TSUKUBA J MATH.

Vol. 28 No. 1 (2004), 193-213

ON ALMOST PARA-COSYMPLECTIC MANIFOLDS

By
Piotr DACKO

Abstract. An almost para-cosymplectic manifold is by definition

an odd-dimensional differentiable manifold endowed with an almost
paracontact stmcture with hyperbolic metric for which the structure
forms are closed. The local structure of an almost para-cosymplectic
manifold is described. We also treat some special subclasses of this
class of manifolds: para-cosymplectic, weakly para-cosymplectic and
almost para-cosymplectic with para-K\"ahlerian leaves. Necessary and
sufficient conditions for an almost para-cosymplectic manifold to be
para-cosymplectic are found. Necessary and sufficient conditions for
an almost para-cosymplectic manifold with para-K\"ahlerian leaves to
be weakly para-cosymplectic are also established. We construct
examples of weakly para-cosymplectic manifolds, which are not para-
cosymplectic. It is proved that in dimensions an almost para- $\geq 5$

cosymplectic manifold cannot be of non-zero constant sectional

curvature. Main curvature identities which are fulfilled by any almost
para-cosymplectic manifold are found.

1. Preliminaries
Letbe a $(2n+1)$ -dimensional differentiable manifold. Suppose that
$M$

is an almost paracontact hyperbolic metric structure on $M$ . This means

$(\varphi, \xi, \eta, g)$

that is a quadruple consisting of a $(1, 1)$ -tensor field , a vector field ,

$(\varphi, \xi, \eta, g)$
$\varphi$ $\xi$

a l-form and a pseudo-Riemannian metric on $M$ satisfying the following

$\eta$
$g$

relations
$\varphi^{2}X=X-\eta(X)\xi$ , $\eta(\xi)=1$ , $g(\varphi X, \varphi Y)=-g(X, Y)+\eta(X)\eta(Y)$ .

Key words and phrases. Almost para-cosymplectic manifold, para-cosymplectic manifold, weakly para-
cosymplectic manifold, manifold of constant curvature.
2000 Mathematics Subject Classification. $53C15,53C25$ .
Received December 9, 2002.
Revised October 6, 2003.
194 Piotr DACKO

In the above and in the sequel, $X,$ $Y,$ denote arbitrary smooth vector fields on $\ldots$

$M$ if it is not otherwise stated. As consequences of the above, we additionally

have
$\varphi\xi=0$ , $\eta(\varphi X)=0$ , $\eta(X)=g(X, \xi)$ , $g(\varphi X, Y)=-g(\varphi Y, X)$ .

Thus, $\Phi(X, Y)=g(\varphi X, Y)$ is a 2-form on $M$ , which will be said the fundamental
form of the structure.
With the above terminology we follow [8]. In the papers [14], [3], [4], [1] the
authors called such structures almost para-coHermitian.
The manifold $M$ endowed with the almost paracontact hyperbolic metric
structure will be called
(a) para-cosymplectic if the forms and are parallel with respect to the $\eta$
$\Phi$

Levi-Civita connection of the metric , that is, and $\nabla\Phi=0$ ;

$\nabla$
$g$ $\nabla\eta=0$

(b) almost para-cosymplectic if the forms and are closed, that is, $d\eta=0$ $\eta$
$\Phi$

and $d\Phi=0$ .
The above notions of (almost) para-cosymplectic manifolds are paracontact–
with a hyperbolic metric–analogue of (almost) cosymplectic manifolds (for
almost cosymplectic manifolds see [2], [9]).
Our definition of the para-cosymplecticity differs from that used in the paper
[8], in which this notion concems even-dimensional indefinite almost Hermitian or
almost para-Hermitian manifolds with coclosed fundamental forms.
For an almost para-cosymplectic manifold, define the $(1, 1)$ -tensor field by $A$

$ AX=-\nabla_{X}\xi$ .

PROPOSITION 1. For an almost para-cosymplectic manifold, we have

$\mathscr{L}_{\xi}\eta=0$ , $\mathscr{L}_{\xi}\Phi=0$ , $g(AX, Y)=g(AY, X)$ , $A\xi=0$ ,

$\eta oA=0$ , $(\mathscr{L}_{\xi}g)(X, Y)=-2g(AX, Y)$ , $\nabla_{\xi}\varphi=0$ ,

$A\varphi+\varphi A=0$ , $g(\varphi AX, Y)=g(\varphi AY, X)$ , $Tr(\varphi A)=Tr(A)=0$ ,

where $\mathscr{L}$

indicates the operator of the Lie $d\iota fferentiation$ .

PROOF. By $d\eta=0,$ $d\Phi=0,$ $i_{\xi}(\eta)=1$ and $i_{\xi}\Phi(X)=g(\varphi\xi, X)=0$ , we have

$\mathscr{L}_{\xi}\eta=di_{\xi}\eta+i_{\xi}d\eta=0$ , $\mathscr{L}_{\xi}\Phi=di_{\xi}\Phi+i_{\xi}d\Phi=0$ .

Moreover, by $(\nabla_{X}\eta)(Y)=g(\nabla_{X}\xi, Y)=-g(AX, Y)$ ,

$0=2d\eta(X, Y)=-g(AX, Y)+g(AY, X)$ ,

 On almost para-cosymplectic manifolds 195

that is, $A$

is a symmetric operator. Consequently,
$(\mathscr{L}_{\xi}g)(X, Y)=g(\nabla_{X}\xi, Y)+g(\nabla_{Y}\xi, X)=-2g(AX, Y)$ ,

$g(A\xi, Y)=g(AY, \xi)=-g(\nabla_{Y}\xi, \xi)=-\frac{1}{2}\nabla_{Y}g(\xi, \xi)=0$ ,

and the last line of equations implies $A\xi=-\nabla_{\xi}\xi=0$ and $\eta\circ A=0$ .
For , we have the decomposition
$\mathscr{L}_{\xi}$
, here indicates the unique $\mathscr{L}_{\xi}=\nabla_{\xi}+A$ $A$

extension of the $(1, 1)$ -tensor field to a derivation of the tensor algebra (see e.g. $A$

[10], p. 30).
From one hand, since , it holds $\mathscr{L}_{\xi}\Phi=0$

$ 0=(\mathscr{L}_{\xi}\Phi)(X, Y)=(\nabla_{\xi}\Phi)(X, Y)-\Phi(AX, Y)-\Phi$ $X,$ ( A ) $Y$

$=g((\nabla_{\xi}\varphi)X, Y)-g((A\varphi+\varphi A)X, Y)$ .

Thus, we have $\nabla_{\xi}\varphi=A\varphi+\varphi A$ .
On the other hand, since $A\xi=0$ , applying $\nabla_{\xi}$
to $\varphi^{2}X=X-\eta(X)\xi$ , we
obtain
$\varphi(\nabla_{\xi}\varphi)X+(\nabla_{\xi}\varphi)\varphi X=g(A\xi, X)\xi+\eta(X)A\xi=0$ .

We note that $ A=\varphi(\varphi A)=(A\varphi)\varphi$ , by $A\xi=0$ and $\eta\circ A=0$ . Hence

$0=\varphi(\varphi(\nabla_{\xi}\varphi)+(\nabla_{\xi}\varphi)\varphi)$

$=\varphi(\varphi(A\varphi+\varphi A)+(A\varphi+\varphi A)\varphi)=2\varphi(A+\varphi A\varphi)=2(\varphi A+A\varphi)$ .

Consequently . Since is skew-symmetric and $A$ symmetric, then
$\nabla_{\xi}\varphi=0$ is $\varphi$
$\varphi A$

traceless. Note that the trace of also vanishes, because $\varphi A+A\varphi=0$ $A=\varphi(\varphi A)$

implies the symmetry of . $\varphi A$

$\square $

2. The Local Structure

In this section, we establish a local equivalence between almost para-
cosymplectic structures and certain special families of almost para-K\"ahlerian
structures.
By an almost para-K\"ahlerian manifold it is meant a $2n$ -dimensional dif-
ferentiable manifold endowed with a pair , where is an almost para-
$\tilde{M}$ $(\tilde{J},\tilde{g})$
$\tilde{J}$

complex structure , is a pseudo-Riemannian metric such that $(\tilde{J}^{2}=\tilde{I})$ $\tilde{g}$

and the fundamental form is closed.

$\tilde{g}(\tilde{J}\tilde{X},\tilde{J}\tilde{Y})=-\tilde{g}(\tilde{X},\tilde{Y})$ $\tilde{\Omega}(\tilde{X},\tilde{Y})=\tilde{g}(\tilde{J}\tilde{X},\tilde{Y})$

An almost para-K\"ahlerian manifold with integrable almost para-complex

structure (equivalently, ) is said to be para-K\"ahlerian. For almost para-
$\tilde{J}$ $\tilde{\nabla}\tilde{J}=0$

K\"ahlerian structures, we refer the survey articles [5], [6].

196 Piotr DACKO

Let $t\in(a, b),$ $a<b$ , be a l-parameter family of almost para-

$(\tilde{J}_{t},\tilde{g}_{t}),$

Kahlerian structures on a -dimensional manifold

$2n$
such that for any $\tilde{M}$ $\tilde{\Omega}_{f}=\tilde{\Omega}$

being a fixed closed 2-form on . This family enables us to define an

$t\in(a, b),\tilde{\Omega}$ $\tilde{M}$

almost para-cosymplectic structure on the product $M=(a, b)\times\tilde{M}$ . In fact, it is

sufficient to assume that are given on $M$ by $\varphi,$
$\xi,$
$\eta,$ $g$

(1) $\varphi_{(t,p)}=(\tilde{J}_{l})_{p}$
, $g=dt\otimes dt+\tilde{g}_{l}$ , $\xi=\frac{\partial}{\partial t}$
, $\eta=dt$ .

The fundamental form $\Phi(X, Y)=g(\varphi X, Y)$ at any $(t, p)\in M$ is given by
, and therefore it is closed. Especially, if the family
$\Phi_{(t,p)}=(\tilde{\Omega}_{t})_{p}=\tilde{\Omega}_{p}$ $(\tilde{J}_{t},\tilde{g}_{t})$

collapses to a single almost para-K\"aler stmcture , that is, for $(\tilde{J},\tilde{g})$ $(\tilde{J}_{t},\tilde{g}_{t})=(\tilde{J},\tilde{g})$

any $t\in(a, b)$ , then we say that the almost para-cosymplectic manifold $M$ is the
product of the open interval $(a, b)$ and the almost para-K\"ahlerian manifold . $\tilde{M}$

We will show that any almost para-cosymplectic structure can locally be seen
as that in formula (1).
In fact, let be an almost para-cosymplectic structure on $M$ and a
$(\varphi, \xi, \eta, g)$
$p$

fixed point of $M$ . Since $d\eta=0$ and $\eta(\xi)=1$ , we choose a coordinate neigh-
bourhood $U$ around , which is diffeomorphic to $a>0,\tilde{U}\subset R^{2n}$
$p$ , $(-a, a)\times\tilde{U},$

with coordinates being the coordinate on $(-a, a)$ , such that

$(x^{0}, x^{1}, \ldots, x^{2n}),$ $x^{0}$

$\xi=\frac{\partial}{\partial x^{0}}$
, $\eta=dx^{0}$ .

Since $g0i=\delta_{0i},$ $g$ can be written as

$g=dx^{0}\otimes dx^{0}+\sum_{i,j=1}^{2n}g_{ij}dx^{i}\otimes dx^{j}$ .

By $\varphi\xi=0$ and $\eta(\varphi X)=0$ , we find

$\varphi=\sum_{i,j=1}^{2n}\varphi_{i}^{\dot{j}}dx^{j}\otimes\frac{\partial}{\partial x^{j}}$
.

Moreover, $\Phi(\xi, \cdot)=0$ and $\mathscr{L}_{\xi}\Phi=0$ yield for the components of $\Phi$

$\Phi_{0i}=0$ , $\frac{\partial\Phi_{ij}}{\partial x^{0}}=0$

Thus the fundamental form $\Phi$

has the shape

$\Phi=2\sum_{1\leq i<j\leq 2n}\Phi_{ij}dx^{j}\wedge dx^{j}$

 On almost para-cosymplectic manifolds 197

and does not depend on . For any fixed $x^{0}$ $x^{0}=t$ , define an almost para-
K\"ahlerian structure on by putting $(\tilde{J}_{l},\tilde{g}_{t})$
$\tilde{U}$

$\tilde{J}_{f}=\sum_{i,j=1}^{2n}\varphi_{i}^{j}(t, \cdot)dx^{i}\otimes\frac{\partial}{\partial x^{j}}$

, $\tilde{g}_{t}=\sum_{i,j=1}^{2n}g_{ij}(t, \cdot)dx^{i}\otimes dx^{j}$
,

with the fundamental form . $\tilde{\Omega}_{t}=\Phi|\tilde{U}$

We have just proved the following theorem.

THEOREM 1. Let $M(\varphi, \xi, \eta, g)$ be an almost para-cosymplectic manifold. Then,
for any poin $tp\in M$ ,
(a) there is a neighbourhood $U=(-a, a)\times\tilde{U}$
of $p$ , where $\tilde{U}$

is a 2n-
dimensional manifold and $a>0$ ; $d_{l}fferentiable$

(b) there exist a l-parameter family of almost lerian structures $para\neg K\ddot{a}h$

$t\in(-a, a)$ , which are

defined on with the fundamental forms
$(\tilde{J}_{f},\tilde{g}_{t}),$ $\tilde{U}$
$\tilde{\Omega}_{f}$

not depending on the parameter ; and $t,\tilde{\Omega}_{f}=\tilde{\Omega}$

(c) on , the structure is given as in formula (1).

$(-a, a)\times\tilde{U}$ $(\varphi, \xi, \eta, g)$ $\square $

Families of almost para-K\"ahlerian structures with the same fundamental

form can be constructed in many ways. Below, we present some of them.

EXAMPLE 1. be a fixed almost para-K\"ahler structure on a 2n-

Let $(\tilde{J},\tilde{g})$

dimensional differentiable manifold $N$ and its fundamental form. Let $V$ be an $\tilde{\Omega}$

open subset of $N$ endowed with a frame of vector fields $(E_{1}, \ldots, E_{2n})$ such that
$\tilde{J}E_{\alpha}=E_{\alpha+n},\tilde{J}E_{\alpha+n}=E_{\alpha},\tilde{g}(E_{\alpha}, E_{\beta})=\delta_{\alpha\beta},\tilde{g}(E_{\alpha+n}, E_{\beta+n})=-\delta_{\alpha\beta},\tilde{g}(E_{\alpha}, E_{\beta+n})=0$

for $\alpha,\beta=1,$
$\ldots,$
$n$
.
Given a family of functions $f_{t}$
: $V\rightarrow R,$ $a<t<b$ , define $(\tilde{J}_{f},\tilde{g}_{t})$

by
$\tilde{J}_{f}E_{\alpha}=\exp(f_{f})E_{\alpha+n}$
, $\tilde{J}_{l}E_{\alpha+n}=\exp(-f_{t})E_{\alpha}$
,
$\tilde{g}_{f}(E_{\alpha}, E_{\beta})=\exp(f_{t})\tilde{g}(E_{\alpha}, E_{\beta})$
, $\tilde{g}_{t}(E_{\alpha+n}, E_{\beta+n})=\exp(-f_{t})\tilde{g}(E_{\alpha+n}, E_{\beta+n})$

for any $t\in(a, b)$ . One checks that $(\tilde{J}_{f},\tilde{g}_{f})$

are almost para-K\"ahlerian structures
with fundamental forms . $\tilde{\Omega}_{f}=\tilde{\Omega}$
$\square $

EXAMPLE 2. Let be an almost para-K\"ahlerian structure on a 2n- $(\tilde{J},\tilde{g})$

dimensional differentiable manifold $N$ . Let $V$ be an open subset of $N$ and assume
that there exist a l-parameter family of diffeomorphisms : , $f_{t}$ $V\rightarrow f_{f}(V)\subset N$

$t\in(-a, a),$ $a>0$ , such that the fundamental form of $N$ is invariant with $\tilde{\Omega}$

respect to all , that is, . [One should note that any point of $N$ has a
$f_{l}\prime s$ $f_{f}^{*}\tilde{\Omega}=\tilde{\Omega}$
198 Piotr DACKO

neighbourhood $V$
with this property.] Define a family of almost para-Hermitian
structures $(\tilde{J}_{l},\tilde{g}_{l})$
on as follows
$V$

$\tilde{J}_{t}=f_{t*}^{-1}J\tilde{f}_{t*}$
, $\tilde{g}_{t}=f_{t}^{*}\tilde{g}$
.

It can be checked that are almost para-K\"ahlerian structures on $V$ with $(\tilde{J}_{l},\tilde{g}_{t})$

fundamental forms for any $t\in(-a, a)$ . $\tilde{\Omega}_{t}=\tilde{\Omega}$

The very special case of the above construction can be obtained when $X$ is a
vector field on $N$ satisfying . Then, any point of $N$ has a neighbourhood $\mathscr{L}_{X}\tilde{\Omega}=0$

$V\subset N$ and there exists a l-parameter group of diffeomorphisms : $f_{t}$

$ V\rightarrow$

$f_{t}(V)\subset N$ generated by $X$ . By , any preserves . $\mathscr{L}_{X}\tilde{\Omega}=0$ $f_{f}$

$\tilde{\Omega}$

$\square $

REMARK 1. An almost para-cosymplectic manifold $M$ possesses a canonical

foliation generated by the $2n$ -dimensional, completely integrable and
$\mathscr{F}$ $\varphi$

invariant distribution . A leaf of is a submanifold of $M$ of $\mathscr{D}=ker\eta$

$\tilde{M}$ $\mathscr{F}$

codimension 1. Since is a vector field normal to , we may treat as a $\xi|\tilde{M}$

$\tilde{M}$ $\tilde{M}$

pseudo-Riemannian hypersurface. Then $ A=-\nabla\xi$ restricted to is the shape $\tilde{M}$

operator of . $\tilde{A}$ $\tilde{M}$

Let be the $(1, 1)$ -tensor field defined by

$\tilde{J}$

and the induced metric $\tilde{J}\tilde{X}=\varphi\tilde{X}$ $\tilde{g}$

on . Then the pair

$\tilde{M}$
is an almost para-Hermitian structure on . In fact, $(\tilde{J},\tilde{g})$
$\tilde{M}$

it is almost para-K\"ahlerian since its fundamental form is closed, as it is the pull-

back of the fundamental form of $M$ .
Fix a point of $M$ and choose a neighbourhood $U=(-a, a)\times\tilde{U}$ , on which
the structure is given as in (1), where is a suitable family of
$(\varphi, \xi, \eta, g)$
$(\tilde{J}_{l},\tilde{g}_{t})$

almost para-K\"ahlerian structures on . Then is an open subset of a $\tilde{U}$

$\{t\}\times\tilde{U}$

leaf. Identifying the set with , we note that is just the induced $\tilde{U}$
$\{t\}\times\tilde{U}$ $(\tilde{J}_{t},\tilde{g}_{t})$

almost para-K\"ahlerian structure on . Moreover, by the equality $(\tilde{J},\tilde{g})$ $\{\iota\}\times\tilde{U}$

$g(AX, Y)=-(1/2)(\mathscr{L}_{\xi}g)(X, Y)$ (see Proposition 1), for the second fundamental

form of $h_{t}$
, we have .
$\{\iota\}\times\tilde{U}$ $h_{t}=-(1/2)(\partial\tilde{g}_{s}/\partial s)|_{s=\iota}$ $\square $

3. Basic Structure Identities

LEMMA 1. For an almost paracontact hyperbolic metric mamfold $M(\varphi, \xi, \eta, g)$

with its fundamental 2-form the following equations hold $\Phi$

(2) $(\nabla_{X}\Phi)(Y, Z)=g((\nabla_{X}\varphi)Y, Z)$ ,

(3) $(\nabla_{X}\Phi)(Z, \varphi Y)+(\nabla_{X}\Phi)(Y, \varphi Z)=-\eta(Z)g(AX, Y)-\eta(Y)g(AX, Z)$ ,

(4) $(\nabla_{X}\Phi)(\varphi Y, \varphi Z)-(\nabla_{X}\Phi)(Y, Z)=-\eta(Z)g(AX, \varphi Y)+\eta(Y)g(AX, \varphi Z)$ ,

where $ A=-\nabla\xi$ .
 On almost para-cosymplectic manifolds 199

PROOF. Equality (2) is obvious. Differentiating the identity $\varphi^{2}=I-\eta\otimes\xi$

covariantly, we obtain
(5) $(\nabla_{X}\varphi)\varphi Y+\varphi(\nabla_{X}\varphi)Y=g(AX, Y)\xi+\eta(Y)AX$

Projecting this equality onto $Z$ , we find (3).

To prove (4), we find at first
(6) $(\nabla_{X}\varphi)\xi=-\varphi\nabla_{X}\xi=\varphi AX$
,

whence it follows
$(\nabla_{X}\Phi)(Y, \xi)=-g((\nabla_{X}\varphi)\xi, Y)=-g(\varphi AX, Y)$ .
Replacing $Z$ by $\varphi Z$
in (3), and applying the last equality, we find (4). $\square $

PROPOSITION 2. For any almost para-cosymplectic mamfold, we have

(7) $(\nabla_{\varphi X}\varphi)\varphi Y-(\nabla_{X}\varphi)Y-\eta(Y)A\varphi X=0$
.

PROOF. Let us define $(0,3)$ -tensor field $B$

as follows
$B(X, Y, Z)=g((\nabla_{\varphi X}\varphi)\varphi Y-(\nabla_{X}\varphi)Y-\eta(Y)A\varphi X, Z)$

$=(\nabla_{\varphi X}\Phi)(\varphi Y, Z)-(\nabla_{X}\Phi)(Y, Z)-\eta(Y)g(\varphi X, AZ)$ .

Antisymmetrizing $B$
with respect to $X,$ $Y$ we have
$B(X, Y, Z)-B(Y, X, Z)=(\nabla_{\varphi X}\Phi)(\varphi Y, Z)-(\nabla_{\varphi Y}\Phi)(\varphi X, Z)$

$-(\nabla_{X}\Phi)(Y, Z)+(\nabla_{Y}\Phi)(X, Z)$

$-\eta(Y)g(\varphi X, AZ)+\eta(X)g(\varphi Y, AZ)$ .

Since the metric connection $\nabla$

is torsionless and $d\Phi=0$ ,

$(\nabla_{X}\Phi)(Y, Z)+(\nabla_{Y}\Phi)(Z, X)+(\nabla_{Z}\Phi)(X, Y)=0$ .

Applying this in the previous formula, we obtain

$B(X, Y, Z)-B(Y, X, Z)=-(\nabla_{Z}\Phi)(\varphi X, \varphi Y)+(\nabla_{Z}\Phi)(X, Y)$

$-\eta(Y)g(\varphi X, AZ)+\eta(X)g(\varphi Y, AZ)$ .

By (4), the right hand side of this equality vanishes identically, so that
$B(X, Y, Z)-B(Y, X, Z)=0$ , i.e. $B$ is symmetric with respect to $X,$ . $Y$
200 Piotr DACKO

Symmetrizing $B$
with respect to $Y,$ $Z$ , we find

$B(X, Y, Z)+B(X, Z, Y)=(\nabla_{\varphi X}\Phi)(\varphi Y, Z)+(\nabla_{\varphi X}\Phi)(\varphi Z, Y)$

$-\eta(Y)g(\varphi X, AZ)-\eta(Z)g$ ( $\varphi X,$ A $Y$ ).

This, with the help of (3), implies $B(X, Y, Z)+B(X, Z, Y)=0$ , i.e. $B$ is anti-
symmetric with respect to $Y,$ $Z$ . The tensor $B$ having such symmetries must
vanish identically, which implies (7). $\square $

LEMMA 2. For an almost para-cosymplectic mamfold, we also have

(8) $(\nabla_{\varphi X}\varphi)Y-(\nabla_{X}\varphi)\varphi Y+\eta(Y)AX=0$ ,

(9) $(\nabla_{\varphi X}\varphi)Y+\varphi(\nabla_{X}\varphi)Y-g(AX, Y)\xi=0$ .

PROOF. Putting $\varphi Y$

instead of $Y$
in (7), we get
$(\nabla_{\varphi X}\varphi)Y-\eta(Y)(\nabla_{\varphi X}\varphi)\xi-(\nabla_{X}\varphi)\varphi Y=0$ .

By (6), $(\nabla_{\varphi X}\varphi)\xi=\varphi A\varphi X=-AX$ , which applied to the above gives (8). Now, (9)
follows from (8) and (5). $\square $

PROPOSITION 3. For the curva ture of an almost para-cosymp lectic manifo $ld$
,
we have the following identities
(10) $R_{XY}\xi=-(\nabla_{X}A)Y+(\nabla_{Y}A)X$ ,

(11) $ R_{\varphi X\varphi Y}\xi+R_{XY}\xi+\varphi R_{\varphi XY}\xi+\varphi R_{X\varphi Y}\xi=-\nabla_{N(X,Y)}\xi$ ,

where $R_{XY}=[\nabla_{X}, \nabla_{Y}]-\nabla_{[X,Y]}$ and $N$ is the Nijenhuis torsion tensor of $\varphi$
,

$N(X, Y)=\varphi^{2}[X, Y]+[\varphi X, \varphi Y]-\varphi[\varphi X, Y]-\varphi[X, \varphi Y]$ .

PROOF. By $ AX=-\nabla_{X}\xi$ , we have

$(\nabla_{X}A)Y=-\nabla_{X}\nabla_{Y}\xi+\nabla_{\nabla_{X}Y}\xi$ .

Hence, we get
$-(\nabla_{X}A)Y+(\nabla_{Y}A)X=[\nabla_{X}, \nabla_{Y}]\xi-\nabla_{[X,Y]}\xi=R_{XY}\xi$ ,

that is, formula (10).

Proposition lleads to
(12) $\nabla_{\varphi X}\xi=-A\varphi X=\varphi AX=-\varphi\nabla_{X}\xi$ .
 On almost para-cosymplectic manifolds 201

Moreover, using (8), we obtain

$-(\nabla_{\varphi X}\varphi)\nabla_{Y}\xi+(\nabla_{X}\varphi)\varphi\nabla_{Y}\xi=0$
.

Hence this, together with (12), gives

(13) $\nabla_{\varphi X}\nabla_{\varphi Y}\xi+\varphi\nabla_{\varphi X}\nabla_{Y}\xi+\nabla_{X}\nabla_{Y}\xi+\varphi\nabla_{X}\nabla_{\varphi Y}\xi=0$

.

Now, using (12) and (13), we find

$-\nabla_{N(X,Y)}\xi=\nabla_{\varphi X}\nabla_{\varphi Y}\xi+\varphi\nabla_{\varphi X}\nabla_{Y}\xi+\nabla_{X}\nabla_{Y}\xi+\varphi\nabla_{X}\nabla_{\varphi Y}\xi$

$-\nabla_{\varphi Y}\nabla_{\varphi X}\xi-\varphi\nabla_{\varphi Y}\nabla_{X}\xi-\nabla_{Y}\nabla_{X}\xi-\varphi\nabla_{Y}\nabla_{\varphi X}\xi$

$-\nabla_{[X,Y]}\xi-\nabla_{[\varphi X,\varphi Y]}\xi-\varphi\nabla_{[\varphi X,Y]}\xi-\varphi\nabla_{[X,\varphi Y]}\xi$

$=R_{\varphi X\varphi Y}\xi+R_{XY}\xi+\varphi R_{\varphi XY}\xi+\varphi R_{X\varphi Y}\xi$ ,

that is (11). $\square $

4. Para-cosymplectic Manifolds
In this section, we prove various necessary and sufficient conditions for an
almost para-cosymplectic manifold to be para-cosymplectic.
At first, we prove the following proposition

PROPOSITION 4. For the Nijenhuis torsion tensor $N$

of an almost para-
cosymplectic manifold, we have the following

(14) $N(X, Y)=2((\nabla_{\varphi X}\varphi)Y-(\nabla_{\varphi Y}\varphi)X)=-2\varphi((\nabla_{X}\varphi)Y-(\nabla_{Y}\varphi)X)$ ,

(15) $N(\varphi X, \varphi Y)=N(X, Y)-2\eta(X)AY+2\eta(Y)AX$ ,

(16) $(\nabla_{\varphi Z}\Phi)(X, Y)=-\frac{1}{2}g(Z, N(X, Y))$ ,

(17) $\eta(N(X, Y))=0$ , $N(\xi, X)=2AX$ .

PROOF. Writing the Nijenhuis torsion tensor of $\varphi$

with the help of the Levi-
Civita connection, we get

$N(X, Y)=-\varphi(\nabla_{X}\varphi)Y+\varphi(\nabla_{Y}\varphi)X+(\nabla_{\varphi X}\varphi)Y-(\nabla_{\varphi Y}\varphi)X$ .

202 Piotr DACKO

Formula (14) follows from the above in view of (9). Using (14) and (7), we find
(15). Moreover, using (14), we compute

$(\nabla_{\varphi Z}\Phi)(X, Y)=-g((\nabla_{X}\varphi)Y-(\nabla_{Y}\varphi)X, \varphi Z)$

$=g(\varphi((\nabla_{X}\varphi)Y-(\nabla_{Y}\varphi)X), Z)=-\frac{1}{2}g(Z, N(X, Y))$ ,

which gives (16). Formulas (17) are immediate consequences of (14) and (15),
respectively. $\square $

THEOREM 2. For an almost para-cosymplectic mamfold $M$, the following

conditions are equivalent
(a) $M$ is para-cosymplectic,
(b) $N=0$ ,
(c) is parallel,
$\varphi$

(d) $M$ is locally a product of an open interval and a para-Kahlerian mamfold,

(e) the leaves of the canonical foliation
$\tilde{M}$

are tota fly geodesic and the $\mathscr{F}$

induced structures are para-Kahlerian. $(\tilde{J},\tilde{g})$

PROOF. : Note that $A=0$ since

$(a)\Rightarrow(e)$
and . $ AZ=\varphi(\nabla_{Z}\varphi)\xi$ $\nabla\varphi=0$

Therefore, for the shape operator of a leaf of , it holds $\tilde{A}=A|\tilde{M}=0$

. Thus, $\tilde{M}$
$\mathscr{F}$

is totally geodesic and V

$\tilde{M}$

by the Gauss equation. Consequently, $=\nabla|\tilde{M}$

, that is,
$\tilde{\nabla}\tilde{J}=0$
is para-K\"ahlerian.
$(\tilde{J},\tilde{g})$

: By Theorem 1, choose a neighbourhood $U=(-a, a)\times\tilde{U}$ on

$(e)\Rightarrow(d)$

which the structure is given as in (1), where

$(\varphi, \xi, \eta, g)$
is a family of $(\tilde{J}_{t},\tilde{g}_{t})$

almost para-K\"ahlerian structures on with not depending on . Restrict $\tilde{U}$ $\tilde{\Omega}_{t}$

$t$

further considerations to the set . As we have already known,

$U$
are the $(\tilde{J}_{t},\tilde{g}_{t})s$

induced structures on leaves. By our assumption, they are para-K\"ahlerian. Since

the leaves are also totally geodesic, their second fundamental forms vanish $h_{t}$

identically and consequently . Hence are independent

$(\partial/\partial t)\tilde{g}_{t}=-2h_{t}=0$ $(\tilde{J}_{t},\tilde{g}_{t})$

of . Thus, $t$
$M$ is locally a product of an open interval and a para-K\"ahlerian
manifold.
: It is obvious.
$(d)\Rightarrow(c)$

: It follows from (2) and (16).

$(c)\Rightarrow(b)$

: Since $N=0,$ (2) and (16) give

$(b)\Rightarrow(a)$
and . But then, $\nabla_{\varphi Z}\Phi=0$ $(\nabla_{\varphi Z}\varphi)=0$

by the virtue of Proposition 1 that , we get which in tum implies $\nabla_{\xi}\varphi=0$ $\nabla\varphi=0$

$\nabla\xi=0$ . Also by (2), so we have $\nabla\Phi=0$ . On the other hand, since

$\nabla_{\xi}\Phi=0$

$(\nabla_{X}\eta)(Y)=g(\nabla_{X}\xi, Y)$ and $\nabla\xi=0$ one gets . Thus (a) follows. $\nabla\eta=0$ $\square $
 On almost para-cosymplectic manifolds 203

Let us call an almost para-cosymplectic manifold, satisfying the condition

(18) $[R_{XY}, \varphi]=R_{XY}\circ\varphi-\varphi\circ R_{XY}=0$ ,

weakly para-cosymplectic.
It is obvious that a para-cosymplectic manifold is weakly para-cosymplectic.
The converse implication does not hold in general. Indeed, the almost para-
cosymplectic manifolds given in the below example fulfill (18) and are not para-
cosymplectic.

EXAMPLE 3. Consider the flat pseudo-Riemannian metric $g$ on $R^{3}$

of sig-
nature $(++-)$ ,

$g=dx^{1}\otimes dx^{1}+dx^{2}\otimes dx^{2}-dx^{3}\otimes dx^{3}$

.

Let $h=1-x^{1}-x^{3}$ and define a frame of vector fields $(E_{0}, E_{1}, E_{2})$ on $R^{3}$
by

$E_{0}=h\frac{\partial}{\partial x^{1}}+\frac{\partial}{\partial x^{2}}-h\frac{\partial}{\partial x^{3}}$

,

$E_{1}=\frac{\partial}{\partial x^{1}}-\frac{\partial}{\partial x^{3}}$

,

$E_{2}=\frac{1}{2}(1-h^{2})\frac{\partial}{\partial x^{1}}-h\frac{\partial}{\partial x^{2}}+\frac{1}{2}(1+h^{2})\frac{\partial}{\partial x^{3}}$

.

For these vector fields, we have $g(E_{0}, E_{0})=g(E_{1}, E_{2})=g(E_{2}, E_{1})=1$ , otherwise
$g(E_{j}, E_{\dot{j}})=0$ . Let be the dual frame of l-forms. Then can be
$(\omega^{0}, \omega^{1}, \omega^{2})$
$g$

written in the form

$g=\omega^{0}\otimes\omega^{0}+\omega^{1}\otimes\omega^{2}+\omega^{2}\otimes\omega^{1}$
.

Let us define $\varphi,$

$\xi,$
$\eta$
by

$\xi=E_{0}$ , $\eta=\omega^{0}$
, $\varphi=\omega^{1}\otimes E_{1}-\omega^{2}\otimes E_{2}$
.

Then $M_{0}=R^{3}(\varphi, \xi, \eta, g)$ is a 3-dimensional flat almost para-cosymplectic man-
ifold. By the flatness, realizes (18). Since $M_{0}$
, $AE_{2}=-\nabla_{E_{2}}\xi=\partial/\partial x^{1}-\partial/\partial x^{3}\neq 0$

then . Thus,
$\nabla\varphi\neq 0$
is weakly para-cosymplectic but not para-cosymplectic.
$M_{0}$

It is interesting to point out that the vector fields form a basis of a $E_{0},$ $E_{1},$ $E_{2}$

3-dimensional Lie algebra isomorphic to the Lie algebra of the Heisenberg group
. Explicitly, the Poisson brackets are the following
$H^{3}$

$[E_{0}, E_{1}]=0$ , $[E_{0}, E_{2}]=E_{1}$ , $[E_{1}, E_{2}]=0$ .

204 Piotr DACKO

Moreover, are complete. Thus, there is a unique Lie group structure

$E_{0},$ $E_{1},$ $E_{2}$ $G$

on with
$R^{3}$
$(O, 0,0)\in R^{3}$ as the identity element, for which are left- $E_{0},$ $E_{1},$ $E_{2}$

invariant [15]. Because the group is connected and simply connected, is $G$ $G$

isomorphic to the Heisenberg group . By the above construction, the structure $H^{3}$

is left-invariant.
$(\varphi, \xi, \eta, g)$

Let be a discrete, cocompact subgroup of

$G_{0}$
and be a $G$ $M_{1}=G_{0}\backslash G$

compact cosets manifold. Via the canonical projection, we obtain a flat non para-
cosymplectic, almost para-cosymplectic structure on , which will be denoted $M_{1}$

also by .
$(\varphi, \xi, \eta, g)$

Examples of strictly weakly para-cosymplectic manifolds in higher dimensions

can be obtained in the following way. Let $M=M_{0}$ or $M=M_{1}$ with the suitable
stmcture defined in the above and
$(\varphi, \xi, \eta, g)$
be an arbitrary para- $\tilde{M}(\tilde{J},\tilde{g})$

K\"ahlerian manifold. On the product manifold , define an almost $M^{\prime}=M\times\tilde{M}$

para-cosymplectic stmcture as the product stmcture $(\varphi^{\prime}, \xi^{\prime}, \eta^{\prime}, g^{\prime})$

$\varphi^{\prime}=(\varphi,\tilde{J})$
, $\xi^{\prime}=(\xi, 0)$
, $\eta^{\prime}=(\eta, 0)$
, $g^{\prime}=(g,\tilde{g})$
.
Then, clearly, $[R_{XY}^{\prime}, \varphi^{\prime}]=0$
and . Thus, is weakly para-cosymplectic
$\nabla^{\prime}\varphi^{\prime}\neq 0$ $M^{\prime}$

non para-cosymplectic. If $M=M_{1}$ and is compact, then $M$ ‘ is compact too.

$\tilde{M}$

$\square $

5. Manifolds with $Para- K\ddot{a}hlerian$ Leaves

In this section, we study almost para-cosymplectic manifolds, whose leaves
of the canonical foliation are para-Kahlerian submanifolds. We will call such
manifolds almost para-cosymplectic with para-K\"ahlerian leaves.

THEOREM 3. An almost para-cosymplectic mamfold $M$ has para-Kahlerian

leaves if and only any of the following equivalent conditions holds
$lf$

(19) $N(X, Y)=2\eta(X)AY-2\eta(Y)AX$ ,

(20) $(\nabla_{X}\varphi)Y=g(A\varphi X, Y)\xi-\eta(Y)A\varphi X$ ,

PROOF. Note that the Nijenhuis tensors $N$ and of and the induced para- $\tilde{N}$

$\varphi$

complex structure of a leaf are related by . If the induced

$\tilde{J}$ $\tilde{M}\in \mathscr{F}$
$N|_{\overline{M}}=\tilde{N}$

structures are para-K\"ahlerian, then

$(\tilde{J},\tilde{g})$
, and consequently $N(X, Y)=0$ $\tilde{N}=0$

for any vector fields $X,$ tangent to . Therefore, $N(\varphi X, \varphi Y)=0$ for any
$Y$ $\tilde{M}$

vector fields on $M$ , whence (19) follows by (15).

Let us assume (19) for an almost para-cosymplectic manifold. Then, by (16),
we have
 On almost para-cosymplectic manifolds 205

$(\nabla_{\varphi Z}\Phi)(X, Y)=-\eta(X)g(AZ, Y)+\eta(Y)g(AZ, X)$ ,

and hence
$(\nabla_{\varphi Z}\varphi)X=g(AZ, X)\xi-\eta(X)AZ$ .
If we replace $Z$ by in the last equation and use , we get (20). $\varphi Z$ $\nabla_{\xi}\varphi=0$

Now we prove that (20) implies the leaves of the manifold $M$ are para-
K\"ahlerian. Let be the induced almost para-K\"ahlerian structure on a leaf
$(\tilde{J},\tilde{g})$ $\tilde{M}$

and be the Levi-Civita connection of . By the Gauss equation

$\tilde{\nabla}$

$\tilde{g}$

$\tilde{\nabla}_{\overline{X}}\tilde{Y}=\nabla_{\overline{X}}\tilde{Y}-g(A\tilde{X},\tilde{Y})\xi$

and (20), we find

$(\tilde{\nabla}_{\tilde{X}}\tilde{J})\tilde{Y}=(\nabla_{\overline{X}}\varphi)\tilde{Y}+g(\varphi A\tilde{X},\tilde{Y})\xi=g(A\varphi\tilde{X},\tilde{Y})\xi+g(\varphi A\tilde{X},\tilde{Y})\xi=0$
,
hence $(\tilde{J},\tilde{g})$
is para-Kahlerian. $\square $

PROPOSITION 5. For an almost para-cosymplectic mamfold with para-

K\"ah lerian leaves, we have the $fo$ llowing curvature identity
(21) $R_{ZX}\varphi Y-\varphi R_{ZX}Y=g(A\varphi Z, Y)AX-g(A\varphi X, Y)AZ+g(AZ, Y)A\varphi X$

$-g(AX, Y)A\varphi Z-g(R_{ZX}\xi, \varphi Y)\xi-\eta(Y)\varphi R_{ZX}\xi$ .

PROOF. By $\varphi A=-A\varphi,$ (20) and $\eta(AX)=0$ we find

$(\nabla_{Z}(A\varphi))X=-(\nabla_{Z}\varphi)AX-\varphi(\nabla_{Z}A)X=-g(A\varphi Z, AX)\xi-\varphi(\nabla_{Z}A)X$ .
Differentiating covariantly (20) and using the relation above, we obtain
$(\nabla_{ZX}^{2}\varphi)Y=-g(A\varphi X, Y)AZ+g(AZ, Y)A\varphi X$

$+g((\nabla_{Z}A)X, \varphi Y)\xi+\eta(Y)\varphi(\nabla_{Z}A)X$ .

Now, the result follows if we antisymmetrize the last relation with respect to $Z,$ $X$

and use (10). $\square $

THEOREM 4. Almost para-cosymplectic mamfolds of constan $t$

non-zero sec-
tional curvature do not exist in dimensions $\geq 5$
.

PROOF. Let $M$ be an almost para-cosymplectic manifold of non-zero

constant sectional curvature of dimension $2n+1\geq 5$ . Then $\lambda\neq 0$

$R_{\xi X}\xi=\varphi R_{\xi\varphi X}\xi=\lambda\eta(X)\xi-\lambda X$ .

206 Piotr DACKO

On the other hand, by (10) and Proposition 1, $R_{\xi X}\xi=A^{2}X-(\nabla_{\xi}A)X$ and

. Hence, $(\nabla_{\xi}A)X=0$ and
$\varphi R_{\xi\varphi X}\xi=A^{2}X+(\nabla_{\xi}A)X$

(22) $ A^{2}X=-\lambda X+\lambda\eta(X)\xi$ .

Formula (11) implies

$2\lambda\eta(Y)X-2\lambda\eta(X)Y=-\nabla_{N(X,Y)}\xi=AN(X, Y)$ .

Applying $A$
to the both sides of the above equation and using (22), (17) and
$\lambda\neq 0$
, we find
$2\eta(Y)AX-2\eta(X)AY=-N(X, Y)$ .

Then by the virtue of Theorem 3, $M$ has para-K\"ahlerian leaves. By $Tr(A)=$

$Tr(A\varphi)=Tr(Z-\rangle\varphi R_{ZX}\xi)=0$ , the trace of (21) with respect to $Z$ gives

$(2n-1)\lambda g(X, \varphi Y)=-g(R_{\xi X}\xi, \varphi Y)=\lambda g(X, \varphi Y)$ .

This is a contradiction since $n\geq 2$ and $\lambda\neq 0$

. $\square $

LEMMA 3. Let be bilinear symmetric forms on a real s-dimensional vector

$u,$ $v$

space $W,$ $s\geq 2$ . If rank $(u)=rank(v)=p,$ and have a common diagonalizing $u$ $v$

basis and
$u(z, y)u(x, w)-u(x, y)u(z, w)+v(z, y)v(x, w)-v(x, y)v(z, w)=0$

for any $x,$ $y,$ $z,$ $w\in W$ , then $p\leq 2$ .

PROOF. Choose a basis $(e_{j}, i=1,2, \ldots, s)$ , so that $v(e_{i}, e_{j})=$ $u(e_{j}, e_{j})=a_{j}\delta_{ij},$

for certain
$b_{i}\delta_{ij}$
. We may assume that
$a_{j},$
$b_{j}$
for $i=1,$ , $a_{j}\neq 0,$ $b_{j}\neq 0$
$\ldots,$ $p$

otherwise $a_{j}=b_{l}=0$ . Let us suppose that $p\geq 3$ . From (3), for $z=y=e_{j}$ ,
$x=w=e_{j},$ $1\leq i\neq j\leq p$ , we obtain $a_{i}a_{j}+b_{i}b_{j}=0$ . Hence $b_{l}=-a_{j}a_{1}/b_{1}$ for
$2\leq i\leq p$ . This applied in the previous equation gives $a_{j}a_{j}=0,2\leq l\neq j\leq p$ ,

which is a contradiction. $\square $

THEOREM 5. Let $M$ be an almost para-cosymplectic manifold with para-

Kahlerian leaves. Then $M$ is weakly para-cosymplectic and only if the following $\iota f$

two conditions (I) and (II) are fulfilled

(I) the tensor field $A$ is a Codazzi tensor, that is, $(\nabla_{X}A)Y=(\nabla_{Y}A)X$ ;
(II) at any point $p\in M$ , the operator $A$ has one of the following shape
 On almost para-cosymplectic manifolds 207

(a) $A=0$ ,
(b) $AX=\epsilon g(X, V)V$ , where and $V$ is a non-zero null vector such
$|\epsilon|=1$

that $\varphi V=V$ or $\varphi V=-V$ ,

(c) $AX=\epsilon_{1}g(X, V_{1})V_{1}+\epsilon_{2}g(X, V_{2})V_{2}$ , where are non-zero $V_{1},$ $V_{2}$

orthogonal null vectors such that $\varphi V_{1}=-V_{1},$ and . $\varphi V_{2}=V_{2}$ $|\epsilon_{j}|=1$

PROOF. Let $M$ be the weakly para-cosymplectic. Then $\varphi R_{XY}\xi=R_{XY}\varphi\xi-$

$[R_{XY}, \varphi]\xi=0$ , and hence $R_{XY}\xi=0$ . Now (I) follows by (10). Observe, $0=$
$\varphi R_{\xi\varphi X}\xi=A^{2}X+(\nabla_{\xi}A)X=A^{2}X+(\nabla_{X}A)\xi=2A^{2}X$ .
Thus $A^{2}X=0$ .
By $R_{XY}\xi=0$ the identity (21) simplifies to
$g(A\varphi Z, Y)AX-g(A\varphi X, Y)AZ+g(AZ, Y)A\varphi X-g(AX, Y)A\varphi Z=0$ .

Projecting the last relation onto $\varphi W$

we find
(23) $g(A\varphi Z, Y)g(A\varphi X, W)-g(A\varphi X, Y)g(A\varphi Z, W)$

$+g(AZ, Y)g(AX, W)-g(AX, Y)g(AZ, W)=0$ .

Now, let $(E_{0}, E_{\alpha}, E_{\alpha+n}),$ $\alpha=1,$
$\ldots,$
$n$ , be a basis of the tangent space at a point
$p\in M$ , such that

(24) $E_{0}=\xi_{p}$ , $\varphi E_{\alpha}=E_{\alpha}$

, $\varphi E_{\alpha+n}=-E_{\alpha+n}$ ,
$g(E_{0}, E_{0})=1$ , $g(E_{\alpha}, E_{\alpha+n})=1$ .
For $X=\sum_{i=0}^{2n}X^{l}E_{i},$ $Y=\sum_{i=0}^{2n}Y^{l}E_{i}$
, put $g(AX, Y)=\sum_{i,j=0}^{2n}c_{ij}X^{t}Y^{j}$ . By $A\xi=0$ ,
$c_{0i}=g(A\xi_{p}, E_{i})=0$ and by $\varphi A=-A\varphi,$ (24)
$c_{\alpha(\beta+n)}=g(AE_{\alpha}, E_{\beta+n})=-g(\varphi AE_{\alpha}, \varphi E_{\beta+n})$

$=g(A\varphi E_{\alpha}, \varphi E_{\beta+n})=-g(AE_{\alpha}, E_{\beta+n})=-c_{\alpha(\beta+n)}=0$ .

Hence

(25) $g(AX, Y)=\sum_{\alpha,\beta=1}^{n}(c_{\alpha\beta}X^{\alpha}Y^{\beta}+c_{(\alpha+n)(\beta+n)}X^{\alpha+n}Y^{\beta+n})$ .

Observe the following

$g(A\varphi E_{\alpha}, E_{\beta})=g(AE_{\alpha}, E_{\beta})=c_{\alpha\beta}$
,

$g(A\varphi E_{\alpha}, E_{\beta+n})=g(AE_{\alpha}, E_{\beta+n})=c_{\alpha(\beta+n)}=0$ ,

$g(A\varphi E_{\alpha+n}, E_{\beta+n})=-g(AE_{\alpha+n}, E_{\beta+n})=-c_{(\alpha+n)(\beta+n)}$ .

208 Piotr DACKO

Thus we have

(26) $g(A\varphi X, Y)=\sum_{\alpha,\beta=1}^{n}(c_{\alpha\beta}X^{\alpha}Y^{\beta}-c_{(\alpha+n)(\beta+n)}X^{\alpha+n}Y^{\beta+n})$ .

The forms have the same rank , common diagonalizing basis

$g(A\cdot, \cdot),$ $g(A\varphi\cdot, \cdot)$ $r$

(by (25) and (26)) and fulfill (23). That means, they realize the assumptions
of Lemma 3, therefore it must hold $r\leq 2$ . Note that $r=rank(c_{\alpha\beta})+$
$rank(c_{(\alpha+n)(\beta+n)})$ . If $r=0$ , then $A=0$ . Let $r=1$ . We will show the assertion
(II)(b). At first, consider the case rank . Then $c_{(\alpha+n)(\beta+n)}=0$ , and $(c_{\alpha\beta})=1$

for and certain . Define a l-form

$c_{\alpha\beta}=\epsilon d_{\alpha}d_{\beta}$
and a vector
$|\epsilon|=1$ by $d_{\alpha}$
$\omega$ $V$

assuming . One checks that $A=$

$\omega(X)=\sum_{\alpha=1}^{n}d_{\alpha}X^{\alpha}$
and $V=\sum_{\alpha=1}^{n}d_{\alpha}E_{\alpha+n}$

$\varphi V=-V,$ $\omega(X)=g(X, V)$ and $\omega(V)=g(V, V)=0$ . Similarly, in the

$\epsilon\omega\otimes V,$

case rank and rank , we find a l-form

$(c_{\alpha\beta})=0$ and a vector $(c_{(\alpha+n)(\beta+n)})=1$ $\omega$ $V$

for which $\varphi V=V,$ $\omega(X)=g(X, V)$ and $g(V, V)=0$ . Let now
$A=\epsilon\omega\otimes V,$

$r=2$ .Suppose rank $(c_{\alpha\beta})=2$ . Then which together with (25) and
$c_{(\alpha+n)(\beta+n)}=0$ ,
(26) implies $g(A\varphi X, Y)=g(AX, Y)$ . Applying the last relation into (23), we find

$g(AZ, Y)g(AX, W)-g(AX, Y)g(AZ, W)=0$ ,

which clearly yields rank $(g(A\cdot, \cdot))=rank(c_{\alpha\beta})\leq 1$ , a contradiction. Hence

rank . Similar arguments show that rank $(c_{(\alpha+n)(\beta+n)})<2$ . Thus
$(c_{\alpha\beta})<2$

rank $(c_{\alpha\beta})=rank(c_{(\alpha+n)(\beta+n)})=1$ ,

that is, $c_{\alpha\beta}=\epsilon_{1}d_{\alpha}d_{\beta},$ $c_{(\alpha+n)(\beta+n)}=\epsilon_{2}h_{\alpha}h_{\beta},$ $|\epsilon_{j}|=1$

. Define l-forms $\omega_{1},$ $\omega_{2}$
and
vectors $V_{1},$ $V_{2}$
by assuming

$\omega_{1}(X)=\sum_{\alpha=1}^{n}d_{\alpha}X^{\alpha}$ , $\omega_{2}(X)=\sum_{\alpha=1}^{n}h_{\alpha}X^{\alpha+n}$ , $V_{1}=\sum_{\alpha=1}^{n}d_{\alpha}E_{\alpha+n}$ , $V_{2}=\sum_{\alpha=1}^{n}h_{\alpha}E_{\alpha}$

.

We verify that
$A=\epsilon_{1}\omega_{1}\otimes V_{1}+\epsilon_{2}\omega_{2}\otimes V_{2}$
, $\omega_{1}(X)=g(X, V_{1})$ , $\omega_{2}(X)=g(X, V_{2})$ ,

$\varphi V_{1}=-V_{1}$ , $\varphi V_{2}=V_{2}$ , $g(V_{1}, V_{1})=g(V_{2}, V_{2})=0$ .

Finally, $A^{2}=0$ implies A , hence $g(V_{1}, V_{2})=0$ . Thus, $V_{1}=\epsilon_{1}\epsilon_{2}g(V_{1}, V_{2})^{2}V_{1}=0$

the assertion (II)(c) holds.

Conversely, by (10) and (I), $R_{XY}\xi=0$ . Moreover, (II) implies (23). Con-
sequently, $R_{ZX}\varphi Y-\varphi R_{ZX}Y=0$ follows from (21). $\square $
 On almost para-cosymplectic manifolds 209

EXAMPLE 4. Let $M=R^{5}$ with coordinates $(z, u_{1}, u_{2}, v_{1}, v_{2})$ . Define a $(1, 1)-$

tensor field , a vector field , a l-form

$\varphi$
and a metric $\xi$
$\eta$ $g$ as follows
$\varphi\frac{\partial}{\partial z}=2u_{1}\frac{\partial}{\partial v_{1}}-2u_{2}\frac{\partial}{\partial v_{2}}$
,

$\varphi\frac{\partial}{\partial u_{1}}=-2u_{1}\frac{\partial}{\partial z}-\frac{\partial}{\partial u_{1}}+4u_{1}u_{2}\frac{\partial}{\partial v_{2}}$

,

$\varphi\frac{\partial}{\partial u_{2}}=2u_{2}\frac{\partial}{\partial z}+\frac{\partial}{\partial u_{2}}-4u_{1}u_{2}\frac{\partial}{\partial v_{1}}$

,

$\varphi\frac{\partial}{\partial v_{1}}=\frac{\partial}{\partial v_{1}}$

, $\varphi\frac{\partial}{\partial v_{2}}=-\frac{\partial}{\partial v_{2}}$

,

$\xi=\frac{\partial}{\partial z}-2u_{1}\frac{\partial}{\partial v_{1}}-2u_{2}\frac{\partial}{\partial v_{2}}$

,

$\eta=dz-2u_{1}du_{1}-2u_{2}du_{2}$ ,
$g=dz^{2}+2du_{1}dv_{1}+2du_{2}dv_{2}$ .
By straightforward computations we verify that $(\varphi, \xi, \eta, g)$
is the almost para-
cosymplectic structure on $M$ with the fundamental form $\Phi$
given by
$\Phi=4u_{1}dz\wedge du_{1}-4u_{2}dz\wedge du_{2}+8u_{1}u_{2}du_{1}\wedge du_{2}-2du_{1}\wedge dv_{1}+2du_{2}\wedge dv_{2}$ .
For the tensor field $ A=-\nabla\xi$ , we have

$A=2du_{1}\otimes\frac{\partial}{\partial v_{1}}+2du_{2}\otimes\frac{\partial}{\partial v_{2}}$

,

so that is of rank 2 everywhere on $M$ . The covariant derivative

$A$
, which is $\nabla\varphi$

nonzero, satisfies the relation (20) and therefore by Theorem 3, $M$ has para-
K\"ahlerian leaves. $M$ is weakly para-cosymplectic since the metric is flat. $g$ $\square $

6. Curvature Identities

PROPOSITION 6. For any almost para-cosymplectic manifold, we have

(27) $[R_{Z\varphi X}, \varphi]+[R_{\varphi ZX}, \varphi]-[R_{\varphi Z\varphi X}, \varphi]\varphi-[R_{ZX}, \varphi]\varphi$

$=\nabla_{\varphi N(Z,X)}\varphi+\eta\otimes(R_{\varphi Z\varphi X}\xi+R_{ZX}\xi)$ .

PROOF. At first, by (8), we have

$(\nabla_{X}\varphi)\varphi Y-(\nabla_{\varphi X}\varphi)Y=\eta(Y)AX$ .
210 Piotr DACKO

Next, differentiating this covariantly, we obtain

(28) $(\nabla_{ZX}^{2}\varphi)\varphi Y-(\nabla_{Z\varphi X}^{2}\varphi)Y$

$=(\nabla_{(\nabla_{Z}\varphi)X}\varphi)Y-(\nabla_{X}\varphi)(\nabla_{Z}\varphi)Y-g(AZ, Y)AX+\eta(Y)(\nabla_{Z}A)X$ .

Replacing in (28) $Z,$ $X,$ $Y$ by $\varphi V,$ $U,$ $\varphi W$

, respectively, we obtain
(29) $(\nabla_{\varphi VU}^{2}\varphi)W-\eta(W)(\nabla_{\varphi VU}^{2}\varphi)\xi-(\nabla_{\varphi V\varphi U}^{2}\varphi)\varphi W$

$=(\nabla_{(\nabla_{\varphi V}\varphi)U}\varphi)\varphi W-(\nabla_{U}\varphi)(\nabla_{\varphi V}\varphi)\varphi W-g(AV, W)AU$ .

On the other hand, using (7) and (9), we find

$(\nabla_{U}\varphi)(\nabla_{\varphi V}\varphi)\varphi W=(\nabla_{U}\varphi)(\nabla_{V}\varphi)W+\eta(W)(\nabla_{U}\varphi)A\varphi V$ ,

$(\nabla_{(\nabla_{\varphi V}\varphi)U}\varphi)\varphi W=(\nabla_{-\varphi(\nabla_{V}\varphi)U+g(AV,U)\xi}\varphi)\varphi W=-(\nabla_{\varphi(\nabla_{V}\varphi)U}\varphi)\varphi W$

$=-(\nabla_{(\nabla_{V}\varphi)U}\varphi)W-\eta(W)A\varphi(\nabla_{V}\varphi)U$ .

By these relations, (29) tums into

(30) $(\nabla_{\varphi^{2}VU}\varphi)W-(\nabla_{\varphi^{2}V\varphi U}\varphi)\varphi W$

$=-(\nabla_{(\nabla_{\nabla}\varphi)U}\varphi)W-(\nabla_{U}\varphi)(\nabla_{V}\varphi)W-g(AV, W)AU$

$+\eta(W)((\nabla_{\varphi^{2}VU}\varphi)\xi-(\nabla_{U}\varphi)A\varphi V-A\varphi(\nabla_{V}\varphi)U)$ .

Putting $V=X,$ $U=Z,$ $W=Y$ in (30) and adding the obtained relation to
(28), we have
$-[R_{z_{\varphi}x,\varphi]Y-(\nabla_{\varphi^{2}X\varphi Z}\varphi)\varphi Y+(\nabla_{ZX}^{2}\varphi)\varphi Y}$

$=-(\nabla_{(\nabla_{X}\varphi)Z}\varphi)Y+(\nabla_{(\nabla_{Z}\varphi)X}\varphi)Y-(\nabla_{X}\varphi)(\nabla_{Z}\varphi)Y-(\nabla_{Z}\varphi)(\nabla_{X}\varphi)Y$

$-g(AX, Y)AZ-g(AZ, Y)AX+\eta(Y)((\nabla_{\varphi^{2}XZ}\varphi)\xi-(\nabla_{Z}\varphi)A\varphi X$

$-A\varphi(\nabla_{X}\varphi)Z+(\nabla_{Z}A)X)$ .

Antisymmetrization of the last equality with respect to $Z,$ $X$ and application of
(14), gives

(31) $[R_{Z\varphi X}, \varphi]Y+[R_{\varphi ZX}, \varphi]Y-[R_{\varphi Z\varphi X}, \varphi]\varphi Y-[R_{ZX}, \varphi]\varphi Y$

$=-2(\nabla_{(\nabla_{Z}\varphi)X}\varphi)Y+2(\nabla_{(\nabla_{X}\varphi)Z}\varphi)Y-\eta(Y)S(Z, X)$

$=(\nabla_{\varphi N(Z,X)}\varphi)Y-\eta(Y)S(Z, X)$ ,

 On almost para-cosymplectic manifolds 211

where $S$
is a $(1,2)$ skew-symmetric tensor field. Put $ Y=\xi$ in (31) and find
$[R_{Z\varphi X\varphi]\xi+[R_{\varphi ZX},\varphi]\xi-(\nabla_{\varphi N(Z,X)\varphi)\xi=-S(Z,X)}}$ .

This implies $g(S(Z, X),$ $\xi$

) $=0$ . Having this in mind and projecting (31) onto , $\xi$

we find
$[R_{Z\varphi X\varphi]\xi+[R_{\varphi ZX},\varphi]\xi+\varphi[R_{\varphi Z\varphi X},\varphi]\xi+\varphi[R_{ZX},\varphi]\xi=(\nabla_{\varphi N(Z,X)\varphi)\xi}}$ ,

which having substituted to the previous equation, gives

$ S(Z, X)=\varphi[R_{\varphi Z\varphi X}, \varphi]\xi+\varphi[R_{ZX}, \varphi]\xi=-R_{\varphi Z\varphi X}\xi-R_{ZX}\xi$ .

But then this reduces (31) to (27). $\square $

Let $Ric$ and $Ric^{*}$

be the Ricci and $*$
-Ricci tensors defined by
$Ric(X, Y)=Tr\{Z\vdash\rightarrow R_{ZX}Y\}$ , $Ric^{*}(X, Y)=Tr\{Z\mapsto\varphi R_{ZX}\varphi Y\}$ .

Let be the Ricci and

$\overline{Ric},\overline{Ric}^{*}$
$*$
-Ricci operators and $r,$
$r^{*}$
be the scalar and $*-$

scalar curvatures given by

$Ric(X, Y)=g(\overline{Ric}X, Y)$ , $Ric^{*}(X, Y)=g(\overline{Ric}^{*}X, Y)$ ,
$r=Tr(\overline{Ric})$
, $r^{*}=Tr(\overline{Ric}^{*})$
.

THEOREM 6. For an almost para-cosymplectic mamfold, we have

(32) $R_{Z\varphi X}\varphi Y-\varphi R_{Z\varphi X}Y+R_{\varphi ZX}\varphi Y-\varphi R_{\varphi ZX}Y$

$-R_{\varphi Z\varphi X}Y+\varphi R_{\varphi Z\varphi X}\varphi Y-R_{ZX}Y+\varphi R_{ZX}\varphi Y=(\nabla_{\varphi N(Z,X)}\varphi)Y$ ,

(33) $Ric^{*}(X, Y)+Ric^{*}(Y, X)-Ric(X, Y)+Ric(\varphi X, \varphi Y)$

$+\frac{1}{2}(R_{\xi XY\xi}-R_{\xi\varphi X\varphi Y\xi})+\sum_{j=0}^{2n}\epsilon_{j}g((\nabla_{E_{j}}\varphi)X, (\nabla_{E_{j}}\varphi)Y)=0$ ,

(34) $r^{*}-r+Ric(\xi, \xi)+\frac{1}{2}g(\nabla\varphi, \nabla\varphi)=0$

.

where $(E_{j}, 0\leq j\leq 2n)$ is an orthonormal frame, $\epsilon_{j^{\prime}}s$

are the indicators of $E_{j}^{J}s$
and

$g(\nabla\varphi, \nabla\varphi)=\sum_{i,j=0}^{2n}\epsilon_{l}\epsilon_{j}g((\nabla_{E_{j}}\varphi)E_{i}, (\nabla_{E_{j}}\varphi)E_{i})$

.

PROOF. Formula (32) is in fact a direct consequence of the identity (27).

212 Piotr DACKO

Taking the trace of (32) with respect to $Z$ , we find

(35) $2Ric^{*}(X, Y)-2Ric(X, Y)+2Ric(\varphi X, \varphi Y)-2Tr\{Z-*\varphi R_{Z\varphi X}Y\}$

$+R_{\xi XY\xi}-R_{\xi\varphi X\varphi Y\xi}=Tr\{Z\mapsto(\nabla_{\varphi N(Z,X)}\varphi)Y\}$ .

Let $(E_{l})$
be a local orthonormal frame and compute

$Tr\{Z\mapsto\varphi R_{Z\varphi X}Y\}=-\sum_{i=0}^{2n}\epsilon_{j}R_{\varphi E_{j}Y\varphi XE_{j}}=-Ric^{*}(Y, X)$

.

By (16), we have

$Tr\{Z\mapsto(\nabla_{\varphi N(Z,X)}\varphi)Y\}=\sum_{i=0}^{2n}\epsilon_{i}g((\nabla_{\varphi N(E_{i},X)}\varphi)Y, E_{i})$

$=\frac{1}{2}\sum_{i=0}^{2n}\epsilon_{i}g(N(E_{i}, X),$ $N(E_{j}, Y))=-\frac{1}{2}\sum_{i=0}^{2n}\epsilon_{j}g(\varphi N(E_{i}, X),$ $\varphi N(E_{j}, Y))$

$=-\frac{1}{2}\sum_{i,j=0}^{2n}\epsilon_{i}\epsilon_{j}g(E_{j}, \varphi N(E_{i}, X))g(E_{j}, \varphi N(E_{i}, Y))$

$=-2\sum_{i,j=0}^{2n}\epsilon_{i}\epsilon_{j}g((\nabla_{E_{j}}\varphi)X, E_{i})g((\nabla_{E_{j}}\varphi)Y, E_{i})=-2\sum_{j=0}^{2n}\epsilon_{j}g((\nabla_{E_{j}}\varphi)X, (\nabla_{E_{i}}\varphi)Y)$ .

Applying these relations into (35), we find (33).

Taking the trace of (33) with respect to , we obtain $g$

(36) $2r^{*}-r+Tr_{g}\{(X, Y)\mapsto Ric(\varphi X, \varphi Y)\}+\frac{1}{2}Ric(\xi, \xi)$

$-\frac{1}{2}Tr_{g}\{(X, Y)-\rangle R_{\xi\varphi X\varphi Y\xi}\}+g(\nabla\varphi, \nabla\varphi)=0$

.

On the other hand, we find

$Tr_{g}\{(X, Y)\leftrightarrow Ric(\varphi X, \varphi Y)\}=-Tr\{X\mapsto\varphi\overline{Ric}\varphi X\}$

$=-Tr\{X-r\overline{Ric}\varphi^{2}X\}=-r+Ric(\xi, \xi)$ .
Moreover
$Tr_{g}\{(X, Y)\mapsto R_{\xi\varphi X\varphi Y\xi}\}=-Tr\{X-\rangle\varphi R_{\varphi X\xi}\xi\}$

$=-Tr\{X\mapsto R_{X\xi}\xi\}=-Ric(\xi, \xi)$ .

The last two relations reduce (36) to (34). $\square $

 On almost para-cosymplectic manifolds 213

FINAL REMARKS. Certain of our results are para-cosymplectic analogies of

theorems conceming almost cosymplectic manifolds proved in [12], [13], [7].

References
[1] C.-L. Bejan, Almost parahermitian structures on the tangent bundle of an almost para-
cohermitian manifold, In: The Proceedmgs of the Fifth National Seminar of Finsler and
Lagrange Spaces (Brasov, 1988), pp. 105-109, Soc. Mat. R.S. Rom\^ania,
$S\iota iinte$

Bucharest, 1989.
[2] D. E. Blair, Riemannian geometry of contact and symplectic manifolds, Progress Math. Vol.
203, Birkh\"auser, Boston-Basel-Berlin, 2002.
[3] K. Buchner and R. Rosca, Vari\’et\’es para-coK\"ahleriennes \‘a champ concirculaire horizontale, C.
R. Acad. Sci. Paris 285 (1977), Ser. A, 723-726.
[4] K. Buchner and R. Rosca, Co-isotropic submanifolds of a para-coKahlerian manifold with
concircular structure vector field, J. Geometry 25 (1985), 164-177.
[5] V. Cruceanu, P. Fortuny and P. M. Gadea, A survey on paracomplex geometry, Rocky
Mountain J. Math. 26 (1996), 83-115.
[6] V. Cruceanu, P. M. Gadea and J. Masqu\’e, Para-Hermitian and para-K\"ahler manifolds,
$Mu\tilde{n}oz$

Quademi dell Istituto di Matematica, Facolt\‘a di Economia, Universit\‘a di Messina, No. 1

(1995), 1-72.
[7] P. Dacko and Z. Olszak, On conformally flat almost cosymplectic manifolds with K\"ahlerian
leaves, Rend. Sem. Mat. Univ. Pol. Torino, Vol. 56 (1998), 89-103.
[8] S. Erdem, On almost (para)contact (hyperbolic) metric manifolds and harmonicity of $(\varphi, \varphi^{\prime})-$

holomorphic maps between them, Houston J. Math. 28 (2002), 21-45.

[9] S. I. Goldberg and K. Yano, Integrability of almost cosymplectic structures, Pacific J. Math. 31
(1969), 373-382.
[10] S. Kobayashi and K. Nomizu, Foundations of differentiable manifolds, Vol. I, Interscience
Publishing, New York-London, 1963.
[11] I. Mihai, R. Rosca and L. Verstraelen, Some aspects of the differential geometry of vector fields.
On skew symmetric Killing and conformal vector fields, and their relations to various
geometrical structures, Centre for Pure and Applied Differential Geometry, Katholieke
Universiteit Leuven/Brussel, 1996.
[12] Z. Olszak, On almost cosymplectic manifolds, Kodai Math. J. 4 (1981), 239-250.
[13] Z. Olszak, Almost cosymplectic manifolds with K\"ahlerian leaves, Tensor N.S. 46 (1987),
117-124.
[14] R. Rosca and L. Vanhecke, S\"ur une vari\’et\’e presque paracok\"ahl\’erienne munie d’une connexion
self-orthogonale involutive, Ann. Sti. Univ. “Al. I. Cuza” Iasi 22 (1976), 49-58.
[15] F. Tricerri and L. Vanhecke, Homogeneous structures on Riemannian manifolds, London Math.
Soc. Lect. Notes Ser. 83, Cambridge Univ. Press, Cambridge, 1983.

Institute of Mathematics
Wroclaw University of Technology
Wybrzeze Wyspia\’{n}skiego 27
50-370 Wroclaw
Poland