Arxiv:2011.00233V2 [Gr-Qc] 6 Apr 2021

Refik Mansuroglu ∗ and Hanno Sahlmann † Institute for Quantum Gravity, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Staudtstraße 7, 91058 Erlangen, Germany (Dated: April 7, 2021) We define and study kinematical observables involving fermion spin, such as the total spin of a collection of particles, in loop quantum gravity. Due to the requirement of gauge invariance, the relevant quantum states contain strong entanglement between gravity and fermionic degrees of freedom. Interestingly we find that properties and spectra of the observables are nevertheless similar to their counterparts from quantum mechanics. However, there are also new effects. Due to the entanglement between gravity and fermionic degrees of freedom, alignment of quantum spins has consequences for quantized geometry. We sketch a particular effect of this kind that may in principle be observable.

I. INTRODUCTION contained in spatial regions. Depending on the quantum state, this can be the spin of a single fermion, or the to- Spin is a quintessential quantum property but at the tal spin of multi-fermion configurations. Specifically, we same time a bit of geometric information – a direction. define operators measuring the squared total spin, and Indeed, spins were at the heart of Penrose’s spin net- components of the total spin in the direction of another works [22, 23], invented as hypothetical quantum states spin, or the component normal to a surface. of spatial geometry. This was indeed prophetic, as in loop To add spins, we need to transport them through quantum gravity (for reviews see [2, 29]), spin networks space. We use the Ashtekar-Barbero connection for this. turn out to describe a basis of eigenstates of spatial ge- Consequently the spin operators modify fermion and ometry [4]. With the present work, we will, in a sense, gravitational states in tandem. Somewhat surprisingly, close a circle, as we will investigate the quantum theory of algebra and spectra have much in common with the sit- particle spins in loop quantum gravity. It will turn out uation in quantum mechanics. However, there are also that one can regard spin networks as flows of physical new effects. For example we show that alignment of spins spin with some justification. in a region can lead to a change of its surface area, which The quantum theory of spinor fields coupled to loop could in principle be observable. It also shows that one quantum gravity is by now fairly well understood [5, 7, 8, can regard spin networks as flows of physical spin with 17, 20, 21, 27, 28]. In particular, we will follow Thiemann some justification. [27] in adopting a discretisation of the classical symplec- When we speak of spin observables, we have to be care- tic structure that is extremely natural in the context of ful, however. In the present work we will not be con- loop quantum gravity. As a consequence, the space of cerned with implementing diffeomorphism and Hamilton fermion states can be described as a fermionic Fock space constraint which encode general covariance. In particular of particle-like excitations. the spin observable that we consider are not observables Loop quantum gravity uses a formulation of general in the sense of Dirac, and it is in fact not clear what relativity as a constrained gauge theory, with a Gauß ultimate significance they have. In this way, they are constraint encoding SU(2) gauge invariance (invariance similar to the geometric operators for area, volume and under spatial frame rotations). It was realised early on length. It should be fairly straightforward to generalise [5, 20, 21] that, to solve the Gauß constraint, gravity and some of the results of the present work to the spatially fermionic excitations have to be coupled. A very com- diffeomorphism invariant level. Moreover, if one consid- arXiv:2011.00233v2 [gr-qc] 6 Apr 2021 pelling solution was first suggested in [20, 21] and later ers material reference systems as described in [11, 12], expanded on in [5]: the fermions sit at the open ends of we expect that the operators we describe correspond to gravitational spin networks. More recently, the coupling physical quantities. of fermion states to gravity has been investigated also An observation that is crucial for the construction of from the perspective of spin foam models for loop quan- the spin observables is that the spatial frame provides tum gravity [6, 13]. The coupled gravity-fermion states an isomorphism between tangent spaces and the inter- we are considering here are precisely the boundary states nal space. This means that in principle we can freely in the spinfoam formalism. move between spins as tangent vectors and spins as in- In the present work, we set out to define and inves- ternal vectors, and scalar products can be evaluated in tigate observables that measure the total fermion spin tangent space using the metric or internally. In prac- tice, the latter picture is the only practical one in loop quantum gravity, as no operator for the spatial metric is known, and neither are any other operators for tensorial ∗ Refi[email protected] quantities. † [email protected] 2

Throughout the paper, we will follow a number of con- The non-vanishing anti-Poisson relations for the Weyl ventions. Spacetime is described by a smooth manifold spinors are then induced by the anti-Poisson relations M foliated into 3-dimensional hypersurfaces M = R × Σ. of Ψ. They read For the metric signature we use the “mostly plus” con- A A (3) vention η = diag(−, +, +, +). We use capital latin let- θ (x), πθB(y) + = δB δ (x, y), (II.7) ters I, J, . . . = 0, 1, 2, 3 for indices of tensor fields in the 4-dimensional internal Minkowski space and lower case and similar for ν. In the following, we will focus on one latin letters i, j, . . . = 1, 2, 3 for indices of tensor fields Weyl component and its momentum (θ, πθ) only. How- in the spatial internal space, which are pulled with the ever, everything works analogously for the other chiral Euclidean metric δ. Tensor fields on M carry indices de- component. scribed by lower case greek letters µ, ν, . . . = 0, 1, 2, 3, The action (II.1) yields contributions to Gauss, dif- tensor fields on Σ carry indices described by lower case feomorphism and Hamilton constraint [7, 28]. For our latin letters a, b, . . . = 1, 2, 3. Finally, Weyl spinors purposes, only the Gauss constraint and spin 1 representations of SU(2) carry indices de- 2 G (x) = D(A)Ea(x) + θ†(x)σ θ(x) + ν†(x)σ ν(x) scribed by capital latin letters A, B, . . . = 1, 2. We de- i a i i i (II.8) fine holonomies h such that they transform under gauge e will be relevant. The first term of (II.8) is the SU(2) transformations as Gauss constraint known from matter-free loop quantum −1 he 7→ g(s(e)) he g (t(e)) (I.1) gravity. The second term is the SU(2) current of the half-density Weyl spinors, which generates SU(2) trans- where t(e) is the endpoint and s(e) the starting point of formations. the edge e. The Hilbert space of gravitational degrees of freedom is constructed the same way as in the matter-free theory. The gravitational observables hence act on cylin- II. A QUICK REVIEW ON LOOP QUANTUM drical functions which form the Ashtekar-Lewandowski GRAVITY WITH FERMIONS Hilbert space,

1 2 On classical level, a theory of spin 2 fermions coupled HAL = L (Σ, dµAL), (II.9) to gravity is described by the Einsten-Cartan-Holst action [7, 17, 28] together with the covariant version of the via multiplication of holonomies and the interior product Dirac action with the derivation XS, respectively [1] Z 1 4 µ ν IJ IJ πj(h)eΨ[A] = πj(h)e[A]Ψ[A], (II.10) S[e, ω, Ψ] = dx | det e|eI eJ P KLFµν (ω) 16πG M Z Z i i 4 I µ EiΨ[A] = i(XSΨ)[A]. (II.11) + dx | det e| Ψγ eI ∇µΨ − c.c. . S 2 M (II.1) The Weyl fields on the other hand are quantized as fermionic creation and annihilation operators, A foliation into three-dimensional hypersurfaces M = × Σ is performed, and the canonical analysis yields A †A R θ (x) = c (x), −iπθB(y) = cA(y) (II.12) the canonical gravitational variables [7, 28], i i i i k l which satisfy canonical anticommutation relations, Aa(x) = Γa + βKk + 2πGβ kleaJ (II.2) 1 †B A a abc i j cA(x), c (y) + = δx,y δB, (II.13) Ei (x) = ijk eaeb(x), (II.3) 2 †A †B c (x), c (y) + = 0, [cA(x), cB(y)]+ = 0. (II.14) which is the analogue of the Ashtekar connection and its conjugate momentum. In (II.2), Γ is the torsion free spin The action of this operator spans the fermionic Fock connection, K the torsion free extrinsic curvature and J l space over a one particle space given by are the spatial components of the fermion current, h = {f :Σ −→ 2 : f(x) 6= 0 only for finitely many x} l l C J = Ψγ Ψ. (II.4) X hf | f 0i = f(x)f 0(x) (II.15) Note that, in contrast to matter-free loop quantum grav- x∈Σ ity, the Ashtekar connection carries non-zero torsion. In the chiral basis, we can split the Dirac fermion Ψ and We note the change between (II.7) and its conjugate momentum into its chiral components and θ(x)A, π (y) = iδAδ , (II.16) define the half-densities [28], θB + B x,y

A p4 † θ (x) = det q ΨR(x), πθ(x) = −iθ (x) (II.5) in that a Dirac delta has become a Kronecker delta. This means that, following [27], and in contrast to [4] we quan- p4 † νA(x) = det q ΨL(x), πν (x) = −iν (x) (II.6) tise a modiﬁed symplectic structure. 3

The combined system of matter ﬁelds and gravitational The two terms correspond to orbital angular momentum degrees of freedom is given by the tensor product of the and spin, respectively. Fixing a reference system by spec- Ashtekar-Lewandowski Hilbert space and the fermionic ifying a timelike normal n, one can recover spatial angu- Fock space over h lar momentum as

µ µν 1 µν µ0ν0 H = HAL ⊗ F−(h), (II.17) J = (∗J) n ≡ 0 0 J n . (III.3) ν 2 µ ν ν so we extend (II.10), (II.12) in the obvious way. µ This is a spatial vector, J nµ = 0, so in adapted coordi- On the way towards a physical Hilbert space, we need nates J a, a = 1, 2, 3 is the angular momentum vector. In to implement the Gauss constraint. For this, we only re- particular, for the spin component, gard quantum states lying in the kernel of the quantum Gauss constraint operator, or equivalently states that are Z 1 Sµν = d3x Ψ(t, x) γ0 [γµ, γν ] Ψ(t, x), (III.4) invariant with respect to the unitary action Ug of gauge 4 transformations generated by the Gauss constraint oper- Z 1 ator. The action of U reads1 Sa = d3x Ψ(t, x) γ0a γa, γb Ψ(t, x), (III.5) g 8 bc −1 −1 Sµν |0i = 0. (III.6) Ugπj(he)Ug = g(s(e)) · πj(he) · g (t(e)), (II.18) −1 Ugθ(x)Ug = g(x) · θ(x), (II.19) In the chiral representation, the spin operator works out −1 −1 to Ugπθ(x)Ug = πθ(x) · g (x). (II.20) Z 1 σa 0 We finally arrive at the Hilbert space of SU(2) invariant Sa = d3x Ψ(t, x) Ψ(t, x), (III.7) 0 −σa states. One suitable basis is given by a generalisation of 2 spin network states, which also admit Weyl spinors at which is nothing but the generator, on spinor space, of the vertices of the underlying spin network graph. rotations as defined by n = (1, 0, 0, 0). In curved spacetime there are generically no isometries, and so no action of global Lorentz transformations. Thus III. FERMION SPIN OBSERVABLES there is no obvious definition of total angular momentum or spin. However, there is a close analogue to spin den- We want to define observables based on the spin of sities in (III.2), (III.7): There are generators for local fermions. To this end, we first have to clarify the defi- Lorentz transformations (i.e. local frame rotations), and nition of angular momentum and spin for fermion fields those can provide a spin bivector density. in curved spacetime. In flat space, angular momentum is To obtain Ashtekar variables, a partial gauge fixing is described by the covariant angular momentum bivector employed. A timelike unit vector nI (x) is introduced to J µν . For a point particle with position xµ and momen- µ I fix three components of the frame eI . n (x) introduces a tum pµ this is given by decomposition of the local Lorentz transformations into µν µ ν µ ν rotations and boosts, and only local rotations survive J = x p − p x , (III.1) as gauge transformations. Their generator is the Gauss constraint (II.8), and so a natural candidate for the local where here and in the following few equations we are spin density contribution from the Weyl field θ is given pulling indices with the Minkowski metric. In quantum by theory, the corresponding operators are nothing but the generators of Lorentz transformations. In quantum field 1 Si(x) = θ(x)σ θ†(x). (III.8) theory, these generators are given as spacetime integrals 2 i of angular momentum densities, and for non-scalar fields, a contribution from the intrinsic spin appears. For a It immediately follows that Dirac field Ψ(t, x), 1 Si(x), c†A(y) = δ σiA c†B(y), (III.9) Z 2 x,y B J µν = d3x Ψ(t, x)(−i)γ0× Si(x) |0i = 0, (III.10) i × [xµ, pν ] + [γµ, γν ] Ψ(t, x). (III.2) and 4 i j ij k S (x),S (y) = iδx,y k S (x). (III.11)

Note that while Si(x) is an internal vector ﬁeld, we can 1 Note that we are using the convention that the ﬁrst index of change freely between internal and tangent space using he transforms at s(e) and the second one at t(e). At the same the 3d frame e, time, we are using Thiemann’s convention for e ◦ f [29] such that he · hf = he◦f holds. a a i S (x)(x) := ei (x)S (x) (III.12) 4 and scalar products can be taken in either space,

a b i j S (x)S (x)qab(x) = S (x)S (x)δij. (III.13)

For the quantum theory, working in the internal space has advantages, as there are no states known that transform like tensors in tangent space, whereas states that t(E) transform nontrivially under gauge transformations are well studied. We will also need to transport spins from one point to the other to compare them. This raises the question how Figure 1. Graphical representation of the total spin operator to translate between transport in tangent and internal SE . Note that edges can intersect and also run on top of each space, and which connection to use for the transport. other, as long as the corresponding intertwiners are trivial. The first question is easy to answer. If B is a connection This is indicated by the virtual networks drawn next to the in tangent space and C is a connection in internal space, intersections in the figure. then the condition Definition III.1. Given a set of edges E with com- D ei ≡ ∂ ei − B c ei + C i ej = 0 (III.14) a b a b a b c a j b mon endpoint and disjoint initial points, t(e) = t(e0) and s(e) 6= t(e0) for all e, e0 ∈ E, define the total spin is equivalent to X (B)a b b (C)j SE = Se (III.19) he b ei (t(e)) = ej(s(e)) he i. (III.15) e∈E Thus, fixing a connection on one bundle will yield a com- For a graphical representation of the total spin see fig- patible connection on the other. The second question is ure 1. The total spin SE is again a quantum mechanical more subtle. The spin connection of e (or equivalently spin, the 3d Levi-Civita connection Γ) or the Ashtekar-Barbero connection A immediately come to mind, but there are h i j i ij k S ,S = i k S . (III.20) undoubtedly others. Ultimately, the connection used is E E E part of the choice of observables that we will define below. Under a gauge transformation g it transforms at the joint We will make use of the the Ashtekar-Barbero connection final point t(E) of the e ∈ E, A, as it is a well defined operator in loop quantum gravity. Then we can define the parallel transport of spin −1 j Ug SE Ug = π1(g)(t(E)) · S . (III.21) operator as E The total spin is not gauge invariant, but it can be used i −1 i j Se = π1(he ) j S (s(e)) (III.16) to construct a large class of gauge invariant observables. Invariant subspaces for the action of SE are con- where s(e) is the initial point (source) of e. Interestingly, structed as follows. Denote by E0 ⊆ E any subset of 0 0 the parallel transported spin is again a quantum theoret- E and the edges in E as E = {e1, e2, . . . e|E0|}. Define ical spin: states,

n A1...A|E0| 0 o where t(e) is the final point (target) of e. HE := span ΨE0 |A1,...,A|E0| ∈ {1, 2},E ⊆ E . (III.23) The space HE is invariant under the action of gauge A. Total Spin transformations. Since its states transform only at t(E), the irreducible subspaces of HE are effectively labeled by In analogy to the theory of angular momentum in flat irreps of SU(2), standard quantum theory, we start by defining the total nj spin of the fermion field evaluated at a number of points. M (j) (j) M (j)(k) HE = HE , HE = HE To add the spins, they all have to be transported to the j k same point. To do so, one has to chose a set of edges E (j)(k) with common endpoint and disjoint initial points. HE ' πj (III.24) 5 with nj denoting possible multiplicity (including nj = 0). spin of the neighbouring holonomy,

Note that we can also easily add two total spin operators 2j 2j and arrive at ...... (III.29) ...... i i i 2j-1 2j+1 SE∪E0 = SE + SE0 (III.25) Depending on the split of the number n = n+ + n− which is again a spin operator, if and only if the set of of fermions into increasing and decreasing spins, the 0 starting points of E and E are disjoint. Also, we have corresponding spin of the eigenstates can take values to impose t(E) = t(E0) in order to get a gauge covariant 1 n n j ∈ { 2 , ..., 2 } for odd n and j ∈ {0, ..., 2 } for even n. object. The eigenstates (III.28) can be found by induction over 2 Let us consider the action of (SE) on HE. We have n, but in this case a more direct route is possible.

2 X 2 (j)(k) (SE) = S (s(e)) + Proposition III.2. The total spin SE acts on HE e∈E like a spin in the j-representation. In particular, X 0 0−1 2 1 + S(s(e)) · he◦e · S(s(e )). (III.26) (SE) (j)(k) = j(j + 1) (III.30) HE (e,e0)∈E×E i Proof. The operator Se is just a grasping operation, it In order to understand the mixed terms, we make use A B acts on the state π 1 (he) B ⊗ θ(s(e)) |0i as of the binor calculus, a graphical notation suggested by 2 Penrose [9, 14, 22, 24]. Using this toolbox we can rewrite i A the mixed terms in the following way A e i A j k C Se e 2Sk(s(e1)) B (he) k S (s(e2)) D = e 7−→ e (III.31)

A C

= e e Then using the fact that we can move the intertwiner B D along edges, A C A C 1 e = + , (III.27) 2 B D B e D e −1 = , (III.32) where we denoted e := e1 ◦ e2 for the sake of clarity. e For technical details and a proof of this identity, see Ap- e e pendix A. Now, given an arbitrary spin network state ψ with n fermions sitting on distinct vertices of the spin network, we can calculate the action of (III.26) by writing ψ using the binor formalism and applying (III.27). In i general, however, the spin network states are not eigen- one sees that Se acts as the generator of gauge transfor- states of (III.26). Instead, we ﬁnd a subset of its eigen- mations at t(E) on HE, i.e. as an angular momentum states with eigenvalues, which are very reminiscent of ﬂat operator on quantum theory, M N |E0| (π 1 ) . (III.33) 2 E0 Decomposing into irreps completes the proof. e Note that the action on any state Ψ in H that has 2 (SE ) no fermionic excitations at the positions s(e), e ∈ E is −−−−→ j(j + 1) trivial, more precisely,

SEΨ = 0. (III.34)

More generally, for {x1, . . . , xm} ∩ {s(e)|e ∈ E} = ∅, 2j consider a gauge invariant operator F built from ar- (III.28) bitrary holonomies and the fermion creation operators θ(x ) . . . θ(x ), i.e., Here, the empty circle represents an arbitrary combina- 1 m B1 Bm tion of fermions adding or respectively subtracting the F = FB1···Bm [A] ⊗ θ (x1) ··· θ (xm). (III.35) 6

state or vice versa, there is a non-trivial basis change in general. The state (III.38), for instance, is a non-trivial linear combination of spin network states and particu- larly not proportional to the spin network state depicted in Figure 2. In fact, we can follow the same procedure also with an arbitrary number of fermions n and two points of outflow. For suitable coefficients λab, ..., λab...cd we can show Figure 2. Sketch of a Spin Network State (blue) consisting that the generalised eigenstates of the squared total spin of two fermions and a holonomy inbetween. Moreover, the (S )2 to the spin j + k have the following form two fermion system admits two points of outflow. The spin E operator SE (black) measures the parallel transported spin of the fermions at the joint endpoint, which coincides with one of the points of outflow in this case.

Then evidently

[SE,F ] = 0. (III.36) Therefore

(j)(k) Corollary III.3. The total spin SE acts on FHE like a spin in the j-representation, where F is any operator constructed as above. In particular,

2 1 (SE) (j)(k) = j(j + 1) (III.37) F HE Corollary III.3 provides for a large number of irreps for SE and consequently a large number of eigenstates for 2 SE. We note however, that there are more eigenstates than those listed already. The states of corollary III.3 have one point in which spin flows out. The idea to construct more general eigenstates is to consider multiple such points of outflow. Let us consider an example. Example III.4. For two fermions and two points of outflowing spin, there is the eigenstate,

1 + e e 2

 

2 (SE )  1  −−−−→ 2  +  , (III.38)  2  which has spin 1. We notice that the first term corre- sponds to the new type of states while the second term is a special case of (III.28) with spin j = 0. The second term arises from the first by connecting each, the two fermions and the two points of outflowing spin, by the same holonomy he. In order to get an eigenstate, taking this linear combination is necessary, as the mixed terms (III.27) couple the disconnected fermions. We have to be aware of the difference between the spin network and the binor representation. Even if it is some- times tempting to read a binor term as a spin network 7

2k 2k-1

X e f + λab + ... + a

2j 2j-1

b d X e e f g + λab...cd (III.39) a

a c

|2j-2k| 2k

where we numbered the fermions by the labels a, b, ..., c, d. gives rise to a more general set of eigenstates. In princi- Because of (III.36) we can even act on it with suitable ple, we can continue adding points of outflow. However, operators F to create further eigenvectors to the same we would only increase the combinatorial complexity of eigenvalue. In (III.39), we assumed that there are no the eigenstates as we can now reduce the outflowing spin pure holonomies between the points of outflowing spin. by connecting any two of the endpoints and two fermions From (III.39) it becomes apparent that the last term is to another using suitable holonomies. Although the num- always of the form (III.28), and the sequence terminates. ber of coefficients grows exponentially, the procedure will 2 By acting on that state with SE we can read off the be exactly the same as in the case of two points of out- coefficients if we require that it is an eigenstate to the flow. eigenvalue j + k. Proposition III.5. The states of the form (III.39) are 2 eigenstates of SE to the eigenvalue j+k if the coefficients are defined by the following recursion relation

λ−1 = λ−1 := m(2j + 2k − m + 1)λ−1 a1b1...ambm m m−1 λ0 = 1, (III.40) where m denotes the number of reductions made. In addition λa1b1...ambm = 0 if at least one pair (ai, bi) of fermions is of the same type (both increasing or reducing the spin from top to bottom). Proof. See Appendix B.

The eigenstates (III.39) can be interpreted as terms Figure 3. Sketch of a generic spin network state (blue) with contributing to a spin network state of the form depicted four fermions and two points of outﬂow. The spin operator SE in Figure 3. Note again that switching from the binor to (black) measures the transported spin of the fermions at the the spin network representation or vice versa in general joint endpoint t(E), which represents one of the two points of requires a non-trivial basis change. Proposoition III.5 outﬂow. 8

In fact, it should be possible to write any spin network where Eˆ is the smeared canonical momentum operator state as a linear combination of states with n fermions of the Ashtekar connection Aˆ and (2)q is the pullback of and up to n points of outﬂow multiplied by a spin network the spatial metric q to the two-dimensional surface S. state with only purely gravitational degrees of freedom. The vector ni is related to the spatial surface normal by i This becomes apparent when writing any spin network na = eani ⊥ S. The operator (III.42) is a measure of the state in the binor formalism and resolving the symmetri- area of the surface S. This motivates the deﬁnition of a sations. This way, we have sketched the process of how new observable to characterise the full spectrum of the squared total spin 2 operators SE.

B. Projection onto a Surface Normal Definition III.6. Let E be a set of edges with a joint Another observable inspired by flat standard quantum endpoint and S a surface that intersects the spin network theory is the projection of the total spin onto some direc- graph defined by E exactly once in t(E). We define the z projection of the total spin onto the surface normal by tion, for example SE. The obvious problem is that this operator is not gauge invariant and hence not a good observable. Instead, we might project the total spin onto a reference field (cf. magnetic field in a Stern-Gerlach experiment). One such reference field may be given by S = Si Eˆ ( ). (III.43) the normal vector field of a surface S S E i S

a b c ni := Ei abct1t2, (III.41) where t1, t2 are tangent vectors to a surface S. To get rid of the density weight, we deﬁne the smeared operator

From the fact that Eî couples also on pure gravita- Z Z tional spin network functions, one can see that it will not ˆ â b c â 2 Ei(S) = dSaEi = dx ∧ dx abcEi have the same eigenbasis as SE. This can be also proven S S by explicitly calculating the commutator. Already for Z q := n det (2)q, (III.42) n = 2 fermion spins, we get a non-trivial contribution of i ˆ S the commutator of he and Eî,

m S2 ,S = Sl(e (0)) h−1 S (e (0))Sk(e (0)) E S 1 e1 l j 1 2 j l −1m k τm −1 + S (e2(0)) he Sj(e1(0))S (e2(0)) he1 he 6= 0. (III.44) 2 l 2 2 k

On the other hand, the two operators in (III.43) do com- inserted at the intersection point of S with e and at the mute. point where the respective fermion is located. The factor is given by the defining relation, which is known from the h i ˆ i X i −1 j k SE, Ei(S) = τi j π1(he ) kS(s(e)) = 0, (III.45) area operator in loop quantum gravity [26], e∈E i 2 2 i 2 2 EîEˆ he = − 8πl γ he τiτ he = 8πl γ j(j + 1)he, which is a result of the total antisymmetry of τ. As p 1 2 p (III.46) a consequence, the self-adjointness of (III.43) is ensured and we do not have to think about the order of the two with e = e1 ◦ e2 being split at the intersection point operators. with S. Note that we assume now and for the following Although there is no hope to get a spin component with discussion that the edge e intersects the surface S exactly exactly the same properties as Sz with (III.43), we are once and transversely, i.e. it punctures the surface non- able to understand some special cases of spin coupling. tangentially. Furthermore, the intersection point equals As Eî acts with the insertion of a Pauli matrix τi, we neither e(0), nor e(1). find that every addend of (III.43) acts very similar to In order for (III.46) to hold, we have to multiply the (III.27). In particular, the binor representation is, up to binor representation (III.27) with the spin j of the holon- 2 a multiplicative factor, identical with the operator being omy he, which intersects S and a constant factor 8πlpγ . 9

Consider now two fermions of the form (III.28) with where we again omitted the ”e” label in the calculation. the spin j = 0. This kind of state involving two fermions (III.47) is straight-forwardly calculated using the basic was first considered in [21]. If we admit an intersection of techniques of binor calculus. This result would be ex- S with an arbitrary point on the edge, we can calculate pected for a component of a vanishing spin projection. the action of the spin projection, More generally, we can also consider the product of the singlet (III.47) with a purely gravitational spin network.   Here, we will use the Leibnitz rule of Eî and the term   where S acts on the singlet vanishes. The latter is van-   S SS 1   ishes by (III.47) and the first term reads −→  −  = 0, e 2      

(III.47)

    SS 1   f · −→  f + f  = 0. (III.48) 2     2j 2j 2j

This can be calculated by dissolving the symmetrisa- flat quantum theory, but now the quantisation axis is tions, pulling two holonomies across each other and con- defined by the orthonormal of the surface S. Indeed, if sequently using the binor identity. Note that we assumed we flip the sign of the surface normal, also the signs of the edge f to intersect the surface S at the same point (III.49) and (III.50) flip, because the operator Eˆ includes as e. Apart from that, f is arbitrary. If we allow f to in- the sign of the inner product of the surface normal and tersect S at any arbitrary point, the projection operator tangent vector of e at the intersection point.

SS will also map the state to zero and the calculation is Obviously, the operator Eî influences the actual eigen- similar to (III.47). The freedom to multiply any edge f 1 value, as it measures the spin j + 2 of the holonomies here is a result of the symmetry of the state. In the fol- he, whose edges intersect the surface S. If we were to lowing however, we have to stick to f = e. Analogously, normalise the surface normal in the definition of the pro- one can show that also the spin 1 states are eigenstates jection operator (III.43) such that of (III.43). ˆ∗ î tr Ei E = 1, (III.51)

SS 1 2 i 1 e · e −→− (j + 1) · we will gather an additional factor √ for Eˆ ∝ τi 2 j(j+1) 2 in the spin j representation2. Since in (III.47 - III.50), 2j 2j 1 we are acting on a spin j + 2 holonomy, the eigenvalues (III.49) would take the form ± j+1 . If we take the limit q 1 3 (j+ 2 )(j+ 2 ) j → ∞, these reproduce the spin projection ±1, which we would expect for the triplet state in ﬂat quantum theory. Indeed, in the literature [18, 19] the limit of large spin is 1 · −→SS (j + 1) · often regarded as a semiclassical limit of loop quantum e 2 gravity. The third spin 1 state (III.38) completes the 2j 2j discussion of the spin projection operator (III.43) on a

. system of two fermions. One can show in an analogous (III.50) Since it is always the same techniques used in the cal- culations, we refrain from writing out all the steps. The structure of the eigenvalues is reminiscent of the projec- 2 We would arrive at the same result, if we would deﬁne the nor- tion of the ”up-up” and ”down-down” triplet states from malised operator by dividing SS by the surface area AS. 10 manner that C. Projection onto the Spin of a Speciﬁc Fermion

In this section, we want to go away from a manifestly gravitational reference frame and draw attention to an- i S other observable measuring the projection of S onto the −→S 0. (III.52) E e spin of one specific fermion of the fermion system. Definition III.8. In the case of n fermions connected by a set of edges E0 with a joint endpoint t(E0) and a 2j st 0 n + 1 fermion forming E = E ∪ {en+1}, we define the holds. With (III.47) and (III.49) we now understand both observable terms in (III.38) in the context of the spin projection. i 2S 0 · S(e (0)) = 2S 0 S . (III.55) Hence, we find that the third triplet state besides (III.49) E n+1 E {en+1} i and (III.50) is represented by (III.38), which vanishes Note that we exclude the self-projection S2 in {en+1} under the action of SS. Moreover, we have found one 0 singlet state (III.47) in complete analogy to flat quantum (III.55) by projection onto the spin defined by E in- theory. stead of E. Again, the addends of the observable (III.55) Note that it is possible to generalise the above state- have the same action as (III.27) with being inserted at ment to the coupling of n fermions in principle. However, the points where the respective fermions are located. Let we cannot change the purely gravitational spin network us have a look at the action of (III.55) on the eigenstates of S2 discussed in (III.28). state to be different from a sole holonomy he, otherwise E it will not be an eigenstate of (III.43). One counter ex- 2 Proposition III.9. The eigenstates (III.28) of SE are ample is also eigenstates of 2SE0 · S(en+1(0)) to the eigenvalue Example III.7. Consider the two fermion state with an j, the spin of the state corresponding to the eigenvalue 2 intertwiner inbetween j(j + 1) of SE, i.e.

2SE0 ·S(en+1(0)) = 1/2 (III.53) ψ · −−−−−−−−−−−→ j · ψ · f

1 ,

2j n+1 n+1 (III.56) where the right hand side represents the spin network of the state which is written in the binor formalism on the If the n + 1st fermion is of the type which reduces the left. We only inserted one non-trivial intertwiner. When total spin of the n fermion system, the eigenvalue reads applying the same techniques as above, we end up with a −(j + 1), i.e. diﬀerent state

1 g 2SE0 ·S(en+1(0)) −→SS ψ · −−−−−−−−−−−→− (j + 1) ψ · 2 e

2j-1 , n+1 n+1 (III.54) (III.57) where we constructed the counter example such that the Here, ψ stands for an arbitrary spin network state ad- intersection point of S with the spin network graph lies mitting gravitational degrees of freedom only. −1 on the edge f and e = f ◦ g . The +1 in the negative eigenvalue seems odd at ﬁrst, We showed that the projection (III.43) can in principle but is consistent with the fact that the mixed term of the 1 mimic the observable SS in very special cases. In general, total spin operator for the singlet state (case j = 2 ) needs ˆ 3 however, there is a non-trivial interaction between Ei a contribution of − 2 in order to yield the total spin 0. As (and, in consequence SS) and gravity. As a consequence, a result, we found a more general subclass of eigenstates 2 SE and SS do not have the same eigenbasis. of a projection operator compared to section III B. 11

2 While the squared total spin SE is well understood, we are having trouble finding a well behaving counterpart for the spin projection Sz. Still, first attempts have brought reminiscent structures in special cases. Anyway, we draw an important conclusion from the above analysis, namely that fermionic systems influ- j j + k ence the spin quantum numbers of the neighbouring holonomies in a way that should be principally measurable. In the following section, we want to sketch a mea- surement which makes this statement more specific.

IV. THE INFLUENCE OF FERMION SPIN ON ( n THE GEOMETRY OF SPACETIME 0, ..., 2 for n even with k ∈ 1 n (IV.1) 2 , ..., 2 for n odd So far, we learned that fermionic spin degrees of freedom influence the spin quantum numbers of the neigh- Figure 4. Sketch of the lateral cross section of a three- bouring holonomies, which are in turn a measure of sur- dimensional cube B containing n fermions. Two egdes are face area [26]. This interaction between geometry and intersecting the surface area of B. Dependent on the cou- matter should in principle be observable by measuring pling of the fermion spins, the spin quantum numbers of the the area of two surfaces in the vicinity of the fermion. holonomies intersecting the surface of B may differ by k, 2 which is bounded by 0 and n . This is reminiscent of the As a quantum of area is of the order of lp and hence 2 small compared to measurable scales of area, we want to total spin of the Ising model in flat spacetime. sketch a macroscopic version of this experiment. There- fore, consider a cube B ⊂ Σ with a number of n fermions each being attached to a single vertex of the spin network. difference reads We will argue later, however, that the details of the ar- 2 p p rangement of the fermions are not so important for the ∆A = 8πlpγ (j + k)(j + k + 1) − j(j + 1) effect we are discussing. 2 ≈ 8πlpγ k for j 1. (IV.3) If there were no fermions, the Gauss law would con- strain the spins flowing out of B to be able to couple to Hence, the maximal difference is proportional to the zero. With fermions being placed inside B, the outflow- number of fermions n. Considering a cubic centimetre ing spins have to be able to couple to the total spin of 23 1 with an electron density of around 10 cm3 (average solid the n fermion system, which we can now measure with state), a rough dimensional analysis yields an approxi- 2 SE. The simplest example of this scenario is described mate maximal area difference of by two outflowing spins shown in Figure 4, which depicts the lateral cross section of the cube. ∆A ≈ γ · 10−42cm2, (IV.4) In general, the quantum state corresponding to Fig- ure 4 can be described by a linear combination of binor being significantly larger than the Planck length squared 2 calculus states like (III.28). The value k can be tuned by lp, but still small compared to area scales of well known aligning neighbouring spins anti-paralelly for k minimal physics. If we assume a homogeneous growth of area, we or paralelly for k maximal, respectively. This scenario can trace it back to the error of length scale is a simple approach to mimic antiferromagnetism (or 2 2 2 ferromagnetism respectively) in loop quantum gravity3. A0 + ∆A = (l + ) = l + 2l + O( ) (IV.5) ∆A Using the well understood action of the area operator on =⇒ ≈ (IV.6) holonomies [26] 2l q where l is the length of the square at total spin 0 and ˆ ˆ î 2 −9 A(S) πj(he) = EiE πj(he) = A0 = l . Only for l = 10 cm, enters the regime = 8πl2γ pj(j + 1) π (h ), (IV.2) of Planck length and would be even smaller than lp for p j e larger l. Although the large number of fermions first ap- we notice an increase of area between the areas of the peared to make the effect large, we end up with an effect two surfaces on the left and on the right of Figure 4. The in length of the order of the Planck length. This comes from the fact that it is apparently harder to measure a 2 quantum of area ∼ lp than a quantum of length ∼ lp. The area change (IV.4) can be additionally increased 3 In principle, k can also take negative values. However, as the by increasing the electron density or starting with a problem is symmetric, negative k values can be equivalently de- longer bar instead of a cube in the first place. Also a scribed by flipping the outflowing spins. more efficient method to measure area (without tracing 12

V. SUMMARY AND OUTLOOK

We have introduced a total spin operator SE for a set of points connected by a set of paths E to a reference point t(E). The parallel transport of each fermionic spin operator to t(E) ensures that we can add all the diﬀerent single particle spin operators to a gauge covariant total spin operator.

Building on SE we further introduced a number of observables describing the squared total spin and various projections of SE. We showed that, in general, spin network states are not eigenstates of these operators. In- stead, it turned out to be convenient to work with the binor formalism. Within this framework, we were able to describe the eigenstates of the squared total spin operator in some detail and hence define a notion of fermionic spin coupling in loop quantum gravity. ES is algebraically a Figure 5. Sketch of the lateral cross section of a compact spin, hence the spectrum of this operator is the standard subset B ⊂ Σ of the spatial hypersurface, which is punctured one. by many holonomies. The total spin flowing out of (or into) B is a measure of the surface area increase compared to the Although we were not able to define a spin projection surface area of B without matter. that commutes with the squared total spin, we introduced two possible observables which mimic the action of Sˆz from quantum mechanics, using on the one hand a fixed surface S and on the other hand the spin of one spe- it back to length) might improve the problem of spa- cific fermion as a reference. A more accurate imitation of tial resolution discussed above. In this way it might in a Stern-Gerlach type observable would be the coupling of principle be possible to experimentally measure an upper the spin operator with an electromagnetic field. The for- bound to the Barbero-Immirzi parameter γ. mulation of loop quantum gravity with charged fermions is in principle available [27], a detailed treatment is still Indeed, this is consistent with the fact that the work in progress [15], however. Barbero-Immirzi parameter in loop quantum gravity is no more arbitrary in the presence of fermions [25]. It At last, we used the above results to sketch an exper- might still be difficult to make a statement about the or- imental setup describing an observable effect as a result der of magnitude of γ. While the Immirzi parameter is of the entanglement between fermion spin and geometry. estimated to be of order of magnitude 1 in order to re- In the example considered, a closed surface B containing produce the Bekenstein-Hawking formula in the theory of n fermions would experience a change in area linear in black hole entropy [10, 16], it is treated as a large quan- n when the spins of the fermions inside B were aligned. tity γ 1 in order to justify the perturbative calculation We note again, that neither area nor spin are Dirac ob- of time evolution [3]. servables, and thus the picture could change in a more complete approach. Note that the above analysis only represents a model which would be more accurate if we would allow more Indeed, it would be very interesting to take diffeomor- than only two points of outflow. This includes states of phism and Hamiltonian constraint into account to under- the form (III.39), which are also eigenstates of the total stand the dynamics of the kinematical states discussed in spin squared. Here, however, the area increase is not this paper. To do this, a reduced phase space quantisa- directed like in (IV.1) but might flow out of the cube in tion could be carried out [11, 12]. every direction. Still the minimal and maximal increase Also, to understand the connection between area and n 2 are again limited by 0 and 2 . total spin better, the eigenstates of SE and the action of the area operator on these are to be understood more More generally, in order to predict the effect it is suffi- precisely. Since the two observables should be measured cient to require B to be a connected and compact subset simultaneously, their commutator has to be investigated. of Σ not necessarily diffeomorphic to a cube. The more general statement when embedding a spin network state We finally note that we chose the Ashtekar-Barbero with n fermions into B is that the surface area of B will connection to parallely transport spins. Other choices of sense the same area increase (IV.3) when coupling the parallel transport are clearly possible in principle. How fermion spins completely parallel compared to the case they could be realised within loop quantum gravity, and of total spin 0. The most general setup is sketched in how this would change the picture could also be investi- Figure 5. gated in the future. 13

ACKNOWLEDGMENTS box with as many ingoing edges as outgoing ones,

We thank Kristina Giesel and Thomas Thiemann for interesting discussions during the completion of this work. Moreover, we thank Thomas Thiemann for very e e e = πj(he). (A.3) detailed and valuable feedback.

R.M. thanks the Elite Graduate Programme of the Friedrich-Alexander-Universit¨at Erlangen-N¨urnberg for 2j its support and for encouraging discussions among its The last element from our toolbox which enables us to members and faculty. H.S. would like to acknowledge mimick spin network states in the binor formalism is the the contribution of the COST Action CA18108. notation of the intertwining operators,

2k 2i

2b 2c = ι : i ⊗ j ⊗ k → 0, (A.4)

N Appendix A: SPIN COUPLING AND THE BINOR where a, b, c ∈ 0 are determined by FORMALISM 2 1 a = (i + k − j) (A.5) 2 1 It turns out that within the spectral analysis of the b = (j + k − i) (A.6) spin operator (III.26), it is convenient to leave the spin 2 1 network basis of loop quantum gravity and describe the c = (i + j − k). (A.7) states of H using the binor calculus [9, 14, 22, 24]. In 2 order to ﬁx the notation, we recall the main deﬁnitions As higher valent intertwiners can always be expanded of the binor-formalism: into a basis of 3-valent intertwiners, (A.4) completes the toolbox of the binor formalism. At last, we want to recall the binor identity,

A A C AC A C + = − ⇐⇒ δBδD − BD = δDδB , A A B AB = δB = iAB = i A B B (A.8) (A.1) which is used to dissolve crossings within the binor calculus. We can use the binor identity, to show the binor A B A A representation of (III.27). We start with the identity, A B A A 1 = −δDδC e = π (he) B = θi A C A C 2 A C i 1 1 C D B σ A σiC = = + (A.2) 2 i B D 2 B D B D B D (A.9) On the other hand, using the intertwining property of For the sake of clarity, we deviate from the standard no- the Pauli matrices, 1 tation by denoting a spin representation of the holon- D 2 i 1 iA B C −1 omy he as a straight line with label ”e”. Moreover, we π1(he) j = σ B π 1 C σj D π 1 (A.10) 2 2 2 A draw a spinor as a star. Higher spin j representations can be constructed by the symmetrization of 2j many and substituting π1(he) in the deﬁnition (III.27), we ﬁ- 1 spin 2 holonomies. We denote the symmetrization by a nally deduce 14

A C H ˆ j ˆk 1 A jE F G −1 kC 2 J1j π1 (he) k J2 = σj B σ F π 1 Gσk H π 1 σ D B D 4 2 2 E

A C B A C 1 e = − = + e e 2 C B D B e D

A C = e (A.11) e B D

1 −1 which shows (III.27). The operator (A.11) acts on the λ1 = 2 , or equivalently λ1 = 2, which is in agreement fermions 1 and 2 by inserting a Pauli matrix at the point with (III.40). where the fermion is located. With this operator, we are 2 able to calculate the action of (SE) on an arbitrary state in H. Induction step

We assume that the proposition holds for a fixed number of fermions n. If we now consider a state with n + 1 Appendix B: PROOF OF PROPOSITION III.5 fermions, there are four possible cases, we have to distin- guish. On the one hand, we can lower or raise the total We proceed with a proof of induction. spin of the state and on the other hand we can either raise (or lower) j or k. However, the problem is highly symmetric such that we can focus on the case where j is raised by the n + 1st fermion and refer to the other cases Induction start to be treated in analogue manner. Moreover, we will reduce the problem further by dis- We will consider example III.4 for the start. Here, we solving all symmetrisations (except at the points of 1 have j = k = 2 and two fermions of different types. outflow) and operating on a product of fermions and Apart from λ0, which can always be set to 1 without holonomies only. Hence, we are left with the following loss of generality, there is only one non-trivial coefficient building blocks

... , ... , ... , (B.1)

out of which the remaining terms are constructed. We fermion is of the form omitted the labels denoting the edges of the connecting st holonomies for the sake of clarity. In our case, the n+1 (B.2) n+1. 15

If we denote the graph connecting fermion 1 to fermion action of (B.3) namely the second term in (III.27). This 0 0 st n to a joint point by E and E = E ∪ {en+1}, then we term connects the n + 1 fermion with each of the build- can split oﬀ the action of the squared total spin in the ing blocks as well as the endpoints which are left over by 1 following way a spin 2 holonomy. Let us start with the building block (B.2) which in- creases j. As the n + 1st fermion is of the same type, 2 2 2 the resulting state vanishes as the symmetrisation and S =S 0 + S (e (0)) E E n+1 an antisymmetrisation meet n ! X i j k + 2 S (el(0))δijπ1(h −1 ) S (en+1(0)). en+1◦el k l=1 (B.3) = 0. (B.4)

... If we act on the n+1 fermion state with the first operator, it will give the eigenvalue jm(jm + 1) by the induction This result is independent of which of the terms we re- hypothesis, where jm = j + k − m denotes the sum of gard. As a next step, let us have a look at the building outflowing spin, which is dependent on the number m of blocks which are factors of singlet states, reductions made. The second operator is also well known 1 1 3 and gives the eigenvalue 2 2 + 1 = 4 . We are left with the mixed terms, which connect the n + 1st fermion with 2 any of the other fermions each in a separate addend. The (B.5) binor representation of this operator is shown in (III.27). One can see that we get two terms for each of its action if we dissolve the symmetrisation in (III.27) together with 1 . using the binor identity (A.8). The first term in (III.27) 1 st acts as an identity operator with a factor 2 . From this The operator connects the n + 1 fermion with each of 3 contribution, we get, additionally to jm(jm + 1) and 4 , the two fermions 1 and 2 by a holonomy. This yields the n also a factor 2 . We will now discuss the only non-trivial state

2 1 2 7→ +

n+1 2 1 1 n + 1 n + 1

1 2 = − −

1 n + 1 2 n + 1

1 2 = = − (B.6)

n+1 n+1 2 1 ,

where we used basic manipulations of and δ and the calculation. In general, they have to be taken care of. binor identity in the second step. The labels of the edges With (B.6) we have another contribution to the eigen- are again omitted, since they are not relevant for this value, which goes linear with the number of the build- 16 ing blocks of the form (B.5). Let us denote the num- then we reduce the outﬂowing spin by 1. This reduction ber of fermions which decrease the number of holonomies yields from top to bottom by n−. Analogously, the number of fermions increasing the number of holonomies, we denote by n+ such that they sum up to the total number st 4 of fermions excluding the n + 1 fermion, 7→ (B.10)

n− + n+ = n. (B.7) n+1 n+1 , On the other hand, it holds n+ − n− = 2j − 2k. The contribution of (B.6) hence will be −(n −2k+m), which − which is equal to one of the other states corresponding to can be combined with n to 2 one more reduction m + 1. This contribution goes with n 1 a factor 1. Finally, we can collect all the pieces to ﬁnd − n + 2k − m = (n − n ) + 2k − m that the coeﬃcients in front of a state with m reductions 2 − 2 + − reads = j + k − m = j0 − m. (B.8) 3 λ + λ (j − m)(j − m + 1) + + j − m = The last building block will change the state non-trivially. m−1 m 0 0 4 0 If we connect the n + 1st fermion with a fermion of the 1 3 type =λ + λ j + j + + m2 − 2m − 2mj , m−1 m 0 2 0 2 0 (B.11) (B.9) , 1 3 which equals to (j0 + 2 )(j0 + 2 ) if and only if 1 λ = λ · m 2 j + − m + 1 . (B.12) m−1 m 0 2

This completes the proof.

[1] A. Ashtekar and J. Lewandowski. Quantum theory of ge- and the loop representation of quantum gravity. Class. ometry. 1: Area operators. Class. Quant. Grav., 14:A55– Quant. Grav., 14:53–70, 1997. A82, 1997. [10] M. Domagala and J. Lewandowski. Black hole entropy [2] A. Ashtekar and J. Lewandowski. Background indepen- from quantum geometry. Class. Quant. Grav., 21:5233– dent quantum gravity: A Status report. Class. Quant. 5244, 2004. Grav., 21:R53, 2004. [11] K. Giesel and T. Thiemann. Algebraic quantum gravity [3] M. Assanioussi, J. Lewandowski, and I. Mäkinen.Time (AQG). IV. Reduced phase space quantisation of loop evolution in deparametrized models of loop quantum quantum gravity. Class. Quant. Grav., 27:175009, 2010. gravity. Phys. Rev. D, 96(2):024043, 2017. [12] K. Giesel and T. Thiemann. Scalar Material Reference [4] J. C. Baez. Spin networks in nonperturbative quantum Systems and Loop Quantum Gravity. Class. Quant. gravity. In The Interface of Knots and Physics, pages Grav., 32:135015, 2015. 167–203, 4 1995. [13] M. Han and C. Rovelli. Spin-foam fermions: PCT [5] J. C. Baez and K. V. Krasnov. Quantization of diffeomor- symmetry, Dirac determinant and correlation functions. phism invariant theories with fermions. J. Math. Phys., Classical and Quantum Gravity, 30(7):075007, Mar 2013. 39:1251–1271, 1998. [14] L. Kauffman. Spin networks and the bracket polynomial. [6] E. Bianchi, M. Han, C. Rovelli, W. Wieland, E. Magliaro, Banach Center Publications, 42, 12 2002. and C. Perini. Spinfoam fermions. Classical and Quan- [15] R. Mansuroglu and H. Sahlmann. Kinematics of arbi- tum Gravity, 30(23):235023, Oct 2013. trary spin matter fields in loop quantum gravity, 2020. [7] M. Bojowald and R. Das. Canonical gravity with [Phys. Rev. D (to be published)]. fermions. Phys. Rev. D, 78:064009, 2008. [16] K. A. Meissner. Black hole entropy in loop quantum [8] M. Bojowald and R. Das. Fermions in Loop Quantum gravity. Class. Quant. Grav., 21:5245–5252, 2004. Cosmology and the Role of Parity. Class. Quant. Grav., [17] S. Mercuri. Fermions in Ashtekar-Barbero connections 25:195006, 2008. formalism for arbitrary values of the Immirzi parameter. [9] R. De Pietri. On the relation between the connection Phys. Rev. D, 73:084016, 2006. [18] A. Mikovic and M. Vojinovic. Effective action and semiclassical limit of spin foam models. Class. Quant. Grav., 28:225004, 2011. [19] A. R. Mikovic. Spin network wavefunction and nonper- 4 Whether we include it or not is a matter of convention. 17

turbative graviton propagator. Fortsch. Phys., 56:475, bridge Univ. Press, Cambridge, UK, 4 2011. 2008. [25] A. Perez and C. Rovelli. Physical effects of the Immirzi [20] H. A. Morales-Tecotl and C. Rovelli. Fermions in quan- parameter. Phys. Rev. D, 73:044013, 2006. tum gravity. Phys. Rev. Lett., 72:3642–3645, 1994. [26] C. Rovelli and L. Smolin. Discreteness of area and volume [21] H. A. Morales-Técotland C. Rovelli. Loop space rep- in quantum gravity. Nucl. Phys. B, 442:593–622, 1995. resentation of quantum fermions and gravity. Nuclear [Erratum: Nucl.Phys.B 456, 753–754 (1995)]. Physics B, 451(1):325–361, Feb. 1995. [27] T. Thiemann. Kinematical Hilbert spaces for Fermionic [22] R. Penrose. Angular momentum: an approach to combi- and Higgs quantum field theories. Class. Quant. Grav., natorial spacetime. In T. Bastin, editor, Quantum The- 15:1487–1512, 1998. ory and Beyond, pages 151–180, Cambridge, 1971. Cam- [28] T. Thiemann. QSD 5: Quantum gravity as the natural bridge University Press. regulator of matter quantum field theories. Class. Quant. [23] R. Penrose. On the nature of quantum geometry. In Grav., 15:1281–1314, 1998. J. Klauder, editor, Magic Without Magic, pages 333–354, [29] T. Thiemann. Modern Canonical Quantum General Rela- San Francisco, 1972. Freeman. tivity. Cambridge Monographs on Mathematical Physics. [24] R. Penrose and W. Rindler. Spinors and Space-Time. Cambridge University Press, 2007. Cambridge Monographs on Mathematical Physics. Cam-