Fermions, differential forms and doubled geometry

We show that all fermions of one generation of the Standard Model (SM) can be elegantly described by a single fixed parity (say even) inhomogeneous real-valued differential form in seven dimensions. In this formalism the full kinetic term of the SM fermionic Lagrangian is reproduced as the appropriate dimensional reduction of (Psi, D Psi) where Psi is a general even degree differential form in R^7, the inner product is as described in the main text, and D is essentially an appropriately interpreted exterior derivative operator. The new formalism is based on geometric constructions originating in the subjects of generalised geometry and double field theory.


Introduction
The gauge group of the Pati-Salam unification model [1] is SO(4) ∼ SU(2) × SU(2) times SO(6) ∼ SU(4). This last group extends the usual SU(3) of the Standard Model (SM) by interpreting the lepton charge as the fourth colour. The groups SO(4) and SO (6) are usually put together into SO(4) ×SO(6) ⊂ SO(10) unified gauge group, see e.g. [2] for a nice exposition of this standard material.
It has long been known that the Pati-Salam group SO(4) × SO (6) can be put together with the Lorentz group SO(3, 1) into a pseudo-orthogonal group in dimension 14 so that one generation of the Standard Model fermions is described by a single spinor representation of this large gauge group. In particular, a model based on SO(3, 1) × SO(10) ⊂ SO (3,11) has been studied in the literature, see [3] and also [4] for a recent review of the related material. This group has signature zero (modulo eight), and so its Weyl representations are real. It can then be seen that the fermions of a single generation of the SM, including their Lorentz spinor components, can be described by a single real Weyl representation of SO (3,11). Further, the Weyl Lagrangian in R 3,11 dimensionally reduces to just the right collection of chiral Dirac Lagrangians for the SM elementary particles in four dimensions.
It has also been previously noted, in particular in [5], that there is another group for which a similar construction is possible. Namely, given that the non-compact Lorentz group SO (3,1) is part of the game, there is no reason to restrict one's attention to only compact gauge groups on the SM side. Thus, we can also embed SO(4)×SO(6) ⊂ SO (4,6), and the latter fits together with the Lorentz group into SO(3, 1) × SO(4, 6) ⊂ SO (7,7). The representation theory in this case works out correctly as well, and it can similarly be expected that the Weyl Lagrangian in R 7,7 dimensionally reduces to the right collection of chiral Dirac Lagrangians in four dimensions.
The split (or neutral) signature groups like SO (7,7) also appear and play central role in completely different subjects of generalised geometry [6], [7] and double field theory [8]. In particular, in the generalised geometry setup it is of central importance that spinors of SO(n, n) have the natural interpretation of differential forms in R n .
The purpose of this article is to combine the referred to above representation theory construction putting together the Lorentz group SO (3,1) with the Pati-Salam group SO(4)×SO (6) into SO (7,7) with the geometric setups of generalised geometry and double field theory. What is achieved as the result is an interpretation of the SM fermions as even degree differential forms in seven dimensions. Further, we achieve a very compact rewriting of the kinetic part of the SM fermionic Lagrangian. Thus, the main result of this article can be summarised by the following relation Here Ψ ∈ Λ even R 7 is a general inhomogeneous even degree real form in R 7 , the SO (7,7) invariant inner product (·, ·) is to be explained below, but is essentially a combination of an involution on Λ • R 7 and restriction to the top form, and D is the Dirac operator on R 7,7 that is essentially an appropriately interpreted exterior derivative operator, see below. All differential form component functions are originally assumed to be functions on R 7,7 , the double of the space on which the differential forms are taken. This makes the subject of double field theory with some of its geometric constructions relevant. To achieve a reduction to the right-hand-side in (1) one assumes that the component functions only depend on 4 of the coordinates on R 7,7 , so that they are in fact functions on a copy of Minkowski space R 3,1 . This involves a choice of which copy of R 3,1 is taken, and it is in this process of selecting SO(3, 1) ⊂ SO (7,7) that the metric to which all fermions on the right-hand-side of (1) couple arises. Then \ ∂ is the chiral Dirac operator that maps unprimed 2-component Lorentz spinors to primed 2-component spinors, and ξ aI ,ξ aI are all unprimed 2component Lorentz spinors. The indices a = 1, 2 are the isospin ones, and I = 1, 2, 3, 4 are the colour ones, with the lepton charge interpreted as the fourth colour. The spinors ξ aI are spinor representations of the SU(2) L , and thus describe left-handed particles, whileξ aI are representations of the SU(2) R . If desired, the right-hand-side in (1) can be further split by choosing SU(3) ⊂ SU (4), and then rewriting everything in terms of the usual leptons and quarks. We explain everything in more details in the main text.
Thus, the main outcome of this article is a geometric construction which makes it obvious that SM fermions are, or at least can be elegantly described by, differential forms.
There have been numerous previous attempts to interpret spinors as differential forms. Indeed, it has been known for a very long time that spinors are related to differential forms, and the Dirac operator is related to the exterior derivative operator. In the physics literature this has been studied under the name of Dirac-Kähler fermions, the approach initiated in [9]. The other well-known references on this approach are [10], [11].
Briefly, the idea is to consider the first-order differential operator d + d * , where d is the exterior derivative and d * is the dual operator. This operator is the square root of the Laplacian operator dd * + d * d on differential forms, and naturally acts on the space of inhomogeneous differential forms (the space of differential forms of all degrees). Given that there is a relation between the Clifford algebra over a vector space and the exterior algebra, see below for a review of this, the space of differential forms is a module for the Clifford algebra, and thus has a spinor interpretation.
There is, however, the following well-known problem with this idea. After differential forms are given spinor interpretation they turn out to carry integer spinor representations. This can be seen in many different ways. A particularly straightforward way is available in four dimensions and uses the 2-component spinor formalism. Indeed, 1-and 3-forms are objects v µ with one spacetime index, which translates into two spinor indices of different type v µ → v M M ′ . The 2-forms v µν can be decomposed into their self-and anti-self-dual parts, and these correspond to rank 2 spinors v M N , v M ′ N ′ . All in all, differential forms are bi-spinors rather than spinors. In the cited above literature this fact is expressed by saying that in four dimensions a Dirac-Kähler fermion is a collection of four Dirac fermions.
While the above mentioned problem makes the old ideas [9] of little use in physics, the relation between differential forms and spinors has found a much more respectful place in mathematics, where it is regarded as classical. Thus, as is well-known, the Clifford algebra over a vector space V with inner product and the exterior algebra over V are isomorphic as vector spaces. Moreover, and this fact will play the central role in the present article, the spinor representations of the orthogonal group SO(2n) can be realised in the space Λ • R n of inhomogeneous differential forms in half the dimension. In the physics literature this construction goes under the name of (fermionic) oscillator realisation of representations. Another classical geometrical construction that makes prominent use of differential forms and the natural Dirac-like operator D = d + d * is the article [12] that used the index theorem for D to prove properties of zeros of vector fields on a manifold.
In this article we combine the mentioned construction of spinor representations of SO(2n) in the space Λ • R n with the geometric setup of double field theory [8]. This leads to a realisation of spinors as differential forms, with the Dirac operator being related to the exterior derivative operator. In the construction that we describe the problem of the Dirac-Kähler approach that differential forms are bi-spinors rather than spinors does not arise. This is achieved by working with differential forms in a space of half the dimension, which explains why the geometric setup of double field theory is particularly relevant.
The main points of our construction are quite simple and can be explained already in the Introduction. Let us start by describing the referred to differential forms construction of spinor representations of orthogonal groups. We will restrict our attention to the split signature orthogonal groups SO(n, n) for which this construction is particularly elegant. Thus, we start with a vector space V of dimension 2n that is equipped with a metric of split (neutral) signature. For any such space we can choose a doubly-null basis, i.e.
so that both T,T are totally null, see more on the geometry of such a choice below. We will represent elements of this vector space as columns The split signature metric is where G(ξ, η) is some (non-degenerate) pairing that provides an identificationT ∼ T * of objects η with duals of objects ξ. This geometric setup is present in double field theory [8], and also in a related but different way in Hitchin's generalised geometry [6], [7]. 1 Given a vector space equipped with an inner product, we form the Clifford algebra. Thus, we define the Clifford algebra for V as the algebra generated by V modulo the defining relation This Clifford algebra has a natural representation in the exterior algebra Λ •T overT . Indeed, the Clifford multiplication of ξ + η ∈ V with a general element φ ∈ Λ •T is defined as follows where the interior product i ξ is defined on elements ofT via i ξ η := 2G(ξ, η), and this is extended to arbitrary forms by linearity. This defines a representation because and so we have (5). This construction identifies spinors of SO(T ⊕T ) with elements of Λ •T . In particular, Weyl spinors are identified with the spaces of even/odd elements in Λ •T . In the generalised geometry approach [6] this construction gives identification of spinors with differential forms, and plays the central role.
We now extend the above linear algebra construction into a differential geometry one. For this we use the geometric setup of double field theory. It is at this point where we start to deviate from the generalised geometry setup. As is explained in [13], see also [14], the geometry of double field theory is that of a flat bi-Lagrangian, or para-Kähler manifold. Thus, let M be a manifold of dimension 2n endowed with a split (neutral) signature metric G and a metriccompatible para-complex structure F such that The para-complex structure F splits T M = T ⊕T the tangent space T M into eigenspaces T,T of eigenvalue ±1. The minus sign in the metric compatibility relation implies that both these eigenspaces are null. "Lowering the index" on F with G we get another tensor W (X, . Then the fact that T,T are null implies that the distributions T,T are Lagrangian with respect to the symplectic form W . All this statements will be verified in the main text. For simplicity, we shall assume throughout this article that the metric G is flat and distributions T,T are integrable. It should be noted that the described double field theory setup is a real version of the usual setup of complex manifolds with their compatible triple of a symplectic form, a Hermitian metric, a complex structure and an integrable distribution of the tangent space into (1, 0) and (0, 1) subspaces. Moreover, in the complex manifold setup it is well-known that the Dirac operator is essentially the Dolbeault operator∂. One has D =∂ +∂ * : Λ 0,even → Λ 0,odd , see e.g. Exercise 2.3.39 in [15]. Thus, in the complex manifolds case the spinors are differential forms in half the dimension, i.e. those generated by elementary forms dz i , where z i ,z i are the complex coordinates.
What we will describe next can be interpreted as the real version of the construction of the previous paragraph. Let M be a para-Kähler manifold with a split signature metric G and a compatible doubly-null integrable distribution T M = T ⊕T . Let T * M = T * ⊕T * be the corresponding distribution of the cotangent space. Let x I be a set of coordinates for M, with x I = (x i ,x i ), i = 1, . . . , n and the corresponding basis of T * M being given by 1-forms so that the metric pairing is given by We now form the Clifford algebra for T * ⊕T * . It is generated by dx i , dx i with the defining relations being and both dx i , dx j anti-commuting between themselves. As we already know, the space of spinors for SO(n, n) is naturally identified with the space Λ • T * whose elements are differential forms where the wedge product is implied. The Dirac operator for the metric G is then explicitly described as follows where c is the Clifford multiplication. Then, as we know from (6), Clifford multiplication c(dx i ) is just the wedge product, while c(dx i ) acts by interior multiplication. Explicitly, where ∂ i := ∂/∂x i ,∂ i := ∂/∂x i . Thus, we see that the Dirac operator on M for a SO(n, n) metric G is just the appropriately interpreted exterior derivative operator. Numerous explicit examples of working with the operator D will be given in the main text. We hope that the reader will appreciate the naturalness of this construction, in particular by following these examples.
What we have described is just the translation of well-known construction of the Dirac operator on Kähler manifolds to the real setup of para-Kähler manifolds, but, as far as we know, this translation has not appeared in the literature before. This is presumably due to the fact that the construction of the spinor representation of SO(n, n) as that in the space Λ • R n is most frequently met in the geometric setup of generalised geometry, where the manifold remains of dimension n and in particular not doubled, while we saw this to be necessary to describe the Dirac operator. On the other hand, in the double field theory framework spinors and differential forms has not played any significant role up to now, and so there was no motivation to consider the Dirac operator.
We can now explain how the other part of the geometric setup of double field theory, namely another metric on M comes into play. This happens in the process of selecting of which SO(3, 1) ⊂ SO(7, 7) is identified with the Lorentz group. Geometrically, this is done by selecting another decomposition V = U ⊕Ũ so that the metric G restricts to a non-degenerate metric on U,Ũ, e.g. positive definite on U and negative definite onŨ . One can then identify the Lorentz group as that of mixing say first 3 of the directions in U and one direction inŨ . Then, as we explain in the main text, the required decomposition V = U ⊕Ũ is the same as a metric G compatible endomorphism J squaring to identity J 2 = I, with U,Ũ being its eigenspaces. As we shall see in the main text, this endomorphism is essentially the generalised metric that arises in both double field theory and generalised geometry. Thus, this second geometric ingredient of double field theory also plays an important role in our construction, and arises in identifying the Lorentz group SO(3, 1) inside SO (7,7).
The organisation of the rest of this paper is as follows. In Section 2, we start by spelling out the sketched in the Introduction geometric constructions in more details. We then study groups SO(n, n) of increasing dimension and work out what the described idea of embedding the Lorentz group SO(3, 1) into SO(n, n) gives in each case. We will see that the properties of the SO(n, n) invariant inner product (·, ·) needed on the left-hand-side of (1) are such that only in very few cases the Lagrangian (Ψ, DΨ) produces something non-trivial. One of these cases is the setup of SO(7, 7) that is related to the Standard Model.
We start in Section 3 with the setup of SO (2,2). There is no Lorentz group inside in this case, and we just work out the spinor representations and explicitly verify that (15) is the correct Dirac operator in this case. Our next example is that of SO(3, 3), which we treat in Section 4. The group SO (3,3) contains the Lorenz group. The subgroup that commutes with Lorenz group is SO(2), and so we expect to see charged fermions in this case. We verify that the Lagrangian (Ψ, DΨ) in this case reduces to the Weyl Lagrangian for a single charged 2component Weyl fermion in four dimensions. As we explicitly verify in Section 5, the Lagrangian (Ψ, DΨ) vanishes in the setup of SO (5,5). The setup of SO(7, 7) is treated in Section 6, where we explicitly verify that the SM fermion content arises and check the relation (1). We conclude with a discussion.
2 Geometric preliminaries 2.1 Pseudo-orthogonal group SO(n, n) The group of transformations preserving the metric (4) is O(n, n). In this paper we are not interested in subtleties related to discrete subgroups, and so we will just denote the relevant group by SO(n, n) everywhere. Its Lie algebra can be described explicitly as follows. We follow [16] closely, making necessary changes to work in the double field theory rather than generalised geometry setup. The Lie algebra of the group SO(n, n) consists of endomorphisms of V with the property To provide the explicit matrix description it is convenient to use the identificatioñ that is provided by the metric G. With this identification in mind and the representation (3) of vectors in V assumed, the Lie algebra so(T ⊕ T * ) consists of the following matrices and both β, B are anti-symmetric tensors. Explicitly, introducing a basis e i ∈ T andẽ i ∈T so that the metric pairing is G(e i ,ẽ j ) = (1/2)δ j i , the general element of V is ξ i e i + η iẽ i . The pairing of T,T that is used in the identification (17) The tensors β, B are objects β ij , B ij , and A T is the endomorphism η i → A j i η j . A proof of the fact that M preserves the metric G is straightforward verification.
The exponentiation of some of the subgroups is easy.
B-transform. Exponentiating the subgroup generated by B we get This acts on X = (ξ, η) T as The minus sign in the above formula is different from that in [16], but is more natural if the action of a 2-form on a vector field is in components B ij ξ j , i.e. the second index of the tensor in O is contracted with the index in X. This convention also agrees more naturally with what we will see in the Clifford algebra.
β-transform. Exponentiating the subgroup generated by β we get This acts on X as Our sign here is also different from that in [16].
GL(n)-transform. Exponentiating the subgroup generated by A we get

Spinors and differential forms
As explained in the Introduction, we construct the Clifford algebra for T ⊕ T * . It is generated by the basis e i ,ẽ i , e i ∈ T,ẽ i ∈ T * with the defining relations being and both e i ,ẽ i mutually anti-commuting. The space Λ • T * is then a module for the above Clifford algebra with the Clifford multiplication being and All this can also be described in a more physics-friendly creation-annihilation operator notation, see below.

The action of SO(n, n) on spinors
Our aim is to describe the action of subgroups of SO(n, n) on spinors as differential forms. Again, we follow [16]. The group SO(T ⊕ T * ) is doubly covered by Spin(T ⊕ T * ), and the latter can be explicitly described as sitting inside the Clifford algebra The Lie algebra so(V ) is Λ 2 V , and this also sits naturally inside the Clifford algebra. Its action on V can then be described as the natural action of Λ 2 V on V by the commutator, both viewed as sitting inside the Clifford algebra Let us work this out in our setup. Using the Clifford algebra relations (24) we havẽ and so for some 2-form B = (1/2)B ijẽ iẽj its action on vectors is This is the same action as we have seen in (20). Let us also see how the bi-vectors act. We have and so which is the action from (22).
Finally, let us check that the action of the Lie algebra of GL(n) is also as we previously described. It corresponds to the commutator of the Clifford algebra element ( which is the correct action. This allows us to write the action of the Lie algebra so(T ⊕ T * ) on spinors from Λ • T * . We have Using (25) and (26) this works out to is the insertion of the bi-vector β into the k-form φ.

Creating-annihilation operators
The above Clifford algebra relations and the formula (34) can be rewritten in more physics friendly notations with the help of creation-annihilation operators. Thus, we identify We then have The Clifford algebra module Λ • T * is then the Hilbert space spanned by all vectors created from the vacuum |Ω via the creation operators The formula (34) takes the following form

The double field theory setup and the Dirac operator
In preparation for the description of the Dirac operator in the above language, we now describe the geometric setup of double field theory in some more detail. As already described in the Introduction, we start with a manifold M of dimension 2n with a split signature metric G on it. We then require that there exists a endomorphism of the tangent bundle F : F 2 = I, which is metric compatible in the sense of (9). This splits the tangent bundle into subspaces T,T of eigenvalues ±1 of F . It is not hard to see that these subspaces are totally null. Indeed, denoting by X ′ , Y ′ eigenvectors of F we have and so G(X ′ , Y ′ ) = 0. The data of F and G define another tensor W (X, Y ) = G(F X, Y ), which can be thought of as the endomorphism F with one of its indices lowered with the metric G. This tensor is in this case anti-symmetric and so is a 2-form. It is easy to see that the subspaces T,T are Lagrangian with respect to W . So, manifolds of this type can be referred to as bi-Lagrangian.
In what follows we assume that the metric G is flat, and that we can work in coordinates x i ,x i in which the basis in T * M is given by (10). We then generate the Clifford algebra as in (12), and realise its spinor representations by elements in Λ • T * , and so by differential forms of the type (13). Then the Dirac operator for the metric G is given by (15). The formula (40) for the action of the Lie algebra of Lorentz group on forms still applies, one just has to identify and keep in mind that the fist operator acts by the usual wedge product, while the second operator acts by interior product

The inner product(s)
We now describe two different inner products in the space Λ • T * , both invariant under the action of SO(T ⊕ T * ). We take the first inner product from [16], see also references therein. Let σ 1 be the main antiautomorphism of the Clifford algebra, i.e. the one determined by the map It is not hard to check that it acts on elements of the Clifford algebra changing signs according to the degree of the corresponding element where ǫ 1 (p) = 1 when p = 0, 1 mod 4 and ǫ 1 (p) = −1 when p = 2, 3 mod 4. Then for Ψ 1 , Ψ 2 ∈ Λ • T * we define the inner product to be where the notation means restriction to the top form in Λ • T * . The invariance of this inner product follows from Taking into account that Spin(T ⊕ T * ) sits inside the Clifford algebra as (27) we see that the inner product (46) is invariant under the identity component of Spin. The described inner product is symmetric when n = 0, 1 mod 4, and anti-symmetric when n = 2, 3 mod 4. We now describe another invariant inner product on spinors. We take this construction from [17]. It is given by the same construction (46), but with the involution σ 1 replaced with a different one σ 2 . The involution σ 2 can again be described as changing signs according to degrees (45), but now with ǫ 2 (p) = 1 when p = 0, 3 mod 4 and ǫ 2 (p) = −1 when p = 2, 3. So, the involutions σ 1 , σ 2 differ by sign in what they do to odd elements of Clifford algebra. The second inner product is symmetric if n = 0, 3 mod 4 and anti-symmetric if n = 1, 2 mod 4. This is opposite symmetry property to (46) for odd n.
It should also be noted that the spaces of even and odd forms in Λ • T * are null with respect to both inner products when n is odd, and orthogonal to each other when n is even.

The Weyl and Dirac Lagrangians
We can now describe natural SO(n, n) invariant Lagrangians that can be constructed with the Dirac operator D and the above invariant inner products on Λ • T * . The construction is to take (Ψ, DΨ) 1,2 with respect to one of the two described inner products.
The construction is different depending on the parity of n. Let us describe the n odd case first. In this case we can restrict the Lagrangian (Ψ, DΨ) 1,2 to the space of even or odd forms in Λ • T * , i.e. to the space of Weyl spinors. Indeed, if we decompose where Ψ + stands for even forms and Ψ − for odd, then we have and so even and odd forms are not mixed by the kinetic term. They would be mixed if we wished to add to the Lagrangian terms like (Ψ, Ψ) 1,2 , which are possible depending on which inner product is used in dimensions n = 1, 3 mod 4. However, we will not be considering these Dirac mass terms, and restrict our attention to Weyl spinors, which we take to be given by even forms. Then no explicit mass terms can be written. It can then be checked that (Ψ + , DΨ + ) 1,2 vanishes modulo surface terms arising by integration by parts when n = 1 mod 4. This happens for both inner products. The Weyl Lagrangian (Ψ + , DΨ + ) 1,2 is only non-trivial for n = 3 mod 4, and in this case both inner products give the same result, modulo an overall sign. A proof of these statements is by explicit verification.
Let us now consider the situation when n is even. In this case the kinetic term mixes the even and odd forms It now does matter which inner product is chosen. It can be checked that the first described inner product gives a non-trivial Lagrangian for n = 0 mod 4 (and vanishes modulo surface terms for n = 2 mod 4), and the second product gives a non-trivial Dirac Lagrangian for n = 2 mod 4 (and vanishes for n = 0 mod 4). The Dirac mass term (Ψ, Ψ) is only non-trivial for n = 0 mod 4. To summarise, the Weyl Lagrangian (Ψ + , DΨ + ) only exists for n = 3 mod 4, in which case it does not matter which inner product is used. The Dirac Lagrangian (Ψ, DΨ) exists for n = 0, 2 mod 4, depending on which inner product is used. We will only consider the Weyl Lagrangian in this paper, and omit the subscript "plus" from Ψ + from now on.

The second metric
Let us now assume that on top of the geometric structure F, G already present on M, we are given another endomorphism J : J 2 = I and that is metric-compatible in the following sense i.e. with no minus sign as in the case of F . Let us denote the eigenspaces of J of eigenvalues ±1 as U,Ũ, so that T M = U ⊕Ũ. Then it is easy to show that the metric G restricts to a non-degenerate metrics on U,Ũ , and U,Ũ are G-orthogonal. Indeed, to show the orthogonality we take X ′ ∈ U, Y ′′ ∈Ũ and compute and so G(X ′ , Y ′′ ) = 0. A similar computation shows that the restriction of G to U,Ũ is non-degenerate.
We can now parametrise such endomorphisms J by what in the double field theory context is usually referred to as the generalised metric. Let us see how this can be done. First, each of the spaces U,Ũ, being of same dimension as T , can be described as a graph of some map T → T * , where we again identifiedT = T * . Each such map is a rank two tensor, and let us denote by g its symmetric part, and by b its anti-symmetric part. So, the space U can be parametrised as consisting of elements where b, g ∈ T * ⊗ T * are some anti-symmetric and symmetric tensors. A moment of reflection shows that the space that is G-orthogonal to U is then where the same tensors are used. The restriction of the metric (4) to U is then and the restriction toŨ is minus this. Let us now use the data g, b to construct an endomorphism of T ⊕T * that squares to identity, is metric compatible and whose eigenspaces are as described above. This endomoprhism is explicitly given by It is easy to see that this endomorphism is designed to square to the identity The endomorphism J is also metric-compatible G(J·, J·) = G(·, ·). To check this, we must compute Let us now pair JX 1 , JX 2 . We have where some transposes where taken and minus signs from b T = −b introduced. Opening up the brackets and seeing the cancellations one verifies G(JX 1 , JX 2 ) = G(X, Y ). Let us also see that the eigenspaces of J are as described above. The eigenvector equation for eigenvalue +1 is Then the second of the arising equations (g − bg −1 b)ξ + bg −1 η = η is automatically satisfied. Thus, we learn that the eigenvectors of eigenvalue +1 are of the form (53) and those of eigenvalue −1 are of the form (54). So, J given by (56) is indeed the required endomorphism.

Another basis for V
We thus assume that in addition to data F, G there is some mechanism that gives rise to an endomorphism J as described above, and thus to tensors g, b ∈ T * ⊗ T * . A possible origin of such mechanism will be discussed in the last section. For simplicity we assume that b = 0 in what follows. We will also usually assume that g is the flat Riemannian signature metric in T . With this assumption a vector ξ + η ∈ V can be decomposed into its U,Ũ parts as The inverse of this transformation, in matrix form In this basis the metric is and the Lie algebra is represented by matrices which have g-anti-symmetric matrices on the diagonal, and have the off-diagonal blocks gtranspose of each other. When g ij = δ ij the factors of g, g −1 can be simply removed from this formula, with positions of indices adjusted appropriately. We will need this result below when we describe embedding of various subgroups into SO(n, n).
3 The case of SO (2,2) In this section we work out the spinor representations of SO(2, 2) and explicitly verify that the Dirac operator as described in (15) is the usual Dirac operator for the split signature metric in four dimensions. Readers that do not need such an explicit verification can skip this section.

Spinor representations of SO(2, 2)
Let us see explicitly how the spinor representations of SO(2, 2) are differential forms in R 2 . To this end, we introduce a pair of creating annihilation operators a 1 , (a 1 ) † and a 2 , (a 2 ) † , with the usual anti-commutation relations a i (a j ) † + (a j ) † a i = δ j i and all other pairs anti-commuting. We can then consider the following operators It is easy to check that the following sl(2) commutation relations hold This gives us one copy of sl(2) Lie algebra. One can form the second copy of sl(2) in the following waȳ Again we get the usual sl(2) commutation relations And it is not hard to check that all barred operators commute with unbarred ones, so we have two commuting copies of sl(2). If we do this construction over reals we get an explicit realisation of the Lie algebra of so(2, 2) as two commuting Lie algebras sl(2, R).
Let us now discuss its action on spinors. The Weyl representations are formed by forms of even and odd degrees. The forms of odd degree are spanned by dx 1 , dx 2 . The action of the first copy of sl(2) is as follows while the second copy acts trivially on these states. The algebra (67) is realised the by the matrices A comparison of this with the action (70) then fixes the identification of forms with 2-component column spinors up to an overall sign, which we choose as follows The even degree forms are spanned by 1 and dx 1 dx 2 . The first copy of sl(2) acts trivially, while the action of the second copy is The identification with 2-column spinors that we choose for this copy is 3.2 The Dirac operator on R 2,2 We start with R (2,2) with metric in the diagonal form In terms of coordinates the metric is Consider the real 2 × 2 matrix Here A, A ′ = 1, 2 are 2-component spinor indices. For the matrix u AA ′ the index A ′ enumerates columns and A enumerates rows. The determinant of the above matrix is minus the squared interval.
Let us now construct the related matrices u A A ′ and u A ′ A . The first one is given by −ǫu, and the second one by −ǫu T , where is the metric in the space of spinors, with the conventions being so that (ǫλ) T is the row representing λ A . Then λ A µ A is the usual matrix product of a row and a column. We have We have We can then form the 4 × 4 matrix that acts on 4-component Dirac spinors and satisfies the desired Clifford algebra relation This construction gives us the two chiral Dirac operators We can write these operators more compactly as These are the two chiral Dirac operators The action on a primed spinor in S − is and on unprimed one we have

The Dirac operator as the exterior derivative
We now show verify that the Dirac operator on R 2,2 is essentially the exterior derivative operator appropriately interpreted. Thus, we identify the space of primed spinors with the space of odd forms as in (72). The exterior derivative operator is then Here the differentials dx 1 , dx 2 are interpreted as annihilation operators that can act on dx 1 , dx 2 non-trivially, and only the terms giving non-zero contribution have been kept. We used a different letter for the exterior derivative to signify the fact that D 2 = 0.
We can write the above result as Taking into account (72), (74) we see that D indeed gives the correct chiral Dirac operator when it acts on odd forms. We can similarly compute the action of D on even forms We see that This verifies that the exterior derivative operator D, interpreted in the sense of Clifford multiplication, matches the Dirac operator on R 2,2 In particular, this shows that where ∆ is the Laplacian on R (2,2) .

Case of SO(3, 3)
The group SO(3, 3) is the smallest of SO(n, n) groups that contains the Lorentz group SO(3, 1).

Embedding of SO(3, 1) × SO(2)
To select a copy of the Lorentz group sitting inside SO(3, 3) we pass to the u,ũ coordinates (63) that make the metric diagonal. We do this by choosing the metric g to be a flat metric of signature all plus. Using this metric we can lower-raise the indices of x i ,x i , and also of matrices β ij , B ij , A i j . We will write all coordinates with indices down so that a † i = dx i and a i = dx i , and matrices with indices up.
We want to embed the Lorentz group SO(3, 1) in the "diagonal" way into SO (3,3). The Lie algebra of SO(n, n) in the basis in which the metric is diagonal is formed by matrices of the form (65). We assume g = δ everywhere. To describe a copy of SO(3, 1) inside let us start with the rotations subgroup. This is embedded into the upper-left corner of the 6 × 6 matrix (65). So we want A + A T + B − β = 0, A − A T − (B + β) = 0 and so A = β, B = β, which in the familiar ξ, η basis corresponds to matrices of the form with β being the usual 3 × 3 matrices representing rotations of the first three coordinates. This leads to the following matrices where we included the prefactors in order to get the correct normalisation, see below.
Let us now discuss the boosts. We choose this to mix the coordinateũ 3 with the coordinates u i . They are thus represented in the u,ũ basis by matrices (65) with zero on the diagonal A − A T ± (B + β) = 0 , and with the off-diagonal block equal to (A + A T + B − β) ij = δ is δ j3 , where s = 1, 2, 3. These leads to A being a symmetric matrix, and B = −β, and the following matrices in the η, ξ representation: The above set of matrices K i , P i can then be checked to have the usual Lorentz group commutation relations These matrices act on differential forms as (34), and so correspond to the following set of operators on differential forms which are all anti-Hermitian, and which are all Hermitian. We now work out the similar embedding of SO(2) subgroup that mixes theũ 1 ,ũ 2 coordinates. In the u,ũ basis this corresponds to matrices with off-diagonal blocks equal to zero, and thus A + A T + (B − β) = 0, and with the upper-diagonal block equal to zero A − A T + B + β = 0. This gives B = −A, β = A T . Thus, these are matrices of the form The particular rotation that we are after is represented by the following matrix which corresponds to the operator This operator is anti-Hermitian, as is appropriate for a rotation. It can be checked that the matrix C commutes with K i , P i as it should.

Change of basis
To describe the action of all the operators on differential forms, we introduce the complex linear combinations We then define a new set of creation and annihilation operators, corresponding to creationannihilation of m,m The anti-commutation relations are now a m a †m + a †m a m = 1, while a m a † m + a † m a m = 0, ama †m + a †m am = 0.
This is of course just the usual Clifford algebra relations corresponding to the fact that the metric in 1, 2 plane in the new basis is dm ⊗ dm + dm ⊗ dm. In the new basis, the operator K 3 takes the following form This immediately gives the eingestates of K 3 We also need the operator C in the new basis with eigenstates being We also list the eigenstate of P 3 It is also convenient to introduce the complex linear combinations K 1 ± i K 2 and P 1 ± i P 2 . We can rewrite these operators in the new basis as Finally, we introduce the usual self-dual/anti-self-dual combinations

2 × 2 matrix realisation of the Lorentz Lie algebra
For reference, we give here the 2 × 2 matrix realisation of the Lie algebra so (3,1). In this realisation the generators are given by respectively, where σ i are the usual Pauli matrices. And so we have which shows that E − ,Ē − are lowering and E + ,Ē + are raising operators, as the notation suggests.

Eigenstates
We now consider the Weyl representations of SO (3,3), which are the spaces of even and odd forms in R 3 . We want to exhibit a basis in this space that diagonalises the operators K 3 ± i P 3 and C. We have 8-dimensional space of forms on R 3 that is spanned by forms m,m, (1 ± mm) and the same forms times dx 3 . They are all eigenstates of K 3 , P 3 , and also of C, and so we just have to divide the states into two groups that transform non-trivially under K 3 + i P 3 and trivially under K 3 − i P 3 and vice versa. Let us first describe the even forms. The corresponding Weyl representation consists of four real states, which can be represented as two complex and their complex conjugates. The states that transform non-trivially under K 3 + i P 3 are and they thus form a single primed spinor of SO(3, 1), and the states that transform non-trivially under The factors of 1/ √ 2 are introduced for future convenience. These two SO(3, 1) spinors are also eigenstates of C, with the primed spinor (120) being eigenstate of eigenvalue +i /2, and the unprimed spinor (121) having the eigenvalue −i /2. So, the content of the Weyl representation of SO(3, 3) on even forms in R 3 is a single electrically charged 2-component Weyl fermion plus its complex conjugate 4 R = 2 C .
We can also check that the states (120), (121) transform correctly under the operators (114), (115). We have for the primed spinor and for the unprimed onē Comparison with (119) then shows that we should identify a linear combination of these forms with the following 2-component spinor To identify the primed spinor in a way that respects the complex conjugation we need to interpret the action of the 2 × 2 matrices on spinors from the right instead of from the left. In this case the identification (125)  Let us also describe the odd forms. We have the states that transform non-trivially under as well as states that transform non-trivially under K 3 − i P 3 These states are also eigenstates of C, with the primed spinor (126) being of eigenvalue −i /2 and the unprimed spinor (127) of eigenvalue +i /2. We thus get another complex 2-component spinor (and its complex conjugate) in the other Weyl representation of SO(3, 3).
If we now select the states of a given eigenvalue of C, e.g. i /2, these are the unprimed spinor from the space of odd forms (127) and the primed spinor from the space of even forms (120). Thus, the states of a given eigenvalue of C form a charged Dirac spinor. Altogether, the Dirac representation of SO(3, 3) is then a Dirac spinor of SO(3, 1) and its complex conjugate spinor 8 R = 4 C , where 4 is the Dirac representation of the Lorentz group.
For completeness, let us also list the action of operators (114), (115) on the odd forms. We have and for the unprimed spinor forms we havē Comparison with (119) fixes the 2-component spinor identification up to an overall sign, which we choose as For the primed spinor, we again choose the identification that respects the complex conjugation

The Dirac operator on R 3,1
Our goal now is to verify that the Dirac operator that arises from the exterior derivative on R 3,3 reduces to the usual 4-dimensional Dirac operator when dimensionally reduced by assuming that all states depend only on u i ,ũ 3 but not onũ 1 ,ũ 2 . To this end, we need to describe the usual Dirac operator, analogously to what we have done for the Dirac operator on R 2,2 in the previous section.
As the first step, we form the 2 × 2 Hermitian matrix This matrix has the property that its determinant is minus the interval Form this we form the matrix with primed index down u A ′ A = −ǫ A ′ B ′ u B ′ A and so in index-free notations the matrix u A ′ A is given by minus (ǫu), where u is the matrix (132). We have We can also take the transpose of u and lower the unprimed index, i.e. consider −ǫu T . This gives Together, these matrices satisfy the properties We then define the two chiral Dirac operators

Change of coordinates
The derivatives appearing in the Dirac operator are with respect to u i andũ 3 , and these coordinates are related to the x,x coordinates as u = (1/2)(x +x),ũ = (1/2)(x −x), and so we have where we omitted the indices for brevity. Thus, we can write all u,ũ derivatives as combinations of x,x derivatives. Now, we introduce the complex linear combinations that appear in (103) This gives So, we can write all u,ũ derivatives as combinations of derivatives with respect to m,m,m,m and x 3 ,x 3 . We now take into account that the functions we want to consider, i.e. components of the differential forms, should depend only on u i ,ũ 3 but not onũ 1 ,ũ 2 . Becauseũ = (1/2)(x −x), this means that all functions depend on x 1 , x 2 ,x 1 ,x 2 only in combination x 1 +x 1 , x 2 +x 2 , but there is no dependence on the differences. This implies the following relations between the partial derivatives Then (140) implies that These are the derivative relations that follow from the assumption of the dimensional reduction. Then the derivatives appearing in the Dirac operator are and their complex linear combinations are We also have the linear combinations of derivatives with respect to u 3 ,ũ 3 Taking into account the above, we get the following expressions for the chiral Dirac operators (137) Note that ∂ T = ∂, where bar denotes the complex conjugation.

The Lagrangian
For future use, we now list the usual Lagrangian for a single Weyl 2-component fermion In index-free notations this reads The resulting action is explicitly real by integration by parts. This of course also follows directly from (147) by Hermiticity of ∂ A ′ A . Neglecting the possible boundary terms we can also write the Lagrangian in an explicitly real form

The Dirac operator as the exterior derivative
Using (142), the D operator applied to the differential form corresponding to a primed spinor is where we only wrote non-vanishing terms. The using the rule that dm eats dm and dm eats dm we see that there are some cancellations in the above expression, and it simplifies to This result can be written as where we have used (146). The comparison with (130) shows that D gives the correction action of the Dirac operator, up to a factor of 1/2.
We now do a similar calculation for the unprimed spinor form, skipping intermediate steps this time. We have This can be written as follows Again, comparison with (131) shows that the correction action of the chiral Dirac operator is reproduced. This formula is just the complex conjugate of (151) as it should be. It can also be similarly verified that and that the complex conjugate version of this formula also holds. Overall, the operator D reproduces the chiral Dirac operators.

The Lagrangian in terms of differential forms
We take a real Weyl spinor of SO (3,3) which is realised in the space of even forms in R 3 . In terms of introduced above basis (124), (125) this corresponds to the following differential form where c.c. stands for the complex conjugate form. To simplify calculations that follow, it is convenient to rewrite this state in index-free notations as where we introduced a column of even forms that correspond to an unprimed Lorentz spinor.
The D derivative of f T even ξ is given by (153). We can rewrite this derivative as where we introduced We now calculate (Ψ, DΨ). We will use the first inner product described in (46). We need the following matrix of inner products where ǫ is the anti-symmetric matrix (79), and we have used dmdm = i dx 1 dx 2 . This means that where we have compared with (149). Thus, the Lagrangian in terms of differential forms (Ψ, DΨ), which is explicitly real, matches the Weyl Lagrangian (147) in its explicitly real form (149).

The case of SO(5, 5)
This case is interesting because the stabiliser subgroup of the Lorentz group in this case is SO (2, 4), the conformal group in 4 dimensions. In particular, the maximally compact subgroup of the stabiliser is SO(2) ∼ U(1) times SO(4) = SU(2) × SU(2)/Z 2 . So, our Lorenz spinors will also transform non-trivially under U(1) × SU(2) × SU(2), which is similar to what is happening in the SM, and so is interesting. However, as we know from general considerations, in this case the Weyl Lagrangian (Ψ, DΨ) vanishes for either of the two inner products. Thus, there is no interesting Lagrangian that can be written in this case. Still, we work it out because there are many similarities with the representation theory of the interesting SO(7, 7) case.

Embedding of SO(3, 1) × SO(2) × SO(4)
We now have two more coordinate pairs u 4,5 ,ũ 4,5 and x 4,5 ,x 4,5 as compared to the previous section. We leave the embedding of the Lorentz Lie algebra unchanged, and given by (98), (99). So, we only need to describe the embedding of SO(2) × SO(4). Let us start with SO(2). This group mixes the coordinates u 4 , u 5 . Its Lie algebra in the x,x basis is then given by matrices of the type (94), where β ij = δ [i4 δ j]5 . The corresponding operator, which we shall call C (for charge), is given by Let us now describe the SO(4) subgroup. This is the group of rotations that mixes the coordinatesũ 1,2,4,5 . In the x,x basis it consists of matrices of the form (100). We will call the generatorsK i ,P i and choose them as follows They generate the SO(4) Lie algebra

Change of basis
We now go to the m,m basis (103). In this basis the operatorK 3 takes the following form This immediately gives the eingestates ofK 3 To describe the other operators, let us introduce a set of complex coordinates z,z mixing the directions 4, 5 and Finally, we introduce the self-dual/anti-self-dual combinations

Eigenstates
We now consider the space of even forms on R 5 . In comparison to the situation analysed in the previous section, we have added two new coordinates, which in the complex basis are z,z. Thus, the even forms on R 5 are even forms on R 3 that we already know how to interpret times even forms on R 2 , of which the most convenient basis is (1 ± dzdz), plus odd forms on R 3 times odd forms on R 2 for which the basis is dz, dz. We should now construct combinations that, in addition to diagonalising operators K 3 ± i P 3 , also diagonaliseK 3 ±P 3 , as well as C. At this stage it is best to pass to particle physics notations, and label states by the eigenstates of all these operators. Let us first work out the eigenstates ofK 3 ±P 3 . All eigenvalues will be ±i , and for the copy of SU(2) to which the generatorK 3 +P 3 belongs (we will refer to this copy as right) we will identify the eigenvalue +i as projection of spin +(1/2), and eigenvalue −i as projection of spin −(1/2). For the copy of SU(2) containing the generatorK 3 −P 3 (we will refer to it as left) the identification is opposite. When listing the eigenvalues, we will list the left eigenvalue first, followed by the right eigenvalue. We will also indicate the eigenvalue of C, in the third position, with eigenvalue +i /2 indicated as +1 and −i /2 as −1.
This conventions produce the following list of states. We start with the eigenstates of SU(2) R dmdz 0, in the space of even forms in R 4 and in the space of odd forms. Note that the complex conjugate of a right doublet is a right doublet, as it should be. Altogether we have 4 SU(2) R doublets, 2 doublets of each charge. The SU(2) L doublets in the space of even forms in R 4 are and in the space of odd forms Again this gives 4 SU(2) L doublets. Finally, by wedging the odd forms from the above list with dx 3 we get all of 16 even forms in R 5 . They are also eigenstates of K 3 ± i P 3 , as we know from the previous section. Let us select from this list of 16 states 8 states that are left (unprimed) spinors, i.e. transform non-trivially under K 3 − i P 3 . The required states can be seen in (124), (130). Another useful piece of notation is to introduce particle names for the listed above states. Thus, we shall refer to eigenstates of SU(2) L using unbarred letters, and will put a bar above the name of a particle for SU(2) R states. This gives Here the bars are just parts of the name and have nothing to do with complex conjugation. With these notations in mind we can write the form that corresponds to an unprimed Lorentz spinor and an SU(2) L spinor. We have where the signs are chosen as in the previous section. This left spinor is of C charge +1 in the conventions chosen. We can similarly write the form that corresponds to an unprimed Lorentz spinor and an SU(2) R spinor. We have The right spinor is of C charge −1. This parametrises the general real even form on R 5 as

Rewriting
We now rewrite the above form in a way more suitable for computations. We have already introduced the Lorentz 2-component states (156) and (158). With these notations we can rewrite the form Ψ L as where it is understood that ν, e are 2-component unprimed Lorentz spinors Similarly,

The Lagrangian
We know from general considerations that the Lagrangian (Ψ, DΨ) should be zero in this case, modulo surface terms. Here we confirm this by an explicit calculation, to become more proficient with the technology.
With results (153), (154) it is easy to compute the action of the D operator. Indeed, the component functions have dependence on only m,m, x 3 andm,m,x 3 , and there is no dependence on the z-coordinates. Thus, the D operator is insensitive to the terms that contain dz, dz. These terms are always written at the end of the forms, and they continue to remain there after the D acts on what stands before such terms according to (153), (154). Thus, we can use This gives and We now need to compute the relevant inner products of forms. An example calculation is where ǫ is the anti-symmetric matrix (79). We note that we could have computed this in steps where we used −(dz, dz) = −dzdz = dzdz = i . This means that the part of the Lagrangian (Ψ, DΨ) that depends on the ν functions is We should notice however that there is no factor of i in this Lagrangian, in contrast to what happened in the SO (3,3) setup. This happened here because we got two factors of i , one coming from dmdm, and another coming from dzdz. But because there is no factor of i here we get a cancellation, modulo surface terms. Indeed, because ∂ A ′ A is a Hermitian matrix we have, for any spinor ξ A and so the complex conjugate of the first term in (194) is minus itself modulo surface terms, which is the reason why in (147) we multiply this expression by i to get a real Lagrangian. But this means that there is cancellation in (194). Thus, we confirmed by an explicit verification that (Ψ, DΨ) = 0 modulo surface terms in the SO(5, 5) setup. Thus, there is no interesting Lagrangian we can write in this case.
6 The case of SO(7, 7) We now come to the case of real interest. We embed the Lorentz group SO(3, 1) as described before. Its commutant in SO(7, 7) is SO (4,6), and its compact subgroup is SO(4) × SO (6), which is the Pati-Salam GUT gauge group. We want to describe how even forms in R 7 give the desired SM fermion representations.

The group SO(4)
Let us start by describing the group that rotates the 4, 5, 6, 7 directions. In the diagonal u,ũ basis it is represented by matrices occupying the upper left corner. As we have already deduced when considering 1, 2, 3 rotations, in the x,x basis this corresponds to matrices of the form (94).
In terms of creation-annihilation operators we get the following set The commutators between these read Let us first list the states transforming non-trivially under SU(2) R , to which we takeK 3 +P 3 as belonging. We list the states in the format similar to that in the previous section. We get dzdw 0, 1 2 in the space of even forms in R 4 and in the space of odd forms. As in the previous section, the complex conjugate of a right doublet is a right doublet, as it should be. The SU(2) L doublets in the space of even forms in R 4 are and in the space of odd forms The above are 16 forms in R 4 and we need to complete them into even forms in R 7 . There are in total 8 forms in R 3 , 4 even and 4 odd. In each even/odd class two of the forms correspond to unprimed Lorentz spinor, and two to a primed one. We will only construct forms corresponding to unprimed spinors. The primed Lorentz spinors will then be obtained by complex conjugation. With this in mind, the forms in R 3 corresponding to unprimed Lorentz spinors are f T even ξ, f T odd ξ, where f even , f odd were introduced in (156), (158). We now multiply the even forms in R 4 by the even form f T even ξ, and odd forms in R 4 by f T odd ξ to get a convenient complex basis in the space of even forms in R 7 . It can then be checked that all the obtained states are also ±(i /2) eigenstates of T 1,2,3 , with the states that are SU(2) R spinors having the property that the product of signs of eigenvalues is always −1, and the product of sign for SU(2) L spinors is +1. Then SU(4) acts by mixing the 4 left states, and the 4 right states.

The state Ψ
We now introduce a convenient way to write the state Ψ that is given by a general even form in R 7 , decomposed into the basis of eigenstates described above. First, the Lorentz spinor parts are described by forms f even , f odd . Let us introduce an analogous basis of states for SU(2) L,R and SU(4). We will label these states by f aI , where a = 1, 2 is an SU(2) index and I = 1, 2, 3, 4 is an SU(4) index. We view the states f aI with fixed I as a 2-component row, and similarly for the state with fixed a, so that f aI ξ aI makes sense. Then the right states are as follows and the left states are The reason why we changed the order and listed odd forms before the even forms for the left states will become clear below.
We can now write the state Ψ as where where every ξ aI ,ξ aI is a 2-component unprimed Lorentz spinor represented by a column. Note that we could not have summed over the index I in the above expressions because for right states I = 1, 2 are multiplied by f even while I = 3, 4 are multiplied by f odd , and vice versa for the left states.

The hypercharge
It is now an interesting exercise to work out the expression for the operator that gives the correct particles hypercharges, as is listed in e.g. Table J.1 of [18]. Some simple guesswork shows that the correct expression is Indeed, let us list the −(1/6)(T 1 + T 2 + T 3 ) charges of the states appearing in (212). We have the following charges The operatorK 3 +P 3 does not act on the left states, and so this is already the correct assignment of hypercharges. This identifies the state with I = 4 as lepton, while I = 1, 2, 3 are the 3 different colours of quarks. For the right states the eigenvalues of −(1/6)(T 1 + T 2 + T 3 ) work out to be the opposite of those in (214), and adding to these the eigenvalues of −(1/2)(K 3 +P 3 ) we get just the correct hypercharges for all the right states as well, as are listed in the Table J.1 of [18]. Again we have that I = 4 describes the right-handed leptonsν,ē, while I = 1, 2, 3 describes the 3 colours of the right-handed quarksū,d.
The exercise that results in the expression for the hypercharge (213) is useful because it allows to identify which states correspond to quarks and which to leptons. It also shows how the choice of the hypercharge U(1) given by (213) explicitly breaks the SU(2) R subgroup of the Pati-Salam group.

The Lagrangian
To compute the Lagrangian we need to compute the inner products of states introduced above. A calculation gives where ǫ(I) = +1 for I = 1, 2 and ǫ(I) = −1 for I = 3, 4. For the right states the sign is opposite This means that the part of (Ψ, DΨ) that depends on say ξ a1 is Here in the second step the extra sign appeared because we interchanged the odd degree forms. On the other hand, the part that depends on ξ a3 is Thus, the final sign is the same as in (217). The final arising Lagrangian is which is the result that has been quoted in (1) with the notation \ ∂ := ǫ∂ used. Note that the SU(2) R particles appear with the opposite sign in front of their kinetic term as compared to the left particles. This Lagrangian is manifestly SO(3, 1) × SU(2) L × SU(2) R × SU(4) invariant as expected, and coincides with the kinetic term of the Standard Model fermion Lagrangian in which the lepton charge is treated as the fourth colour. If desired, it can be further split by separating I = 1, 2, 3 states from I = 4 states, which chooses an SU(3) subgroup of SU(4), and separates leptons from quarks. For more details on the used here 2-component spinor description of the SM we refer the reader to [18], see Appendix J of this reference.

Discussion
In this paper we carried out an exercise in SO(n, n) representation theory and worked out how the real Weyl spinor representation of this group splits under an SO(3, 1) subgroup. We also dimensionally reduced the R n,n Dirac operator to the Minkowski space R 3,1 . The outcome is a new interpretation of the fermion content of a single generation of the SM as a general real element of Λ even R 7 , i.e. as a real inhomogeneous even degree differential form in R 7 . Thus, we have demonstrated that fermions, in particular fermions of relevance for the Standard Model, can be described by differential forms, in spite of the difficulties that are usually associated with this idea in the context of Dirac-Kähler fermions [9]. At the most basic level, the reason why everything works properly is that differential forms in half the dimension, i.e. in R n rather than in R n,n , are used to describe spinors of SO(n, n). We also hope to have convinced the reader in the elegance of this formalism. In particular, the kinetic part of the SM fermion Lagrangian is correctly reproduced (1) by simple operations of exterior differentiation on R 7,7 and applying the Clifford algebra relations (12). We believe this new interpretation is comparable in its elegance to something very wellknown, namely the interpretation of the basic operators of vector calculus in R 3 in terms of exterior differentiation of differential forms. In the vector calculus context, the 3 different differential operators turned out to be all just different incarnations of the same exterior derivative operator acting on different degree differential forms in R 3 . Something similar arises as the result of our construction. Indeed, different representations of the SM Pati-Salam gauge group SU(2) × SU(2) × SU(4) with their relevant Dirac operators are all seen to be different incarnations of a single even degree differential form in R 7 , with the Dirac operator being just an appropriately interpreted exterior derivative operator. The naturalness of our construction leads us to suggest that it is possibly the split signature group SO(7, 7) that should be taken as the full group of "internal" symmetries of particle physics.
Another interesting and simple model that we described is based on the SO(3, 3) setup. In this case, the Weyl Lagrangian in R 3,3 dimensionally reduces to a single charged Weyl fermion in the Minkowski space R 3,1 . This model may be a good playground for testing generalisations that are necessary to make the SO(7, 7) model realistic.
Even though we worked with just fermions, setting all the bosonic fields (apart from the metric) to zero, the construction that we described is clearly in the direction of unifying gravity with other interactions. This is clear from the fact that this construction puts together the Lorentz group, which is the group of local frame rotations that preserve the metric, with the gauge groups of the Standard Model. So, our spinor construction effectively unified the local gravitational symmetry with the usual gauge symmetry, by realising both as commuting subgroups of SO (7,7).
A related remark is that the group SO(7, 7) contains GL (7). This means that there is a link to the group of diffeomorphisms is seven dimensions. This is the reason why SO(7, 7) transformations are related to the generalised diffeomorphisms in the double field theory context, see e.g. [19] for a recent reference addressing this point. And if diffeomorphisms are symmetry, then gravity is present.
The Lagrangian (Ψ, DΨ) that we obtained as the result of our construction only reproduces the kinetic part of the SM Lagrangian, with all bosonic fields (apart from the metric) set to zero. It is clear that the necessary step towards making the described in this paper ideas realistic is to understand how also the bosonic fields of the SM can be described. The most natural way to do this appears to be to make the global symmetry SO(7, 7) of the Lagrangian (Ψ, DΨ) local, by extending the operator D to some version of the covariant derivative. A mathematically natural way of doing this is to enlarge where C is a general odd degree element of Cliff(T ⊕ T * ). The reason why the restriction to odd degree elements arises is clear: The operator D maps even forms to odd forms, and this is why the inner product in (Ψ, DΨ) gives a non-trivial result in odd dimensions. The extended covariant derivative operator should work in the same way, and hence C must have odd degree. Further, it is probably not necessary to work with the most general odd degree element of Cliff(T ⊕ T * ) because we act on Weyl spinors, and on these there is a relation between the action of an element of Cliff(T ⊕ T * ) and the element that corresponds to its Hodge dual.
Thus, it appears natural to restrict to say self-dual odd degree elements. The Lie algebra of SO(7, 7), viewed as a subalgebra of Cliff(T ⊕ T * ), then acts on the "connection" C by the commutator. With this action the decomposition of the space of odd elements in Cliff(T ⊕ T * ) into irreducibles under the Lorenz group SO(3, 1) produces integer spins, which is as desired. Thus, the prolongation of the Dirac operator (220) by an odd degree element of the Clifford algebra appears to be the most mathematically natural way of coupling our fermions to bosons. It remains to be seen if these ideas can be made realistic.
Let us also remark that it is the fact that C in (220) must be of odd degree that made us choose to describe fermions by even degree forms in the first place. Thus, the proposal is that fermions are to be described by forms in Λ even R 7 , while bosons C should be described by Λ odd R 7,7 (possibly self-dual elements thereof, and possibly with some additional restrictions imposed), where we used the identification of the Clifford algebra with the exterior algebra.
If one is to take the described geometric construction seriously, the main open question is what mechanism breaks the SO(7, 7) symmetry down to what is seen in Nature. We do not yet have any satisfactory answer to this question, but there are some hints indicating that it may be possible and natural for such a mechanism to exist, as we now describe.
We first remark that SO(7, 7) is special from all SO(n, n) groups, being the group of largest dimension where a Weyl spinor has a non-trivial stabiliser so that (a dense subset of) the space of Weyl spinors can be viewed as an R * × SO(n, n) orbit, where R * acts by rescaling. The fact that the space of spinors can be given the interpretation of a group coset plays the crucial role in the generalised geometry setup [6], where it is SO(6, 6) that is relevant. In the SO(7, 7) case studied in [17] a unit Weyl spinor is shown to have the stabiliser G 2 × G 2 or G 2 (C), where in the former case both real forms of the exceptional Lie group G 2 can appear. The largest dimension when this phenomenon occurs is n = 7, as a simple comparison of dimensions of relevant spaces shows. We find the coincidence between the dimension of the largest SO(n, n) that admits an orbit interpretation of the space of spinors, and the dimension of SO(n, n) that appears from the SM fermions interpretation striking.
This suggests that the group SO(7, 7) can be broken by a fixed Weyl spinor. As we have said, the possible unbroken subgroups in this case are G 2 × G 2 or G 2 (C). The last of these contains the Lorentz group SL(2, C). Moreover, because G 2 × G 2 ⊂ SO(7) × SO(7) for the compact G 2 (for the non-compact real form of G 2 this relation is G 2 × G 2 ⊂ SO(4, 3) × SO(4, 3)) and G 2 (C) ⊂ SO(7, C), in all cases the stabiliser group defines a split V = U ⊕Ũ and the metric G restricts to non-degenerate metrics in U,Ũ , or a complex metric in U in the case of G 2 (C) stabiliser. This naturally brings the metric geometry into play, something exceptional that only happens in the SO(7, 7) setup. Thus, after the SO(7, 7) symmetry is broken by the split V = U ⊕Ũ, gravity automatically arises as a part of the present geometric construction.
A related way of looking at the SO(7, 7) symmetry breaking is the already discussed necessity of choosing a Lorentz subgroup SO(3, 1) ⊂ SO (7,7). There are many ways of doing this, and it is in the process of making this choice that the metric (=gravity) appears, and also 4 out of 7 coordinates in R 7 are chosen as special. However, to make all these ideas concrete, one has to make the mechanism by which the Lorentz group is selected dynamical. This is only possible if we have a dynamical principle for the bosonic fields C in (220). Presumably, it should be of the first order type CDC, plus a potential for C, but details are still to be worked out.
There is yet another geometric construction that shows that gravity and metric interpretation arises naturally in the context of odd degree differential forms in R 7 is that in [20], [21]. In this references it was shown that (Euclidean signature) gravity in four dimensions can be understood as the dimensional reduction of a theory of differential 3-forms in seven dimensions. The metric interpretation arises because of the classical fact that a generic 3-form in seven dimensions defines a metric, see [20], [21] for more details.
All in all, there are several not obviously related geometric constructions that all suggest that there is a natural metric interpretation of the SO(7, 7) setup, and this interpretation is only possible for n = 7, which is also the largest dimension where the group SO(7, 7) has certain desirable properties. There also appears to be a natural mechanism in place to break the SO(7, 7) symmetry, and the metric interpretation becomes possible after this symmetry is broken. If one adds to this list the geometric construction we described in this paper, which shows that also the SM fermions have an SO(7, 7) interpretation, one is forced to conclude that all roads lead to SO (7,7).