arXiv:1608.00296v2 [gr-qc] 15 Aug 2016 fitgodffrnileutosi o e,adteeis there and new, not is symmetries classical equations to integro-differential the approach of in algebra least Lie The at approximation. motion, integro- theories the of of field analysis equations symmetry quantum perspective differential the to these this mapped of be From corre- could analysis the of symmetry motion. case of the of the structure equations in integro-differential non-locality spondent the the following. is the of GFT in definition refer possible often will work A previous we a which are in to mathemati- analysis which [11], present the theory, our- the of of the some limiting foundations laid cal of and had We aspects it, enough. classical on defining interesting the truly mainly the to focus as selves char- we out non-local stands so the that challenge, is theory it the of but of acter following, analysis types the whole both in tackle the manifold will difficulties We and base involved. GFT more computations considerably the exact of the nature makes and curved analysis generally symmetry the group-based Lie non- of well-known intrinsically adaptation non-trivial the the the requires of time, GFTs of goal same structure the the local At is This article. from present theories. adapta- techniques field the analysis for quantum symmetry suitable ordinary standard is the it such, of Quan- As tion Loop 6]. of [4, Gravity quantized, reformulation tum 2nd theoretical a terms field as in and quantum gravity [7–10], tensor dimensional triangulations and higher random matrix of and grav- of two generalization quantum for as of seen models be theory can fundamental It ity. a candidate for promising a formalism is [1–6] (GFT) calculations. Theory Field analytical tech- Group computa- and mathematical numerical many for powerful as niques well and as simplifications systems, tional of quantum macroscopic of terms of conceptual characterization phases powerful the in in offer example, interactions also for tools, such They of laws. conservation features and particles, them, key in between and interactions capture fields crucially allowed the fundamental enter dictate of they They definition formulation very theory the interactions. field quantum fundamental the of in especially physics, and modern in omni-present are principles Symmetry ∗ † [email protected] [email protected] oeso unu rvt n rvd eea nlsso t of analysis general a conservation provide f generalized then and the We gravity including transformations, perspective. quantum geometric of a models from motion, of equations edsustento fsmere nnnlclfil theori field non-local in symmetries of notion the discuss We Introduction otnospitsmere nGopFedTheories Field Group in symmetries point Continuous a lnkIsiuefrGaiainlPyis(letEin (Albert Physics Gravitational for Institute Planck Max mMhebr ,146PtdmGl,Gray EU Germany, Potsdam-Golm, 14476 1, Mühlenberg Am lxne Kegeles Alexander ∗ n ail Oriti Daniele and zdcnevto as h ae nwihmte fields matter which in case, The laws. general- conservation their the the ized with present along we for currents III groups Noether section can correspondent symmetry In reader investigation. of the under section table models this summarized of the end the how find the to clarify At theories we field case. of this, non-local treatment doing geometric Group In the in II. extend analysis to section symmetry the- in the Theory field with Field classical proceed of symme- we formulation of types geometric ories, various the of in definition Then, the tries, I. of section recap in with Theories a Field begin after Group We on follows. review of brief as a features organized is peculiar presentation the The to groups, theories. adapt extent, field symmetry large group to of the calculations modified to is, algebra suitably paper Lie this usual in the use we the method in The the case) anomalies. in possible theory or of identities field Ward study corresponding quantum is the local (with work of derivation difficulty the such course additional to any postponing of anticipate respect is for not do analysis reason we detailed that main a the the at While needed, identify we level. symmetries study the quantum the of work consequences further the discussed for of generalization be leave we the to Similarly, moti- symmetries leave elsewhere. Lie-Baecklund be we or will and contact following, symmetries to the point in to laws. vated restriction conservation generalized can the the same charges in Also the “conserved” included from of that, derived are definition show be fields possible also matter a will of models, We types special [11]. when the in of basis performed the in framework on symmetries laws, conservation derived generalized the of terms of consequences the only vestigate consider (wider) We the symmetries motion. not point of and equations action the the of The- symmetries of Field symmetries Group is in that the ory, models prominent investigate various we of no groups paper but [15] this in In there. calculated proposed was were field treatment group models systematic methods of GFT calculations known several analysis symmetry the for for the examples of for Particular which adapted theories. priori suitable result, a a be As know can yet. not to known do is case method from we universal vary no strongly and analysis case, meth- of the strategies theories, field and [12– local ods to subject contrast this in However, on available 14]. literature extensive quite a asfloigfo them. from following laws scaatrzdb integro-differential by characterized es cso ru il hoy(GFT) Theory Field Group on ocus ercniuu on symmetry point continuous heir † o eea neetn oes n in- and models, interesting several for , ti Institute), stein aitoa symmetry variational 2 are present, is treated in Section IV, with emphasis on kernel is the Lagrangian. In the geometrical interpreta- scalar fields that can be used as relational clocks, and the tion a Lagrangian is a function on a vector bundle which correspondent definition of relational charges. is usually a jet bundle over a principle G-bundle where G is the fundamental symmetry group of the theory. In usual field theories of fundamental interactions, the base I. Group Field Theory manifold of the vector bundle is interpreted as space time and the fiber is a vector space that carries a representa- tion of G. In the first part of this section we give an informal defi- In non-local theories we want to maintain this geomet- nition of the group field theory formalism and its general rical picture as far as possible, even if in our GFT con- features, emphasizing also the connections to other the- text the base manifold will not have the interpretation ories of gravity. In the second part we introduce the of spacetime (but is rather related to superspace, the specific models that we are going to study in the rest of space of smooth spatial geometries, or minisuperspace, this paper. The notation and conventions introduced in the space of homogeneous spatial geometries; see [26– this section will be used throughout the paper. 29]). We assume that a non-local function can be treated A Group Field Theory [2] is a quantum field theory on as a local function on a higher dimensional space. The a group manifold defined by a non-local action at the drawback of this picture is that different non-local therms classical level, and the corresponding path integral at are described by different geometrical bundles, and the the quantum level. The fundamental fields of the the- geometrical structure of the theory strongly depends on ory are functions from a Lie Group to some vector space the model in question, in contrast to the local case. (usually, the complex numbers)1. In models directly re- In a local theory the action is an integral of a Lagrangian lated to lattice gravity and loop quantum gravity, they over some domain Ω. In non-local theory it becomes a are often assumed to satisfy a specific gauge invariance sum of integrals whose domains are different base mani- condition, which indeed provides the perturbative Feyn- folds (in particular, of different dimension). In the spe- man amplitudes of the model with a lattice gauge theory cific types of non-local field theories we will be concerned structure. This possibility is, at its root, allowed by the with, they correspond to different numbers of copies of a choice of peculiar non-local pairing of field arguments in given base manifold. The action can therefore be gener- the GFT interaction terms. In fact, this has the immedi- ally written as ate result that the perturbative expansion of the quantum theory gives Feynman diagrams dual not just to graphs i but to cellular complexes, which can also be understood S = L voli, (1) ZMi as discretization of some smooth manifold. The Feynman Xi amplitudes, whose explicit expression is of course model- where i ranges over the number of different base man- i dependent, can be given the form of lattice gravity path ifolds Mi with the correspondent Lagrangians L . The integrals [16–18] or, equivalently, spin foam models [19– measure of integration on each manifold defines the vol- 23]. The latter are a covariant definition of the quantum ume density, and in the GFT case, will be given by the dynamics of spin networks, the quantum states of Loop Haar measure or some other invariant measure, depend- Quantum Gravity (LQG) [24, 25]. In fact, GFTs can be ing on the Lie group chosen as base manifold. understood as a reformulation of the kinematics and dy- Lie group structure of the base manifold. Indeed, in namics of LQG degrees of freedom in a 2nd quantized Group Field Theory, we require the local base manifold, framework [4]. The GFT Hilbert space re-organizes the which matches by definition the domain of each individ- same type of spin networks in a Fock space, whose funda- ual GFT field, to be given by some number of copies of mental quanta are spin network vertices and are created a Lie group G. In GFT models of quantum gravity, this (annihilated) from (into) a Fock vacuum (a state with no is usually chosen to be the local gauge group of gravity, geometric nor topological structure) by the action of the i.e. the Lorentz group or it double cover SL (2, C) (or its GFT field operators, absent any embedding information Riemannian counterpart Spin(4) for models of gravity into any ambient smooth continuum manifold. in Euclidean signature). For models directly related to In the following we will discuss each point in more detail, Loop Quantum Gravity, the rotation subgroup SU (2) of in order to convey the general idea and motivation behind the Lorentz group becomes the relevant base manifold, each of the above features. via appropriate conditions imposed on the GFT fields Non-local action. In local theories the action is an inte- (and quantum states) at the dynamical level. gral functional on the (appropriate) space of fields, whose The number of copies of the group G defining the lo- cal base manifold is usually the topological dimension of the cellular complexes dual to the Feynman diagrams of the model, chosen to match the dimension of the contin- 1 In fact, the GFT formalism includes also the case in which the uum spacetime one aims to reconstruct from the quan- fields are defined on a finite group, reducing to the tensor models tum dynamics of the model. Clearly, to have a physical formalism. However, for obvious reasons our symmetry analysis connection to General Relativity we need to develop and would not apply to this case, so we stick to the Lie group setting. understand models in four dimensions. However, since 3 these tend to be naturally more involved than their lower Non-local structure of the Lagrangian - combinatorics dimensional counterparts, several two and three dimen- As said, the main feature distinguishing GFTs from or- sional models have been studied, which allow to inves- dinary local field theories is the combinatorial pattern tigate important mathematical and conceptual problems of relations between field arguments in the GFT interac- of the theory, in the setting with reduced complexity. tions. In principle many different non-local interaction Gauge invariance of fields In some GFT models the terms can be included in the action. A preference of one fields are required to satisfy a so-called gauge invariance model over another can be given conclusively only by ex- condition. Explicitly it means that for any h in the diag- tracting its physical consequences. We will discuss the ×n onal subgroup GD = (g, ,g) G g G the fields possible combinatorics and their consequences rather ex- satisfy { ··· ∈ | ∈ } tensively in the following. Here we only stress that a de- tailed analysis of symmetries of the corresponding models φ Rh = φ, (2) ◦ is going to be useful also for choosing one combinatorial ×n ×n where Rh : G G denotes the right multiplication structure over another. → by h GD. Due to this gauge invariance condition, the Number of different fundamental fields - colored theo- base manifold∈ of the GFT fields is effectively reduced to ries One can also consider GFT models involving more a quotient of n copies of the group under the diagonal than a single fundamental field. If a Group Field Theory group action model involves more than one field, we call the model colored and distinguish different fields by an additional G G/GD. ×···× index, calling it the color index. Colored GFTs were in- This condition is imposed for many different physical con- troduced for the first time in [32] for models aimed at siderations. In particular, in gauge invariant GFT models describing simplicial quantum gravity, and topological the perturbative Feynman amplitudes of the theory take BF theories discretized on simplicial complexes. Indeed, the form of lattice gauge theories (on the cellular complex this step immediately led to a large number of interest- dual to each Feynman diagram) and the quantum states ing mathematical results and powerful new techniques, in become those of a lattice gauge theory with gauge group the GFT context as well as for the simpler tensor models G (in particular, for G = SU(2), a complete orthonormal [8]. In particular, Feynman diagrams generated by non- basis is given by spin networks, and the same amplitudes colored simplicial GFT models can be dual to very sin- can be equivalently written as spin foam models). gular simplicial complexes, while Feynman diagrams of colored models are much more regular, and their topol- ogy can be reconstructed to a much greater extent [33]. Other defining features of different GFT models Quantum statistics An additional assumption on the theory is its quantum statistics, i.e. whether the GFT Beside the choice of base manifold, there are various other quanta of a given model are bosonic, fermionic or of ingredients that have to be specified, in order to fully other nonstandard statistics. In local, spacetime-based define a GFT model, and this even before one chooses quantum field theories the quantum statistics is highly a functional form for the interaction kernels. The main constrained by powerful spin-statistics theorems, linking differences between various models include: the quantum statistics to the spin of the quanta, under Presence (absence) of derivatives in the local La- the assumption of Lorentz or Poincare invariance of the grangian - dynamical (static) models The GFT mod- theory. No similar spin-statistics theorem is available, els proposed at first, for a study of topological field the- yet, in the GFT framework. First and foremost, this is ories of BF type and, later, for 4d quantum gravity de- due to the fact that the base manifold in group field the- scribed as a constrained BF theory and in absence of mat- ories is not directly associated with space time, thus we ter fields, only possessed a “mass” term, and no deriva- have no obvious symmetry requirement to impose, like tives of the fields in their quadratic, local part of the Lorentz or Poincare invariance. Second, as we have al- Lagrangian. In local QFT, this would imply a trivial ready stressed as a motivation for our work, very little is dynamics, and therefore one could label these models known about GFT symmetries, from a full classification "static". However, in contrast to local field theories, the of them in specific models to their general consequences, transition functions, and more generally both the classi- on the statistics of the same models and on their physics. cal and quantum dynamics in "static" group field theo- This paper is meant to partially fill this gap. ries are still highly non-trivial due to their non-local na- ture. More recently, GFT models which include deriva- tive terms in the quadratic part of the action have been Notation studied extensively. They can be motivated in various ways, the first being that renormalizability seems to re- quire them, at least for some models in which a non- Throughout the paper we will denote an element of the n trivial dynamical term will be generated by the RG-flow Lie group G× as [30, 31], and therefore needs to be included in the theory space. ~g = (g ,g ,g , ,gn) . (3) 1 2 3 ··· 4

Differential operators with an index will refer to opera- a simple example: a GFT model for BF theory in three tors that act on the correspondent copy of the group. For dimensions and the GFT fields subjected to the gauge example, invariance condition. If the group is chosen to be SU(2), this is a model for 3d gravity in euclidean signature. In 1 (~g123) = ( g1,g2, ,gn) . (4) this case the simplicial interaction is given by ∇ ∇ ··· Differential operators with more than one index refer to I c S [φ ]= λ φ1,2,3φ3,4,5φ5,2,6φ6,4,1 + c.c. , (10) a sum of individual operators as ZΩ ×6 123 = 1 + 2 + 3. (5) where the integral domain is Ω= G . If we associate to ∇ ∇ ∇ ∇ each field in the interaction a triangle with edges labelled If used without further clarification, an integral symbol by the three arguments of the same field, the interaction (without the explicit measure) denotes an integral over all will be associated to a tetrahedron (a 3-simplex) formed variables that appear under the symbol. Integration over by the 4 triangles glued pairwise along common edges, as each single group element is performed with the Haar shown in figure (1). φ measure on G

φ (g1,g2,g3)= dg1dg2dg3 φ (g1,g2,g3) . (6) φ Z Z φ φ This notation will be used a bit differently in section φ III, where we will point out the differences explicitly. The fields are complex scalar fields. The upper script of the field denotes the color and the subscript denotes (a) 3 dim field (b) Simplicial associated with interaction in three the field’s dependence on the variables a triangle dimensions associated with a tetrahedron c c φ := φ (g ,g , ,gn) . (7) 1,2,··· ,n 1 2 ··· Figure 1: Topological interpretation of the field and the simplicial interaction in three dimensions A. Overview of the models discussed in the paper The Feynman diagrams will similarly be in correspon- We are going to present the major distinctions of combi- dence with the simplicial complexes obtained by gluing natorial structures of models discussed in the following. the different tetrahedra associated to the interaction ver- The general structure of the GFT actions has the form tices in the Feynman diagram, along shared triangles, the of (1) as gluing being identified by a propagator line. Depending on the details of the Feynman diagram, the resulting sim- loc nloc S [φ]= S [φ]+ S [φ] . (8) plicial complex may or may not be a simplicial manifold and can be quite singular. Moreover, the data present We assume that the local, quadratic part of the action is in the GFT diagram are not, in general, sufficient for re- defined as constructing the topology of the same simplicial complex in its entirety. Such technical difficulties are cured by Sloc [φ]= κ φ¯ φ + m φφ,¯ (9) introducing colored fields [33]. Z ×n ∇ · ∇ G The extension to higher dimensions, via extension of the where is the gradient on the group G and the de- base manifold of the fields, and appropriate pairing of notes the∇ contraction of the two vectors. We will· also their arguments in the interactions, follows the same cri- treat cases in which κ is zero, meaning that the model teria and it is straightforward. is a static one. In the following, we distinguish between three different types of models - simplicial, tensorial and geometrical. The corresponding interaction parts SI are Remarks on the combinatorial structure of simplicial models presented below and a concise summery of the interaction terms used is given in the table (I). The original Boulatov model [34] for 3d gravity has a slightly different combinatorial structure from the one we introduced above. It is given by 1. Simplicial I c S [φ ]= λ φ1,2,3φ1,4,5φ2,5,6φ3,6,4 + c.c., (11) Z The interaction part of simplicial models is constructed Ω such that the Feynman diagrams have a particular topo- with an additional invariance of fields under cyclic per- logical interpretation, i.e. they are simplicial complexes. mutations of the variables. The quantum geometric con- Let us illustrate in one example how this comes about, in tent of the model is not affected, as it can be seen both in 5 the group representation, and in the spin representation. contracted to give unitary invariants [8, 35]. The gener- In fact, in the Peter-Weyl decomposition both combina- alization of the same invariance characterizes the inter- torial structures lead to a 6J symbol, which encodes both actions of tensorial GFTs [10, 36]. Hereby a monomial in the gauge invariance properties and the piece-wise flat- fields belongs to the theory space if it is invariant under ness of the simplicial complex generated in the pertur- a unitary transformation defined as follows bative expansion. However, while the original Boulatov 1 2 3 model produces a usual 6J symbol, the interaction pat- Uφ (~g)= d~hU (g1,h1) U (g2,h2) U (g3,h3) φ ~h , tern we introduced above produces an additional factor Z   ( 1) that alternates with the representations involved. with the requirement on the kernels U i to satisfy In− the case of colored models this difference can be ab- ˜i i i sorbed in the redefinition of the fields as φ = φ P , dhU (g,h) U † (h, q)= δ (g, q) , (14) i ◦ where P being some permutation of the group elements. ZΩ In this way the exact order of the variables in the field where U † (h,g) := U¯ (g,h). This conditions requires that becomes unimportant. We will choose the following com- two fields which share a group element need to be com- binatorics, since it leads, as we will show, to the largest plex conjugate of each other. It is easy to verify that symmetry group: this excludes the simplicial combinatorics. It is also im- portant to mention that the kernels U i do not need to I c 1 2 3 4 S [φ ]= λ φ1,2,3φ1,4,5φ6,2,5φ6,4,3 + c.c.. (12) be smooth, differentiable or even continuous and for this Z Ω reason they may include delta distributions.We will come It is important to mention that if the models are indeed back to this point in the next section, when we discuss equivalent, their symmetry group should not differ as the symmetries of the tensorial models. well. This implies that a particular choice of the com- Note that a dynamical term will in general break the binatorics may simply help to discover symmetries that unitary invariance. Therefore, when we refer to tensorial would still be there for different combinatorics, but would dynamical models in the following, we imply that the be more difficult to identify. In the next section we will unitary invariance characterizes only the interaction part show how these minor combinatorial differences affect the and not of the whole action. Note also that one can have symmetry group. a very similar type of invariance for real GFT fields, with Notice that, while in the action above we have chosen the unitary group replaced by an orthogonal group. The four GFT fields to appear in one term, with their com- construction proceeds in analogous way. plex conjugates appearing in the other, our focus here was only the combinatorial structure, and one can de- vise simplicial interactions involving both the field and 3. Extended Barrett-Crane model its complex conjugate in the same monomial. For exam- ple, we can start with the action from equation (10) and In four and higher dimensions, gravity can be formulated color the fields in the way φ1 = φ3 = φ and φ2 = φ4 = φ¯ as a BF theory plus appropriate constraints [37], which such that the interaction part coincides with its complex are labeled simplicity constraints. This goes under the conjugate name of Plebanksi formulation of gravity. This formula- tion provides also the conceptual and technical starting I c point for the construction of spin foam and group field S [φ ]= λ φ1,2,3φ¯3,4,5φ5,2,6φ¯6,4,1. (13) ZΩ theory models for 4d quantum gravity. One may call the corresponding GFT models geometric, even though In this case we refer to the above action as colored, with one has a direct control only on the discrete (simplicial) two colors, even though the model involves only the field geometric interpretation of states and amplitudes, while φ and its complex conjugate. This convention will be- the reconstruction of continuum geometry requires more come handy in the classification of the symmetries in the work. As an example of these constructions, we deal with following. the so-called Barrett-Crane model [38], whose detailed treatment in the language of extended Group Field The- ory was presented in [18], for the euclidean signature. 2. Tensorial models Here we show just the main features of the model and refer to the cited literature for more details. Tensor models are characterized by an U (N) invariance. The starting point is the GFT model for 4d BF theory

Given a rank-n complex tensor Ti1...in with index set of based on simplicial interactions, in which the fundamen- dimension N, it transforms naturally under the group tal GFT field is associated to a tetrahedron in 4d, and U(N)×n, where U(N) is a unitary N N matrix, acting the interactions involve five GFT fields, paired to rep- on each of its indices. This is also the× natural symme- resent the gluing of five tetrahedra to form a 4-simplex. try of tensor interactions, so that the full theory space The base group manifold of the model is Spin (4). is defined to be spanned by all possible monomials in Simplicity constraints are characterized by a vector in the tensor and its complex conjugate, with their indices S3 SU (2), interpreted as a unit normal vector (in R4) ≃ 6

Gauge variant Gauge invariant

Dimension 3D and 4D

κ ∇φ ·∇φ + mφφ Kinetic part κ 6= 0 κ 6= 0 κ = 0

Group SU (2) SU (2) SU (2) G

Combinatorics Tensorial Simplicial Simplicial

Colors - (un)colored (un)colored

Simplicity constrains - - - Barrett-Crane

Table I: Overview of the models discussed in this paper of the tetrahedron represented by the field φ. In order to keep track of this additional normal vector the local base manifold is extended, so that the field becomes a function : (~g; k) 1; h−1 (~g; k) k~uk−1, ~u, ; 1 (h ; h ) , S 7→ − · · · − + on four copies of Spin (4) and one copy of SU (2)


where h Spin (4). One can indeed verify that the fields 1 2 3 4 ∈ φ g ,g ,g ,g , k =: φ1,2,3,4,k, (15) invariant under the above transformation satisfy
where gi Spin (4) and k SU (2). φ = φ. (21) The interaction∈ of the model∈ becomes an extended ver- ◦ S sion of the Ooguri interaction given as Notice that, since the simplicity and gauge invariance conditions are imposed on the fields via a projector, the imposition of these conditions on all fields appearing in I S [φ]= φ1,2,3,4,k1 φ4,5,6,7,k2 φ7,3,8,9,k3 φ9,6,2,10,k4 φ10,8,5,1,k5 the action is the most natural choice, but any other Z choice, e.g. imposing them only on the fields appearing + c.c.. (16) in the interactions, would result in the same Feynman The gauge invariance is again written in the usual form amplitudes (but not the same theory, as for example the as classical equations of motion would be different). This is not true for other 4d gravity models, where the simplicity

φ Rh = φ, (17) constraints take a different form [17]. ◦ with h Spin (4)4D. Additionally, the simplicity con- straints∈ are imposed by requiring invariance of the fields II. Symmetries

φ S = φ, (18) ◦ In this section we will present the different notions of under the transformation symmetry transformations in non-local field theories in general (recalling the geometric construction in the local S : (~g, k) (~g; k) k~uk−1, ~u ; 1 , (19) case, first), and then apply them to Group Field Theory 7→ ·

in particular, and derive the symmetry groups for the where uj SU (2). If we write a Spin (4) element in models introduced above, showing the main steps of the its selfdual∈ and anti-selfdual SU (2) components as g = calculations for three-dimensional models. (g−,g+), the above transformation takes the form

j j −1 j j A. Transformations of local field theory S (~g; k)= g− ku k ,g+u ; k . (20)  

In [18] it has been shown that S and Rh commute as The geometrical construction of local field theory is very projectors acting on the space of fields, which allows to well known, but we will briefly review its main points here combine them into a single transformation, which is itself because they will be essential in the following discussion a projector, acting on the GFT fields as of the non-local case. 7

In the geometrical picture, the Lagrangian is a differen- B. Transformations of a non-local field theory tiable (in a sense that needs to be further specified) func- tion on an nth order jet bundle. In order to bring the We now apply this construction to non-local field theo- main idea across without complicating it with technical ries. As we have pointed out earlier, the action is given details, we will assume that the Lagrangian is a function by the sum of integrals over Lagrangians just on a vector bundle. The full construction can be found in usual text books on this subject some of which i are [14, 39]. S = L . (28) ZMi We call the relative vector bundle E, the base manifold of E being M and the fiber being V. Locally, we can Hereby, each of the Lagrangians is a function from a vec- i think about E as a cross product of M V, which we tor bundle Ei to R. For i = j the vector bundles E assume for the rest of this discussion. The× points on E and Ej are assumed to be different.6 If they are not, we are then given by x M and u V, we write (x, u) E. can combine the Lagrangians Li and Lj into a single La- Hence, the values of∈ the Lagrangian∈ can be denoted∈ as grangian Lij = Li + Lj. L (x, u) R. Following the general construction from the previous sec- We then∈ introduce the physical fields φ in the construc- tion we define a transformation of the theory as transfor- tion. This is done by choosing points of the vector bundle mation of the corresponding vector bundles. Neverthe- which are given by a smooth section of E. In other words less, in the non-local case we need to transform different we assume that u = φ (x). bundles, which is why we say that a group action is given Assuming that the set of transformations of the theory by functions Ci,Qi such that, for each i, Ci and Qi are i forms a group GT , we can write the action of the group transformations of E in the above (local) sense. It is im- on E as portant to realize that these transformations can not be independent from one another, since they represent the g (x, u)=(˜x, u˜) = (C (x, u) ,Q (x, u)) . (22) same transformation g GT . Instead, their mutual re- · lations should be given∈ by the relation between different The action is thus specified by two functions C and Q. vector bundles Ei. Note, that in general both functions depend on x and Assuming that E0 denotes the vector bundle, whose sec- u = φ (x) and are not invertible. However, locally around tions are identified with physical fields φ0 = φ we can i each point of the bundle these transformations are diffeo- quite generally write each E as a pull back of ni copies 0 i ×ni morphisms, due the the fact that they represent an action of E by some embedding fi : M M . These of a Lie group. functions f i encode the combinatorial→ structure of non- We ask for the transformed sections φ˜ that corresponds local Lagrangians and provide a relation between differ- i to a new point of the bundle, that is x,˜ φ˜ (˜x) = (˜x, u˜). ent E ’s. Therefore they also give the relations between ×n   the sections φi as φi = φ0 i f. Knowing how the field The transformed fields φ˜ can then be seen as transformed ◦ φ0 transforms under gautomatically
implies the trans- sections under the group action of g. It is a well known i result that the transformed fields are given by formations of φ as

−1 −1 ×ni φ˜ (˜x)= Q C−1 (˜x) , φ C−1 (˜x) , (23) Qi φi Ci = Q0 φ0 C0 f. (29) ◦ ◦ ◦ ◦ ◦ ◦

h
i or, in short, This relation implicitly defines Ci as g −1 ×n φ Q φ C , (24) f Ci = C0 i f, (30) 7→ ◦ ◦ ◦ ◦
at least as long as C is invertible. Hereby, the transfor- and Qi as mation Q is defined along the fiber and the C−1 accounts ×n for the transformation of the base manifold. Qi φi = Q0 φ0 i . (31) We summarize the main properties of the maps Q and C ◦ ◦   before finishing this part. For a fixed φ the base manifold The above equations provide the missing link between transformation C is a local automorphism the group actions on different vector bundles. However, equation (30) does not always define a local automor- Cφ :M M (25) → phism Ci on M i. If Ci were an automorphism, equation x C (x, φ (x)) . i i 7→ (30) would imply that, for any x M , there exists an x˜i M i such that ∈ And for a given point x M, the fiber transformation Q ∈ ∈ ×n is a local automorphism f x˜i = C0 i f xi . (32)


Qx : V V (26) → This however, is not always possible as we will see in the φ (x) Q (x, φ (x)) . (27) 7→ following section. 8

If equation (32) is not satisfied, the group action can not These transformations are of the type (22) and are called be chosen consistently as a transformation of the vector Lie point symmetries or “geometrical” symmetries, be- bundle. In this case, we can define the action of the group cause they admit a geometrical interpretation of a flow, directly on the space of fields, using equation (29) as being generated by vector fields on the vector bundle of the theory. −1 g (x, u) = (x, Q (x,u,ux, )) = x, φ C (x) . The requirement that the symmetry does not change the · ··· ◦
action for any sub domain Ω′ is essential, in order to be Here, ux denotes the coordinates of the Jet space and able to make point-wise statements, i.e. to derive truly refers to derivatives of fields at the point x, i.e. ux = local statements from the existence of the symmetry it- Dφ x. That the transformation Q needs to depend on | self. In the physical literature, this statement is often re- the derivatives of fields φ is easily seen from the Taylor ferred to as the Noether theorem, which allows the deriva- expansion, since tion of point-wise equations of the form div (J)= EL δφ, · −1 2 i.e. the conservation laws of the corresponding field the- φ C (x)= φ (x) Dφ (XM )+ X (33) ◦ − O M ory. 2
= u ux XM + X , (34) − · O M
where XM is the infinitesimal generator of the transfor- 2. Generalized symmetries mation C. Such transformations Q generalize the notion of point transformations from equation (22) to the so A generalization of the symmetry concept (already in- called Lie-Baecklund transformations, which are trans- troduced by Noether in her original paper [41]) leads formations from the Jet bundle to the vector bundle. to the so called generalized symmetries, which are Lie- In the next section, we will briefly explain how the Lie- Baecklund transformations that can change the action by Baecklund transformations represent a more general no- an arbitrary divergence term. That is, for all Ω′ Ω tion of symmetry. However, for reasons that will also ⊂ become apparent in the next section, we will restrict our ′ S ′ [φ] S ′ φ˜ = S ′ [φ]+ div (Γ) Ω Ω . analysis to Lie point symmetries. Ω C(Ω ) Ω → h i ZΩ′ ∀ ⊂ (36) A restriction of such transformations to those that can C. Notions of symmetry transformations depend at most on the first order derivatives of fields defines the so called contact symmetries. There are many different notions of a continuous sym- In general, a set of Lie-Baecklund symmetries is infinitely metry in local field theories. Almost all of them are for- large, but it is often the case that also infinitely many mulated as diffeomorphisms of the vector bundle of the such transformations are equivalent, leading to a finite theory. In order to distinguish between different notions number of inequivalent transformations. Computational of symmetries we first point out that an action (in local algorithms for finding Lie-Baecklund symmetries to a theories) is defined as an integral, and therefore intrin- fixed order of derivative dependencies are known and are sically depends on the domain of integration over which implemented in a large variety of computer algebra pro- grams [42]. Nevertheless, already for flat base manifolds the Lagrangian is integrated, i.e. SΩ [φ]= Ω L. In this sense we can talk about a family of actionsR SΩ′ for all (and of course, local theories) the explicit calculations Ω′ Ω. In the discussion of symmetries, the{ dependence} are quite challenging. of the⊆ action on the domain plays a very important role, The reason for looking for a generalized notion of sym- which we are going to highlight in the following. metry is the observation that two actions are physically equivalent if and only if they differ by a divergence term [39], because the correspondent equations of motion are 1. Point symmetries the same, and this is all that matters in the classical regime. This implies that the physically relevant object is not the action but rather an equivalence class of ac- The simplest notion of a symmetry of an action is a dif- 2 feomorphism on the vector bundle of the theory [39, 40]. tions . As we discussed above choosing a section of the bundle As in the previous cases, this class of symmetries gives (physical field φ) it is possible to locally project the dif- rise to local, point-wise equations. Even more, only in feomorphism on the fiber and the base manifold obtain- this case, the correspondence between symmetries and ing the transformation function Q and C, which define a transformation of the fields and of the base manifold respectively. A symmetry is then a transformation which 2 Is important to distinguish between the symmetries of the ac- does not change the action functional SΩ′ for any subdo- tion and symmetries of the correspondent equations of motion main Ω′ Ω ⊂ (which correspond to extrema of the same action): generalized variational symmetries form a subgroup of Lie-Baecklund trans- ˜ ′ SΩ′ [φ] SC(Ω′) φ = SΩ′ [φ] Ω Ω . (35) formations of the equations of motion. → h i ∀ ⊂ 9 divergence-free quantities like div (J +Γ) = EL δφ is One approach to overcome this difficulty would be to dis- one to one, which is the actual statement of the original· cuss directly the symmetries of the corresponding equa- Noether theorem. tions of motion, which are integro-differential equation (see [14] for the standard approach of Lie algebra meth- ods in integro-differential equations). However, this is 3. Integral symmetries highly non-trivial, in the GFT case. Gauge invariance condition, the structure of the curved base manifold, as An entirely different notion of symmetry arise if we drop well as its large dimension make the usual Lie algebra the requirement that a symmetry transformation should approach, even more involved. leave the family of actions S ′ invariant, and instead Also, there is not much more to say about the integral { Ω } require the invariance of SΩ only for a single, fixed inte- symmetry transformations defining tensorial group field gral domain Ω, theories, beside what we remarked already, i.e that they provide a natural characterization of the corresponding ˜ theory space. For these reasons, we limit our analysis SΩ [φ] SC(Ω) φ = SΩ [φ] . (37) → h i to the Lie point symmetry analysis and postpone the analysis of Lie-Baecklund symmetries to future work. This kind of transformations does not lead to point-wise In the following we will use the definition of a symmetry statements, like conservation laws. Clearly every symme- for a non-local action as follows: try of the previous type is also a symmetry of this type but not the other way around. Definition. A symmetry of a non-local action is a trans- It is interesting to observe that tensor models and ten- formation that is a Lie-point symmetries of each func- sorial GFTs invoke exactly this type of symmetries in tional in the action-sum. order to define the theory space, since we require that the corresponding unitary transformations satisfy D. Symmetry analysis of gauge variant models dbU (g,h) U † (h,s)= δ gs−1 , (38) ZG
We start by performing the standard Lie group analysis only after the integration over the whole group G, and of point symmetries [39] in the case of gauge-variant GFT there is no reason to assume that changing the integral models. domain to a subspace of G would preserve the above In [11] we have shown, that a symmetry condition of the equality. action can be equivalently formulated on the level of its Lagrangians, leading to a generalized version of Noether theorem, with respect to the local case. More precisely, 4. Symmetries in non-local field theories the symmetry relation can be formulated in the following way So far the definition of a symmetry was introduced for an Theorem 1. action which is given by an integral over a Lagrangian. G is a symmetry group of the action iff the In the non-local case, as we have explained, the action generators of the symmetry (XV ,XM ) satisfy the relation is given by a sum of such actions each defined on a dif- 0=DJ L DXQ + DV L XQ + Div (LXM ) , (39) ferent base manifold. Therefore, we need to extend the · · above notions of symmetries to transformations, which where X = X X (φ) and it is assumed that every are symmetries (in one of the above senses) of each and Q V M term is evaluated− at some point z of the correspondent all individual functions in the action-sum. base manifold. This is the only generalization we need, to start analyz- ing symmetries of the non-local GFT models introduced The notation that is used in the above equation needs to above. be further explained: However, it is important to stress here, that this is not enough to study generalized symmetries of the Lie- XM is a vector field on the base manifold which Baecklung type. The same motivation that lead to con- • coefficients depend on a point of the base manifold sidering them in the local case would apply as well for and a point on the fiber. In local coordinates (U, x) non-local models. However, contrary to the local case, of the base manifold the vector field can be written i for non-local field theories the equivalence class of ac- as XM = XM (x, φ (x)) ∂i. tions that yields the same equations of motion is not under control (to our knowledge). The only thing that XV is a vector field along the fiber of the bundle we can say is that it does not coincide with the one de- • that in local coordinates (U, x) (V,u) can be de- i i × i i fined in the local case, because as we pointed out in [11], noted as XV = XV (x, φ (x)) ∂ui where u = φ (x). a divergence term will, in general, change the equations We will sometimes use the simpler notation ∂φ(x) of motion. or even ∂φ, always referring to ∂u. 10

The assumption of dealing with a geometrical sym- equation by different powers in derivatives of fields. Since • i i metry translates into the restriction of the coef- the coefficients XM and XV do not depend on derivatives ficients of vector fields depending only on x and of the fields, it is possible to extract all powers explicitly, φ (x), but not on ∂φ and higher order derivatives. iii) different powers of derivatives of φ are linearly inde- pendent since the condition (39) has to be satisfied for XM (φ) is the Lie derivative of φ along XM . • all fields. For this reason the coefficients in front of each term have to vanish separately. This results in simple dif- XQ is the characteristic vector field, which corre- • sponds to the effective transformation of the fields ferential equations for the coefficients of the vector field from equation (24), given by which can then be easily solved. Since the GFT models of interest, here, are defined on −1 many copies of SU (2) the notation can quickly become XQ = ∂ǫ 0 Qǫ φ Cǫ = XV XM (φ) (40) | ◦ ◦ − unreadable. For this reason we summarize the notation It is also important to spend few words on the differ- used in the rest of this section in the table (II). ent types of derivatives that are used in this geometrical construction. Field value φc (~g) uc The derivative DV denotes a derivative of the La- c - color of the field, • grangian along the coordinates of the fiber. In the common notation we can write ∂φ or δφ Derivative of the field φc at the point ~g

DJ denotes the derivative of the Lagrangian with i - chart component of the single copy of SU (2) • c respect to the jet coordinates. In the above nota- uiA

tion we can write ∂∂iφ or δ∂iφ. A - number of the copy of SU (2)

The derivative D refers to the total derivative with iA direction of the derivative ∂iAφ|~g • respect to the base manifold. This means that the implicit dependence on the base manifold through Vector field that acts on the base manifold M fields needs to be taken into account. iA XM i - chart component of the single copy of SU (2) The partial derivative ∂i is instead a derivative • purely on the explicit dependence of the coordi- A - number of the copy of SU (2) nates. Using above notation we can write Component of XV in ∂uc direction X c Df (x, φ (x)) = ∂xf + DV f ∂xφ. (41) u · c - color of the transformed field We also use the capital letter in Div (L XM ) for · the total derivatives used in the divergence and Table II: Usage of indices in this section div (XM ) to denote the divergence taken only with respect to the explicit coordinates. We denote the vector fields by XM and XV and refer iA Equation (39) holds for local as well as non-local La- to their coefficients in a specific chart by XM and Xuc grangians. By partial integration equation (39) becomes respectively, i.e.

iA EL [XQ]+ Div (DJ L XQ + L XM )=0. (42) XM = X ∂iA XV = Xuc ∂uc + Xu¯c ∂u¯c . (43) · · M

Where EL is the Euler operator acting on the Lagrangian For the rest of this section we assume the summation L3. convention over repeated indices. Having clarified the terminology and the notation, we can The local part of the action is given by use (39) to derive the most general geometric symmetries of the various GFT models. c c c c c L (u ,uJ )= κu¯iAuiA + m u¯ u,¯ (44) We will use a rather standard procedure, based on the XiA following steps: i) We assume a most general vector field on the vector where the sum over A ranges in 1, ,n (the SU (2) bundle and insert it in (39), ii) we rearrange the resulting copies of the local base manifold){ ··· }

3 For local theories EL coincides with the equations of motion and further work needs to be done to provide a connection between the above equation becomes the usual Noether identity. For non- the equations of motion and the divergence terms. The resulting local theories, however, EL does not coincide with the equations relation is shown in the next section of this paper and is carefully of motion, due to their integro-differential structure. In this case derived in [11]. 11

The above symmetry condition equation (39) implies

c iA c XM (L)+ L Div (XM )+2κ ∂iAφ D (XcQ)+2mφ (XcQ)=0. (45) · c Explicitly sorting the terms by powers of uiA we get

c 2 c c 0= m u div (XM )+ mu¯ Xuc + mu Xu¯c (46) h | | i t c 2 nA nA nm + Re u m u D t X + D t X + κg (∂mAX t + ∂mAX t ) (47) nA | | u¯ M u M A u u¯   h
i t c 2 nA nA nm +ı Im u m u D t X D t X + κg (∂mAX t ∂mAX t ) (48) nA | | u M − u¯ M A u¯ − u   h
i c c 1 iB nm ni mA nm +κ Re[¯u u ] X ∂iB g 2g ∂iAX + g (Duc Xuc + Duc Xuc )+ div (XM ) (49) nA mA 2 M A − A M A { ¯ ¯ } c c ni mB6=A 2κ Re u¯ u g ∂iAX (50) − nA mB6=A A M   h i nm c t=6 c +κgA Re u¯nAumA [Dut6=c XcV + Du¯c Xu¯t6=cV ] (51) h i nm c t=6 c +ıκgA Im u¯nAumA [Dut6=c Xuc Du¯c Xu¯t6=c ] (52) h i − nm c t +κgA Re u¯nAu¯mA [Du¯t Xc + Dut Xu¯c ] (53) nm  c t  † +ıκgA Im u¯nAu¯mA Du¯t Xuc Dut Xu¯c (54) h − i c c  t  nm iA iA ni mA mA +2κu¯ u Re u 2g D t X + D t X + g D t X + D t X (55) nA iA mA − A u M u¯ M A u M u¯ M c c  t  nm iA iA
ni mA mA
 +ı2κu¯ u Im u¯ 2g D t X D t X + g D t X D t X (56) nA iA mA − A u M − u¯ M A u M − u¯ M c c  t  nm iB6=A
iB6=A
 +¯u u Re u 2κg D t X + D t X (57) nA iB6=A mA − A u M u¯ M   h  i c c t nm iB6=A † iB6=A +ıu¯ u Im u¯ 2κg D t X D t X . (58) nA iB6=A mA − A u M − u¯ M   h  i

c c A A This equation has to hold true for arbitrary fields u and the field values u i.e. XM = XM (~g). From this c uiA. However, the parts in brackets do not depend on condition, equations (55) and (56) are automati- c uiA, which implies that each line has to vanish individu- cally satisfied. ally4. The consequences of these equations read: 4. Due to the above, equations (47) and (48) reduce 1. Equations (51) and (52) imply that the vector field to component Xuc depend only on the field colors they transform that is (no summation) ∂mAXut =0= ∂mAXu¯c . (59)

c c c c That is, the vector fields do not explicitly depend X c = X c (~g, u , u¯ ) X c = X c (~g, u , u¯ ) . u u u¯ u¯ on the points in the base manifold c c 2. Equations (53) and (54) additionally imply that the XcV = Xuc (u ) Xu¯c = Xu¯c (¯u ) . vector fields Xuc do not depend on the complex conjugate of the field, that is 5. Equation (46), together with the above conclusion, restricts the vector fields to a specific form c c X c = X c (~g, u ) X c = X c (~g, u¯ ) . u u u¯ u¯ c c Xuc = Cu Xu¯c = C¯ u¯ , (60) 3. Equations (57) and (58) tell us that the vector fields where C is an arbitrary constant that satisfies that transform the base manifold do not depend on div (XM )= C C.¯ (61) − − 6. The above condition reduces equations (49) and 4 Notice that, if we allowed for derivative dependence of the co- (50) to efficients χ = χ (x,φ (x) ,Dφ|x) and similar for the ξ, we could not argue that the terms with different powers of derivatives of φ iB nm ni mA mi nA XM ∂iBgA 2 gA ∂iAXM 2 gA ∂iAXM =0 have to vanish independently, since the terms in brackets would − − ni mB6=A ni mA6=B also contain derivatives of the fields. gA ∂iAXM + gB ∂iAXM =0. 12

These two equations are the only ones that are not trivial v1 v2 v3 v4 v5 v6 to solve. However, although lengthy, their solution can be found in a straightforward way. The solution in Hopf v1 0 v5 v6 0 −v2 −v3 coordinates (η,ξ,χ)5 reads as v2 −v5 0 0 v6 v1 −v4

ηA v3 −v6 0 0 v5 −v4 v1 XM = C1 sin ξA sin χA (62)

+C2 cos ξA sin χA v4 0 −v6 −v5 0 v3 v2 +C3 sin ξA cos χA v5 v2 −v1 v4 −v3 0 0 +C4 cos ξA cos χA cos ηA ξA ηA v6 v3 v4 −v1 −v2 0 0 XM = ∂ξAXM + C5 (63) sin ηA χA sin ηA ηA Table III: Lie algebra of symmetry vector fields XM = ∂χAXM + C6, (64) −cos ηA where Ci’s are arbitrary constants. We can split this algebra into su (2) su (2) by taking × Setting subsequently Ci to one and the rest of the coef- the following linear combinations ficients to zero we obtain, for each copy of the group A, v v v + v six linearly independent vector fields given by l = 5 − 6 r = 5 6 1 2 1 2 v v v + v sin (ξ) sin (χ) l = 3 − 2 r = 3 2 (71) 2 √ 2 √ v1 =  cot(η) sin (ξ)cos(χ)  (65) 2 2 v + v v v tan (η) sin (ξ)cos(χ) l = 4 1 r = 4 − 1 .  −  3 √ 3 √ cos(ξ) sin (χ) 2 2 v2 =  cot(η) sin (ξ) sin (χ)  (66) The commutators for li and ri become −tan (η)cos(ξ)cos(χ)  −  [l ,l ]= l [r , r ]= r (72) sin (ξ)cos(χ) 1 2 3 1 2 3 [l ,l ]= l [r , r ]= r (73) v3 =  cot(η)cos(ξ)cos(χ)  (67) 1 3 − 2 1 3 − 2 tan (η) sin (ξ) sin (χ) [l2,l3]=2l1 [r2, r3]=2r1 (74)   cos(ξ)cos(χ) v4 =  cot(η) sin (ξ)cos(χ)  , (68) [li, rj ]=0 i, j 1, 2, 3 (75) −tan (η)cos(ξ) sin (χ) ∀ ∈{ }   A closer inspection shows that li and ri form a set of left and and right invariant vector fields on SU (2), respectively [43]. 0 0 Since the above algebra was derived for each copy of the v5 =  1  v6 =  0  . (69) group A, the whole symmetry group of the local part of 0 1 the action becomes     [SU (2) SU (2)]×3 U (1)×Nc , (76) It is a direct calculation to check that these vector fields × × are divergence free, div (Vi)=0. This fact, together with acting on the base manifold by left and right multiplica- equation (61), implies tion as k k X k = ıCk u X k = ıCk u¯ , (70) u u¯ − L~η R~µ (~g)= η1g1µ1, η2g2µ2, η3g3µ3 , (77) ◦   which generates the usual U (1) symmetry of fields for ×3 for any ~η, ~µ SU (2) and on fields by multiplication each color. with a U (1) ∈phase. In order to find the symmetry group generated by the It is now trivial to insert this transformations in the in- fields v1, , v6 , we look at their algebra. The six di- teraction part of the action in order to verify which of mensional··· Lie algebra of v , , v is given in table (III) 1 ··· 6 the transformations remains a symmetry. It is easy to see that the symmetry group is preserved for tensorial interactions. Indeed, this is a remarkable feature of tensorial GFTs, 5 In this coordinates the metric on SU (2) is given by g = dη2 + which can be also stated as follows: by their very def- 2 2 2 2 sin η dξ + cos η dχ . inition, the symmetry group of tensorial interactions is 13

×6 the same as that of the local part of the action. In this the relation (30). Therefore, for any (g1, ,g6) G , ···×6 ∈ sense, this is a confirmation of the very motivation for in- there should exist a point (˜g1, g˜6) G such that troducing tensorial interactions as encoding the correct ··· ∈ ×4 new notion of locality for tensorial field theories [9]. C f (g1, ,g6)= f (˜g1, , g˜6) . (80) ◦ ··· ··· This is due to the fact that both of the symmetry groups we have found above form particular cases of the unitary Writing C in components as transformations characterizing tensorial interactions, as 1 2 1 the U (1) transformations are implemented by C (~g)= C (~g) , C (~g) , C (~g) , (81)
U (g,h)= δ gh−1 eıθ, (78) condition (80) implies
1 3 and the left (right) multiplication by the group is ob- C (g1,g2,g3)= C (g6,g4,g1) (82) 2 2 tained as C (g1,g2,g3)= C (g5,g2,g6) , (83)

−1 −1 U (g,h)= δ L1gh U (g,h)= δ gR1h . (79) which suggests the following decomposition of C,

1 2 3 In the case of simplicial models the status of both the C (g1,g2,g3)= C (g1) C (g2) C (g3) . (84) SU (2) and the U (1) group as symmetries of the full the- ory depend on the specific interaction in question. We Notice that in this case the diffeomorphism properties of need, then, to check explicitly the condition on the group C carry over to the components Ci. action (32). We postpone the verification of this condi- According to equation (29), the fields transform under C tion to the next section, since it will be the main tool in as the analysis of gauge invariant models. φ φ˜ = φ C−1. (85) 7→ ◦ E. Gauge invariant models The field φ˜ needs to be gauge invariant as well, otherwise the transformation C would leave the allowed space of ˜ In this section we study more in detail the symmetry fields. The gauge invariance of φ reads group of simplicial GFT interactions, and we show how ! the treatment can be significantly simplified in the pres- φ˜ Rh = φ C Rh = φ C = φ.˜ (86) ◦ ◦ ◦ ◦ ence of gauge invariance. Contrary to the previous case, we will use the interaction part to classify the symmetry Since this has to be true for all gauge invariant fields φ, the point C Rh (~g) needs to be in the same orbit (under group, subsequently checking which of the symmetries ◦ represent also a symmetry of the local action. The first the multiplication from the right by the diagonal group) as the point C (~g). This means that, for any h GD, part of the treatment is independent of the local part of ∈ the action and holds for a large number of base group there should exist an h˜ GD such that ∈ manifolds, which is why we do not specify the group at the beginning of the section. However, in order to verify C Rh = Rh˜ C, (87) ◦ ◦ the symmetry group for the local part we need to know the exact structure of the differential operator involved or point-wise and so we need to specify the underlying group as well. ˜ From this point onwards, we specialize the notation to C (~gh)= C (~g) h. (88) the n = 3 case for simplicity of exposition. The exten- As we show in the appendix (A), this restricts the C, up sion to generic n is straightforward. to discrete transformations, to be of the form

C (~g)= L~ h−1 ~gh, (89) 1. Admissible base manifold transformations · ×2 for some L G and h GD. The combinatorics of the interaction part is encoded in In the end,∈ the symmetry∈ group of the interaction part the function f from equations (29) and (30). For example becomes the combinatorial structure of a 3d simplicial interaction ×2 is given by G GD. (90) × f : (g1, ,g6) It is evident that this group already forms a symmetry ··· (g1,g2,g3) (g3,g4,g5) (g5,g2,g6) (g6,g4,g1) . group, due to the left invariance of the Haar measure. 7→ We can summarize the role of combinatorial structure Admissible transformations of the base manifold are and the gauge invariance on the transformation group of given by those functions C : G×3 G×3 that satisfy the base manifolds as follows → 14

3 × ×2 ×2 Diff G Diff (G) G G3D. (91) h i combinatorics−→ gauge−→ invariance ×

For higher dimensional models such as the Ooguri model 2. Symmetries of gauge invariant models with the interaction given by Following the procedure of the previous section, the in- f : (g1, ,g10) finitesimal symmetry condition for the simplicial interac- ··· 7→ (g ,g ,g ,g ) (g ,g ,g ,g ) (g ,g ,g ,g ) tion in three dimensions takes the form 1 2 3 4 4 5 6 7 7 3 8 9 × (g9,g6,g2,g10) (g10,g8,g5,g1) , × 0= DV L XQ + Div (LXM ) , (98) · we observe that the above treatment still results in the Writing the same condition in terms of ui, and using the transformation group fact that the only admissible base manifold transforma- tions are generated by divergence free vector fields, we 2 G G4D, (92) obtain × 2 2 3 4 1 1 3 4 2 where G acts by left multiplication as u u u XV + u u u XV 1 2 4 3 1 2 3 4 +u u u XV + u u u XV (99) (G , G , G , G )~g = G g1, G g2, G g3, G g4 . (93) 1 2 2 1 1 2 2 1 +c.c. =0.
i Hereby X is evaluated at the point (~gi, ~u) with ~g1 = Note on differences between simplicial combinatorics (g1,g2,g3), ~g2 = (g3,g4,g5), ~g3 = (g5,g2,g6) and ~g4 = i (g6,g4,g1) and ~u = φ (~g) i∈{1,··· ,4}. Notice that ~u As we mentioned above, the combinatorial structure for is not u1,u2,u3,u4 since the latter tuple is given by i simplicial models can vary. This variation is captured φ (~gi) i∈{1,··· ,4}. Equation
(99) needs to hold true for by different functions f, as used in the beginning of this any φi and any point of the base manifold. section. i Inserting the formal power series expansion of XV For the original Boulatov interaction we get Xi (~g, ~u)= Θi (~g) φ1 (~g)m1 φ4 (~g)m4 φ¯1 (~g) φ¯4 (~g) , V ~m ··· ··· f : (g1, ,g6) X~m ··· (g ,g ,g ) (g ,g ,g ) (g ,g ,g ) (g ,g ,g ) . 7→ 1 2 3 1 4 5 2 5 6 3 6 4 in equation (99) we observe that all the coefficient func- tions Θi vanish except for one, such that The resulting transformations become Xi (~g, ~u)=Θi (~g) ui. (100) 1 1 1 V C (~g)= C (g1) C (g2) C (g3) , (94) Equation (99) becomes where all the components are the same. It is easy to check 1 2 3 4 that this transformation also respect the cyclic permuta- Θ (~g1)+Θ (~g2)+Θ (~g3)+Θ (~g4)=0. (101) tion condition and therefore we get the symmetry group of the Boulatov model as As we show in the appendix (C), the only functions that are gauge invariant and satisfy the above equation are constants, G G. (95) × θi = const. θi =0 . (102) In the colored case, instead, we get Xi

i #c−1 f : (g1, ,g6) Therefore X generate the symmetry group U (1) , ··· V (g1,g2,g3) (g1,g4,g5) (g6,g2,g5) (g6,g4,g3) , where #c is the number of colors in the interaction part 7→ of the model. Notice, that if the model is not colored, and the resulting admissible transformations are that is the number of colors is one, the U (1) symmetry is not present. The overall symmetry group for simplicial 1 2 3 C (~g)= C (g1) C (g2) C (g3) . (96) models becomes G×n G U (1)#c−1 , (103) The group of admissible transformations is therefore × × where n depends on the actual combinatorial pattern, as G×3 G. (97) × we have shown. This classification of symmetries also fits 15

Model Symmetry Group Action

~g 7→ LC (~g)

1 2 3 4 ×3 ×3 φ1,2,3φ1,4,5φ6,2,5φ6,4,3 G × U (1) c ıθc c c φ 7→ e φ Pc θ = 0 1 2 3 with C = c , c , c

×2 ~g 7→ LC (~g) 1 2 3 3 4 5 5 2 6 6 4 1 φ , , φ , , φ , , φ , , G 1 2 1 with C = c , c , c

P P P P ~g 7→ LC (~g) φ1,2,3 φ1,4,5 φ2,5,6 φ3,6,4 G with C =(c, c, c)

~g 7→ LC (~g) ¯ ¯ ×2 φ1,2,3 φ3,4,5 φ5,2,6 φ6,4,1 G × U (1) φ 7→ eıθφ 1 2 1 with C = c , c , c

×2 ~g 7→ LC (~g) 1 2 3 4 4 5 6 7 7 3 8 9 9 6 2 10 10 8 5 1 φ , , , φ , , , φ , , , φ , , , φ , , , SU (2) 1 2 2 1 with C = c , c , c , c

~g 7→ LS (~g)

×2 Barrett-Crane Spin (4) × SU (2) ~g 7→ c · (~g; k) · c−1 1 2 2 1 with S = s ,s ,s ,s

Table IV: Models and their symmetry groups excluding gauge symmetries of the fields. Hereby, ci SU (2) and i ∈ s Spin (4) and LC denotes the left multiplication by C. ∈ the model with the interaction part of the type φφφ¯ φ¯, be that of an extended Ooguri model from equation (92) since we defined it as a model with two differentR colors. where the group G is now specified to Spin (4) The symmetry group of colored models, which is indepen- dent from the precise combinatorial pattern of field argu- Spin (4)×2 Spin (4) G (k) . (106) ments, is the largest compatible with the one of the local × × part of the action (and with gauge invariance), and coin- The group G (k) denotes a group of transformations of cides with the one of the corresponding tensorial model. the SU (2) element ki. However, remember that the ex- This gives a different perspective, and confirms, the close tension of the GFT field the SU (2) variable k was needed relation between colored simplicial models and tensorial for consistent implication of simplicity constrains and ones, highlighted first in [35] in terms of properties of the therefore the actual meaning of G (k) is relevant only corresponding functional integrals. after the imposition of simplicity constrains. Equation (106) provides the symmetry group of extended Ooguri model with gauge invariance, in order to obtain 3. Barrett-Crane model the symmetry group of the BC model simplicity con- straints need to be further imposed. We refer to the We now briefly discuss the implication of the simplic- appendix (B) for explicit calculations and state here just ity constraints, in the Barrett-Crane formulation, on the the result of imposing the simplicity constrains on the symmetry group. field φ, by imposing invariance under the projector Applying the above analysis to the BC model from equa- from equation (20). As we show in the appendix (B), theS tion (16) defined by the following function f simplicity constrains f : (g , g ; k , , k ) (104) 1 ··· 10 1 ··· 4 7→ φ = φ, (107) (g1,2,3,4; k1) (g4,5,6,7; k2) (g7,3,8,9; k3) ◦ S reduce the symmetry group of the Ooguri model (for the (g9,6,2,10; k4) (g10,8,5,1; k5) , (105) × chosen combinatorics) down to we realize that the symmetry group for the gauge in- variant BC model without simplicity constrains would Spin (4)×2 SU (2) , (108) × 16 where the SU (2) group replaces the Spin (4) G (k) part fields (XM ,XV ) then the following identity holds for all from equation (106) and acts on the elements× of the local i base manifold of the BC model Spin (4)×4 SU (2) by × I c i conjugation as EL [XQ]= DF L (XcQ) δ δ c Z − −1 −1 −1 X   c (~g−,~g+; k) := (c,c; c) (~g−,~g+; k) c ,c ; c . L L ◦ · · DivM DJ L [XQ]+ L XM
− · ×2
And Spin(4) acts by the left multiplication I c c DivM DJ L X δ (109) − Z Q  1 2 3 4 Ω
(G1, G2, G2, G1) (~g; k)= G1g , G2g , G2g , G1g ; k . I i
DivM˜ L XM˜ δ . − Z ˜ · The same considerations we have made regarding the de- Ω
pendance of the symmetry group on the combinatorics Here denotes the vector field of base manifold trans- and on the use of colors apply also to the Barrett-Crane XM˜ ˜ case. formations of M generated by XM as we discussed in the In table (IV) we summarize the symmetries of different previous section, DF L (XcQ) denotes the Fréchet deriva- c interaction terms. tives of the Lagrangian in the direction of XcQ, δ denotes the delta distribution on the domain of the field of color c I I and the non-local Lagrangian L = L (x, φ (x) , ∂φ x) is | III. Classical currents assumed to be a function on the base manifold, fields at the point, and first derivatives of the fields at the same We will now derive the (generalized) conservation laws for point. The left hand side denotes the equations of motion the symmetries we identified in the last section. Once contracted with the vector field XQ. more we limit ourselves to the classical regime of the In the case when the non-local Lagrangian is indepen- GFTs, postponing the analysis of the full quantum the- dent of derivatives of the fields, and the generators of ory. Also, we stress again that the conservation laws the symmetry group of base manifold transformations and corresponding currents, just like the whole kinemat- are divergence-free, div (X )=0, and when the trans- ics and dynamics of such quantum field theories, should M formations of the field values is proportional to the field not be interpreted in spatiotemporal or geometric terms, value itself, Q (φ ) φ , the above identity simpli- at least in general. Even for GFT models with a direct 1,2,3 1,2,3 fies significantly to ∝ quantum gravity interpretation, the spatiotemporal and geometric meaning of the various aspects and regimes of each model should be extracted and analyzed with care. L L EL [XQ] = ∆ DivM DJ L [XQ]+ L XM , (110) On this note, we point out that the classical GFT equa- − · tions of motion of 4d quantum gravity models, which
capture the hydrodynamics of special condensate states where of the theory, have been given a cosmological interpre- L c c tation and have been studied in some detail and with ∆= DcV L X δ , (111) Z Q remarkable results in a series of recent works [26–29, 44– Xc
48]. is referred to as correction term. This result explicitly shows that the currents associated to symmetries of the A. Conservation laws in non-local field theories non-local action are no longer conserved. Instead their divergence are dictated by the non-local part of the ac- tion. In local field theories there is a conserved current asso- ciated to every continuous symmetry of the action given After imposing equations of motion on the fields, we get by the famous Noether theorem. However, for non-local the identity that replaces the usual Noether theorem theories this result does not hold as such, and must be L L DivM DJ L [XQ]+ L XM =: Div (J) = ∆. (112) generalized, due to the fact that the equations of motion · become integro-differential equations.
In [11] we derived an equivalent expression for Noether The quantity in brackets on the left hand side is the currents for the case of non-local field theories, and for Noether current of the local part of the action and the the associated generalized conservation laws. In order right hand side of the equation is the non-vanishing di- to keep the notation simple we present here a simplified vergence of the current due to non-local structure of the version of the theorem, referring to the original work for theory. the full statement. It becomes now a straightforward calculation to apply the equation (112) to models and symmetries introduced Theorem 2. L I If a non-local action S = M S + M˜ S is in the previous section. In the rest of this section we symmetric under a group action generatedR by theR vector summarize the resulting identities. 17

Model Symmetry Group Action

SU ×3 λ δc X⊕3 φ1φ2φ3φ4 1 2 3 4 (2) ∆= −2 Pc R Re  cM
 φ1,2,3φ1,4,5φ6,2,5φ6,4,3 ×3 c 1 2 3 4 U (1) ∆= −ı2λ Pc θc R δ Im φ φ φ φ

×2 ⊕2 φ1,2,3 φ3,4,5 φ5,2,6 φ6,4,1 G ∆= −8λ Re XM (φ) R φφφ

P P P P φ1,2,3 φ4,3,5 φ5,2,6 φ6,4,1 G ∆= −8λ Re XM (φ) R φφφ

×2 G ∆= −4λ Re XM (φ) R φφ¯ φ¯
φ1,2,3 φ¯3,4,5 φ5,2,6 φ¯6,4,1 U (1) ∆= −4λθ Im φ R φφ¯ φ¯

×2 ⊕2 φ1,2,3,4 φ4,5,6,7 φ7,3,8,9 φ9,6,2,10 φ10,8,5,1 SU (2) ∆= −10λ Re XM (φ) R φφφφ

×2 ⊕2 Spin (4) ∆= −10λ Re XM (φ) R φφφφ
Barrett-Crane

SU (2) ∆= −10λ Re XM (φ) R φφφφ

Table V: Models and their correspondent correction terms. The vector fields XM are the left invariant vector fields given in equation (71)

Since the local part of all our models is given by equation IV. Conserved charges in presence of matter (9), the Noether current does not change and can be written as In this section we will discuss the consequences of the c c L matter coupling introduced in [29] and show that such J = κ φ X¯Q + φ¯ XQ + L XM . (113) ∇ · ∇ · · coupling implies the existence of quantities which are con- Xc
stant in the matter field variable, and can be interpreted Note that for the U (1) symmetry XM = 0, and the as conserved charges. Noether current becomes proportional to κ. This au- While this could be taken as a fact of purely mathemat- tomatically implies that for all static models the Noether ical interest, it may also indicate some underlying inter- current associated to the U (1) symmetry is zero. The esting physics, for quantum gravity models. The reason correction term, however, may not trivially vanish. Apart is the following. The type of matter field introduced in from the values for κ, the models introduced earlier will [29] was a free, massless, minimally coupled real scalar differ only by the correction term in equation (112). In field, entering as an additional variable in the domain table (V) we will present the correction terms for dis- of the GFT fields, for 4d gravity models, whose classical cussed models. dynamics was then studied. The same classical dynam- The notation in table (V) is as follows. For brevity we ics was given an interpretation as cosmological dynamics do not indicate the base points and write φ1φ2φ3φ4 for continuum homogeneous universes, emerging from the in order to refer to the non-local part of theR Boulatov GFT system as quantum condensates. As customary in action. We also write φ1 φ2φ3φ4 for quantum cosmology, and to some extent compulsory in R background independent, diffeomorphism invariant the- 1 2 3 4 ories, the dynamics was expressed in terms of relational φ1,2,3 dg4,5,6 φ3,4,5φ5,2,6φ6,4,1. (114) Z observables [49–51]. In particular, the added scalar field was chosen to play the role of relational clock, i.e. the The integral φ2φ3φ4 can be seen as a function evaluated physical time variable in terms of which describing the at the point (g ,g ,g ). We denote the Lie derivative R 1 2 3 evolution of all the geometric observables, e.g. the vol- of this function with respect to the vector field X as M ume of the universe. We refer to [29] for more details. X φφφ . For brevity we denote the expression in M Remarkably, the same variable enters the GFT action (114) also by the formal delta distribution R
just as a standard, local time coordinate would enter an

1 2 3 4 1 1 2 3 4 ordinary field theory. This suggests a deeper physical φ φ φ φ δ = φ dg4,5,6 φ φ φ , Z 1,2,3 Z 3,4,5 5,2,6 6,4,1 meaning for the charges that, following some symmetry of the corresponding GFT model, are in fact conserved meaning that the integral over the domain of the field φ1 with respect to the same relational time variable/clock. is to be excluded. We do not discuss further the possible physical interpre- 18 tation and confine ourselves to the mathematical analysis simplifies to of such extended models. loc c loc c The domain of the GFT field is extended to become EL ϕ [XQ] ∂ϕ ∂∂ φc L X + ∂ ¯c L X¯ Z | ≃ ϕ Q ∂tφ Q M    
loc φ : M R C, (115) + ∂ϕ L Xϕ × → ·
int c int c + ∂ϕ D∂ φc L X + D ¯c L X¯ where M is the base manifold of the correspondent GFT ϕ Q ∂ϕφ Q ZM   model without matter and R describes the degree of free-

int dom of a real scalar field. We call the new base mani- + ∂ϕ L Xϕ , (117) ZM · fold Mmat = M R, and denote a point on Mmat as
× (g , ,gn, ϕ). The field value at this point is then de- 1 ··· where the equality is true up to a minus sign. Taking the noted φ (g1, ,gn, ϕ) = φ1,··· ,n,ϕ. Intuitively we can ϕ component of the current we get think of GFT··· field φ as describing a “chunk” of space in loc c loc c loc which the scalar field takes the value ϕ. Q (ϕ):= ∂∂ φc L X + ∂ ¯c L X¯ + L Xϕ ϕ Q ∂ϕφ Q ·     The dynamics is then described by an action which is int c int c + D∂ φc L X + D ¯c L X¯ non-local in the group variables, but local in the addi- Z ϕ Q ∂ϕφ Q tional matter field variable. This means that every La- M 

 int grangian in the non-local action sum is evaluated at the + L Xϕ . (118) Z · same value ϕ: M
Due to equation (117), this satisfies on shell S = φ (~g, ϕ) K (~g, ϕ) φ (~g, ϕ) ZM×R ∂ϕQ (ϕ)=0. (119) + φ (~g, ϕ) φ ~h, ϕ V ~g, ,~h, ϕ .,(116) Since the interaction Lagrangian does not depend on ZM×R ···    ···  derivatives of φ, the conserved charge becomes ϕ with the dependence of the various terms in the action on Q (ϕ):= S ϕ XR (120) the additional scalar field being motivated by an analy- | · loc c loc c c ¯ sis of the GFT Feynman amplitudes and their relation + ∂∂ϕφ L XQ + ∂∂ φ¯c L XQ , Z ϕ with simplicial gravity path integrals, and by the re- M      loc int quired symmetries of the scalar field dynamics. The fur- where S ϕ = L ϕ + L ϕ is the action in equa- | M | M | ther requirements that the scalar field is free, massless tion (116) atR a fixed valueR of the parameter ϕ. and minimally coupled, plus some further approxima- For example in the case of a U (1) symmetry which is c c tion motivated by the hydrodynamics setting [29], lead generated by XQ = ıθcφ , with c θ = 0, we get the to K (~g, ϕ)= (~g)+∆ϕ, and to a vertex function V that conserved charge P is independentK of ϕ.

c c loc c loc c ¯ c Q (ϕ)= ı θ φ ∂∂tφ L φ ∂∂tφ¯ L . (121) ZM − A. Conserved charges and symmetries Xc

For the SU (2) symmetry, with XQ = XM (φ) and XM − Locality in the matter field allows to define a local slic- being left invariant generators of SU (2) as in equation ing with respect to which we can construct conserved (71), Q takes instead the form quantities Q (ϕ) for any symmetry of the action, such loc c c that ∂ϕQ (ϕ)=0. This is easily seen from the equa- Q (ϕ)= ∂∂ϕφ L [XcM (φ )] tion (109), where the integral domain is now replaced by − ZM c Mmat = M R and the delta function δ that acts on the loc c c ¯ × c ϕ ∂∂ϕφ¯ L XcM φ . domain of the field with color c can be written as δM δR , − ZM ϕ 
 where δc acts on the group part of the domain and δ M R This shows that we can easily calculate “conserved” quan- fixes the value of the matter field. Integrating the above tities for the symmetries we found earlier in the paper. equation over M and taking into account that the action However, in addition to the symmetries on the group is local in the parameter ϕ, as well as the fact that the space we may also have symmetries on R which corre- base manifold M has no boundary6, the above equation spond to symmetries of the matter field, so the symmetry group of the models will be larger. In general, the symmetry of the matter field will also be strongly model dependent, and have to be investigated 6 If the underling group of the model has a boundary, then bound- on a case by case basis. However, in the case of free, ary terms need to taken into account. massless, minimally coupled scalar matter, the action is 19

(and should be) invariant under matter field translations “conservation” laws that correspond to continuous sym- of the form ϕ ϕ + µ. The charge for this symmetry metries. And were able to show that in particular cases of will take the following7→ form matter coupling to GFT fields our construction provides a notion of conserved charges, the same way Noether the- loc c loc c orem does in local field theories. An existence of con- c c ¯ Q (ϕ)= ∂∂ϕφ L ∂ϕφ + ∂∂ϕφ¯ L ∂ϕφ + S ϕ. − ZM   | served quantities shows, that once a matter field satisfies a notion of a “good” clock it also obtains the usual “time” c loc c Defining Π := ∂∂ϕφ L , this takes the form of the Leg- properties in the field theoretical frame-work. As we al- endre transform of the Lagrangian defined by S t ready mentioned, a lot more should be understood about | such conserved charges in GFT models. c c c c It is an exciting and important task to understand Q (ϕ)= Π ∂ϕφ + Π¯ ∂ϕφ¯ + S ϕ. (122) − ZM | the consequences of the GFT symmetry groups on the
physics of these models. This is what needs primarily to This is of course extremely suggestive of a GFT Hamil- be addressed in the future. tonian with respect to the evolution defined by the rela- On the one hand, an understanding of conservation laws tional “time” ϕ, and this is certainly an important point in terms of geometrical objects could be a very impor- to be investigated further, in both its mathematical and tant step in the development of the theory. Conserva- physical consequences. tion laws and conserved charge equations could provide It is important to note that there are very special con- a field theoretical explanation of cosmological features ditions that the matter field ϕ has to satisfy to repre- stemming from the underlying quantum gravity models, sent a good relational clock. It is interesting to investi- in the context of GFT condensate cosmology [29, 47, 52]. gate further also how these conditions, and their relax- The very existence of a condensate phase in GFT mod- ation, reflect on the dependence of the GFT action on els, and more generally their macroscopic phase dia- the same matter field variable, and what field-theoretic gram, currently being explored mainly by FRG methods consequences they have, in particular concerning the ex- [36, 53, 54], can now be studied also on the basis of GFT istence and form of the conserved charges we have found. symmetries and corresponding symmetry breaking. Moreover, it is easy to realize that, if the model has more On the other hand a classification of symmetry groups than one matter field that enters the action locally, the in GFT could be used as a better characterization of the above treatment can be performed for any of the matter theory space, a crucial ingredient for systematic renor- fields. In this case, however, above equations will contain malization group studies [10, 30, 55–57]. additional boundary terms. We leave further analysis of In particular this could help clarifying the connection be- these points to future work. tween simplicial and tensorial GFT models. As we noted, a further indication of their close connection has been found already in our analysis, showing that only colored V. Conclusion and outlook GFT models of simplicial type appear to have an U (1) symmetry as well as the unrestricted translation invari- In this paper we provided an extensive symmetry analysis ance that is found in tensorial GFTs. for various models in Group Field Theory. From a more mathematical point of view, it appears to be very interesting to understand the extension of the We have elucidated the symmetry group of various GFT symmetry groups we considered to Lie-Baecklund or gen- models, and how it is affected by the various ingredients eralized symmetries, which requires a better characteri- entering their definition: rank, base group, color, combi- zation of the equivalence class of GFT actions leading to natorial structure. the same classical equations of motion. Our main result shows that, apart form the expected Finally, we need to go beyond the purely classical analysis symmetry groups of left multiplication and U (1), the performed in this paper, and move to the analysis of the discussed models do not posses any other continuous Lie same symmetries we have identified at the quantum level, point symmetries. This holds even in the case of static, deriving and studying in detail the corresponding Ward gauge invariant, models, in which the Lagrangian does identities, and the issue of possible anomalies. not depend on derivatives of fields. This is not obvious since an ordinary local field theory without dynamical terms would possess a fairly large gauge group of diffeo- morphisms of the base manifold. However, the presence of the interaction term with a particular combinatorial structure as well as the requirement of gauge invariance insures that the symmetry group becomes very small. In this sense our treatment provides a complete set of point symmetries of discussed models. Using our previous result on conservation laws for non- local theories we were then able to calculate generalized 20

A. Reduction of transformations due to gauge In the following we will denote a group element of invariance Spin (4) by its two copies of SU (2) a

Spin (4) g = (g ,g ) , From equation (88) the requirement on the transforma- ∋ − + tion reads a tuple of four elements is referred to by the vector no- C (~gh)= C (~g) h.˜ (A1) tation

Writing out this equation in components we get ~g = (~g−,~g+) .

1 1 We will also sometimes write g , , , for the tuple of ele- C (g1h)= C (g1) h˜ 1 2 3 4 ments (g ,g ,g ,g ). 2 2 ˜ 1 2 3 4 C (g2h)= C (g2) h (A2) A base manifold transformation of the model is denoted 1 1 ×4 ×4 C (g3h)= C (g3) h,˜ by C : Spin (4) SU (2) Spin (4) SU (2). We denote the components× of this→ transformation× as with Ci being a diffeomorphism on the group G. At this − + − + point we employ the known relation C (~g, k)= C , C , , C , C , Ck . 1 1 ··· 4 4


Diff (G) G Diff1 (G) , (A3) ± ≃ × Here all the component functions Ci are functions on the base manifold and therefore depend on points of the that states that the group of diffeomorphisms on G form (~g, k). However, the combinatorial structure of the is diffeomorphic (as a manifold) to the group G itself BC model dictates the following conditions on the com- (that acts by left multiplication) times a group of dif- ponents feomorphisms that stabilizes the identity of G, denoted i 1 Diff (G). This implies that every C can be writ- C1 (g1,2,3,4, k1)= C4 (g10,8,5,1, k5) i i ten by some c G and Diff1 (G) such that C (g , k )= C (g , k ) . Ci (g) = ci i (g)∈with i (1D) =∈1. Inserting this re- 2 1,2,3,4 1 3 9,6,2,10 4 D D lation in the equations (A2) and evaluating it a the point From the above relations we see that the components of 1 g1 = g2 = g3 = we observe the transformation have the following dependences c1 1 (h)= c1 h˜ (A4) · D · C (g1,2,3,4, k) = (C1 (g1) , C2 (g2) , C3 (g3) , C4 (g4) , Ck (k)) . c2 2 (h)= c2 h˜ (A5) · D · A priori we do not have any additional constrains on the c1 3 (h)= c1 h,˜ (A6) · D · component Ck. However, since C is a diffeomorphism which, in tern, implies and Ci are diffeomorphisms, the transformation of the normal has to be a diffeomorphism as well 7. Again in- 1 (h)= 2 (h)= 3 (h)= h˜ =: (h) . (A7) voking the diffeomorphism of manifolds Diff (Spin (4)) ≃ D D D D Spin (4) Diff1 we denote the components of C that be- × Inserting this relation again in (A2) at an arbitrary point long to Diff1 by the lower case c. ~g we get for At this point we remind the reader that in the Barrett- D Crane model the gauge invariance of the fields was ex- (gih)= (gi) (h) . (A8) D D D tended to incorporates simplicity constrains In other words is an homomorphism and therefore an : (~g; k) 1; h−1 (~g; k) k~uk−1, ~u, ; 1 (h; h ) . D S 7→ − · · · + automorphism. On G however, the group of automor-


Where stands for the group multiplication and ; sep- phisms splits in the inner automorphisms which are given · by a conjugation with a fixed group element and outer arates the Spin (4) elements from SU (2). This means that the fields of the model are invariant under , automorphisms which are given by automorphisms of the S Dynkin diagram of the group and relate to the discreet φ = φ. symmetries. Focusing on continuos transformations we ◦ S get Since the fields are transformed under C as φ φ C−1 we again get the following relations for the transforma-7→ ◦ (g)= d g d−1 (A9) D · · tion C for some fixed d SU (2). ∈ φ C S = φ C, ◦ ◦ ◦

B. Barrett-Crane model 7 Notice, that it would not be true if we didn’t have restriction In this section we are going to show what are the admis- on Ci, since then Ci would not be a diffeomorphism and hence sible transformation in the Barrett-Crane model. neither needs to be Ck. 21

×4 or equivalently for each h Spin (4), u SU (2) and C. Constancy of the phase ∈ ×4 ∈ gi Spin (4) there exist u˜ SU (2) and h˜ Spin (4) ∈ ∈ ∈ and k˜ SU (2) such that In this section we are going to prove the following state- ∈ ment Ci (g uk h)= Ci (g) u˜C h˜ (B1) · · · k · −1 ˜−1 ˜ Theorem. If for any point g1, ,g6 G where G is a Ck h− kh+ = h− Ck (k) h+, (B2) simple Lie group the following equation··· ∈ holds
−1 where we write uk = kuk ,u . It is again obvious that the left multiplication by Spin
(4) is untouched by 4 i this transformation, however this is not true for normal θ (~gi)=0. component Ck. We first focus on the transformations Ci Xi and treat the normal component C afterwards. k i 1 And for any h GD (2) the functions θ satisfy From the from of uk we notice that for u = the left ∈ hand side does not depend on k and so should’t the right i i 1 1 θ Rh = θ , (C1) hand side. It follows that for u = we have u˜ = . ◦ Equation (B1) then reads for the Diff1 part, then the functions θi are constants that add up to zero, 1 2 3 4 ci (g h)= ci (g) h.˜ θ + θ + θ + θ =0 . · · It follows that ci is a homomorphism on Spin (4) and We first prove the following lemma therefore is either conjugation by a fixed element of Lemma 3. R Spin (4) or a flip of the SU (2) parts, which is a dis- Let θ be a function from a Lie group G to such that for all g G the difference crete transformation. Hence, if ci is continuous it can be ∈ written as θ (gh) θ (g)= f (h) −1 − ci (g)= s g s , · · is a function only on the “distance” of the points h. Then where g,s Spin (4). This implies R ∈ f is a homomorphism from the group G to ( , +). h˜ = s h s−1. 1 · · Proof. From the definition it follows that f ( ) = 0. −1 Inserting this relation now in equation (B2) we obtain Choosing g =˜gh we get −1 −1 −1 −1 −1 −1 Ck h kh = s h s Ck (k) s h s . f (h)= θ (gh) θ (g)= θ (˜g) θ gh˜ = f h . − + − − − + + + − − −




Splitting C in the left multiplication by SU (2) and Diff1 k Choosing g = ghh˜ we also get we get for some fixed w SU (2) ∈ −1 −1 −1 −1 f hh˜ = θ ghh˜ θ (g) wck h− kh+ = s−h− s− wck (k) s+h+s+ . − (B3)    


˜ Choosing h = h and setting k = 1 we get = θ ghh θ (gh) θ (g) − +   ± − −1 −1 −1 ˜ w = s−h− s− w s+h−s+ , = f h + f (h) .  
1
which can only be satisfied if w = . Which concludes the proof. Inserting equation (B3) in (B1) and using the fact that ci is a homomorphism yields We now prove the above theorem. −1 ci (uk)= ci (k, 1) ci (u,u) ci k , 1 · · Proof. The above equation then reads −1
= ck (k) ci (u,u) ck k . 1 2 3 4
θ (~g1)+ θ (~g2)+ θ (~g3)+ θ (~g4)=0, (C2) Hence, ci (a,b) = (ck (a) ,ck (b)) and ck is a homomor- phisms itself. Therefor where ~g1 = (g1,g2,g3), ~g2 = (g3,g4,g5), ~g3 = (g5,g2,g6) −1 ci (g) = (s,s) g (s,s) , and ~g4 = (g6,g4,g1). Than for any differentiable curve · · c : R I SU (2) with c (0) = 1 the above equation and c (k)= sks−1. ⊃ → k is true if we replace g1 by the curve c (t). Deriving the These are the only admissible transformations that pre- resulting equation with respect to the parameter t we get serve the combinatorial structure of the theory and re- 1 4 spect the simplicity constraints together with gauge in- ∂tθ (c (t) ,g2,g3)+ ∂tθ (g6,g4,c (t))=0. variance. Notice that itself would fail the requirement S (32) and therefor can not be seen as a base manifold By integration we obtain transformations, which is why we dot not obtain the sym- metry under in this approach. θ1 (c (t) ,g ,g ) θ1 (1,g ,g )= θ1 (c (t)) , S 2 3 − 2 3 1 22

1 ×3 for some function θ1 (c (t)). Applying the same argument Since this equation needs to hold for any ~g G we get 4 i ∈ to θ we gain the following relations, for each θj the following condition 1 1 1 1 θ (g1,g2,g3)= θ1 (g1)+ θ ( ,g2,g3) 4 1 4 i i i θ (g ,g ,g )= θ (gd )+ θ (1,g ,g ) . θj (gh) θj (g)= fj (h) , (C4) 6 4 1 − 1 1 2 3 − Inserting these relations into equation (C2) yields i with some functions fj . From the above lemma it fol- 1 2 3 4 θ (1,g2,g3)+ θ (~g2)+ θ (~g3)+ θ (g6,g4, 1)=0. i R lows that fj is a homomorphism from G to ( , +). Since (R, +) is abelian and G is simple f is a constant zero Successively performing the same step for all other group function, f =0. elements g eventually leads to the separation of the func- i 1 tions θi as follows, Evaluating equation (C4) at g = proves that

i i i i i θ (g1,g2,g3)= θ (g1)+ θ (g2)+ θ (g3)+ const. , (C3) i i 1 2 3 θj = const , i where θj ’s satisfy which together with equation (C3) proves 1 4 1 3 1 2 θ1 = θ3 θ2 = θ2 θ3 = θ1 − − − i c i θ2 = θ4 θ2 = θ3 θ (~g, φ )= θ , 2 − 2 3 − 1 3 4 θ3 = θ1. − for some constants θi. The conditions on the constants Using the requirement on gauge invariance (equation follows. (C1)) yields for any h G D ∈ 3 i i i i θ1 (g1h)+ θ2 (g2h)+ θ3 (g3h)= θ (~g) .

[1] D. Oriti, “The Group field theory approach to Equations”, eConf C0107094 (2001) 263–270. quantum gravity”, arXiv:gr-qc/0607032. [13] N. H. Ibragimov, V. Kovalev, and V. Pustovalov, [2] D. Oriti, “The Group field theory approach to “Symmetries of integro-differential equations: A quantum gravity: Some recent results”, AIP Conf. survey of methods illustrated by the benny equations”, Proc. 1196 (2009) 209–218, arXiv:0912.2441. Nonlinear Dynamics 28 (2002), no. 2, 135–153. [3] D. Oriti, “The microscopic dynamics of quantum [14] S. V. Meleshko, Y. N. Grigoriev, N. K. Ibragimov, and space as a group field theory”, in “Proceedings, V. F. Kovalev, “Symmetries of integro-differential Foundations of Space and Time: Reflections on equations: with applications in mechanics and plasma Quantum Gravity: Cape Town, South Africa”, physics”, Springer Science & Business Media, 2010. pp. 257–320. 2011. arXiv:1110.5606. [15] J. Ben Geloun, “Classical Group Field Theory”, J. [4] D. Oriti, “Group field theory as the 2nd quantization Math. Phys. 53 (2012) 022901, arXiv:1107.3122. of Loop Quantum Gravity”, Class. Quant. Grav. 33 [16] A. Baratin and D. Oriti, “Group field theory with (2016), no. 8, 085005, arXiv:1310.7786. non-commutative metric variables”, Phys. Rev. Lett. [5] T. Krajewski, “Group field theories”, PoS 105 (2010) 221302, arXiv:1002.4723. QGQGS2011 (2011) 005, arXiv:1210.6257. [17] A. Baratin and D. Oriti, “Group field theory and [6] D. Oriti, “Group Field Theory and Loop Quantum simplicial gravity path integrals: A model for Gravity”, 2014. arXiv:1408.7112. Holst-Plebanski gravity”, Phys. Rev. D85 (2012) [7] M. Gross, “Tensor models and simplicial quantum 044003, arXiv:1111.5842. gravity in > 2-D”, Nucl. Phys. Proc. Suppl. 25A [18] A. Baratin and D. Oriti, “Quantum simplicial (1992) 144–149. geometry in the group field theory formalism: [8] R. Gurau and J. P. Ryan, “Colored Tensor Models - a reconsidering the Barrett-Crane model”, New J. Phys. review”, SIGMA 8 (2012) 020, arXiv:1109.4812. 13 (2011) 125011, arXiv:1108.1178. [9] V. Rivasseau, “Quantum Gravity and [19] J. C. Baez, “Spin foam models”, Class. Quant. Grav. Renormalization: The Tensor Track”, AIP Conf. Proc. 15 (1998) 1827–1858, arXiv:gr-qc/9709052. 1444 (2011) 18–29, arXiv:1112.5104. [20] S. Alexandrov and P. Roche, “Critical Overview of [10] S. Carrozza, “Tensorial methods and renormalization Loops and Foams”, Phys. Rept. 506 (2011) 41–86, in Group Field Theories”, PhD thesis, Orsay, LPT, arXiv:1009.4475. 2013. arXiv:1310.3736. [21] J. Ben Geloun, R. Gurau, and V. Rivasseau, [11] A. Kegeles and D. Oriti, “Generalized conservation “EPRL/FK Group Field Theory”, Europhys. Lett. 92 laws in non-local field theories”, J. Phys. A49 (2016), (2010) 60008, arXiv:1008.0354. no. 13, 135401, arXiv:1506.03320. [22] S. Alexandrov, M. Geiller, and K. Noui, “Spin Foams [12] Z. J. Zawistowski, “Symmetries of Integro-Differential and Canonical Quantization”, SIGMA 8 (2012) 055, 23

arXiv:1112.1961. Left and right invariant vector fields and one-forms”, [23] A. Perez, “The Spin Foam Approach to Quantum ArXiv e-prints, 2010 arXiv:1003.2708. Gravity”, Living Rev. Rel. 16 (2013) 3, [44] S. Gielen, D. Oriti, and L. Sindoni, “Cosmology from arXiv:1205.2019. Group Field Theory Formalism for Quantum [24] C. Rovelli, “Loop quantum gravity”, Living Reviews in Gravity”, Phys. Rev. Lett. 111 (2013), no. 3, 031301, Relativity 11 (2008), no. 5,. arXiv:1303.3576. [25] D.-W. Chiou, “Loop Quantum Gravity”, Int. J. Mod. [45] S. Gielen, “Perturbing a quantum gravity condensate”, Phys. D24 (2014), no. 01, 1530005, arXiv:1412.4362. Phys. Rev. D91 (2015), no. 4, 043526, [26] S. Gielen, D. Oriti, and L. Sindoni, “Homogeneous arXiv:1411.1077. cosmologies as group field theory condensates”, JHEP [46] S. Gielen, “Identifying cosmological perturbations in 06 (2014) 013, arXiv:1311.1238. group field theory condensates”, JHEP 08 (2015) 010, [27] S. Gielen and D. Oriti, “Quantum cosmology from arXiv:1505.07479. quantum gravity condensates: cosmological variables [47] D. Oriti, L. Sindoni, and E. Wilson-Ewing, “Bouncing and lattice-refined dynamics”, New J. Phys. 16 cosmologies from quantum gravity condensates”, (2014), no. 12, 123004, arXiv:1407.8167. arXiv:1602.08271. [28] S. Gielen, “Quantum cosmology of (loop) quantum [48] M. de Cesare, A. G. A. Pithis, and M. Sakellariadou, gravity condensates: An example”, Class. Quant. “Cosmological implications of interacting Group Field Grav. 31 (2014) 155009, arXiv:1404.2944. Theory models: cyclic Universe and accelerated [29] D. Oriti, L. Sindoni, and E. Wilson-Ewing, “Emergent expansion”, arXiv:1606.00352. Friedmann dynamics with a quantum bounce from [49] B. Dittrich, “Partial and complete observables for quantum gravity condensates”, arXiv:1602.05881. Hamiltonian constrained systems”, Gen. Rel. Grav. 39 [30] J. Ben Geloun and V. Bonzom, “Radiative corrections (2007) 1891–1927, arXiv:gr-qc/0411013. in the Boulatov-Ooguri tensor model: The 2-point [50] B. Dittrich, “Partial and complete observables for function”, Int. J. Theor. Phys. 50 (2011) 2819–2841, canonical general relativity”, Class. Quant. Grav. 23 arXiv:1101.4294. (2006) 6155–6184, arXiv:gr-qc/0507106. [31] J. Ben Geloun, “On the finite amplitudes for open [51] A. Baratin, B. Dittrich, D. Oriti, and J. Tambornino, graphs in Abelian dynamical colored Boulatov–Ooguri “Non-commutative flux representation for loop models”, J. Phys. A46 (2013) 402002, quantum gravity”, Class. Quant. Grav. 28 (2011) arXiv:1307.8299. 175011, arXiv:1004.3450. [32] R. Gurau, “Colored Group Field Theory”, Commun. [52] S. Gielen and L. Sindoni, “Quantum cosmology from Math. Phys. 304 (2011) 69–93, arXiv:0907.2582. group field theory condensates: a review”, [33] R. Gurau, “Lost in Translation: Topological arXiv:1602.08104. Singularities in Group Field Theory”, Class. Quant. [53] J. Ben Geloun, R. Martini, and D. Oriti, “Functional Grav. 27 (2010) 235023, arXiv:1006.0714. Renormalization Group analysis of a Tensorial Group [34] D. V. Boulatov, “A Model of three-dimensional lattice Field Theory on R3”, Europhys. Lett. 112 (2015), gravity”, Mod. Phys. Lett. A7 (1992) 1629–1646, no. 3, 31001, arXiv:1508.01855. arXiv:hep-th/9202074. [54] J. Ben Geloun, R. Martini, and D. Oriti, “Functional [35] V. Bonzom, R. Gurau, and V. Rivasseau, “Random Renormalisation Group analysis of Tensorial Group tensor models in the large N limit: Uncoloring the Field Theories on Rd”, Phys. Rev. D94 (2016), no. 2, colored tensor models”, Phys. Rev. D85 (2012) 024017, arXiv:1601.08211. 084037, arXiv:1202.3637. [55] J. Ben Geloun and V. Rivasseau, “A Renormalizable [36] D. Benedetti, J. Ben Geloun, and D. Oriti, 4-Dimensional Tensor Field Theory”, Commun. Math. “Functional Renormalisation Group Approach for Phys. 318 (2013) 69–109, arXiv:1111.4997. Tensorial Group Field Theory: a Rank-3 Model”, [56] S. Carrozza, D. Oriti, and V. Rivasseau, JHEP 03 (2015) 084, arXiv:1411.3180. “Renormalization of Tensorial Group Field Theories: [37] L. Freidel, K. Krasnov, and R. Puzio, “BF description Abelian U(1) Models in Four Dimensions”, Commun. of higher dimensional gravity theories”, Adv. Theor. Math. Phys. 327 (2014) 603–641, arXiv:1207.6734. Math. Phys. 3 (1999) 1289–1324, [57] J. Ben Geloun, “Renormalizable Tensor Field arXiv:hep-th/9901069. Theories”, in “18th International Congress on [38] J. W. Barrett and L. Crane, “Relativistic spin Mathematical Physics (ICMP2015) Santiago de Chile, networks and quantum gravity”, J. Math. Phys. 39 Chile, July 27-August 1, 2015”. 2016. (1998) 3296–3302, arXiv:gr-qc/9709028. arXiv:1601.08213. [39] P. J. Olver, “Applications of lie groups to differential equations”, Springer Science & Business Media, 2000. [40] S. Lie, “Theorie der transformationsgruppen i”, Mathematische Annalen 16 (1880), no. 4, 441–528. [41] E. Noether, “Invariant variation problems”, Transport Theory and Statistical Physics 1 (1971) 186–207, physics/0503066. [42] W. Fushchich and V. V. Kornyak, “Computer algebra application for determining lie and lie-bäcklund symmetries of differential equations”, Journal of Symbolic Computation 7 (1989), no. 6, 611–619. [43] S. J. Akhtarshenas, “Differential geometry on SU(N):