The Geometry of Generalised Spinʳ Spinors on Projective Spaces

In this paper, we adapt the characterisation of the spin representation via exterior forms to the generalised spinʳ context. We find new invariant spinʳ spinors on the projective spaces ℂℙⁿ, ℍℙⁿ, and the Cayley plane ℙ² for all their homogeneous realisations. Specifically, for each of these realisat...

Full description

Saved in:
Bibliographic Details
Published in:Symmetry, Integrability and Geometry: Methods and Applications
Date:2025
Main Authors: Artacho, Diego, Hofmann, Jordan
Format: Article
Language:English
Published: Інститут математики НАН України 2025
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/212874
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:The Geometry of Generalised Spinʳ Spinors on Projective Spaces. Diego Artacho and Jordan Hofmann. SIGMA 21 (2025), 017, 32 pages

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860297403780825088
author Artacho, Diego
Hofmann, Jordan
author_facet Artacho, Diego
Hofmann, Jordan
citation_txt The Geometry of Generalised Spinʳ Spinors on Projective Spaces. Diego Artacho and Jordan Hofmann. SIGMA 21 (2025), 017, 32 pages
collection DSpace DC
container_title Symmetry, Integrability and Geometry: Methods and Applications
description In this paper, we adapt the characterisation of the spin representation via exterior forms to the generalised spinʳ context. We find new invariant spinʳ spinors on the projective spaces ℂℙⁿ, ℍℙⁿ, and the Cayley plane ℙ² for all their homogeneous realisations. Specifically, for each of these realisations, we provide a complete description of the space of invariant spinʳ spinors for the minimum value of for which this space is non-zero. Additionally, we demonstrate some geometric implications of the existence of special spinʳ spinors on these spaces.
first_indexed 2026-03-21T18:30:54Z
format Article
fulltext Symmetry, Integrability and Geometry: Methods and Applications SIGMA 21 (2025), 017, 32 pages The Geometry of Generalised Spinr Spinors on Projective Spaces Diego ARTACHO a and Jordan HOFMANN b a) Imperial College London, London SW7 2AZ, UK E-mail: d.artacho21@imperial.ac.uk b) King’s College London, London WC2R 2LS, UK E-mail: jordan.2.hofmann@kcl.ac.uk Received July 01, 2024, in final form March 01, 2025; Published online March 11, 2025 https://doi.org/10.3842/SIGMA.2025.017 Abstract. In this paper, we adapt the characterisation of the spin representation via ex- terior forms to the generalised spinr context. We find new invariant spinr spinors on the projective spaces CPn, HPn, and the Cayley planeOP2 for all their homogeneous realisations. Specifically, for each of these realisations, we provide a complete description of the space of invariant spinr spinors for the minimum value of r for which this space is non-zero. Additionally, we demonstrate some geometric implications of the existence of special spinr spinors on these spaces. Key words: special spinors; projective spaces; generalized spin structures; spinc; spinh 2020 Mathematics Subject Classification: 53C27; 15A66; 57R15 1 Introduction A topic of major interest in differential geometry is the existence or non-existence of special G-structures on a given smooth manifold M ; classical examples include Riemannian, complex, symplectic, and spin structures. Spin geometry, in particular, gives a way of accessing global geometric information about Riemannian spin manifolds via sections of a certain naturally defined vector bundle called the spinor bundle. Indeed, for a Riemannian spin manifold M , there are a number of results of the form: M carries a spinor satisfying equation E =⇒ M has geometric property P. For example, it is well known that a manifold carrying a non-zero parallel spinor is Ricci- flat, and, more generally, that the existence of a non-zero real (resp. purely imaginary) Killing spinor implies that the metric is Einstein with positive (resp. negative) scalar curvature [9, 15]. Other notable examples include the bijection between generalised Killing spinors in dimension 5 (resp. 6, resp. 7) and hypo SU(2)-structures (resp. half-flat SU(3)-structures, resp. co-calibrated G2-structures) [1, 12], and the spinorial description of isometric immersions into Riemannian space forms [10, 16, 30]. However, not every manifold can be endowed with a spin structure; the question then natu- rally arises as to whether one can apply the tools of spin geometry to non-spin manifolds. The answer is affirmative, and there are several possible approaches. The unifying idea is to consider suitable enlargements of the spin groups, i.e., Lie groups Ln equipped with homomorphisms Spin(n) ιn−→ Ln pn−→ SO(n) such that pn ◦ ιn is the usual two-sheeted covering Spin(n) → SO(n). Hence, an oriented Riemannian n-manifold admitting a lift of the structure group to Ln is a weaker condition than mailto:d.artacho21@imperial.ac.uk mailto:jordan.2.hofmann@kcl.ac.uk https://doi.org/10.3842/SIGMA.2025.017 2 D. Artacho and J. Hofmann being spin. Following ideas by Friedrich and Trautman [20], the so-called spinorial Lipschitz structures have garnered much attention in recent years [33, 34, 35]. These naturally arise by following the inverse approach: starting with a suitable generalisation of the concept of spinor bundle, one investigates the enlargement Ln of Spin(n) to which this bundle corresponds. These Ln are called Lipschitz groups. Another choice of Ln was introduced by Espinosa and Herrera [14], who had the idea of spinorially twisting the spin group. In our setting, this corresponds to taking, for r ∈ N, the groups Lrn = Spinr(n) := (Spin(n)× Spin(r))/⟨(−1,−1)⟩ with the obvious homomorphisms. We say that an oriented Riemannian n-manifold is spinr if it admits a lift of the structure group to Spinr(n). The case r = 1 is the classical spin case, and the cases r = 2, 3 give rise to spinC and spinH geometry respectively, which have been a fruitful field of study over the past decades – see [17, 37] for spinC and [25, 31] and references therein for spinH. These structures have been characterised topologically by Albanese and Milivojević in [4], where they show that a manifold is spinr if, and only if, it can be embedded into a spin manifold with codimension r. In [5], Lawn and the first author focused on the study of spinr structures on homogeneous spaces, establishing a bijection between G-invariant spinr structures on G/H and certain representation-theoretical data – see Theorem 2.6. Analogously to the usual spin case, from a given spinr structure one can construct, for each odd m ∈ N, the so-called m-twisted spinr spinor bundle – see Definition 2.8. Its sections are called m-twisted spinr spinors or simply spinr spinors, and, as in the classical case, they encode geometric information: for a spinr manifold M , there are results of the form: M carries a spinr spinor satisfying equation E =⇒ M has property P. Some of these results can be found in [14, 26], for example (see Theorem 2.16 for more explicit results): � The existence of a generalised Killing spinr spinor ensures a certain decomposition of the Ricci tensor [14, Theorem 3.3]. � The existence of a parallel pure spin2 ( resp. spin3 ) spinor implies that the manifold is Kähler (resp. quaternionic Kähler) [26, Corollaries 4.10 and 4.12]. In this paper, we illustrate how invariant twisted spinr spinors on the projective spaces CPn, HPn and the Cayley plane OP2 encode different geometric properties of these manifolds. To this end, we proceed as follows: (1) Consider a homogeneous realisation M = G/H of the corresponding space, equipped with a generic G-invariant metric. (2) Find the minimum value of r ∈ N such that M has a G-invariant spinr structure carrying non-trivial invariant m-twisted spinr spinors, for some odd m ∈ N, and describe the space of such spinors. (3) Study the geometric properties ofM encoded by those invariant spinr spinors which satisfy additional algebraic properties. The realm of projective spaces provides a fruitful ground for study. In particular, we prove that Friedrich’s construction of generalised Killing spinors on CP3 [9, p. 146] cannot be gener- alised to higher complex dimensions, showing that this is the only dimension for which CP2k+1 has an Sp(k + 1)-invariant metric carrying non-trivial invariant generalised Killing spinors. We The Geometry of Generalised Spinr Spinors on Projective Spaces 3 M n G r m dim(Σm∗,r)inv Special spinors Geometry CPn k SU(k + 1) 2 1 2 pure, parallel Kähler–Einstein 2k + 1 Sp(k + 1) 2, if k even 1 2 pure, parallel Kähler–Einstein 1, if k odd 1 2 generalised Killing Einstein, nearly Kähler (n = 3 (†)) HPn 2k + 1 Sp(2k + 2) 3 2k + 1 1 pure, parallel quaternionic Kähler 2k Sp(2k + 1) – – – – – OP2 – F4 9 3 4 – – Table 1. For each compact, simple and simply connected Lie group G acting transitively on M : the minimum values of r, m such that M admits a G-invariant spinr structure carrying a non-zero invariant m-twisted spinr spinor, and the geometric significance of these. For †, see [9, p. 146]. also find the spinH spinor on HPn inducing the standard quaternionic Kähler structure (see [26, Corollary 4.12]) with new representation-theoretical methods, and we show that this is, up to scaling, the only Sp(n+ 1)−invariant spinH spinor on this space. Finally, we find that the min- imum values of r and m such that OP2 carries non-trivial F4-invariant m-twisted spinr spinors are r = 9 and m = 3, and the space of such spinors is four-dimensional. The computations carried out in this paper illustrate an extension of the differential forms approach to the spin representation (see, e.g., [2]) to the context of spinr structures. These techniques allow us to express and manipulate complicated twisted spinors in an easy and readable way, finding new examples of special spinr spinors. The main contribution of this paper is, then, the fusion of the differential forms approach with the power of spinr geometry to encode geometric properties of manifolds which are not necessarily spin. Our results are summarised as follows. Theorem. Let G be a compact, simple and simply connected Lie group acting transitively on the projective space M = CPn,HPn or OP2. Then, the minimum values of r,m ∈ N (with m odd) such that M admits a G-invariant spinr structure carrying a non-zero invariant m- twisted spinr spinor are shown in Table 1, together with the geometric information such spinors encode. 2 Preliminaries We begin by introducing the necessary background definitions and results concerning spin and spinr geometry within the context of homogeneous spaces. For an introduction to spin geometry we refer the reader to [9, 32], for spinr manifolds to [4, 5, 14], and for homogeneous spaces to [6]. 2.1 Invariant metrics on reductive homogeneous spaces Let G/H be a reductive homogeneous space with reductive decomposition g = h ⊕ m, where g is the Lie algebra of G, h is the Lie algebra of H and m is an AdH -invariant complement of h in g. Suppose that the adjoint representation of H on m – which, under the usual identifications, corresponds to the isotropy representation of the homogeneous space – decomposes as a direct sum of irreducible components m = m1 ⊕ · · · ⊕mk. We would like to find all the AdH -invariant inner products on m (which correspond to G-invariant metrics on the homogeneous space G/H, see, e.g., [6]). Of course, such metrics need not exist. However, ifH is compact, using Weyl’s trick one readily sees that they do exist. We would like to show that invariant inner products on each irreducible component are unique up to positive scaling, and that any invariant inner product on m is a positive linear combination of invariant inner products on the irreducible components, 4 D. Artacho and J. Hofmann yielding a k-parameter family of invariant metrics. This is of course false in general (consider two copies of the same irreducible representation admitting an invariant metric). However, this is essentially the only obstruction, as we shall see now. This is a very important result which appears to be well known, but it is surprisingly hard to find in the literature. We include it here with a full proof. Proposition 2.1. Let g be a finite-dimensional real Lie algebra and ρ : g → EndR(V ) a finite- dimensional irreducible real representation of g. Suppose that there exists a ρ-invariant inner product on V . Then, it is unique up to positive scaling. Proof. Let B1, B2 : V ×V → R be two ρ-invariant inner products on V , i.e., two inner products on V satisfying ∀X ∈ g, ∀v, w ∈ V : Bi(ρ(X)v, w) +Bi(v, ρ(X)w) = 0, i = 1, 2. Then, for each i = 1, 2, Bi defines an isomorphism of representations φi between ρ and its dual representation ρ∗ : g → EndR(V ∗): φi : V → V ∗, v 7→ Bi(v,−). In particular, ρ is self-dual. Now consider the endomorphism of representations given by φ−1 1 ◦ φ2 : V → V . By Schur’s lemma, the endomorphism ring of an irreducible representation (over any ground field) is a division ring. In particular, the endomorphism ring of ρ is a finite- dimensional associative division algebra over R. By the Frobenius theorem, these are, up to isomorphism, R, C and H. So we need to consider these three cases separately. (1) If End(ρ) ∼= R, then there exists λ ∈ R such that φ−1 1 ◦ φ2 = λ IdV , which implies that φ2 = λφ1, which in turn means that B2 = λB1. As both B1, B2 are inner products, we must have that λ > 0. This completes the proof for the case End(ρ) ∼= R. (2) If End(ρ) ∼= C, then there exists J ∈ End(ρ), with J2 = − IdV . And any φ ∈ End(ρ) is of the form φ = a IdV +bJ , for some a, b ∈ R. In particular, φ2 = aφ1 + bφ1 ◦ J . But this implies that B2(−,−) = aB1(−,−)+bB1(J−,−). Now we claim that, for every v, w ∈ V , B1(Jv, Jw) = B1(v, w). Indeed, define B̃ : V × V → R as B̃(v, w) = B1(Jv, Jw). As J2 = − IdV , B̃ is non- degenerate. And, as J and B1 are ρ-invariant and B1 is symmetric, B̃ is ρ-invariant and symmet- ric. Now define φ̃ : V → V ∗ by φ̃(v) = B̃(v,−). Consider the endomorphism of representations φ−1 1 ◦ φ̃. Then, there exist c, d ∈ R such that φ−1 1 ◦ φ̃ = c IdV +dJ . Hence, for every v, w ∈ V , B̃(v, w) = cB1(v, w) + dB1(Jv,w). As B̃, B1 are symmetric, we have that, for every v, w ∈ V , dB1(Jv,w) = dB1(v, Jw). Suppose d ̸= 0. Then, for every v, w ∈ V , B1(Jv,w) = B1(v, Jw). In particular, if v ̸= 0, we would have that B1(Jv, Jv) = B1 ( v, J2v ) = −B1(v, v) < 0, which con- tradicts positive-definiteness of B1. Hence, B̃ = cB1, for some c ∈ R. By positive-definiteness and non-degeneracy, c > 0. Moreover, if v ̸= 0, B1(v, v) = B1 ( J2v, J2v ) = c2B1(v, v). Hence, c = 1. This completes the proof of the fact that, for every v, w ∈ V , B1(Jv, Jw) = B1(v, w). Equivalently, B1(J−,−) is skew-symmetric. Now, back to our previous situation. As B1 and B2 are symmetric and B1(J−,−) is skew- symmetric, b = 0. And now positive-definiteness implies that a > 0, finishing the proof for the case End(ρ) ∼= C. (3) Finally, suppose End(ρ) ∼= H. Then, there exist J1, J2, J3 ∈ End(ρ) linearly independent, with J2 i = − IdV for i = 1, 2, 3 and J1J2J3 = − IdV , such that any element φ ∈ End(ρ) is of the form a IdV +b1J1 + b2J2 + b3J3, for some a, b1, b2, b3 ∈ R. Hence, in particular, φ−1 1 ◦ φ2 is of this form, and thus B2(−,−) = aB1(−,−) + b1B1(J1−,−) + b2B1(J2−,−) + b3B1(J3−,−), for some a, b1, b2, b3 ∈ R. As in the previous case, one shows that, for every i ∈ {1, 2, 3}, B1(Ji−,−) is skew-symmetric. For completeness, let us show it for J1. Define B̃ : V × V → R as B̃(v, w) = B1(J1v, J1w). As J2 1 = − IdV , B̃ is non-degenerate. And, as J1 and B1 are The Geometry of Generalised Spinr Spinors on Projective Spaces 5 ρ-invariant and B1 is symmetric, B̃ is ρ-invariant and symmetric. Now define φ̃ : V → V ∗ by φ̃(v) = B̃(v,−). Consider the endomorphism of representations φ−1 1 ◦ φ̃. Then, there exist c, d1, d2, d3 ∈ R such that φ−1 1 ◦ φ̃ = c IdV +d1J1 + d2J2 + d3J3. Hence, for every v, w ∈ V , B̃(v, w) = cB1(v, w)+d1B1(J1v, w)+d2B1(J2v, w)+d3B1(J3v, w). As B̃ and B1 are symmetric, d1B1(J1−,−) + d2B1(J2−,−) + d3B1(J3−,−) is symmetric. Define Q = d1J1 + d2J2 + d3J3, so that B1(Q−,−) is symmetric. Note that Q2 = − ( d21 + d22 + d23 ) IdV . Now, for 0 ̸= v ∈ V , B1(Qv,Qv) = B1 ( v,Q2v ) = − ( d21 + d22 + d23 ) B1(v, v) ≤ 0. By positive-definiteness of B1, we must have d21+d 2 2+d 2 3 = 0, and hence B̃(v, w) = cB1(v, w). Again by positive definiteness of B1, c > 0. And, for 0 ̸= v ∈ V , B1(v, v) = B1 ( J2 1v, J 2 1v ) = B̃(J1v, J1v) = cB1(J1v, J1v) = cB̃(v, v) = c2B1(v, v), so c = 1. This shows that B1(J1−, J1−) = B1(−,−) or, equivalently, that B1(J1−,−) is skew-symmetric. One can repeat this reasoning with J2 and J3 to show that B1(J2−,−) and B1(J3−,−) are skew-symmetric. Going back to our previous situation, we have that b1B1(J1−,−) + b2B1(J2−,−) + b3B1(J3−,−) is skew-symmetric. And, as B1, B2 are sym- metric, this bilinear form is also symmetric. Hence, it is 0. Therefore, by positive-definiteness of B1, b1J1 + b2J2 + b3J3 = 0, showing that b1 = b2 = b3 = 0, and hence that B2 = aB1. And, by positive-definiteness, a > 0. This concludes the proof. ■ Proposition 2.2. Let g be a finite-dimensional Lie algebra and (ρ1, V1), (ρ2, V2) two non- isomorphic self-dual irreducible real representations of g. Then, V1 ⊥ V2 with respect to any (ρ1 ⊕ ρ2)-invariant inner product on V1 ⊕ V2. Proof. Let B be a (ρ1 ⊕ ρ2)-invariant inner product on V1 ⊕ V2. If V1 is not B-orthogonal to V2, then we get a non-zero morphism of representations V1 → V ∗ 2 , namely v 7→ B(v,−)|V2 . By Schur’s lemma, this would be an isomorphism, which is a contradiction since V2 is self-dual and V1 is not isomorphic to V2. ■ Finally, we have the result we were after. Theorem 2.3. Let G/H be a reductive homogeneous space with reductive decomposition g = h ⊕ m, where g is the Lie algebra of G, h is the Lie algebra of H and m is an AdH-invariant complement of h in g. Suppose that the adjoint (isotropy) representation of H on m decomposes as a direct sum of pairwise non-isomorphic irreducible components m = m1 ⊕ · · · ⊕mk. Suppose that m admits an AdH-invariant inner product. Then, AdH-invariant inner products on each irreducible component exist and are unique up to positive scaling, and any AdH-invariant inner product on m is a positive linear combination of AdH-invariant inner products on the irreducible components. Proof. An AdH -invariant inner product on m restricts to an AdH -invariant inner product on each irreducible component mi. Hence, mi is self-dual. Now, apply Propositions 2.1 and 2.2. ■ 2.2 Some notation (1) If V is a representation of a Lie group G, and H ⊆ G is a subgroup, then we shall denote the restricted representation by V |H . We shall use analogous notation for restrictions of Lie algebra representations to Lie subalgebras. (2) Throughout the computations carried out in this paper, we will repeatedly use some matrix notation and identities, taken from [2, p. 9]. We will denote by E (n) i,j (resp. F (n) i,j ) the 6 D. Artacho and J. Hofmann elementary n× n skew-symmetric (resp. symmetric) matrix given by E (n) i,j = i j  ... i −1 . . . j . . . 1 ... , F (n) i,j = i j  ... i 1 . . . j . . . 1 ... . By convention, the matrix F (n) i,i has all the entries equal to zero except for the (i, i) entry, which is 1. We will denote by B0 the bilinear form on the space of matrices of appropriate size given by B0(X,Y ) := −Re(tr(XY )), where Re(z) denotes the real part of z and tr(A) is the trace of the matrix A. Finally, if {ei}i is an orthonormal basis for some vector space V with respect to an inner prod- uct B, we shall denote by ei,j := ei ∧ ej the standard basis elements for so(V,B) ∼= Λ2V , sending ei 7→ ej and ej 7→ −ei. 2.3 Invariant spinr structures Denote by SO(n) the special orthogonal group, and let λn : Spin(n) → SO(n) be the standard two-sheeted covering. This map induces an isomorphism at the level of Lie algebras, and its inverse ρ : so(n) ∼= Λ2Rn → spin(n) ⊆ Cl(n) is given by 2ei ∧ ej 7→ ei · ej . If f is any map with codomain so(n), we will refer to f̃ := ρ ◦ f as the spin lift of f . For r ∈ N, we define the group Spinr(n) := (Spin(n)× Spin(r))/Z2, where Z2 = ⟨(−1,−1)⟩ ⊆ Spin(n)×Spin(r). Note that Spin1(n) = Spin(n), Spin2(n) = SpinC(n) and Spin3(n) = SpinH(n), and that there are natural homomorphisms λrn : Spinr(n) → SO(n), [µ, ν] 7→ λn(µ), ξrn : Spinr(n) → SO(r), [µ, ν] 7→ λr(ν). We recall a topological result that we will use multiple times throughout the text. Proposition 2.4 ([5]). The map φr,n : Spinr(n) → SO(n)× SO(r) defined by λrn × ξrn is a two- sheeted covering. Moreover, (1) φ2,2 ♯ ( π1 ( Spin2(2) )) = ⟨(1,±1)⟩ ⊆ Z× Z ∼= π1(SO(2)× SO(2)), (2) for n ≥ 3, φ2,n ♯ ( π1 ( Spin2(n) )) = ⟨(1, 1)⟩ ⊆ Z2 × Z ∼= π1(SO(n)× SO(2)), (3) for r, n ≥ 3, φr,n♯ (π1(Spin r(n))) = ⟨(1, 1)⟩ ⊆ Z2 × Z2 ∼= π1(SO(n)× SO(r)), where we always take the identifications π1(SO(n)× SO(r)) ∼= π1(SO(n))× π1(SO(r)). These enlargements or twistings of the spin group give rise to the main players in this paper. Definition 2.5. Let M be an oriented Riemannian n-manifold with principal SO(n)-bundle of positively oriented orthonormal frames FM . A spinr structure onM is a reduction of the struc- ture group of FM along the homomorphism λrn. In other words, it is a pair (P,Φ) consisting of The Geometry of Generalised Spinr Spinors on Projective Spaces 7 � a principal Spinr(n)-bundle P over M , and � a Spinr(n)-equivariant bundle homomorphism Φ: P → FM , where Spinr(n) acts on FM via λrn. If there is no risk of confusion and Φ is clear from the context, we shall simply denote such a structure by P . The principal SO(r)-bundle associated to P along ξrn is called the auxiliary bundle of the spinr structure, and it is denoted by P̂ . If (P1,Φ1) and (P2,Φ2) are spinr structures on M , an equivalence of spinr structures from (P1,Φ1) to (P2,Φ2) is a Spinr(n)-equivariant diffeomorphism f : P1 → P2 such that Φ1 = Φ2 ◦f . If, moreover, M = G/H is a Riemannian homogeneous space, we say that a spinr struc- ture (P,Φ) on M is G-invariant if G acts smoothly on P by Spinr(n)-bundle homomorphisms and Φ is G-equivariant. G-invariant spinr structures on G/H are in one-to-one correspondence with representation- theoretical data. Theorem 2.6 ([5]). Let G/H be an n-dimensional oriented Riemannian homogeneous space with H connected and isotropy representation σ : H → SO(n). Then, there is a bijective corre- spondence between � G-invariant spinr structures on G/H modulo G-equivariant equivalence of spinr structures, and � Lie group homomorphisms φ : H → SO(r) such that σ × φ : H → SO(n) × SO(r) lifts to a homomorphism ϕ : H → Spinr(n) along λrn, modulo conjugation by an element of SO(r). Explicitly, to such a φ corresponds the spinr structure (P,Φ) with P = G ×ϕ Spinr(n) and Φ: P → FM ∼= G×σ SO(n) given by [g, x] 7→ [g, λrn(x)]. Definition 2.7. For an oriented Riemannian homogeneous space M = G/H, its G-invariant spin type Σ(M,G) is defined by Σ(M,G) := min{r ∈ N |M admits a G-invariant spinr structure}. 2.4 Exterior forms approach to the spin representation It is well known – see, e.g., [32] – that, for n ∈ N, the complexification Cl(n) of the real Clifford algebra Cl(n) satisfies Cl(n) ∼= { M2k(C) if n = 2k, M2k(C)⊕M2k(C) if n = 2k + 1. (2.1) For n = 2k or 2k + 1, define Σn = ( C2 )⊗k , and let sk : M2k(C) → EndC(Σn) be the standard representation of M2k(C). The spin represen- tation ∆n : Cl(n) → EndC(Σn) is defined by ∆n := { sk if n = 2k, sk ◦ prj if n = 2k + 1, where prj is the projection onto the j-th factor. Note that, for odd n, there are two non- isomorphic irreducible representations of Cl(n), and we are choosing one of them (we shall 8 D. Artacho and J. Hofmann specify which one below). We will also denote by ∆n its restriction to Spin(n) when there is no risk of confusion. This restriction is independent of the choice of representation for odd n. It is useful to have an explicit description of the spin representation which does not use the isomorphism (2.1). We will describe one here, which we refer to as the exterior forms approach to the spin representation. Similar realizations have appeared, e.g., in [23, 32] and in the early supergravity literature (see, e.g., [22]). More details and examples, using our precise conventions, can be found, e.g., in [2, 28]. Suppose n = 2k + 1, and let (e0, . . . , e2k) be the standard basis of Rn. Its complexification decomposes as C0 ⊕ L⊕ L′, where C0 = spanC{u0 := ie0} and L := spanC { xj := 1√ 2 (e2j−1 − ie2j) }k j=1 , L′ := spanC { yj := 1√ 2 (e2j−1 + ie2j) }k j=1 . Note that dimC(Λ •L′) = 2k, and Cl(n) acts on Λ•L′ by extending xj · η := i √ 2xj⌟η, yj · η := i √ 2yj ∧ η, u0 · η := −ηeven + ηodd, where ηeven and ηodd are, respectively, the even and odd parts of η ∈ Λ•L′. Hence, e2j−1 · η := i(xj⌟η + yj ∧ η), e2j · η := yj ∧ η − xj⌟η, e0 · η := iηeven − iηodd. (2.2) This representation is isomorphic to ∆n for n = 2k + 1 (the other possible choice of irreducible representation of Cl(n) corresponds to letting e0 act by the negative of what is established in (2.2)). To obtain it for n = 2k, repeat all the above ignoring everything with a zero subscript. These representations have an invariant Hermitian product, which we shall denote by ⟨·, ·⟩, and an associated norm |·|, for which the basis {yj1,...,jk := yj1 ∧ · · · ∧ yjk | 0 ≤ k ≤ n, 1 ≤ j1 < · · · < jk ≤ n} is orthonormal.1 2.5 Invariant spinr spinors A classical spin structure allows us to build a spinor bundle. Similarly, a spinr structure naturally induces a family of complex vector bundles as follows: Definition 2.8. Let M be an n-dimensional oriented Riemannian manifold admitting a spinr structure (P,Φ). For m ∈ N odd, its m-twisted spinr spinor bundle is defined by Σmn,rM := P ×∆m n,r Σmn,r, with the natural projection to M induced by that of P , where ∆m n,r := ∆n ⊗∆⊗m r , Σmn,r := Σn ⊗ Σ⊗m r , where ∆m n,r is viewed as a representation of Spinr(n). The requirement that m be odd in Definition 2.8 comes from the fact that ∆m n,r, which is a representation of Spin(n)× Spin(r), descends to a representation of Spinr(n) if and only if m is odd. 1Here we use the convention that the empty wedge product (i.e., the case k = 0) is equal to 1. The Geometry of Generalised Spinr Spinors on Projective Spaces 9 If, moreover, M = G/H is a homogeneous space and (P,Φ) is G-invariant, then there ex- ists a homomorphism φ : H → SO(r) such that σ × φ lifts to a map ϕ : H → Spinr(n) (see Theorem 2.6), and this bundle takes the form Σmn,rM = G×∆m n,r◦ϕ Σ m n,r. Sections of Σmn,rM are called m-twisted spinr spinors – if there is no risk of confusion, we will just refer to them as spinr spinors or simply spinors. They are identified with H-equivariant maps ψ : G → Σmn,r, and G acts on the space of spinors by (g · ψ)(g′) := ψ ( g−1g′ ) . G-invariant spinors correspond, then, to constant H-equivariant maps G → Σmn,r, which in turn correspond to elements of Σmn,r which are stabilised by H. If H is connected, these are just elements of Σmn,r which are annihilated by the differential action of the Lie algebra of H. We denote the space of invariant m-twisted spinr spinors by (Σmn,r)inv. Remark 2.9. It should be noted that the exterior forms approach to the spin representation described above does not in general give an identification of (classical) spinors with globally de- fined differential forms; the isotropy action on Σn is not generally equivalent (as representations) to the isotropy action on Λ0,•m. Rather, this realisation of spinors via exterior forms is purely algebraic, and is often non-canonical (i.e., it depends on the choice of basis for m). Notable ex- ceptions exist in the presence of certain special geometric structures – see, e.g., [2, Remark 3.9] and the proof of [28, Theorem 5.10]. On the other hand, there are other various constructions (so-called squaring constructions) which associate (real) differential forms to spinors (see, e.g., [13, 32, 41]). One common such construction is to associate to a spinor ψ the k-form ω(k)(X1, . . . , Xk) := Re⟨(X1 ∧ . . . Xk) · ψ,ψ⟩ for all X1, . . . , Xk ∈ TM, and it is well known that if ψ is a Killing spinor then ω(1) (or, more precisely, its dual vector field) is a Killing vector field (see, e.g., [9, Section 1.5]). It should be noted that one obtains quite often ω(k) = 0, even for non-vanishing spinors ψ (see, e.g., [2, Table 6]). In particular, the differential forms associated to an invariant spinor in this manner do not seem to be heavily influenced by its realisation in the exterior form model of the spin representation (which itself may be non-canonical). In a similar spirit to Definition 2.7, we make the following definition. Definition 2.10. For an oriented Riemannian homogeneous space M = G/H, the G-invariant spinor type of M is defined by σ(M,G) := min { r ∈ N ∣∣∣∣M admits a G-invariant spinr structure with (Σmn,r)inv ̸= 0 for some odd m } . Remark 2.11. The G-invariant spinor type σ(Mn, G) is well defined, and it satisfies 1 ≤ σ(M,G) ≤ n. This is because the G-invariant Spinn structure on M determined by tak- ing φ : H → SO(n) to be equal to the isotropy representation always carries a non-zero invariant 1-twisted spinn spinor – see [14, Proposition 3.3]. The requirement that r be minimal in the definition of the G-invariant spinor type is moti- vated by the next proposition, which shows that passing from a spinr structure to any spinr ′ structure (r′ > r) induced by it via the obvious inclusion Spinr(n) ↪→ Spinr ′ (n) leads to redun- dancies. Before stating the proposition, we introduce some terminology which will be useful in describing the relationship between the structures: 10 D. Artacho and J. Hofmann Definition 2.12. Let Mn = G/H be an oriented Riemannian homogeneous space. We say that a spinr structures Pr and a spinr ′ structure Pr′ (r ≤ r′) on M are in the same lineage if Pr′ ∼= Pr ×ι Spin r′(n), where ι : Spinr(n) ↪→ Spinr ′ (n) is the natural inclusion map induced by the inclusion SO(r) ↪→ SO(r′) as the lower right-hand r × r block. Proposition 2.13. Let Mn = G/H be an oriented Riemannian homogeneous space with con- nected isotropy group H, equipped with a G-invariant spinr structure Pr. Furthermore, for any r′ ≥ r consider the invariant spinr ′ structure Pr′ in the lineage of Pr. If ψ ∈ (Σmn,r)inv is an invariant m-twisted spinr spinor, then it induces an invariant m′-twisted spinr ′ spinor for any m′ ≥ m (m, m′ odd), i.e., there is an inclusion (Σmn,r)inv ↪→ ( Σm ′ n,r′ ) inv for all r′ ≥ r, m′ ≥ m (m, m′ odd). Proof. It suffices to prove the result for r′ ∈ {r, r + 1}. Suppose first that r′ = r + 1, and let φ : H → SO(r) be an auxiliary homomorphism corresponding to Pr in the sense of Theorem 2.6. Denoting by σ : H → SO(n) the isotropy representation, we begin by observ- ing that the invariant spinr+1 structure in the lineage of Pr is induced by the lift of the homomorphism σ × φ′ : H → SO(n)× SO(r + 1) given by the composition of σ × φ : H → SO(n) × SO(r) with the inclusion SO(n) × SO(r) ↪→ SO(n) × SO(r + 1). In particular, h acts on Σm ′ n,r+1 = Σn ⊗ Σ⊗m′ r+1 by the (tensor product action associated to the) lift of σ∗ on the Σn factor and the lift of φ∗ on the Σr+1 factors. We now split into two cases based on the parity of r. Supposing first that r is even, we have Σr+1|spin(r)C ≃ Σr as spin(r)C representations, and therefore Σm ′ n,r+1|hC ≃ Σm ′ n,r as h C-representations. Since m, m′ are both odd we have m′−m = 2k for some k ≥ 0, and therefore Σ⊗m′ r |spin(r)C ≃ Σ⊗m r ⊗ Σr ⊗ · · · ⊗ Σr︸ ︷︷ ︸ 2k copies . But Σr is a self-dual representation of spin(r)C, hence also a self-dual representation of hC, so Σ⊗2k r contains a copy of the trivial h-representation. The corresponding H-representation thus also contains a trivial subrepresentation since H is connected. In particular, there is a copy of Σmn,r|H inside Σm ′ n,r|H and the result in this case follows. Suppose now that r is odd, and denote by Σr+1 = Σ+ r+1 ⊕ Σ− r+1 the splitting into positive and negative half-spinor spaces. Then we have Σ+ r+1|spin(r)C ≃ Σr, hence Σm ′ n,r+1|hC contains a copy of Σm ′ n,r, and the result in this case then follows by the same argument as in the even case. We have shown the result holds for r′ = r + 1 (hence for all r′ > r), and all that remains is to consider the case r′ = r. The result in this case follows by arguing exactly as above, using the fact that Σr is a self-dual spin(r)C representation to find a copy of the trivial representation in Σ ⊗(m′−m) r . ■ 2.6 Special spinr spinors In the classical spin setting, it is well known that spinors satisfying certain additional prop- erties carry geometric information about the manifold. Some of the most widely studied ex- amples are the so-called Riemannian Killing spinors, which are solutions of the differential equation ∇g Xψ = λX · ψ for all X ∈ TM (here ∇g denotes the spinorial connection induced by the Levi-Civita connection, and λ ∈ R). We refer the reader to [8, 9, 19, 24], among others, for a detailed exposition of their basic properties and relationship to geometric structures in low dimensions. Another class of important special spinors are the pure spinors, which are defined by the algebraic condition that their annihilator inside TCM (with respect to Clifford multipli- cation) is a maximal isotropic subbundle. Such spinors correspond, uniquely up to scaling, with orthogonal almost complex structures on the manifold – see [32, Chapter 9] for details. As in the classical spin case, special spinr spinors also encode geometric properties. In analogy with pure spinors, we define. The Geometry of Generalised Spinr Spinors on Projective Spaces 11 Definition 2.14 ([26]). Let ψ ∈ Σmn,r, X,Y ∈ Rn and 1 ≤ k < l ≤ r, and let (ê1, . . . , êr) be the standard basis of Rr. The real 2-form ηψkl and the skew-symmetric endomorphism η̂ψkl associated to ψ are defined by ηψkl(X,Y ) := Re⟨(X ∧ Y ) · (êk · êl) · ψ,ψ⟩, η̂ψkl(X) := ( ηψkl(X, ·) )♯ , where X ∧ Y = X · Y + ⟨X,Y ⟩ ∈ spin(n) and êk · êl ∈ spin(r). We say that ψ is pure if( η̂ψkl )2 = − IdRn and ( ηψkl + 2êk · êl ) · ψ = 0 (only for r ≥ 3), for all 1 ≤ k < l ≤ r. An m-twisted spinr spinor on a manifold is pure if it is pure at every point. It is clear that an invariant spinr spinor on a homogeneous space is pure if, and only if, it is pure at one point. We are also interested in various differential equations that a spinr spinor might satisfy. Recall that the Levi-Civita connection on a spin manifold naturally induces a connection on the spinor bundle. Similarly, the Levi-Civita connection of a spinr manifold together with a connection θ on the auxiliary bundle defines a connection ∇θ on each twisted spinr spinor bundle. There are obvious analogues of the usual special spinorial field equations (including the classical Killing spinor equation mentioned above) to the spinr setting. Definition 2.15. Let ψ be a twisted spinr spinor on M and θ a connection on the auxiliary bundle of the spinr structure. (1) ψ is θ-parallel if ∇θψ = 0; (2) ψ is θ-Killing if for all vector fields X one has that ∇θ Xψ = λX ·ψ, for some constant λ ∈ R; (3) ψ is θ-generalised Killing if there exists a symmetric endomorphism field A ∈ End(TM) such that, for all vector fields X, one has ∇θ Xψ = A(X) · ψ. We collect here a number of results that relate the existence of special spinr spinors to geometric properties of the manifold: Theorem 2.16 ([14, 26]). Let M be an n-dimensional spinr manifold, and let θ be a connection on its auxiliary bundle. (1) If M carries a θ-parallel spinor ψ, then the Ricci tensor decomposes as Ric = 1 |ψ|2 ∑ k<l Θ̂kl ◦ η̂ψkl, where Θ̂kl is the skew-symmetric endomorphism associated to the 2-form on TM given by Θkl(X,Y ) := ⟨Ω(X,Y )(êk), êl⟩, where Ω is the curvature 2-form of the connection θ on the auxiliary bundle. (2) If θ is flat and M carries a θ-Killing spinor, then M is Einstein. (3) If M carries a θ-parallel m-twisted pure spinor for some m ∈ N, r ≥ 3, r ̸= 4, n ̸= 8, n+ 4r − 16 ̸= 0 and n+ 8r − 16 ̸= 0, then M is Einstein. (4) If r = 2 and M carries a θ-parallel pure spinor, then M is Kähler. (5) If r = 3 and M admits a θ-parallel pure spinor, then M is quaternionic Kähler. If M = G/H is a Riemannian homogeneous space, invariant connections on homogeneous bundles over M are described by algebraic data [38] (see, e.g., [3] for a modern treatment). 12 D. Artacho and J. Hofmann Proposition 2.17. Let G/H be a homogeneous space, and let ϕ : H → K be a Lie group homo- morphism. There is a one-to-one correspondence between G-invariant connections on G×ϕ K and linear maps Λ : g → k satisfying2 (1) Λ(X) = ϕ∗(X), X ∈ h; (2) Λ ◦AdH(h) = AdK(ϕ(h)) ◦Λ, h ∈ H. The map Λ corresponding to a connection is called the Nomizu map of said connection. For the connections of interest in this article, the Nomizu maps are particularly easy to describe: Proposition 2.18 ([38, Theorem 13.1]). Let (G/H, g) be an n-dimensional oriented Rieman- nian homogeneous space, where the metric g corresponds to an invariant inner product B on a reductive complement m of h. The Nomizu map Λ : g → so(m) of the Levi-Civita connection of g is given by Λ(X)(Y ) = 1 2 [X,Y ]m + U(X,Y ), X ∈ g, Y ∈ m, where U is defined by B(U(X,Y ),W ) = 1 2 (B([W,X]m, Y ) +B(X, [W,Y ]m)). The following proposition describes how the correspondence in Proposition 2.17 works in the particular situation we are interested in. Proposition 2.19. Let (G/H, g) be an n-dimensional Riemannian homogeneous space equipped with a G-invariant spinr structure P . Let Λ : g → so(n) be the Nomizu map of the Levi-Civita connection of g, and let Λ′ : g → so(r) be the Nomizu map of an invariant connection θ on the associated bundle P̂ . Let Λ̃ be the spin lift of Λ to spin(n) and let Λ̃′ be the spin lift of Λ′ to spin(r). Then, Λ̃⊗ ( Λ̃′ )⊗m is the Nomizu map of the invariant connection ∇θ on the m- twisted spinr spinor bundle. Moreover, if ψ ∈ (Σmn,r)inv and X̂ is the fundamental vector field on G/H defined by X ∈ m, then( ∇θ X̂ ψ ) eH = ( Λ̃⊗ Λ̃′⊗m ) (X) · ψ. In particular, an invariant m-twisted spinr spinor ψ is θ-parallel if, and only if, it satisfies the equation ∀X ∈ m : ( Λ̃⊗ Λ̃′⊗m ) (X) · ψ = 0. As we shall see later in the setting of spinC structures, the second condition in Proposition 2.17 is quite restrictive. Indeed, the auxiliary bundles of invariant spinC structures are principal bundles of the abelian group SO(2). Hence, the second condition becomes Λ ◦ AdH(h) = Λ for all h ∈ H. This will force the kernel of Λ|m to be quite large in most of our cases. The following is a useful criterion. Lemma 2.20. Let G/H be a homogeneous space with a reductive decomposition g = h⊕m, and let ϕ : H → K be a Lie group homomorphism. Let H0 ⊆ H be the kernel of AdK ◦ϕ : H → GL(k). If X ∈ spanR[h0,m], then Λ(X) = 0 for the Nomizu map Λ : g → k associated to any invariant connection on G×ϕ K. 2Note that AdH in condition (2) refers to the restriction of the adjoint representation of G to H ⊆ G, whereas AdK refers to the adjoint representation of K. The Geometry of Generalised Spinr Spinors on Projective Spaces 13 Proof. By linearity of Λ, it suffices to consider X = [v, Y ] for some v ∈ h0 and Y ∈ m. Let γ : R → H0 be a curve with γ(0) = eG and γ′(0) = v. By Proposition 2.17, the Nomizu map Λ of any invariant connection on G×ϕ K satisfies Λ ◦AdH(γ(t)) = Λ, and hence 0 = d dt ∣∣ t=0 Λ(Y ) = d dt ∣∣ t=0 Λ(AdH(γ(t))Y ) = Λ([v, Y ]) = Λ(X). ■ Finally, we examine the differential equations satisfied by invariant spinr spinors on symmetric spaces. The following proposition is analogous to the familiar fact in the spin setting that invariant spinors on symmetric spaces are ∇g-parallel, since the Levi-Civita and the Ambrose– Singer connections coincide. Proposition 2.21. Let (M = G/H, g) be a Riemannian symmetric space admitting a G- invariant spinr structure P . Let g = h ⊕ m be a reductive decomposition such that [m,m] ⊆ h. Let E be the natural vector bundle associated to the auxiliary bundle P̂ , and let ∇E be the unique G-invariant connection on E whose Nomizu map vanishes identically on m. Let ∇ := ∇g ⊗ ( ∇E )⊗m be the corresponding twisted connection on Σmn,rM . If ψ ∈ Σmn,r is a G-invariant spinr spinor, then ∇ψ = 0. Proof. With respect to the reductive decomposition g = h⊕m, the Nomizu map associated to the Levi-Civita connection vanishes identically on the reductive complement m, i.e., Λg|m ≡ 0. The Nomizu map of ∇ then vanishes identically on m, and the result follows. ■ This result will be useful for several of the cases in our classification, where the limited number of low-dimensional representations of the isotropy groups will force the auxiliary bundles to be isomorphic to familiar tensor (sub)bundles. 3 Projective spaces Onǐsčik [39, p. 163] classified the compact, simple and simply connected Lie groups which act transitively on the projective spaces CPn, HPn and OP2 – see also [44, p. 356]. We exhibit them in Table 2. Space G H CPn SU(n+ 1) S(U(1)×U(n)) CP2n+1 Sp(n+ 1) U(1)× Sp(n) HPn Sp(n+ 1) Sp(1)× Sp(n) OP2 F4 Spin(9) Table 2. Compact, simple and simply connected Lie groups G acting transitively with isotropy H on projective spaces – see, e.g., [44, p. 356]. 3.1 Hermitian complex projective space In this section, we consider the complex projective space realised as the quotient CPn ∼= SU(n+ 1)⧸S(U(1)×U(n)), where S(U(1)×U(n)) = {( z 0 0 B ) ∈ Mn+1(C) | z ∈ U(1), B ∈ U(n), z det(B) = 1 } . 14 D. Artacho and J. Hofmann In order to study SU(n + 1)-invariant spinr structures and spinors on this space, we need to establish some notation and properties of the Lie algebras involved. Let us denote by h the Lie algebra of H := S(U(1) × U(n)), and consider the copy of SU(n) included in SU(n + 1) as the lower right-hand n× n block. Letting h′ := su(n) ⊆ su(n+ 1), we have the decomposition h = Rξ ⊕ h′ (as Lie algebras), where ξ := i ( −nF (n+1) 1,1 + ∑n+1 l=2 F (n+1) l,l ) and h′ = su(n) = spanR { iF (n+1) p,q , E(n+1) p,q , i ( F (n+1) r,r − F (n+1) r+1,r+1 )} 2≤p<q≤n+1 r=2,...,n . The isotropy subalgebra h ⊆ su(n+ 1) has a reductive complement m := (h)⊥B0 = spanR { iF (n+1) 1,p , E (n+1) 1,p } p=2,...,n+1 , and the adjoint representation of h on m is irreducible. Hence, by Theorem 2.3, the SU(n+ 1)- invariant metrics on CPn correspond to the inner products on m in the one-parameter fam- ily ga := aB0|m×m, a > 0, and a ga-orthonormal basis of m is given by{ e2p−1 := i√ 2a F (n+1) 1,p+1 , e2p := 1√ 2a E (n+1) 1,p+1 } p=1,...,n . We take the orientation defined by the ordering (e1, e2, . . . , e2n−1, e2n). 3.1.1 Invariant spinr structures We are now ready to determine the SU(n+ 1)-invariant spin type of CPn. By [27, p. 327], it is clear that CPn admits an SU(n+1)-invariant spin structure if, and only if, n is odd. Moreover, one has the following. Theorem 3.1. The SU(n+ 1)-invariant spinC structures on CPn are given by SU(n+ 1)×ϕs Spin C(2n), s ∈ Z : n ̸≡ s mod 2, where ϕs is the unique lift of σ×φs to SpinC(2n), σ : H → SO(2n) is the isotropy representation and φs : H → SO(2) ∼= U(1) is given by( z 0 0 B ) 7→ det(B)s. In particular, the SU(n+ 1)-invariant spin type of CPn is Σ(CPn, SU(n+ 1)) = { 1, n odd, 2, n even. Proof. Note that H ∼= U(n), and that every Lie group homomorphism U(n) → U(1) is of the form B 7→ det(B)s, for some s ∈ Z. The loop α(t) = diag ( e−2πit, 1, . . . , 1, e2πit ) gener- ates π1(H) ∼= Z, and (σ × φs)♯(α) = (n− 1, s) ∈ π1(SO(2n))× π1(SO(2)). This can be seen as follows: the image of α(t) under the isotropy representation σ is easily seen to be σ(α(t)) = diag ( e2πit, . . . , e2πit, e4πit ) ∈ U(n) ⊆ SO(2n), where e2πit appears n− 1 times. This can be seen using the realisation of σ as the action of H on m by matrix conjugation. Hence, by Proposition 2.4, σ × φs : H → SO(2n) × U(1) lifts to SpinC(2n) if, and only if, n ̸≡ s mod 2. Finally, as U(1) is abelian, the representations φs are pairwise non-equivalent. The result now follows from Theorem 2.6. ■ The Geometry of Generalised Spinr Spinors on Projective Spaces 15 3.1.2 Invariant spinr spinors The classical spin case r = 1 does not yield any non-trivial invariant spinors, as we show in the following theorem. Theorem 3.2. For n odd, there are no non-trivial SU(n+ 1)-invariant spinors on CPn. Proof. We need the explicit expression of the action of ξ ∈ h on m. Letting σ : H → SO(2n) be the isotropy representation and σ̃ its lift to Spin(2n), and, e.g., using the commutation relations in [2, p. 9], one can readily see that, for each p = 1, . . . , n, ad(ξ)|m(e2p) = [ξ, e2p]m = (n+ 1)e2p−1, ad(ξ)|m(e2p−1) = [ξ, e2p−1]m = −(n+ 1)e2p. Hence, σ∗(ξ) = ad(ξ)|m = (n+ 1) n∑ p=1 e2p ∧ e2p−1 ∈ so(2n), and the spin lift is given by σ̃∗(ξ) = n+ 1 2 n∑ p=1 e2p · e2p−1 ∈ spin(2n) ⊆ Cl(2n). A direct computation using (2.2) shows that, for each 1 ≤ k ≤ n and 1 ≤ j1 < · · · < jk ≤ n, σ̃∗(ξ) · (yj1 ∧ · · · ∧ yjk) = i(n+ 1) 2 (2k − n)yj1 ∧ · · · ∧ yjk . From this we observe that, if n is odd, there are no non-trivial invariant spinors. ■ The fact that no non-trivial invariant spinors exist motivates the investigation of spinC spinors. Theorem 3.3. For n, s ∈ N with n ̸≡ s mod 2, the space of SU(n+1)-invariant 1-twisted spinC spinors on CPn associated to the spinC structure SU(n+ 1)×ϕs Spin C(2n) is given by ( Σ1 2n,2 ) inv =  spanC{1⊗ 1̂, (y1 ∧ · · · ∧ yn)⊗ ŷ1}, s = n+ 1, spanC{(y1 ∧ · · · ∧ yn)⊗ 1̂, 1⊗ ŷ1}, s = −(n+ 1), 0, otherwise. In particular, the SU(n+ 1)-invariant spinor type of CPn is σ(CPn, SU(n+ 1)) = 2. Proof. Recall that h = Rξ ⊕ h′ as Lie algebras, and note that, for ψ ∈ Σ1 2n,2, (∀X ∈ h′ : (ϕs)∗(X) · ψ = 0) ⇐⇒ ψ ∈ spanC{1, y1 ∧ · · · ∧ yn} ⊗ Σ2, by [2, Theorem 3.7] and the definition of φs. Moreover, (ϕs)∗(ξ) = ( n+ 1 2 n∑ p=1 e2p · e2p−1, sn 2 ê1 · ê2 ) ∈ spin(2n)⊕ spin(2) ∼= spinC(2n). Finally, an easy calculation shows that, for 0 ≤ k ≤ n and 1 ≤ j1 < · · · < jk ≤ n, (ϕs)∗(ξ) · ( (yj1 ∧ · · · ∧ yjk)⊗ 1̂ ) = i 2 ((n+ 1)(2k − n) + sn) ( (yj1 ∧ · · · ∧ yjk)⊗ 1̂ ) , (ϕs)∗(ξ) · ((yj1 ∧ · · · ∧ yjk)⊗ ŷ1) = i 2 ((n+ 1)(2k − n)− sn)((yj1 ∧ · · · ∧ yjk)⊗ ŷ1). From this it is straightforward to conclude the result. ■ 16 D. Artacho and J. Hofmann 3.1.3 Special spinr spinors The aim now is to show that the SU(n+1)-invariant spinC spinors on CPn found in Theorem 3.3 are pure and parallel with respect to a suitable connection on the auxiliary bundle. Proposition 3.4. For s = n+1 (resp. s = −(n+1)), the SU(n+1)-invariant spinC spinors 1⊗ 1̂ and (y1∧· · ·∧yn)⊗ ŷ1 ( resp. (y1∧· · ·∧yn)⊗ 1̂ and 1⊗ ŷ1 ) on CPn are pure. Moreover, they are parallel with respect to the invariant connection on the corresponding auxiliary bundle determined by the Nomizu map Λ|m = 0. Proof. We will only show the calculations for the spinor ψ = (y1 ∧ · · · ∧ yn) ⊗ 1̂, as the other three are analogous. Since r = 2 < 3, we only need to show that ( η̂ψ12 )2 = − Id. Indeed, a straightforward calculation shows that ηψ12(e2p, e2q) = Re⟨e2p · e2q · ê1 · ê2 · ψ,ψ⟩+ δp,q Re⟨ê1 · ê2 · ψ,ψ⟩ = Re⟨ie2p · e2q · ψ,ψ⟩+ δp,q Re⟨−iψ,ψ⟩ = 0, ηψ12(e2p−1, e2q−1) = 0, ηψ12(e2p, e2q−1) = Re⟨e2p · e2q−1 · ê1 · ê2 · ψ,ψ⟩ = Re⟨ie2p · e2q−1 · ψ,ψ⟩ = −δp,q. Hence, η̂ψ12(e2p) = −e2p−1 and η̂ψ12(e2p−1) = e2p. The last assertion of the proposition follows by noting that, as CPn ∼= SU(n + 1)/ S(U(1) × U(n)) is a symmetric space, the Levi-Civita connection coincides with the Ambrose–Singer connection, whose Nomizu map satisfies Λg|m ≡ 0. ■ In light of the Ricci decomposition in [14, Theorem 3.1], the existence of parallel pure spinC spinors encodes a very well-known fact – see, e.g., [44]. Theorem 3.5. The SU(n+ 1)-invariant metrics ga on CPn are Kähler–Einstein. Proof. Take s = −(n+1). Consider the spinC structure defined by φs, and endow its auxiliary bundle with the connection described in Proposition 3.4. We have seen that this spinC structure carries a non-zero parallel pure spinC spinor ψ = (y1∧ · · ·∧yn)⊗ 1̂. This implies that the metric is Kähler [26, Corollary 4.10] with respect to the invariant complex structure defined by η̂ψ12. Now, by Theorem 2.16 (1), the Ricci tensor decomposes as Ric = 1 |ψ|2 Θ̂12 ◦ η̂ψ12, where Θ̂12 is the endomorphism associated to the 2-form on m Θ12(X,Y ) := ⟨Ω(X,Y )(ê1), ê2⟩, X,Y ∈ m, where Ω is the curvature 2-form of the connection on the auxiliary bundle. Recall [3] that, if Λ is the Nomizu map of the connection on the auxiliary bundle, then ∀X,Y ∈ m : Ω(X,Y ) = [Λ(X),Λ(Y )]so(2) −Λ([X,Y ]) = −Λ([X,Y ]). It is now easy to see that, for all 1 ≤ p, q ≤ n, Ω(e2p−1, e2q) = δp,q s a ê1 ∧ ê2, Ω(e2p−1, e2q−1) = Ω(e2p, e2q) = 0. Hence, Θ̂12 = s a n∑ p=1 e2p−1 ∧ e2p, The Geometry of Generalised Spinr Spinors on Projective Spaces 17 and finally, using the expression of η̂ψ12 obtained in the proof of Proposition 3.4, we obtain Ric = 1 |ψ|2 Θ̂12 ◦ η̂ψ12 = n+ 1 a Id . (3.1) This proved the theorem. ■ Remark 3.6. Recall that the Fubini–Study metric gFS on CPn is SU(n+1)-invariant, and that its Ricci constant is 2(n+ 1). From equation (3.1), we can deduce that gFS = g1/2. 3.2 Symplectic complex projective space Consider, for n ≥ 1, the homogeneous realisation of odd-dimensional complex projective space CP2n+1 ∼= Sp(n+ 1)⧸U(1)× Sp(n), where H := U(1) × Sp(n) is realised as a subgroup of Sp(n + 1) by the upper left-hand 1 × 1 block for U(1) and the lower right-hand n×n block for Sp(n). Denote by h the Lie algebra of H and h′ := sp(n) ⊆ sp(n+ 1). Then, h = Rξ1 ⊕ h′ (as Lie algebras), where ξ1 := iF (n+1) 1,1 and h′ = sp(n) = spanR { iF (n+1) p,p , jF (n+1) p,p , kF (n+1) p,p , iF (n+1) r,s , jF (n+1) r,s , kF (n+1) r,s , E(n+1) r,s } 2≤r<s≤n+1 p=2,...,n+1 . The Lie subalgebra h ⊆ sp(n+ 1) has a reductive complement m := (h)⊥B0 = V ⊕H, where V := spanR { ξ2 := −kF (n+1) 1,1 , ξ3 := jF (n+1) 1,1 } , H := spanR { e4p := jF (n+1) 1,p+1 , e4p+1 := kF (n+1) 1,p+1 , e4p+2 := iF (n+1) 1,p+1 , e4p+3 := E (n+1) 1,p+1 } p=1,...,n , and this is the decomposition of m into irreducible subrepresentations3 of the adjoint represen- tation of h. We have, therefore, by Theorem 2.3, a two-parameter family of invariant metrics ga,t := aB0|H×H + 2atB0|V×V , a, t > 0, and a ga,t-orthonormal basis of m is given by{ ξa,t2 := 1√ 2ta ξ2, ξ a,t 3 := 1√ 2ta ξ3, e a,t 4p+ε := 1√ 2a e4p+ε } ε=0,1,2,3 p=1,...,n . We take the orientation defined by the ordering ( ξa,t2 , ξa,t3 , ea,t4 , . . . , ea,t4n+3 ) . 3.2.1 Invariant spinr structures We begin by determining the Sp(n+1)-invariant spin type of CP2n+1. By [27, p. 327], it is clear that CP2n+1 admits a unique spin structure, and this structure is Sp(n + 1)-invariant. Using the algebraic characterisation in Theorem 2.6, we can explicitly obtain all Sp(n + 1)-invariant spinC structures on CP2n+1: Theorem 3.7. The Sp(n+ 1)-invariant spinC structures on CP2n+1 are given by Sp(n+ 1)×ϕs Spin C(4n+ 2), s ∈ 2Z, where ϕs is the unique lift of σ × φs to SpinC(4n + 2), σ : H → SO(4n + 2) is the isotropy representation and φs : H → SO(2) ∼= U(1) is defined by (z,A) 7→ zs. 3Note that this decomposition into V and H corresponds to the vertical and horizontal distributions of the generalised Hopf fibration S2 ↪→ CP2n+1 → HPn. 18 D. Artacho and J. Hofmann Proof. This follows from Theorem 2.6, together with the fact that Sp(n) is simple and that all Lie group homomorphisms U(1) → U(1) are of the form z 7→ zs, for some s ∈ Z. Using Proposition 2.4 as in the proof of Theorem 3.1, one sees that σ × φs lifts to SpinC(4n + 2) if, and only if, s is even. As U(1) is abelian, the representations φs are pairwise non-equivalent. Hence, these spinC structures are pairwise non-Sp(n+ 1)-equivariantly equivalent. ■ 3.2.2 Invariant spinr spinors First, we classify the Sp(n+ 1)-invariant spinors for the unique spin structure of CP2n+1. Theorem 3.8. The space Σinv of Sp(n+ 1)-invariant spinors on CP2n+1 is given by Σinv = { spanC { ψ+ := ω(n+1)/2, ψ− := y1 ∧ ω(n−1)/2 } , n odd, 0, n even, where ω := ∑n p=1 y2p ∧ y2p+1. Proof. By [2, Theorem 4.11], the space of invariant spinors is quite restricted: Σinv ⊆ spanC { ωk, y1 ∧ ωk } k=0,...,n . We only need to determine which of these are annihilated by ξ1. A computation analogous to the one in the proof of Theorem 3.2 shows that, if σ̃ is the lift to Spin(4n + 2) of the isotropy representation σ : H → SO(4n+ 2), σ̃∗(ξ1) = ξa,t2 · ξa,t3 + 1 2 n∑ p=1 ( ea,t4p · ea,t4p+1 + ea,t4p+2 · e a,t 4p+3 ) . In particular, σ̃∗(ξ1) · ωk = i(n+ 1− 2k)ωk, σ̃∗(ξ1) · ( y1 ∧ ωk ) = i(n− 1− 2k) ( y1 ∧ ωk ) , and the result follows. ■ We now turn to the study of spinC spinors. Using the same argument as in the Hermitian case (see Theorem 3.3), one obtains. Theorem 3.9. For n ∈ N and s = 2s′ ∈ 2Z, the space ( Σ1 4n+2,2 ) inv of Sp(n + 1)-invariant 1-twisted spinC spinors on CP2n+1 associated to the spinC structure Sp(n+1)×ϕs Spin C(4n+2) is given by ( Σ1 4n+2,2 ) inv =  spanC { ω(n+1+s′)/2, y1 ∧ ω(n−1+s′)/2 } ⊗ 1̂⊕ ⊕ spanC { ω(n+1−s′)/2, y1 ∧ ω(n−1−s′)/2}⊗ ŷ1, n ̸≡ s′ mod 2, 0, otherwise, where negative powers of ω are defined to be zero. In particular, the Sp(n+ 1)-invariant spinor type of CP2n+1 satisfies σ ( CP2n+1,Sp(n+ 1) ) = { 1, n odd, 2, n even. Remark 3.10. The spinC structure corresponding to s = 0 is the one induced by the usual spin structure. Indeed, taking s = 0 in Theorem 3.9, one recovers the spinors in Theorem 3.8 tensored with Σ2. The Geometry of Generalised Spinr Spinors on Projective Spaces 19 3.2.3 Special spinr spinors In order to differentiate these spinors, one can see, using the formulas for the Nomizu map from [9, p. 141], that the spin lift Λ̃a,t of the Nomizu map of the Levi-Civita connection of ga,t is given by Λ̃a,t ( ξa,t2 ) = 1− t 2 √ 2at n∑ p=1 ( ea,t4p · ea,t4p+2 − ea,t4p+1 · e a,t 4p+3 ) , Λ̃a,t ( ξa,t3 ) = 1− t 2 √ 2at n∑ p=1 ( ea,t4p · ea,t4p+3 + ea,t4p+1 · e a,t 4p+2 ) , Λ̃a,t ( ea,t4p ) = 1 2 √ t 2a ( −ξa,t2 · ea,t4p+2 − ξa,t3 · ea,t4p+3 ) , Λ̃a,t ( ea,t4p+1 ) = 1 2 √ t 2a ( ξa,t2 · ea,t4p+3 − ξa,t3 · ea,t4p+2 ) , Λ̃a,t ( ea,t4p+2 ) = 1 2 √ t 2a ( ξa,t2 · ea,t4p + ξa,t3 · ea,t4p+1 ) , Λ̃a,t ( ea,t4p+3 ) = 1 2 √ t 2a ( −ξa,t2 · ea,t4p+1 + ξa,t3 · ea,t4p ) . (3.2) Baum et al. proved in [9, p. 146] that CP3 admits non-trivial Sp(2)-invariant generalised Killing spinors, given by ψ+ ± iψ−. For the Fubini–Study metric, the two eigenvalues of these generalised Killing spinors coincide, yielding real Killing spinors which are related to the nearly Kähler geometry of CP3. We now show that this does not occur in higher dimensions. Theorem 3.11. The spaces CP2n+1 admit non-trivial Sp(n + 1)-invariant generalised Killing spinors if, and only if, n = 1. Proof. First, we recall from Theorem 3.8 that there are no invariant spinors when n is even, so it remains only to consider the case where n is odd. Let n be odd, and suppose that n ≥ 3 so that ω(n±3)/2 ̸= 0. Using the above formulas (3.2) for the Nomizu map, we get, for α, β ∈ C, Λ̃a,t ( ea,t4p ) · (αψ+ + βψ−) = − √ t 2a { α n+ 1 2 y1 ∧ y2p + βy2p+1 } ∧ ω(n−1)/2. Writing a general element X ∈ m as a (real) linear combination X = µ2ξ a,t 2 + µ3ξ a,t 3 + n∑ p=1 ( µ4pe a,t 4p + µ4p+1e a,t 4p+1 + µ4p+2e a,t 4p+2 + µ4p+3e a,t 4p+3 ) of the basis vectors, we find that the Clifford product with an arbitrary invariant spinor is given by X · (αψ+ + βψ−) = α { (iµ2 + µ3)y1 + n∑ p=1 [(iµ4p + µ4p+1)y2p + (iµ4p+2 + µ4p+3)y2p+1] } ∧ ω(n+1)/2 + α n+ 1 2 { n∑ p=1 [(iµ4p − µ4p+1)y2p+1 + (−iµ4p+2 + µ4p+3)y2p] } ∧ ω(n−1)/2 + β{iµ2 − µ3}ω(n−1)/2 20 D. Artacho and J. Hofmann + βy1 ∧ { n∑ p=1 [(−iµ4p − µ4p+1)y2p + (−iµ4p+2 − µ4p+3)y2p+1] } ∧ ω(n−1)/2 + β n− 1 2 { n∑ p=1 [(−iµ4p + µ4p+1) + (iµ4p+2 − µ4p+3)] } y1 ∧ ω(n−3)/2. Hence, by equating coefficients in Λ̃a,t ( ea,t4p ) · (αψ+ + βψ−) = X · (αψ+ + βψ−), one easily concludes (using crucially that n ≥ 3) that the only possibility is α = β = 0, and the result then follows from the preceding discussion about the case n = 1. ■ We now turn to the study of the Sp(n + 1)-invariant spinC spinors on CP2n+1 found in Theorem 3.9. The aim is to show that, when s′ = s/2 = ±n±1 and t = 1, there is a pure spinor which is parallel with respect to a suitable connection on the auxiliary bundle. This encodes the fact that, for t = 1, the metric ga,t is Kähler. Lemma 3.12. Let k ∈ N and ψ ∈ { ωk⊗ 1̂, ( y1∧ωk ) ⊗ 1̂, ωk⊗ ŷ1, ( y1∧ωk ) ⊗ ŷ1 } . Then, a scalar multiple of ψ is pure if, and only if, k = 0 or k = n. Proof. We will only prove it for ψ = ωk ⊗ 1̂, as the other cases are analogous. Since r = 2 < 3, we only need to show that ( η̂ψ12 )2 = − Id. Indeed, for all 1 ≤ p, q ≤ n and ε ∈ {0, 1, 2, 3}, one calculates: ηψ12 ( ξa,t2 , ξa,t3 ) = Re 〈 ξa,t2 · ξa,t3 · ê1 · ê2 · ψ,ψ 〉 = Re 〈 iξa,t2 · ξa,t3 · ψ,ψ 〉 = −Re 〈 ωk ⊗ 1̂, ωk ⊗ 1̂ 〉 = −(k!)2 ( n k ) , ηψ12 ( ea,t4p , e a,t 4q+1 ) = Re 〈 ea,t4p · ea,t4q+1 · ê1 · ê2 · ψ,ψ 〉 = Re 〈 i · ea,t4p · ea,t4q+1 · ω k, ωk 〉 = −δp,q 〈 ωk − 2ky2p ∧ y2p+1 ∧ ωk−1, ωk 〉 = −δp,q(k!)2 [( n k ) − 2 ( n− 1 k − 1 )] , ηψ12 ( ea,t4p+2, e a,t 4q+3 ) = −δp,q(k!)2 [( n k ) − 2 ( n− 1 k − 1 )] , ηψ12 ( ξa,t2 , ea,t4p+ε ) = ηψ12 ( ξa,t3 , ea,t4p+ε ) = ηψ12 ( ea,t4p , e a,t 4q+2 ) = ηψ12 ( ea,t4p , e a,t 4q+3 ) = ηψ12 ( ea,t4p+1, e a,t 4q+3 ) = 0, where ( n−1 k−1 ) is understood to be 0 if k = 0. Altogether, we have η̂ψ12 = −(k!)2 [( n k ) ξa,t2 ∧ ξa,t3 + [( n k ) − 2 ( n− 1 k − 1 )] n∑ p=1 ( ea,t4p ∧ ea,t4p+1 + ea,t4p+2 ∧ e a,t 4p+3 )] , which is easily seen to square to a multiple of − Id if, and only if, k = 0 or k = n. ■ The preceding lemma, together with Theorem 3.9 describing the invariant spinC spinors, implies that the Sp(n + 1)-invariant spinC structure corresponding to s = 2s′ ∈ 2Z admits invariant pure spinC spinors if, and only if, s′ ∈ {n+ 1,−n− 1, n− 1,−n+ 1}, which are given in each case by{ (y1 ∧ ωn)⊗ 1̂, 1⊗ ŷ1 } , s′ = n+ 1, { 1⊗ 1̂, (y1 ∧ ωn)⊗ ŷ1 } , s′ = −n− 1,{ ωn ⊗ 1̂, y1 ⊗ ŷ1 } , s′ = n− 1, { y1 ⊗ 1̂, ωn ⊗ ŷ1 } , s′ = −n+ 1. The Geometry of Generalised Spinr Spinors on Projective Spaces 21 In order to differentiate these spinors, one needs to fix a connection on the auxiliary bundle. Applying the criterion in Lemma 2.20, one sees that the only Sp(n+1)-invariant connection on the auxiliary bundle is the one with Nomizu map Λ|m = 0. This connection, together with the Levi-Civita connection of the metric ga,t (with Nomizu map Λa,t), induces a connection ∇a,t on the corresponding spinC spinor bundle. The following lemma is a straightforward calculation using the expression of the spin lift of the Nomizu map (3.2): Lemma 3.13. The invariant pure spinC spinor 1⊗ 1̂ is ∇a,t-parallel if, and only if, t = 1. These spinC spinors encode some well-known geometric information of CP2n+1 – see, e.g., [44]. Theorem 3.14. The metric ga,1 on CP2n+1 is Kähler–Einstein, for all a > 0. Proof. Let s = −2(n+ 1), and consider the spinC structure on CP2n+1 determined by φs. By Lemmas 3.12 and 3.13, this structure carries a ∇a,1 parallel pure spinC spinor ψ = 1⊗ 1̂. Hence, by [26, Corollary 4.10], the metric ga,1 is Kähler, for all a > 0. Let us now see that these metrics are also Einstein. By Theorem 2.16 (1), the Ricci tensor decomposes as Ric = 1 |ψ|2 Θ̂12 ◦ η̂ψ12. (3.3) By calculations similar to those in the proof of Theorem 3.5, one finds that, for 1 ≤ p, q ≤ n, 0 ≤ ε ≤ 3 and 2 ≤ l ≤ 3, Ω ( ξa,12 , ξa,13 ) = −Λ ([ ξa,12 , ξa,13 ]) = −1 a Λ(ξ1) = −1 a ( φ−2(n+1) ) ∗(ξ1) = 2(n+ 1) a ê1 ∧ ê2, Ω ( ea,14p , e a,1 4q+1 ) = Ω ( ea,14p+2, e a,1 4q+3 ) = δp,q 2(n+ 1) a ê1 ∧ ê2, Ω ( ea,14p , e a,1 4q+2 ) = Ω ( ea,14p , e a,1 4q+3 ) = Ω ( ea,14p+1, e a,1 4q+2 ) = Ω ( ea,14p+1, e a,1 4q+3 ) = Ω ( ξa,1l , ea,14p+ε ) = 0. Hence, using the definition of Θ̂12 in terms of Ω and taking k = 0 in the proof of Lemma 3.12, Θ̂12 = 2(n+ 1) a ( ξa,12 ∧ ξa,13 + n∑ p=1 ( ea,14p ∧ ea,14p+1 + ea,14p+2 ∧ ea,14p+3 )) , η̂ψ12 = − ( ξa,12 ∧ ξa,13 + n∑ p=1 ( ea,14p ∧ ea,14p+1 + ea,14p+2 ∧ ea,14p+3 )) . Finally, substituting everything into equation (3.3), we get Ric = 2(n+1) a Id, which completes the proof. ■ 3.3 Quaternionic projective space Consider the homogeneous realisation of quaternionic projective space given by HPn ∼= Sp(n+ 1)⧸Sp(1)× Sp(n), where H := Sp(1) × Sp(n) is realised as a subgroup of Sp(n + 1) by the upper left-hand 1 × 1 block for Sp(1) and the lower right-hand n × n block for Sp(n). Denote by h the Lie algebra of H and h′ := sp(n) ⊆ h. Then, h = sp(1)⊕h′ (as Lie algebras), and explicit bases are given by sp(1) = spanR { ξ1 := iF (n+1) 1,1 , ξ2 := −kF (n+1) 1,1 , ξ3 := jF (n+1) 1,1 } , 22 D. Artacho and J. Hofmann h′ = sp(n) = spanR { iF (n+1) p,p , jF (n+1) p,p , kF (n+1) p,p , iF (n+1) r,s , jF (n+1) r,s , kF (n+1) r,s , E(n+1) r,s } 2≤r<s≤n+1 p=2,...,n+1 . The isotropy subalgebra h ⊆ sp(n+ 1) has a reductive complement m := spanR { e4p := jF (n+1) 1,p+1 , e4p+1 := kF (n+1) 1,p+1 , e4p+2 := iF (n+1) 1,p+1 , e4p+3 := E (n+1) 1,p+1 } p=1,...,n , and the adjoint representation of h on m is irreducible. Therefore, by Theorem 2.3, the invariant metrics come in a one-parameter family ga := aB0|m×m, a > 0, and one easily verifies that the above basis of m rescaled by 1/ √ 2a is ga-orthonormal. Without virtually any loss of generality, in order to simplify the notation we will only consider g := g1/2. We take the orientation defined by the ordering (e4, e5, . . . , e4n+3). 3.3.1 Invariant spinr structures As HP1 is just the sphere S4, we will suppose throughout this section that n > 1. By Theo- rem 2.6, in order to understand Sp(n + 1)-invariant spinr structures on HPn, we need to find all Lie group homomorphisms φ : H → SO(r) such that σ × φ lifts to Spinr(4n). Since H is simply-connected, any such homomorphism lifts. Note also that, for r = 2, using simplicity of Sp(1) and Sp(n), the only Lie group homomorphism Sp(1) × Sp(n) → SO(2) is the trivial one. The corresponding Sp(n+ 1)-invariant spinC structure on HPn is naturally induced by its unique spin structure. The first interesting case is r = 3, which corresponds to spinH structures. In order to classify them, we need to describe all homomorphisms Sp(1) → SO(3): Proposition 3.15. Up to conjugation by elements of SO(3) there are exactly two Lie group homomorphisms Spin(3) → SO(3), namely the trivial homomorphism and the double covering λ3. Proof. Let φ : Sp(1) → SO(3) be a non-trivial homomorphism, and recall that Sp(1) ∼= Spin(3). As the Lie algebra so(3) is simple, the only non-trivial normal subgroups of Sp(1) are discrete. Hence, φ has discrete kernel. As Sp(1) is compact, the image of φ is a closed subgroup of SO(3). By the first isomorphism theorem for Lie groups, the image of φ is a 3-dimensional Lie subgroup of SO(3), and hence it is open in SO(3). As SO(3) is connected, φ must be surjective. In particular, the representation of Sp(1) on R3 induced by φ must be irreducible, since oth- erwise the image of φ would be contained inside a subgroup isomorphic to {1}×SO(2). There is only one real irreducible 3-dimensional representation of Sp(1) up to isomorphism [29, Proposi- tion 11] (namely, the standard spin double-cover φ0), hence φ is conjugate to φ0 inside GL(3,R). It follows that any two non-trivial homomorphisms φ1, φ2 : Sp(1) → SO(3) are conjugate to each other inside GL(3,R), and it remains only to show that they are conjugate inside SO(3). Fix T ∈ GL(3,R) such that, for all A ∈ Sp(1), we have T−1φ1(A)T = φ2(A). We claim that there exists T̂ ∈ SO(3) such that T̂−1φ1T̂ = T−1φ1T . Indeed, let B := T−1φ1(A)T = φ2(A) ∈ SO(3). Then, TT t = φ1(A) −1TBT t = φ1(A) −1TBBtT t ( φ1(A) −1 )t = φ1(A) −1TT tφ1(A). As φ1 is surjective, TT t commutes with all elements of SO(3), hence it is a scalar multiple of the identity. The result then follows by taking T̂ = det(T )−1/3T ∈ SO(3). ■ This allows us to classify invariant spinH structures on quaternionic projective spaces. Theorem 3.16. For n > 1, the Sp(n+ 1)-invariant spinH structures on HPn are given by Sp(n+ 1)×ϕi Spin H(4n), i = 0, 1, where σ : H → SO(4n) is the isotropy representation, φ0 is the trivial homomorphism Sp(1) × Sp(n) → SO(3), φ1(x, y) = λ3(x) and ϕi is the unique lift of σ × φi to SpinH(4n). The Geometry of Generalised Spinr Spinors on Projective Spaces 23 The invariant spinH structure corresponding to φ0 is simply the one induced by the unique spin structure, so for the rest of our discussion of HPn we fix the spinH structure corresponding to φ1. Remark 3.17. Observe that the auxiliary vector bundle of the spinH structure corresponding to φ1 is Sp(n+1)-equivariantly isomorphic to the rank-3 vector subbundle of End(THPn) induced by the standard quaternionic Kähler structure on HPn. 3.3.2 Invariant spinr spinors To begin, it is easy to see that this homogeneous realisation carries no invariant spinors: as the homogeneous realisation of HPn that we are considering is that of a symmetric space, invariant spinors are parallel, and we know that HPn cannot have any non-trivial parallel spinor, since it is not Ricci-flat. However, we shall see in the next proposition that there are always non-trivial invariant spinH spinors, for sufficient twistings of the spinor bundle, when n is odd. Proposition 3.18. The Sp(n+ 1)-invariant spinor type of HPn (n > 1) is σ(HPn, Sp(n+ 1)) = { 3, n odd, > 3, n even. Furthermore, for n odd, the number of twistings m ≥ 0 of the spinor bundle which realises this is m = n. Proof. The preceding discussion shows that there are no invariant spinors. As noted above, the only invariant spinC structure is the one coming from the spin structure, and it is clear that there are also no invariant spinC spinors in this case (since H acts trivially on Σ2 and hence each Σm4n,2 is equivalent as H-modules to a direct sum of copies of Σ4n). This shows that σ(HPn,Sp(n+ 1)) ≥ 3. Denote by Vt := V (tω1) the irreducible representation of sp(2,C) ∼= sl(2,C) with highest weight tω1 (and dimension t+ 1). Arguing as in [2, Section 4.1.6], we con- sider the structure of S := (Σ4n|hC)sp(2n,C) = spanC{ωk}nk=0 (ω := ∑n j=1 y2j ∧ y2j+1) as a module for sp(2,C) ⊂ hC (here we adopt the usual convention sp(k)C ∼= sp(2k,C)). The action of sp(2,C) on S follows from [28, Lemma 5.13] and is given explicitly by ãd(ξ1)|m · ωk = i(n− 2k)ωk, (3.4) ãd(ξ2)|m · ωk = k(n− k + 1)ωk−1 − ωk+1, ãd(ξ3)|m · ωk = i ( ωk+1 + k(n− k + 1)ωk−1 ) . (3.5) The standard basis element for the Cartan subalgebra of sp(2,C) ∼= sl(2,C) is −iξ1 ∼ diag[1,−1], and by (3.4) we see that the action of this element on S has highest eigenvalue n. In particular, S contains a copy of Vn, and by reason of dimension we have S ≃ Vn as sp(2,C)-modules. Note that this representation is self-dual (see, e.g., [40]). Thus, it suffices to show that the smallest odd tensor power of Σ3 which contains a copy of Vn is Σ⊗n 3 . Recalling the well-known decomposition Vs ⊗ Vt ≃ Vs+t ⊕ Vs+t−2 ⊕ · · · ⊕ Vs−t, s ≥ t (3.6) of sl(2,C)-representations (see, e.g., [21, Sections 11.1 and 11.2]), the result for the case where n is odd follows by repeatedly using (3.6) to decompose tensor powers of Σ3 ≃ V1. For the case where n is even, one sees from (3.6) that the decompositions of odd tensor powers of Σ3 ≃ V1 into irreducible representations contain only factors of the form Vt with t odd, and in particular cannot contain a copy of Vn. ■ 24 D. Artacho and J. Hofmann Remark 3.19. The difference in behaviour between the even and odd cases in the preced- ing proposition occurs as something of a technicality rather than a manifestation of any sig- nificant geometric difference; indeed, the argument presented in the odd case also produces representation-theoretic invariants in the even case if we allow even twistings of the spinor bun- dle. The reason to exclude even twistings is that the twisted spinor module would then fail to be well defined as a representation of Spinr(n), since [−1,−1] wouldn’t act by the identity map. In order to obtain a notion of spinor bundles with even numbers of twistings, one needs to consider instead the alternative structure groups Spin(n)× Spin(r) described in [26, Remark 2.3]. In the next proposition we describe explicitly the invariant n-twisted spinH spinors which re- alise the equality σ(HPn,Sp(n + 1)) = 3 for n odd. Recall that, as noted in Remark 3.17, the auxiliary vector bundle E of the spinH structure corresponding to φ1 can be seen as a subbundle of the endomorphism bundle End(THPn). In particular, E inherits a natural connection ∇E from the Levi-Civita connection ∇g on THPn, and the former induces a con- nection ∇g,E := ∇g ⊗ ( ∇E )⊗m on the spinor bundle Σm4n,3HPn for any odd m ≥ 1. Recall that {ξ1, ξ2, ξ3} is a basis of sp(1) ⊂ sp(1) ⊕ sp(n) = h, and denote by (Φ1 := ad(ξ1)|m, Φ2 := ad(ξ2)|m,Φ3 := ad(ξ3)|m) the standard basis of the invariant rank-3 subspace of End(m) corresponding to E. The action of ξi on this subspace is given by Φi 7→ 0, Φj 7→ 2Φk, Φk 7→ −2Φj , where (i, j, k) is an even permutation of (1, 2, 3), and the spin lift of this representation acts on Σ3 ∼= C2 by the standard basis matrices for su(2): ρ(ξ1) := ( i 0 0 −i ) , ρ(ξ2) := ( 0 1 −1 0 ) , ρ(ξ3) := ( 0 i i 0 ) (these matrices are taken relative to the standard basis 1̂ := (1, 0), ŷ1 := (0, 1) for Σ3 ∼= C2). In order to relate these to the usual presentation of the Lie algebra sl(2,C) ∼= sp(2,C), we introduce H := −iξ1, X := 1 2(ξ2 − iξ3), Y := −1 2(ξ2 + iξ3), so that ρ(H) = ( 1 0 0 −1 ) , ρ(X) = ( 0 1 0 0 ) , ρ(Y ) = ( 0 0 1 0 ) (3.7) act in the representation Σ3 ∼= C2 by the usual operators. Using this setup, we obtain the following. Theorem 3.20. If n > 1 is odd, the space of Sp(n + 1)-invariant n-twisted spinH spinors is spanned over C by ψ := n∑ j=0 (−1)jωj ⊗ ( ρ(Y )n−j .1 ) , where 1 := 1̂⊗ n· · · ⊗ 1̂. Proof. As in the proof of Proposition 3.18, we note that S := (Σ4n|hC)sp(2n,C) ≃ Vn as mod- ules for sl(2,C) ∼= sp(2,C) ⊂ hC; explicitly we have S = spanC { ωℓ }n ℓ=0 , with the action of sp(2,C) = spanC{ξ1, ξ2, ξ3} by the formulas (3.4)–(3.5). It is clear from (3.7) that 1 ∈ Σ⊗n 3 is a highest weight vector for sp(2,C), and that the sp(2,C)-submodule U(sp(2,C)).1 that it gen- erates4 is isomorphic to Vn (since ρ(H) = diag[1,−1] acts on 1 by multiplication by n). There- fore, we have U(sp(2,C)).1 = spanC { ρ(Y )k.1 }n k=0 . On the other hand, we see from (3.4)–(3.5) 4Here, U(sp(2,C)) refers to the universal enveloping algebra of sp(2,C) and the . product refers to the action via the representation. The Geometry of Generalised Spinr Spinors on Projective Spaces 25 that 1 = ω0 ∈ S is a highest weight vector and Y.ωk = ωk+1 for all k = 0, . . . , n. In particular, the isomorphism T : S → U(sp(2,C)).1 is given (up to rescaling) by T ( ωk ) = ρ(Y )k.1. We have T ∈ Homsp(2,C) ( S,Σ⊗n 3 ) ≃ S∗ ⊗ Σ⊗n 3 , and the invariant spinH spinor we are seeking is the corresponding element of S ⊗ Σ⊗n 3 obtained via the musical isomorphism ♯ : S∗ ≃ S associated to the sp(2,C)-invariant symplectic form Ω ( ωj , ωk ) = { (−1)j , j + k = n, 0, j + k ̸= n on S. Defining ω̂j ∈ S∗ by ω̂j ( ωk ) = δj,k, one sees that ( ω̂j )♯ = (−1)j+1ωn−j , and the result then follows by noting that T = ∑n j=0 ω̂ j ⊗ ( ρ(Y )j .1 ) . ■ Remark 3.21. The spinor in the statement of Theorem 3.20 corresponds to the one in [26, Section 3.4.1], which the authors show to be pure. Finally, we give the differential equation satisfied by ψ. Recall that HPn = Sp(n+1)/Sp(1)× Sp(n) is a symmetric space, and that the auxiliary bundle of the spinH structure under consid- eration is isomorphic to the rank-3 bundle spanned by the (locally-defined) endomorphisms Φ1, Φ2, Φ3. This bundle inherits a connection ∇E from the Levi-Civita connection, and its Nomizu map vanishes identically when restricted to m. The following is an immediate consequence of Proposition 2.21. Corollary 3.22. For n > 1 odd, the invariant n-twisted spinH spinor ψ in Theorem 3.20 is parallel with respect to the invariant connection ∇g,E := ∇g ⊗ ( ∇E )⊗n . This spinH spinor encodes, via [26, Corollary 4.12], a well-known geometric fact – see, e.g., [11]. Theorem 3.23. The metric ga on HPn is quaternionic Kähler. 3.4 Octonionic projective plane Consider now the octonionic projective plane, realised as a homogeneous space via OP2 ∼= F4⧸Spin(9). A description of the isometric action of F4 can be found, e.g., in [7], and, importantly, the isotropy representation is just the real spin representation of Spin(9) on R16: Spin(9) ↪−→ Cl09,0 ∼= Cl8,0 ∼= M16(R). As in [42, Theorems 1.4.3 and 1.4.4, Proposition 1.4.5], at the level of Lie algebras this inclusion is given by spin(9) ↪−→ Cl09,0 ψ1∼= Cl8,0 ψ2∼= Cl0,6⊗Cl2,0 ψ3∼= Cl4,0⊗Cl0,2⊗Cl2,0 ψ4∼= Cl0,2⊗Cl2,0⊗Cl0,2⊗Cl2,0 ψ5∼= M2(R)⊗H⊗M2(R)⊗H ψ6∼= M2(R)⊗M2(R)⊗H⊗H ψ7∼= M4(R)⊗M4(R) ψ8∼= M16(R). (3.8) The algebra isomorphisms ψ1, ψ2, ψ3, ψ4, ψ5 are given explicitly in [42]; ψ6 is the obvious permutation of the second and third factors; ψ7 is the tensor product of the Kronecker product of the first two factors and the isomorphism H⊗R H → M4(R), q1 ⊗ q2 7→ (x 7→ q1 · x · q2), 26 D. Artacho and J. Hofmann where x = (x1, x2, x3, x4) ∈ R4 is thought of as the quaternion x1 + ix2 + jx3 + kx4; and ψ8 is the Kronecker product. Letting {e0, e1, . . . , e8} be the canonical basis of R9, a basis of spin(9) is given by spin(9) = spanR{ei · ej}0≤i<j≤8. With a slight abuse of notation, we will denote by e1, . . . , en the elements of the canonical basis of Rn, for n = 2, 4, 6, 8, 16. Now we give the images in M16(R) of each of the elements of our basis of spin(9). For the first basis vector e0 · e1, one computes, following the chain of maps in (3.8) e0 · e1 7→ e0 · e1 7→ e1 7→ e1 ⊗ (e1 · e2) 7→ e1 ⊗ (e1 · e2)⊗ (e1 · e2) 7→ e1 ⊗ (e1 · e2)⊗ (e1 · e2)⊗ (e1 · e2) 7→ ( 1 0 0 −1 ) ⊗ k ⊗ ( 0 1 −1 0 ) ⊗ k 7→ ( 1 0 0 −1 ) ⊗ ( 0 1 −1 0 ) ⊗ k ⊗ k 7→  0 1 0 0 −1 0 0 0 0 0 0 −1 0 0 1 0 ⊗  1 0 0 0 0 −1 0 0 0 0 −1 0 0 0 0 1  7→ −E(16) 1,5 + E (16) 2,6 + E (16) 3,7 − E (16) 4,8 + E (16) 9,13 − E (16) 10,14 − E (16) 11,15 + E (16) 12,16. (3.9) The others are computed similarly, giving e0 · e2 7→ −E(16) 1,13 + E (16) 2,14 + E (16) 3,15 − E (16) 4,16 + E (16) 5,9 − E (16) 6,10 − E (16) 7,11 + E (16) 8,12 , e0 · e3 7→ −E(16) 1,7 − E (16) 2,8 − E (16) 3,5 − E (16) 4,6 − E (16) 9,15 − E (16) 10,16 − E (16) 11,13 − E (16) 12,14, e0 · e4 7→ E (16) 1,6 + E (16) 2,5 − E (16) 3,8 − E (16) 4,7 + E (16) 9,14 + E (16) 10,13 − E (16) 11,16 − E (16) 12,15, e0 · e5 7→ −E(16) 1,4 + E (16) 2,3 + E (16) 5,8 − E (16) 6,7 − E (16) 9,12 + E (16) 10,11 + E (16) 13,16 − E (16) 14,15, e0 · e6 7→ −E(16) 1,8 + E (16) 2,7 − E (16) 3,6 + E (16) 4,5 − E (16) 9,16 + E (16) 10,15 − E (16) 11,14 + E (16) 12,13, e0 · e7 7→ −E(16) 1,2 + E (16) 3,4 − E (16) 5,6 + E (16) 7,8 − E (16) 9,10 + E (16) 11,12 − E (16) 13,14 + E (16) 15,16, e0 · e8 7→ −E(16) 1,3 − E (16) 2,4 − E (16) 5,7 − E (16) 6,8 − E (16) 9,11 − E (16) 10,12 − E (16) 13,15 − E (16) 14,16. (3.10) The images of the other basis vectors ei · ej (1 ≤ i ≤ 8) for spin(9) are then determined by taking products of the above generators inside Cl09,0 using the Clifford algebra identities ei · ej = (e0 · ei) · (e0 · ej). 3.4.1 Invariant spinr structures By Theorem 2.6, in order to understand F4-invariant spin r structures on OP2, we need to find all Lie group homomorphisms φ : Spin(9) → SO(r) such that σ×φ lifts to Spinr(16) (of course, as the group Spin(9) is simply connected, the lifting condition is automatically satisfied). As the Lie algebra spin(9) ∼= so(9) is simple, for 1 ≤ r ≤ 8 the only homomorphism Spin(9) → SO(r) is the trivial one. The corresponding F4-invariant spin r structures are just the ones in the lineage of the invariant spin structure. The first non-trivial case is r = 9, where we have the covering homomorphism λ9 : Spin(9) → SO(9). By essentially the same argument as in Proposition 3.15, there are only two Lie group homomorphisms Spin(9) → SO(9) up to conjugation by elements of SO(9), namely the trivial one φ0 and the double covering φ1 := λ9, leading to two possible invariant spin9 structures up to equivalence. Theorem 3.24. The F4-invariant spin9 structures on OP2 are given by F4 ×ϕi Spin 9(16), i = 0, 1, where ϕi is the unique lift of σ × φi to Spin9(16) and σ : Spin(9) → SO(16) is the isotropy representation. The Geometry of Generalised Spinr Spinors on Projective Spaces 27 3.4.2 Invariant spinr spinors As in the case of HPn, the octonionic projective plane OP2 does not admit any invariant spinors (cf. the discussion before Proposition 3.18), so from this point forward we consider the non-trivial invariant spin9 structure (i.e., the one corresponding to i = 1 in the preceding theorem). In order to describe its twisted spin9 spinors we first need a small lemma describing the decomposition of the spin lift of the isotropy representation as a direct sum of highest weight modules. This result may be found, written in a slightly different form and without proof, in [18, Section 7]; we include a sketch of the proof here as the notation and formulas will be useful for subsequent discussion. Lemma 3.25 ([18]). As modules for Spin(9)C, the spin lift σ̃ of the isotropy representation decomposes as Σ16 ≃ V (ω1 + ω4)︸ ︷︷ ︸ Σ− 16 ⊕V (ω3)⊕ V (2ω1)︸ ︷︷ ︸ Σ+ 16 , (3.11) where V (µ) denotes the irreducible representation of highest weight µ and ωi, i = 1, 2, 3, 4 denote the fundamental weights of spin(9)C ∼= so(9,C). Proof. In order to take advantage of the explicit operators calculated in (3.9)–(3.10), we view spin(9)C ∼= so(9,C) as the set of 9 × 9 skew-symmetric matrices in gl(9,C). We take the (real form of the) Cartan subalgebra spanned by hj := −iE (9) 2j−1,2j , j = 1, 2, 3, 4, together with the (positive) re-scaling of the Killing form such that the hj ’s are orthogonal and unit length. Letting vj := h♭j , we have the simple roots α1 = v1 − v2, α2 = v2 − v3, α3 = v3 − v4, α4 = v4, and the corresponding fundamental weights ωj := 2αj ||αj ||2 are given by ω1 = v1, ω2 = v1 + v2, ω3 = v1 + v2 + v3, ω4 = 1 2 (v1 + v2 + v3 + v4). The root vectors Xi := Xαi associated to the simple roots αi are X1 = E (9) 1,3 + E (9) 2,4 + i ( −E(9) 2,3 + E (9) 1,4 ) , X2 = E (9) 3,5 + E (9) 4,6 + i ( −E(9) 4,5 + E (9) 3,6 ) , X3 = E (9) 5,7 + E (9) 6,8 + i ( −E(9) 6,7 + E (9) 5,8 ) , X4 = E (9) 7,9 − iE (9) 8,9 , and the root vectors associated to the roots −αi, i = 1, 2, 3, 4 are given by Yi := Yαi := Xi. We note that this setup is slightly unusual5 and can be found, e.g., in [43]. Using the explicit formulas for σ from (3.9)–(3.10), we find that the Cartan subalgebra generators hi and simple root vectors Xi act in the complexified isotropy representation mC ≃ Σ9 by the operators σ(h1) = − i 2 (−e1,5 + e2,6 + e3,7 − e4,8 + e9,13 − e10,14 − e11,15 + e12,16), σ(h2) = − i 2 (e1,11 − e2,12 − e3,9 + e4,10 + e5,15 − e6,16 − e7,13 + e8,14), σ(h3) = − i 2 (e1,7 − e2,8 − e3,5 + e4,6 + e9,15 − e10,16 − e11,13 + e12,14), σ(h4) = − i 2 (−e1,7 − e2,8 + e3,5 + e4,6 − e9,15 − e10,16 + e11,13 + e12,14), 5One usually chooses a different realization of the Lie algebra so(9,C) in order to make the elements of the Cartan subalgebra diagonal matrices, but that realization is less convenient for our purposes here. 28 D. Artacho and J. Hofmann and 2σ(X1) = e1,3 − ie1,7 − ie1,9 − e1,13 − e2,4 − ie2,8 − ie2,10 + e2,14 − ie3,5 − ie3,11 + e3,15 − ie4,6 − ie4,12 − e4,16 + e5,7 + e5,9 − ie5,13 − e6,8 − e6,10 − ie6,14 − e7,11 − ie7,15 + e8,12 − ie8,16 − e9,11 − ie9,15 + e10,12 − ie10,16 − ie11,13 − ie12,14 − e13,15 + e14,16, 2σ(X2) = −ie1,4 + e1,6 − e1,10 + ie1,16 − ie2,3 − e2,5 + e2,9 + ie2,15 + e3,8 − e3,12 − ie3,14 − e4,7 + e4,11 − ie4,13 − ie5,8 + ie5,12 − e5,14 − ie6,7 + ie6,11 + e6,13 − ie7,10 − e7,16 − ie8,9 + e8,15 − ie9,12 + e9,14 − ie10,11 − e10,13 + e11,16 − e12,15 − ie13,16 − ie14,15, 2σ(X3) = −2e1,3 − 2ie1,5 − 2ie3,7 + 2e5,7 − 2e9,11 − 2ie9,13 − 2ie11,15 + 2e13,15, 2σ(X4) = ie1,4 + e1,6 − ie2,3 − e2,5 − e3,8 + e4,7 + ie5,8 − ie6,7 + ie9,12 + e9,14 − ie10,11 − e10,13 − e11,16 + e12,15 + ie13,16 − ie14,15. Considering the action of the lifts σ̃(hi), σ̃(Xi) ∈ spin(16)C in the spin representation, a straight- forward calculation using computer algebra software yields three linearly independent joint eigenvectors for the σ̃(hi) which are simultaneously annihilated by the action of each σ̃(Xi) (i.e., highest weight vectors). The corresponding weights are 1 2 (3v1 + v2 + v3 + v4) = ω1 + ω4, v1 + v2 + v3 = ω3, 2v1 = 2ω1, and the assertion that Σ− 16 ≃ V (ω1 + ω4) and Σ+ 16 ≃ V (ω3)⊕ V (2ω1) may be deduced from [18, Section 7]. ■ Note that the preceding lemma immediately recovers the fact that the invariant spin struc- ture carries no invariant spinors. It also allows us to readily describe the smallest twisting for which OP2 admits invariant twisted spin9 spinors. Theorem 3.26. The F4-invariant spinor type of OP2 is σ ( OP2, F4 ) = 9, and the twisting of the spinor bundle which realises this is m = 3. Furthermore, the space of invariant 3-twisted spin9 spinors has complex dimension 4. Proof. First, we recall that every representation of so(9,C) is self-dual (see, e.g., [40]). There- fore, using a similar argument as in the proof of Proposition 3.18, and in light of the preceding lemma, it suffices to show that Σ⊗3 9 is the smallest odd tensor power of Σ9 ≃ V (ω4) which con- tains a copy of V (ω1 + ω4), V (ω3), or V (2ω1). It is easily verified using, e.g., the LiE software package [36] that Σ⊗3 9 ≃ 5V (ω4)⊕ V (3ω4)⊕ 2V (ω3 + ω4)⊕ 3V (ω2 + ω4)⊕ 4V (ω1 + ω4). (3.12) Finally, using self-duality, it follows from (3.11) and (3.12) that dimC ( Σ3 16,9 ) inv = dimC ( Σ16 ⊗ Σ⊗3 9 )Spin(9) = dimCHomSpin(9) ( Σ16,Σ ⊗3 9 ) = 4. ■ Now we examine more closely the invariant 3-twisted spin9 spinors from the preceding theo- rem. This 4-dimensional space corresponds to the pairings of the 4 copies of V (ω1+ω4) in (3.12) with the single copy in (3.11), so in order to obtain formulas for the spinors we first need to clarify the algebraic structure of this representation. With all notation as above, one finds using computer algebra software an explicit highest weight vector w0 (unique up to scaling) gener- ating Σ− 16 ≃ V (ω1 + ω4) ⊆ Σ16, and one may verify furthermore that any other weight vector The Geometry of Generalised Spinr Spinors on Projective Spaces 29 can be obtained from w0 by applying at most 18 lowering operators Yi, i = 1, 2, 3, 4. Writing YI := Yi1 .Yi2 . . . Yik for a multi-index I = {i1, . . . , ik}, one possible minimal choice of multi- indices ⋃18 k=0{Ik,ℓ} µk ℓ=1 generating V (ω1 + ω4) is given in Table 3, where µk denotes the number of k-multi-indices in the generating set. In what follows we describe explicitly the invariant spinors, using a more sophisticated version of the technique from the proof of Proposition 3.20. Theorem 3.27. A basis for the space of F4-invariant 3-twisted spin9 spinors on OP2 is given by ψp := 18∑ k=0 µk∑ ℓ=1 ( ̂YIk,ℓ .w0 )♯ ⊗ (YIk,ℓ .wp), p = 1, 2, 3, 4, (3.13) where ♯ : Σ∗ 16 → Σ16 is the musical isomorphism, wp (p = 1, 2, 3, 4) denote highest weight vectors for the four copies of Σ− 16 inside Σ⊗3 9 , the indices Ik,ℓ are as in Table 3, and for any (YIk,ℓ .w0) ∈ Σ− 16 ⊆ Σ16 we denote by ̂YIk,ℓ .w0 ∈ Σ∗ 16 the corresponding dual map sending YIk′,ℓ′ .w0 7→ δk,k′δℓ,ℓ′ and Σ+ 16 7→ 0. Proof. From the preceding discussion and Table 3, we have the highest weight vector w0 for Σ− 16 ⊆ Σ16, together with explicit sequences of lowering operators Yi generating this subrepre- sentation. Altogether this gives four spin(9)C-module isomorphisms Tp : Σ − 16 → Σ⊗3 9 , p = 1, 2, 3, 4 defined by Tp : YIk,ℓ .w0 7→ YIk,ℓ .wp, k = 0, . . . , 18, ℓ = 1, . . . , µk, where the Ik,ℓ are as in Table 3 and we use the convention Y∅ = Id. By abuse of notation, we also denote by Tp the extensions to all of Σ16 by Σ+ 16 7→ 0. The spinors ψp are the images of the Tp under the spin(9)C-module isomorphism (Σ− 16) ∗ ⊗ Σ⊗3 9 ≃ Σ− 16 ⊗ Σ⊗3 9 , which are precisely given by (3.13). ■ Finally, we give the differential equation satisfied by the spinors from Theorem 3.27. To begin, we need to first specify a connection on the vector bundle A associated to the auxil- iary SO(9)-bundle of the spin9 structure. Note that A is associated to the principal Spin(9) bundle F4 → F4/ Spin(9) by the composition of the covering map λ9 : Spin(9) → SO(9) with the standard representation ρstd : SO(9) → GL ( R9 ) . Indeed, there is a natural choice of in- variant connection defined on A as follows. The structure of m ≃ ΣR 9 as a Clifford module for Cl9 gives 9 linearly independent endomorphisms, corresponding to Clifford multiplication by an orthonormal set of basis vectors for R9. By slightly modifying the Clifford multiplication (see [18, Section 2]), one obtains endomorphisms Ti : m → m, i = 1, . . . , 9 satisfying the mod- ified Clifford relations Ti ◦ Tj + Tj ◦ Ti = 2δi,j Id, T ∗ i = Ti, trTi = 0 for i = 1, . . . , 9. In this description, the isotropy image σ(Spin(9)) ⊆ SO(m) ⊆ End(m) coincides with the normaliser of the 9-dimensional subspace T := spanR{T1, . . . , T9} ⊆ End(m) [18, p. 132]: σ(Spin(9)) = { g ∈ SO(m) | gT g−1 = T } . The 9-dimensional Spin(9)-representation T (action via conjugation) is isomorphic to ρstd ◦ λ9 (since there is only one irreducible real 9-dimensional representation of Spin(9) up to isomor- phism), hence we have A ∼= F4 ×Spin(9) T ⊆ F4 ×Spin(9) End(m) ∼= End ( T ( OP2 )) . In particular, A naturally inherits a connection ∇End from the extension of the Levi-Civita connection to the endomorphism bundle, whose Nomizu map vanishes on m. In light of Propo- sition 2.21, we finally see that the invariant twisted spinors found above are parallel: Theorem 3.28.The 4-dimensional space of F4-invariant 3-twisted spin9 spinors on OP2 is span- ned by parallel spinors for the connection ∇g,End := ∇g ⊗ ( ∇End )⊗3 . 30 D. Artacho and J. Hofmann k µk Ik,ℓ (ℓ = 1, . . . , µk) 0 1 ∅ 1 2 {1}, {4} 2 3 {2, 1}, {1, 4}, {3, 4} 3 5 {3, 2, 1}, {2, 1, 4}, {1, 3, 4}, {2, 3, 4}, {4, 3, 4} 4 6 {4, 3, 2, 1}, {3, 2, 1, 4}, {2, 1, 3, 4}, {4, 1, 3, 4}, {1, 2, 3, 4}, {4, 2, 3, 4} 5 8 {3, 4, 3, 2, 1}, {4, 4, 3, 2, 1}, {2, 3, 2, 1, 4}, {4, 3, 2, 1, 4}, {1, 2, 1, 3, 4}, {4, 2, 1, 3, 4}, {4, 1, 2, 3, 4}, {3, 4, 2, 3, 4} 6 10 {2, 3, 4, 3, 2, 1}, {4, 3, 4, 3, 2, 1}, {3, 4, 4, 3, 2, 1}, {4, 4, 4, 3, 2, 1}, {1, 2, 3, 2, 1, 4}, {4, 2, 3, 2, 1, 4}, {4, 1, 2, 1, 3, 4}, {3, 4, 2, 1, 3, 4}, {3, 4, 1, 2, 3, 4}, {4, 3, 4, 2, 3, 4} 7 11 {1, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 3, 2, 1}, {4, 4, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 3, 2, 1}, {4, 3, 4, 4, 3, 2, 1}, {4, 1, 2, 3, 2, 1, 4}, {3, 4, 2, 3, 2, 1, 4}, {3, 4, 1, 2, 1, 3, 4}, {4, 3, 4, 2, 1, 3, 4}, {2, 3, 4, 1, 2, 3, 4}, {4, 3, 4, 1, 2, 3, 4} 8 12 {4, 1, 2, 3, 4, 3, 2, 1}, {3, 4, 2, 3, 4, 3, 2, 1}, {4, 4, 2, 3, 4, 3, 2, 1}, {3, 4, 4, 3, 4, 3, 2, 1}, {4, 4, 4, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 3, 2, 1}, {3, 4, 1, 2, 3, 2, 1, 4}, {4, 3, 4, 2, 3, 2, 1, 4}, {2, 3, 4, 1, 2, 1, 3, 4}, {4, 3, 4, 1, 2, 1, 3, 4}, {4, 2, 3, 4, 1, 2, 3, 4} 9 12 {3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 2, 3, 4, 3, 2, 1}, {3, 4, 4, 2, 3, 4, 3, 2, 1}, {4, 4, 4, 2, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 3, 2, 1}, {2, 3, 4, 1, 2, 3, 2, 1, 4}, {4, 3, 4, 1, 2, 3, 2, 1, 4}, {4, 2, 3, 4, 1, 2, 1, 3, 4}, {3, 4, 2, 3, 4, 1, 2, 3, 4} 10 12 {2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {3, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 3, 4, 2, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 2, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 1, 2, 3, 2, 1, 4}, {3, 4, 2, 3, 4, 1, 2, 1, 3, 4}, {4, 3, 4, 2, 3, 4, 1, 2, 3, 4} 11 11 {4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {3, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1}, {4, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 2, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 3, 4, 3, 2, 1}, {3, 4, 2, 3, 4, 1, 2, 3, 2, 1, 4}, {4, 3, 4, 2, 3, 4, 1, 2, 1, 3, 4} 12 10 {3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {3, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 2, 3, 4, 1, 2, 3, 2, 1, 4} 13 8 {4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {3, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 1, 2, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1} 14 6 {4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {2, 3, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 3, 4, 2, 3, 4, 3, 2, 1} 15 5 {3, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {1, 2, 3, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 3, 4, 1, 2, 3, 4, 3, 2, 1} 16 3 {2, 3, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 3, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 1, 2, 3, 4, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1} 17 2 {1, 2, 3, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1}, {4, 2, 3, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1} 18 1 {4, 1, 2, 3, 4, 4, 3, 4, 2, 3, 4, 1, 2, 3, 4, 3, 2, 1} Table 3. Ordered sequences of lowering operators generating V (ω1 + ω4). Acknowledgements The authors are grateful to Travis Schedler for his contributions to the representation-theoretical aspect of the paper, and to Marie-Amélie Lawn for her comments and fruitful discussions. We The Geometry of Generalised Spinr Spinors on Projective Spaces 31 are grateful to the referees for their helpful comments. D. Artacho is funded by the UK Engi- neering and Physical Sciences Research Council (EPSRC), grant EP/W5238721. J. Hofmann was supported by the Engineering and Physical Sciences Research Council [EP/L015234/1, EP/W522429/1]; the EPSRC Centre for Doctoral Training in Geometry and Number Theory (The London School of Geometry and Number Theory: University College London, King’s College London, and Imperial College London); and a DAAD Short Term Research Grant for a research stay at Philipps-Universität Marburg. References [1] Agricola I., Chiossi S.G., Friedrich T., Höll J., Spinorial description of SU(3)- and G2-manifolds, J. Geom. Phys. 98 (2015), 535–555, arXiv:1411.5663. [2] Agricola I., Hofmann J., Lawn M.-A., Invariant spinors on homogeneous spheres, Differential Geom. Appl. 89 (2023), 102014, 57 pages, arXiv:2023.10201. [3] Agricola I., Naujoks H., Theiss M., Geometry of principal fibre bundles, in preparation. [4] Albanese M., Milivojević A., Spinh and further generalisations of spin, J. Geom. Phys. 164 (2021), 104174, 13 pages, arXiv:2021.10417. [5] Artacho D., Lawn M.-A., Generalised Spinr structures on homogeneous spaces, arXiv:2303.05433. [6] Arvanitoyeorgos A., An introduction to Lie groups and the geometry of homogeneous spaces, Stud. Math. Libr., Vol. 22, American Mathematical Society, Providence, RI, 2003. [7] Baez J.C., The octonions, Bull. Amer. Math. Soc. (N.S.) 39 (2002), 145–205, arXiv:math/0105155. [8] Bär C., Real Killing spinors and holonomy, Comm. Math. Phys. 154 (1993), 509–521. [9] Baum H., Friedrich T., Grunewald R., Kath I., Twistors and Killing spinors on Riemannian manifolds, Teubner-Texte Math., Vol. 124, B.G. Teubner Verlagsgesellschaft mbH, Stuttgart, 1991. [10] Bayard P., Lawn M.-A., Roth J., Spinorial representation of submanifolds in Riemannian space forms, Pacific J. Math. 291 (2017), 51–80, arXiv:1602.02919. [11] Besse A.L., Einstein manifolds, Class. Math., Springer, Berlin, 2008. [12] Conti D., Salamon S., Generalized Killing spinors in dimension 5, Trans. Amer. Math. Soc. 359 (2007), 5319–5343, arXiv:math.DG/0508375. [13] Cortés V., Lazaroiu C., Shahbazi C.S., Spinors of real type as polyforms and the generalized Killing equation, Math. Z. 299 (2021), 1351–1419, arXiv:1911.08658. [14] Espinosa M., Herrera R., Spinorially twisted spin structures, I: Curvature identities and eigenvalue estimates, Differential Geom. Appl. 46 (2016), 79–107, arXiv:1409.6246. [15] Friedrich T., Der erste Eigenwert des Dirac-Operators einer kompakten, Riemannschen Mannigfaltigkeit nichtnegativer Skalarkrümmung, Math. Nachr. 97 (1980), 117–146. [16] Friedrich T., On the spinor representation of surfaces in Euclidean 3-space, J. Geom. Phys. 28 (1998), 143–157, arXiv:dg-ga/9712021. [17] Friedrich T., Dirac operators in Riemannian geometry, Grad. Stud. Math., Vol. 25, American Mathematical Society, Providence, RI, 2000, arXiv:gr-qc/9507046. [18] Friedrich T., Weak Spin(9)-structures on 16-dimensional Riemannian manifolds, Asian J. Math. 5 (2001), 129–160, arXiv:math/9912112. [19] Friedrich T., Kath I., Moroianu A., Semmelmann U., On nearly parallel G2-structures, J. Geom. Phys. 23 (1997), 259–286. [20] Friedrich T., Trautman A., Spin spaces, Lipschitz groups, and spinor bundles, Ann. Global Anal. Geom. 18 (2000), 221–240, arXiv:math/9901137. [21] Fulton W., Harris J., Representation theory, Grad. Texts in Math., Vol. 129, Springer, New York, 1991. [22] Gillard J., Gran U., Papadopoulos G., The spinorial geometry of supersymmetric backgrounds, Classical Quantum Gravity 22 (2005), 1033–1076, arXiv:hep-th/0410155. [23] Goodman R., Wallach N.R., Symmetry, representations, and invariants, Grad. Texts in Math., Vol. 255, Springer, Dordrecht, 2009. https://doi.org/10.1016/j.geomphys.2015.08.023 https://doi.org/10.1016/j.geomphys.2015.08.023 https://arxiv.org/abs/1411.5663 https://doi.org/10.1016/j.difgeo.2023.102014 https://arxiv.org/abs/2023.10201 https://doi.org/10.1016/j.geomphys.2021.104174 https://arxiv.org/abs/2021.10417 https://arxiv.org/abs/2303.05433 https://doi.org/10.1090/stml/022 https://doi.org/10.1090/S0273-0979-01-00934-X https://arxiv.org/abs/math/0105155 https://doi.org/10.1007/BF02102106 https://doi.org/10.2140/pjm.2017.291.51 https://arxiv.org/abs/1602.02919 https://doi.org/10.1007/978-3-540-74311-8 https://doi.org/10.1090/S0002-9947-07-04307-3 https://arxiv.org/abs/math.DG/0508375 https://doi.org/10.1007/s00209-021-02726-6 https://arxiv.org/abs/1911.08658 https://doi.org/10.1016/j.difgeo.2016.01.008 https://arxiv.org/abs/1409.6246 https://doi.org/10.1002/mana.19800970111 https://doi.org/10.1016/S0393-0440(98)00018-7 https://arxiv.org/abs/dg-ga/9712021 https://doi.org/10.1090/gsm/025 https://doi.org/10.1090/gsm/025 https://arxiv.org/abs/gr-qc/9507046 https://doi.org/10.4310/AJM.2001.v5.n1.a9 https://arxiv.org/abs/math/9912112 https://doi.org/10.1016/S0393-0440(97)80004-6 https://doi.org/10.1023/A:1006713405277 https://arxiv.org/abs/math/9901137 https://doi.org/10.1007/978-1-4612-0979-9 https://doi.org/10.1088/0264-9381/22/6/009 https://doi.org/10.1088/0264-9381/22/6/009 https://arxiv.org/abs/hep-th/0410155 https://doi.org/10.1007/978-0-387-79852-3 32 D. Artacho and J. Hofmann [24] Grunewald R., Six-dimensional Riemannian manifolds with a real Killing spinor, Ann. Global Anal. Geom. 8 (1990), 43–59. [25] Herrera H., Herrera R., Spinq manifolds admitting parallel and Killing spinors, J. Geom. Phys. 57 (2007), 1525–1539. [26] Herrera R., Santana N., Spinorially twisted Spin structures. II: Twisted pure spinors, special Riemannian holonomy and Clifford monopoles, SIGMA 15 (2019), 072, 48 pages, arXiv:1506.07681. [27] Hirzebruch F., Slodowy P., Elliptic genera, involutions, and homogeneous spin manifolds, Geom. Dedicata 35 (1990), 309–343, arXiv:0712.57010. [28] Hofmann J., Homogeneous Sasakian and 3-Sasakian structures from the spinorial viewpoint, Adv. Math. 439 (2024), 109493, 39 pages, arXiv:2024.10949. [29] Itzkowitz G., Rothman S., Strassberg H., A note on the real representations of SU(2,C), J. Pure Appl. Algebra 69 (1991), 285–294. [30] Kusner R., Schmitt N., The spinor representation of surfaces in space, arXiv:dg-ga/9610005. [31] Lawson Jr. H.B., Spinh manifolds, SIGMA 19 (2023), 012, 7 pages, arXiv:2301.09683. [32] Lawson Jr. H.B., Michelsohn M.-L., Spin geometry, Princeton Math. Ser., Vol. 38, Princeton University Press, Princeton, NJ, 1989. [33] Lazaroiu C.I., Shahbazi C.S., Complex Lipschitz structures and bundles of complex Clifford modules, Dif- ferential Geom. Appl. 61 (2018), 147–169, arXiv:1711.07765. [34] Lazaroiu C.I., Shahbazi C.S., Real pinor bundles and real Lipschitz structures, Asian J. Math. 23 (2019), 749–836, arXiv:1606.07894. [35] Lazaroiu C.I., Shahbazi C.S., Dirac operators on real spinor bundles of complex type, Differential Geom. Appl. 80 (2022), 101849, 53 pages, arXiv:2022.10184. [36] Leeuwen M., Cohen A.M., Lisser B., LiE: a computer algebra package for Lie group computations, 1988, available at http://wwwmathlabo.univ-poitiers.fr/~maavl/LiE/. [37] Moroianu A., Parallel and Killing spinors on Spinc manifolds, Comm. Math. Phys. 187 (1997), 417–427. [38] Nomizu K., Invariant affine connections on homogeneous spaces, Amer. J. Math. 76 (1954), 33–65. [39] Onǐsčik A., Transitive compact transformation groups, in Eleven Papers on Topology and Algebra, 2nd ed., Amer. Math. Soc. Transl., Vol. 55, Springer, New York, 1966, 153–194. [40] Samelson H., Notes on Lie algebras, 2nd ed., Universitext, Springer, New York, 1990. [41] Wang M.Y., Parallel spinors and parallel forms, Ann. Global Anal. Geom. 7 (1989), 59–68. [42] Wernli K., Lecture notes on spin geometry, arXiv:1911.09766. [43] Ziller W., Lie groups. Representation theory and symmetric spaces, 2010, available at https://www2.math. upenn.edu/~wziller/math650/LieGroupsReps.pdf. [44] Ziller W., Homogeneous Einstein metrics on spheres and projective spaces, Math. Ann. 259 (1982), 351–358. https://doi.org/10.1007/BF00055017 https://doi.org/10.1016/j.geomphys.2007.01.002 https://doi.org/10.3842/SIGMA.2019.072 https://arxiv.org/abs/1506.07681 https://doi.org/10.1007/BF00147351 https://arxiv.org/abs/0712.57010 https://doi.org/10.1016/j.aim.2024.109493 https://arxiv.org/abs/2024.10949 https://doi.org/10.1016/0022-4049(91)90023-U https://doi.org/10.1016/0022-4049(91)90023-U https://arxiv.org/abs/dg-ga/9610005 https://doi.org/10.3842/SIGMA.2023.012 https://arxiv.org/abs/2301.09683 https://doi.org/10.1016/j.difgeo.2018.08.006 https://doi.org/10.1016/j.difgeo.2018.08.006 https://arxiv.org/abs/1711.07765 https://doi.org/10.4310/AJM.2019.v23.n5.a3 https://arxiv.org/abs/1606.07894 https://doi.org/10.1016/j.difgeo.2022.101849 https://doi.org/10.1016/j.difgeo.2022.101849 https://arxiv.org/abs/2022.10184 http://wwwmathlabo.univ-poitiers.fr/~maavl/LiE/ https://doi.org/10.1007/s002200050142 https://doi.org/10.2307/2372398 https://doi.org/10.1090/trans2/055 https://doi.org/10.1007/978-1-4613-9014-5 https://doi.org/10.1007/BF00137402 https://arxiv.org/abs/1911.09766 https://www2.math.upenn.edu/~wziller/math650/LieGroupsReps.pdf https://www2.math.upenn.edu/~wziller/math650/LieGroupsReps.pdf https://doi.org/10.1007/BF01456947 1 Introduction 2 Preliminaries 2.1 Invariant metrics on reductive homogeneous spaces 2.2 Some notation 2.3 Invariant spin^r structures 2.4 Exterior forms approach to the spin representation 2.5 Invariant spin^r spinors 2.6 Special spin^r spinors 3 Projective spaces 3.1 Hermitian complex projective space 3.1.1 Invariant spin^r structures 3.1.2 Invariant spin^r spinors 3.1.3 Special spin^r spinors 3.2 Symplectic complex projective space 3.2.1 Invariant spin^r structures 3.2.2 Invariant spin^r spinors 3.2.3 Special spin^r spinors 3.3 Quaternionic projective space 3.3.1 Invariant spin^r structures 3.3.2 Invariant spin^r spinors 3.4 Octonionic projective plane 3.4.1 Invariant spin^r structures 3.4.2 Invariant spin^r spinors References
id nasplib_isofts_kiev_ua-123456789-212874
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1815-0659
language English
last_indexed 2026-03-21T18:30:54Z
publishDate 2025
publisher Інститут математики НАН України
record_format dspace
spelling Artacho, Diego
Hofmann, Jordan
2026-02-13T13:49:09Z
2025
The Geometry of Generalised Spinʳ Spinors on Projective Spaces. Diego Artacho and Jordan Hofmann. SIGMA 21 (2025), 017, 32 pages
1815-0659
2020 Mathematics Subject Classification: 53C27; 15A66; 57R15
arXiv:2406.18337
https://nasplib.isofts.kiev.ua/handle/123456789/212874
https://doi.org/10.3842/SIGMA.2025.017
In this paper, we adapt the characterisation of the spin representation via exterior forms to the generalised spinʳ context. We find new invariant spinʳ spinors on the projective spaces ℂℙⁿ, ℍℙⁿ, and the Cayley plane ℙ² for all their homogeneous realisations. Specifically, for each of these realisations, we provide a complete description of the space of invariant spinʳ spinors for the minimum value of for which this space is non-zero. Additionally, we demonstrate some geometric implications of the existence of special spinʳ spinors on these spaces.
The authors are grateful to Travis Schedler for his contributions to the representation-theoretical aspect of the paper, and to Marie-Am´elie Lawn for her comments and fruitful discussions. We are grateful to the referees for their helpful comments. D. Artacho is funded by the UK Engineering and Physical Sciences Research Council (EPSRC), grant EP/W5238721. J. Hofmann was supported by the Engineering and Physical Sciences Research Council [EP/L015234/1, EP/W522429/1]; the EPSRC Centre for Doctoral Training in Geometry and Number Theory (The London School of Geometry and Number Theory: University College London, King’s College London, and Imperial College London); and a DAAD Short Term Research Grant for a research stay at Philipps-Universität Marburg.
en
Інститут математики НАН України
Symmetry, Integrability and Geometry: Methods and Applications
The Geometry of Generalised Spinʳ Spinors on Projective Spaces
Article
published earlier
spellingShingle The Geometry of Generalised Spinʳ Spinors on Projective Spaces
Artacho, Diego
Hofmann, Jordan
title The Geometry of Generalised Spinʳ Spinors on Projective Spaces
title_full The Geometry of Generalised Spinʳ Spinors on Projective Spaces
title_fullStr The Geometry of Generalised Spinʳ Spinors on Projective Spaces
title_full_unstemmed The Geometry of Generalised Spinʳ Spinors on Projective Spaces
title_short The Geometry of Generalised Spinʳ Spinors on Projective Spaces
title_sort geometry of generalised spinʳ spinors on projective spaces
url https://nasplib.isofts.kiev.ua/handle/123456789/212874
work_keys_str_mv AT artachodiego thegeometryofgeneralisedspinrspinorsonprojectivespaces
AT hofmannjordan thegeometryofgeneralisedspinrspinorsonprojectivespaces
AT artachodiego geometryofgeneralisedspinrspinorsonprojectivespaces
AT hofmannjordan geometryofgeneralisedspinrspinorsonprojectivespaces