t−→−∞ −it−H0 ×U(t2, t1)A1U(t1, 0)U(0, t−)e Φ0 = Φ0|UInt(∞, tk)Ak,I(tk)UInt(tk, tk−1) ··· UInt(t2, t1)A1,I(t1)UInt(t1, −∞)Φ0 , which equals the right hand side of (1.3). Abstract LSZ approach
Setting some of ti = ±∞ in Green’s functions (1.3) has the obvious meaning – we take the corresponding limit limti→±∞. Green’s functions can be used to compute the scattering operator. In fact, if B1,...,Bm,Ak,...,A1 are some operators, then
(Bm ··· B1Φ0|SAk ··· A1Φ0) ∗ ∗ = G B1, ∞, ··· ,Bm, ∞,Ak, −∞, ··· A1, −∞ . (1.4) Adiabatic dynamics. −|t| Let V be a self-adjoint operator and > 0. We define V(t) := e V . Then we will write ± H(t), U(t+, t−), S(t+, t−), S , S for the corresponding time-dependent Hamiltonian, etc. Proposition 1.6 We have H(t+)U(t+, t−) − U(t+, t−)H(t−) 0 ≥ t+ ≥ t−; iλ∂λU(t+, t−) = −H(t+)U(t+, t−) + U(t+, t−)H(t−) t+ ≥ t− ≥ 0;
HInt(t+)S(t+, t−) − S(t+, t−)HInt(t−) 0 ≥ t+ ≥ t−; iλ∂λUInt(t+, t−) = −HInt(t+)S(t+, t−) + S(t+, t−)HInt(t−) t+ ≥ t− ≥ 0;
± ± ± iλ∂λS = ±HS ∓ S H0,
iλ∂λS = −H0S + SH0. Proof. Display the dependence on λ by writing U,λ(t+, t−). For t+ ≥ t− ≥ 0 we have
U,λ(t+, t−) = U,λe−θ (t+ + θ, t− + θ).
Hence, d 0 = U,λe−θ (t+ + θ, t− + θ) dθ θ=0 d d = −λ∂λU,λ(t+, t−) + U,λ(t+, t−) + U,λ(t+, t−) dt+ dt− = −λ∂λU,λ(t+, t−) − iH,λ(t+)U,λ(t+, t−) + iU,λ(t+, t−)H,λ(t−).
This proves the first identity, from which the remaing follow. 2 Assume that Φ0 is an eigenvector of H0 with H0Φ0 = E0Φ0. Set ± ± S Φ0 Φ := ± , (Φ0|S Φ0) ± ± ± (Φ0|HS Φ0) (Φ0|HΦ ) E := ± = ± . (Φ0|S Φ0) (Φ0|Φ ) Proposition 1.7
± (H − E0 ∓ iλ∂λ)S Φ0 = 0, ± ± ±iλ∂λ log(Φ0|S Φ0) = E − E0, ± ± H − E ∓ iλ∂λ Φ = 0. The Gell-Mann and Low formula for eigenvectors.
Theorem 1.8 Assume that there exist
± ± ± ΦGL := lim |(Φ0|S Φ0)|Φ . &0 and ± lim λ∂λΦ = 0. &0 Then there exist ± ± EGL := lim E &0 and ± ± ± HΦGL = EGLΦGL. Suppose that E0 is a discrete nondegenerate eigenvalue of H0. Then for small enough λ0 > 0 and |λ| < λ0, H has a unique nondegenerate eigenvalue close to E0, which we denote E. If ± + − + EGL depends continuously on λ, then we see that EGL = EGL = E and ΦGL is proportional − to ΦGL. Suppose now that the we have time reversal invariance. More precisely, assume that T is −1 −1 an antiunitary operator such that TH0T = H0, TVT = V , T Φ0 = Φ0. Then, obviously, −1 + −1 − + − + − TU(t+, t−)T = U(−t+, −t−), TS T = S . Therefore, T ΦGL = ΦGL, EGL = EGL and
+ − (Φ0|ΦGL) = (Φ0|ΦGL).
If we assume both the discreteness of E0 and the time reversal invariance, then we can + − conclude the ΦGL is equal ΦGL. + − We will assume in the following that ΦGL = ΦGL =: Φ.
Lemma 1.9 Let B be an operator. Then (S+Φ |BS−Φ ) (Φ|BΦ) = lim 0 0 . (1.5) &0 (Φ0|SΦ0) Proof. The right hand side of (1.5) equals
+ − + − + − (S Φ0|BS Φ0) (Φ |BΦ ) (ΦGL|BΦGL) lim + − = lim + − = + − . &0 (S Φ0|S Φ0) &0 (Φ |Φ ) (ΦGL|ΦGL) ± But ΦGL = Φ and (Φ|Φ) = 1. 2 The Sucher formula
Theorem 1.10 iλ E − E0 = lim ∂λ log(Φ0|SΦ0). (1.6) &0 2 Proof.
+∗ − iλ∂λS = H0S + SH0 − 2S HS ,
We sandwich it with Φ0 and divide with (Φ0|SΦ0) obtaining
+ − (S Φ0|HS Φ0) iλ∂λ log(Φ0SΦ0) = 2E0 − 2 + − . (S Φ0|S Φ0) The last term, by Lemma 1.9, converges to 2E. 2 The Gell-Mann and Low formula for the N-point function
Theorem 1.11
G (Ak, tk, ··· ,A1, t1) (1.7) ∞ n Z ∞ Z ∞ −1 X (iλ) = lim Z dsn ··· ds1 &0 n! n=0 −∞ −∞ ×G0 (V, sn, ··· V, s1,Ak, tk, ··· A1, t1) , where ∞ X (iλ)n Z ∞ Z ∞ Z = ds ··· ds n! n 1 n=0 −∞ −∞ ×G0 (V, sn, ··· ,V, s1) . Proof. The left-hand side of (1.7) equals
lim (Φ|Ak,H(tk) ··· A1,H(t1)Φ) &0 (S+Φ |A (t ) ··· A (t )S−Φ ) = lim 0 k,H k 1,H 1 0 &0 (Φ0|SΦ0)
But Z = (Φ0|SΦ0). 2 Adiabatic Møller wave operators
Theorem 1.12 1. Assume that there exists the adiabatic or Gell-Mann–Low wave operator
± ± |(Φ0|S Φ0)| ± SGL := w− lim ± S , (1.8) &0 (Φ0|S Φ0) and ± |(Φ0|S Φ0)| ± w− lim λ∂λ ± S = 0. (1.9) &0 (Φ0|S Φ0) Then
± ± SGL(H0 − E0) = (H − E)SGL, (1.10) ± ± SGLΦ0 = ΦGL. (1.11) 2. Define the wave function renormalization operators
± ±∗ ± Z := SGLSGL.
Clearly Z± ≥ 0,
± ± Z H0 = H0Z , ± Z Φ0 = Φ0.
3. Assume that KerZ± = {0}.Define the renormalized Møller operators
± ± ± −1/2 Sren := SGL(Z ) .
± Then Sren are isometric and
± ± Sren(H0 − E0) = (H − E)SGL., (1.12) ± ± SrenΦ0 = ΦGL. (1.13) ± ± 4. Assume that (1.8) is true in the sense of strong limits. Then SGL are isometric and Z = 1l, so that there is no need to renormalize Gell-Mann – Low wave operators. q |f| f |f| |f| i∂λf Proof of 1. Using f = f , we obtain i∂λ f = − f Re f . Therefore, setting f := ± (Φ0|S Φ0) we compute ± ± ± ± |(Φ0|S Φ0)| |(Φ0|S Φ0)| (Φ0|(HS − S H0)Φ0) ±iλ∂λ ± = − ± Re ± (Φ0|S Φ0) (Φ0|S Φ0) (Φ0|S Φ0) ± |(Φ0|S Φ0)| ± = − ± Re E − E0 . (Φ0|S Φ0) Therefore,
± ± |(Φ0|S Φ0)| ± |(Φ0|S Φ0)| ± ± ± ±iλ∂λ ± S = − ± HS − S H0 + Re(E − E0) . (Φ0|S Φ0) (Φ0|S Φ0) ± ± Using ReEGL = EGL, we obtain (1.10). (1.11) follows by definition. 2 Adiabatic scattering operator ± Assume that SGL exist. Define the adiabatic or Gell-Mann–Low scattering operator
+∗ − SGL = SGLSGL. (1.14)
Set
+ −1/2 − 1/2 Sren := (Z ) SGL(Z ) +∗ − = SrenSren.
+ − Assume that ΦGL = ΦGL. Theorem 1.13 1. We have
H0SGL = SGLH0,
SGLΦ0 = Φ0.
2. We have
H0Sren = SrenH0,
SrenΦ0 = Φ0.
+ − 3. If RanSren = RanSren, then Sren is unitary.
± 4. If SGL exist as strong limits, then
S SGL := w− lim . &0 (Φ0|SΦ0)
Chapter 2
Axioms of quantum field theory 65
Consider the Minkowski space R1,3 with variables xµ, µ = 0, 1,..., 3. By definition, it is the vector space R4 of signature (− + ++). In particular, it is equipped with the pseudo-Euclidean quadratic form for x = (xµ) ∈ R1,3 equal to
3 X −(x0)2 + (xi)2. i=1 Writing R3 we will mean the spatial part of the Minkowski space obtained by setting x0 = 0. If x ∈ R1,3, then ~x will denote the projection of x onto R3. Latin letters i, j, k will sometimes denote the spatial indices of a vector. On R1,3 we have the standard Lebesgue measure denoted dx. The notation d~x will be used for the Lebesgue measure on R3 ⊂ R1,3. A nonzero vector x ∈ R1,3 is called
timelike if x2 < 0, causal if x2 ≤ 0, lightlike if x2 = 0, spacelike if x2 > 0.
A causal vector x is called
future oriented if x0 > 0, past oriented if x0 < 0. The set of future/past oriented causal vectors is denoted J ±. We set J := J + ∪ J −. If O ⊂ R1,3, its causal shadow is defined as J(O) := O + J. We also define its future/past shadow J ±(O) := O + J ±. 1,3 Let Oi ⊂ R . We will write O1 × O2 iff J(O1) ∩ O2 = ∅, or equivalently, O1 ∩ J(O2) = ∅.
We then say that O1 and O2 are causally separated. A function on R1,3 is called space-compact iff there exists a compact K ⊂ R1,3 such that suppf ⊂ J(K). It is called future/past space-compact iff there exists a compact K ⊂ R1,3 such that suppf ⊂ J ±(K). ∞ 1,3 The set of space-compact smooth functions will be denoted Csc (R ). The set of fu- ∞ 1,3 ture/past space-compact smooth functions will be denoted C±sc(R ). Lorentz and Poincar´egroup The pseudo-Euclidean group O(1, 3) called the Lorentz group. Its connected component of unity is denoted SO↑(1, 3). The affine extension of the Lorentz group R1,3 o O(1, 3) is called the Poincar´egroup. These groups have double coverings P in(1, 3), Spin↑(1, 3) and R1,3 o P in(1, 3). Note that Spin↑(1, 3) happens to be isomorphic to Sl(2, C). An element of R1,3 o P in(1, 3). will be often written as (a, Λ)˜ and then the corresponding element of R1,3 o O(1, 3) will be written as (a, Λ). Representations of Spin↑(1, 3) can be divided into two categories. The integer spin repre- sentations induce a representation of SO↑(1, 3) and the half-integer representations do not. Basic rules of relativistic quantum mechanics Physical states are described by a Hilbert space H equipped with a strongly continuous unitary representation of the double cover of the Poincar´egroup (or one of its subgroups)
1,3 ↑ R o Spin (1, 3) 3 (x, Λ) 7→ U(x, Λ) ∈ U(H).
The self-adjoint generator of space-time translations is denoted P = (H, P~ ). H is called the Hamiltonian and P~ – the momentum. Thus
U((t, ~x), 1) = e−itH+i~xP~ . It is natural to impose the following conditions:
1. Existence of a Poincar´ecovariant vacuum: There exists a (normalized) vector Ω invariant with respect to R1,3 × Spin↑(1, 3). 2. Spectral condition: sp(P ) ⊂ J +.
3. Uniqueness of the vacuum: The vector Ω is unique up to a phase factor.
Note that (2) implies H ≥ 0. Conversely, the Poincar´einvariance and H ≥ C implies (2). (2) implies also that Ω is the ground state of H. (3) implies that this groundstate is nondegenerate. In the mathematical literature one can find two basic sets of axioms for a quantum theory, which try to express the concept of causality: Haag-Kastler and Wightman axioms. Even though the Wightman axioms were formulated earlier, it is in my opinion more natural to start with the Haag-Kastler axioms. Note that we keep the “basic rules of relativistic quantum mechanics” as a part of both sets of axioms. Haag-Kastler axioms To each open bounded set O ⊂ R1,3 we associate a distinguished von Neumann algebra A(O) ⊂ B(H). The family A(O), O ⊂ R1,3, satisfies the following conditions
1. Isotony: O1 ⊂ O2 imples A(O1) ⊂ A(O1).
2. Covariance: for (a, Λ)˜ ∈ R1,3×Spin↑(1, 3) we have U(a, Λ)˜ A(O)U(a, Λ)˜ ∗ = A ((a, Λ)O)).
3. Einstein causality: O1 × O2 and Ai ∈ A(Oi), i = 1, 2 imply A1A2 = A2A1. This means that measurements in spatially separated regions are independent. S 0 4. Irreducibility: ( O A(O)) = C1l. Local algebras The algebras A(O) describe algebras of observables in O. This means that in principle an observer contained in O can perturb the dynamics by a self-adjoint operator from A(O), and only from A(O). The least obvious axiom is that of irreduciblity. Quantum fields In practical computations of quatum field theory the information is encoded in a family of 1,3 functions R 3 x 7→ φa(x), where a = 1, . . . , n is a finite index set enumerating the “internal degrees of freedom”, typically, species of particles and the value of their spin projected on a distinguished axis. These functions can be viewed as “operator valued distributions”, which become (possibly unbounded) self-adjoint operators when smeared out with test functions in ∞ 1,3 Cc (R ). Some of the particles are bosons, some are fermions. They commute or anticom- mute for causally separated points, which is expressed with the commutation/anticommutation relations 2 [φa(x), φb(y)]± = 0, (x − y) > 0. Smeared quantum fields We can organize the internal degrees of freedom into a finite dimensional vector space n ∞ 1,3 n V = R . Thus for any f = (fa) ∈ Cc (R , R ) we obtain an operator X Z φ[f] := fa(x)φa(x)dx. a From the mathematical point of view it is natural to treat φ[f] as the basic objects. This leads to the following set of axioms. Wightman axioms
We assume that V = Vs ⊕ Va is a finite dimensional real vector space. Vs, resp. Va is called the space of bosonic, resp. fermionic particles. We set := 1l ⊕ (−1l) on V. We will write |s| = 0, |a| = 1. The space V is equipped with a representation
Spin↑(1, 3) 3 Λ˜ 7→ S(Λ)˜ , (2.1) which preserves the decomposition V = Vs ⊕ Va.
We assume that H = Hs ⊕ Ha. We set I := 1l ⊕ (−1l) on H. Hs, resp. Ha is called even fermionic, resp. odd fermionic. We suppose that D is a dense subspace of H containing Ω. We have a map
∞ 1,3 Cc (R , V) 3 f 7→ φ[f] (2.2) into linear operators on D satisfying the following conditions:
1. Continuity: For any Φ, Ψ ∈ D,
∞ 1,3 Cc (R , V) 3 f 7→ (Φ|φ[f]Ψ) (2.3)
is continuous. h i 2. Covariance: for (x, Λ)˜ ∈ R1,3×Spin↑(1, 3) we have U(x, Λ)˜ φ[f]U(x, Λ)˜ ∗ = φ S(Λ)˜ f ◦ (x, Λ)−1 . ∞ |j1||j2| 3. Einstein causality: O1×O2 and fi ∈ Cc (Oi, Vji ), i = 1, 2, imply φ[f1]φ[f2] = (−1) φ[f2]φ[f1]. nS o 4. Irreducibility: Span f φ[f]Ω is dense in H. 5. Self-adjointness: φ[f]∗ = φ[f].
6. Fermionic fields change the fermionic parity: φ[f] = Iφ[f]I
One can show the theorem about the connection of spin and statistics saying that the representation (2.1) has integer spin in Vs and half-integer in Va. Wightman axioms are satisfied by free fields. Neutral vs charged fields In practice, only the so-called neutral fields are assumed to be self-adjoint. One has operator 1,3 ∗ valued distributions R 3 x 7→ ψa(x), ψa(x), a = 1, . . . , m. They are the so-called charged fields. After smearing with complex test functions we obtain operators X Z ψ[g] := ga(x)ψa(x)dx, a Z ∗ X ∗ ψ [g] := ga(x)ψa(x)dx, a such that ψ[g]∗ = ψ∗[g]. Clearly, by setting
1 ∗ φa,R(x) := √ (ψa(x) + ψ (x)), 2 a 1 ∗ φa,I(x) := √ (ψa(x) − ψ (x)), i 2 a for any charged field we obtain a pair of neutral fields. Thus one can organize the space describing the species of particles into two finite dimensional spaces: a real space Vn describing neutral fields and a complex space Vc describing charged fields. Relationship between Haag-Kastler and Wightman axioms ”Morally”, Wightman axioms are stronger than the Haag-Kastler axioms. In fact, let Falg(O) be the ∗-algebra in L(D) (linear operators on D) generated by φ[f] with suppf ⊂ O. It is equipped with an involution α given by α(φ[f]) = φ[f], so that Falg(O) as a vector space is alg alg alg a direct sum of its even and odd part, F0 (O) and F1 (O). Then the family F0 (O) almost alg satisfies the Haag-Kastler axioms, except that elements of F0 (O) are defined only on D and not on the whole H, and often do not extend to bounded operators on H. We know that the fields φ[f] are hermitian on D. Suppose they are essentially self-adjoint. Then we could consider the von-Neumann algebra F(O) generated by bounded functions of
φ[f], suppf ⊂ O. Then the family F0(O) satisfies the the Haag-Kastler axioms.
In practice, observable algebras A(O) can be even smaller than F0(O). This is so in particular if we have a group acting G on the space V responsible for a (global) symmetry of the fields. For example, in the case of charged fields the group equals U(1) and this symmetry is obtained by multiplying fields with a phase factor. The action of G on fields induces an action of G on
F0(O). One can argue that the subalgebra of fixed points of this action should be taken as the observable algebra. (Note that what we described is the case of a global symmetry and not of a local gauge invariance). N-point Wightman functions Under the Wightman axioms we obtain a multilinear map
1,3 1,3 Cc(R , V) × · · · × Cc(R , V)
3 (fN , . . . , f1) 7→ (Ω|φ[fN ] ··· φ[f1]Ω) ∈ C, (2.4) which is separately continuous in its arguments. By the Schwartz Kernel Theorem, it can be extended to a linear map Z 1,3 N ⊗N Cc (R ) , V 3 F 7→ W (xN , . . . , x1)F (xN , . . . , x1)dxN ··· dx1,
(1,3)N (1,3)N where R 3 (xN , . . . , x1) 7→ W (xN , . . . , x1) is a distribution on R with values in the space dual to V⊗N , called the N-point Wightman function, so that (2.4) equals Z W (xN , . . . , x1)fN (x1) ··· f1(x1)dxN ... dx1. From the point of view of the Wightman axioms, the collection of Wightman functions WN , N = 0, 1,... , contains all the information about a given quantum field theory. In particular,
(φ[fN ] ··· φ[f1]Ω|φ[gM ] ··· φ[g1]Ω) Z = W (y1, . . . , yN , xM , . . . , x1)
×f1(x1) ··· fN (xN )gM (yM ) ··· g1(y1)dx1 ··· dxN dyM ··· dy1. N-point Green’s functions In practical computations, the functions that play an important role are not Wightman functions but the so-called (time-orderded) Green’s functions. Their formal definition is as follows:
G(xN , . . . , x1) (2.5) X 0 0 0 0 := sgn(σ)θ xσ(N) − xσ(N−1) ··· θ xσ(2) − xσ(1) W (xσ(N), . . . , xσ(1)),
σ∈SN where sgna(σ) is the sign of the permutation of the fermionic elements among N,..., 1. Note that in (2.5), we multiply a distribution with a discontinuous function, which strictly speaking is illegal. Disregarding this problem, the Green’s function is a covariant, due to the (anti-)commutativity of fields at spacelike separations. Chapter 3
Neutral scalar bosons 89
In this chapter we consider the Klein-Gordon equation with an external source
(−2 + m2)φ(x) = j(x). (3.1)
We will quantize the space of its real solutions. Special solutions of the homogeneous equation Every function ζ that solves the Klein-Gordon equation
(−2 + m2)ζ(x) = 0. (3.2) can be written as 1 Z ζ(x) = eikxg(k)δ(k2 + m2)dk (2π)3 Z ~ q √ 1 X dk 0 ~ 2 2 ~ = g(± ~k2 + m2,~k)e±ix k +m ∓i~xk, (2π)3 p 2 2 ± ~k + m where g is a function on the hyperboloid (mass shell) k2 + m2 = 0. A special role is played by the following 3 solutions of the homogeneous Klein-Gordon equa- tion:
1. The standard positive, resp. negative frequency solution. i Z D(±)(x) = ∓ eikxθ(±k0)δ(k2 + m2)dk (2π)3 Z ~ √ i dk 0 ~ 2 2 ~ = ∓ e±ix k +m ∓i~xk 3 p (2π) 2 ~k2 + m2 2 2 0 √ 1 0 2 mθ(−x ) ±sgnx p 2 miθ(x ) 2 = sgnx δ(x ) ∓ √ H (m −x ) ∓ √ K1(m x ). 4π 8π −x2 1 4π2 x2 ± where H1 are the Hankel functions and K1 is the MacDonald function of the 1st order. 2. The Pauli-Jordan function (or the commutator function) i Z D(x) = e−ikxsgn(k0)δ(k2 + m2)dk (2π)3 1 Z d~k ~ q = ei~xk sin x0 ~k2 + m2 3 p (2π) ~k2 + m2 2 1 0 2 mθ(−x ) p 2 = sgnx δ(x ) − √ J1(m −x ). 2π 4π −x2 D(x) is the unique solution of the homogeneous Klein-Gordon equation satisfying
D(0, ~x) = 0, D˙ (0, ~x) = δ(~x).
We have, suppD ⊂ J. Green’s functions Solutions of (−2 + m2)ζ(x) = δ(x), (3.3) are called Green’s functions or fundamental solutions of the Klein-Gordon equation. In partic- ular, let us introduce the following important Green’s functions:
1. advanced, resp. retarded Green’s function. 1 Z e−ikx D±(x) = dk (2π)4 k2 + m2 ∓ i0sgnk0 2 0 1 0 2 mθ(−x )θ(±x ) p 2 = θ(±x )δ(x ) − √ J1(m −x ). 2π 4π −x2 We have suppD± ⊂ J ±. In the literature, D+(x) is usually denoted Dret(x) and D−(x) is usually denoted Dadv(x). 2. causal or Feynman(-Stueckelberg) Green’s function.
1 Z e−ikx Dc(x) = dk (2π)4 k2 + m2 − i0 2 2 √ 1 2 mθ(−x ) − p 2 miθ(x ) 2 = δ(x ) − √ H (m −x ) + √ K1(m x ). 4π 8π −x2 1 4π x2 Relationships between special solutions and Green’s functions
D(x) = −D(−x) = D(+)(x) + D(−)(x) = D+(x) − D−(x), D(+)(x) = −D(−)(−x), D+(x) = D−(−x) = θ(x0)D(x) D−(x) = θ(−x0)D(x) Dc(x) = Dc(−x) = θ(x0)D(−)(x) − θ(−x0)D(+)(x). Let us prove the last identity. 1 Z e−ikx Dc(x) = dk (2π)4 k2 + m2 − i0 1 Z e−ikx = 4 dk (2π) p~ 2 2 p~ 2 2 2 k + m k + m − |k0| − i0 1 Z e−ikx + 4 dk (2π) p~ 2 2 p~ 2 2 2 k + m k + m + |k0| + i0 ~ 1 Z eik0x0−ik~x = 4 dk (2π) p~ 2 2 p~ 2 2 2 k + m k + m − k0 − i0 ~ 1 Z eik0x0+ik~x + 4 dk (2π) p~ 2 2 p~ 2 2 2 k + m k + m + k0 + i0 ~ 1 Z eik0x0−ik~x = 4 dk (2π) p~ 2 2 p~ 2 2 2 k + m k + m − k0 − i0 ~ 1 Z eik0x0+ik~x + 4 dk (2π) p~ 2 2 p~ 2 2 2 k + m k + m + k0 + i0 This equals
~ ~ i Z eik0x0−ik~x i Z e−ik0x0+ik~x = θ(x ) d~k + θ(x ) d~k 3 0 p 3 0 p (2π) 2 ~k2 + m2 (2π) 2 ~k2 + m2 (−) (−) = θ(x0)D (x) + θ(−x0)D (−x), where we used the identity Z eits ds = 2πie−itaθ(t). a − s − i0 Retarded and advanced solutions of the inhomogeneous problem Among solutions of the inhomogeneous Klein-Gordon equation one can distinguish the ad- vanced and the retarded solution. They can be obtained by the convolution with the ad- vanced/retarded Green’s function.
∞ 1,3 ± ∞ 1,3 Theorem 3.1 For any f ∈ Cc (R ) there exist unique functions ζ ∈ C±sc(R ), solutions of (−2 + m2)ζ± = f, suppζ± ⊂ J ±(suppf).
Moreover, ζ±(x) = R D±(x − y)f(y)dy.
Note that by duality D± can be applied to distributions of compact support. Therefore, in the above theorem we can assume that f is a distribution of a compact support, and then ζ± is a distribution. The Cauchy problem Solutions of the Cauchy problem are uniquely parametrized by their Cauchy data (the value and the normal derivative on a Cauchy surphace). They can be expressed by the Cauchy data with help of the Jordan-Pauli function.
∞ 3 ∞ 1,3 Theorem 3.2 Let ς, ϑ ∈ Cc (R ). Then there exists a unique ζ ∈ Csc (R ) that solves
(−2 + m2)ζ = 0 (3.4) with initial conditions ζ(0, ~x) = ς(~x), ζ˙(0, ~x) = ϑ(~x). It satisfies suppζ ⊂ J(suppς ∪ suppϑ) and is given by Z Z ζ(t, ~x) = − D˙ (t, ~x − ~y)ς(~y)d~y + D(t, ~x − ~y)ϑ(~y)d~y. R3 R3 Conserved current
Let YKG,R, resp. YKG denote the space of real, resp. complex, space-compact solutions of the Klein Gordon equation. Clearly, YKG = CYKG,R. ∞ 1,3 Let ζ1, ζ2 ∈ C (R , R). We define
µ µ µ µ j (x) = j (ζ1, ζ2, x) := ∂ ζ1(x)ζ2(x) − ζ1(x)∂ ζ2(x).
We easily check that
µ 2 2 ∂µj (x) = −(−2 + m )ζ1(x)ζ2(x) + ζ1(x)(−2 + m )ζ2(x),
Therefore, if ζ1, ζ2 ∈ YKG, then µ ∂µj (x) = 0. One says that jµ(x) is a conserved current. Symplectic form The flux of jµ across a space-like subspace S of codimension 1 does not depend on its choice. It defines a symplectic form on YKG,R Z µ ζ1ωζ2 = j (ζ1, ζ2, x)dsµ(x) S Z ˙ ˙ = ζ1(t, ~x)ζ2(t, ~x) − ζ1(t, ~x)ζ2(t, ~x) d~x. (3.5)
Clearly, the form (3.5) is well defined if only ζ2 ∈ YKG,R, and ζ1 is a distributional solution of the Klein-Gordon equation. Poincar´ecovariance
The Poincar´egroup acts on YKG,R by