nLab
renormalization

Context

Algebraic Quantum Field Theory

algebraic quantum field theory (perturbative, on curved spacetimes, homotopical)

Introduction

Concepts

field theory: classical, pre-quantum, quantum, perturbative quantum

Lagrangian field theory

quantization

quantum mechanical system, quantum probability

free field quantization

gauge theories

interacting field quantization

renormalization

Theorems

States and observables

Operator algebra

Local QFT

Perturbative QFT

Contents

Idea

The construction of a perturbative quantum field theory from a given local Lagrangian density (rigorously via causal perturbation theory) involves ambiguities associated with the detailed nature of the quantum processes at point interactions. What is called renormalization is making a choice of fixing these ambiguities to produce a perturbative quantum field theory (Wightman et al. 76).

In causal perturbation theory the renormalization ambiguities are understood as the freedom of extending the time-ordered product of operator-valued distributions to the interaction points, in the sense of extension of distributions. The notorious “infinities that plague quantum field theory” arise only if this extension is not handled correctly and has to be fixed. Hence what historically is called “renormalization” could from a mathematical point of view just be called “normalization” (a point made vividly for instance in Scharf 95, Scharf 01).

The main theorem of perturbative renormalization theory (theorem 2 below) states that any two choices of such (re-)normalizations are uniquely related by a re-definition of the interaction Lagrangian density which introduces further point interactions of higher order (“counter terms”),

The extension of distributions of the time-ordered product may naturally be organized via graphs, the Feynman graphs (Garcia-Bondia & Lazzarini 00, Keller 10, chapter IV), and hence the renormalized perturbative S-matrix defining the perturbative quantum field theory is expressed as a formal power series in renormalized Feynman graphs, the Feynman perturbation series (Keller 10 (IV.12)).

Historically the Feynman perturbation series was motivated from intuition about the would-be path integral, and this is still a popular point of view, albeit its lack of rigorous formulation. One may understand the axiomatics on the S-matrix in causal perturbation theory as defining the result of the path integral without actually doing an integration over field configurations.

But while path integral quantization for perturbative quantum field theory remains elusive, it has been shown that the (re-)normalized perturbative quantum field theory thus constructed via causal perturbation theory is, at least under favorable circumstances, equivalently the (Fedosov) formal deformation quantization of the covariant phase space induced by the given interacting Lagrangian density (Collini 16). This identifies the (re-)normalization freedom with the usual freedom in choosing formal deformation quantization.

This also suggest that the construction of the full non-perturbative quantum field theory ought to be given by a strict deformation quantization of the covariant phase space. But presently no example of such for non-trivial interaction in spacetime dimension 4\geq 4 is known. In particular the phenomenologically interesting case of a complete construction of interacting field theories on 4-dimensional spacetimes is presently unknown. For the case of Yang-Mills theory this open problem is one of the “Millenium Problems” (see at quantization of Yang-Mills theory).

The following is taken from geometry of physics -- A first idea of quantum field theory. See there for more backround.

Details

There are different formulations of renormalization:

  1. In causal perturbation theory

  2. BPHZ and Hopf-Algebraic renormalization

  3. Of theories in BV-CS form

In causal perturbation theory

We discuss (re-)normalization of relativistic perturbative QFT in the rigorous formulation of causal perturbation theory/perturbative AQFT.

In this rigorous discussion no “infinite divergent quantities” (as in the original informal discussion due to Schwinger-Tomonaga-Feynman-Dyson) that need to be “re-normalized” to finite well-defined quantities are ever considered, instead finite well-defined quantities are considered right away, and the available space of choices is determined. Therefore making such choices is really a normalization of the time-ordered products/Feynman amplitudes (as prominently highlighted in Scharf 95, see title, introduction, and section 4.3). Actual re-normalization is the the change of such normalizations:

The construction of perturbative QFTs may be explicitly described by an inductive extension of distributions of time-ordered products/Feynman amplitudes to coinciding interaction points. This type of construction is called

This inductive construction has the advantage that it gives accurate control over the space of available choices of renormalizations (theorem 1 below) but it leaves the nature of the “new interactions” that are to be chosen at coinciding interaction points somwewhat implicit.

Alternatively, one may re-define the interactions explicitly (by adding “counterterms”), depending on a chosen UV cutoff-scale, and construct the limit as the “cutoff is removed”. This is called (“re”-)normalization by

This still leaves open the question how to choose the counterterms. For this it serves to understand the relative effective action induced by the choice of UV cutoff at any given cutoff scale. The infinitesimal change of these relative effective actions follows a universal differential equation, Polchinski's flow equation (prop. 12 below). This makes the problem of (“re”-)normalization be that of solving this differential equation subject to chosen initial data. This is the perspective on (“re”-)normalization called

The main theorem of perturbative renormalization states that different S-matrix schemes are precisely related by vertex redefinitions. This yields the

If a sub-collection of renormalization schemes is parameterized by some group RGRG, then the main theorem implies vertex redefinitions depending on pairs of elements of RGRG this is known as

Specifically scaling transformations on Minkowski spacetime yields such a collection of renormalization schemes; the corresponding renormalization group flow is known as

\,

Epstein-Glaser normalization

The construction of perturbative quantum field theories around a given gauge fixed relativistic free field vacuum is equivalently, by this prop., the construction of S-matrices 𝒮(gS int+jA)\mathcal{S}(g S_{int} + j A) in the sense of causal perturbation theory (this def.) for the given local interaction gS int+jAg S_{int} + j A. By prop. 1 the construction of these S-matrices is inductively in kk \in \mathbb{N} a choice of extension of distributions (remark 1 and def. 3 below) of the corresponding kk-ary time-ordered products of the interaction to the locus of coinciding interaction points. An inductive construction of the S-matrix this way is called Epstein-Glaser-("re"-)normalization (this def.).

By paying attention to the scaling degree (def. 2 below) one may precisely characterize the space of choices in the extension of distributions (prop. 5 below): For a given local interaction gS int+jAg S_{int} + j A it is inductively in kk \in \mathbb{N} a finite-dimensional affine space. This conclusion is theorem 1 below.

\,

Proposition

(("re"-)normalization is inductive extension of time-ordered products to diagonal)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge-fixed relativistic free vacuum according to this def..

Assume that for nn \in \mathbb{N}, time-ordered products {T k} kn\{T_{k}\}_{k \leq n} of arity knk \leq n have been constructed in the sense of this def.. Then the time-ordered product T n+1T_{n+1} of arity n+1n+1 is uniquely fixed on the complement

Σ n+1diag(n)={(x iΣ) i=1 n|i,j(x ix j)} \Sigma^{n+1} \setminus diag(n) \;=\; \left\{ (x_i \in \Sigma)_{i = 1}^n \;\vert\; \underset{i,j}{\exists} (x_i \neq x_j) \right\}

of the image of the diagonal inclusion ΣdiagΣ n\Sigma \overset{diag}{\longrightarrow} \Sigma^{n} (where we regarded T n+1T_{n+1} as a generalized function on Σ n+1\Sigma^{n+1} according to this remark).

This statement appears in (Popineau-Stora 82), with (unpublished) details in (Stora 93), following personal communication by Henri Epstein (according to Dütsch 18, footnote 57). Following this, statement and detailed proof appeared in (Brunetti-Fredenhagen 00).

Proof

We will construct an open cover of Σ n+1Σ\Sigma^{n+1} \setminus \Sigma by subsets 𝒞 IΣ n+1\mathcal{C}_I \subset \Sigma^{n+1} which are disjoint unions of non-empty sets that are in causal order, so that by causal factorization the time-ordered products T n+1T_{n+1} on these subsets are uniquely given by T k() HT nk()T_{k}(-) \star_H T_{n-k}(-). Then we show that these unique products on these special subsets do coincide on intersections. This yields the claim by a partition of unity.

We now say this in detail:

For I{1,,n+1}I \subset \{1, \cdots, n+1\} write I¯{1,,n+1}I\overline{I} \coloneqq \{1, \cdots, n+1\} \setminus I. For I,I¯I, \overline{I} \neq \emptyset, define the subset

𝒞 I{(x i) i{1,,n+1}Σ n+1|{x i} iI{x j} j{1,,n+1}I}Σ n+1. \mathcal{C}_I \;\coloneqq\; \left\{ (x_i)_{i \in \{1, \cdots, n+1\}} \in \Sigma^{n+1} \;\vert\; \{x_i\}_{i \in I} {\vee\!\!\!\wedge} \{x_j\}_{j \in \{1, \cdots, n+1\} \setminus I} \right\} \;\subset\; \Sigma^{n+1} \,.

Since the causal order-relation involves the closed future cones/closed past cones, respectively, it is clear that these are open subsets. Moreover it is immediate that they form an open cover of the complement of the diagonal:

I{1,,n+1}I,I¯𝒞 I=Σ n+1diag(Σ). \underset{ { I \subset \{1, \cdots, n+1\} \atop { I, \overline{I} \neq \emptyset } } }{\cup} \mathcal{C}_I \;=\; \Sigma^{n+1} \setminus diag(\Sigma) \,.

(Because any two distinct points in the globally hyperbolic spacetime Σ\Sigma may be causally separated by a Cauchy surface, and any such may be deformed a little such as not to intersect any of a given finite set of points. )

Hence the condition of causal factorization on T n+1T_{n+1} implies that restricted to any 𝒞 I\mathcal{C}_{I} these have to be given (in the condensed generalized function-notation from this remark) on any unordered tuple X={x 1,,x n+1}𝒞 I\mathbf{X} = \{x_1, \cdots, x_{n+1}\} \in \mathcal{C}_I with corresponding induced tuples I{x i} iI\mathbf{I} \coloneqq \{x_i\}_{i \in I} and I¯{x i} iI¯\overline{\mathbf{I}} \coloneqq \{x_i\}_{i \in \overline{I}} by

(1)T n+1(X)=T(I)T(I¯)AAforA𝒳𝒞 I. T_{n+1}( \mathbf{X} ) \;=\; T(\mathbf{I}) T(\overline{\mathbf{I}}) \phantom{AA} \text{for} \phantom{A} \mathcal{X} \in \mathcal{C}_I \,.

This shows that T n+1T_{n+1} is unique on Σ n+1diag(Σ)\Sigma^{n+1} \setminus diag(\Sigma) if it exists at all, hence if these local identifications glue to a global definition of T n+1T_{n+1}. To see that this is the case, we have to consider any two such subsets

I 1,I 2{1,,n+1},AAI 1,I 2,I 1¯,I 2¯. I_1, I_2 \subset \{1, \cdots, n+1\} \,, \phantom{AA} I_1, I_2, \overline{I_1}, \overline{I_2} \neq \emptyset \,.

By definition this implies that for

X𝒞 I 1𝒞 I 2 \mathbf{X} \in \mathcal{C}_{I_1} \cap \mathcal{C}_{I_2}

a tuple of spacetime points which decomposes into causal order with respect to both these subsets, the corresponding mixed intersections of tuples are spacelike separated:

I 1I 2¯><I 1¯I 2. \mathbf{I}_1 \cap \overline{\mathbf{I}_2} \; {\gt\!\!\!\!\lt} \; \overline{\mathbf{I}_1} \cap \mathbf{I}_2 \,.

By the assumption that the {T k} kn\{T_k\}_{k \neq n} satisfy causal factorization, this implies that the corresponding time-ordered products commute:

(2)T(I 1I 2¯)T(I 1¯I 2)=T(I 1¯I 2)T(I 1I 2¯). T(\mathbf{I}_1 \cap \overline{\mathbf{I}_2}) \, T(\overline{\mathbf{I}_1} \cap \mathbf{I}_2) \;=\; T(\overline{\mathbf{I}_1} \cap \mathbf{I}_2) \, T(\mathbf{I}_1 \cap \overline{\mathbf{I}_2}) \,.

Using this we find that the identifications of T n+1T_{n+1} on 𝒞 I 1\mathcal{C}_{I_1} and on 𝒞 I 2\mathcal{C}_{I_2}, accrding to (1), agree on the intersection: in that for X𝒞 I 1𝒞 I 2 \mathbf{X} \in \mathcal{C}_{I_1} \cap \mathcal{C}_{I_2} we have

T(I 1)T(I 1¯) =T(I 1I 2)T(I 1I 2¯)T(I 1¯I 2)T(I 1¯I 2¯) =T(I 1I 2)T(I 1¯I 2)T(I 1I 2¯)T(I 1¯I 2¯) =T(I 2)T(I 2¯) \begin{aligned} T( \mathbf{I}_1 ) T( \overline{\mathbf{I}_1} ) & = T( \mathbf{I}_1 \cap \mathbf{I}_2 ) T( \mathbf{I}_1 \cap \overline{\mathbf{I}_2} ) \, T( \overline{\mathbf{I}_1} \cap \mathbf{I}_2 ) T( \overline{\mathbf{I}_1} \cap \overline{\mathbf{I}_2} ) \\ & = T( \mathbf{I}_1 \cap \mathbf{I}_2 ) \underbrace{ T( \overline{\mathbf{I}_1} \cap \mathbf{I}_2 ) T( \mathbf{I}_1 \cap \overline{\mathbf{I}_2} ) } T( \overline{\mathbf{I}_1} \cap \overline{\mathbf{I}_2} ) \\ & = T( \mathbf{I}_2 ) T( \overline{\mathbf{I}_2} ) \end{aligned}

Here in the first step we expanded out the two factors using (1) for I 2I_2, then under the brace we used (2) and in the last step we used again (1), but now for I 1I_1.

To conclude, let

(3)(χ IC cp (Σ n+1),supp(χ I)𝒞 i) I{1,,n+1}I,I¯ \left( \chi_I \in C^\infty_{cp}(\Sigma^{n+1}), \, supp(\chi_I) \subset \mathcal{C}_i \right)_{ { I \subset \{1, \cdots, n+1\} } \atop { I, \overline{I} \neq \emptyset } }

be a partition of unity subordinate to the open cover formed by the 𝒞 I\mathcal{C}_I:

Then the above implies that setting for any XΣ n+1diag(Σ)\mathbf{X} \in \Sigma^{n+1} \setminus diag(\Sigma)

(4)T n+1(X)I{1,,n+1}I,I¯χ i(X)T(I)T(I¯) T_{n+1}(\mathbf{X}) \;\coloneqq\; \underset{ { I \in \{1, \cdots, n+1\} } \atop { I, \overline{I} \neq \emptyset } }{\sum} \chi_i(\mathbf{X}) T( \mathbf{I} ) T( \overline{\mathbf{I}} )

is well defined and satisfies causal factorization.

Remark

(time-ordered products of fixed interaction as distributions)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge-fixed relativistic free vacuum according to this def., and assume that the field bundle is a trivial vector bundle (this example)

and let

gS int+jALoObs(E BV-BRST)[[,g,j]]g,j g S_{int} + j A \;\in\; LoObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle

be a polynomial local observable, to be regarded as a adiabatically switched interaction action functional. This means that there is a finite set

{L int,i,α iΩ Σ p+1,0(E BV-BRST)} i,i \left\{ \mathbf{L}_{int,i}, \mathbf{\alpha}_{i'} \in \Omega^{p+1,0}_\Sigma(E_{\text{BV-BRST}}) \right\}_{i,i'}

of Lagrangian densities which are monomials in the field and jet coordinates, and a corresponding finite set

{g sw,iC cp (Σ)g,j sw,iC cp (Σ)j} \left\{ g_{sw,i} \in C^\infty_{cp}(\Sigma)\langle g \rangle \,,\, j_{sw,i'} \in C^\infty_{cp}(\Sigma)\langle j \rangle \right\}

of adiabatic switchings, such that

gS int+jA=τ Σ(ig sw,iL int,i+ij sw,iα i) g S_{int} + j A \;=\; \tau_{\Sigma} \left( \underset{i}{\sum} g_{sw,i} \mathbf{L}_{int,i} \;+\; \underset{i'}{\sum} j_{sw,i'} \mathbf{\alpha}_{i'} \right)

is the transgression of variational differential forms (this def.) of the sum of the products of these adiabatic switching with these Lagrangian densities.

In order to discuss the S-matrix 𝒮(gS int+jA)\mathcal{S}(g S_{int} + j A) and hence the time-ordered products of the special form T k(gS int+jA,,gS int+jAkfactors)T_k\left( \underset{k \, \text{factors}}{\underbrace{g S_{int} + j A, \cdots, g S_{int} + j A }} \right) it is sufficient to restrict attention to the restriction of each T kT_k to the subspace of local observables induced by the finite set of Lagrangian densities {L int,i,α i} i,i\{\mathbf{L}_{int,i}, \mathbf{\alpha}_{i'}\}_{i,i'}.

This restriction is a continuous linear functional on the corresponding space of bump functions {g sw,i,j sw,i}\{g_{sw,i}, j_{sw,i'}\}, hence a dstributional section of a corresponding trivial vector bundle.

In terms of this, prop. 1 says that the choice of time-ordered products T kT_k is inductively in kk a choice of extension of distributions to the diagonal.

If Σ= p,1\Sigma = \mathbb{R}^{p,1} is Minkowski spacetime and we impose the renormalization condition “translation invariance” (this def.) then each T kT_k is a distribution on Σ k1= (p+1)(k1)\Sigma^{k-1} = \mathbb{R}^{(p+1)(k-1)} and the extension of distributions is from the complement of the origina 0 (p+1)(k1)0 \in \mathbb{R}^{(p+1)(k-1)}.

Therefore we now discuss extension of distributions (def. 3 below) on Cartesian spaces from the complement of the origin to the origin. Since the space of choices of such extensions turns out to depend on the scaling degree of distributions, we first discuss that (def. 2 below).

Definition

(rescaled distribution)

Let nn \in \mathbb{N}. For λ(0,)\lambda \in (0,\infty) \subset \mathbb{R} a positive real number write

n s λ n x λx \array{ \mathbb{R}^n &\overset{s_\lambda}{\longrightarrow}& \mathbb{R}^n \\ x &\mapsto& \lambda x }

for the diffeomorphism given by multiplication with λ\lambda, using the canonical real vector space-structure of n\mathbb{R}^n.

Then for u𝒟( n)u \in \mathcal{D}'(\mathbb{R}^n) a distribution on the Cartesian space n\mathbb{R}^n the rescaled distribution is the pullback of uu along m λm_\lambda

u λs λ *u𝒟( n). u_\lambda \coloneqq s_\lambda^\ast u \;\in\; \mathcal{D}'(\mathbb{R}^n) \,.

Explicitly, this is given by

𝒟( n) u λ, b λ nu,b(λ 1()). \array{ \mathcal{D}(\mathbb{R}^n) &\overset{ \langle u_\lambda, - \rangle}{\longrightarrow}& \mathbb{R} \\ b &\mapsto& \lambda^{-n} \langle u , b(\lambda^{-1}\cdot (-))\rangle } \,.

Similarly for X nX \subset \mathbb{R}^n an open subset which is invariant under s λs_\lambda, the rescaling of a distribution u𝒟(X)u \in \mathcal{D}'(X) is is u λs λ *uu_\lambda \coloneqq s_\lambda^\ast u.

Definition

(scaling degree of a distribution)

Let nn \in \mathbb{N} and let X nX \subset \mathbb{R}^n be an open subset of Cartesian space which is invariant under rescaling s λs_\lambda (def. 1) for all λ(0,)\lambda \in (0,\infty), and let u𝒟(X)u \in \mathcal{D}'(X) be a distribution on this subset. Then

  1. The scaling degree of uu is the infimum

    sd(u)inf{ω|limλ0λ ωu λ=0} sd(u) \;\coloneqq\; inf \left\{ \omega \in \mathbb{R} \;\vert\; \underset{\lambda \to 0}{\lim} \lambda^\omega u_\lambda = 0 \right\}

    of the set of real numbers ω\omega such that the limit of the rescaled distribution λ ωu λ\lambda^\omega u_\lambda (def. 1) vanishes. If there is no such ω\omega one sets sd(u)sd(u) \coloneqq \infty.

  2. The degree of divergence of uu is the difference of the scaling degree by the dimension of the underlying space:

deg(u)sd(u)n. deg(u) \coloneqq sd(u) - n \,.
Example

(scaling degree of non-singular distributions)

If u=u fu = u_f is a non-singular distribution given by bump function fC (X)𝒟(X)f \in C^\infty(X) \subset \mathcal{D}'(X), then its scaling degree (def. 2) is non-positive

sd(u f)0. sd(u_f) \leq 0 \,.

Specifically if the first non-vanishing partial derivative αf(0)\partial_\alpha f(0) of ff at 0 occurs at order |α|{\vert \alpha\vert} \in \mathbb{N}, then the scaling degree of u fu_f is |α|-{\vert \alpha\vert}.

Proof

By definition we have for bC cp ( n)b \in C^\infty_{cp}(\mathbb{R}^n) any bump function that

λ ω(u f) λ,n =λ ωn nf(x)g(λ 1x)d nx =λ ω nf(λx)g(x)d nx, \begin{aligned} \left\langle \lambda^{\omega} (u_f)_\lambda, n \right\rangle & = \lambda^{\omega-n} \underset{\mathbb{R}^n}{\int} f(x) g(\lambda^{-1} x) d^n x \\ & = \lambda^{\omega} \underset{\mathbb{R}^n}{\int} f(\lambda x) g(x) d^n x \end{aligned} \,,

where in last line we applied change of integration variables.

The limit of this expression is clearly zero for all ω>0\omega \gt 0, which shows the first claim.

If moreover the first non-vanishing partial derivative of ff occurs at order |α|=k{\vert \alpha \vert} = k, then Hadamard's lemma says that ff is of the form

f(x)=(iα i!) 1( αf(0))i(x i) α i+β n|β|=|α|+1i(x i) β ih β(x) f(x) \;=\; \left( \underset{i}{\prod} \alpha_i ! \right)^{-1} (\partial_\alpha f(0)) \underset{i}{\prod} (x^i)^{\alpha_i} + \underset{ {\beta \in \mathbb{N}^n} \atop { {\vert \beta\vert} = {\vert \alpha \vert} + 1 } }{\sum} \underset{i}{\prod} (x^i)^{\beta_i} h_{\beta}(x)

where the h βh_{\beta} are smooth functions. Hence in this case

λ ω(u f) λ,n =λ ω+|α| n(iα i!) 1( αf(0))i(x i) α ib(x)d nx =+λ ω+|α|+1 ni(x i) β ih β(x)b(x)d nx. \begin{aligned} \left\langle \lambda^{\omega} (u_f)_\lambda, n \right\rangle & = \lambda^{\omega + {\vert \alpha\vert }} \underset{\mathbb{R}^n}{\int} \left( \underset{i}{\prod} \alpha_i ! \right)^{-1} (\partial_\alpha f(0)) \underset{i}{\prod} (x^i)^{\alpha_i} b(x) d^n x \\ & \phantom{=} + \lambda^{\omega + {\vert \alpha\vert} + 1} \underset{\mathbb{R}^n}{\int} \underset{i}{\prod} (x^i)^{\beta_i} h_{\beta}(x) b(x) d^n x \end{aligned} \,.

This makes manifest that the expression goes to zero with λ0\lambda \to 0 precisely for ω>|α|\omega \gt - {\vert \alpha \vert}, which means that

sd(u f)=|α| sd(u_f) = -{\vert \alpha \vert}

in this case.

Example

(scaling degree of derivatives of delta-distributions)

Let α n\alpha \in \mathbb{N}^n be a multi-index and αδ𝒟(X)\partial_\alpha \delta \in \mathcal{D}'(X) the corresponding partial derivatives of the delta distribution δ 0𝒟( n)\delta_0 \in \mathcal{D}'(\mathbb{R}^n) supported at 00. Then the degree of divergence (def. 2) of αδ 0\partial_\alpha \delta_0 is the total order the derivatives

deg( αδ 0)=|α| deg\left( {\, \atop \,} \partial_\alpha\delta_0{\, \atop \,} \right) \;=\; {\vert \alpha \vert}

where |α|iα i{\vert \alpha\vert} \coloneqq \underset{i}{\sum} \alpha_i.

Proof

By definition we have for bC cp ( n)b \in C^\infty_{cp}(\mathbb{R}^n) any bump function that

λ ω( αδ 0) λ,b =(1) |α|λ ωn( |α| α 1x 1 α nx nb(λ 1x)) |x=0 =(1) |α|λ ωn|α| |α| α 1x 1 α nx nb(0), \begin{aligned} \left\langle \lambda^\omega (\partial_\alpha \delta_0)_\lambda, b \right\rangle & = (-1)^{{\vert \alpha \vert}} \lambda^{\omega-n} \left( \frac{ \partial^{{\vert \alpha \vert}} }{ \partial^{\alpha_1} x^1 \cdots \partial^{\alpha_n}x^n } b(\lambda^{-1}x) \right)_{\vert x = 0} \\ & = (-1)^{{\vert \alpha \vert}} \lambda^{\omega - n - {\vert \alpha\vert}} \frac{ \partial^{{\vert \alpha \vert}} }{ \partial^{\alpha_1} x^1 \cdots \partial^{\alpha_n}x^n } b(0) \end{aligned} \,,

where in the last step we used the chain rule of differentiation. It is clear that this goes to zero with λ\lambda as long as ω>n+|α|\omega \gt n + {\vert \alpha\vert}. Hence sd( αδ 0)=n+|α|sd(\partial_{\alpha} \delta_0) = n + {\vert \alpha \vert}.

Example

(scaling degree of Feynman propagator on Minkowski spacetime)

Let

Δ F(x)=limϵ(0,)ϵ0+i(2π) p+1 e ik μx μk μk μ(mc) 2+iϵdk 0d pk \Delta_F(x) \;=\; \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu x^\mu} }{ - k_\mu k^\mu - \left( \tfrac{m c}{\hbar} \right)^2 + i \epsilon } \, d k_0 \, d^p \vec k

be the Feynman propagator for the massive free real scalar field on n=p+1n = p+1-dimensional Minkowski spacetime (this prop.). Its scaling degree is

sd(Δ F) =n2 =p1. \begin{aligned} sd(\Delta_{F}) & = n - 2 \\ & = p -1 \end{aligned} \,.

(Brunetti-Fredenhagen 00, example 3 on p. 22)

Proof

Regarding Δ F\Delta_F as a generalized function via the given Fourier-transform expression, we find by change of integration variables in the Fourier integral that in the scaling limit the Feynman propagator becomes that for vannishing mass, which scales homogeneously:

limλ0(λ ωΔ F(λx)) =limλ0(lamba ωlimϵ(0,)ϵ0+i(2π) p+1 e ik μλx μk μk μ(mc) 2+iϵdk 0d pk) =limλ0(λ ωnlimϵ(0,)ϵ0+i(2π) p+1 e ik μλx μ(λ 2)k μk μ(mc) 2+iϵdk 0d pk) =limλ0(λ ωn+2limϵ(0,)ϵ0+i(2π) p+1 e ik μλx μk μk μ+iϵdk 0d pk). \begin{aligned} \underset{\lambda \to 0}{\lim} \left( \lambda^\omega \; \Delta_F(\lambda x) \right) & = \underset{\lambda \to 0}{\lim} \left( \lamba^{\omega} \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu \lambda x^\mu} }{ - k_\mu k^\mu - \left( \tfrac{m c}{\hbar} \right)^2 + i \epsilon } \, d k_0 \, d^p \vec k \right) \\ & = \underset{\lambda \to 0}{\lim} \left( \lambda^{\omega-n} \; \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu \lambda x^\mu} }{ - (\lambda^{-2}) k_\mu k^\mu - \left( \tfrac{m c}{\hbar} \right)^2 + i \epsilon } \, d k_0 \, d^p \vec k \right) \\ & = \underset{\lambda \to 0}{\lim} \left( \lambda^{\omega-n + 2 } \; \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu \lambda x^\mu} }{ - k_\mu k^\mu + i \epsilon } \, d k_0 \, d^p \vec k \right) \,. \end{aligned}
Proposition

(basic properties of scaling degree of distributions)

Let X nX \subset \mathbb{R}^n and u𝒟(X)u \in \mathcal{D}'(X) be a distribution as in def. 1, such that its scaling degree is finite: sd(u)<sd(u) \lt \infty (def. 2). Then

  1. For α n\alpha \in \mathbb{N}^n, the partial derivative of distributions α\partial_\alpha increases scaling degree at most by |α|{\vert \alpha\vert }:

    deg( αu)deg(u)+|α| deg(\partial_\alpha u) \;\leq\; deg(u) + {\vert \alpha\vert}
  2. For α n\alpha \in \mathbb{N}^n, the product of distributions with the smooth coordinate functions x αx^\alpha decreases scaling degree at least by |α|{\vert \alpha\vert }:

    deg(x αu)deg(u)|α| deg(x^\alpha u) \;\leq\; deg(u) - {\vert \alpha\vert}
  3. Under tensor product of distributions their scaling degrees add:

    sd(uv)sd(u)+sd(v) sd(u \otimes v) \leq sd(u) + sd(v)

    for v𝒟(Y)v \in \mathcal{D}'(Y) another distribution on Y nY \subset \mathbb{R}^{n'};

  4. deg(fu)deg(u)kdeg(f u) \leq deg(u) - k for fC (X)f \in C^\infty(X) and f (α)(0)=0f^{(\alpha)}(0) = 0 for |α|k1{\vert \alpha\vert} \leq k-1;

(Brunetti-Fredenhagen 00, lemma 5.1, Dütsch 18, exercise 3.34)

Proof

The first three statements follow with manipulations as in example 1 and example 2.

For the fourth…

Proposition

(scaling degree of product distribution)

Let u,v𝒟( n)u,v \in \mathcal{D}'(\mathbb{R}^n) be two distributions such that

  1. both have finite degree of divergence (def. 2)

    deg(u),deg(v)< deg(u), deg(v) \lt \infty
  2. their product of distributions is well-defined

    uv𝒟( n) u v \in \mathcal{D}'(\mathbb{R}^n)

    (in that their wave front sets satisfy Hörmander's criterion)

then the product distribution has degree of divergence bounded by the sum of the separate degrees:

deg(uv)deg(u)+deg(v). deg(u v) \;\leq\; deg(u) + deg(v) \,.

With the concept of scaling degree of distributions in hand, we may now discuss extension of distributions:

Definition

(extension of distributions)

Let XιX^X \overset{\iota}{\subset} \hat X be an inclusion of open subsets of some Cartesian space. This induces the operation of restriction of distributions

𝒟(X^)ι *𝒟(X). \mathcal{D}'(\hat X) \overset{\iota^\ast}{\longrightarrow} \mathcal{D}'(X) \,.

Given a distribution u𝒟(X)u \in \mathcal{D}'(X), then an extension of uu to X^\hat X is a distribution u^𝒟(X^)\hat u \in \mathcal{D}'(\hat X) such that

ι *u^=u. \iota^\ast \hat u \;=\; u \,.
Proposition

(unique extension of distributions with negative degree of divergence)

For nn \in \mathbb{N}, let u𝒟( n{0})u \in \mathcal{D}'(\mathbb{R}^n \setminus \{0\}) be a distribution on the complement of the origin, with negative degree of divergence at the origin

deg(u)<0. deg(u) \lt 0 \,.

Then uu has a unique extension of distributions u^𝒟( n)\hat u \in \mathcal{D}'(\mathbb{R}^n) to the origin with the same degree of divergence

deg(u^)=deg(u). deg(\hat u) = deg(u) \,.

(Brunetti-Fredenhagen 00, theorem 5.2, Dütsch 18, theorem 3.35 a))

Proof

Regarding uniqueness:

Suppose u^\hat u and u^ {\hat u}^\prime are two extensions of uu with deg(u^)=deg(u^ )deg(\hat u) = deg({\hat u}^\prime). Both being extensions of a distribution defined on n{0}\mathbb{R}^n \setminus \{0\}, this difference has support at the origin {0} n\{0\} \subset \mathbb{R}^n. By this prop. this implies that it is a linear combination of derivatives of the delta distribution supported at the origin:

u^ u^=α nc α αδ 0 {\hat u}^\prime - \hat u \;=\; \underset{ {\alpha \in \mathbb{N}^n} }{\sum} c^\alpha \partial_\alpha \delta_0

for constants c αc^\alpha \in \mathbb{C}. But by this example the degree of divergence of these point-supported distributions is non-negative

deg( αδ 0)=|α|0. deg( \partial_\alpha \delta_0) = {\vert \alpha\vert} \geq 0 \,.

This implies that c α=0c^\alpha = 0 for all α\alpha, hence that the two extensions coincide.

Regarding existence:

Let

bC cp ( n) b \in C^\infty_{cp}(\mathbb{R}^n)

be a bump function which is 1\leq 1 and constant on 1 over a neighbourhood of the origin. Write

χ1bC ( n) \chi \coloneqq 1 - b \;\in\; C^\infty(\mathbb{R}^n)

graphics grabbed from Dütsch 18, p. 108

and for λ(0,)\lambda \in (0,\infty) a positive real number, write

χ λ(x)χ(λx). \chi_\lambda(x) \coloneqq \chi(\lambda x) \,.

Since the product χ λu\chi_\lambda u has support of a distribution on a complement of a neighbourhood of the origin, we may extend it by zero to a distribution on all of n\mathbb{R}^n, which we will denote by the same symbols:

χ λu𝒟( n). \chi_\lambda u \in \mathcal{D}'(\mathbb{R}^n) \,.

By construction xhi λu\xhi_\lambda u coincides with uu away from a neighbourhood of the origin, which moreover becomes arbitrarily small as λ\lambda increases. This means that if the following limit exists

u^limλχ λu \hat u \;\coloneqq\; \underset{\lambda \to \infty}{\lim} \chi_\lambda u

then it is an extension of uu.

To see that the limit exists, it is sufficient to observe that we have a Cauchy sequence, hence that for all bC cp ( n)b\in C^\infty_{cp}(\mathbb{R}^n) the difference

(χ n+1uχ nu)(b)=u(b)(χ n+1+χ n) (\chi_{n+1} u - \chi_n u)(b) \;=\; u(b)( \chi_{n+1} + \chi_n )

becomes arbitrarily small.

It remains to see that the unique extension u^\hat u thus established has the same scaling degree as uu. This is shown in (Brunetti-Fredenhagen 00, p. 24).

Proposition

(space of point-extensions of distributions)

For nn \in \mathbb{N}, let u𝒟( n{0})u \in \mathcal{D}'(\mathbb{R}^n \setminus \{0\}) be a distribution of degree of divergence deg(u)<deg(u) \lt \infty.

Then uu does admit at least one extension (def. 3) to a distribution u^𝒟( n)\hat u \in \mathcal{D}'(\mathbb{R}^n), and every choice of extension has the same degree of divergence as uu

deg(u^)=deg(u). deg(\hat u) = deg(u) \,.

Moreover, any two such extensions u^\hat u and u^ {\hat u}^\prime differ by a linear combination of partial derivatives of distributions of order deg(u)\leq deg(u) of the delta distribution δ 0\delta_0 supported at the origin:

u^ u^=α n|α|deg(u)q α αδ 0, {\hat u}^\prime - \hat u \;=\; \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha\vert} \leq deg(u) } }{\sum} q^\alpha \partial_\alpha \delta_0 \,,

for a finite number of constants q αq^\alpha \in \mathbb{C}.

This is essentially (Hörmander 90, thm. 3.2.4). We follow (Brunetti-Fredenhagen 00, theorem 5.3), which was inspired by (Epstein-Glaser 73, section 5). Review of this approach is in (Dütsch 18, theorem 3.35 (b)), see also remark 2 below.

Proof

For fC ( n)f \in C^\infty(\mathbb{R}^n) a smooth function, and ρ\rho \in \mathbb{N}, we say that ff vanishes to order ρ\rho at the origin if all partial derivatives with multi-index α n\alpha \in \mathbb{N}^n of total order |α|ρ{\vert \alpha\vert} \leq \rho vanish at the origin:

αf(0)=0AAA|α|ρ. \partial_\alpha f (0) = 0 \phantom{AAA} {\vert \alpha\vert} \leq \rho \,.

By Hadamard's lemma, such a function may be written in the form

(5)f(x)=α n|α|=ρ+1x αr α(x) f(x) \;=\; \underset{ {\alpha \in \mathbb{N}^n} \atop { {\vert \alpha \vert} = \rho + 1 } }{\sum} x^\alpha r_\alpha(x)

for smooth functions r αC cp ( n)r_\alpha \in C^\infty_{cp}(\mathbb{R}^n).

Write

𝒟 ρ( n)𝒟( n)C cp ( n) \mathcal{D}_\rho(\mathbb{R}^n) \hookrightarrow \mathcal{D}(\mathbb{R}^n) \coloneqq C^\infty_{cp}(\mathbb{R}^n)

for the subspace of that of all bump functions on those that vanish to order ρ\rho at the origin.

By definition this is equivalently the joint kernel of the partial derivatives of distributions of order |α|{\vert \alpha\vert} of the delta distribution δ 0\delta_0 supported at the origin:

b𝒟 ρ( n)AAAAα n|α|ρ αδ 0,b=0. b \in \mathcal{D}_\rho(\mathbb{R}^n) \phantom{AA} \Leftrightarrow \phantom{AA} \underset{ {\alpha \in \mathbb{N}^n} \atop { {\vert \alpha\vert} \leq \rho } } {\forall} \left\langle \partial_\alpha \delta_0, b \right\rangle = 0 \,.

Therefore every continuous linear projection

(6)p ρ:𝒟( n)𝒟 ρ( n) p_\rho \;\colon\; \mathcal{D}(\mathbb{R}^n) \longrightarrow \mathcal{D}_\rho(\mathbb{R}^n)

may be obtained from a choice of dual basis to the { αδ 0}\{\partial_\alpha \delta_0\}, hence a choice of smooth functions

{w βC cp ( n)} β n|β|ρ \left\{ w^\beta \in C^\infty_{cp}(\mathbb{R}^n) \right\}_{ { \beta \in \mathbb{N}^n } \atop { {\vert \beta\vert} \leq \rho } }

such that

αδ 0,w β=δ α βAAAAAA αw β(0)=δ α βAAAAfor|α|ρ, \left\langle \partial_\alpha \delta_0 \,,\, w^\beta \right\rangle \;=\; \delta_\alpha^\beta \phantom{AAA} \Leftrightarrow \phantom{AAA} \partial_\alpha w^\beta(0) \;=\; \delta_\alpha^\beta \phantom{AAAA} \text{for}\, {\vert \alpha\vert} \leq \rho \,,

by setting

(7)p ρidα n|α|ρw α αδ 0,(), p_\rho \;\coloneqq\; id \;-\; \left\langle \underset{ {\alpha \in \mathbb{N}^n} \atop { {\vert \alpha\vert} \leq \rho } }{\sum} w^\alpha \partial_\alpha \delta_0 \,,\, (-) \right\rangle \,,

hence

p ρ:bbα n|α|ρ(1) |α|w α αb(0). p_\rho \;\colon\; b \mapsto b - \underset{ {\alpha \in \mathbb{N}^n} \atop { {\vert \alpha\vert} \leq \rho } }{\sum} (-1)^{{\vert \alpha\vert}} w^\alpha \partial_\alpha b(0) \,.

Together with Hadamard's lemma in the form (5) this means that every b𝒟( n)b \in \mathcal{D}(\mathbb{R}^n) is decomposed as

(8)b(x) =p ρ(b)(x)+(idp ρ)(b)(x) =α n|α|=ρ+1x αr α(x)+α n|α|ρ(1) |α|w α αb(0) \begin{aligned} b(x) & = p_\rho(b)(x) \;+\; (id - p_\rho)(b)(x) \\ & = \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha \vert} = \rho + 1 } }{\sum} x^\alpha r_\alpha(x) \;+\; \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha \vert} \leq \rho } }{\sum} (-1)^{{\vert \alpha \vert}} w^\alpha \partial_\alpha b(0) \end{aligned}

Now let

ρdeg(u). \rho \;\coloneqq\; deg(u) \,.

Observe that (by this prop.) the degree of divergence of the product of distributions x αux^\alpha u with |α|=ρ+1{\vert \alpha\vert} = \rho + 1 is negative

deg(x αu) =ρ|α|1 \begin{aligned} deg\left( x^\alpha u \right) & = \rho - {\vert \alpha \vert} \leq -1 \end{aligned}

Therefore prop. 4 says that each x αux^\alpha u for |α|=ρ+1{\vert \alpha\vert} = \rho + 1 has a unique extension x αu^\widehat{ x^\alpha u} to the origin. Accordingly the composition up ρu \circ p_\rho has a unique extension, by (8):

(9)u^,b =u^,p ρ(b)+u^,(idp ρ)(b) =α n|α|=ρ+1x αu^,r αunique+α n|α|ρu^,w αq αchoice αδ 0,b \begin{aligned} \left\langle \hat u \,,\, b \right\rangle & = \left\langle \hat u , p_\rho(b) \right\rangle + \left\langle \hat u , (id - p_\rho)(b) \right\rangle \\ & = \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha \vert} = \rho + 1 } }{\sum} \underset{ \text{unique} }{ \underbrace{ \left\langle \widehat{x^\alpha u} \,,\, r_\alpha \right\rangle } } \;+\; \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha\vert} \leq \rho } }{\sum} \underset{ { q^\alpha } \atop { \text{choice} } }{ \underbrace{ \langle \hat u \,,\, w^\alpha \rangle } } \left\langle \partial_\alpha \delta_0 \,,\, b \right\rangle \end{aligned}

That says that u^\hat u is of the form

u^=up ρ^unique+α n|α|ρc α αδ 0 \hat u \;=\; \underset{ \text{unique} }{ \underbrace{ \widehat{ u \circ p_\rho } } } + \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha\vert} \leq \rho } }{\sum} c^\alpha \, \partial_\alpha \delta_0

for a finite number of constants c αc^\alpha \in \mathbb{C}.

Notice that for any extension u^\hat u the exact value of the c αc^\alpha here depends on the arbitrary choice of dual basis {w α}\{w^\alpha\} used for this construction. But the uniqueness of the first summand means that for any two choices of extensions u^\hat u and u^ {\hat u}^\prime, their difference is of the form

u^ u^=α n|α|ρ((c) αc α) αδ 0, {\hat u}^\prime - \hat u \;=\; \underset{ { \alpha \in \mathbb{N}^n } \atop { {\vert \alpha\vert} \leq \rho } }{\sum} ( (c')^\alpha - c^\alpha ) \, \partial_\alpha \delta_0 \,,

where the constants q α((c) αc α)q^\alpha \coloneqq ( (c')^\alpha - c^\alpha ) \in \mathbb{C} are independent of any choices.

It remains to see that all these u^\hat u in fact have the same degree of divergence as uu.

By this example the degree of divergence of the point-supported distributions on the right is deg( αδ 0)=|α|ρdeg(\partial_\alpha \delta_0) = {\vert \alpha\vert} \leq \rho.

Therefore to conclude it is now sufficient to show that

deg(up ρ^)=ρ. deg\left( \widehat{ u \circ p_\rho } \right) \;=\; \rho \,.

This is shown in (Brunetti-Fredenhagen 00, p. 25).

Remark

(“W-extensions”)

Since in Brunetti-Fredenhagen 00, (38) the projectors (7) are denoted “WW”, the construction of extensions of distributions via the proof of prop. 5 has come to be called “W-extensions” (e.g Dütsch 18).

In conclusion we obtain the central theorem of causal perturbation theory:

Theorem

(existence and choices of ("re"-)normalization of S-matrices/perturbative QFTs)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge-fixed relativistic free vacuum, according to this def., such that the underlying spacetime is Minkowski spacetime and the Wightman propagator Δ H\Delta_H is translation-invariant.

Then:

  1. an S-matrix scheme 𝒮\mathcal{S} (this def.) around this vacuum exists;

  2. for gS int+jALocObs(E BV-BRST)[[,g,j]]g,jg S_{int} + j A \in LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle a local observable, regarded as an adiabatically switched interaction action functional, the space of possible choices of S-matrices

    𝒮(gS int+jA)PolyObs(E BV-BRST) mc[[,g,j]] \mathcal{S}(g S_{int} + j A) \;\in\; PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]

    hence of the corresponding perturbative QFTs, by this prop., is, inductively in kk \in \mathbb{N}, a finite dimensional affine space, parameterizing the extension of the time-ordered product T kT_k to the locus of coinciding interaction points.

Proof

By prop. 3 the Feynman propagator is finite scaling degree of a distribution, so that by prop. 3 the binary time-ordered product away from the diagonal T 2(,)| Σ 2diag(Σ)=() F()T_2(-,-)\vert_{\Sigma^2 \setminus diag(\Sigma)} = (-) \star_{F} (-) has finite scaling degree.

By prop. 3 this implies that in the inductive description of the time-ordered products by prop. 1, each induction step is the extension of distributions of finite scaling degree of a distribution to the point. By prop. 5 this always exists.

This proves the first statement.

Now if a polynomial local interaction is fixed, then via remark 1 each induction step involved extending a finite number of distributions, each of finite scaling degree. By prop. 5 the corresponding space of choices is in each step a finite-dimensional affine space.

\,

Stückelberg-Petermann renormalization group

A genuine re-normalization is the passage from one S-matrix ("re"-)normalization scheme 𝒮\mathcal{S} to another such scheme 𝒮\mathcal{S}'. The inductive Epstein-Glaser ("re"-normalization) construction (prop. 1) shows that the difference between any 𝒮\mathcal{S} and 𝒮\mathcal{S}' is inductively in kk \in \mathbb{N} a choice of extra term in the time-ordered product of kk factors, equivalently in the Feynman amplitudes for Feynman diagrams with kk vertices, that contributes when all kk of these vertices coincide in spacetime (prop. 5).

A natural question is whether these additional interactions that appear when several interaction vertices coincide may be absorbed into a re-definition of the original interaction gS int+jAg S_{int} + j A. Such an interaction vertex redefinition (def. 4 below)

𝒵:gS int+jAgS int+jA+higher order corrections \mathcal{Z} \;\colon\; g S_{int} + j A \;\mapsto\; g S_{int} + j A \;+\; \text{higher order corrections}

should perturbatively send local interactions to local interactions with higher order corrections.

The main theorem of perturbative renormalization (theorem 2 below) says that indeed under mild conditions every re-normalization 𝒮𝒮\mathcal{S} \mapsto \mathcal{S}' is induced by such an interaction vertex redefinition in that there exists a unique such redefinition 𝒵\mathcal{Z} so that for every local interaction gS int+jAg S_{int} + j A we have that scattering amplitudes for the interaction gS int+jAg S_{int} + j A computed with the ("re"-)normalization scheme 𝒮\mathcal{S}' equal those computed with 𝒮\mathcal{S} but applied to the re-defined interaction 𝒵(gS int+jA)\mathcal{Z}(g S_{int} + j A):

𝒮(gS int+jA)=𝒮(𝒵(gS int+jA)). \mathcal{S}' \left( {\, \atop \,} g S_{int} + j A {\, \atop \,} \right) \;=\; \mathcal{S}\left( {\, \atop \,} \mathcal{Z}(g S_{int} + j A) {\, \atop \,} \right) \,.

This means that the interaction vertex redefinitions 𝒵\mathcal{Z} form a group under composition which acts transitively and freely, hence regularly, on the set of S-matrix ("re"-)normalization schemes; this is called the Stückelberg-Petermann renormalization group (theorem 2 below).

\,

Definition

(perturbative interaction vertex redefinition)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H) be a gauge fixed free field vacuum (this def.).

A perturbative interaction vertex redefinition (or just vertex redefinition, for short) is an endofunction

𝒵:LocObs(E BV-BRST)[[,g,j]]g,jLocObs(E BV-BRST)[[,g,j]]g,j \mathcal{Z} \;\colon\; LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle \longrightarrow LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle

on local observables with formal parameters adjoined (this def.) such that there exists a sequence {Z k} k\{Z_k\}_{k \in \mathbb{N}} of continuous linear functionals, symmetric in their arguments, of the form

(LocObs(E BV-BRST)[[,g,j]]g,j) [[,g,j]] kLocObs(E BV-BRST)[[,g,j]]g,j \left( {\, \atop \,} LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle {\, \atop \,} \right)^{\otimes^k_{\mathbb{C}[ [ \hbar, g, j] ]}} \longrightarrow LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle

such that for all gS int+jALocObs(E BV-BRST)[[,g,j]]g,jg S_{int} + j A \in LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle the following conditions hold:

  1. (perturbation)

    1. Z 0(gS int+jA)=0Z_0(g S_{int + j A}) = 0

    2. Z 1(gS int+jA)=gS int+jAZ_1(g S_{int} + j A) = g S_{int} + j A

    3. and

      𝒵(gS int+jA) =Zexp (gS int+jA) k1k!Z k(gS int+jA,,gS int+jAkargs) \begin{aligned} \mathcal{Z}(g S_{int} + j A) & = Z \exp_\otimes( g S_{int} + j A ) \\ & \coloneqq \underset{k \in \mathbb{N}}{\sum} \frac{1}{k!} Z_k( \underset{ k \, \text{args} }{ \underbrace{ g S_{int} + j A , \cdots, g S_{int} + j A } } ) \end{aligned}
  2. (field independence) The local observable 𝒵(gS int+jA)\mathcal{Z}(g S_{int} + j A) depends on the field histories only through its argument gS int+jAg S_{int} + j A , hence by the chain rule:

    (10)δδΦ a(x)𝒵(gS int+jA)=𝒵 gS int+jA(δδΦ a(x)(gS int+jA)) \frac{\delta}{\delta \mathbf{\Phi}^a(x)} \mathcal{Z}(g S_{int} + j A) \;=\; \mathcal{Z}'_{g S_{int} + j A} \left( \frac{\delta}{\delta \mathbf{\Phi}^a(x)} (g S_{int} + j A) \right)

The following proposition should be compared to the axiom of causal additivity of the S-matrix scheme (this equation):

Proposition

(local additivity of vertex redefinitions)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H) be a gauge fixed free field vacuum (this def.) and let 𝒵\mathcal{Z} be a vertex redefinition (def. 4).

Then for all local observables O 0,O 1,O 2LocObs(E BV-BRST)[[,g,j]]g,jO_0, O_1, O_2 \in LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j ] ]\langle g, j\rangle with spacetime support denoted supp(O i)Σsupp(O_i) \subset \Sigma (this def.) we have

  1. (local additivity)

    (supp(O 1)supp(O 2)=) AA𝒵(O 0+O 1+O 2)=𝒵(O 0+O 1)𝒵(O 0)+𝒵(O 0+O 2). \begin{aligned} & \left( supp(O_1) \cap supp(O_2) = \emptyset \right) \\ & \Rightarrow \phantom{AA} \mathcal{Z}( O_0 + O_1 + O_2) = \mathcal{Z}( O_0 + O_1 ) - \mathcal{Z}(O_0) + \mathcal{Z}(O_0 + O_2) \end{aligned} \,.
  2. (preservation of spacetime support)

    supp(𝒵(O 0+O 1)𝒵(O 0))supp(O 1) supp \left( {\, \atop \,} \mathcal{Z}(O_0 + O_1) - \mathcal{Z}(O_0) {\, \atop \,} \right) \;\subset\; supp(O_1)

    hence in particular

    supp(𝒵(O 1))=supp(O 1) supp \left( {\, \atop \,} \mathcal{Z}(O_1) {\, \atop \,} \right) = supp(O_1)

(Dütsch 18, exercise 3.98)

Proof

Under the inclusion

LocObs(E BV-BRST)PolyObs(E BV-BRST) LocObs(E_{\text{BV-BRST}}) \hookrightarrow PolyObs(E_{\text{BV-BRST}})

of local observables into polynomial observables we may think of each Z kZ_k as a generalized function, as for time-ordered products in this remark.

Hence if

O j=Σj Σ (L j) O_j = \underset{\Sigma}{\int} j^\infty_\Sigma( \mathbf{L}_j )

is the transgression of a Lagrangian density L\mathbf{L} we get

Z k((O 1+O 2+O 3),,(O 1+O 2+O 3))=j 1,,j k{0,1,2}Σ kZ(L j 1(x 1),,L j k(x k)). Z_k( (O_1 + O_2 + O_3) , \cdots , (O_1 + O_2 + O_3) ) = \underset{ j_1, \cdots, j_k \in \{0,1,2\} }{\sum} \underset{\Sigma^{k}}{\int} Z( \mathbf{L}_{j_1}(x_1) , \cdots , \mathbf{L}_{j_k}(x_k) ) \,.

Now by definition Z k()Z_k(\cdots) is in the subspace of local observables, i.e. those polynomial observables whose coefficient distributions are supported on the diagonal, which means that

δδΦ a(x)δδΦ b(y)Z k()=0AAforAAxy \frac{\delta}{\delta \mathbf{\Phi}^a(x)} \frac{\delta}{\delta \mathbf{\Phi}^b(y)} Z_{k}(\cdots) = 0 \phantom{AA} \text{for} \phantom{AA} x \neq y

Together with the axiom “field independence” (10) this means that the support of these generalized functions in the integrand here must be on the diagonal, where x 1==x kx_1 = \cdots = x_k.

By the assumption that the spacetime supports of O 1O_1 and O 2O_2 are disjoint, this means that only the summands with j 1,,j k{0,1}j_1, \cdots, j_k \in \{0,1\} and those with j 1,,j k{0,2}j_1, \cdots, j_k \in \{0,2\} contribute to the above sum. Removing the overcounting of those summands where all j 1,,j k{0}j_1, \cdots, j_k \in \{0\} we get

Z k((O 1+O 2+O 3),,(O 1+O 2+O 3)) =j 1,,j k{0,1}Σ kZ(L j 1(x 1),,L j k(x k)) =j 1,,j k{0}Σ kZ(L j 1(x 1),,L j k(x k)) =j 1,,j k{0,2}Σ kZ(L j 1(x 1),,L j k(x k)) =Z k((O 0+O 1),,(O 0+O 1))Z k(O 0,,O 0)+Z k((O 0+O 2),,(O 0+O 2)). \begin{aligned} & Z_k\left( {\, \atop \,} (O_1 + O_2 + O_3) , \cdots , (O_1 + O_2 + O_3) {\, \atop \,} \right) \\ & = \underset{ j_1, \cdots, j_k \in \{0,1\} }{\sum} \underset{\Sigma^{k}}{\int} Z( \mathbf{L}_{j_1}(x_1) , \cdots , \mathbf{L}_{j_k}(x_k) ) \\ & \phantom{=} - \underset{ j_1, \cdots, j_k \in \{0\} }{\sum} \underset{\Sigma^{k}}{\int} Z( \mathbf{L}_{j_1}(x_1) , \cdots , \mathbf{L}_{j_k}(x_k) ) \\ & \phantom{=} - \underset{ j_1, \cdots, j_k \in \{0,2\} }{\sum} \underset{\Sigma^{k}}{\int} Z( \mathbf{L}_{j_1}(x_1) , \cdots , \mathbf{L}_{j_k}(x_k) ) \\ & = Z_k\left( {\, \atop \,} (O_0 + O_1), \cdots, (O_0 + O_1) {\, \atop \,}\right) - Z_k\left( {\, \atop \,} O_0, \cdots, O_0 {\, \atop \,} \right) + Z_k\left( {\, \atop \,} (O_0 + O_2), \cdots, (O_0 + O_2) {\, \atop \,} \right) \end{aligned} \,.

This directly implies the claim.

As a corollary we obtain:

Proposition

(composition of S-matrix scheme with vertex redefinition is again S-matrix scheme)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H) be a gauge fixed free field vacuum (this def.) and let 𝒵\mathcal{Z} be a vertex redefinition (def. 4).

Then for

𝒮:LocObs(E BV-BRST)[[,g,j]]g,jPolyObs(E BV-BRST) mc(())[[g,j]] \mathcal{S} \;\colon\; LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle \longrightarrow PolyObs(E_{\text{BV-BRST}})_{mc}((\hbar))[ [ g,j ] ]

and S-matrix scheme (this def.), the composite

𝒮𝒵:LocObs(E BV-BRST)[[,g,j]]g,j𝒵LocObs(E BV-BRST)[[,g,j]]g,j𝒮PolyObs(E BV-BRST) mc(())[[g,j]] \mathcal{S} \circ \mathcal{Z} \;\colon\; LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle \overset{\mathcal{Z}}{\longrightarrow} LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g, j\rangle \overset{\mathcal{S}}{\longrightarrow} PolyObs(E_{\text{BV-BRST}})_{mc}((\hbar))[ [ g,j ] ]

is again an S-matrix scheme.

Moreover, if 𝒮\mathcal{S} satisfies the renormalization condition “field independence” (this prop.), then so does 𝒮𝒵\mathcal{S} \circ \mathcal{Z}.

(e.g Dütsch 18, theorem 3.99 (b))

Proof

It is clear that causal order of the spacetime supports implies that they are in particular disjoint

(supp(O 1)supp(O 2))AAAA(supp(O 1)supp(O )=) \left( {\, \atop \,} supp(O_1) {\vee\!\!\!\wedge} supp(O_2) {\, \atop \,} \right) \phantom{AA} \Rightarrow \phantom{AA} \left( {\, \atop \,} supp(O_1) \cap supp(O_) \;=\; \emptyset {\, \atop \,} \right)

Therefore the local additivity of 𝒵\mathcal{Z} (prop. 6) and the causal factorization of the S-matrix (this remark) imply the causal factorization of the composite:

𝒮(𝒵(O 1+O 2)) =𝒮(𝒵(O 1)+𝒵(O 2)) =𝒮(𝒵(O 1))𝒮(𝒵(O 2)). \begin{aligned} \mathcal{S} \left( {\, \atop \,} \mathcal{Z}(O_1 + O_2) {\, \atop \,} \right) & = \mathcal{S} \left( {\, \atop \,} \mathcal{Z}(O_1) + \mathcal{Z}(O_2) {\, \atop \,} \right) \\ & = \mathcal{S} \left( {\, \atop \,} \mathcal{Z}(O_1) {\, \atop \,} \right) \, \mathcal{S} \left( {\, \atop \,} \mathcal{Z}(O_2) {\, \atop \,} \right) \,. \end{aligned}

But by this prop. this implies in turn causal additivity and hence that 𝒮𝒵\mathcal{S} \circ \mathcal{Z} is itself an S-matrix scheme.

Finally that 𝒮𝒵\mathcal{S} \circ \mathcal{Z} satisfies “field indepndence” if 𝒮\mathcal{S} does is immediate by the chain rule, given that 𝒵\mathcal{Z} satisfies this condition by definition.

Proposition

(any two S-matrix renormalization schemes differ by unique vertex redefinition)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H) be a gauge fixed free field vacuum (this def.).

Then for 𝒮,𝒮\mathcal{S}, \mathcal{S}' any two S-matrix schemes (this def.) which both satisfy the renormalization condition “field independence”, the there exists a unique vertex redefinition 𝒵\mathcal{Z} (def. 4) relating them by composition, i. e. such that

𝒮=𝒮𝒵. \mathcal{S}' \;=\; \mathcal{S} \circ \mathcal{Z} \,.
Proof

By applying both sides of the equation to linear combinations of local observables of the form κ 1O 1++κ kO k\kappa_1 O_1 + \cdots + \kappa_k O_k and then taking derivatives with respect to κ\kappa at κ j=0\kappa_j = 0 (as in this example) we get that the equation in question implies

(i) k kκ 1κ k𝒮(κ 1O 1++κ kO k)| κ 1,,κ k=0=(i) k kκ 1κ k𝒮𝒵(κ 1O 1++κ kO k)| κ 1,,κ k=0 (i \hbar)^k \frac{ \partial^k }{ \partial \kappa_1 \cdots \partial \kappa_k } \mathcal{S}'( \kappa_1 O_1 + \cdots + \kappa_k O_k ) \vert_{\kappa_1, \cdots, \kappa_k = 0} \;=\; (i \hbar)^k \frac{ \partial^k }{ \partial \kappa_1 \cdots \partial \kappa_k } \mathcal{S} \circ \mathcal{Z}( \kappa_1 O_1 + \cdots + \kappa_k O_k ) \vert_{\kappa_1, \cdots, \kappa_k = 0}

which in components means that

T k(O 1,,O k) =2nk1n!(i) knI 1I n={1,,k},I 1,,I nT n(Z |I 1|((O i 1) i 1I 1),,Z |I n|((O i n) i nI n),) =+Z k(O 1,,O k) \begin{aligned} T'_k( O_1, \cdots, O_k ) & = \underset{ 2 \leq n \leq k }{\sum} \frac{1}{n!} (i \hbar)^{k-n} \underset{ { { I_1 \sqcup \cdots \sqcup I_n } \atop { = \{1, \cdots, k\}, } } \atop { I_1, \cdots, I_n \neq \emptyset } }{\sum} T_n \left( {\, \atop \,} Z_{{\vert I_1\vert}}\left( (O_{i_1})_{i_1 \in I_1} \right), \cdots, Z_{{\vert I_n\vert}}\left( (O_{i_n})_{i_n \in I_n} \right), {\, \atop \,} \right) \\ & \phantom{=} + Z_k( O_1,\cdots, O_k ) \end{aligned}

where {T k} k\{T'_k\}_{k \in \mathbb{N}} are the time-ordered products corresponding to 𝒮\mathcal{S}' (by this example) and {T k} k𝒩\{T_k\}_{k \in \mathcal{N}} those correspondong to 𝒮\mathcal{S}.

Here the sum on the right runs over all ways that in the composite 𝒮𝒵\mathcal{S} \circ \mathcal{Z} a kk-ary operation arises as the composite of an nn-ary time-ordered product applied to the |I i|{\vert I_i\vert}-ary components of 𝒵\mathcal{Z}, for ii running from 1 to nn; except for the case k=nk = n, which is displayed separately in the second line

This shows that if 𝒵\mathcal{Z} exists, then it is unique, because its coefficients Z kZ_k are inductively in kk given by the expressions

(11) Z k(O 1,,O k) =T k(O 1,,O k)2nk1n!(i) knI 1I n={1,,k},I 1,,I nT n(Z |I 1|((O i 1) i 1I 1),,Z |I n|((O i n) i nI n),)(T𝒵 <k) k \begin{aligned} & Z_k( O_1,\cdots, O_k ) \\ & = T'_k( O_1, \cdots, O_k ) \;-\; \underset{ (T \circ \mathcal{Z}_{\lt k})_k }{ \underbrace{ \underset{ 2 \leq n \leq k }{\sum} \frac{1}{n!} (i \hbar)^{k-n} \underset{ { { I_1 \sqcup \cdots \sqcup I_n } \atop { = \{1, \cdots, k\}, } } \atop { I_1, \cdots, I_n \neq \emptyset } }{\sum} T_n \left( Z_{{\vert I_1\vert}}( (O_{i_1})_{i_1 \in I_1} ), \cdots, Z_{{\vert I_n\vert}}( (O_{i_n})_{i_n \in I_n} ), \right) } } \end{aligned}

(The symbol under the brace is introduced as a convenient shorthand for the term above the brace.)

Hence it remains to see that the Z kZ_k defined this way satisfy the conditions in def. 4.

The condition “perturbation” is immediate from the corresponding condition on 𝒮\mathcal{S} and 𝒮\mathcal{S}'.

Similarly the condition “field independence” follows immediately from the assumoption that 𝒮\mathcal{S} and 𝒮\mathcal{S}' satisfy this condition.

It only remains to see that Z kZ_k indeed takes values in local observables. Given that the time-ordered products a priori take values in the larrger space of microcausal polynomial observables this means to show that the spacetime support of Z kZ_k is on the diagonal.

But observe that, as indicated in the above formula, the term over the brace may be understood as the coefficient at order kk of the exponential series-expansion of the composite 𝒮𝒵 <k\mathcal{S} \circ \mathcal{Z}_{\lt k}, where

𝒵 <kn{1,,k1}1n!Z n \mathcal{Z}_{\lt k} \;\coloneqq\; \underset{ n \in \{1, \cdots, k-1\} }{\sum} \frac{1}{n!} Z_n

is the truncation of the vertex redefinition to degree <k\lt k. This truncation is clearly itself still a vertex redefinition (according to def. 4) so that the composite 𝒮𝒵 <k\mathcal{S} \circ \mathcal{Z}_{\lt k} is still an S-matrix scheme (by prop. 7) so that the (T𝒵 <k) k(T \circ \mathcal{Z}_{\lt k})_k are time-ordered products (by this example).

So as we solve 𝒮=𝒮𝒵\mathcal{S}' = \mathcal{S} \circ \mathcal{Z} inductively in degree kk, then for the induction step in degree kk the expressions T <kT'_{\lt k} and (T𝒵) <k(T \circ \mathcal{Z})_{\lt k} agree and are both time-ordered products. By this prop. this implies that T kT'_{k} and (T𝒵 <k) k(T \circ \mathcal{Z}_{\lt k})_{k} agree away from the diagonal. This means that their difference Z kZ_k is supported on the diagonal, and hence is indeed local.

In conclusion this establishes the following pivotal statement of perturbative quantum field theory:

Theorem

(main theorem of perturbative renormalizationStückelberg-Petermann renormalization group of vertex redefinitions)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H) be a gauge fixed free field vacuum (this def.).

  1. the vertex redefinitions 𝒵\mathcal{Z} (def. 4) form a group under composition;

  2. the set of S-matrix ("re"-)normalization schemes (this def., this remark) satisfying the renormalization condition “field independence” (this prop.) is a torsor over this group, hence equipped with a regular action in that

    1. the set of S-matrix schemes is non-empty;

    2. any two S-matrix ("re"-)normalization schemes 𝒮\mathcal{S}, 𝒮\mathcal{S}' are related by a unique vertex redefinition 𝒵\mathcal{Z} via composition:

      𝒮=𝒮𝒵. \mathcal{S}' \;=\; \mathcal{S} \circ \mathcal{Z} \,.

This group is called the Stückelberg-Petermann renormalization group.

Typically one imposes a set of renormalization conditions (this def.) and considers the corresponding subgroup of vertex redefinitions preserving these conditions.

Proof

The group-structure and regular action is given by prop. 7 and prop. 8. The existence of S-matrices follows is the statement of Epstein-Glaser ("re"-)normalization in theorem 1.

\,

UV-Regularization via Conterterms

While Epstein-Glaser renormalization (prop. 1) gives a transparent picture on the space of choices in ("re"-)normalization (theorem 1) the physical nature of the higher interactions that it introduces at coincident interaction points (via the extensions of distributions in prop. 5) remains more implicit. But the main theorem of perturbative renormalization (theorem 2), which re-expresses the difference between any two such choices as an interaction vertex redefinition, suggests that already the choice of ("re"-)normalization itself should have an incarnation in terms of interaction vertex redefinitions.

This may be realized via a construction of ("re"-)normalization in terms of UV-regularization (prop. 9 below): For any choice of “UV-cutoff”, given by an approximation of the Feynman propagator Δ F\Delta_F by non-singular distributions Δ F,Λ\Delta_{F,\Lambda} (def. 5 below) there is a unique “effective S-matrix𝒮 Λ\mathcal{S}_\Lambda induced at each cutoff scale (def. 6 below). While the “UV-limit” limΛ𝒮 Λ\underset{\Lambda \to \infty}{\lim} \mathcal{S}_\Lambda does not in general exist, it may be “regularized” by applying suitable interaction vertex redefinitions 𝒵 Λ\mathcal{Z}_\Lambda; if the higher-order corrections that these introduce serve to “counter” (remark 3 below) the coresponding UV-divergences.

This perspective of ("re"-)normalization via via counterterms is often regarded as the primary one. Its elegant proof in prop. 9 below, however relies on the Epstein-Glaser renormalization via inductive extensions of distributions and uses the same kind of argument as in the proof of the main theorem of perturbative renormalization (theorem 2) that establishes the Stückelberg-Petermann renormalization group.

\,

Definition

(UV cutoffs)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum over Minkowski spacetime Σ\Sigma (according to this def.), where Δ H=i2(Δ +Δ )+H\Delta_H = \tfrac{i}{2}(\Delta_+ - \Delta_-) + H is the corresponding Wightman propagator inducing the Feynman propagator

Δ FΓ Σ×Σ(E BV-BRSTE BV-BRST) \Delta_F \in \Gamma'_{\Sigma \times \Sigma}(E_{\text{BV-BRST}} \boxtimes E_{\text{BV-BRST}})

by Δ F=i2(Δ ++Δ )+H\Delta_F = \tfrac{i}{2}(\Delta_+ + \Delta_-) + H.

Then a choice of UV cutoffs for perturbative QFT around this vacuum is a collection of non-singular distributions Δ F,Λ\Delta_{F,\Lambda} parameterized by positive real numbers

(0,) Γ Σ×Σ,cp(E BV-BRSTE BV-BRST) Λ Δ F,Λ \array{ (0, \infty) &\overset{}{\longrightarrow}& \Gamma_{\Sigma \times \Sigma,cp}(E_{\text{BV-BRST}} \boxtimes E_{\text{BV-BRST}}) \\ \Lambda &\mapsto& \Delta_{F,\Lambda} }

such that:

  1. each Δ F,Λ\Delta_{F,\Lambda} satisfies the following basic properties

    1. (translation invariance)

      Δ F,Λ(x,y)=Δ F,Λ(xy) \Delta_{F,\Lambda}(x,y) = \Delta_{F,\Lambda}(x-y)
    2. (symmetry)

      Δ F,Λ ba(y,x)=Δ F,Λ ab(x,y) \Delta^{b a}_{F,\Lambda}(y, x) \;=\; \Delta^{a b}_{F,\Lambda}(x, y)

      i.e.

      Δ F,Λ ba(x)=Δ F,Λ ab(x) \Delta_{F,\Lambda}^{b a}(-x) \;=\; \Delta_{F,\Lambda}^{a b}(x)
  2. the Δ F,Λ\Delta_{F,\Lambda} interpolate between zero and the Feynman propagator, in that, in the Hörmander topology:

    1. the limit as Λ0\Lambda \to 0 exists and is zero

      limΛΔ F,Λ=0. \underset{\Lambda \to \infty}{\lim} \Delta_{F,\Lambda} \;=\; 0 \,.
    2. the limit as Λ\Lambda \to \infty exists and is the Feynman propagator:

      limΛΔ F,Λ=Δ F. \underset{\Lambda \to \infty}{\lim} \Delta_{F,\Lambda} \;=\; \Delta_F \,.

(Dütsch 10, section 4)

example: relativistic momentum cutoff with ϵ\epsilon-regularization (Keller-Kopper-Schophaus 97, section 6.1, Dütsch 18, example 3.126)

Definition

(effective S-matrix scheme)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5).

We say that the effective S-matrix scheme 𝒮 Λ\mathcal{S}_\Lambda at cutoff scale Λ[0,)\Lambda \in [0,\infty)

PolyObs(E BV-BRST) mc[[,g,j]] 𝒮 Λ PolyObs(E BV-BRST) mc[[,g,j]] O 𝒮 Λ(O) \array{ PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ] &\overset{\mathcal{S}_{\Lambda}}{\longrightarrow}& PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ] \\ O &\mapsto& \mathcal{S}_\Lambda(O) }

is the exponential series

(12)𝒮 Λ(O) exp F,Λ(1iO) =1+1iO+121(i) 2O F,ΛO+13!1(i) 3O F,ΛO F,Λ0+. \begin{aligned} \mathcal{S}_\Lambda(O) & \coloneqq \exp_{F,\Lambda}\left( \frac{1}{i \hbar} O \right) \\ & = 1 + \frac{1}{i \hbar} O + \frac{1}{2} \frac{1}{(i \hbar)^2} O \star_{F,\Lambda} O + \frac{1}{3!} \frac{1}{(i \hbar)^3} O \star_{F,\Lambda} O \star_{F,\Lambda} 0 + \cdots \end{aligned} \,.

with respect to the star product F,Λ\star_{F,\Lambda} induced by the Δ F,Λ\Delta_{F,\Lambda} (this def.).

This is evidently defined on all polynomial observables as shown, and restricts to an endomorphism on microcausal polynomial observables as shown, since the contraction coefficients Δ F,Λ\Delta_{F,\Lambda} are non-singular distributions, by definition of UV cutoff.

(Dütsch 10, (4.2))

Proposition

(("re"-)normalization via UV regularization)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let gS int+jALocObs(E BV-BRST)[[,g,j]]g,jg S_{int} + j A \in LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle a polynomial local observable, regarded as an adiabatically switched interaction action functional.

Let moreover {Δ F,Λ} Λ[0,)\{\Delta_{F,\Lambda}\}_{\Lambda \in [0,\infty)} be a UV cutoff (def. 5); with 𝒮 Λ\mathcal{S}_\Lambda the induced effective S-matrix schemes (12).

Then

  1. there exists a [0,)[0,\infty)-parameterized interaction vertex redefinition {𝒵 Λ} Λ 0\{\mathcal{Z}_\Lambda\}_{\Lambda \in \mathbb{R}_{\geq 0}} (this def.) such that the limit of effective S-matrix schemes 𝒮 Λ\mathcal{S}_{\Lambda} (12) applied to the 𝒵 Λ\mathcal{Z}_\Lambda-redefined interactions

    𝒮 limΛ(𝒮 Λ𝒵 Λ) \mathcal{S}_\infty \;\coloneqq\; \underset{\Lambda \to \infty}{\lim} \left( \mathcal{S}_\Lambda \circ \mathcal{Z}_\Lambda \right)

    exists and is a genuine S-matrix scheme around the given vacuum (this def.);

  2. every S-matrix scheme around the given vacuum arises this way.

These 𝒵 Λ\mathcal{Z}_\Lambda are called counterterms (remark 3 below) and the composite 𝒮 Λ𝒵 Λ\mathcal{S}_\Lambda \circ \mathcal{Z}_\Lambda is called a UV regularization of the effective S-matrices 𝒮 Λ\mathcal{S}_\Lambda.

Hence UV-regularization via counterterms is a method of ("re"-)normalization of perturbative QFT (this def.).

This was claimed in (Brunetti-Dütsch-Fredenhagen 09, (75)), a proof was indicated in (Dütsch-Fredenhagen-Keller-Rejzner 14, theorem A.1).

Proof

Let {p ρ k} k\{p_{\rho_{k}}\}_{k \in \mathbb{N}} be a sequence of projection maps as in (6) defining an Epstein-Glaser ("re"-)normalization (prop. 1) of time-ordered products {T k} k\{T_k\}_{k \in \mathbb{N}} as extensions of distributions of the T kT_k, regarded as distributions via remark 1, by the choice q k α=0q_k^\alpha = 0 in (9).

We will construct that 𝒵 Λ\mathcal{Z}_\Lambda in terms of these projections p ρp_\rho.

First consider some convenient shorthand:

For nn \in \mathbb{N}, write 𝒵 n1{1,,n}1n!Z n\mathcal{Z}_{\leq n} \coloneqq \underset{1 \in \{1, \cdots, n\}}{\sum} \frac{1}{n!} Z_n. Moreover, for kk \in \mathbb{N} write (T Λ𝒵 n) k(T_\Lambda \circ \mathcal{Z}_{\leq n})_k for the kk-ary coefficient in the expansion of the composite 𝒮 Λ𝒵 n\mathcal{S}_\Lambda \circ \mathcal{Z}_{\leq n}, as in equation (11) in the proof of the main theorem of perturbative renormalization (theorem 2).

In this notation we need to find 𝒵 Λ\mathcal{Z}_\Lambda such that for each nn \in \mathbb{N} we have

(13)limΛ(T Λ𝒵 n,Λ) n=T n. \underset{\Lambda \to \infty}{\lim} \left( T_\Lambda \circ \mathcal{Z}_{\leq n, \Lambda} \right)_n \;=\; T_n \,.

We proceed by induction over nn \in \mathbb{N}.

Since by definition T 0=const 1T_0 = const_1, T 1=idT_1 = id and Z 0=const 0Z_0 = const_0, Z 1=idZ_1 = id the statement is trivially true for n=0n = 0 and n=1n = 1.

So assume now nn \in \mathbb{N} and {Z k} kn\{Z_{k}\}_{k \leq n} has been found such that (13) holds.

Observe that with the chosen renormalizing projection p ρ n+1p_{\rho_{n+1}} the time-ordered product T n+1T_{n+1} may be expressed as follows:

(14)T n+1(O,,O) =I{1,,n+1}I,I¯χ i(X)(T |I|(I)) F(T |I¯|(I¯)),p ρ k(OO) =limΛI{1,,n+1}I,I¯χ i(X)(T |I|(I)) F,Λ(T |I¯|(I¯)),p ρ k(OO). \begin{aligned} T_{n+1}(O, \cdots, O) & = \left\langle \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} T_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F} \left( {\, \atop \,} T_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_k}(O \otimes \cdots \otimes O) \right\rangle \\ & = \left\langle \underset{\Lambda \to \infty}{\lim} \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} T_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F,\Lambda} \left( {\, \atop \,} T_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_k}(O \otimes \cdots \otimes O) \right\rangle \end{aligned} \,.

Here in the first step we inserted the causal decomposition (4) of T n+1T_{n+1} in terms of the {T k} kn\{T_k\}_{k \leq n} away from the diagonal, as in the proof of prop. 1, which is admissible because the image of p ρ n+1p_{\rho_{n+1}} vanishes on the diagonal. In the second step we replaced the star-product of the Feynman propagator Δ F\Delta_F with the limit over the star-products of the regularized propagators Δ F,Λ\Delta_{F,\Lambda}, which converges by the nature of the Hörmander topology (which is assumed by def. 5).

Hence it is sufficient to find Z n+1,ΛZ_{n+1,\Lambda} and K n+1,ΛK_{n+1,\Lambda} such that

(15)(T Λ𝒵 Λ) n+1,(,,) =I{1,,n+1}I,I¯χ i(X)(T |I|(I)) F,Λ(T |I¯|(I¯)),p ρ k(,,) =+K n+1,Λ(,,) \begin{aligned} \left\langle \left( T_{\Lambda} \circ \mathcal{Z}_{\Lambda} \right)_{n+1} \,,\, (-, \cdots, -) \right\rangle & = \left\langle \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} T_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F,\Lambda} \left( {\, \atop \,} T_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_{k}}\left( -, \cdots, - \right) \right\rangle \\ & \phantom{=} + K_{n+1,\Lambda}(-, \cdots, -) \end{aligned}

subject to these two conditions:

  1. 𝒵 n+1,Λ\mathcal{Z}_{n+1,\Lambda} is local;

  2. limΛK n+1,Λ=0\underset{\Lambda \to \infty}{\lim} K_{n+1,\Lambda} = 0.

Now by expanding out the left hand side of (15) as

(T Λ𝒵 Λ) n+1=Z n+1,Λ+(T ΛZ n,Λ) n+1 (T_\Lambda \circ \mathcal{Z}_\Lambda)_{n+1} \;=\; Z_{n+1,\Lambda} \;+\; (T_\Lambda \circ Z_{\leq n, \Lambda})_{n+1}

(which uses the condition T 1=idT_1 = id) we find the unique solution of (15) for Z n+1,ΛZ_{n+1,\Lambda}, in terms of the {Z n,Λ}\{Z_{\leq n,\Lambda}\} and K n+1,ΛK_{n+1,\Lambda} (the latter still to be chosen) to be:

(16)Z n+1,Λ,(,,) =I{1,,n+1}I,I¯χ i(X)(T |I|(I)) F,Λ(T |I¯|(I¯)),p ρ n+1(,,) =(T Λ𝒵 n,Λ) n+1,(,,) =+K n+1,Λ,(,,). \begin{aligned} \left\langle Z_{n+1,\Lambda} , (-,\cdots, -) \right\rangle & = \left\langle \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} T_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F,\Lambda} \left( {\, \atop \,} T_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_{n+1}}(-, \cdots, -) \right\rangle \\ & \phantom{=} - \left\langle \left( T_{\Lambda} \circ \mathcal{Z}_{\leq n,\Lambda} \right)_{n+1} \,,\, (-, \cdots, -) \right\rangle \\ & \phantom{=} + \left\langle K_{n+1, \Lambda}, (-, \cdots, -) \right\rangle \end{aligned} \,.

We claim that the following choice works:

(17)K n+1,Λ(,,) (T Λ𝒵 n,Λ) n+1,p ρ n+1(,,) =I{1,,n+1}I,I¯χ i(X)(T |I|(I)) F,Λ(T |I¯|(I¯)),p ρ n+1(,,). \begin{aligned} K_{n+1, \Lambda}(-, \cdots, -) & \coloneqq \left\langle \left( T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda} \right)_{n+1} \,,\, p_{\rho_{n+1}}(-, \cdots, -) \right\rangle \\ & \phantom{=} - \left\langle \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} T_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F,\Lambda} \left( {\, \atop \,} T_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_{n+1}}(-, \cdots, -) \right\rangle \end{aligned} \,.

To prove this, we need to show that 1) the resulting Z n+1,ΛZ_{n+1,\Lambda} is local and 2) the limit of K n+1,ΛK_{n+1,\Lambda} vanishes as Λ\Lambda \to \infty.

First regarding the locality of Z n+1,ΛZ_{n+1,\Lambda}: By inserting (17) into (16) we obtain

Z n+1,Λ,(,,) =(T Λ𝒵 n) n+1,p(,,)(T Λ𝒵 n) n+1,(,,) =(T Λ𝒵 n) n+1,(p ρ n+1id)(,,) \begin{aligned} \left\langle Z_{n+1,\Lambda} \,,\, (-,\cdots,-) \right\rangle & = \left\langle \left( T_{\Lambda} \circ \mathcal{Z}_{\leq n} \right)_{n+1} \,,\, p(-, \cdots, -) \right\rangle - \left\langle \left( T_{\Lambda} \circ \mathcal{Z}_{\leq n} \right)_{n+1} \,,\, (-, \cdots, -) \right\rangle \\ & = \left\langle \left( T_{\Lambda} \circ \mathcal{Z}_{\leq n} \right)_{n+1} \,,\, ( p_{\rho_{n+1}} - id)(-, \cdots, -) \right\rangle \end{aligned}

By definition p ρ n+1idp_{\rho_{n+1}} - id is the identity on test functions (adiabatic switchings) that vanish at the diagonal. This means that Z n+1,ΛZ_{n+1,\Lambda} is supported on the diagonal, and is hence local.

Second we need to show that limΛK n+1,Λ=0\underset{\Lambda \to \infty}{\lim} K_{n+1,\Lambda} = 0:

By applying the analogous causal decomposition (4) to the regularized products, we find

(18) (T Λ𝒵 n,Λ) n+1,p ρ n+1(,,) =I{1,,n+1}I,I¯χ i(X)((T Λ𝒵 n,Λ) |I|(I)) F,Λ((T Λ𝒵 n,Λ) |I¯|(I¯)),p ρ n+1(,,). \begin{aligned} & \left\langle (T_\Lambda \circ \mathcal{Z}_{\leq n, \Lambda})_{n+1} \,,\, p_{\rho_{n+1}}(-,\cdots,-) \right\rangle \\ & = \left\langle \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} (T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda})_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F,\Lambda} \left( {\, \atop \,} (T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda})_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_{n+1}}(-,\cdots,-) \right\rangle \,. \end{aligned}

Using this we compute as follows:

(19) limΛ(T Λ𝒵 n,Λ) n+1,p ρ n+1(,,) =limΛI{1,,n+1}I,I¯χ i(X)((T Λ𝒵 n,Λ) |I|(I)) F,Λ((T Λ𝒵 n,Λ) |I¯|(I¯)),p ρ n+1(,,) =I{1,,n+1}I,I¯χ i(X)(limΛ(T Λ𝒵 n,Λ) |I|(I))T |I|(I)(limΛ F,Λ)(limΛ(T Λ𝒵 n,Λ) |I¯|(I¯))T |I¯|(I¯),p ρ n+1(,,) =limΛI{1,,n+1}I,I¯χ i(X)T |I|(I) F,ΛT |I¯|(I¯),p ρ n+1(,,). \begin{aligned} & \left\langle \underset{\Lambda \to \infty}{\lim} (T_\Lambda \circ \mathcal{Z}_{\leq n, \Lambda})_{n+1} \,,\, p_{\rho_{n+1}}(-,\cdots,-) \right\rangle \\ & = \left\langle \underset{\Lambda \to \infty}{\lim} \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { \mathbf{I}, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \!\!\chi_i(\mathbf{X})\, \left( {\, \atop \,} (T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda})_{{\vert \mathbf{I}\vert}} (\mathbf{I}) {\, \atop \,} \right) \star_{F,\Lambda} \left( {\, \atop \,} (T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda})_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) {\, \atop \,} \right) \,,\, p_{\rho_{n+1}}(-,\cdots,-) \right\rangle \\ & = \left\langle \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { I, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \chi_i(\mathbf{X})\, \underset{ T_{{\vert \mathbf{I}\vert}}(\mathbf{I}) }{ \underbrace{ \left( \underset{\Lambda \to \infty}{\lim} (T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda})_{{\vert \mathbf{I}\vert}} (\mathbf{I}) \right) }} \left( \underset{\Lambda \to \infty}{\lim} \star_{F,\Lambda} \right) \underset{ T_{{\vert \overline{\mathbf{I}}\vert}}(\overline{\mathbf{I}}) }{ \underbrace{ \left( \underset{\Lambda \to \infty}{\lim} (T_{\Lambda} \circ \mathcal{Z}_{\leq n, \Lambda})_{ {\vert \overline{\mathbf{I}} \vert} } ( \overline{\mathbf{I}} ) \right) }} \,,\, p_{\rho_{n+1}}(-,\cdots,-) \right\rangle \\ & = \left\langle \underset{\Lambda \to \infty}{\lim} \underset{ {\mathbf{I} \in \{1, \cdots, n+1\} } \atop { I, \overline{\mathbf{I}} \neq \emptyset } }{\sum} \chi_i(\mathbf{X})\, T_{ { \vert \mathbf{I} \vert } }( \mathbf{I} ) \star_{F,\Lambda} T_{ {\vert \overline{\mathbf{I}} \vert} }( \overline{\mathbf{I}} ) \,,\, p_{\rho_{n+1}}(-,\cdots,-) \right\rangle \end{aligned} \,.

Here in the first step we inserted (18); in the second step we used that in the Hörmander topology the product of distributions preserves limits in each variable and in the third step we used the induction assumption (13) and the definition of UV cutoff (def. 5).

Inserting this for the first summand in (17) shows that limΛK n+1,Λ=0\underset{\Lambda \to \infty}{\lim} K_{n+1, \Lambda} = 0.

In conclusion this shows that a consistent choice of counterterms 𝒵 Λ\mathcal{Z}_\Lambda exists to produce some S-matrix 𝒮=limΛ(𝒮 Λ𝒵 Λ)\mathcal{S} = \underset{\Lambda \to \infty }{\lim} (\mathcal{S}_\Lambda \circ \mathcal{Z}_\Lambda).

It just remains to see that for every other S-matrix 𝒮˜\widetilde{\mathcal{S}} there exist counterterms 𝒵˜ λ\widetilde{\mathcal{Z}}_\lambda such that 𝒮˜=limΛ(𝒮 Λ𝒵˜ Λ)\widetilde{\mathcal{S}} = \underset{\Lambda \to \infty }{\lim} (\mathcal{S}_\Lambda \circ \widetilde{\mathcal{Z}}_\Lambda).

But by the main theorem of perturbative renormalization (theorem 2) we know that there exists a vertex redefinition 𝒵\mathcal{Z} such that

𝒮˜ =𝒮𝒵 =limΛ(𝒮 Λ𝒵 Λ)𝒵 =limΛ(𝒮 Λ(𝒵 Λ𝒵𝒵˜ Λ)) \begin{aligned} \widetilde{\mathcal{S}} & = \mathcal{S} \circ \mathcal{Z} \\ & = \underset{\Lambda \to \infty}{\lim} \left( \mathcal{S}_\Lambda \circ \mathcal{Z}_\Lambda \right) \circ \mathcal{Z} \\ & = \underset{\Lambda \to \infty}{\lim} ( \mathcal{S}_\Lambda \circ ( \underset{ \widetilde{\mathcal{Z}}_\Lambda }{ \underbrace{ \mathcal{Z}_\Lambda \circ \mathcal{Z} } } ) ) \end{aligned}

and hence with counterterms 𝒵 Λ\mathcal{Z}_\Lambda for 𝒮\mathcal{S} given, then counterterms for any 𝒮˜\widetilde{\mathcal{S}} are given by the composite 𝒵˜ Λ𝒵 Λ𝒵\widetilde{\mathcal{Z}}_\Lambda \coloneqq \mathcal{Z}_\Lambda \circ \mathcal{Z}.

Remark

(counterterms)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5).

Consider

gS int+jALocObs(E BV-BRST)[[,g,j]]g,j g S_{int} + j A \;\in\; LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle

a local observable, regarded as an adiabatically switched interaction action functional.

Then prop. 9 says that there exist vertex redefinitions of this interaction

𝒵 Λ(gS int+jA)LocObs(E BV-BRST)[[,g,j]]g,j \mathcal{Z}_\Lambda(g S_{int} + j A) \;\in\; LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle

parameterized by Λ[0,)\Lambda \in [0,\infty), such that the limit

𝒮 (gS int+jA)limΛ𝒮 Λ(𝒵 Λ(gS int+jA)) \mathcal{S}_\infty(g S_{int} + j A) \;\coloneqq\; \underset{\Lambda \to \infty}{\lim} \mathcal{S}_\Lambda\left( \mathcal{Z}_\Lambda( g S_{int} + j A )\right)

exists and is an S-matrix for perturbative QFT with the given interaction gS int+jAg S_{int} + j A.

In this case the difference

S counter,Λ (gS int+jA)𝒵 Λ(gS int+jA)Loc(E BV-BRST)[[,g,j]]g 2,j 2,gj \begin{aligned} S_{counter, \Lambda} & \coloneqq \left( g S_{int} + j A \right) \;-\; \mathcal{Z}_{\Lambda}(g S_{int} + j A) \;\;\;\;\;\in\; Loc(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g^2, j^2, g j\rangle \end{aligned}

(which by the axiom “perturbation” in this def. is at least of second order in the coupling constant/source field, as shown) is called a choice of counterterms at cutoff scale Λ\Lambda. These are new interactions which are added to the given interaction at cutoff scale Λ\Lambda

𝒵 Λ(gS int+jA)=gS int+jA+S counter,Λ. \mathcal{Z}_{\Lambda}(g S_{int} + j A) \;=\; g S_{int} + j A \;+\; S_{counter,\Lambda} \,.

In this language prop. 9 says that for every free field vacuum and every choice of local interaction, there is a choice of counterterms to the interaction that defines a corresponding ("re"-)normalized perturbative QFT, and every (re"-)normalized perturbative QFT arises from some choice of counterterms.

\,

Wilson-Polchinski effective QFT flow

We have seen above that a choice of UV cutoff induces effective S-matrix schemes 𝒮 Λ\mathcal{S}_\Lambda at cutoff scale Λ\Lambda (def. 6). To these one may associated non-local relative effective actions S eff,ΛS_{eff,\Lambda} (def. 7 below) which are such that their effective scattering amplitudes at scale Λ\Lambda coincide with the true scattering amplitudes of a genuine local interaction as the cutoff is removed. This is the Wilsonian picture of effective quantum field theory at a given cutoff scale (remark 4 below). Crucially the “flow” of the relative effective actions with the cutoff scale satisfies a differential equation that in itself is independent of the full UV-theory; this is Polchinski's flow equation (prop. 12 below). Solving this equation for given choice of initial value data is hence another way of choosing ("re"-)normalization constants.

\,

Proposition

(effective S-matrix schemes are invertible functions)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5).

Write

PolyObs(E BV-BRST) mc[[,g,j]]g,jPolyObs(E BV-BRST) mc[[,g,j]] PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]\langle g,j\rangle \hookrightarrow PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]

for the subspace of the space of formal power series in ,g,j\hbar, g, j with coefficients polynomial observables on those which are at least of first order in g,jg,j, i.e. those that vanish for g,j=0g, j = 0 (as in this def.).

Write moreover

1+PolyObs(E BV-BRST) mc[[,g,j]]g,jPolyObs(E BV-BRST) mc[[,g,j]] 1 + PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]\langle g,j\rangle \hookrightarrow PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]

for the subspace of polynomial observables which are the sum of 1 (the multiplicative unit) with an observable at least linear n g,jg,j.

Then the effective S-matrix schemes 𝒮 Λ\mathcal{S}_\Lambda (def. 6) restrict to linear isomorphisms of the form

PolyObs(E BV-BRST) mc[[,g,j]]g,jAA𝒮 ΛAA1+PolyObs(E BV-BRST) mc[[,g,j]]g,j. PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]\langle g,j\rangle \underoverset{\simeq}{\phantom{AA}\mathcal{S}_\Lambda \phantom{AA} }{\longrightarrow} 1 + PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]\langle g,j\rangle \,.

(Dütsch 10, (4.7))

Proof

Since each Δ F,Λ\Delta_{F,\Lambda} is symmetric (def. 5) if follows by general properties of star products (this prop.) just as for the genuine time-ordered product on regular polynomial observables (this prop.) that eeach the “effective time-ordered product” F,Λ\star_{F,\Lambda} is isomorphic to the pointwise product ()()(-)\cdot (-) (this def.)

A 1 F,ΛA 2=𝒯 Λ(𝒯 Λ 1(A 1)𝒯 Λ 1(A 2)) A_1 \star_{F,\Lambda} A_2 \;=\; \mathcal{T}_\Lambda \left( \mathcal{T}_\Lambda^{-1}(A_1) \cdot \mathcal{T}_\Lambda^{-1}(A_2) \right)

for

𝒯 Λexp(12ΣΔ F,Λ ab(x,y)δ 2δΦ a(x)δΦ b(y)) \mathcal{T}_\Lambda \;\coloneqq\; \exp \left( \tfrac{1}{2}\hbar \underset{\Sigma}{\int} \Delta_{F,\Lambda}^{a b}(x,y) \frac{\delta^2}{\delta \mathbf{\Phi}^a(x) \delta \mathbf{\Phi}^b(y)} \right)

(as in this equation).

In particular this means that the effective S-matrix 𝒮 Λ\mathcal{S}_\Lambda arises from the exponential series for the pointwise product by conjugation with 𝒯 Λ\mathcal{T}_\Lambda:

𝒮 Λ=𝒯 Λexp (1i())𝒯 Λ 1 \mathcal{S}_\Lambda \;=\; \mathcal{T}_\Lambda \circ \exp_\cdot\left( \frac{1}{i \hbar}(-) \right) \circ \mathcal{T}_\Lambda^{-1}

(just as for the genuine S-matrix on regular polynomial observables in this def.).

Now the exponential of the pointwise product on 1+PolyObs(E BV-BRST)[[,g,j]]g,j1 + PolyObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle has as inverse function the natural logarithm power series, and since 𝒯\mathcal{T} evidently preserves powers of g,jg,j this conjugates to an inverse at each UV cutoff scale Λ\Lambda:

(20)𝒮 Λ 1=𝒯 Λln(i())𝒯 Λ 1. \mathcal{S}_\Lambda^{-1} \;=\; \mathcal{T}_\Lambda \circ \ln\left( i \hbar (-) \right) \circ \mathcal{T}_\Lambda^{-1} \,.
Definition

(relative effective action)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5).

Consider

gS int+jALocObs(E BV-BrST)[[,g,j]]g,j g S_{int} + j A \;\in\; LocObs(E_{\text{BV-BrST}})[ [ \hbar, g, j] ]\langle g, j\rangle

a local observable regarded as an adiabatically switched interaction action functional.

Then for

Λ,Λ vac(0,) \Lambda,\, \Lambda_{vac} \;\in\; (0, \infty)

two UV cutoff-scale parameters, we say the relative effective action S eff,Λ,Λ 0S_{eff, \Lambda, \Lambda_0} is the image of this interaction under the composite of the effective S-matrix scheme 𝒮 Λ 0\mathcal{S}_{\Lambda_0} at scale Λ 0\Lambda_0 (12) and the inverse function 𝒮 Λ 1\mathcal{S}_\Lambda^{-1} of the effective S-matrix scheme at scale Λ\Lambda (via prop. 10):

(21)S eff,Λ,Λ 0𝒮 Λ 1𝒮 Λ 0(gS int+jA)AAAΛ,Λ 0[0,). S_{eff,\Lambda, \Lambda_0} \;\coloneqq\; \mathcal{S}_{\Lambda}^{-1} \circ \mathcal{S}_{\Lambda_0}(g S_{int} + j A) \phantom{AAA} \Lambda, \Lambda_0 \in [0,\infty) \,.

For chosen counterterms (remark 3) hence for chosen UV regularization 𝒮 \mathcal{S}_\infty (prop. 9) this makes sense also for Λ 0=\Lambda_0 = \infty and we write:

(22)S eff,ΛS eff,Λ,𝒮 Λ 1𝒮 (gS int+jA)AAAΛ[0,) S_{eff,\Lambda} \;\coloneqq\; S_{eff,\Lambda, \infty} \;\coloneqq\; \mathcal{S}_{\Lambda}^{-1} \circ \mathcal{S}_{\infty}(g S_{int} + j A) \phantom{AAA} \Lambda \in [0,\infty)

(Dütsch 10, (5.4))

Remark

(effective quantum field theory)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.), let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5), and let 𝒮 =limΛ𝒮 Λ𝒵 Λ\mathcal{S}_\infty = \underset{\Lambda \to \infty}{\lim} \mathcal{S}_\Lambda \circ \mathcal{Z}_\Lambda be a corresponding UV regularization (prop. 9).

Consider a local observable

gS int+jALocObs(E BV-BrST)[[,g,j]]g,j g S_{int} + j A \;\in\; LocObs(E_{\text{BV-BrST}})[ [ \hbar, g, j] ]\langle g, j\rangle

regarded as an adiabatically switched interaction action functional.

Then def. 5 and def. 7 say that for any Λ(0,)\Lambda \in (0,\infty) the effective S-matrix (12) of the relative effective action (21) equals the genuine S-matrix 𝒮 \mathcal{S}_\infty of the genuine interaction gS int+jAg S_{int} + j A:

𝒮 Λ(S eff,Λ)=𝒮 (gS int+jA). \mathcal{S}_\Lambda( S_{eff,\Lambda} ) \;=\; \mathcal{S}_\infty\left( g S_{int} + j A \right) \,.

In other words the relative effective action S eff,ΛS_{eff,\Lambda} encodes what the actual perturbative QFT defined by 𝒮 (gS int+jA)\mathcal{S}_\infty\left( g S_{int} + j A \right) effectively looks like at UV cutoff Λ\Lambda.

Therefore one says that S eff,ΛS_{eff,\Lambda} defines effective quantum field theory at UV cutoff Λ\Lambda.

Notice that in general S eff,ΛS_{eff,\Lambda} is not a local interaction anymore: By prop. 10 the image of the inverse 𝒮 Λ 1\mathcal{S}^{-1}_\Lambda of the effective S-matrix is microcausal polynomial observables in 1+PolyObs(E BV-BRST) mc[[,g,j]]g,j1 + PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ]\langle g,j\rangle and there is no guarantee that this lands in the subspace of local observables.

Therefore effective quantum field theories at finite UV cutoff-scale Λ[0,)\Lambda \in [0,\infty) are in general not local field theories, even if their limit as Λ\Lambda \to \infty is, via prop. 9.

Proposition

(effective action is relative effective action at Λ=0\Lambda = 0)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5).

Then the relative effective action (def. 7) at Λ=0\Lambda = 0 is the actual effective action (this def.) in the sense of the the Feynman perturbation series of Feynman amplitudes Γ(gS int+jA)\Gamma(g S_{int} + j A) (this def.) for connected Feynman diagrams Γ\Gamma:

S eff,0 S eff,0, =S effΓΓ connΓ(gS int+jA). \begin{aligned} S_{eff,0} & \coloneqq\; S_{eff,0,\infty} \\ & = S_{eff} \;\coloneqq\; \underset{\Gamma \in \Gamma_{conn}}{\sum} \Gamma(g S_{int} + j A) \,. \end{aligned}

More generally this holds true for any Λ[0,){}\Lambda \in [0, \infty) \sqcup \{\infty\}

S eff,0,Λ =ΓΓ connΓ Λ(gS int+jA), \begin{aligned} S_{eff,0,\Lambda} & = \underset{\Gamma \in \Gamma_{conn}}{\sum} \Gamma_\Lambda(g S_{int} + j A) \,, \end{aligned}

where Γ Λ(gS int+jA)\Gamma_\Lambda( g S_{int} + j A) denotes the evident version of the Feynman amplitude (this def.) with time-ordered products replaced by effective time ordered product at scale Λ\Lambda as in (def. 6).

(Dütsch 18, (3.473))

Proof

Observe that the effective S-matrix scheme at scale Λ=0\Lambda = 0 (12) is the exponential series with respect to the pointwise product (this def.)

𝒮 0(O)=exp (O). \mathcal{S}_0(O) = \exp_\cdot( O ) \,.

Therefore the statement to be proven says equivalently that the exponential series of the effective action with respect to the pointwise product is the S-matrix:

exp (1iS eff)=𝒮 (gS int+jA). \exp_\cdot\left( \frac{1}{i \hbar} S_{eff} \right) \;=\; \mathcal{S}_\infty\left( g S_{int} + j A \right) \,.

That this is the case is the statement of this prop..

The definition of the relative effective action 𝒮 eff,Λ𝒮 eff,Λ,\mathcal{S}_{eff,\Lambda} \coloneqq \mathcal{S}_{eff,\Lambda, \infty} in def. 7 invokes a choice of UV regularization 𝒮 \mathcal{S}_\infty (prop. 9). While (by that proposition and the main theorem of perturbative renormalization (theorem 2) this is guaranteed to exist, in practice one is after methods for constructing this without specifying it a priori.

But the collection relative effective actions 𝒮 eff,Λ,Λ 0\mathcal{S}_{eff,\Lambda, \Lambda_0} for Λ 0<\Lambda_0 \lt \infty “flows” with the cutoff-parameters Λ\Lambda and in particular also with Λ 0\Lambda_0 (remark 5 below) which suggests that examination of this flow yields information about full theory at 𝒮 \mathcal{S}_\infty.

This is made precise by Polchinski's flow equation (prop. 12 below), which is the infinitesimal version of the “Wilsonian RG flow” (remark 5). As a differential equation it is independent of the choice of 𝒮 \mathcal{S}_{\infty} and hence may be used to solve for the Wilsonian RG flow without knowing 𝒮 \mathcal{S}_\infty in advance.

The freedom in choosing the initial values of this differential equation corresponds to the ("re"-)normalization freedom in choosing the UV regularization 𝒮 \mathcal{S}_\infty. In this sense “Wilsonian RG flow” is a method of ("re"-)normalization of perturbative QFT (this def.).

Remark

(Wilsonian groupoid of effective quantum field theories)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.) and let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5).

Then the relative effective actions 𝒮 eff,Λ,Λ 0\mathcal{S}_{eff,\Lambda, \Lambda_0} (def. 7) satisfy

S eff,Λ,Λ 0=(𝒮 Λ 1𝒮 Λ)(S eff,Λ,Λ 0)AAAforΛ,Λ[0,),Λ 0[0,){}. S_{eff, \Lambda', \Lambda_0} \;=\; \left( \mathcal{S}_{\Lambda'}^{-1} \circ \mathcal{S}_\Lambda \right) \left( S_{eff, \Lambda, \Lambda_0} \right) \phantom{AAA} \text{for} \, \Lambda,\Lambda' \in [0,\infty) \,,\, \Lambda_0 \in [0,\infty) \sqcup \{\infty\} \,.

This is similar to a group of UV-cutoff scale-transformations. But since the composition operations are only sensible when the UV-cutoff labels match, as shown, it is really a groupoid action.

This is often called the Wilsonian RG.

We now consider the infinitesimal version of this “flow”:

Proposition

(Polchinski's flow equation)

Let (E BV-BRST,L,Δ H)(E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H ) be a gauge fixed relativistic free vacuum (according to this def.), let {Δ F,Λ} Λ[0,)\left\{ \Delta_{F,\Lambda}\right\}_{\Lambda \in [0,\infty)} be a choice of UV cutoffs for perturbative QFT around this vacuum (def. 5), such that ΛΛ F,Λ\Lambda \mapsto \Lambda_{F,\Lambda} is differentiable.

Then for every choice of UV regularization 𝒮 \mathcal{S}_\infty (prop. 9) the corresponding relative effective actions S eff,ΛS_{eff,\Lambda} (def. 7) satisfy the following differential equation:

ddΛS eff,Λ=121iddΛ(S eff,Λ F,ΛS eff,Λ)| Λ=Λ, \frac{d}{d \Lambda} S_{eff,\Lambda} \;=\; - \frac{1}{2} \frac{1}{i \hbar} \frac{d}{d \Lambda'} \left( S_{eff,\Lambda} \star_{F,\Lambda'} S_{eff,\Lambda} \right)\vert_{\Lambda' = \Lambda} \,,

where on the right we have the star product induced by Δ F,Λ\Delta_{F,\Lambda'} (this def.).

This goes back to (Polchinski 84, (27)). The rigorous formulation and proof is due to (Brunetti-Dütsch-Fredenhagen 09, prop. 5.2, Dütsch 10, theorem 2).

Proof

First observe that for any polynomial observable OPolyObs(E BV-BRST)[[,g,j]]O \in PolyObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ] we have

1(k+2)!ddΛ(O F,Λ F,ΛOk+2factors) =1(k+2)!ddΛ(prodexp(1i<jkΔ F,Λ,δδΦ iδδΦ j)(OOk+2factors)) =1(k+2)!(k+22)=121k!(ddΛO F,ΛO) F,ΛO F,Λ F,ΛOkfactors \begin{aligned} & \frac{1}{(k+2)!} \frac{d}{d \Lambda} ( \underset{ k+2 \, \text{factors} }{ \underbrace{ O \star_{F,\Lambda} \cdots \star_{F,\Lambda} O } } ) \\ & = \frac{1}{(k+2)!} \frac{d}{d \Lambda} \left( prod \circ \exp\left( \hbar \underset{1 \leq i \lt j \leq k}{\sum} \left\langle \Delta_{F,\Lambda} , \frac{\delta}{\delta \mathbf{\Phi}_i} \frac{\delta}{\delta \mathbf{\Phi}_j} \right\rangle \right) ( \underset{ k + 2 \, \text{factors} }{ \underbrace{ O \otimes \cdots \otimes O } } ) \right) \\ & = \underset{ = \frac{1}{2} \frac{1}{k!} }{ \underbrace{ \frac{1}{(k+2)!} \left( k + 2 \atop 2 \right) }} \left( \frac{d}{d \Lambda} O \star_{F,\Lambda} O \right) \star_{F,\Lambda} \underset{ k \, \text{factors} }{ \underbrace{ O \star_{F,\Lambda} \cdots \star_{F,\Lambda} O } } \end{aligned}

Here δδΦ i\frac{\delta}{\delta \mathbf{\Phi}_i} denotes the functional derivative of the iith tensor factor of OO, and the binomial coefficient counts the number of ways that an unordered pair of distinct labels of tensor factors may be chosen from a total of k+2k+2 tensor factors, where we use that the star product F,Λ\star_{F,\Lambda} is commutative (by symmetry of Δ F,Λ\Delta_{F,\Lambda}) and associative (by this prop.).

With this and the defining equality 𝒮 Λ(S eff,Λ)=𝒮(gS int+jA)\mathcal{S}_\Lambda(S_{eff,\Lambda}) = \mathcal{S}(g S_{int} + j A) (22) we compute as follows:

0 =ddΛ𝒮(gS int+jA) =ddΛ𝒮 Λ(S eff,Λ) =(1iddΛS eff,Λ) F,Λ𝒮 Λ(S eff,Λ)+(ddΛ𝒮 Λ)(S eff,Λ) =(1iddΛS eff,Λ) F,Λ𝒮 Λ(S eff,Λ)+12ddΛ(1iS eff,Λ F,Λ1iS eff,Λ)| Λ=Λ F,Λ𝒮 Λ(S eff,Λ) \begin{aligned} 0 & = \frac{d}{d \Lambda} \mathcal{S}(g S_{int} + j A) \\ & = \frac{d}{d \Lambda} \mathcal{S}_\Lambda(S_{eff,\Lambda}) \\ & = \left( \frac{1}{i \hbar} \frac{d}{d \Lambda} S_{eff,\Lambda} \right) \star_{F,\Lambda} \mathcal{S}_\Lambda(S_{eff,\Lambda}) + \left( \frac{d}{d \Lambda} \mathcal{S}_{\Lambda} \right) \left( S_{eff, \Lambda} \right) \\ & = \left( \frac{1}{i \hbar} \frac{d}{d \Lambda} S_{eff,\Lambda} \right) \star_{F,\Lambda} \mathcal{S}_\Lambda(S_{eff,\Lambda}) \;+\; \frac{1}{2} \frac{d}{d \Lambda'} \left( \frac{1}{i \hbar} S_{eff,\Lambda} \star_{F,\Lambda'} \frac{1}{i \hbar} S_{eff, \Lambda} \right) \vert_{\Lambda' = \Lambda} \star_{F,\Lambda} \mathcal{S}_\Lambda \left( S_{eff, \Lambda} \right) \end{aligned}

Acting on this equation with the multiplicative inverse () F,Λ𝒮 Λ(S eff,Λ)(-) \star_{F,\Lambda} \mathcal{S}_\Lambda( - S_{eff,\Lambda} ) (using that F,Λ\star_{F,\Lambda} is a commutative product, so that exponentials behave as usual) this yields the claimed equation.

\,

Renormalization group flow

In perturbative quantum field theory the construction of the scattering matrix 𝒮\mathcal{S}, hence of the interacting field algebra of observables for a given interaction gS intg S_{int} perturbing around a given free field vacuum, involves choices of normalization of time-ordered products/Feynman diagrams (traditionally called "re"-normalizations) encoding new interactions that appear where several of the original interaction vertices defined by gS intg S_{int} coincide.

Whenever a group RGRG acts on the space of observables of the theory such that conjugation by this action takes ("re"-)normalization schemes into each other, then these choices of ("re"-)normalization are parameterized by – or “flow with” – the elements of RGRG. This is called renormalization group flow (prop. 13 below); often called RG flow, for short.

The archetypical example here is the group RGRG of scaling transformations on Minkowski spacetime (def. 4 below), which induces a renormalization group flow (prop. 15 below) due to the particular nature of the Wightman propagator resp. Feynman propagator on Minkowski spacetime (example 5 below). In this case the choice of ("re"-)normalization hence “flows with scale”.

Now the main theorem of perturbative renormalization states that (if only the basic renormalization condition called “field independence” is satisfied) any two choices of ("re"-)normalization schemes 𝒮\mathcal{S} and 𝒮\mathcal{S}' are related by a unique interaction vertex redefinition 𝒵\mathcal{Z}, as

𝒮=𝒮𝒵. \mathcal{S}' = \mathcal{S} \circ \mathcal{Z} \,.

Applied to a parameterization/flow of renormalization choices by a group RGRG this hence induces an interaction vertex redefinition as a function of RGRG. One may think of the shape of the interaction vertices as fixed and only their (adiabatically switched) coupling constants as changing under such an interaction vertex redefinition, and hence then one has coupling constants g jg_j that are parameterized by elements ρ\rho of RGRG:

𝒵 ρ vac ρ:{g j}{g j(ρ)} \mathcal{Z}_{\rho_{vac}}^\rho \;\colon\; \{g_j\} \mapsto \{g_j(\rho)\}

This dependendence is called running of the coupling constants under the renormalization group flow (def. 8 below).

One example of renormalization group flow is that induced by scaling transformations (prop. 15 below). This is the original and main example of the concept (Gell-Mann & Low 54)

In this case the running of the coupling constants may be understood as expressing how “more” interactions (at higher energy/shorter wavelength) become visible (say to experiment) as the scale resolution is increased. In this case the dependence of the coupling g j(ρ)g_j(\rho) on the parameter ρ\rho happens to be differentiable; its logarithmic derivative (denoted “ψ\psi” in Gell-Mann & Low 54) is known as the beta function (Callan 70, Symanzik 70):

β(g)ρg jρ. \beta(g) \coloneqq \rho \frac{\partial g_j}{\partial \rho} \,.

The running of the coupling constants is not quite a representation of the renormalization group flow, but it is a “twisted” representation, namely a group 1-cocycle (prop. 14 below). For the case of scaling transformations this may be called the Gell-Mann-Low renormalization cocycle (Brunetti-Dütsch-Fredenhagen 09).

\,

Proposition

(renormalization group flow)

Let

vac(E BV-BRST,L,Δ H) vac \;\coloneqq\; (E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H )

be a relativistic free vacuum (according to this def.) around which we consider interacting perturbative QFT.

Consider a group RGRG equipped with an action on the Wick algebra of off-shell microcausal polynomial observables with formal parameters adjoined (as in this def.)

rg ():RG×PolyObs(E BV-BRST) mc(())[[g,j]]PolyObs(E BV-BRST) mc(())[[,g,j]], rg_{(-)} \;\colon\; RG \times PolyObs(E_{\text{BV-BRST}})_{mc}((\hbar))[ [ g, j ] ] \longrightarrow PolyObs(E_{\text{BV-BRST}})_{mc}((\hbar))[ [ \hbar, g, j ] ] \,,

hence for each ρRG\rho \in RG a continuous linear map rg ρrg_\rho which has an inverse rg ρ 1RGrg_\rho^{-1} \in RG and is a homomorphism of the Wick algebra-product (the star product H\star_H induced by the Wightman propagator of the given vauum vacvac)

rg ρ(A 1 HA 2)=rg ρ(A 1) Hrg ρ(A 2) rg_\rho( A_1 \star_H A_2 ) \;=\; rg_\rho(A_1) \star_H rg_\rho(A_2)

such that the following conditions hold:

  1. the action preserves the subspace of off-shell polynomial local observables, hence it restricts as

    rg ():RG×LocObs(E BV-BRST)[[,g,j]]g,jLocObs(E BV-BRST)[[,g,j]]g,j rg_{(-)} \;\colon\; RG \times LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j ] ]\langle g,j\rangle \longrightarrow LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j ] ]\langle g,j\rangle
  2. the action respects the causal order of the spacetime support (this def.) of local observables, in that for O 1,O 2LocObs(E BV-BRST)[[,g,j]]O_1, O_2 \in LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ] we have

    (supp(O 1)supp(O 2))AA(supp(rg ρ(O 1))supp(rg ρ(O 2))) \left( supp(O_1) \,{\vee\!\!\!\wedge}\, supp(O_2) \right) \phantom{A} \Rightarrow \phantom{A} \left( supp(rg_\rho(O_1)) \,{\vee\!\!\!\wedge}\, supp(rg_\rho(O_2)) \right)

    for all ρRG\rho \in RG.

Then:

The operation of conjugation by this action on observables induces an action on the set of S-matrix renormalization schemes (this def., this remark), in that for

𝒮:LocObs(E BV-BRST)[[,g,j]]PolyObs(E BV-BRST)(())[[g,j]] \mathcal{S} \;\colon\; LocObs(E_{\text{BV-BRST}})[ [ \hbar , g, j] ] \longrightarrow PolyObs(E_{\text{BV-BRST}})( (\hbar) )[ [ g, j] ]

a perturbative S-matrix scheme around the given free field vacuum vacvac, also the composite

𝒮 ρrg ρ𝒮rg ρ 1:LocObs(E BV-BRST)[[,g,j]]PolyObs(E BV-BRST)(())[[g,j]] \mathcal{S}^\rho \;\coloneqq\; rg_\rho \circ \mathcal{S} \circ rg_{\rho}^{-1} \;\colon\; LocObs(E_{\text{BV-BRST}})[ [ \hbar , g, j] ] \longrightarrow PolyObs(E_{\text{BV-BRST}})( (\hbar) )[ [ g, j] ]

is an S-matrix scheme, for all ρRG\rho \in RG.

More generally, let

vac ρ(E BV-BRST,L ρ,Δ H,ρ) vac_\rho \;\coloneqq\; (E_{\text{BV-BRST}}, \mathbf{L}'_\rho, \Delta_{H,\rho} )

be a collection of gauge fixed free field vacua parameterized by elements ρRG\rho \in RG, all with the same underlying field bundle; and consider rg ρrg_\rho as above, except that it is not an automorphism of any Wick algebra, but an isomorphism between the Wick algebra-structures on various vacua, in that

(23)rg ρ(A 1 H,ρ 1ρ vacA 2)=rg ρ(A 1) H,ρ vacrg ρ(A 2) rg_{\rho}( A_1 \star_{H, \rho^{-1} \rho_{vac}} A_2 ) \;=\; rg_{\rho}(A_1) \star_{H, \rho_{vac}} rg_{\rho}(A_2)

for all ρ,ρ vacRG\rho, \rho_{vac} \in RG

Then if

{𝒮 ρ} ρRG \{ \mathcal{S}_{\rho} \}_{\rho \in RG}

is a collection of S-matrix schemes, one around each of the gauge fixed free field vacua vac ρvac_\rho, it follows that for all pairs of group elements ρ vac,ρRG\rho_{vac}, \rho \in RG the composite

(24)𝒮 ρ vac ρrg ρ𝒮 ρ 1ρ vacrg ρ 1 \mathcal{S}_{\rho_{vac}}^\rho \;\coloneqq\; rg_\rho \circ \mathcal{S}_{\rho^{-1}\rho_{vac}} \circ rg_\rho^{-1}

is an S-matrix scheme around the vacuum labeled by ρ vac\rho_{vac}.

Since therefore each element ρRG\rho \in RG in the group RGRG picks a different choice of normalization of the S-matrix scheme around a given vacuum at ρ vac\rho_{vac}, we call the assignment ρ𝒮 ρ vac ρ\rho \mapsto \mathcal{S}_{\rho_{vac}}^{\rho} a re-normalization group flow.

(Brunetti-Dütsch-Fredenhagen 09, sections 4.2, 5.1, Dütsch 18, section 3.5.3)

Proof

It is clear from the definition that each 𝒮 ρ vac ρ\mathcal{S}^{\rho}_{\rho_{vac}} satisfies the axiom “perturbation” (in this def.).

In order to verify the axiom “causal additivity”, observe, for convenience, that by this prop. it is sufficient to check causal factorization.

So consider O 1,O 2LocObs(E BV-BRST)[[,g,j]]g,jO_1, O_2 \in LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]\langle g,j\rangle two local observables whose spacetime support is in causal order.

supp(O 1)supp(O 2). supp(O_1) \;{\vee\!\!\!\wedge}\; supp(O_2) \,.

We need to show that the

𝒮 ρ vac ρ(O 1+O 2)=𝒮 ρ vac ρ(O 1) H,ρ vac𝒮 vac e ρ(O 2) \mathcal{S}_{\rho_{vac}}^{\rho}(O_1 + O_2) = \mathcal{S}_{\rho_{vac}}^\rho(O_1) \star_{H,\rho_{vac}} \mathcal{S}_{vac_e}^\rho(O_2)

for all ρ,ρ vacRG\rho, \rho_{vac} \in RG.

Using the defining properties of rg ()rg_{(-)} and the causal factorization of 𝒮 ρ 1ρ vac\mathcal{S}_{\rho^{-1}\rho_{vac}} we directly compute as follows:

𝒮 ρ vac ρ(O 1+O 2) =rg ρ𝒮 ρ 1ρ vacrg ρ 1(O 1+O 2) =rg ρ(𝒮 ρ 1ρ vac(rg ρ 1(O 1)+rg ρ 1(O 2))) =rg ρ((𝒮 ρ 1ρ vac(rg ρ 1(O 1))) H,ρ 1ρ vac(𝒮 ρ 1ρ vac(rg ρ 1(O 2)))) =rg ρ(𝒮 ρ 1ρ vac(rg ρ 1(O 1))) H,ρ vacrg ρ(𝒮 ρ 1ρ vac(rg ρ 1(O 2))) =𝒮 ρ vac ρ(O 1) H,ρ vac𝒮 ρ vac ρ(O 2). \begin{aligned} \mathcal{S}_{\rho_{vac}}^\rho(O_1 + O_2) & = rg_\rho \circ \mathcal{S}_{\rho^{-1} \rho_{vac}} \circ rg_\rho^{-1}( O_1 + O_2 ) \\ & = rg_\rho \left( {\, \atop \,} \mathcal{S}_{\rho^{-1}\rho_{vac}} \left( rg_\rho^{-1}(O_1) + rg_\rho^{-1}(O_2) \right) {\, \atop \,} \right) \\ & = rg_\rho \left( {\, \atop \,} \left( \mathcal{S}_{\rho^{-1}\rho_{vac}}\left(rg_\rho^{-1}(O_1)\right) \right) \star_{H, \rho^{-1} \rho_{vac}} \left( \mathcal{S}_{ \rho^{-1} \rho_{vac} }\left(rg_\rho^{-1}(O_2)\right) \right) {\, \atop \,} \right) \\ & = rg_\rho \left( {\, \atop \,} \mathcal{S}_{\rho^{-1} \rho_{vac}}\left(rg_{\rho^{-1}}(O_1)\right) {\, \atop \,} \right) \star_{H, \rho_{vac}} rg_\rho \left( {\, \atop \,} \mathcal{S}_{\rho^{-1} \rho_{vac}}\left( rg_\rho^{-1}(O_2)\right) {\, \atop \,} \right) \\ & = \mathcal{S}^\rho_{\rho_{vac}}( O_1 ) \, \star_{H, \rho_{vac}} \, \mathcal{S}_{\rho_{vac}}^\rho(O_2) \,. \end{aligned}
Definition

(running coupling constants)

Let

vacvac e(E BV-BRST,L,Δ H) vac \coloneqq vac_e \;\coloneqq\; (E_{\text{BV-BRST}}, \mathbf{L}', \Delta_H )

be a relativistic free vacuum (according to this def.) around which we consider interacting perturbative QFT, let 𝒮\mathcal{S} be an S-matrix scheme around this vacuum and let rg ()rg_{(-)} be a renormalization group flow according to prop. 13, such that each re-normalized S-matrix scheme 𝒮 vac ρ\mathcal{S}_{vac}^\rho satisfies the renormalization condition “field independence”.

Then by the main theorem of perturbative renormalization (this prop.) there is for every pair ρ 1,ρ 2RG\rho_1, \rho_2 \in RG a unique interaction vertex redefinition

𝒵 ρ vac ρ:LocObs(E BV-BRST)[[,g,j]]LocObs(E BV-BRST)[[,g,j]] \mathcal{Z}_{\rho_{vac}}^{\rho} \;\colon\; LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ] \longrightarrow LocObs(E_{\text{BV-BRST}})[ [ \hbar, g, j] ]

which relates the corresponding two S-matrix schemes via

(25)𝒮 ρ vac ρ=𝒮 ρ vac𝒵 ρ vac ρ. \mathcal{S}_{\rho_{vac}}^{\rho} \;=\; \mathcal{S}_{\rho_{vac}} \circ \mathcal{Z}_{\rho_{vac}}^\rho \,.

If one thinks of an interaction vertex, hence a local observable gS int+jAg S_{int}+ j A, as specified by the (adiabatically switched) coupling constants g jC cp (Σ)gg_j \in C^\infty_{cp}(\Sigma)\langle g \rangle multiplying the corresponding interaction Lagrangian densities L int,jΩ Σ p+1,0(E BV-BRST)\mathbf{L}_{int,j} \in \Omega^{p+1,0}_\Sigma(E_{\text{BV-BRST}}) as

gS int=jτ Σ(g jL int,j) g S_{int} \;=\; \underset{j}{\sum} \tau_\Sigma \left( g_j \mathbf{L}_{int,j} \right)

(where τ Σ\tau_\Sigma denotes transgression of variational differential forms) then 𝒵 ρ 1 ρ 2\mathcal{Z}_{\rho_1}^{\rho_2} exhibits a dependency of the (adiabatically switched) coupling constants g jg_j of the renormalization group flow parameterized by ρ\rho. The corresponding functions

𝒵 ρ vac ρ(gS int):(g j)(g j(ρ)) \mathcal{Z}_{\rho_{vac}}^{\rho}(g S_{int}) \;\colon\; (g_j) \mapsto (g_j(\rho))

are then called running coupling constants.

(Brunetti-Dütsch-Fredenhagen 09, sections 4.2, 5.1, Dütsch 18, section 3.5.3)

Proposition

(running coupling constants are group cocycle over renormalization group flow)

Consider running coupling constants

𝒵 ρ vac ρ:(g j)(g j(ρ)) \mathcal{Z}_{\rho_{vac}}^{\rho} \;\colon\; (g_j) \mapsto (g_j(\rho))

as in def. 8. Then for all ρ vac,ρ 1,ρ 2RG\rho_{vac}, \rho_1, \rho_2 \in RG the following equality is satisfied by the “running functions” (25):

𝒵 ρ vac ρ 1ρ 2=𝒵 ρ vac ρ 1(σ ρ 1𝒵 ρ 1ρ vac ρ 2σ ρ 1 1). \mathcal{Z}_{\rho_{vac}}^{\rho_1 \rho_2} \;=\; \mathcal{Z}_{\rho_{vac}}^{\rho_1} \circ \left( \sigma_{\rho_1} \circ \mathcal{Z}_{\rho^{-1} \rho_{vac}}^{\rho_2} \circ \sigma_{\rho_1}^{-1} \right) \,.

(Brunetti-Dütsch-Fredenhagen 09 (69), Dütsch 18, (3.325))

Proof

Directly using the definitions, we compute as follows:

𝒮 ρ vac𝒵 ρ vac ρ 1ρ 2 =𝒮 ρ vac ρ 1ρ 2 =σ ρ 1σ ρ 2𝒮 ρ 2 1ρ 1 1ρ vacσ ρ 2 1=𝒮 ρ 1 1ρ vac ρ 2=𝒮 ρ 1 1ρ vac𝒵 ρ 1 1ρ vac ρ 2σ ρ 1 1 =σ ρ 1𝒮 ρ 1 1ρ vacσ ρ 1 1σ ρ 1=id=𝒮 ρ vac𝒵 ρ vac ρ 1σ ρ 1𝒵 ρ 1 1ρ vac ρ 2σ ρ 1 1 =𝒮 ρ vac𝒵 ρ vac ρ 1σ ρ 1𝒵 ρ 1 1ρ vac ρ 2σ ρ 1 1 \begin{aligned} \mathcal{S}_{\rho_{vac}} \circ \mathcal{Z}_{\rho_{vac}}^{\rho_1 \rho_2} & = \mathcal{S}_{\rho_{vac}}^{\rho_1 \rho_2 } \\ & = \sigma_{\rho_1} \circ \underset{ = \mathcal{S}_{\rho_1^{-1} \rho_{vac}}^{\rho_2} = \mathcal{S}_{\rho_1^{-1} \rho_{vac}} \circ \mathcal{Z}_{\rho_1^{-1} \rho_vac}^{\rho_2} }{ \underbrace{ \sigma_{\rho_2} \circ \mathcal{S}_{\rho_2^{-1}\rho_1^{-1}\rho_{vac}} \circ \sigma_{\rho_2}^{-1} }} \circ \sigma_{\rho_1}^{-1} \\ & = \underset{ = \mathcal{S}_{\rho_{vac}} \circ \mathcal{Z}_{\rho_{vac}}^{\rho_1} \circ \sigma_{\rho_1} }{ \underbrace{ \sigma_{\rho_1} \circ \mathcal{S}_{\rho_1^{-1} \rho_{vac}} \circ \overset{ = id }{ \overbrace{ \sigma_{\rho_1}^{-1} \circ \sigma_{\rho_1} } } }} \circ \mathcal{Z}_{\rho_1^{-1} \rho_{vac}}^{\rho_2} \circ \sigma_{\rho_1}^{-1} \\ & = \mathcal{S}_{\rho_{vac}} \circ \mathcal{Z}_{\rho_{vac}}^{\rho_1} \circ \underbrace{ \sigma_{\rho_1} \circ \mathcal{Z}_{\rho_1^{-1} \rho_{vac}}^{\rho_2} \circ \sigma_{\rho_1}^{-1} } \end{aligned}

This demonstrates the equation between vertex redefinitions to be shown after composition with an S-matrix scheme. But by the uniqueness-clause in the main theorem of perturbative renormalization (theorem 2) the composition operation 𝒮 ρ vac()\mathcal{S}_{\rho_{vac}} \circ (-) as a function from vertex redefinitions to S-matrix schemes is injective. This implies the equation itself.

\,

Gell-Mann & Low RG flow

We discuss (prop. 15 below) that, if the field species involved have well-defined mass dimension (example 5 below) then scaling transformations on Minkowski spacetime (example 4 below) induce a renormalization group flow (def. 13). This is the original and main example of renormalization group flows (Gell-Mann& Low 54).

Example

(scaling transformations and mass dimension)

Let

EfbΣ E \overset{fb}{\longrightarrow} \Sigma

be a field bundle which is a trivial vector bundle over Minkowski spacetime Σ= p,1 p+1\Sigma = \mathbb{R}^{p,1} \simeq_{\mathbb{R}} \mathbb{R}^{p+1}.

For ρ(0,)\rho \in (0,\infty) \subset \mathbb{R} a positive real number, write

Σ ρ Σ x ρx \array{ \Sigma &\overset{\rho}{\longrightarrow}& \Sigma \\ x &\mapsto& \rho x }

for the operation of multiplication by ρ\rho using the real vector space-structure of the Cartesian space p+1\mathbb{R}^{p+1} underlying Minkowski spacetime.

By pullback this acts on field histories (sections of the field bundle) via

Γ Σ(E) ρ * Γ Σ(E) Φ Φ(ρ()). \array{ \Gamma_\Sigma(E) &\overset{\rho^\ast}{\longrightarrow}& \Gamma_\Sigma(E) \\ \Phi &\mapsto& \Phi(\rho(-)) } \,.

Let then

ρvac ρ(E BV-BRST,L ρ,Δ H,ρ) \rho \mapsto vac_\rho \;\coloneqq\; (E_{\text{BV-BRST}}, \mathbf{L}'_{\rho}, \Delta_{H,\rho} )

be a 1-parameter collection of relativistic free vacua on that field bundle, according to this def., and consider a decomposition into a set SpecSpec of field species (this def.) such that for each spSpecsp \in Spec the collection of Feynman propagators Δ F,ρ,sp\Delta_{F,\rho,sp} for that species scales homogeneously in that there exists

dim(sp) dim(sp) \in \mathbb{R}

such that for all ρ\rho we have (using generalized functions-notation)

(26)ρ 2dim(sp)Δ F,1/ρ,sp(ρx)=Δ F,sp,ρ=1(x). \rho^{ 2 dim(sp) } \Delta_{F, 1/\rho, sp}( \rho x ) \;=\; \Delta_{F,sp, \rho = 1}(x) \,.

Typically ρ\rho rescales a mass parameter, in which case dim(sp)dim(sp) is also called the mass dimension of the field species spsp.

Let finally

PolyObs(E) σ ρ PolyObs(E) Φ sp a(x) ρ dim(sp)Φ a(ρ 1x) \array{ PolyObs(E) & \overset{ \sigma_\rho }{\longrightarrow} & PolyObs(E) \\ \mathbf{\Phi}_{sp}^a(x) &\mapsto& \rho^{- dim(sp)} \mathbf{\Phi}^a( \rho^{-1} x ) }

be the function on off-shell polynomial observables given on field observables Phi a(x)\mathbf{Phi}^a(x) by pullback along ρ 1\rho^{-1} followed by multiplication by ρ\rho taken to the negative power of the mass dimension, and extended from there to all polynomial observables as an algebra homomorphism.

This constitutes an action of the group

RG( +,) RG \coloneqq \left( \mathbb{R}_+, \cdot \right)

of positive real numbers (under multiplication) on polynomial observables, called the group of scaling transformations for the given choice of field species and mass parameters.

(Dütsch 18, def. 3.19)

Example

(mass dimension of scalar field)

Consider the Feynman propagator Δ F,m\Delta_{F,m} of the free real scalar field on Minkowski spacetime Σ= p,1\Sigma = \mathbb{R}^{p,1} for mass parameter m(0,)m \in (0,\infty); a Green function for the Klein-Gordon equation.

Let the group RG( +,)RG \coloneqq (\mathbb{R}_+, \cdots) of scaling transformations ρ +\rho \in \mathbb{R}_+ on Minkowski spacetime (def. 4) act on the mass parameter by inverse multiplication

(ρ,Δ F,m)Δ F,ρ 1m(ρ()). (\rho , \Delta_{F,m}) \mapsto \Delta_{F,\rho^{-1}m}(\rho (-)) \,.

Then we have

Δ F,ρ 1m(ρ())=ρ (p+1)+2Δ F,1(x) \Delta_{F,\rho^{-1}m}(\rho (-)) \;=\; \rho^{-(p+1) + 2} \Delta_{F,1}(x)

and hence the corresponding mass dimension (def. 4) of the real scalar field on p,1\mathbb{R}^{p,1} is

dim(scalar field)=(p+1)/21. dim(\text{scalar field}) = (p+1)/2 - 1 \,.
Proof

By (this prop.) the Feynman propagator in question is given by the Cauchy principal value-formula (in generalized function-notation)

Δ F,m(x) =limϵ(0,)ϵ0+i(2π) p+1 e ik μx μk μk μ(mc) 2±iϵdk 0d pk. \begin{aligned} \Delta_{F,m}(x) & = \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu x^\mu} }{ - k_\mu k^\mu - \left( \tfrac{m c}{\hbar} \right)^2 \pm i \epsilon } \, d k_0 \, d^p \vec k \,. \end{aligned}

By applying change of integration variables kρ 1kk \mapsto \rho^{-1} k in the Fourier transform this becomes

Δ F,ρ 1m(ρx) =limϵ(0,)ϵ0+i(2π) p+1 e ik μρx μk μk μ(ρ 1mc) 2±iϵdk 0d pk =ρ (p+1)limϵ(0,)ϵ0+i(2π) p+1 e ik μx μρ 2k μk μρ 2(mc) 2±iϵdk 0d pk =ρ (p+1)+2limϵ(0,)ϵ0+i(2π) p+1 e ik μx μk μk μ(mc) 2±iϵdk 0d pk =ρ (p+1)+2Δ F,m(x) \begin{aligned} \Delta_{F,\rho^{-1}m}(\rho x) & = \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu \rho x^\mu} }{ - k_\mu k^\mu - \left( \rho^{-1} \tfrac{m c}{\hbar} \right)^2 \pm i \epsilon } \, d k_0 \, d^p \vec k \\ & = \rho^{-(p+1)} \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu x^\mu} }{ - \rho^{-2} k_\mu k^\mu - \rho^{-2} \left( \tfrac{m c}{\hbar} \right)^2 \pm i \epsilon } \, d k_0 \, d^p \vec k \\ & = \rho^{-(p+1)+2} \underset{ {\epsilon \in (0,\infty)} \atop {\epsilon \to 0} }{\lim} \frac{+i}{(2\pi)^{p+1}} \int \int_{-\infty}^\infty \frac{ e^{i k_\mu x^\mu} }{ - k_\mu k^\mu - \left( \tfrac{m c}{\hbar} \right)^2 \pm i \epsilon } \, d k_0 \, d^p \vec k \\ & = \rho^{-(p+1) + 2} \Delta_{F,m}(x) \end{aligned}
Proposition

(scaling transformations are renormalization group flow)

Let

vacvac m(E BV-BRST,L,Δ H,m) vac \coloneqq vac_m \coloneqq (E_{\text{BV-BRST}}, \mathbf{L}', \Delta_{H,m})

be a relativistic free vacua on that field bundle, according to this def. equipped with a decomposition into a set SpecSpec of field species (this def.) such that for each spSpecsp \in Spec the collection of Feynman propagators the corresponding field species has a well-defined mass dimension dim(sp)dim(sp) (def. 4)

Then the action of the group RG( +,)RG \coloneqq (\mathbb{R}_+, \cdot) of scaling transformations (def. 4) is a renormalization group flow in the sense of this prop..

(Dütsch 18, exercise 3.20)

Proof

It is clear that rescaling preserves causal order and the renormalization condition of “field indepencen”.

The condition we need to check is that for A 1,A 2PolyObs(E BV-BRST) mc[[,g,j]]A_1, A_2 \in PolyObs(E_{\text{BV-BRST}})_{mc}[ [ \hbar, g, j] ] two microcausal polynomial observables we have for any ρ,ρ vac +\rho, \rho_{vac} \in \mathbb{R}_+ that

σ ρ(A 1 H,ρ 1ρ vaccA 2)=σ ρ(A 1) H,ρ vacσ ρ(A 2). \sigma_\rho \left( A_1 \star_{H, \rho^{-1} \rho_{vac} c} A_2 \right) \;=\; \sigma_\rho(A_1) \star_{H,\rho_{vac}} \sigma_\rho(A_2) \,.

By the assumption of decomposition into free field species spSpecsp \in Spec, it is sufficient to check this for each species Δ H,sp\Delta_{H,sp}. Moreover, by the nature of the star product on polynomial observables, which is given by iterated contractions with the Wightman propagator, it is sufficient to check this for one such contraction.

Observe that the scaling behaviour of the Wightman propagator Δ H,m\Delta_{H,m} is the same as the behaviour (26) of the correspponding Feynman propagator. With this we directly compute as follows:

σ ρ(Φ(x)) F,ρ vacmσ ρ(Φ(y) =ρ 2dimΦ(ρ 1x) F,ρ vacmΦ(ρ 1y) =ρ 2dimΔ F,ρ vacm(ρ 1(xy)) =Δ F,ρ 1ρ vacm(x,y)1 =rg ρ(Δ F,ρ 1ρ vacm(x,y)1) =rg ρ(Φ(x) F,ρ 1ρ vacmΦ(y)). \begin{aligned} \sigma_\rho (\mathbf{\Phi}(x)) \star_{F, \rho_{vac} m} \sigma_\rho (\mathbf{\Phi}(y) & = \rho^{-2 dim } \mathbf{\Phi}(\rho^{-1} x) \star_{F, \rho_{vac} m} \mathbf{\Phi}(\rho^{-1} y) \\ & = \rho^{-2 dim } \Delta_{F, \rho_{vac} m}(\rho^{-1}(x-y)) \\ & = \Delta_{F, \rho^{-1}\rho_{vac}m }(x,y) \mathbf{1} \\ & = rg_{\rho}\left( \Delta_{F, \rho^{-1}\rho_{vac}m }(x,y) \mathbf{1} \right) \\ & = rg_{\rho} \left( \mathbf{\Phi}(x) \star_{F, \rho^{-1} \rho_{vac} m} \mathbf{\Phi}(y) \right) \end{aligned} \,.

\,

Dimensional regularization

Discussion of renormalization via dimensional regularization in the rigorous context of causal perturbation theory is due to Dütsch-Fredenhagen-Keller-Rejzner 14, section 4.

Connes-Kreimer renormalization

Discussion of the Connes-Kreimer Hopf algebraic renormalization in causal perturbation theory is in Dütsch-Fredenhagen-Keller-Rejzner 14, section 5.

(…)

\,

BPHZ and Hopf-algebraic renormalization

The phenomenon

In the study of perturbative quantum field theory one is concerned with functions – called amplitudes – that take a collection of graphs – called Feynman graphs – to Laurent polynomials in a complex variable zz – called the (dimensional) regularization parameter

Amplitude:CertainGraphsLaurentPolynomials Amplitude : CertainGraphs \to LaurentPolynomials

and wishes to extract a “meaningful” finite component when evaluated at vanishing regularization parameter z=0z = 0.

A prescription – called renormalization scheme – for adding to a given amplitude in a certain recursive fashion further terms – called counterterms – such that the resulting modified amplitude – called the renormalized amplitude – is finite at z=0z=0 was once given by physicists and is called the BPHZ-procedure .

This procedure justifies itself mainly through the remarkable fact that the numbers obtained from it match certain numbers measured in particle accelerators to fantastic accuracy.

Its combinatorial Hopf-algebraic interpretation

The combinatorial Hopf algebraic approach to perturbative quantum field theory, see for instance

starts with the observation that the BPHZ-procedure can be understood

  • by noticing that there is secretly a natural group structure on the collection of amplitudes;

  • which is induced from the fact that there is secretly a natural Hopf algebra structure on the vector space whose basis consists of graphs;

  • and with respect to which the BPHZ-procedure is simply the Birkhoff decomposition of group valued functions on the circle into a divergent and a finite part.

The Hopf algebra structure on the vector space whose basis consists of graphs can be understood most conceptually in terms of pre-Lie algebras.

The Connes-Kreimer theorem

A Birkhoff decomposition of a loop ϕ:S 1G\phi : S^1 \to G in a complex group GG is a continuation of the loop to

  • a holomorphic function ϕ +\phi_+ on the standard disk inside the circle;

  • a holomorphic function ϕ \phi_- on the complement of this disk in the projective complex plane

  • such that on the unit circle the original loop is reproduced as

    ϕ=ϕ +(ϕ ) 1, \phi = \phi_+ \cdot (\phi_-)^{-1} \,,

    with the product and the inverse on the right taken in the group GG.

    Notice that by the assumption of holomorphicity ϕ +(0)\phi_+(0) is a well defined element of GG.

Theorem

(Connes-Kreimer)

  1. If GG is the group of characters on any graded connected commutative Hopf algebra HH

    G=Hom(H,) G = Hom(H,\mathbb{C})

    then the Birkhoff decomposition always exists and is given by the formula

    ϕ :(XH)Counit(X)PolePartOf(Product(ϕ ϕ)(1(1Counit))Coproduct(X)). \phi_- : (X \in H) \mapsto Counit(X) - PolePartOf( Product(\phi_- \otimes \phi) \circ (1 \otimes (1 - Counit)) \circ Coproduct (X) ) \,.
  2. There is naturally the structure of a Hopf algebra, H=GraphsH = Graphs, on the graphs considered in quantum field theory. As an algebra this is the free commutative algebra on the “1-particle irreducible graphs”. Hence QFT amplitudes can be regarded as characters on this Hopf algebra.

  3. The BPHZ renormalization-procedure for amplitudes is nothing but the first item applied to the special case of the second item.

Proof

The proof is given in

The Hopf-algebra perspective on QFT

This result first of all makes Hopf algebra an organizational principle for (re-)expressing familiar operations in quantum field theory.

Computing the renormalization ϕ +\phi_+ of an amplitude ϕ\phi amounts to using the above formula to compute the counterterm ϕ \phi_- and then evaluating the right hand side of

ϕ + renormalizedamplitude=ϕ amplitude convolutionproductϕ counterterm, \underbrace{\phi_+}_{renormalized amplitude} = \underbrace{\phi}_{amplitude} \underbrace{\cdot}_{convolution product} \underbrace{\phi_-}_{counterterm} \,,

where the product is the group product on characters, hence the convolution product of characters.

Every elegant reformulation has in it the potential of going beyond mere reformulation by allowing to see structures invisible in a less natural formulation. For instance Dirk Kreimer claims that the Hopf algebra language allows him to see patterns in perturbative quantum gravity previously missed.

Gauge theory and BV-BRST with Hopf algebra

Walter von Suijlekom is thinking about the Hopf-algebraic formulation of BRST-BV methods in nonabelian gauge theory

In his nicely readable

  • Walter von Suijlekom?, Renormalization of gauge fields using Hopf algebra, (arXiv)

he reviews the central idea: the BRST formulation of Yang-Mills theory manifests itself at the level of the resulting bare i.e. unnormalized amplitudes in certain relations satisfied by these, the Slavnov-Taylor identities .

Renormalization of gauge theories is consistent only if these relations are still respected by renormalized amplitudes, too. We can reformulate this in terms of Hopf algebra now:

the relations between amplitudes to be preserved under renormalization must define a Hopf ideal in the Hopf algebra of graphs.

Walter von Suijlekom proves this to be the case for Slavnov-Taylor in his theorem 9 on p. 12

As a payoff, he obtains a very transparent way to prove the generalization of Dyson’s formula to nonabelian gauge theory, which expresses renormalized Green’s functions in terms of unrenormalized Green’s functions “at bare coupling”. This is his corollary 12 on p. 13.

In the context of BRST-BV quantization these statements are subsumed, he says, by the structure encoded in the Hopf ideal which corresponds to imposing the BV-master equation. See also (Suijlekom).

Lattice renormalization

Of theories in BV-CS form

In (Costello 07) a renormalization procedure is discussed that applies to theories that are given by action functionals which can be given in the form

S(ϕ)=ϕ,Qϕ+I(ϕ) S(\phi) = \langle \phi , Q \phi \rangle + I(\phi)

where

  1. the fields ϕ\phi are sections of a graded field bundle EE on which QQ is a differential, ,\langle -,-\rangle a compatible antibracket pairing such that (E,Q,)(E,Q, \langle \rangle) is a free field theory (as discussed there) in BV-BRST formalism;

  2. II is an interaction that is at least cubic.

These are action functionals that are well adapted to BV-BRST formalism and for which there is a quantization to a factorization algebra of observables.

Most of the fundamental theories in physics are of this form, notably Yang-Mills theory. In particular also all theories of infinity-Chern-Simons theory-type coming from binary invariant polynomials are perturbatively of this form, notably ordinary 3d Chern-Simons theory.

For a discussion of just the simple special case of 3d CS see (Costello 11, chapter 5.4 and 5.14).

For comparison of the following with other renormalization schemes, see at (Costello 07, section 1.7).

  1. The setup

  2. Operator (heat) kernels and propagators

  3. The renormalization group operator

  4. The path integral

  5. Renormalized action

  6. Renormalization

The setup

Definition

A free field theory (E,,,.Q)(E, \langle, -,-\rangle. Q):

kinematics:

Write cΓ cp(E)\mathcal{E}_c \coloneqq \Gamma_{cp}(E) for the space of sections of the field bundle of compact support. Write

,: c c \langle -,-\rangle \;\colon\; \mathcal{E}_c \otimes \mathcal{E}_c \to \mathbb{C}

for the induced pairing on sections

ϕ,ψ= xXϕ(x),ψ(x) loc. \langle \phi, \psi\rangle = \int_{x \in X} \langle \phi(x), \psi(x)\rangle_{loc} \,.

The paring being non-degenerate means that we have an isomorphism EE *Dens XE \stackrel{\simeq}{\to} E^* \otimes Dens_X and we write

E !E *Dens X. E^! \coloneqq E^* \otimes Dens_X \,.

dynamics

  • A differential operator on sections of the field bundle

    Q:𝒞 Q \;\colon\; \mathcal{E} \to \mathcal{C}

    of degree 1 such that

    1. (,Q)(\mathcal{E}, Q) is an elliptic complex;

    2. QQ is self-adjoint with respect to ,\langle -,-\rangle in that for all fields ϕ,ψ c\phi,\psi \in \mathcal{E}_c of homogeneous degree we have ϕ,Qψ=(1) |ϕ|Qϕ,ψ\langle \phi , Q \psi\rangle = (-1)^{{\vert \phi\vert}} \langle Q \phi, \psi\rangle.

From this data we obtain:

  • The action functional S: cS \colon \mathcal{E}_c \to \mathbb{C} of this corresponding free field theory is

    S:ϕ Xϕ,Qϕ. S \;\colon\; \phi \mapsto \int_X \langle \phi, Q \phi\rangle \,.
  • The classical BV-complex is the symmetric algebra Sym !Sym \mathcal{E}^! of sections of E !E^! equipped with the induced action of the differential QQ and the pairing

    {α,β} xXα(x),β(x). \{\alpha,\beta\} \coloneqq \int_{x \in X} \langle \alpha(x), \beta(x)\rangle \,.
Definition

An interaction term II \in \cdots

(…)

Definition

A gauge fixing operator Q GFQ^{GF} such that

H[Q,Q GF] H \coloneqq [Q, Q^{GF}]

is a generalized Laplace operator.

Operator (heat) kernels and propagators

Definition

For K=KKK = \sum K' \otimes K'' \in \mathcal{E} \otimes \mathcal{E} the corresponding convolution operator K:K \star \colon \mathcal{E} \to \mathcal{E} is

Ke=(1) |e|KK,e. K \star e = (-1)^{\vert e\vert} \sum K' \otimes \langle K'', e\rangle \,.
Definition

For H:H \colon \mathcal{E} \to \mathcal{E} a linear operator, a heat kernel for it is a function K (): >0K_{(-)} \colon \mathbb{R}_{\gt 0} \to \mathcal{E}\otimes \mathcal{E} such that for each t >0t \in \mathbb{R}_{\gt 0} the convolution with K tK_t, def. 12, reproduces the exponential of tcdotHt\cdotH:

K t=exp(tH)t >0. K_t \star = \exp(-t H) \;\;\;\;\; \forall t \in \mathbb{R}_{\gt 0} \,.
Proposition

For HH a generalized Laplace operator such as the [Q,Q GF][Q,Q^{GF}] of def. 11 there is a unique heat kernel which is moreover a smooth function of tt.

Definition

For ϕ=ϕϕSym 2\phi = \sum \phi' \otimes \phi'' \in Sym^2 \mathcal{E} write

ϕ12 ϕ ϕ. \partial_{\phi} \coloneqq \frac{1}{2} \sum \partial_{\phi''} \partial_{\phi'} \,.

This is (Costello 07, p. 32).

Proposition
[ ϕ,Q]= Qϕ. [\partial_\phi, Q] = \partial_{Q \phi} \,.

The renormalization group operator

Definition

For Q GFQ^{GF} the gauge fixing operator of def. 11, and K tK_t the heat kernel of the corresponding generalized Laplace operator H=[Q,Q GF]H = [Q, Q^{GF}] by prop. 16, write for ϵ,T >0\epsilon, T \in \mathbb{R}_{\gt 0}

P(ϵ,T) ϵ T(Q GF1)K tdt. P(\epsilon, T) \coloneqq \int_{\epsilon}^T (Q^{GF} \otimes 1) K_t \; d t \;\;\;\;\; \in \mathcal{E} \otimes \mathcal{E}\,.

(Costello 07, p. 33)

Definition

Write

Γ(P(ϵ,T),S)log(exp( P(ϵ,T))exp(I/))𝒪(,[[]]), \Gamma(P(\epsilon,T), S) \coloneqq \hbar log \left( \exp\left(\hbar \partial_{P(\epsilon,T)}\right) \exp\left(I/\hbar\right) \right) \;\; \in \mathcal{O}(\mathcal{E}, \mathbb{C}[ [ \hbar ] ]) \,,

where \partial_{\cdots} is given by def. 14.

(Costello 07, def. 6.6.1)

The path integral

Proposition

If X=*X = * is the point, then the path integral over the action functional exists as an ordinary integral and is equal to

log x(ImQ GF) exp(12x,Qx+I(x+a)dμ)=Γ(P(0,),I)(a) \hbar log \int_{x \in (Im Q^{GF})_{\mathbb{R}}} \exp\left( \tfrac{1}{2} \langle x, Q x\rangle + I(x + a) d \mu \right) = \Gamma(P(0,\infty), I)(a)

This is (Costello 07, lemma 6.6.2).

Remark

For XX of positive dimension, the limit

limϵ0Γ(P(ϵ,),I)(a) \underset{\epsilon \to 0}{\lim} \Gamma(P(\epsilon,\infty), I)(a)

does not in general exist. Renormalization is the process of adding \hbar-corrections to the action – the counterterms – such as to make it exist after all. In this case we may regard the limit, by prop. 18, as the definition of the path integral.

Renormalized action

Definition

Given the action functional S:ϕ12ϕ,Qphi+I(ϕ)S \colon \phi \mapsto \frac{1}{2}\langle \phi, Q phi\rangle + I(\phi), a renormalization is a power series

I R(,ϵ)=I()i>0kk iI i,k CT(ϵ) I^R(\hbar, \epsilon) = I(\hbar) - \underset{{i \gt 0} \atop { k \geq k}}{\sum} \hbar^i I^{CT}_{i,k}(\epsilon)

such that the limit

limϵ0Γ(P(ϵ,T),I R(ϵ)), \underset{\epsilon \to 0}{\lim} \Gamma(P(\epsilon,T), I^R(\epsilon)) \,,

by def. 16, exists.

The I i,k CTI^{CT}_{i,k} are called the counterterms.

Renormalization

Definition

A renormalization scheme is a decomposition of functions 𝒜\mathcal{A} on (0,1)(0,1), as a vector space, into a direct sum

𝒜𝒜 0𝒜 >0 \mathcal{A} \simeq \mathcal{A}_{\geq 0} \oplus \mathcal{A}_{\gt 0}

such that the functions f𝒜 0f \in \mathcal{A}_{\geq 0} are non-singular in that limϵ0f(ϵ)\underset{\epsilon \to 0}{\lim} f(\epsilon) exists.

Hence this is a choice of picking the singularities in functions that are not necessarily defined at ϵ=0\epsilon = 0.

Theorem

Given any choice of renormalization scheme, def. 18, there exists a unique choice of counterterms {I k,i CT}\{I^{CT}_{k,i}\}, def. 17 such that

  • each counterterm is in the chosen 𝒜 >0\mathcal{A}_{\gt 0} as a function of ϵ\epsilon;

  • for k>0k \gt 0 the germ of a counterterm at xXx \in X depends only on the germ of the given field theory data (E,Q,Q GF)(E, Q, Q^{GF}) at that point.

This is (Costello 07, theorem B, p. 38).

Definition

Given a renormalization I RI^{R}, write for all T ht0T \in \mathbb{R}_{\ht 0}

Γ R(P(0,T),I)ϵ0Γ(P(ϵ,R),II CT(ϵ)). \Gamma^R(P(0,T), I) \coloneqq \underset{\epsilon \to 0}{\to} \Gamma(P(\epsilon, R), I - I^{CT}(\epsilon)) \,.
Remark

We think of Γ R(P(0,T),I)\Gamma^R(P(0,T), I) as the renormalized effective action of the original action at scale TT.

Proposition

The renormalization group flow

Γ R(P(0,T),I)=Γ(P(T,T),Γ R(P(0,T),I)) \Gamma^R(P(0,T'), I) = \Gamma(P(T,T'), \; \Gamma^R(P(0,T), I ))

holds.

(Costello 07, lemma 9.0.6).

Examples

Vacuum energy and Cosmological constant

The renormalization freedom in perturbative quantization of gravity (perturbative quantum gravity) induces freedom in the choice of vacuum expectation value of the stress-energy tensor and hence in the cosmological constant.

For details see there.

References

After the original informal suggestions by Schwinger-Tomonaga-Feynman-Dyson

  • Freeman Dyson, The raditation theories of Tomonaga, Schwinger and Feynman, Phys. Rev. 75, 486, 1949 (pdf)

the mathematics of renormalization was finally understood and summarized in the 1975 Erice Majorana School:

  • G. Velo and Arthur Wightman (eds.) Renormalization Theory Proceedings of the 1975 Erice summer school, NATO ASI Series C 23, D. Reidel, Dordrecht, 1976

which included causal perturbation theory, BPHZ renormalization, proof of the forest formula? and the BRST complex method for gauge theory.

Little advancement happened until the identification of Hopf algebra structure in the forest formula? due to

  • Dirk Kreimer, On the Hopf algebra structure of perturbative quantum field theories, Adv. Theor. Math. Phys. 2 , 303 (1998) (q-alg/9707029)

This finally triggered the formulation of causal perturbation theory in terms of Feynman diagrams in

Thus the original “dark art” of the construction of perturbative quantum field theory via renormalization finds a complete rigorous formulation.

In

it is shown that (at least under favorable circumstances) the construction of perturbative quantum field theory via causal perturbation theory is equivalently Fedosov deformation quantization of the given interacting Lagrangian density, thus identifying the renormalization freedom with the Freedom in choosing a deformation quantization.

Review

A textbook account of ("re"-)normalization in causal perturbation theory is in

Informal introductions include

In causal perturbation theory

The idea of Epstein-Glaser renormalization is due to

following precursors in

This was formulated in terms of splittings of distributions. The equivalent formulation in terms of extensions of distributions is due to

  • G. Popineau and Raymond Stora, A pedagogical remark on the main theorem of perturbative renormalization theory, Nucl. Phys. B 912 (2016), 70–78, preprint: LAPP–TH, Lyon (1982).

  • Raymond Stora, Differential algebras in Lagrangean field theory, Lectures at ETH, Zürich, 1993, unpublished

  • Romeo Brunetti, Klaus Fredenhagen, Microlocal Analysis and Interacting Quantum Field Theories: Renormalization on Physical Backgrounds, Commun. Math. Phys. 208:623-661 (2000) (arXiv:math-ph/9903028)

Exposition includes

The resulting Stückelberg-Petermann renormalization group is due to

The relation of Epstein-Glaser/Stückelberg-Petermann to the renormalization group flow of

  • Murray Gell-Mann and F. E. Low, Quantum Electrodynamics at Small Distances, Phys. Rev. 95 (5) (1954), 1300–1312 (pdf)

and the effective quantum field theory of

  • Kenneth Wilson, Renormalization group and critical phenomena , I., Physical review B 4(9) (1971).

  • Joseph Polchinski, Renormalization and effective Lagrangians , Nuclear Phys. B B231, 1984 (pdf)

is due to

Review is in

For more see at perturbation theory – In AQFT.

Applications to the renormalization of Yang-Mills theory on curved background spacetimes is accomplished in

BPHZ Renormalization

BPHZ renormalization was introduced in particular in

  • Klaus Hepp, Théorie de la Renormalisation Lect. Notes in Phys. Springer (1969)

Review includes

The original articles on this are

An introduction and review to the Hopf-algebraic description of renormalization is in

A textbook treatment is

  • Dirk Kreimer, Knots and Fenyman diagrams , Cambridge Lecture Notes in Physics. 13. Cambridge: Cambridge University Press.
  • Joseph C. Varilly, The interface of noncommutative geometry and physics, hep-th/0206007

Some heavywheight automated computations using this formalism are discussed in

See also

  • W. van Suijlekom, Representing Feynman graphs on BV-algebras , (arXiv)

In BV formalism

Discussion of renormalization in the context of BV-BRST formalism is for relativistic field theory in the rigorous framework of causal perturbation theory/perturbative AQFT due to

and surveyed in

  • Kasia Rejzner, section 7 of Perturbative algebraic quantum field theory Springer 2016 (web)

Discussion for Euclidean field theory is in

building on

See also at factorization algebra of observables.

A vaguely related approach earlier appeared in

Operadic description

Relations to motives, polylogarithms, positivity

Last revised on April 4, 2018 at 08:50:30. See the history of this page for a list of all contributions to it.