On the integrability of generalized almost para-Norden and para-Hermitian structures (Q1679175)

From an adapted version of the crucial Newlander-Nirenberg theorem, it is well known that the definition of a kind of `complex-like structure' on a smooth manifold \(M\) of dimension \(n\) involves two steps; firstly, we have to specify an algebraic condition on the tangent bundle (obtaining an `almost complex-like structure'), as well as the eigenbundles have to be closed under a `Lie-like bracket' (the so-called integrability condition). Usually, in computations, it is easier to prove the integrability condition by the vanishing of the `associated Nijenhuis-like tensor'. An \textit{almost complex structure} consists of an endomorphism of vector bundles \(I: TM\to TM\) such that \(I^2=-\text{Id}\); it gives rise to a direct sum decomposition \(T_{\mathbb{C}}X=T^{1,0}_M\oplus T^{0,1}_M\). Then the almost complex structure \(I\) is \textit{integrable} if and only if the Lie bracket of vector fields preserves \(T^{0,1}_X\), i.e., \([T^{0,1}_M,T^{0,1}_M]\subset T^{0,1}_M\). In this case, we can rephrase the integrability condition by the vanishing of the associated Nijenhuis tensor, defined as \[ N_I(X,Y)=[X,Y]-[I(X),I(Y)]+I\left([I(X),Y]+[X,I(Y)]\right). \] Furthermore, let \(\mathbf{T}M=TM\oplus T^ *M\) be the generalized tangent bundle of \(M\) endowed with its structure of a Courant algebroid (the bracket is denoted \([-,-]_C\), and the pairing of signature \((n,n)\) is \(g_0\)). A \textit{generalized almost complex structure} on \(M\) is a bundle map \(I: \mathbf{T}M\to \mathbf{T}M\) satisfying \(I^{2}=-\text{Id}\). Now, a \textit{generalized complex structure} on \(\mathbf{T}M\) is a generalized almost complex structure \(I\) on \(\mathbf{T}M\), whose \(i\)-eigenbundle is closed with respect to the Courant bracket. Again, it is useful to rewrite the integrability condition in terms of the vanishing of the generalized Nijenhuis tensor: \[ N^{I}_{C}(\alpha,\beta)=[I(\alpha),I(\beta)]_{C}-[\alpha,\beta]_C-I\left([I(\alpha),\beta]_{C}+[\alpha,I(\beta)]_{C}\right), \] for all sections \(\alpha,\beta\) of \(\mathbf{T}M\). The article under review deals with integrability conditions of differentiable manifolds with an additional compatible metric structure. Firstly, a \textit{paracomplex vector space} is a pair \((V,F)\), where \(V\) is real finite-dimensional vector space, and \(F\) is an endomorphism \(F\in\text{End}(V)\) satisfying \(F^2=\text{Id}\), \(F\neq \text{Id}\), and the two eigenspaces \(V^{\pm}:=\ker(\text{Id}\pm F)\) have the same dimension. Consequently, an \textit{almost paracomplex structure} on a \(2n\)-dimensional smooth manifold \(M\) is defined as an endomorphism \(F\in\text{End}(TM)\) such that \((T_mM,F_m)\) is a paracomplex vector space for all \(m\in M\). A \textit{paracomplex structure} on \(M\) is an integrable almost paracomplex structure \(F\), that is, the Nijenhuis tensor \(N_F\) associated to \(F\) vanishes, where \[ N_F(X,Y):=[F(X),F(Y)]+[X,Y]-F\left([F(X),Y]+[X,F(Y)]\right), \] for every \(X,Y\in\Gamma(TM)\). Furthermore (Definition 2.1), an \textit{almost para-Norden manifold} (resp. an \textit{almost para-Hermitian manifold}) is a triple \((M,F,g)\), where \((M,F)\) is an almost paracomplex manifold, and \(g\) is a Riemannian metric satisfying \(g(F(X),Y)=g(X,F(Y))\) (resp. \(g(F(X),Y)=-g(X,F(Y))\)), for all \(X,Y\in\Gamma(TM)\). Notably (see Proposition 2.2), almost para-Norden/para-Hermitian manifolds \((M,F,g)\) are endowed with a natural canonical connection \(D\) on \(M\) such that \(DF=0\) and \(Dg=0\). Naturally, an almost para-Norden/para-Hermitian manifold \((M,g,F)\), with \(F\) an integrable paracomplex structure, is called a \textit{para-Norden/para-Hermitian manifold}. As a way to unify symplectic forms, paracomplex structures, and Poisson structures, in [C. R., Math., Acad. Sci. Paris 338, No. 11, 889--894 (2004; Zbl 1057.53019)], \textit{A. Wade} defined a \textit{generalized almost paracomplex structure} of a manifold \(M\) as an endomorphism \(I\in\text{End}(\mathbf{T}M)\), which satisfies \(I^2=\text{Id}\), \(I\neq \pm \text{Id}\). Now, if \(\mathcal{L}_+ \) and \(\mathcal{L}_{-}\) are the eigendistributions of \(I\), the integrability of the generalized almost paracomplex structure \(I\) is defined by the involutivity of \(\mathcal{L}_+,\mathcal{L}_{-}\); this condition is equivalent to the vanishing of \(N^I_C\) (as defined above). Now, a generalized almost paracomplex structure \(I\) is called a \textit{generalized almost para-Norden structure} (resp. a \textit{generalized almost para-Hermitian structure}) if \(g_0(I(\sigma),\tau)=g_0(\sigma,I(\tau))\) (resp. \(g_0(I(\sigma),\tau)=-g_0(\sigma,I(\tau))\)), for all \(\sigma,\tau\in\Gamma(\mathbf{T}M)\). In [\textit{A. Nannicini}, J. Geom. Phys. 56, No. 6, 903--916 (2006; Zbl 1102.53018); J. Geom. Phys. 99, 244--255 (2016; Zbl 1341.53055)], using a linear connection \(\nabla\) on \(M\), Nannicini introduced a new bracket \([-,-]_{\nabla}\) on \(\mathbf{T}M\), which is skew-symmetric, and the Leibniz and Jacobi identities hold. In this way, we obtain an alternative and interesting notion of integrability (Definition 2.2): the generalized almost paracomplex structure \(I\) is called \textit{\(\nabla\)-integrable} if and only if its eigendistributions are involutive with respect to \([-,-]_\nabla\). Again, it is convenient to introduce a new Nijenhuis tensor whose vanishing is equivalent to the \(\nabla\)-integrability of I (see Corollary 2.1): for all \(\alpha,\beta\in\Gamma(\mathbf{T}M)\), we define \[ N^{\nabla}_I(\alpha,\beta):=[I(\alpha),I(\beta)]_\nabla+[\alpha,\beta]_{\nabla}-I\left([I(\alpha),\beta]_\nabla+[\alpha,I(\beta)]_{\nabla}\right). \] As it is written in the article, this work is mainly inspired by the study of generalized paracomplex structures in [Zbl 1057.53019] and the generalized geometry of Norden manifolds from [Zbl 1341.53055], and the integrability of generalized almost para-Norden/para-Hermitian structures. Using the description in the block matrix form (see Equation (3.3)), the authors are able to give necessary and sufficient conditions for the \(\nabla\)-integrability of these structures in Proposition 3.3, assuming that \(\nabla\) is a torsion-free connection on \(M\). Furthermore, given a Riemannian manifold \((M,g)\), one can define the \textit{canonical generalized almost para-Norden structure} \(I_g\). In application of Proposition 3.4, \(I_g\) is \(\nabla\)-integrable if and only if \(g\) is a Codazzi tensor with respect to a torsion-free connection \(\nabla\) on \((M,g)\). Now, the \textit{canonical generalized almost para-Hermitian structure} \(I_\omega\) is defined from an almost symplectic manifold \((M,\omega)\), and it is \(\nabla\)-integrable if and only if \(\omega\) is a Codazzi tensor with respect to \(\nabla\). Moreover, if \((M,F)\) is an almost paracomplex manifold, one can define the \textit{generalized almost para-Norden structure} induced by \(F\) (resp. the \textit{generalized almost para-Hermitian structure} induced by \(F\)) \(I_{1,F}\) (resp. \(I_{-1,F}\)). Then in Propositions 3.6 and 3.8 the authors study the \(\nabla\)-integrability and \(D\)-integrability of these structures, showing their different behavior. Finally, in Proposition 3.9, the \(D\)-integrability of \(I_{1,F}\circ I_g\) and \(I_{1,F}\circ I_g\) is discussed. Moreover, in Sections 3.3 and 3.4, the invariance under \(B\)-field transformations of the first type and second type of the canonical generalized almost para-Norden structure \(I_g\), the canonical generalized almost para-Hermitian structure \(I_{\omega}\), the induced generalized almost para-Norden \(I_{1,F}\), and the induced generalized para-Hermitian structure \(I_{-1,F}\) are studied. Finally, in Section 4, the authors face a natural question about the relation between the integrability in terms of the Courant bracket and the \(\nabla\)-integrability of a generalized almost para-Norden/para-Hermitian structure \(I\). In Proposition 4.1, an explicit expression of the difference between the Nijenhuis tensors \(N^C_I\), \(N^{\nabla}_I\) is given, which leads to Propositions 4.2 and 4.3 that provide conditions in some particular cases when \(N^C_I=N^{\nabla}_I\) holds.