Arxiv:2105.11129V1 [Math.DG] 24 May 2021 Eri Aiod.Te,I Sitrsigt N Nenti Aiodwit Manifold Einstein an ﬁnd to ﬁelds

arXiv:2105.11129v1 [math.DG] 24 May 2021 eri aiod.Te,i sitrsigt n nEnti aiodwit manifold Einstein an find to fields. interesting Rarita-Schwinger has is which it constant Einstein Then, manifolds. cons Einstein gebraic positive Rarita symmetric Kähler with parallel manifolds, quaternionic manifolds Einstein with fields, some Rarita-Schwinger manifolds find the also Eins They negative classified with fields. [11] manifold Semmelmann compact and connected simply Homma a as manifold this aiod eefis tde yA ry[] eryKalrmnfl sc is Kähler manifold nearly N A proposed. [6]. Gray [11] A. Semmelmann by and studied Homma first which were constant questions manifolds Einstein open positive the have of which Kähler 6-manifolds, nearly on Rarita-Sc many with manifold B compact dimension a [12]. given Tomihisa exists and any oper there Rarita-Schwinger Homma that by the proved ass computed of [1] was in spectrum spaces important The symmetric is [23]). crete operator (cf. Rarita-Schwinger Einst genus The infinitesimal elliptic and spinors. fields Killing these with Rarita-Schwin between about relationship studies the some examined been have grav there hypothesized example su mathematics, the in for and of partner because, relativity supersymmetric role in the important gravitino, especially an physics, play In fields Rarita-Schwinger Dirac free. classical divergence the with in W of sections fields, by generalization as introduced a fields Rarita-Schwinger fermions, operator, consider Rarita-Schwinger spin-3/2 can of We equation [17]. Schwinger field relativistic the is AIASHIGRFED NNAL K NEARLY ON FIELDS RARITA-SCHWINGER aiaShigrfilsaesltoso h aiaShigrequa Rarita-Schwinger the of solutions are fields Rarita-Schwinger nti ae,w netgt hte hr xs nntiil Rar (non-trivial) exist there whether investigate we paper, this In onie ihtedrc u facranegnpc fteLpaeo Laplace the of eigenspace spinors. certain Killing to a deformation the technique of infinitesimal of sum the same space direct of the space the Applying the with fi that coincides Rarita-Schwinger 3-forms. find the also harmonic of we space the theory, the of L that the space show relat we and the result, the connection a clarify Hermitian As the we tion. for them, operators investigate differential to some order In Kähler manifolds. Abstract. esuyRrt-cwne ed n6dmninlcmatsrc n strict compact 6-dimensional on fields Rarita-Schwinger study We n OAON N AUATOMIHISA TAKUMA AND OHNO SOMA ≥ .I atclr if particular, In 4. 1. Introduction 1 n sdvsbeb ,te ecntake can we then 4, by divisible is HE MANIFOLDS AHLER ¨ tn nteohrhand, other the On iton. fKligspinors Killing of s lscicdswith coincides elds v-iiaconnec- evi-Civita prtrfrspin-1/2 for operator osi between ionship eao n the and perator e ed:Wn [22] Wang fields: ger pcs n oeal- some and spaces, t-cwne fields ita-Schwinger in hsequation This tion. deformation a hydsrb the describe they , tro oecon- some on ator hssle one solves This . esrn theory, perstring i deformations ein wne ed in fields hwinger h enlo the of kernel the aiaadJ. and Rarita . a n Mazzeo är and atadmitting tant enconstant. tein cainwith ociation al Kählerearly haracterized early -Schwinger positive h 2 SOMAOHNOANDTAKUMATOMIHISA

as an almost Hermitian manifold (M,g,J) with (∇X J)X = 0 for any vector field X, where ∇ is the Levi-Civita connection. The 6-dimensional case stands out because of the lowest dimension in which non-Kähler nearly Kähler manifolds appear, and because of the existence of real Killing spinors. Thus, 6-dimensional non-Kähler nearly Kähler manifolds are Einstein manifolds with positive scalar curvature. Furthermore, in this case, they have SU(3)-structures. We have three major ideas to study Rarita-Schwinger fields on nearly Kähler manifolds. The first in Section 2 is to decompose the tangent bundle, the spinor bundle S1/2, and S3/2 by using the Killing spinors and the SU(3)-structures. Here, in Section 2, we also define Rarita-Schwinger fields and nearly Kähler manifolds. The second in Section 3 is to investigate relationships between various differential operators for the Levi-Civita connection and the Hermitian connection. The third in Section 4 is to rewrite the Rarita-Schwinger equation into the simultaneous equation consisting of 2- forms and 3-forms by using these decompositions and relationships. As a result, we get the main result in Section 4. Theorem A. The space of the Rarita-Schwinger fields is isomorphic to the space of the harmonic 3-forms on 6-dimensional compact non-Kähler nearly Kähler manifolds.

In particular, in Section 5, we know that there is a two dimensional space of the Rarita-Schwinger fields on the nearly Kähler S3 × S3. This is the first example of a six-dimensional Einstein manifold with positive scalar curvature admitting Rarita- Schwinger fields. In the same way as we prove Theorem A, we can obtain a theorem about deformation theory of Killing spinors. The general theory of deformations of Killing spinors was first developed by Wang [22]. van Coevering [20] recently studied deformations of Killing spinors on 3-Sasakian manifolds. We study deformations of Killing spinors on nearly Kähler manifolds. Let E(λ) be the λ-eigenspace of the Laplace operator restricted to co-closed primitive (1, 1)-forms, and K+ be the space of the Killing spinors with the 1 Killing number 2 . Then our result in Section 5 is Theorem B. Let (M 6,g,J) be a 6-dimensional compact non-Kähler nearly Kähler manifold. Then the space of the infinitesimal deformations of Killing spinors is isomorphic to the space E(12) ⊕ K+.

There have been some works about deformation theory on nearly Kähler 6-manifolds ([5], [13], [15]). These works have been conducted on 6-dimensional compact non-Kähler nearly Kähler manifolds. Moroianu, Nagy and Semmelmann [13] showed that, except for the round sphere S6, the space of the infinitesimal deformations of nearly Kähler structures is isomorphic to the space E(12). As described in the paper [8], there is a one-to-one correspondence between the space of the Killing spinors (modulo constant rescaling) and the space of nearly Kähler structures. Since the space of the Killing 6 spinors K+ is one dimensional except for S , Theorem B is a reproof of the result of [13]. RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 3

2. Preliminalies

2.1. Rarita-Schwinger fields. In this subsection, we introduce Rarita-Schwinger fields and some related operators. Details such as definitions and Weitzenböck formulas are listed in [11], [22].

Let (M,g)bean n-dimensional Riemannian spin manifold with the spinor bundle S1/2 and the complexiﬁed tangent bundle T M C. We consider the twisted Dirac operator on C S1/2 ⊗ T M , n

C DTM = (ek · ⊗ IdTM ) ◦∇ek , k=1 X C where ∇ is the covariant derivative on S1/2 ⊗ T M induced from the Levi-Civita connection and ek· is the Clifford multiplication by an orthonormal frame {ek} of T M. We C define the vector bundle S3/2 as ker Π for the bundle map Π : S1/2 ⊗ T M ∋ ζ ⊗ X 7→ C ∼ X · ζ ∈ S1/2. With respect to the Spin(n) decomposition S1/2 ⊗ T M = S1/2 ⊕ S3/2, we can write DTM as the 2 × 2 matrix 2−n ∗ n D 2P DTM = 2 , n P Q ! where D : Γ(S1/2) → Γ(S1/2) is the Dirac operator in the ordinary sense, P : Γ(S3/2) → ∗ Γ(S1/2) is the Penrose operator, and P is the formal adjoint operator of P . The operator Q : Γ(S3/2) → Γ(S3/2) is called the Rarita-Schwinger operator, which is a formally self-adjoint elliptic differential operator of first order. C A Rarita-Schwinger field is a section φ of S1/2 ⊗ T M that satisfies ∗ φ ∈ Γ(S3/2), P φ =0, and Qφ =0.

These equations are equivalent to φ ∈ Γ(S3/2) and DTM φ = 0. The result of Homma and Semmelmann [11, Proposition 4.1] tells us the following. If (M n,g), n ≥ 3, is a compact Einstein spin manifold with non-negative scalar curvature, then we have ker Q ∩ ker P ∗ = ker Q.

2.2. Nearly K¨ahler manifolds. An almost Hermitian manifold (M 2m,g,J) is called a nearly K¨ahler manifold if

(∇X J)X = 0 for all X ∈ Γ(T M), (2.1) where ∇ denotes the Levi-Civita connection of g. The canonical Hermitian connection ∇¯ , defined by 1 ∇¯ Y := ∇ Y − J(∇ J)Y for all X,Y ∈ Γ(T M), (2.2) X X 2 X satisfies ∇¯ g = 0 and ∇¯ J = 0. Note that the torsion of ∇¯ given by T¯(X,Y ) = −J(∇X J)Y vanishes iff (M,g,J) is a Kähler manifold. 4 SOMAOHNOANDTAKUMATOMIHISA

Nagy [16] proved that every compact simply connected nearly Kähler manifold is isometric to the product of several Riemannian manifolds. Each manifold is in one of the following classes of nearly Kähler manifold: Kähler manifolds, naturally reductive 3-symmetric spaces, twistor spaces over compact quaternion-Kähler manifolds with positive scalar curvature, and 6-dimensional nearly Kähler manifolds. From now on, we will consider only compact 6-dimensional strict (i.e. non-Kähler) nearly Kähler manifolds. It is an important fact that the Ricci curvature satisfies Ric =5g. In particular, these manifolds are Einstein manifolds. Note that compactness and completeness are equivalent on 6-dimensional strict nearly Kähler manifolds, because the Ricci curvature is positive. We denote as usual the Kähler form of M by ω := g(J·, ·). The tensor ψ+ := ∇ω is a 3-form by (2.1), and the tensor ψ− := ∗ψ+ is also a 3-form, where ∗ is the Hodge star operator. The ∇¯ -parallel complex volume form is represented as ψ+ + iψ−. Also, 1 + − we know the (real) volume form vol = volg coincides with 4 ψ ∧ ψ . It is a prominent property that the pair (ω, ψ+) leads to an SU(3)-structure on T M (cf. [9]).

For a vector field X, let AX denote the section AX = J(∇X J) of End T M. By ¯ 1 definition, the canonical Hermitian connection is written as ∇X = ∇X − 2 AX on the tangent bundle. For every endomorphism or 2-tensor B in End T M ∼= T ∗M ⊗ T ∗M, we denote by B⋆ in End T M the natural extension of B (cf. [15, p.3059]), where T M is a ¯ 1 tensor bundle. Then a similar equation ∇X = ∇X − 2 AX⋆ holds on T M. From now on, we identify T M with T ∗M using the metric without notice. As stated in [7], a spin structure exists on a 6-dimensional strict nearly Kähler manifold. The spin connection ∇ is obtained by pulling back the Levi-Civita connection to the spinor bundle. Similarly, we pull back the canonical Hermitian connection to the spinor bundle and also denote this connection on the spinor bundle as ∇¯ , which is written explicitly as 1 ∇¯ ζ = ∇ ζ − X y ψ− · ζ for all X ∈ Γ(T M),ζ ∈ Γ(S ). (2.3) X X 4 1/2 2.3. Algebraic results on nearly Kähler manifolds. In the subsections 2.3 and 2.4, we mainly summarize the contents of the articles [5] and [15] which we will use in later calculations. Assume that (M 6,g,J) is a 6-dimensional strict nearly Kähler manifold with the normalized scalar curvature scal = 30. We decompose the exterior bundles 2 3 ∧ M, ∧ M, and the spinor bundle S1/2. Also, we write the action of differential forms on the Killing spinors by the Clifford multiplication explicitly. The exterior bundle ∧2M decomposes into SU(3) irreducible components as follows: 2 ∼ R (2,0)+(0,2) (1,1) ∧ M = ω ⊕ ∧ M ⊕ ∧0 M, (1,1) y + where ∧0 M is the bundle of primitive (1, 1)-forms. The map X 7→ X ψ identifies the second summand with T M. The map h 7→ g(Jh·, ·) defines an isomorphism between + + Sym0 M and the third summand, where Sym M is the bundle of symmetric endomor- + + phisms commuting with J, and Sym0 M is the trace-free part of Sym M. Similarly, RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 5 one can decompose the exterior bundle ∧3M into SU(3) irreducible components 3 ∼ 1 (3,0)+(0,3) (2,1)+(1,2) ∧ M = ∧ M ∧ ω ⊕ ∧ M ⊕ ∧0 M. The second summand is a rank 2 trivial bundle spanned by the forms ψ±. The isomor- + − phism S 7→ S⋆ψ identifies Sym M, the bundle of symmetric endomorphisms anticom- muting with J, with the third summand. Here we note that the bundle of symmetric endomorphisms Sym M is equal to Sym+ M ⊕ Sym− M. Since the scalar curvature is normalized as scal = 30, M admits a unit Killing spinor 1 κ with the Killing number 2 . The Killing spinor κ defines a bundle map γ 7→ γ · κ from 0 1 6 ∧ M ⊕∧ M ⊕∧ M to S1/2. This map preserves the inner product, so must be injective. Since its domain and target have equal rank, it is an isomorphism: 0 1 6 ∼ ∧ M ⊕ ∧ M ⊕ ∧ M = S1/2. (2.4)

We give various actions on the Killing spinor, which is used in later calculations. From the construction of the Killing spinor (see [8]), the equation is obtained: JX · κ = vol ·X · κ = −X · vol ·κ for all X ∈ Γ(T M). (2.5) Next, the form ψ− ∈ Ω(3,0)+(0,3)M acts as follows (cf. [3, Lemma 2]): ψ− · κ =4κ. (2.6)

y + ∞ (2,0)+(0,2) (1,1) For any 2-form η = λω + Y ψ + η0 ∈ C (M)ω ⊕ Ω M ⊕ Ω0 M, we have η · κ =3λ vol ·κ +2JY · κ. (2.7) The proof of (2.7) is in [5, Lemma 3.3]. Moreover, for any vector ﬁelds X and Y , the identity holds:

X · Y · κ = −g(X,Y )κ + ω(X,Y )κ + AX Y · κ. (2.8)

2.4. The curvature endomorphism. In this subsection, we consider several operators related to the curvatures for the Levi-Civita connection ∇ and the canonical Hermitian connection ∇¯ . Furthermore, we introduce properties of these operators and relationships between them. Please refer to [15] for detailed definitions and proofs. Note that in the remaining part of this paper, we adopt the Einstein convention of summation on the repeated subscripts. We review the definition of the basic curvature tensor. The curvature tensor is defined by

R(W, X)Y := ∇W ∇X Y −∇X ∇W Y −∇[W,X]Y for all W,X,Y ∈ Γ(T M).

Also, for a vector ﬁeld Z, we denote g(R(W, X)Y,Z) by RWXYZ. Replacing the Levi- Civita connection ∇ with the canonical Hermitian connection ∇¯ , we deﬁne the curvature tensor R¯. The following relation holds for these two curvature tensors R and R¯. 6 SOMAOHNOANDTAKUMATOMIHISA

Lemma 2.1 ([7]). For any vector fields W,X,Y,Z, one has 1 RWXYZ = R¯WXYZ − g(Y, W )g(X,Z) − g(X,Y )g(Z, W ) 4 (2.9) − 3g(Y,JW )g(JX,Z)+3g(Y,JX)g(JW,Z)+2g(X,JW )g(JY,Z) . From the equation (2.9) and the fact that (M 6,g,J) is an Einstein manifold with Ric = 5g, we see that the Ricci curvature of R¯ satisfies Ric = 4g. Also, using the formula (2.9) and the first Bianchi identity, for any vector fields X,Y,Z, we have R¯(X,Y )Z + R¯(Y,Z)X + R¯(Z,X)Y (2.10) =2(g(JX,Y )JZ + g(JY,Z)JX + g(JZ,X)JY ) .

Next, we write the curvature operator R : ∧2M → ∧2M properly: 1 1 R(e ∧ e )= R e ∧ e = e ∧ R(e , e )e , i j 2 ijkl k l 2 k i j k where {ei} is a local orthonormal frame of T M. Let EM be a vector bundle associated to the oriented orthonormal frame bundle or the spin bundle on (M,g). We now deﬁne the curvature endomorphism q(R)= qE(R) ∈ End(EM) as 1 q(R)= (e ∧ e ) R(e ∧ e ) . (2.11) 2 i j ⋆ i j ⋆ In particular, we have q(R) = Ric = 5 Id on 1-forms. We can consider the curvature endomorphism for the canonical Hermitian connection q(R¯). It is straightforward to show that the curvature endomorphism q(R¯) satisﬁes q(R¯)= Ric = 4 Id on 1-forms. An interesting property of the curvature endomorphism q(R¯) is that it preserves all tensor bundles associated to SU(3)-representations. Refer to [15, p.3061] for the reason.

2.5. The actions of A. Let X be a vector field, and AX be the tensor field AX = J(∇X J). In Subsection 2.2, we introduced the extension AX⋆ to tensor fields Γ(End T M). The result of Moroianu and Semmelmann [15, Lemma 4.3] gives the following equations ¯ (1,1) related to the extension AX⋆ and the canonical Hermitian connection ∇ on ∧0 M, + − (2,1)+(1,2) Sym0 M, Sym M, and ∧0 M. We will essentially use these equations in the proof of Theorem A. (1,1) − Proposition 2.2 ([15]). Let ϕ and S be sections of ∧0 M and Sym M, respectively. + Sections h and σ are defined in Subsection 2.3 by g(Jh·, ·) := ϕ(·, ·) and σ := S⋆ψ . Then we have ¯ y + Aei⋆∇ei ϕ = −(Jδϕ) ψ . (2.12) ¯ Aei⋆∇ei σ = −2δS ∧ ω. (2.13) ¯ + (Aei⋆∇ei h)⋆ψ = 2δh ∧ ω − 4dϕ. (2.14) ¯ y + Aei⋆∇ei S = (δS ψ + δσ) ◦ J. (2.15) RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 7

Here δ denotes the co-diﬀerential on diﬀerential forms or the divergence operator when- ever applied to symmetric endomorphisms.

We remark that Moroianu and Semmelmann [15] make a mistake regarding a formula (2.14), but it does not aﬀect the results in [15] and ours.

Next, we introduce the other extension AX⋆ to 2-tensor ﬁelds, which also plays an important role in the proof of Theorem A. The extension AX⋆ is expressed for a 2-tensor ﬁeld α ⊗ β ∈ Γ(T M ⊗ T M) as g

AX⋆(α ⊗ β)= AX α ⊗ β + α ⊗ AX β.

On the other hand, we deﬁne AX⋆ for α ⊗ β ∈ Γ(T M ⊗ T M) as A (α ⊗ β)= A α ⊗ β − α ⊗ A β. X⋆ g X X To make formulas for A similar to those in Proposition 2.2, we prepare various tools. X⋆g First, for vector ﬁelds X, Y , Z, the identity holds: g AZ AX Y = −g(X,Z)Y + g(Y,Z)X + g(JX,Z)JY − g(JY,Z)JX. This identity yields

AAX Y = X ∧ Y − JX ∧ JY. (2.16) Next, we consider the induced action of a endomorphism B ∈ Γ(End T M) introduced in Subsection 2.2. The induced action of B on a p-form u is written as ∗ B⋆u = −B (ei) ∧ ei y u, (2.17) ∗ where B is the metric adjoint of B and {ei} is a local orthonormal frame of T M. + Taking AX as B and ψ as u, we get by a easy calculation (cf. [15, equation (2.9)]) + AX⋆ψ = −2X ∧ ω. (2.18) 2 2 (p,q)+(q,p) Taking J as B, we have (J⋆) = −(q − p) on Ω M. By this characterization, we ﬁnd that + ± (3,0)+(0,3) Γ(Sym M) ∋ h 7→ h⋆ψ ∈ Ω M, − ± (2,1)+(1,2) Γ(Sym M) ∋ S 7→ S⋆ψ ∈ Ω M, (1,1) ± (3,0)+(0,3) Ω M ∋ w 7→ w⋆ψ ∈ Ω M. ± (3,0)+(0,3) + − A section h⋆ψ ∈ Ω M is expressed as aψ + bψ at each point by using some constants a, b. By substituting some bases and calculating a, b, we get 1 h ψ± = − (tr h)ψ±. (2.19) ⋆ 2 + In particular, for a section h of Sym0 M, we have ± h⋆ψ =0. (2.20) In the same fashion, we have ± w⋆ψ =0. (2.21) 8 SOMAOHNOANDTAKUMATOMIHISA

± We consider the Hodge star operator ∗. The action of ∗ on S⋆ψ is shown in [13, equation (12)]:

+ − ∗ (S⋆ψ )= −S⋆ψ . (2.22)

For a primitive (1, 1)-form γ, we have

∗ (ei ∧ γ)= −ei y ∗γ = ei y (γ ∧ ω)=(ei y γ) ∧ ω + Jei ∧ γ. (2.23)

(1,1) Here, refer [5, Lemma 2.3] for the reason why ∗γ = −γ ∧ ω is valid for γ ∈ Ω0 M. Finally, by applying Schur’s Lemma to SU(3)-decompositions, we get some relations (cf. [15, Lemma 4.2]).

Lemma 2.3. The following relations hold:

(1,1) ei ∧ (Aei⋆w) = 0 for all w ∈ ∧0 M, (2.24) y (1,1) ei (Aei⋆w) = 0 for all w ∈ ∧0 M, (2.25) + (Aei⋆h)(ei) = 0 for all h ∈ Sym0 M, (2.26) − (Aei⋆S)(ei) = 0 for all S ∈ Sym M, (2.27) y − − ei (Aei⋆(S⋆ψ )) = 0 for all S ∈ Sym M. (2.28)

Now, we have all the tools to prove the proposition below.

(1,1) + − Proposition 2.4. Let w, h, and S be sections of ∧0 M, Sym0 M, and Sym M, + respectively. A section σ is deﬁned by σ = S⋆ψ . Then we have

¯ + (Aei⋆∇ei w)⋆ψ = 2δw ∧ ω +4 ∗ dw. (2.29) ¯ y − Aei⋆∇ei h = −δh ψ . (2.30) g − Ae ⋆∇¯ e S = ∗dσ − δS y ψ . (2.31) gi i g Proof. Before checking (2.29), we ﬁrst calculate

(2.24) ¯ ∗dw = ∗(ei ∧∇ei w) = ∗(ei ∧ ∇ei w) (2.23),(2.25) ¯ = Jei ∧ ∇ei w − δw ∧ ω. (2.32) RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 9

Let us prove the equation (2.29).

¯ + ¯ + ¯ + (Aei⋆∇ei w)⋆ψ = −(Aei ◦ ∇ei w)⋆ψ − (∇ei w ◦ Aei )⋆ψ ¯ + ¯ + = −(Aei⋆∇ei w)⋆ψ − 2(∇ei w ◦ Aei )⋆ψ g (2.12),(2.17) y + + ¯ y + = (Jδw ψ )⋆ψ +2Aei ∇ei w(ek) ∧ ek ψ + ¯ y + = −Aδw⋆ψ +2Aei⋆(∇ei w(ek) ∧ ek ψ ) ¯ y + −2∇ei w(ek) ∧ Aei⋆(ek ψ ) (2.17),(2.18) ¯ + = 2δw ∧ ω +2Aei⋆((∇ei w)⋆ψ ) ¯ y + y + −2∇ei w(ek) ∧ (−Aei ek ψ + ek Aei⋆ψ ) (2.18) = 2δw ∧ ω − 2∇¯ w(e ) ∧ JA + 4(∇¯ w) (e ∧ ω) ei k Aei ek ei ⋆ i (2.16) ¯ = 2δw ∧ ω − 2∇ei w(ek) ∧ (ei ∧ Jek + Jei ∧ ek) ¯ ¯ +4(∇ei w)ei ∧ ω +4ei ∧ (∇ei w)⋆ω (2.21),(2.25) ¯ = 4Jei ∧ ∇ei w − 2δw ∧ ω (2.32) = 2δw ∧ ω +4 ∗ dw.

Next, for any vector ﬁeld X, we get ¯ ¯ ¯ (Aei⋆∇ei h)(X) = −Aei ∇ei h(X) − ∇ei h(Aei X) ¯ y y − ¯ y y − = −(∇ei h)(X) ei ψ − (∇ei h)(X ei ψ ) g ¯ y y − − ¯ = (∇ei h)(ej,X)ei ej ψ − ψ (ei,X,ej)(∇ei h)ej ¯ y y − = ((∇ei h)ej ∧ (ei ej ψ ))(X). Thus, we have ¯ ¯ y y − Aei⋆∇ei h =(∇ei h)ej ∧ (ei ej ψ ). We proceed further with the calculation. g ¯ y ¯ y − y ¯ y − Aei⋆∇ei h = −ei ((∇ei h)ej ∧ ej ψ )+(ei (∇ei h)ej)ej ψ (2.26) = −e y ∇¯ (h(e ) ∧ e y ψ−) − δh(e )e y ψ− g i ei j j j j y ¯ − y − = ei ∇ei (h⋆ψ ) − δh ψ (2.20) = −δh y ψ−. This proves (2.30). In the same way that we prove the identity (2.30), we know that

¯ ¯ y y − (2.27),(2.28) − y − Aei⋆∇ei S =(∇ei S)ej ∧ (ei ej ψ ) = −δ(S⋆ψ ) − δS ψ . (2.33)

g 10 SOMAOHNOANDTAKUMATOMIHISA

The ﬁrst term of the equation (2.33) is

− − (2.22) 2 + δ(S⋆ψ )= −∗ d ∗ (S⋆ψ ) = ∗d ∗ (S⋆ψ )= −∗ dσ. Thus, we obtain the equation (2.31).

3. Comparison of differential operators

Assume that (M 6,g,J) is a 6-dimensional strict nearly Kähler manifold with the normalized scalar curvature scal = 30. We defined the twisted Dirac operator DTM in Section 2. Then we have a natural 2nd order differential operator: the standard Laplace operator ∆ := ∇∗∇ + q(R), that is the sum of the rough Laplacian and the curvature endomorphism in (2.11). Similarly, we define the twisted Dirac operator for C ¯ the canonical Hermitian connection as DTM :=(ek · ⊗ IdTM ) ◦ ∇ek and the Hermitian Laplace operator as ∆¯ := ∇¯ ∗∇¯ + q(R¯). These Laplace operators are introduced in [14] and [18]. We shall study the relationship between the twisted Dirac operator and the Hermitian Laplace operator. We get the difference below between the twisted Dirac operator for the Levi-Civita connection and the one for the canonical Hermitian connection.

Theorem 3.1. On the sections of S1/2 ⊗ T M, the relation holds: 3 1 D = D − ψ− · ⊗ Id − e · ⊗A . (3.1) TM TM 4 2 i ei Proof. The theorem is a straightforward consequence of (2.2) and (2.3).

We next see the relationship between the Hermitian Laplace operator and the twisted Dirac operator for the canonical Hermitian connection. Lemma 3.2. The identity holds: 1 e · R¯ (X, e )ζ = − Ric(X) · ζ − X · ζ + JX · ω · ζ, (3.2) j S j 2 for any vector ﬁeld X and spinor ζ, where R¯S denotes the curvature tensor for the canonical Hermitian connection on the spinor bundle S1/2.

Proof. Using the Cliﬀord relation, we get

12ej · R¯S(X, ej)=3g(R¯(X, ej)ek, el)ejekel =3g(X, ei)R¯ijklejekel

= g(X, ei)(R¯ijklejekel + R¯ikljekelej + R¯iljkelejek)

= g(X, ei)R¯ijklejekel + g(X, ei)R¯iklj(−2δljek +2δkjel + ejekel)

+ g(X, ei)R¯iljk(−2δljek +2δlkej + ejekel)

= g(X, ei)(R¯ijkl + R¯iklj + R¯iljk)ejekel − 6Ric(X, ek)ek. RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 11

From the equation (2.10), it is straightforward to show that

g(X, ei)(R¯ijkl + R¯iklj + R¯iljk)ejekel

= −2g(Jej, ek)g(Jel,X)ejekel − 2g(Jek, el)g(Jej,X)ejekel

− 2g(Jel, ej)g(Jek,X)ejekel = −12X + 12JX · ω. Combining the above equations, we arrive at the required identity (3.2).

The next equation follows immediately from (3.2).

ei · ej · R¯S(ei, ej)=18+2ω · ω · . (3.3)

From this equation, for the curvature endomorphism qS(R¯) on the spinor bundle S1/2, it follows that 9 1 q (R¯)= + ω · ω · . S 2 2 Theorem 3.3. Between the square of the twisted Dirac operator and the Hermitian Laplace operator, we obtain the relation:

2 1 1 − D = ∆¯ ⊗ + + ω · ω · ⊗ Id+(e y ψ · ⊗ Id)∇¯ . (3.4) TM S T 2 2 j ej Proof. Calculating the same as the Lichnerowicz formula (cf. [10, p.107]), we get

2 1 D = ∇¯ ∗∇¯ +(e y ψ− · ⊗ Id)∇¯ + e e {R¯ (e , e ) ⊗ Id + Id ⊗R¯(e , e )}. TM j ej 2 j k S j k j k Applying the equation (3.3), we ﬁnd that the above equation becomes

2 1 D = ∇¯ ∗∇¯ +9+ ω · ω · ⊗ Id+(e y ψ− ⊗ Id)∇¯ + e e ⊗ R¯(e , e ). (3.5) TM j ej 2 j k j k

Next, we compute the curvature endomorphism qS⊗T (R¯), which is the curvature term of the Hermitian Laplace operator ∆¯ S⊗T . 1 1 q ⊗ (R¯)= R¯(e ∧ e ) ⊗ (e ∧ e ) + (e ∧ e ) ⊗ R¯(e ∧ e ) S T 2 i j ⋆ i j ⋆ 2 i j ⋆ i j ⋆ + qS(R¯) ⊗ Id + Id ⊗qT (R¯) 9 1 1 = + ω · ω ⊗ Id + e e ⊗ R¯(e , e ) + Id ⊗Ric. 2 2 2 j k j k ¯ 9 1 ¯ Here, we use the equations qS(R) = 2 + 2 ω · ω· and qT (R) = Ric = 4 Id. Substituting the above equation into the equation (3.5), we obtain the relation (3.4).

The following lemma is obtained directly from (2.2), (2.3), and the deﬁnition of the rough Laplacian ∇∗∇ = −∇ ∇ + ∇ . ei ei ∇ei ei

Lemma 3.4. On the sections of S1/2 ⊗ T M, we have 5 1 1 ∇¯ ∗∇¯ = ∇∗∇ + + ω · ω · ⊗ Id +(Id ⊗A )∇ + (e y ψ− · ⊗ Id)∇¯ . (3.6) 8 8 ei ei 2 i ei 12 SOMAOHNOANDTAKUMATOMIHISA

Finally, we see the diﬀerence in the square of the twisted Dirac operator for each connection.

Theorem 3.5. On the sections of S1/2⊗T M, we obtain the relation between the squares of the two twisted Dirac operators DTM and DTM :

2 2 17 9 3 y − ¯ DTM =DTM + + ω · ω · ⊗ Id + (ej ψ · ⊗ Id)∇ej 8 8 2 (3.7) 1 + (Id ⊗A )∇ + e · e · ⊗(R¯(e , e ) − R(e , e )). ej ej 2 j k j k j k

(i) In particular, for a local section α ⊗ ei of S1/2 ⊗ T M, we have

2 7 9 1 D (α(i) ⊗ e )= D 2(α(i) ⊗ e ) − α(i) ⊗ e + ω · ω · α(i) ⊗ e + ω · α(i) ⊗ Je TM i TM i 8 i 8 i 2 i 3 + (e y ψ− · ⊗ Id)∇¯ (α(i) ⊗ e ) + (Id ⊗A )∇¯ (α(i) ⊗ e ) 2 j ej i ej ej i 1 + (e y ψ− · ⊗A )(α(i) ⊗ e ) 4 j ej i 1 3 − e · e · α(i) ⊗ e − Je · Je · α(i) ⊗ e . 4 j i j 4 i j j

Proof. Due to [10, p.107], the Lichnerowicz formula for the twisted Dirac operator is

15 1 D 2 = ∇∗∇ + + e e ⊗ R(e , e ). TM 2 2 j k j k

Employing the equations (3.4) and (3.6), we obtain

2 1 D = ∇¯ ∗∇¯ +9+ ω · ω · ⊗ Id+(e y ψ− · ⊗ Id)∇¯ + e e ⊗ R¯(e , e ) TM j ej 2 j k j k 77 9 3 = ∇∗∇ + + ω · ω · ⊗ Id + (e y ψ− ⊗ Id)∇¯ 8 8 2 j ej 1 + e e ⊗ R¯(e , e ) + (Id ⊗A )∇ 2 j k j k ej ej 17 9 3 = D 2 + + ω · ω ⊗ Id + (e y ψ− ⊗ Id)∇¯ TM 8 8 2 j ej 1 + (Id ⊗A )∇ + e e ⊗ (R¯(e , e ) − R(e , e )). ej ej 2 j k j k j k RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 13

(i) We calculate the curvature term in the above equation by using (2.9). For α ⊗ ei in Γ(S1/2 ⊗ T M), we have 1 e e α(i) ⊗ (R¯(e , e )e − R(e , e )e ) 2 j k j k i j k i 1 = e e α(i) ⊗ δ e − δ e − 3g(Je , e )Je +3g(Je , e )Je +2g(Je , e )e 8 j k ji k ki j j i k k i j j k i 1 1 3 = e · e · α(i)⊗ e − e · e · α(i) ⊗ e + Je · e · α(i) ⊗ Je 8 i j j 8 j i j 8 i j j 3 1 − e · Je · α(i) ⊗ Je + e · Je · α(i) ⊗ Je 8 j i j 4 j j i 1 3 1 = − e · e · α(i) ⊗ e − Je · Je · α(i) ⊗ e − α(i) ⊗ e + ω · α(i) ⊗ Je . 4 j i j 4 i j j i 2 i This gives the theorem.

4. Rarita-Schwinger fields on nearly Kahler¨ manifolds

The goal of this section is to ﬁnd the conditions under which Rarita-Schwinger ﬁelds exist, and this is the main result in this paper.

We recall that the deﬁnition of a section φ ∈ Γ(S1/2⊗T M) being the Rarita-Schwinger ﬁeld is φ ∈ Γ(S3/2) and DTM φ = 0. These are, of course, equivalent to φ ∈ Γ(S3/2) and 2 DTM φ = 0. (i) Any elements of Γ(S1/2 ⊗ T M) are represented locally as α ⊗ ei using a local or- (i) thonormal frame {ei}. By the isomorphism (2.4), let α ∈ Γ(S1/2) be decomposed into (i) (i) (i) (i) 0 1 6 α = α0 + α1 + α6 vol · κ ∈ Ω M ⊕ Ω M ⊕ Ω M · κ. (4.1) We use this decomposition liberally in the calculations of this section. (i) First, we rewrite the condition that α ⊗ ei ∈ Γ(S1/2 ⊗ T M) is an element of Γ(S3/2).

(i) (i) Lemma 4.1. Let α ⊗ ei be in S1/2 ⊗ T M. Then, α ⊗ ei is in S3/2 if and only if (i) α1 ⊙ ei ∈ Sym0 M, (4.2) (i) (1,1) (2,0)+(0,2) α1 ∧ ei ∈ ∧0 M ⊕ ∧ M, (4.3)  (i) (i) (i) α0 ei − α6 Jei + Aei α1 =0, (4.4) M M where Sym0 denotes the trace-free part of Sym . (i) (i) Proof. By deﬁnition, for α ⊗ ei in S3/2, we know ei · α = 0. Using the equations (2.5) and (2.8), we get (i) (i) (i) (i) (i) (i) ei · α = −g(ei,α1 )κ + ω(ei,α1 ) vol ·κ + α0 ei − α6 Jei + Aei α1 · κ. (i) (i) (i) Using the isomorphism (2.4) again, we get g(ei,α1 ) = 0, ω(ei,α1 ) = 0, andα0 ei − (i) (i) (i) (i) α6 Jei + Aei α1 = 0. Here, the equations g(ei,α1 ) = 0 and ω(ei,α1 ) = 0 are 14 SOMAOHNOANDTAKUMATOMIHISA equivalent to (4.2) and (4.3), respectively. The converse is obvious from the above discussion.

We already know the decomposition of the spinor bundle S1/2 (2.4). Also, we have an SU(3) irreducible decomposition of S1/2 ⊗ T M: ∼ 1 R (2,0)+(0,2) (1,1) + R − S1/2 ⊗T M = 2∧ M ⊕ ω ⊕∧ M ⊕∧0 M ⊕(Sym0 M ⊕ g)⊕Sym M, (4.5) where Rg is the trace part of Sym+ M. Keep in mind that the two ∧1M above are derived from ∧0M ⊗ ∧1M and ∧6M ⊗ ∧1M, respectively. (i) Next, we will see what conditions we get when α ⊗ ei is both in Γ(S3/2) and in the kernel of the twisted Dirac operator.

(i) Lemma 4.2. We denote a Rarita-Schwinger ﬁeld φ locally as α ⊗ ei, then we obtain (i) (i) α0 ei = 0 and α6 ei = 0.

Proof. We take a local orthonormal frame {ei} which is ∇¯ -parallel at a point. By the (i) formula (3.1), for a Rarita-Schwinger ﬁeld α ⊗ ei, the following holds: 3 1 D α(i) ⊗ e + ψ− · α(i) ⊗ e + e · α(i) ⊗ A e =0. (4.6) TM i 4 i 2 j ej i Using the decomposition (4.1) and the formulas (2.5)-(2.8), we compute each term in the left-hand side of the equation (4.6). For the ﬁrst term of the equation (4.6), we have (i) ¯ (i) DTM α0 κ ⊗ ei = (ej · ⊗ Id)∇ej (κ ⊗ α0 ei), (i) ¯ (i) DTM α6 vol ·κ ⊗ ei = −(Jej · ⊗ Id)∇ej (κ ⊗ α6 ei), (i) ¯ (i) ¯ (i) DTM α1 · κ ⊗ ei = −∇ej (g(ej,α1 )κ ⊗ ei) − ∇ej (g(ej,Jα1 ) vol ·κ ⊗ ei) ¯ (i) ¯ (i) +∇ej (Aej α1 · κ ⊗ ei)+ g(∇ej ej,α1 )κ ⊗ ei (i) (i) +g(∇¯ e ,Jα ) vol ·κ ⊗ e − A ¯ α · κ ⊗ e . ej j 1 i ∇ej ej 1 i Similarly, for the second term of the equation (4.6), we have

− (i) (i) ψ · α0 κ ⊗ ei = −4κ ⊗ α0 ei, − (i) (i) ψ · α6 vol ·κ ⊗ ei = 4 vol ·κ ⊗ α6 ei, − (i) ψ · α1 · κ ⊗ ei = 0. Finally, for the third term of the equation (4.6), we have

(i) (i) ej · α0 κ ⊗ Aej ei = ej · κ ⊗ α0 Aej ei, (i) (i) ej · α6 vol ·κ ⊗ Aej ei = −Jej · κ ⊗ α6 Aej ei, (i) (i) (i) ej · α1 · κ ⊗ Aej ei = κ ⊗ Aei α1 − vol ·κ ⊗ JAei α1 (i) (i) −ei · κ ⊗ α1 − Jei · κ ⊗ Jα1 . RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 15

Summarizing the above calculations, we ﬁnd that the equation (4.6) becomes 1 0= κ ⊗ −3α (i)e − ∇¯ (g(e ,α (i))e )+ A α (i) 0 i ej j 1 i 2 ei 1 1 + vol ·κ ⊗ 3α (i)e − ∇¯ (g(e ,Jα (i))e ) − JA α (i) 6 i ej j 1 i 2 ei 1 1 + ∇¯ (e · κ ⊗ α (i)e )+ e · κ ⊗ α (i)A e − ∇¯ (Je · κ ⊗ α (i)e ) (4.7) ej j 0 i 2 j 0 ej i ej j 6 i 1 − Je · κ ⊗ α (i)A e +(A ⊗ Id)∇¯ (α (i) · κ ⊗ e ) 2 j 6 ej i ej ej 1 i 1 1 − e · κ ⊗ α (i) − Je · κ ⊗ Jα (i). 2 i 1 2 i 1 Our main idea is to project the equation (4.7) onto each space of the irreducible decomposition (4.5). First, projecting the equation (4.7) onto the bundle ∧0M ⊗∧1M ∼= ∧1M, we get 1 0 = −3α (i)e − ∇¯ (g(e ,α (i))e )+ A α (i) 0 i ej j 1 i 2 ei 1 (4.4) 7 1 = − α (i)e + α (i)Je − ∇¯ (g(e ,α (i))e ). (4.8) 2 0 i 2 6 i ej j 1 i In the same way, projecting onto ∧6M ⊗ ∧1M ∼= ∧1M, we get

(4.4) 7 1 0 = α (i)e + α (i)Je − ∇¯ (g(e ,Jα (i))e ). (4.9) 2 6 i 2 0 i ej j 1 i As a reminder, using the symbols that will be deﬁned later in (4.12) and (4.13), the ¯ (i) term ∇ej (g(ej,α1 )ei) in the equation (4.8) is expressed as 1 1 ∇¯ (g(e ,α (i))e )= − δ(α (i) ∧ e ) − δ(α (i) ⊙ e ). ej j 1 i 2 1 i 2 1 i ¯ (i) Analogously, the term ∇ej (g(ej,Jα1 )ei) in the equation (4.9) is expressed as 1 1 ∇¯ (g(e ,Jα (i))e )= Jδ(Jα (i) ∧ Je )+ Jδ(Jα (i) ⊙ Je ). ej j 1 i 2 1 i 2 1 i We leave the other projection in (4.7) to the proof of Theorem A. 2 (i) Now let’s consider another equation DTM α ⊗ ei = 0. According to (3.4) and (3.7), the equation D 2 α(i) ⊗ e = 0 is equivalent to TM i 11 5 1 ∆¯ α(i) ⊗ e = − α(i) ⊗ e + ω · ω · α(i) ⊗ e + ω · α(i) ⊗ Je i 8 i 8 i 2 i 1 3 + e y ψ− · α(i) ⊗ A e − Je · Je · α(i) ⊗ e (4.10) 4 j ej i 4 i j j 1 + (e y ψ− · ⊗ Id)∇¯ (α(i) ⊗ e ) + (Id ⊗A )∇¯ (α(i) ⊗ e ). 2 j ej i ej ej i 16 SOMAOHNOANDTAKUMATOMIHISA

Using the decomposition (4.1) and the formulas (2.5)-(2.8) again, we compute each term in the right-hand side of the equation (4.10).

11 11 11 − α(i) ⊗ e = − κ ⊗ α (i)e − vol ·κ ⊗ α (i)e 8 i 8 0 i 8 6 i 11 − α (i) · κ ⊗ e , 8 1 i 5 45 45 ω · ω · α(i) ⊗ e = − κ ⊗ α (i)e − vol ·κ ⊗ α (i)e 8 i 8 0 i 8 6 i 5 − α (i) · κ ⊗ e , 8 1 i 1 3 3 ω · α(i) ⊗ Je = − κ ⊗ α (i)Je + vol ·κ ⊗ α (i)Je 2 i 2 6 i 2 0 i 1 − Jα (i) · κ ⊗ Je , 2 1 i 1 1 1 e y ψ− · α(i) ⊗ A e = κ ⊗ A α (i) + vol ·κ ⊗ JA α (i) 4 j ej i 2 ei 1 2 ei 1 1 1 + α (i)e · κ ⊗ A e + α (i)Je · κ ⊗ A e , 2 0 j ej i 2 6 j ej i 3 3 − Je · Je · α(i) ⊗ e = κ ⊗ (α (i)e + α (i)Je + A α (i)) 4 i j j 4 0 i 6 i ei 1 3 + vol ·κ ⊗ (−α (i)Je + α (i)e + JA α (i)) 4 0 i 6 i ei 1 +{∧1M ⊗ ∧1M part}, 1 (e y ψ− · ⊗ Id)∇¯ (α(i) ⊗ e ) = −κ ⊗ ∇¯ (g(e ,α (i))e ) 2 j ej i ej j 1 i ¯ (i) +g(∇ej ej,α1 )κ ⊗ ei ¯ (i) − vol ·κ ⊗ ∇ej (g(Jej,α1 )ei) ¯ (i) +g(J∇ej ej,α1 ) vol ·κ ⊗ ej +{∧1M ⊗ ∧1M part}, ¯ (i) ¯ (i) ¯ (i) (Id ⊗Aej )∇ej (α ⊗ ei) = κ ⊗ Aej ∇ej (α0 ei) + vol ·κ ⊗ Aej ∇ej (α6 ei) ¯ (i) +(Id ⊗Aej )∇ej (α1 · κ ⊗ ei).

We project the equation (4.10) onto ∧0M ⊗ ∧1M and ∧6M ⊗ ∧1M again. Considering the projection onto ∧0M ⊗ ∧1M ∼= ∧1M, we get 25 3 5 ∆¯ α (i)e = − α (i)e − α (i)Je + A α (i) 0 i 4 0 i 4 6 i 4 ei 1 ¯ (i) ¯ (i) −∇ej g(ej,α1 )ei + Aej ∇ej α0 ei (4.8) (i) ¯ (i) = −4α0 ei + Aej ∇ej α0 ei . RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 17

¯ ¯ From the fact that ∆ − ∆ = Aej ∇ej − 2 Id over the tangent bundle T M, the above equation becomes (i) (i) ∆ α0 ei = −2α0 ei. Since the Laplace operator on a tangent bundle is a non-negative operator, the equation (i) 6 1 ∼ 1 α0 ei = 0 follows from this equation. Similarly for ∧ M ⊗ ∧ M = ∧ M, we know (i) α6 ei = 0.

(i) (i) Applying the results of Lemma 4.2, α0 ei = 0 and α6 ei = 0, to the equation (4.7), we have ¯ (i) ¯ (i) 0= −κ ⊗ ∇ej (g(ej,α1 )ei) − vol ·κ ⊗ ∇ej (g(ej,Jα1 )ei) 1 1 (4.11) − e · κ ⊗ α (i) − Je · κ ⊗ Jα (i) +(A ⊗ Id)∇¯ (α (i) · κ ⊗ e ). 2 i 1 2 i 1 ej ej 1 i As already mentioned in the proof of Lemma 4.2, we want to project the equation (4.6) (brieﬂy the equation (4.11)) onto bundles other than ∧1M of the decomposition (4.5). For this purpose, we introduce several symbols:

(i) 1 (i) 1 (i) (i) w := pr 2 α ⊗ e = α ∧ e = α ⊗ e − e ⊗ α , (4.12) ∧ M 1 i 2 1 i 2 1 i i 1 1 1 H := pr α (i) ⊗ e = α (i) ⊙ e = α (i) ⊗ e + e ⊗ α (i) , (4.13) Sym M 1 i 2 1 i 2 1 i i 1 1 (i) (i) h := pr + H = α ⊙ e + Jα ⊙ Je , (4.14) Sym M 4 1 i 1 i 1 (i) (i) S := pr − H = α ⊙ e − Jα ⊙ Je , (4.15) Sym M 4 1 i 1 i ϕ(·, ·) := g(Jh·, ·) ∈ Ω(1,1)M, (4.16) + (2,1)+(1,2) σ := S⋆ψ ∈ Ω0 M, (4.17) 2 R (2,0)+(0,2) (1,1) where ∧ M denotes ω ⊕ ∧ M ⊕ ∧0 M. Theorem A. Let (M,g,J) be a 6-dimensional compact strict nearly K¨ahler manifold, then the space of the Rarita-Schwinger ﬁelds is isomorphic to the space of the harmonic 3-forms. In particular, we get

dim ker Q = b3(M), where b3(M) denotes the 3rd Betti number.

(i) Proof. Let α ⊗ei be a local expression of a Rarita-Schwinger ﬁeld. Employing Lemma (i) (i) ∞ 4.2 and the equation (4.4), we get Aei α1 = 0. This implies α1 ∧ ei ∈ C (M)ω ⊕ (1,1) (1,1) Ω0 M. Combining this with (4.3), we obtain w ∈ Ω0 M. Furthermore, we know + (1,1) h ∈ Γ(Sym0 M), and of course ϕ ∈ Ω0 M. ¯ (i) Next, the equation (4.8) becomes ∇ej (g(ej,α1 )ei) = 0 from Lemma 4.2, and this is equal to δw + δh + δS = 0. In the same way, δw + δh − δS = 0 follows immediately 18 SOMAOHNOANDTAKUMATOMIHISA from the equation (4.9). Putting these two equations together, we obtain δw + δh =0, δS =0. (4.18)

Now let us project the equation (4.11) onto the bundle ∧1M ⊗ ∧1M. Then we get 1 1 (A ⊗ Id)∇¯ (α (i) ⊗ e )= e ⊗ α (i) + Je ⊗ Jα (i) = −w + h. ej ej 1 i 2 i 1 2 i 1

Furthermore, using the two actions Aej ⋆ and Aej ⋆ on 2-tensors introduced in Subsection 2.5, we ﬁnd that the above equation becomes 1 1 g A ∇¯ (α (i) ⊗ e )+ A ∇¯ (α (i) ⊗ e )= −w + h. (4.19) 2 ej ⋆ ej 1 i 2 ej ⋆ ej 1 i

(i) Note that α1 ⊗ ei is denoted as w + hg+ S. Applying the results of Lemma 2.2 and Proposition 2.4 to the equation (4.19), we get y − y − ¯ −2w +2h = δw ψ +(JδS ψ + δσ) ◦ J + Aej ⋆∇ej h (4.20) y − y − ¯ − δh ψ + ∗dσ − δS ψ + Aej ⋆∇ej w. Therefore, the equations (4.18) and (4.20) are equivalent to the system g ∗dσ = −2w,  δσ ◦ J =2h,  ¯ ¯  Aej ⋆∇ej ⋆h + Aej ⋆∇ej w =0, (4.21)  − −  (δw − δh − δS) y ψ +(JδS y ψ ) ◦ J =0, δw + δh =0,g δS =0.    By combining with J, the second equation of (4.21) becomes − δσ =2ϕ. (4.22) Let the third equation of (4.21) act on ψ+. Employing Lemma 2.2 and Proposition 2.4 again, we ﬁnd 2(δh + δw) ∧ ω − 4dϕ +4 ∗ dw =0. (4.23) For the fourth equation of (4.21), taking note of (JδS y ψ−) ◦ J = −δS y ψ−, we get δw − δh − 2δS =0. (4.24) Putting this equation together with the ﬁfth equation of (4.21), we get δw =0, δh =0, and δS =0. (4.25) Based on these calculations (4.22)-(4.25), the system (4.21) is equivalent to ∗dσ = −2w, δσ = −2ϕ,  (4.26)  ∗dw = dϕ,  δw =0, δϕ =0.   RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 19

Here, we use the relation δh = −Jδϕ. It is a little diﬃcult to derive the equation δS = 0 when we show (4.21) from (4.26). We solve this problem by checking in which component of the decomposition (4.5) each term of the formula (2.31) is contained. Finally, combining the equations in this system (4.26), we obtain the system equivalent to (4.26):

∆σ =0, ϕ =0, w =0. (4.27)

Here, we note that M is compact. Conversely, if there exists a harmonic primitive (2, 1)+(1, 2)-form σ, then we have a (i) Rarita-Schwinger ﬁeld α ⊗ ei immediately. By Verbitsky’s theorem (cf.[21]), arbitrary harmonic 3-form is primitive and of type (2, 1)+(1, 2), so Theorem A is proved.

5. Examples and Applications

5.1. Examples. As shown in Theorem A, we revealed that the space of the Rarita- Schwinger fields isomorphic to the space of the harmonic 3-form on 6-dimensional compact strict nearly Kähler manifolds. It is an important problem whether there actually exist manifolds with non-zero Rarita-Schwinger fields. We find out how many Rarita- Schwinger fields exist for a specific manifold. There are only six known examples of complete simply connected strict nearly Kähler 6 3 3 6-manifolds. Four of them are homogeneous: S = G2 / SU(3), S × S = (SU(2) × SU(2) × SU(2))/∆ SU(2), CP3 = Sp(2)/(U(1) × Sp(1)), F(1, 2) = SU(3)/T 2. They were first constructed by J. A. Wolf and A. Gray [24]. Moreover, Butruille [2] showed that there is no other homogeneous example in dimension 6. The other two are inhomogeneous examples, S6 and S3 × S3, found by Foscolo and Haskins [4]. We actually calculate the 3rd betti numbers for these six nearly Kähler manifolds, 3 3 we get b3(S × S ) = 2 (both for homogeneous and inhomogeneous) and the other cases vanish. Therefore we know that Rarita-Schwinger fields exist in two S3 × S3 and not in the other. We remark that the manifold S3 × S3 with the standard metric, which is not a nearly Kähler manifold, does not have Rarita-Schwinger fields. This is easily indicated by the following.

Let {e1, e2, e3} be a local orthonormal tangent frame of the ﬁrst factor of the Rie- 3 3 mannian product S × S , and {e4, e5, e6} be a local orthonormal tangent frame of the 3 3 (i) second factor. Any sections of S1/2 ⊗ T (S × S ) are written locally as i α ⊗ ei. P 20 SOMAOHNOANDTAKUMATOMIHISA

Using the Lichnerowicz formula for the twisted Dirac operator, we get scal D 2(α(i) ⊗ e )= ∇∗∇(α(i) ⊗ e )+ α(i) ⊗ e TM i i 4 i i i i X X X 1 + e e α(i) ⊗ R(e , e )e 2 j k j k i Xi,j,k ∗ (i) (i) = ∇ ∇(α ⊗ ei)+4 α ⊗ ei i i X X (i) (i) + ejeiα ⊗ ej + ejeiα ⊗ ej. 1≤i,j≤3 4≤i,j≤6 X X 2 (i) By taking the L -inner product with i α ⊗ ei, the above equation becomes 2 P 2 2 (i) (i) (i) DTM α ⊗ ei = ∇ α ⊗ ei +4 α ⊗ ei ! ! i i i X X 2 X 2 (5.1) (i) (i) − ei · α − ei · α .

1≤i≤3 4≤i≤6 X X

Evaluating the third term of the equation (5.1), we have 2 2 2 (i) (i) (i) ei · α ≤ ei · α = α 1≤i≤3 1≤i≤3 ! 1≤i≤3 ! X X X 2 = α (i) +2 α(1) α(2) +2 α (2) α(3) +2 α(3) α(1)

1≤i≤3 X 2 ≤ 3 α( i) . 1≤i≤3 X

We evaluate the fourth term in the same way. Putting these calculations together, we get the inequality

2 2 2 (i) (i) (i) DTM α ⊗ ei ≥ ∇ α ⊗ ei + α ⊗ ei . ! ! i i i X X X

From this equation, we deduce ker DTM = {0}. In other words, we know that there is no Rarita-Schwinger field on S3 × S3 with the standard metric. This is the first example in which the existence of non-trivial Rarita-Schwinger fields depends not only on topological conditions but also on metrics. In both of the nearly Kähler and the standard S3 × S3, there is no harmonic spinor field. However, only in the nearly Kähler S3 × S3, there are Rarita-Schwinger fields. RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 21

5.2. Linear stability. The following theorem about the linear stability of nearly Kähler manifolds is known by Semmelmann, Wang, and Wang [19]. Theorem 5.1 ([19]). Let (M,g,J) be a complete strict nearly Kähler 6-manifold. If b2(M) or b3(M) is nonzero, then g is linearly unstable with respect to the Einstein- Hilbert action restricted to the space of Riemannian metrics with constant scalar curvature and fixed volume.

Since dim ker Q = b3(M) is obtained in our main result (Theorem A), we know that g is linearly unstable in the above sense on six-dimensional complete strict nearly K¨ahler manifolds with non-zero Rarita-Schwinger ﬁelds.

5.3. Infinitesimal deformation of Killing spinors. For detailed information about Killing spinor variations, please refer to [22] and other sources. We define an infinitesimal deformation of Killing spinors for a general n-dimensional (β,κ0) ∗ Riemannian manifold (M,g). Before that, we provide a tensor Ψ in Γ(T M ⊗ S1/2) for β : T M → T M and κ0 in Γ(S1/2):

(β,κ0) Ψ (X)= β(X) · κ0. Deﬁnition 5.2. We call a pair (β, κ) an inﬁnitesimal deformation of the Killing spinor κ0 with constant c if β : T M → T M symmetric and κ in Γ(S1/2) satisfy (i) κ is a Killing spinor with constant c. (ii) tr β = δβ = 0. (β,κ0) (β,κ0) (iii) DTM Ψ = ncΨ .

We now consider a compact 6-dimensional strict nearly Kähler manifold (M 6,g,J) with scalar curvature scal = 30 except the round sphere S6. In this case, of course, 1 n = 6 and c = 2 in Definition 5.2. Furthermore, there is a one-to-one correspondence between a Killing spinor (modulo constant rescaling) and a nearly Kähler structure (cf. [8]). So, the Killing spinor κ0 is corresponding to the nearly Kähler structure (g,J). For (i) in the above definition, we take κ as a constant multiplication of κ0 because the 1 space of the Killing spinors with the Killing number 2 is one-dimensional. We adjust the symbols to the ones we used in Section 4. First, since β is an endomorphism of (i) T M, it is expressed locally as α1 ⊗ ei. Furthermore, since β is symmetric, it becomes H defined in (4.13). This means that w, defined in (4.12), is zero. Therefore, (ii) in the above definition is rewritten as

H ∈ Sym0 M, δH =0. The condition (iii) becomes (i) (i) DTM (α1 · κ0 ⊗ ei)=3α1 · κ0 ⊗ ei. (5.2) Projecting the left-hand side of the equation (5.2) onto the bundle ∧1M ⊗ ∧1M, we get ¯ (i) (Aej ⊗ Id)∇ej (α1 ⊗ ei) − h. 22 SOMAOHNOANDTAKUMATOMIHISA

Projecting the right-hand side of the equation (5.2), we get

3h +3S.

Putting the above equations together, we have

¯ (i) (Aej ⊗ Id)∇ej (α1 ⊗ ei)=4h +3S. Continuing with the calculation similar to the one in Section 4, we ﬁnd that 3 δσ = −8ϕ, dϕ = − σ, ϕ ∈ Ω(1,1)M, σ ∈ Ω(2,1)+(1,2)M. 2 0 0 This is equivalent to 2 ∆ϕ = 12ϕ, δϕ =0, σ = − dϕ, ϕ ∈ Ω(1,1)M, 3 0 so we obtain the following result. Note that the diﬀerential of a co-closed promitive (1, 1)-form is contained in the space of primitive (2, 1)+(1, 2)-forms. In the case of the round sphere S6, the same arguments hold and the dimension of the space of the Killing spinors is eight.

Theorem B. Let (M 6,g,J) be a compact strict nearly K¨ahler 6-manifold. Then the space of the inﬁnitesimal deformations of Killing spinors is isomorphic to the direct sum of the space of the primitive co-closed (1, 1)-eigenforms of the Laplace operator for the 1 eigenvalue 12 and the space of the Killing spinors with constant 2 .

We denote by E(λ) the λ-eigenspace of ∆ restricted to the space of primitive co- closed (1, 1)-forms. Besides the round sphere S6, Moroianu, Nagy, and Semmelmann [13, Theorem 4.1] showed that the space of the infinitesimal deformations of nearly Kähler structures is E(12), which is a part of the space of the essential infinitesimal Einstein deformations E(2) ⊕ E(6) ⊕ E(12) [15, Lemma 5.2]. As described in the paper [8], the space of the Killing spinors (modulo constant rescaling) is one-to-one correspondence with the space of the nearly Kähler structures except S6. So, the space of the infinitesimal deformations of Killing spinors (up to a constant) and the infinitesimal deformations of nearly Kähler structures are consistent. Thus, Theorem B is a reproof of the result of [13] through the Killing spinor.

Acknowledgement

We wish to oﬀer our immeasurable gratitude to Professor Yasushi Homma for useful discussions, advices, and providing us with the topic. This paper is a part of the outcome of research performed under a Waseda University Grant for Special Research Projects (Project number: 2020C-614). RARITA-SCHWINGER FIELDS ON NEARLY KAHLERMANIFOLDS¨ 23

References [1] C. Bär, R. Mazzeo. Manifolds with Many Rarita-Schwinger Fields. Comm. Math. Phys. 384 (2021), no. 1, 533-548. [2] J.-B. Butruille. Classification des variétés approximativement kähleriennes homogènes. Ann. Global Anal. Geom. 27 (2005), no. 3, 201-225. [3] B. Charbonneau, D. Harland. Deformations of nearly Kähler instantons. Comm. Math. Phys. 348 (2016), no. 3, 959-990. 6 [4] L. Foscolo, M. Haskins. New G2-holonomy cones and exotic nearly Kähler structures on S and S3 × S3. Ann. of Math. (2) 185 (2017), no. 1, 59-130. [5] L. Foscolo. Deformation theory of nearly Kähler manifolds. J. Lond. Math. Soc. (2) 95 (2017), no.2, 586-612. [6] A. Gray. Nearly Kähler manifolds. J. Differential Geometry 4 (1970), 283–309. [7] A. Gray. The structure of nearly Kähler manifolds. Math. Ann. 223 (1976), no. 3, 233-248. [8] R. Grunewald. Six-dimensional Riemannian manifolds with a real Killing spinor. Ann. Global Anal. Geom. 8 (1990), no.1, 43-59. [9] N. Hitchin. Stable forms and special metrics. Global differential geometry: the mathematical legacy of Alfred Gray (Bilbao, 2000), 70–89, Contemp. Math., 288, Amer. Math. Soc., Providence, RI, 2001. [10] Y. Homma. Twisted Dirac operators and generalized gradients. Ann. Global Anal. Geom. 50 (2016), no. 2, 101-127. [11] Y. Homma, U. Semmelmann. The kernel of Rarita-Schwinger operator on Riemannian spin manifolds. Comm. Math. Phys. 370 (2019), no. 3, 853-871. [12] Y. Homma, T. Tomihisa. Spectra of the Rarita-Schwinger operator on some symmetric spaces.J. Lie Theory 31 (2021), no. 1, 249-264. [13] A. Moroianu, P.-A. Nagy, U. Semmelmann. Deformations of nearly Kähler structures. Pacific J. Math. 235 (2008), no. 1, 57-72. [14] A. Moroianu, U. Semmelmann. The Hermitian Laplace operator on nearly Kähler manifolds. Comm. Math. Phys. 294 (2010), no. 1, 251-272. [15] A. Moroianu, U. Semmelmann. Infinitesimal Einstein deformations of nearly Kähler metrics. Trans. Amer. Math. Soc. 363 (2011), no. 6, 3057-3069. [16] P.-A. Nagy. Nearly Kähler geometry and Riemannian foliations. Asian J. Math. 6 (2002), no. 3, 481-504. [17] W. Rarita, J. Schwinger. On a theory of particles with half-integral spin. Phys. Rev. (2) 60, (1941), 61. [18] U. Semmelmann, G. Weingart The standard Laplace operator. Manuscripta Math. 158 (2019), no. 1-2, 273-293. [19] U. Semmelmann, C. Wang, M. Y. Wang. On the linear stability of nearly Kähler 6-manifolds. Ann. Global Anal. Geom. 57 (2020), no. 1, 15-22. [20] C. van Coevering. Deformations of Killing spinors on Sasakian and 3-Sasakian manifolds. J. Math. Soc. Japan 69 (2017), no. 1, 53-91. [21] M. Verbitsky. Hodge theory on nearly Kähler manifolds. Geom. Topol. 15 (2011), no. 4, 2111-2133. [22] M. Y. Wang. Preserving parallel spinors under metric deformations. Indiana Univ. Math. J. 40 (1991), no. 3, 815-844. [23] E. Witten. Elliptic genera and quantum field theory. Comm. Math. Phys. 109 (1987), no. 4, 525- 536. [24] J. A. Wolf, A. Gray. Homogeneous spaces defined by Lie group automorphisms. II. J. Differential Geometry 2 (1968), 115-159. 24 SOMAOHNOANDTAKUMATOMIHISA

Soma Ohno, Department of Pure and applied Mathematics, Graduate school of fundamental science and engineering, Waseda University, 3-4-1 Ohkubo, Shinjuku-ku, Tokyo, 169-8555, Japan. Email address: [email protected]

Takuma Tomihisa, Department of Applied Mathematics, Faculty of science and engineering, Waseda University, 3-4-1 Ohkubo, Shinjuku-ku, Tokyo, 169-8555, Japan. Email address: [email protected]