The propagator and diffeomorphisms of an interacting field theory
Master thesis
submitted to the Institut f¨urPhysik Mathematisch-Naturwissenschaftliche Fakult¨at Humboldt-Universtit¨atzu Berlin
by
Paul-Hermann Balduf
in partial fulfillment of the requirements for the degree Master of Science.
Adlershof, July 5, 2018
Referees: Prof. Dr. Dirk Kreimer (supervisor) Dr. Christian Bogner We consider a scalar quantum field theory where the field φ is replaced by a diffeomor- 2 3 phism φ = ρ+a1ρ +a2ρ +.... The field ρ constitutes a modified quantum field theory defined implicitly by the diffeomorphism coefficients {aj}j and the Lagrange density of the underlying field φ. For a generic diffeomorphism, ρ is a non-renormalizable quan- tum field with infinitely many interaction vertices, even if φ itself is a free field. In the case that φ is an interacting field itself, ρ obtains additional vertices proportional to the coupling parameters in the Lagrangian density of φ. We examine the general Lagrangian density of a scalar field with local interactions,
∞ 1 µ 1 2 2 X λs s L = ∂µφ∂ φ − m φ − φ , 2 2 s! s=3 and show that the S-matrix elements of ρ coincide with the ones of φ. This implies the fields φ and ρ are indistinguishable in experiments. In this sense diffeomorphisms form equivalence classes of interacting scalar quantum field theories. On the other hand, n-point functions differ between φ and ρ if their momenta are offshell. Tuning the diffeomorphism coefficients {aj}j allows to change the behaviour of the offshell n-point functions of ρ. We use this freedom to eliminate all loop-corrections to the propagator of ρ for a fixed offshell four-momentum p. If φ has φs-type interaction,
1 µ 1 2 2 λ s L = ∂µφ∂ φ − m φ − φ , 2 2 s! and all tadpole graphs vanish, then the field formally given by λ ρ(x) = φ(x) − φs−1(x) (s − 1)!(p2 − m2) has a free propagator to all orders in perturbation theory, i.e.
ρ(p)ρ(−p) = −i p2 − m2.
Contents
1 Introduction5 1.1 Minkowski spacetime...... 5 1.2 Fuss-Catalan numbers...... 6 1.2.1 Generating function of A(2, 1)...... 7 1.2.2 Generating function of A(3, 1)...... 8 1.3 Bell polynomials...... 10 1.4 Axiomatic quantum field theory...... 11 1.4.1 Wightman axioms...... 11 1.4.2 Wightman distributions...... 12 1.4.3 Haag‘s theorem...... 13 1.5 Perturbative quantum field theory...... 14 1.5.1 Scattering theory in the interaction picture...... 14 1.5.2 Feynman graphs...... 17 1.5.3 Renormalization...... 18 1.6 Diffeomorphisms...... 21 1.7 Motivation for this work...... 21 1.8 Organization of the text...... 23
2 Free theory 25 2.1 Diffeomorphism of a free scalar theory...... 25 2.1.1 Lagrangian density...... 25 2.1.2 Feynman rules...... 27 2.1.3 Renormalizability...... 29 2.2 Tree sums...... 30 2.3 Multiple external offshell edges...... 33 2.4 Loop amplitudes...... 37
3 Phi3-theory 39 3.1 Diffeomorphism vertices of Phi3-theory...... 39 3.2 Cancellation of higher interaction vertices...... 41 3.2.1 Four external edges...... 41 3.2.2 General structure...... 42 3.2.3 Formal definition of Sn...... 43 3.2.4 Vanishing of Sn...... 46 3.3 Tree sums with interaction vertices...... 47 3.3.1 General form...... 47 3.3.2 Example: Tree sum with four external edges...... 48 3.3.3 Example: Tree sum with five external edges...... 51 3.4 Explicit cancellation of the 2-point function for one loop...... 57
3 3.5 Explicit calculation for two loops...... 60 3.5.1 Topology A...... 61 3.5.2 Topology B...... 62 3.5.3 Topology C...... 64 3.5.4 Topology D...... 68 3.5.5 Total result...... 72 3.6 All-orders cancellation of corrections...... 72 3.6.1 Decomposition of non-vanishing graphs...... 73 3.6.2 Factorization of tree sums...... 74 3.6.3 Parameters of the adiabatic diffeomorphism...... 79
4 Higher order interaction 83 4.1 Feynman rules...... 83 4.2 Tree sums and higher interaction vertices...... 85 4.3 Cancellation of corrections to the 2-point function...... 87 4.4 Multiple interaction monomials...... 94 4.4.1 Feynman rules...... 94 4.4.2 Tree sums...... 95
5 Properties of the adiabatic diffeomorphism 97 5.1 Rho as a free field...... 97 5.1.1 Rho in momentum space...... 97 5.1.2 Classical fields...... 98 5.1.3 Cancellation of higher correlation functions...... 99 5.2 An identity for PhiS-theory...... 99 5.3 Conclusion and outlook...... 101
Bibliography 103
A Lemmas 107 A.1 Sums of partitions of momenta...... 107 A.2 Compatibility of symmetry factors...... 109 A.3 Combinatoric Lemmas...... 111
B Umgangssprachliche Erl¨auterungder Arbeit 123
C Danksagung 127
D Selbstst¨andigkeitserkl¨arung 129
4 1 Introduction
1.1 Minkowski spacetime
Four-dimensional Minkowski spacetime M with speed of light c = 1 is used throughout. Four-vectors are underlined and three-vectors are bold such that the four-momentum is 0 1 2 3 p = p , p , p , p = (E, px, py, pz) = (E, p). The metric is flat and has negative sign, 1 0 0 0 0 −1 0 0 η = . 0 0 −1 0 0 0 0 −1 Four-dimensional spacetime indices are greek and summation is implicit,
2 µ µ ν 2 2 p = pµp = ηµνp p = E − p . (1.1) Other indices are not summed over unless explicitly denoted. The spacetime derivative ∂ is ∂µ = ∂xµ , especially the second derivative µ 2 ∂µ∂ = ∂0∂0 − ∂1∂1 − ∂2∂2 − ∂3∂3 = ∂t − 4 = . Physical particles (i.e. “really existent” as opposed to virtual particles used in interme- diate steps in computations) obey the relativistic energy-momentum-relation
E2 = p2 + m2. (1.2)
By eq. (1.1) the square of the four-momentum of a physical particle is its mass squared. In four-dimensional momentum space, eq. (1.2) forms a surface called the mass shell. Hence if a particle fulfills eq. (1.2) it is called onshell, otherwise offshell. Definition 1. For a four-momentum p assigned to a particle with mass m, the offshell parameter is defined as
2 2 xp = p − m . This parameter is zero if the momentum p belongs to a physical particle with mass m. The notation here follows slightly nontrivial rules: If the index is a letter like xp, this letter indicates the name of the corresponding four-momentum. If the index is a number, it is the running number of some (usually canonically) numbered momenta like x p2 − m2. 1 = 1 (1.3)
5 If sums and differences appear in the index, these indicate the corresponding operations with momenta, not indices themselves. This is, 2 x p p − m2, 1+2 = 1 + 2 (1.4) but generally x1+2 =6 x3. Note that xp only depends on the magnitude of the momen- tum, for any four-momentum p it is xp = x−p. The equation xp = xq does not imply p = q.
1.2 Fuss-Catalan numbers
Definition 2. For fixed a ∈ N0 and b ∈ N the Fuss-Catalan numbers are defined as b ma + b (ma + b − 1)! Am(a, b) = = b . ma + b m (ma + b − m)!m! Some useful values are
A0(a, b) = 1
Am(1, 1) = 1
Am(1, 2) = m b A (0, b) = . m m
Example 1: Fuss-Catalan numbers used in subsequent examples
The (ordinary) Catalan numbers {Cm}m∈N = {1, 1, 2, 5, 14, 42, 132, 429,...} are
1 2m (2m)! Am(2, 1) = Cm = = . (1.5) m + 1 m m!(m + 1)!
In the context of φ4-theory we encounter the choice a = 3, b = 1,
1 3m {Am(3, 1)}m∈N = = {1, 1, 3, 12, 55, 273, 1428, 7752 ...}. 2m + 1 m m∈N
Countless interpretations of the Fuss-Catalan numbers are known, see for example their entries A000108 (Am(2, 1)) or A001764 (Am(3, 1)) in the OEIS [Slo18]. Noteworthy, the Catalan number Cm is the number of different planar trees with n + 1 leaves built of 3-valent vertices. Similarly, Am(s, 1) counts the number of such trees if they are made of s-valent vertices. The generating function of the Fuss-Catalan numbers is defined by ∞ X m Ca,b(t) = t Am(a, b). (1.6) m=0
6 Lemma 1.1. Iff for q ∈ R a function R(t) obeys
R(t) − 1 = tRq(t) then for b ∈ R ∞ X b mq + b Rb(t) = tm . mq + b m m=0
Proof. This is [MSV06, theorem 2.1] with the replacement R(t) = 1 + w.
By lemma 1.1 the generating function eq. (1.6) of the Fuss-Catalan-numbers is defined equivalently via its functional equation
a b b Ca,b(t) = tCa,b(t) + 1 . (1.7)
From eq. (1.6) and A0(a, b) = 1 the boundary condition is
Ca,b(0) = 1. (1.8)
For b = 1 and a = 2 or a = 3 the functional equation can be solved explicitly.
1.2.1 Generating function of A(2, 1) Set a = 2, b = 1, this gives the generating function of ordinary Catalan numbers defined implicitly via eq. (1.7)
2 C2,1(t) ≡ C(t) = tC (t) + 1. (1.9)
Solving the quadratic equation, there are (for any given t) two solutions √ 1 ± 1 − 4t C (t) = . (1.10) 1/2 2t As can be seen in fig. 1.1 or computed explicitly, only one of them fulfills the boundary condition eq. (1.8) C(0) = 1, consequently the generating function of Catalan numbers is √ 1 − 1 − 4t C(t) = . (1.11) 2t
As a generating function according to eq. (1.6), C(t) = C2,1(t) has to be real-valued. This is the case for 1 t ≤ . 4
7 Figure 1.1: The two solutions Cj(t) from eq. (1.10) of the functional equation (1.9). Solid real part, dashed imaginary part. To distinguish them, the imaginary parts have been slightly shifted off the x-axis. Only C2(t) (red) fulfills the boundary condition C(1) = 1 from eq. (1.8).
1.2.2 Generating function of A(3, 1)
For a = 3, b = 1 set C(t) ≡ C3,1(t), then eq. (1.7) becomes the cubic equation
1 1 C3(t) − C(t) + = 0. (1.12) t t |{z} |{z} p q
It can be solved with the Cardanic formulae [Car68]. Define
p3 q 2 27t − 4 ∆ = + = 3 2 108t3 and two auxiliary functions
s r r q √ 3 1 −4 + 27t u(t) = 3 − + ∆ = − + , 2 2t 108t3 s r r q √ 3 1 −4 + 27t v(t) = 3 − − ∆ = − − . 2 2t 108t3
p 1 The complex cube roots have to be chosen to guarantee u · v = − 3 = 3t as shown in fig. 1.2. √ √ −1+i 3 −1−i 3 Introducing the cube roots of unity 1 = 2 and 2 = 2 , the three solutions
8 Figure 1.2: Functions u(t) and v(t). of eq. (1.12) are
C1(t) = u + v
C2(t) = 1u + 2v (1.13)
C3(t) = 2u + 1v.
Figure 1.3: The three solutions Cj(t) from eq. (1.13). The solution fulfilling eq. (1.8), C(1) = 1, can be constructed joining C1(t) and C2(t).
To fulfill the boundary condition eq. (1.8), the correct generating function is C1(t)
9 for t < 0 and and C(0) = 1 and C2(t) for t > 0. Explicitly, this is C (t) t < 0 1 C3,1(t) = 1 t = 0 (1.14) C2(t) t > 0 r q r q 3 1 27t−4 3 1 27t−4 − 2t + 108t3 + − 2t − 108t3 t < 0 = √ r q √ r q −1+i 3 3 1 27t−4 1+i 3 3 1 27t−4 2 − 2t + 108t3 − 2 − 2t − 108t3 t ≥ 0. The distinction of the two cases just means taking different cube roots. The generating function C3,1(t) has to be real-valued, this is the case only for 4 t ≤ . 27 1.3 Bell polynomials
We follow the definitions of [KY17]. See also [FG05, chapter 7] for details and proofs of the properties listed below.
Definition 3. For k ∈ N0, n ∈ N0 and k ≤ n, the Bell polynomials are
X n! x1 j1 x2 j2 x3 j3 xn jn Bn,k(x1, x2, x3,...) = ··· , j1!j2! ··· jn! 1! 2! 3! n! S where the sum runs over all {ji}i∈{1,...,k} such that
S = {ji ≥ 0 ∀i, j1 + j2 + j3 + ... + jk = k j1 + 2j2 + 3j3 + ... + (n − k)jn−k = n}. Special values are
B0,0 = 1
B0,k = 0, k > 0
Bn,0 = 0, n > 0 (1.15)
Bn,k = 0, k > n
Bn,1 = xn, n > 0 n Bn,n = x1 , n > 0. We will frequently need the generating function of Bell polynomials ∞ n n ∞ j X X k t X t Bn,k(x1, x2,...)u = expu xj . (1.16) n! j! n=0 k=0 j=1 Expanding the exponential function, exchanging summation order on the l.h.s. and extracting the coefficients of uk this becomes k ∞ ∞ X tn 1 X tj Bn,k(x1, x2,...) = xj . (1.17) n! k! j! n=k j=1
10 Bell polynomials have a striking combinatoric meaning: They count partitions of {1, . . . , n} into k nonempty disjoint subsets, i.e. X Bn,k(x1, x2, x3,...) = x|P1| ··· x|Pk|, P where
P = {∅= 6 Pi ⊆ {1, . . . , n} ∀i, Pi ∩ Pj = ∅ ∀i =6 j, P1 ∪ ... ∪ Pk = {1, . . . , n}}. (1.18) A consequence of eq. (1.16) is that Bell Polynomials represent the coefficients of composed diffeomorphisms. Fa`adie Bruno’s Formula [Wei18a] for the n-th derivative n ∂t of a composed function is n n X k 2 3 ∂t f(g(t)) = ∂t f (g(t)) · Bn,k ∂tg(t), ∂t g(t), ∂t g(t),... . k=0 P∞ tn P∞ tn In terms of power series f(t) = n=1 fn n! , g(t) = n=0 gn n! and ∞ X tn f(g(t)) =: h(t) = hn n! n=0 this becomes n X hn = fk · Bn.k(g1, . . . , gn+1−k). (1.19) k=1
1.4 Axiomatic quantum field theory
We consider scalar quantum fields exclusively, these are fields with spin zero. The following part follows Lutz Klaczynski’s PhD thesis [Kla16].
1.4.1 Wightman axioms The Wightman axioms are a mathematically rigorous way of defining quantum fields. They read 1. The states of the physical system are described by vectors in a separable Hilbert space h equipped with a strongly unitary representation (a, Λ) → U(a, Λ) of the ↑ connected Poincar´egroup P+ (i.e. the group of orthochronous proper Lorentz transformations and shifts). Moreover there is a unique vacuum Ψ0 ∈ h which is invariant under these Poincar´etransformations, U(a, Λ)Ψ0 = Ψ0. 2. The generator of the translation subgroup ∂ i U(a, 1) = pµ ∂aµ a=0
has its spectrum inside the closed forward light cone: σ(p) ⊂ V + and the generator of time translations (the Hamiltonian) has nonnegative eigenvalues, H = p0 ≥ 0.
11 3. For every Schwartz function f ∈ S(M) there are operators φ1(f), . . . , φn(f) (called † † quantum fields) and their adjoints φ1(f), . . . , φn(f) on h such that the polynomial algebra
D † E A(M) = φj(f), φj(f): j ∈ {1, . . . , n} C has a stable common dense domain D ⊂ h. This means
A(M)D ⊂ D ↑ U(a, Λ)D ⊂ D ∀(a, Λ) ∈ P+.
Further, the vacuum is cyclic for A(M), this means
Ψ0 ∈ D
D0 := A(M)Ψ0 ⊆ D is dense in h.
Finally, the maps
0 f → Ψ φj(f)Ψ
are tempered distributions on S(M) for all Ψ, Ψ0 ∈ D and j ∈ {1, . . . , n}.
4. The quantum fields transform under the unitary representation of the Poincar´e group according to
† U(a, Λ)φj(f)U (a, Λ) = φj({a, Λ}f)
on the domain D where
({a, Λ}f)(x) = f Λ−1(x − a)
is the Poincar´e-transformedtest function.
5. For f, g ∈ S(M) with mutually spacelike-separated support, this is
f(x)g(y) =6 0 ⇒ (x − y)2 < 0,
the commutator of quantum fields vanishes:
[φj(f), φl(g)] = φj(f)φl(g) − φl(g)φj(f) = 0 ∀j, l ∈ {1, . . . , n}.
1.4.2 Wightman distributions Wightman distributions are the n-fold vacuum expectation values of quantum fields
Wn(f1, . . . , fn) = hΨ0 φ(f1) ··· φ(fn)Ψ0i.
They fulfill a set of axioms equivalent to the Wightman axioms for the quantum fields themselves in the sense that given a set of Wightman distributions, there exists a
12 corresponding quantum field theory fulfilling the Wightman axioms (Wightman’s re- construction theorem, [Kla16, Theorem 10.1]). Especially, if φ is a free field without interaction, the 2-point distribution (where Ω0 instead of Ψ0 is used to stress the fact this state is the vacuum of a free theory) is Z 4 d p ∗ 0 2 2 hΩ0 φ(f)φ(h)Ω0i = f˜ (p)2πθ(p )δ(p − m )h˜(p), (1.20) (2π)4 where f˜ is the Fourier transform of f. If one takes the test functions f and h to have “sufficiently” small support around some points x, y such that they can be approximated by delta distributions
f(z) → δ(z − x), h(z) → δ(z − y) (1.21) then their Fourier transformations become
f˜∗(p) → e−ipx h˜(p) → eipy and one can use the notation
φ(x) := φ(f) f(y)≈δ(y−x)
φ(p) := φ(f˜) . f˜(p)≈exp(ipy)
Under this condition eq. (1.20) reduces to the function [Zwi17, eq. (7)], [Kla16, eq. (9.5)]