2-spinors, Dirac spinors and spacetime geometry
(private notes - November 19, 2017)
D. Canarutto Dipartimento di Matematica e Informatica “U. Dini”, http://www.dma.unifi.it/˜canarutto
November 19, 2017
Contents
1 Two-spinors 1 1.1 Complex spaces and conjugate spaces ...... 1 1.2 Hermitian tensors ...... 2 1.3 Two-spinor space ...... 4 1.4 2-spinors and Minkowski space ...... 6 1.5 The natural Dirac map ...... 8 1.6 Riemann sphere and celestial sphere ...... 10
2 2-spinors and 4-spinors 12 2.1 4-spinor bases ...... 13 2.2 Conjugation and Dirac adjoint ...... 13 2.3 The Dirac map in 4-spinor terms ...... 14
3 Additional structures 16 3.1 Observers and positive Hermitian metrics ...... 16 3.2 Charge conjugation ...... 18 3.3 Parity and time reversal ...... 19 3.4 Spin operators ...... 20 3.5 Particle momentum in 2-spinor form ...... 24 3.6 Helicity of Dirac Fermions ...... 27
4 Two-spinor bundles 31 4.1 Two-spinor connections ...... 31 4.2 Two-spinor tetrad ...... 36 4.3 Cotetrad ...... 37 4.4 Tetrad and connections ...... 38 4.5 The Dirac operator ...... 40 4.6 Fermi transport of spinors ...... 41 4.7 Flat spacetime ...... 47
I II CONTENTS
5 Optical geometry in curved spacetime 48 5.1 Real optical algebra ...... 48 5.2 Complexified optical algebra ...... 51 5.3 Optical spin ...... 53 5.4 Optical bundles ...... 54
6 2-spinor groups and 4-spinor groups 55 6.1 2-spinor groups ...... 55 6.2 2-spinor groups and Lorentz group ...... 60 6.3 Action of 2-spinor groups on 4-spinor space ...... 65 6.4 Dirac algebra in 2-spinor terms ...... 67 6.5 Clifford group in 2-spinor terms ...... 76 6.6 4-spinor groups in 2-spinor terms ...... 83 6.7 Bundle structure of W ...... 86
A Spinors: further details and calculations 93 A.1 Complex spaces as real spaces ...... 93 A.2 Real volume forms on spinor spaces ...... 94 A.3 Two-spinor exchange formulas ...... 98 A.4 Pauli matrices ...... 99 A.5 Spinors as endomorphisms ...... 101 A.6 Spin groups ...... 102 A.7 4-momentum and spinors ...... 109 A.8 World-spinors ...... 115 A.9 Tensor algebra of world-spinors ...... 117 A.10 Spinor connection ...... 120 A.11 Spinor Laplacian ...... 123
B Clifford algebra 125 B.1 Clifford algebra ...... 125 B.2 Practical Clifford product calculations ...... 126 B.3 Clifford group ...... 129 B.4 Dirac algebra ...... 130 B.5 4-spinors ...... 131 B.6 Complexified spinor groups ...... 133 1
1 Two-spinors
1.1 Complex spaces and conjugate spaces
If A is a set and f : A → C is any map, then the conjugated map f¯ : A → C is defined ¯ 1 ¯ 1 ¯ by f(a) ≡ f(a), a ∈ A. Moreover one sets
V F → V F : λ 7→ λ¯ .
Next, consider bi(anti)dual spaces. Set V ≡ V F F, and call this the conjugate space of V . One obtains the isomorphisms
V =∼ V FF =∼ V F F , V ≡ V F F =∼ V FF .
In fact, any v ∈ V can be viewed as the linear map V F → C : λ 7→ hλ, vi or as the antilinear map V F → C : α 7→ hα,¯ vi; andv ¯ ∈ V can be viewed as the linear map V F → C : α 7→ hα, v¯i ≡ α(v). Summarizing, one has the four spaces
V ↔ V , V F ↔ V F , where the arrows indicate the conjugation anti-isomorphisms. Conjugated, dual and antidual spaces of these yield, up to natural isomorphisms, no new space: when the symbols F , F and are attached to V , in any number and in any order, an odd (even) number of ‘barred’ symbols yields a (non-)barred space, an odd (even) number of starred symbols yields a (non-)starred space. In particular, V F can be written as V F.
Remark. The symbols F , F and together with the identity constitute the Abelian group Z2 × Z2 (§??). Remark. According to another definition, the conjugate space of V is introduced as the complex vector space, on the same underlying set as V , where the product of a number c by a vector v is given byc ¯ v. This works also for infinite dimensional spaces; when the dimension is finite this definition is equivalent to the above one (the two spaces are isomorphic), but gives rise to notational complications, in particular with regard to coordinate expressions. Also, it seems preferable to maintain the standard use that, for any complex-valued function f and complex number c, the notation c f stands for the map a 7→ c f(a) , and this would conflict with the alternative definition. On the other hand, note that one could also introduce V as the vector space of all antilinear maps C → V . One has the usual identifications between tensor products and spaces of (multi)linear maps. In particular
L(V , W ) =∼ W ⊗ V F , A(V , W ) =∼ L(V , W ) =∼ W ⊗ V F . 2 1 TWO-SPINORS
The conjugation map can be extended to any tensor product by means of the universal property (namely it is applied to each factor). Contractions in tensor products are defined when the contracted factors are either V and V F, or V and V F. A F Let (bA), 1 ≤ A ≤ dim V , be a basis of V , and (b ) its dual basis of V . The corresponding indices in the conjugate spaces are distinguished by a dot, namely one writes
¯ ¯A˙ A bA˙ ≡ bA , b ≡ b ,
¯ ¯A˙ F F so that {bA˙} is a basis of V and {b } its dual basis of V . For v ∈ V and λ ∈ V one has
A A˙ ¯ v = v bA , v¯ =v ¯ bA˙ , A ¯ ¯ ¯A˙ λ = λAb , λ = λA˙b ,
A˙ A ¯ wherev ¯ = v , λA˙ ≡ λA and Einstein summation convention is used. Remark. In some literature one finds also the convention of writing vA˙ instead ofv ¯A˙, and the like. This makes many formulas appear simpler, but in some cases may give rise to inconsistencies. On the other hand, one may question the utility of the ‘dotted indices’ con- ¯A ¯ vention: one might write simplyv ¯ = λ bA and the like, possibly avoiding misunderstandings, in complicate formulas, by explicitely carrying along the involved basis. However any index convention, including the standard one for Riemannian geometry (high and low indices), is never of a fundamental nature: its role is just that of making calculations easier and formulas more readable. If K is the matrix of an automorphism of V , then K¯ , K F (transposed matrix) and K¯ F (antitransposed matrix) are the matrices of the induced automorphisms of V , V F and V F, respectively. If τ is a tensor with indices of various types, then its components transform with the above matrices, each one for the appropriate type. The indices ofτ ¯ are all of dot-reversed type. Contractions in tensor products are indicated by the usual Einstein summation convention. Observe that dotted indices cannot be contracted with non-dotted indices.
1.2 Hermitian tensors Consider the space V ⊗ V ; this has a natural real linear (complex anti-linear) involution w 7→ w†, which on decomposable tensors reads
(u ⊗ v¯)† = v ⊗ u¯ .
Hence one has the natural decomposition of V ⊗ V into the direct sum of the real eigenspaces of the involution with eigenvalues ±1, respectively called the Hermitian and anti-Hermitian subspaces, namely V ⊗ V = (V ∨¯ V ) ⊕ i (V ∨¯ V ) . In other terms, the Hermitian subspace V ∨¯ V is constituted by all w ∈ V ⊗ V such that w† = w, while an arbitrary w is uniquely decomposed into the sum of an Hermitian and an anti-Hermitian tensor as 1 † 1 † w = 2 (w + w ) + 2 (w − w ) . 1 Note that w is Hermitian (resp. anti-Hermitian) iff its real part (V ∨¯ V )∗ =∼ V F ∨¯ V F , (i V ∨¯ V )∗ =∼ i V F ∨¯ V F , where ∗ denotes the real dual. A Hermitian 2-form is defined to be a Hermitian tensor h ∈ V F ∨¯ V F. The associated quadratic form v 7→ h(v, v) is real-valued. The notions of signature and non-degeneracy of Hermitian 2-forms are introduced similarly to the case of real bilinear forms. If h is non- degenerate then it yields the isomorphism h[ : V → V F :v ¯ 7→ h(¯v, ); its conjugate map is an anti-isomorphism V → V F which, via composition with the canonical conjugation, can be seen as the isomorphism h¯[ : V → V F : v 7→ h( , v). The inverse isomorphisms h# and h¯# are similarly related to a Hermitian tensor h−1 ∈ V ∨¯ V . One has the coordinate expressions ¯A˙ A −1 AA˙ ¯ h = hAA˙b ⊗ b , h = h bA˙⊗ bA , [ A˙ ¯[ B ¯ B (h (¯v))B = hA˙B v¯ , (h (v))A˙ = hA˙B v = hBA˙v , # ¯ B A˙B ¯ ¯# A˙ A˙B ¯BA˙ (h (λ)) = h λA˙ , (h (λ)) = h λB = h λB , where C˙A A A˙C A˙ h hC˙B = δ B , h hB˙C = δ B˙ . Namely one has an index lowering and raising formalism similar to the pseudo-Euclidean case, but the operation changes the index’ kind from dotted to non-dotted and vice-versa. I’ll use the notations1 v† ≡ v¯[ ≡ v[ ∈ V F , λ† ≡ λ¯# ≡ λ† ∈ V . The Hermitian conjugate of f ∈ End(V ) =∼ V ⊗ V F with respect to h is defined to be f † ≡ h# ◦ f¯F ◦ h¯[ ∈ End(V ) , where f¯F ∈ End(V F) is the anti-transpose of f. In coordinates † A C˙A ¯D˙ (f ) B = h hD˙B f C˙ . Equivalently, the definition of f † can be expressed as h(f †u, v) = h(u, fv) , u, v ∈ V , h(f †u)†, vi = hu†, fvi , h(f †u)†, vi = hf Fu†, vi i.e. (f †u)† = f Fu† . Obviously, one has (f †)† = f . 1Usually, particularly in the physics literature, t† and t denote objects which are related by their matrices being the Hermitian conjugate (i.e. transposed conjugate) of each other. For the Hermitian involution in V ⊗ V this is true in any basis. On the other hand, for Hermitian conjugation determined by a Hermitian metric, this relation between matrices of corresponding objects holds in an orthonormal basis. 4 1 TWO-SPINORS The linear map f ∈ End(V ) is said to be unitary with respect to h if it fulfils h ◦ (f × f) = h , i.e. h(fu)†, fvi = hu†, vi . Unitarity of f can be equivalently expressed as † † f ◦ f = f ◦ f = 11V , since2 hu†, vi = h(fu)†, fvi = hu†, f †fvi . Moreover, a unitary f fulfils | det f|2 = 1 . 1.3 Two-spinor space Let S be a 2-dimensional complex vector space. Set Λ2 ≡ ∧2S, a 1-dimensional complex vector space, and identify Λ−2 ≡ Λ2F with ∧2SF through the rule3 0 1 0 2 F 0 ω(s ∧ s ) ≡ 2 ω(s, s ) , ∀ ω ∈ ∧ S , s, s ∈ S , 0 1 0 0 s ∧ s ≡ 2 (s⊗s −s ⊗s) . Then Λ−2 can be viewed as the space of ‘symplectic’ forms on S. Any ω ∈ Λ−2 \{0} has a unique ‘inverse’ or ‘dual’ element ω−1 ≡ ωF, i.e. ω(ω−1) = 1 . Denote by ω[ : S → SF the linear map given by hω[(s), ti ≡ ω(s, t) , and by ω# : SF → S the linear map given by hµ, ω#(λ)i ≡ ω−1(λ, µ) . Then ω# = −(ω[)−1 . Namely the assignment of a non-zero element of Λ2 determines an index lowering and raising formalism, in which however one has to pay some attention to signs. The Hermitian subspace of Λ2 ⊗ Λ2 is a real vector space with a distinguished orientation, whose positively oriented semispace 2 2 2 + 2 L ≡ (Λ ∨¯ Λ ) ≡ {w ⊗ w,¯ w ∈ Λ } has the square root semi-space L, called the space of length units. Next, consider the complex 2-dimensional space −1/2 U ≡ L ⊗ S . This is our 2-spinor space. Observe that the 1-dimensional space 2 −1 2 Q ≡ ∧ U = L ⊗ ∧ S 2 If V is a finite-dimensional vector space and A, B ∈ End(V ) are such that A ◦ B = 11V , then obviously A −1 −1 −1 −1 −1 and B are automorphisms of V and B ◦ A = (A ◦ B) = 11V . Then also B ◦ A = B ◦ (B ◦ A ) ◦ A = 11V . 3 This contraction, defined in such a way to respect the usual conventions in two-spinor literature, corre- sponds to half standard tensor-algebra contraction and 1/4 standard exterior-algebra contraction. 1.3 Two-spinor space 5 has a distinguished Hermitian metric, defined as the unit element in 2 F 2 F 2 2 2 ∼ (∧ U ) ∨¯ (∧ U ) = L ⊗ (Λ ∨¯ Λ ) = R . Hence there is the distinguished set of normalized symplectic forms on U, any two of them dif- fering by a phase factor. One says that elements of U and of its tensor algebra are ‘conformally r invariant’, while tensorializing by L one obtains ‘conformal densities’ of weight r. Consider an arbitrary basis (ξA) of S and its dual basis (xA) of SF . This determines the mutually dual bases AB 2 w ≡ ε ξA ∧ ξB = 2 ξ1 ∧ ξ2 of Λ , ∗ A B −2 ∼ 2F w ≡ εAB x ∧ x = 2 x1 ∧ x2 of Λ = Λ . AB where ε and εAB both denote the antisymmetric Ricci matrix, and the basis √ l ≡ w ⊗ w¯ of L . Then one also has the induced normalized basis −1/2 (ζA) ≡ (l ⊗ ξA) of U , and its dual basis (zA) ≡ (l1/2 ⊗ xA) of U F ; the induced mutually dual bases −1 A B −1 2 F ε ≡ l ⊗ w = εAB z ∧ z ∈ Q ≡ ∧ U and ∗ −1 −1 AB 2 ε ≡ ε ≡ l ⊗ w = ε ζA ∧ ζB ∈ Q ≡ ∧ U are then normalized too. The mutually inverse symplectic forms ε and ε∗ determine isomorphisms ε[ : U → U F , ε# : U F → U by the rules hε[(u), vi ≡ ε(u, v) , hε#(λ), µi ≡ ε∗(λ, µ) , namely4 [ 1 # ∗ 1 ∗ ε (u) = ucε = 2 u|ε , ε (λ) = λcε = 2 λ|ε . These conventions can be checked by coordinate expressions. We have5 A B 1 A B B A A B ε = εAB z ∧ z = 2 εAB (z ⊗ z − z ⊗ z ) = εAB z ⊗ z = = 2 z1 ∧ z2 = z1 ⊗ z2 − z2 ⊗ z1 , [ A B 1 2 2 1 A B 1 2 2 1 ε (u) = εAB u z = u z − u z , ε(u, v) = εAB u v = u v − u v , 4See also the above comment about duality and footnote 3. 5Remember (see [Go69], §I.6) that u|(λ ∧ µ) = (u|λ) ∧ µ + (−1)|λ| λ ∧ (u|µ). 6 1 TWO-SPINORS and similar expressions for ε#, with inverted index positions. In particular we have [ 2 [ 1 # 1 # 2 ε (ζ1) = z , ε (ζ2) = −z , ε (z ) = ζ2 , ε (z ) = −ζ1 . For u ∈ U and λ ∈ U F we write u[ ≡ ε[(u) ∈ U F , λ# ≡ ε#(λ) ∈ U , [ A C # C BC uB ≡ (u )B = u εAB , λ ≡ (λ ) = λB ε . Noting that6 BC C A BC A εAB ε = −δA ⇒ u εAB ε = −u we see that (u[)# = −u . Similarly (λ#)[ = −λ . Remark. In contrast to the usual 2-spinor formalism, the ‘symplectic’ 2-form ε is not a- AB priori fixed. Two-spinor indices may be lowered and raised by εAB and its inverse ε , but this operation depends from the chosen ε, hence on the chosen two-spinor basis (ξA). Moreover note that we use the convention that indices are moved by contraction with the first index in ε —indeed we’ll consistently use the same conventions also for other instances of index-moving operations determined by invertible 2-tensors. Then we find that ε[ and ε# are minus inverse isomorphisms. 4 Remark. An even more specialized space is the neutrino space, obtained by choosing a square root Q1/2 of Q as −1/2 −1 Uν ≡ Q ⊗ U = Λ ⊗ S . On Uν , one has a distinguished symplectic 2-form. 4 1.4 2-spinors and Minkowski space The fact that a normalized element ε ∈ QF is unique up to a phase factor, implies that there exists an intrinsic, well-defined object ε ⊗ ε¯ ∈ QF ⊗ QF; this can be also seen as a natural bilinear formg ˜ on U ⊗ U, given for decomposable elements by g˜(p ⊗ q,¯ r ⊗ s¯) = ε(p, r)ε ¯(¯q, s¯) . The fact that any ε is non-degenerate implies thatg ˜ is non-degenerate too. In terms of index formalism, one can express this by saying that lowering and raising of couples (AA˙) of mutually conjugate indices, in a normalized basis of U, corresponds to an intrinsic operation. In a normalized 2-spinor basis (ζA) , one writes g˜AA˙BB˙ = εAB ε¯A˙B˙ . AA˙ ¯ Namely, let w = w ζA ⊗ ζA˙ ∈ U ⊗ U ; then AA˙ BB˙ AA˙ 11 22 12 21 g˜(w, w) = εAB ε¯A˙B˙w w = 2 det w = 2 (w w − w w ) . 1 Remark. The fact thatg ˜ (w) ≡ 2 g˜(w, w) can be written as the determinant of the matrix of w leads one to consider how is this determinant an intrinsecally defined object. First, observe 6In the usual index moving formalism of (pseudo)Riemannian formalism one tends to regard tensors as objects defined independently of the positions of their indices. In other situations, like in the formalism under consideration, such identifications must be handled with greater care. 1.4 2-spinors and Minkowski space 7 that a linear map between different vector spaces has a well-defined determinant relatively to any fixed isomorphism between the two spaces. The case here is just slightly more involved: an element w ∈ U ⊗ U can be viewed as an anti-linear map U F → U, and these two spaces are also related by the anti-isomorphism induced by ε ; the latter is unique up to a phase factor, which is actually compensated by the anti-linearity. Next, consider the Hermitian subspace H ≡ U ∨¯ U ⊂ U ⊗ U . This is a 4-dimensional real vector space; for any given normalized basis (ζA) of U consider, in particular, the Pauli basis (τλ) of H associated with (ζA), namely √1 τλ ≡ 2 σλ ≡ AA˙ ¯ √1 AA˙ ¯ ≡ τ λ ζA ⊗ ζA˙ ≡ 2 σλ ζA ⊗ ζA˙ , AA˙ where (σλ ) denotes the λ-th Pauli matrix (§A.4). The restriction ofg ˜ to the Hermitian subspace H (denoted still with the same symbol) turns out to be a Lorentz metric with signature (+, −, −, −); actually, a Pauli basis is readily seen to be orthonormal,7 that is 0 0 g˜λµ ≡ g˜(τλ , τµ) = ηλµ ≡ 2 δλδµ − δλµ . Moreover the Lorentz metricg ˜ determines, up to sign, a normalized volume formη ˜ on H. AA˙ ¯ One might also say that (σλ ) is the matrix of 11H in the bases (σλ) and (ζA ⊗ ζA˙) ; sim- AA˙ ¯ ilarly, (τ λ ) is the matrix of 11H in the bases (τλ) and (ζA ⊗ ζA˙) . Then one can apply the above said ε-related index moving formalism to these components, and also move (“space- time”) greek indices viag ˜λµ . λ Let (t ) be the (real) dual basis of (τλ) . Then one obtains λ √1 λ A A˙ t ≡ 2 σ AA˙z ⊗ ¯z , λ √1 λ λ λ AA˙ λ namely t AA˙ = 2 σ AA˙ = τ AA˙, since σ AA˙σµ = 2δ µ as can be easily checked using the formulas reported in §A.4. One also writes AA˙ BB˙ g˜λµ = τ λ τ µ εAB ε¯A˙B˙ , org ˜AA˙BB˙ = εAB ε¯A˙B˙ . Proposition 1.1 An element w ∈ U ⊗ U = C ⊗ H is null, that is g˜(w, w) = 0 , iff it is a decomposable tensor: w = u ⊗ s¯, u, s ∈ U . proof: A decomposable element u ⊗ s¯ ∈ U ⊗ U is immediately seen to be null (isotropic). In AA˙ ¯ order to show the converse, write w = w ζA ⊗ ζA˙ in a normalized 2-spinor basis. Then (see above) 0 =g ˜(w, w) = det wAA˙ , hence the two rows of the matrix wAA˙ are proportional, 1A˙ 2A˙ 1A˙ ¯ say w = c w , c ∈ C . One then gets w = (ζ1 + c ζ2) ⊗ (w ζA˙). 7 Note that τ0 is a Hermitian metric on U F; conversely, any ε-normalized positive Hermitian metric can be seen as a unit timelike future-pointing vector in H, namely an ‘observer’. 8 1 TWO-SPINORS A null element in U ⊗ U is also in H iff it is of the form ±u ⊗ u¯. Hence the null cone N ⊂ H is constituted exactly by such elements. Note how this fact yields a way of distinguish between time orientations: by convention, one chooses the future and past null-cones in H to be, respectively, N + ≡ {u ⊗ u,¯ u ∈ U} , N − ≡ {−u ⊗ u,¯ u ∈ U} . Proposition 1.2 For each orthonormal positively oriented basis (eλ) , such that e0 is future- oriented, there exists a normalized 2-spinor basis (ζA) whose associated Pauli basis (τλ) coin- cides with (eλ) . proof: The elements e0+e3 and e0−e3 are isotropic vectors, hence there exist u, v ∈ U such that √ √ e0 + e3 = 2 u ⊗ u¯ , e0 − e3 = 2 v ⊗ v¯ , and one gets 1 √1 e0 = 2 (e0 + e3) + (e0 − e3) = 2 (u ⊗ u¯ + v ⊗ v¯) , 1 √1 e3 = 2 (e0 + e3) − (e0 − e3) = 2 (u ⊗ u¯ − v ⊗ v¯) . 2 Moreover, one has 1 =g ˜(e0 , e0) = |ε(u, v)| . By multiplying either u or v by an appropriate phase factor, one may redefine them so that ε(u, v) = 1 . The elements 0 √1 0 √i e1 ≡ 2 (u ⊗ v¯ + v ⊗ u¯) , e2 ≡ 2 (−u ⊗ v¯ + v ⊗ u¯) are in the 2-dimensional space orthogonal to e0 and e3 . Now one can redefine u and v , while iθ/2 −iθ/2 leaving all the above relations untouched, by multiplying u by e and v by e , θ ∈ R . One gets the transformation 0 √1 iθ −iθ 0 0 e1 7→ 2 (e u ⊗ v¯ + e v ⊗ u¯) = cos θ e1 − sin θ e2 , 0 √i iθ −iθ 0 0 e2 7→ 2 (−e u ⊗ v¯ + e v ⊗ u¯) = sin θ e1 + cos θ e2 . 0 0 Since this is an arbitrary orientation preserving orthogonal transformation in the (e1 , e2)- plane, for an appropriate value of θ one gets (e1 , e2). One point worth stressing, related to the above proposition, is that any future-pointing timelike vector can√ be written as u ⊗ u¯ + v ⊗ v¯ , for suitable u, v ∈ U . Its Lorentz pseudo-norm is then given by 2 |ε(u, v)| . 1.5 The natural Dirac map Next observe that an element of U ⊗ U can be seen as a linear map U F → U, while an element of U F ⊗ U F can be seen as a linear map U → U F. Then one defines the linear map √ γ˜ : U ⊗ U → End(U ⊕ U F): y 7→ γ˜(y) ≡ 2 y, y[F , √ i.e. γ˜(y)(u, χ) = 2 ycχ , ucy[ , where y[ ≡ g˜[(y) . 1.5 The natural Dirac map 9 In particular for a decomposable y = p ⊗ q¯ one has √ γ˜(p ⊗ q¯)(u, χ) = 2 hχ, q¯i p , hp[, ui q¯[ = √ = 2 −ε¯(χ#, q¯) p , ε(p, u)ε ¯[(¯q) . One obtains the coordinate expression √ AB˙ B A˙ γ˜(y)(u, χ) = 2 y χB˙ζA , y¯A˙B u ¯z = √ AB˙ B A˙ C˙D = 2 y χB˙ζA , yBA˙u ¯z , yBA˙ =y ¯A˙B ≡ ε¯C˙A˙ε¯DB y¯ . Proposition 1.3 γ˜ is a Clifford map; then, its restriction to H is a Dirac map. proof: From formula 2 in §A.3, for any p, q, r ∈ U one has ε(p, q) r[ + ε(q, r) p[ + ε(r, p) q[ = 0 . In order to prove thatγ ˜ is a Clifford map it suffices to show that the Clifford identity γ˜(y) ◦ γ˜(y0) +γ ˜(y0) ◦ γ˜(y) = 2g ˜(y, y0) 11holds for any two elements y, y0 of one basis of U ⊗ U ; since there exist bases constituted by null (isotropic) elements,8 it suffices to show that the Clifford identity holds for any two null elements in U ⊗ U ; such elements are just the decom- posable ones, so set y = p ⊗ q¯, y0 = r ⊗ s¯, with p, q, r, s ∈ U . One gets (see also the following footnote 9) √ γ˜(p ⊗ q¯) ◦ γ˜(r ⊗ s¯)(u, χ) = 2γ ˜(p ⊗ q¯) −ε¯(χ#, s¯) r, ε(r, u)s ¯[ = = 2 ε(r, u)ε ¯(¯s, q¯) p +ε ¯(¯s, χ#) ε(p, r)q ¯[ . Hence [˜γ(p ⊗ q¯) ◦ γ˜(r ⊗ s¯) +γ ˜(r ⊗ s¯) ◦ γ˜(p ⊗ q¯)](u + χ) = = 2 ε(r, u)ε ¯(¯s, q¯) p +ε ¯(¯s, χ#) ε(p, r)q ¯[ + ε(p, u)ε ¯(¯q, s¯) r +ε ¯(¯q, χ#) ε(r, p)s ¯[ = = 2ε ¯(¯q, s¯)[ε(u, r) p + ε(p, u) r] + 2 ε(p, r) [¯ε(¯s, χ#)q ¯[ +ε ¯(χ#, q¯)s ¯[] = = 2ε ¯(¯q, s¯)[−ε(r, p) u] + 2 ε(p, r)[−ε¯(¯q, s¯)(χ#)[] = = 2 ε(p, r)ε ¯(¯q, s¯)(u, χ) = 2g ˜(p ⊗ q,¯ r ⊗ s¯)(u, χ) . Recalling §B.1, we see that the condition thatγ ˜ is a Clifford map can be equivalently stated as γ˜[w] ◦ γ˜[w] =g ˜(w, w) 11W , ∀ w ∈ U ⊗ U . The above relation can be easily checked in components as one gets