0% found this document useful (0 votes)

50 views

Tensor Operators and Their Applications

Uploaded by

rustomhabtai

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

50 views

Tensor Operators and Their Applications

Uploaded by

rustomhabtai

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 200

Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

TENSOR OPERATORS
AND THEIR APPLICATIONS
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.

Additional books in this series can be found on Nova’s website

under the Series tab.

Additional e-books in this series can be found on Nova’s website

under the e-book tab.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

TENSOR OPERATORS
AND THEIR APPLICATIONS

ARIF SALIMOV
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

New York

All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.

For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com

NOTICE TO THE READER

The Publisher has taken reasonable care in the preparation of this book, but makes no expressed
or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.

Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.

This publication is designed to provide accurate and authoritative information with regard to the
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.

Additional color graphics may be available in the e-book version of this book.

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Salimov, Arif, 1956- author.

Tensor operators and their applications / Arif Salimov (Ataturk University, Faculty of Science, Dep.
of Mathematics, Erzurum, Turkey).
pages cm
Includes bibliographical references and index.
ISBN: (eBook)
1. Tensor algebra. I. Title.
QA200.S25 2012
515'.724--dc23
2012020865

Published by Nova Science Publishers, Inc. † New York

Preface xi

1 On Operators Applied to Pure Tensor Fields 1

1.1. Pure Tensor Fields . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2. Tachibana Operators . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1. ϕφ −operator Applied to a Tensor Field of Type (1,1) . . 6
1.2.2. ϕφ −operator Applied to a Tensor Field of Type (1,s),
s≥2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.3. ϕφ −operator Applied to a 1-form . . . . . . . . . . . . 9
1.2.4. ϕφ −operator Applied to a Tensor Field of Type
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(0, s), s ≥ 2 . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.5. ϕφ −operator Applied to a Tensor Field of Type (r, s) . . 13
1.3. Vishnevskii Operators . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.1. ψφ −operator Applied to a Tensor Field of Type (1,s),
s≥0 . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.2. ψφ −operator Applied to a Tensor Field of Type (0,s) . . 16
1.3.3. ψφ −operator Applied to a Tensor Field of Type (r,s),
r>1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4. ψφ −operator Applied to a Pure Connection . . . . . . . . . . . 19
1.5. Tachibana Operators Applied to a Mixed Tensor Field . . . . . . 21
1.5.1. Pure Tensor Fields of Mixed Kind on Submanifolds . . . 21
1.5.2. ϕφ,φ̃ −operators . . . . . . . . . . . . . . . . . . . . . . 22
1.5.3. ϕφ,φ̃ −operator Applied to Tensor Fields of Type
(1, s, 0, q) and (0, s, 1, q) . . . . . . . . . . . . . . . . . 24
1.5.4. ϕφ,φe −operator Applied to Tensor Fields of Type
(0, s, 0, 0) and (0, 0, 0, q) . . . . . . . . . . . . . . . . . 26

1.6. Yano-Ako Operators . . . . . . . . . . . . . . . . . . . . . . . 28

1.6.1. Definitions . . . . . . . . . . . . . . . . . . . . . . . . 28
1.6.2. ϕS −operator Applied to a Tensor Field of Type (1, s) . . 29
1.6.3. ϕS −operator Applied to a Tensor Field of Type (0, s) . . 30
1.7. ψS −operators . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.7.1. ψS −operator Applied to a Tensor Field of Type (1, s),
s≥0 . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.7.2. ψS −operator Applied to a Tensor Field of Type (0, s) . . 34
1.8. Generalizations . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2 Algebraic Structures on Manifolds 39

2.1. Algebraic Theory . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.1.1. Associative Algebras . . . . . . . . . . . . . . . . . . . 40
2.1.2. Commutative Algebras . . . . . . . . . . . . . . . . . . 41
2.1.3. Holomorphic Functions . . . . . . . . . . . . . . . . . 43
2.2. Algebraic Π−structures on Manifolds . . . . . . . . . . . . . . 46
2.3. Integrable Regular Π−structure . . . . . . . . . . . . . . . . . 50
2.4. Pure Tensors with Respect to the Regular Structure . . . . . . . 54
2.5. A-holomorphic Tensors in Real Coordinate
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.6. Pure Connections . . . . . . . . . . . . . . . . . . . . . . . . . 58
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

2.7. Torsion Tensors of Pure Π−connections . . . . . . . . . . . . . 61

2.8. A−holomorphic Hypercomplex Connection . . . . . . . . . . . 62
2.9. Some Properties of Pure Curvature Tensors . . . . . . . . . . . 64

3 Applications to the Norden Geometry 69

3.1. Hyper-Kähler-Norden Manifolds . . . . . . . . . . . . . . . . . 70
3.2. Complex Kähler-Norden Manifolds . . . . . . . . . . . . . . . 76
3.3. Almost Product Riemannian Manifolds . . . . . . . . . . . . . 79
3.3.1. Decomposable Riemannian Manifolds . . . . . . . . . . 79
3.3.2. Para-Kähler-Norden Manifolds . . . . . . . . . . . . . . 80
3.3.3. Nonexistence of Para-Kähler-Norden Warped Metrics . 83
3.4. Dual-Kähler-Norden Manifolds . . . . . . . . . . . . . . . . . . 86
3.5. Norden–Hessian Structures . . . . . . . . . . . . . . . . . . . . 89
3.6. Norden-Walker Manifolds with Proper Structures . . . . . . . . 92
3.6.1. Almost Norden-Walker Metrics . . . . . . . . . . . . . 93
3.6.2. Integrability of Proper Almost Complex Structures . . . 95

3.6.3. Holomorphic Norden-Walker (Kähler-Norden-Walker)

Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.6.4. Curvature Properties of Norden-Walker Manifolds . . . 98
3.6.5. Isotropic Kähler-Norden-Walker Structures . . . . . . . 99
3.6.6. Quasi-Kähler-Norden-Walker Structures . . . . . . . . . 101
3.6.7. On the Goldberg Conjecture . . . . . . . . . . . . . . . 102
3.7. Opposite Almost Complex Structure . . . . . . . . . . . . . . . 105
3.8. Para-Norden-Walker Metrics . . . . . . . . . . . . . . . . . . . 107
3.9. Some Notes Concerning Norden-Walker
8-manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

4 Applications to the Theory of Lifts 117

4.1. Tensor Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.2. Horizontal and Complete Lifts of Vector Fields . . . . . . . . . 120
4.2.1. Vertical Lifts of Tensor Fields and γ−operator . . . . . . 120
4.2.2. Complete Lifts of Vector Fields . . . . . . . . . . . . . 121
4.2.3. Horizontal Lifts of Vector Fields . . . . . . . . . . . . . 122
4.2.4. Complete Lifts of Derivations . . . . . . . . . . . . . . 122
4.2.5. Derivations DKX Y and Formulas on Lie Derivations . . . 124
4.3. Cross-sections in the Tensor Bundle . . . . . . . . . . . . . . . 126
4.4. Lifts of Affinor Fields . . . . . . . . . . . . . . . . . . . . . . . 128
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

4.4.1. Complete Lifts of Affinor Fields . . . . . . . . . . . . . 128

4.4.2. Almost Complex Structures on Tqp (Mn ) . . . . . . . . . 133
4.4.3. Almost Hyperholomorphic Pure Submanifolds in the
Tensor Bundle . . . . . . . . . . . . . . . . . . . . . . 135
4.4.4. Horizontal Lifts of Affinor Fields . . . . . . . . . . . . 136
4.4.5. Diagonal Lifts of Affinor Fields . . . . . . . . . . . . . 138
4.5. Lifts of Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.5.1. Adapted Frames . . . . . . . . . . . . . . . . . . . . . 141
4.5.2. Sasakian Metrics on the Tensor Bundles . . . . . . . . . 142
4.5.3. Geodesics on Tensor Bundles . . . . . . . . . . . . . . 145
4.5.4. Jacobi Tensor Fields . . . . . . . . . . . . . . . . . . . 151
4.6. Some Special Cases . . . . . . . . . . . . . . . . . . . . . . . . 152
4.6.1. Para-Nordenian Structures in Cotangent Bundles . . . . 153
4.6.2. Paraholomorphic Cheeger-Gromoll Metric in the Tan-
gent Bundle . . . . . . . . . . . . . . . . . . . . . . . . 160
4.6.3. On Almost Complex Structures in Tangent Bundles . . . 166

4.7. Complete Lift of a Skew-Symmetric Tensor Field . . . . . . . . 168

References 175

Index 185
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

The notion of derivation of tensor fields is one of the central concepts and tools
of differential geometry. Typical examples of these derivations include Lie and
covariant differentiations with respect to a vector field on manifolds. A ten-
sor operator applied to a pure tensor field with respect to a fixed tensor field of
type (1, q), q > 0 is a generalization of Lie and covariant differentiations. An
especially important class of tensor operators is the class of Tachibana opera-
tors associated with a considering fixed tensor field of type (1, 1). The study of
Tachibana operators was started in the early 1960s by Tachibana [97], Tachibana
and Koto [98] and Sato [88]. Shirokov [93] and Kruchkovich [36] developed
the theory of Tachibana operators associated with a commutative hypercom-
plex structure. If affinors of commutaive hypercomplex structure are covariantly
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

constant with respect to the torsion-free connection, then the class of Tachibana
operators coincides with the class of Vishnevskii operators. This operator was
first considered by Vishnevskii [103] and devoloped by Vishnevskii, Shirokov
and Shurygin [107]. Lather the Tachibana and Vishnevskii operators was ques-
tioned in Norden geometry and in theory of complete and horizontal lifts by
Salimov [64], [67]. Other important class of tensor operators is the class of
Yano-Ako operators associated with a tensor field of type (1, 2). An affirmative
answer to the application of Yano-Ako operators was obtained in [68], [86]. On
the other hand, tensor operator can be applied to a pure connection and these
operators were lately investigated in [71].
The author believe that differential geometric applications of tensor opera-
tors is a very fruitful research domain and povides many new problems in the
study of modern differential geometry. However, in spite of its importance,
tensor operators and related topics are not as yet so well-known, and there is
no reference book covering this field. This was the motivation for publishing
the present monograph. This research monograph is intended to provide a sys-

tematic introduction to the theory of tensor operators. Let us first detailedly

outline tensor operators and describe some of the areas in which they find ap-
plications. Our goal is to expound the recent developments in applications of
tensor operators in hypercomplex, Norden geometry and theory of lifts. We plan
to present all these developments in four chapters. Preliminaries needed in this
monograph are intermediate level aspects of differential geometry. Instead of
collecting them in a single chapter, we prefer to introduce them briefly as they
are encountered in each section.
In Chapter 1 we introduce tensor operators. We treat pure tensor fields with
respect to the fixed tensor field, introduce several different families of tensor op-
erators applied to pure tensor fields, and also treat the theory of tensor operators
applied to pure connections.
Commutative hypercomplex structures on manifolds defined by a fixed ten-
sor field of type (1, 1) are discussed in Chapter 2. We introduce basic algebraic
theory, hypercomplex structures and holomorphic manifold theory. Formulas
are given for the holomorphic pure objects in the context of Tachibana opera-
tors associated with integrable commutative hypercomplex structures which are
central to our subsequent discussion.
Chapter 3 deals with special Riemannian metrics which is not necessarily
positive definite. For example, complex Norden manifolds are defined by a pure
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

metrics of neutral signature. We examine Kahler metrics of Norden type in the

context of Tachibana operators. The second part of this chapter is devoted to a
study of Kahler metrics of Norden type on the Walker manifolds.
Chapter 4 treats the theory of complete, horizontal and diagonal lifts of ten-
sor fields on a pure cross-sections in the tensor bundle. Various tensor operators
are used for interpretation and construction of these lifts.
This work was supported through the grants TBAG-108T590 and TBAG-
105T551 from TUBITAK (Turkey). I am very grateful to Professors V. N.
Berestovskii and V. V. Shurygin for their valuable suggestions and encourage-
ments.

Arif Salimov
Ataturk University, Turkey
January 2012

On Operators Applied to Pure

Tensor Fields

In this chapter we introduce the tensor operators which are used throughout this
book. In Section 1.1, we give the definition of pure tensor field with respect to
the affinor structure (i.e. tensor structure of type (1, 1) ) and its pure product.
In Section 1.2, we define the Tachibana operator applied to a pure tensor field.
Explicit expressions of these operators for different pure tensors are presented.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

In Section 1.3, by using a linear connection, we define the Vishnevskii operator

and we show that if the affinor structure is covariantly constant with respect to
the torsion-free connection, then the Tachibana operator reduces to the Vish-
nevskii operator. Section 1.4 devoted to the study of tensor operators applied to
a pure connection. By using these operators we discuss the purity conditons of
curvature tensors. In Section 1.5, we discuss some tensor operators concerning
pure tensor fields of mixed kind. In Section 1.6, we consider the case when the
tensor structure on manifolds is a tensor structure of type (1, 2) and we define
the Yano-Ako operator applied to a pure tensor field with respect to the tensor
structure of type (1, 2). Also explicit expressions of Yano-Ako operators are
presented. In Section 1.7, by using a linear connection, we define an operator
associated with a given tensor structure of type (1, 2) and applied to an arbitrary
tensor field. We prove that if the tensor structure of type (1, 2) is covariantly
constant with respect to the torsion-free connection, then the Yano-Ako oper-
ator reduces to this operator. Section 1.8 deals with generalizations of some
tensor operators.

1.1. Pure Tensor Fields

Let M be a C∞-manifold of finite dimension n. We denote by ℑrs (M) the module

over F(M) of all C∞ -tensor fields of type (r, s) on M, i.e. of contravariant degree
r and covariant degree s, where F(M) is the algebra of C∞ -functions on M.

Definition 1. Let ϕ be an affinor field on M, i.e. ϕ ∈ ℑ11 (M). A tensor field t of

type (r, s) is called pure tensor field with respect to ϕ if
1 2 r 1 2 r
t(ϕX1 , X2 , ..., Xs, ξ, ξ, ..., ξ) = t(X1 , ϕX2 , ..., Xs, ξ, ξ, ..., ξ)
...
1 2 r
= t(X1 , X2 , ..., ϕXs , ξ, ξ, ..., ξ)
1 2 r
= t(X1 , X2 , ..., Xs,´ϕξ, ξ, ..., ξ) (1.1)
1 2 r
= t(X1 , X2 , ..., Xs, ξ,´ϕξ, ..., ξ)
..
.
1 2 r
= t(X1 , X2 , ..., Xs, ξ, ξ, ...,´ϕξ)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1 2 r
for any X1 , X2 , ..., Xs ∈ ℑ10 (M) and ξ, ξ, ..., ξ ∈ ℑ01 (M), where ´ϕ is the adjoint
operator of ϕ defined by

(´ϕξ)(X) = ξ(ϕX) = (ξ ◦ ϕ)(X), X ∈ ℑ10 (M), ξ ∈ ℑ01 (M).

Let x1 , x2 , ..., xn be a local coordinate system in M. By setting X1 =

1 r
∂
∂xi1
= ∂x∂is and ξ = dx j1 , ..., ξ = dx jr in (1.1), we see that the condition
, ..., Xs
j ... j
of pure tensor fields may be expressed in terms of the components ϕij and ti11...is r
as follows:

j1 ... jr m ... jr m
tmi ϕ = tij11m...i
2 ...is i1
ϕ = ... = tij11i...
s i2
jr m
ϕ =
2 ...m is
(1.2)
m j ... j j m... j j j ...m
ti1 ...i2 s r ϕmj1 = ti11...is r ϕmj2 = ... = ti11...i2 s ϕmjr .

We consider for convenience’s sake the vector, covector and scalar fields as
pure tensor fields.

Example 1. In particular, let now t ∈ ℑ11 (M) be a pure tensor field of type (1, 1).
Then (1.1) may be written as:

ϕ(tX) = t(ϕX).

Thus if t ∈ ℑ11 (M) and ϕ ∈ ℑ11 (M) satisfies the commutativity condition

ϕ ◦ t = t ◦ ϕ, (1.3)
C C
where (ϕ ◦t)X = (ϕ ⊗ t)X = ϕ(tX) (⊗ is a tensor product with a contraction C),
then t is a pure with respect to ϕ, and conversely ϕ is also a pure with respect
to t.

From (1.3) it follows easily that ϕ itself and the unit affinor field I are exam-
ples of the pure tensor field. Also, from (1.3) we have: if ϕ is a regular affinor
field, i.e. det(ϕij ) 6= 0, then the affinor field ϕ−1 whose components are given by
the elements of the inverse matrix of ϕ is also pure.

Example 2. We consider the matrix equation

T
ϕg = gϕ, (1.4)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

where g = (gi j ), ϕ = (ϕij ), T ϕ is a transpose matrix. From (1.4) we obtain

gim ϕmj = gm j ϕm
i ,

which by virtue of (1.2) is a purity condition of a tensor field g ∈ ℑ02 (M).

Remark 1. Note that, when g is a (pseudo-)Riemannian metric, a linear op-

erator ϕ satisfying the purity condition g (ϕX,Y ) = g (X, ϕY ) is called a self-
adjoint. For example, the Jacobi operator and the Ricci operator on a (pseudo-)
Riemannian manifold, also an affinor of an almost complex structure on a Nor-
den manifold are self-adjoint operators.

Pure tensor fields have been studied in [19], [36], [37], [58], [66], [77], [88],
[93], [97], [98], [105], [108], [112], [113] from different view points.
From (1.1) we easily see that if K and L are both pure tensor fields of type
(r, s), then K + L and f K ( f ∈ F(M)) are also pure tensor fields. We denote by
∗
ℑrs (M) the module of all pure tensor fields of type (r, s) on M with respect to

the affinor field ϕ. If K and L are pure tensor fields of types (p1 , q1 ), p1 ≥ 1 and
(p2 , q2 ), q2 ≥ 1 respectively, then the tensor product of K and L with contraction
C mi ...i r ...r
2 p 1 p
K ⊗ L = K j1 ... jq 1 Lms2 ...s2q2
1

is also a pure tensor field. For simplicity, we shall prove only the case when
∗ ∗
K ∈ ℑ11 (M) and L ∈ ℑ02 (M). In fact,
C
(K ⊗ L)(ϕX,Y ) = K(L(ϕX,Y )) = K(L(X, ϕY ))
C
= (K ⊗ L)(X, ϕY ), X,Y ∈ ℑ10 (M).
∗ ∞ ∗
We shall now make the direct sum ℑ(M) = ∑ ℑrs (M) into an alegebra over
r,s=0
C
the real number R by defining the pure product (denoted by ⊗ or ” ◦ ”) of K ∈
∗ ∗
ℑqp11 (M) and L ∈ ℑqp22 (M) as follows:
 mi2 ...i p r1 ...r p

 K j1 ... jq 1 Lms2 ...s2q2 , i f p1 , q2 ≥ 1,

 i1 ...i p1
1
mr2 ...r p2
C C
⊗ : (K, L) → (K ⊗ L) = K m j L
2 jq1 s1 ...sq2
... , if p2 , q1 ≥ 1,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.


 0 , if p1 , p2 = 0,


0 , if q1 , q2 = 0.

In particular, let K = X ∈ ℑ10 (M), and L ∈ Λq (M) be a q−form. Then the pure
C
product X ⊗ L coincides with the usual interior product ιX L.

1.2. Tachibana Operators

Definition 2. [71], [97] Let ϕ ∈ ℑ11 (M), and ℑ(M) = ∑∞ r
r,s=0 ℑs (M) be a tensor
∗
algebra over R. A map φϕ |r+s>0 : ℑ(M) → ℑ(M) is called a Tachibana operator
or φϕ −operator on M if
(a) φϕ is linear with respect to constant coefficients,
∗
(b) φϕ : ℑrs (M) → ℑrs+1 (M) for all r and s,
C C C ∗
(c) φϕ (K ⊗ L) = (φϕ K) ⊗ L + K ⊗ φϕ L for all K, L ∈ ℑ(M).

(d) φϕX Y = −(LY ϕ)X for all X,Y ∈ ℑ10 (M), where LY is the Lie derivation
with respect to Y .
(e) (φϕX ω)Y = (d(ιY ω))(ϕX)−(d(ιY (ω◦ϕ)))X +ω((LY ϕ)X) = (ϕX)(ιY ω)
−X(ιϕY ω) + ω((LY ϕ)X) for all ω ∈ ℑ01 (M) and X,Y ∈ ℑ10 (M), where ιY ω =
C
ω(Y ) = ω ⊗ Y .

Remark 2. If r = s = 0, then from (c), (d) and (e) of Definition 2 we have

φϕX (ιY ω) = (ϕX)(ιY ω) − X(ιϕY ω) for ιY ω ∈ ℑ00 (M), which is not well-defined
φϕ −operator. Different choices of Y and ω leading to same function f = ιY ω do
2
get
the same values. Consider M = R with standard coordinates x, y. Let ϕ =
0 1
. Consider the function f ≡ 1. This may be written in many different
1 0
∂ ∂ ∂
ways as ιY ω. Indeed taking ω = dx, we may choose Y = ∂x or Y = ∂x + x ∂y .
Now the right-hand side of φϕX (ιY ω) = (ϕX)(ιY ω) − X(ιϕY ω) is (ϕX)1 − 0 = 0
∂
in the first case, and (ϕX)1 − Xx = −Xx in the second case. For X = ∂x , the
latter expression is −1 6= 0. Therefore, we put r + s > 0.

Remark 3. From (d) of Definition 2 we have

φϕX Y = [ϕX,Y ] − ϕ [X,Y ] .

By virtue of
[ f X, gY ] = f g [X,Y ] + f (Xg)Y − g (Y f ) X
for any f , g ∈ F(M), we see that φϕX Y is linear in X, but not Y.

We have in fact

φϕ( f X)Y = [ f ϕX,Y ] − ϕ [ f X,Y ]

= f [ϕX,Y ] − (Y f )ϕX − ϕ( f [X,Y ]) + ϕ(Y f )X
= f ([ϕX,Y ] − ϕ [X,Y ])
= f φϕX Y,

φϕX (gY ) = g φϕX Y + ((ϕX) g)Y − (Xg)ϕY.

In the sequel we write φϕY X for φϕX Y .

1.2.1. φϕ−operator Applied to a Tensor Field of Type (1,1)

∗
Let t ∈ ℑ11 (M), i.e. t ◦ ϕ = ϕ ◦ t (see Example 1). Then tY ∈ ℑ10 (M) for any
Y ∈ ℑ10 (M) and by (c) of Definition 2 we have
C C
φϕtY X = φϕt X ⊗ Y + t ⊗ φϕY X
(1.5)
= φϕt (X,Y ) + t φϕY X .

Using (1.5), we have from (d) of Definition 2

φϕt (X,Y ) = φϕtY X − t φϕY X
= (−LtY ϕ + t (LY ϕ))X (1.6)
= [ϕX,tY ] − ϕ [X,tY ] − t [ϕX,Y ] + ϕt [X,Y ]
= Qϕ,t (X,Y ) .

The tensor field Qϕ,t ∈ ℑ12 (M) is called the Nijenhuis-Shirokov tensor field [35],
[36]. Thus we have
∗
Theorem 1. Let t ∈ ℑ11 (M). Then φϕt is the Nijenhuis-Shirokov tensor field.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Since t ◦ ϕ = ϕ ◦ t is trivially satisfied for t = ϕ, we from (1.6) obtain

φϕ ϕ (X,Y ) = −LϕY ϕ + ϕ (LY ϕ) X
= [ϕX, ϕY ] − ϕ [X, ϕY ] − ϕ [ϕX,Y ] + ϕ2 [X,Y ]
= Nϕ (X,Y ),

where Nϕ is the Nijenhuis tensor field constructed from ϕ [57].

Thus we have

Theorem 2. If Nϕ is the Nijenhuis tensor field of ϕ, then

Nϕ = φϕ ϕ.

which is nothing but the torsion tensor field Sφ,t of φ and t [57], [33, p.38].
Thus we have

Theorem 3. Let ϕ,t ∈ ℑ11 (M). The expression φϕt + φt ϕ without the purity con-
dition t ◦ϕ = ϕ◦t defines a tensor field Sϕ,t of type (1, 2), where Sϕ,t is the torsion
tensor field of ϕ and t.

From the expression (1.6) and from Theorem 3 we have

Corollary 4. Let t be a tensor field of type (1, 1) without the purity condition.
Then
Sϕ,t (X,Y ) = Qϕ,t (X,Y ) − Qϕ,t (Y, X) + (ϕt − tϕ) [X,Y ] .
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

By a straightforward computation, we have

(a) φϕ I = 0,

(b) φϕX (tY ) = Qϕ,t (X,Y ) + t ◦ φϕY X,

(c) φϕt (X,Y ) = Qϕ,t (X,Y ) = −Qt,ϕ (Y, X) = − (φt ϕ) (Y, X),
(d) Qϕ,t1 ◦t2 = Qϕ,t1 ◦ t2 + t1 ◦ Qϕ,t2 ,
∗
(e) Qϕ,ϕ◦t2 = Nϕ ◦ t2 + ϕ ◦ Qϕ,t2 for any t,t1,t2 ∈ ℑ11 (M) and X,Y ∈ ℑ10 (M),
where I is an identity affinor and

C
Qϕ,t Y X = (Qϕ,t ⊗Y )X = Qϕ,t (X,Y ) ,
C
t1 ◦ Qϕ,t2 (X,Y ) = (t1 ⊗ Qϕ,t2 ) (X,Y ) = Qϕ,t2 (X,t1Y ) ,

Qϕ,t1 ◦ t2 (X,Y ) = t2 (Qϕ,t1 (X,Y )).

1.2.2. φϕ−operator Applied to a Tensor Field of Type (1,s), s ≥ 2

∗
Let t ∈ ℑ1s (M), that is

ϕ(t(Y1,Y2 , ...,Ys)) = t(ϕY1 ,Y2 , ...,Ys)

..
.
= t(Y1,Y2 , ..., ϕYs).
Using (c) and (d) of Definition 2 we have

φϕt (X,Y1 ,Y2 , ...,Ys) = φϕ (t (Y1 ,Y2 , ...,Ys)) X
s
− ∑ t Y1 ,Y2 , ..., φϕYλ X, ...,Ys
λ=1

= − Lt(Y1 ,Y2 ,...,Ys ) ϕ X (1.7)
s
+ ∑ t Y1 ,Y2 , ..., LYλ ϕ X, ...,Ys .
λ=1

The expression (1.7) has components

h s
φϕ t = ϕm h h m m h h m
k ∂m t j1 ... js − ϕm ∂k t j1 ... js − t j1 ... js ∂m ϕk + ∑ t j1 ...m... js ∂ jλ ϕk (1.8)
k j1 ... js
λ=1

with respect to natural coordinate system in M, which are components of a

tensor field of type (1, s + 1) with the purity condition of t.

Remark 4. In particular, let t ∈ ℑ12 (M). Using the alternation of φϕt
k j1 j2
with respect to k, j1 and j2 , we can check by an elementary tensor computation
h
that φϕt are components of a tensor field of type (1, 3) without the purity
[k j1 j2 ]
condition of t [113].
h
Willmore [111] proved that the tensor field with components φϕt
[k j1 j2 ]
reduces to the tensor field introduced by Slebodzinski [95], when t = Nφ , ϕ2 =
−id, and the Slebodzinski tensor field is identically zero.

1.2.3. φϕ−operator Applied to a 1-form

Let ω ∈ ℑ01 (M). Using Definition 2, we have

(φϕ ω)(X,Y) = (φϕX ω)Y (1.9)

= (ϕX) (ιY ω) − X ιϕY ω + ω ((LY ϕ)X)
= (LϕX ω − LX (ω ◦ ϕ))Y

for any X,Y ∈ ℑ10 (M), where the 1-form ω ◦ ϕ is defined by

C
(ω ◦ ϕ)Y = (´ϕω)Y = (ω ⊗ ϕ)Y = ω (ϕY ) .
From (1.9) we see that the tensor field φϕ ω ∈ ℑ02 (M) has components

φϕ ω = ϕm m m
i ∂m ω j − ∂i (ωm ϕ j ) + ωm ∂ j ϕi
ij

with respect to the natural frame. The φϕ ω is not skew-symmetric in i
ij
and j in general. The the tensor field φϕ ω coincides with the tensor field
[i j]
introduced by Frolicher and Nijenhuis [29], [30].
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 5. Let ω be an exact 1-form. Then ω ∈ Kerφϕ if and only if an asso-

ciated 1-form ω ◦ ϕ is a closed 1-form.

Proof. Let ω ∈ ℑ01 (M). Using [33, p.36]

1
(dω)(X,Y) = {X(ω(Y )) −Y ω(X) − ω([X,Y])}
2
for any X,Y ∈ ℑ10 (M) and ω ∈ ℑ01 (M),we have

1
(dω)(Y, ϕX) = {Y (ω(ϕX)) − (ϕX)(ω(Y )) − ω([Y, ϕX])}
2
1
= {Y (ω(ϕX)) − (ϕX)(ω(Y )) + ω([ϕX,Y ])}
2
1
= {Y (ω(ϕX)) − (ϕX)(ω(Y )) + ω([ϕX,Y ]
2
−ϕ[X,Y ]) + ω(ϕ[X,Y ])}. (1.10)

From (1.9), we have

(φϕ ω)(X,Y ) = (ϕX)(ω(Y )) − X(ω(ϕY )) − ω([ϕX,Y ] − ϕ[X,Y ]). (1.11)

Substituting (1.11) into (1.10), we obtain

1
(dω)(Y, ϕX) = {− φϕ ω (X,Y ) +Y (ω(ϕX)) − X(ω(ϕY )) + ω(ϕ[X,Y ])}
2
1
= {− φϕ ω (X,Y ) +Y ((ω ◦ ϕ)(X))
2
−X((ω ◦ ϕ)(Y )) − (ω ◦ ϕ)([Y, X])}
1
= − φϕ ω (X,Y ) + (d(ω ◦ ϕ))(Y, X).
2
From this we see that equation φϕ ω = 0 is equivalant to

(d(ω ◦ ϕ))(Y, X) = (dω)(Y, ϕX),

which for ω = d f turns into the following simple form:

(d(d f ◦ ϕ))(Y, X) = (d 2 f )(Y, ϕX) = 0,

i.e. d f ◦ ϕ is a closed 1-form.

Theorem 6. Let ω ∈ ℑ01 (M) and ϕ2 = −id. Then ω ◦ ϕ ∈ Ker φϕ if and only if
ω ∈ Ker φϕ .

Proof. If we substitute ω ◦ ϕ into ω and ϕX into X, then the equation (1.9) may
be written as

(φϕ (ω ◦ ϕ))(ϕX,Y ) = (Lϕ2 X (ω ◦ ϕ) − LϕX (ω ◦ ϕ2 ))Y

= −(LX (ω ◦ ϕ) + LϕX ω)Y
= −(φϕ ω)(X,Y )

or
((φϕ (ω ◦ ϕ)) ◦ ϕ)(X,Y ) = −(φϕ ω)(X,Y),
from which by virtue of detϕ 6= 0, we see that φϕ (ω ◦ ϕ) = 0 if and only if
φϕ ω = 0.

Corollary 7. Let ω ∈ ℑ01 (M), and ω ∈ Kerφϕ . If ϕ2 = −id, then ω ◦ Nϕ = 0,

where ω ◦ Nϕ (X,Y ) = ω Nϕ (X,Y ) .

Proof. It follows immediately from Theorem 6 and the following formula:

φϕ (ω ◦ ϕ) = (φϕ ω) ◦ ϕ + ω ◦ Nϕ .

1.2.4. φϕ−operator Applied to a Tensor Field of Type (0, s), s ≥ 2

∗
Theorem 8. Let ω ∈ ℑ0s (M). Then

(φϕ ω) (X,Y1 , ...,Ys) = (ϕX) (ω (Y1 , ...,Ys)) − X (ω (ϕY1 , ...,Ys))

s
+ ∑ ω(Y1 , ..., (LYλ ϕ)X, ...,Ys).
λ=1

Proof. For simplicity, let s = 2 and ωY1 be an 1-form such that ωY1 (Y2 ) =
ω (Y1 ,Y2 ) . By (e) of Definition 2 we have

(φϕX ωY1 )Y2 = (ϕX)(ωY1 (Y2 )) − X(ωY1 (ϕY2 )) + ωY1 ((LY2 ϕ)X)
= (ϕX)(ω (Y1 ,Y2 )) − X(ω (Y1 , ϕY2 )) + ω(Y1 , (LY2 ϕ)X),
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

from which, by virtue of

C C C
(φϕX (ω ⊗Y1 ))Y2 = ((φϕX ω) ⊗ Y1 + ω ⊗ (φϕX Y1 ))Y2
= (φϕX ω) (Y1 ,Y2 ) + ω(φϕX Y1 ,Y2 )

it follows that

(φϕX ω) (Y1 ,Y2 ) = (ϕX)(ω(Y1 ,Y2 )) − X(ω (ϕY1 ,Y2 ))

+ω(Y1 , (LY2 ϕ)X) + ω((LY1 ϕ)X,Y2 )

Thus the proof is completed.

∗
Let now ω ∈ ℑ02 (M), i.e.

ω(ϕX,Y ) = ω(X, ϕY ). (1.12)

Taking account of (1.12), Theorem 8, and φϕX Y = −(LY ϕ)X we now have

(φϕ ω)(X,Y, Z) = (ϕX)(ω(Y, Z)) − X(ω(ϕY, Z)) (1.13)

+ω((LY ϕ)X, Z) + ω(Y, (LZ ϕ)X)
= (LϕX ω − LX (ω ◦ ϕ))(Y, Z),

where the tensor field ω ◦ ϕ is defined by

(ω ◦ ϕ)(X,Y ) = ω(ϕX,Y ).
∗
Remark 5. We note that if ω ∈ ℑ02 (M) is a symmetric (skew-symmetric) pure
∗
tensor field, then ω ◦ ϕ ∈ ℑ02 (M) is also symmetric (skew-symmetric) pure tensor
field.

From (1.13) we see that the tensor field φϕ ω ∈ ℑ03 (M) has components

φϕ ω = ϕm m m
i ∂m ωi j − ∂k (ω ◦ ϕ)i j + ωm j ∂i ϕk + ωim ∂ j ϕk
ki j

with respect to the natural frame, where

(ω ◦ ϕ)i j = ωm j ϕm m
i = ωim ϕ j . (1.14)

The expression φϕ ω are components of a tensor field of type (0,3)
[ki j]
without the purity condition (1.14) , and if ω is a 2-form, then φϕ ω coin-
[ki j]
cides with the tensor field introduced by Frolicher and Nijenhuis [29], [30].
∗
If ω ∈ ℑ0s (M), s ≥ 2, then taking account of Theorem 8, we have

φϕ ω (X,Y1 , ...,Ys) = (ϕX) (ω (Y1 , ...,Ys)) − X (ω (ϕY1 , ...,Ys))
s
+ ∑ ω Y1 , ..., LYλ ϕ X, ...,Ys (1.15)
λ=1

= LϕX ω − LX (ω ◦ ϕ) (Y1 , ...,Ys) ,

where ω ◦ ϕ is defined by

(ω ◦ ϕ)(Y1 , ...,Ys) = ω (ϕY1 ,Y2 , ...,Ys)

..
.
= ω (Y1 ,Y2 , ..., ϕYs ).
Using (1.15), in a way similar to that of the proof of Theorem 6, we can prove
∗
Theorem 9. Let ω ∈ ℑ0s (M), s ≥ 2, and ϕ2 = −id. Then ω ◦ ϕ ∈ Kerφϕ if and
only if ω ∈ Kerφϕ .

∗
Corollary 10. Let ω ∈ ℑ0s (M), s ≥ 2, ω ∈ Kerφϕ . If ϕ2 = −id, then ω ◦ Nϕ = 0,

where ω ◦ Nϕ (X,Y1 , ...,Ys) = ω Nϕ (X,Y1 )Y2 , ...,Ys .

1.2.5. φϕ−operator Applied to a Tensor Field of Type (r, s)

∗
Let t ∈ ℑrs (M), r > 1, s ≥ 1. We now define a pure tensor field of type
∗ 1 2 r
(0,s) t 1 2 r ∈ ℑ0s (M) by t 1 2 r (Y1 ,Y2 , ...,Ys) = t(Y1 ,Y2 , ...,Ys, ξ, ξ, ..., ξ), where
ξ,ξ,...,ξ ξ,ξ,...,ξ
t1 2 r has components of the form:
ξ,ξ,...,ξ
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1 2 r
(t 1 2 r ) j1 j2 ... js = t ij11...i
... js ξi1 ξi2 ...ξir .
r

ξ,ξ,...,ξ

According to Theorem 8, we find

1 2 r r 1 µ r
φϕX t(Y1 , ...,Ys, ξ, ξ, ..., ξ)+ ∑ t(Y1, ...,Ys, ξ, ..., φϕX ξ, ..., ξ)
µ=1

= (φϕX t 1 2 r ) (Y1 , ...,Ys) = (ϕX) (t 1 2 r (Y1 , ...,Ys))

ξ,ξ,...,ξ ξ,ξ,...,ξ
s
−X(t 1 2 r (ϕY1 , ...,Ys))+ ∑ t 1 2 r Y1 , ..., (LYλ ϕ)X, ...,Ys .
ξ,ξ,...,ξ λ=1 ξ,ξ,...,ξ

µ µ
µ
Then, using φϕX ξ = LϕX ξ − LX ξ ◦ ϕ (see (1.9)), we see that φϕt for t ∈
∗
ℑrs (M), r > 1, s ≥ 1, is by definition, a tensor field of type (r, s + 1) given by

1 2 r
1 2 r

φϕ t X,Y1 , ...,Ys, ξ, ξ, ..., ξ = φϕX t Y1 , ...,Ys, ξ, ξ, ..., ξ
1 2 r
1 2 r

= (ϕX)t Y1 , ...,Ys, ξ, ξ, ..., ξ − Xt ϕY1 , ...,Ys, ξ, ξ, ..., ξ
s 1 2 r

+ ∑ t Y1 , ..., LYλ ϕ X, ...,Ys, ξ, ξ, ..., ξ
λ=1
r 1 µ
µ r

− ∑ t Y1 , ...,Ys, ξ, ..., LϕX ξ − LX ξ ◦ ϕ , ..., ξ . (1.16)
µ=1

µ
By setting X = ∂k , Yλ = ∂ jλ , ξ = dxiµ , λ = 1, ..., s; µ = 1, ..., r in the equa-
i1 ...ir
tion (1.16), we see that the components φϕt of φϕt with respect to local
j1 ... js
coordinate system x1 , ..., xn may be expressed as follows:

i1 ...ir
i1 ...ir i1 ...ir
φϕt = ϕm
k ∂m t j1 ... js − ∂k (t ◦ ϕ) j1 ... js (1.17)
k j1 ... js
s i1 ...ir r
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

iµ iµ
+ ∑ ∂ jλ ϕm
k t j1 ...m... js + ∑ k m m k t ij11...m...i
∂ ϕ − ∂ ϕ ... js
r
,
λ=1 µ=1

where

(t ◦ ϕ)ij11...i i1 ...ir m i1 ...ir m m...ir i1 i1 ...m ir

... js = tm... js ϕ j1 = ... = t j1 ...m ϕ js = t j1 ... js ϕm = ... = t j1 ... js ϕm .
r

The operator (1.17) first introduced by Tachibana [97].

1.3. Vishnevskii Operators

Suppose now that ∇ is a linear connection on M, and let ϕ ∈ ℑ11 (M).We can
replace the condition (d) of Definition 2 by
0
(d ) ψϕX Y = ∇ϕX Y − ϕ (∇X Y )
for any X,Y ∈ ℑ10 (M). Then we can consider a new operator as follows.

Definition 3. By a Vishnevskii operator or ψϕ −operator on M, we shall mean a

∗
map ψϕ : ℑ(M) → ℑ(M), which satisfies conditions (a), (b), (c), (e) of Definition
0
2 and the condition (d ).

Remark 6. Taking account of definition of connection ∇, we easily see that

ψϕX Y is linear in X, but not Y . In the sequal we write ψϕY X for ψϕX Y .

1.3.1. ψϕ −operator Applied to a Tensor Field of Type (1,s), s ≥ 0

∗
Let t ∈ ℑ1s (M). Since t (Y1 ,Y2 , ...,Ys) ∈ ℑ10 (M), then using Definition 3, we have

ψϕX t (Y1 ,Y2 , ...,Ys) = ψϕt (X,Y1 ,Y2 , ...,Ys)
s
= ψϕX t (Y1 ,Y2 , ...,Ys) − ∑ t Y1 , ..., ψϕYλ X, ...,Ys
λ=1
= ∇ϕX t (Y1 ,Y2 , ...,Ys) − ϕ∇X (t (Y1 ,Y2 , ...,Ys))
s
− ∑ t Y1 , ..., ψϕYλ X, ...,Ys ,
λ=1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

s
ψϕX t (Y1 ,Y2 , ...,Ys) = ∇ϕX t (Y1 , ...,Ys) + ∑t Y1 , ..., ∇ϕX Yλ , ...,Ys
λ=1
s
−ϕ (∇X t) (Y1 , ...,Ys) − ϕ( ∑ t (Y1 , ..., ∇X Yλ , ...,Ys))
λ=1

s
− ∑ t Y1 , ..., ψϕYλ X, ...,Ys . (1.18)
λ=1

Since t is a pure tensor field, we have

s s
ϕ( ∑ t (Y1 , ..., ∇X Yλ , ...,Ys)) = ∑ t (Y1, ..., ϕ(∇X Yλ) , ...,Ys) . (1.19)
λ=1 λ=1

Substituting (1.19) into (1.18), we obtain

ψϕ t (X,Y1 , ...,Ys) = ∇ϕX t − ϕ (∇X t) (Y1 , ...,Ys)

Thus

ψϕt (X,Y1 , ...,Ys) = ∇ϕX t − ϕ (∇X t) (Y1 , ...,Ys) . (1.20)

From (1.20) we have

∗
Theorem 11. Let t ∈ ℑ1s (M). If ∇t = 0, then t ∈ Kerψϕ .

1.3.2. ψϕ −operator Applied to a Tensor Field of Type (0,s)

Let ω ∈ ℑ01 (M). Using Definition 3, we have

ψϕ ω (X,Y ) = (ψϕX ω)Y (1.21)

= (ϕX)(ιY ω) − X(ιϕY ω) − ω ∇ϕX Y − ϕ (∇X Y )

= ∇ϕX ω − ∇X (ω ◦ ϕ) Y
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

for any X,Y ∈ ℑ10 (M),where (ω ◦ ϕ)Y = ω (ϕY ). From (1.21) we see that
ψϕX ω = ∇ϕX ω − ∇X (ω ◦ ϕ) is a 1−form.
∗
Let now ω ∈ ℑ0s (M), s > 1. Using Theorem 8, by similar devices we have

ψϕ ω (X,Y1 , ...,Ys) = ψϕX ω (Y1 , ...,Ys)
s
= (ϕX)(ω (Y1 , ...,Ys)) − Xω (ϕY1 , ...,Ys) − ∑ (ω Y1 , ..., ∇ϕX Yλ , ...,Ys
λ=1
− ω (Y1 , ..., ϕ(∇X Yλ ) , ...,Ys))
s
= ∇ϕX ω (Y1 , ...,Ys) − Xω (ϕY1 ,Y2 , ...,Ys) + ∑ ω (Y1 , ..., ϕ∇X Yλ, ...,Ys)
λ=1

= ∇ϕX ω (Y1 , ...,Ys) − Xω (ϕY1 ,Y2 , ..,Ys) + ω (ϕ(∇X Y1 ),Y2, ...,Ys)
+ ω (ϕY1 , ∇X Y2 , ...,Ys) + ... + ω (ϕY1 ,Y2 , ..., ∇X Ys )
= (∇ϕX ω − ∇X (ω ◦ ϕ) (Y1 , ...,Ys) .

Thus

ψϕ ω (X,Y1 , ...,Ys) = (∇ϕX ω − ∇X (ω ◦ ϕ) (Y1 , ...,Ys) . (1.22)

From (1.22) we have

∗
Theorem 12. Let ω ∈ ℑ0s (M). If ∇ω = 0, then ω ∈
/ Kerψϕ .

1.3.3. ψϕ −operator Applied to a Tensor Field of Type (r,s), r > 1

∗
Let t ∈ ℑrs (M). By similar devices (see Section 1.2.5) we have

1 2 r

ψϕt X,Y1 , ...,Ys, ξ, ξ, ..., ξ
1 2 r
1 2 r

= (ϕX)t Y1 , ...,Ys, ξ, ξ, ..., ξ − Xt ϕY1 ,Y2 , ...,Ys, ξ, ξ, ..., ξ
s 1 2 r

− ∑ t Y1 , ..., ∇ϕX Yλ − ϕ (∇X Yλ ) , ...,Ys, ξ, ξ, ..., ξ
λ=1
r µ
µ
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1 r
− ∑ t Y1 , ...,Ys, ξ, ..., ∇ϕX ξ − ∇X ξ ◦ ϕ , ..., ξ . (1.23)
µ=1

Substituting
1 µ
µ r

t Y1 , ...,Ys, ξ, ..., ∇ϕX ξ − ∇X ξ ◦ ϕ , ..., ξ
1 µ
µ µ r

= t Y1 , ...,Ys, ξ, ..., ∇ϕX ξ − ∇X ξ ◦ ϕ − ξ ◦ ∇X ϕ, ..., ξ
1 µ µ r

0
= t Y1 , ...,Ys, ξ, ..., ∇ϕX ξ − ϕ∇X ξ, ..., ξ
1 µ r

− t Y1 , ...,Ys, ξ, ..., ξ◦ ∇X ϕ, ..., ξ

into (1.23), we have

1 2 r

1 2 r

ψϕt X,Y1 , ...,Ys, ξ, ξ, ..., ξ = ∇ϕX t − ∇X (t ◦ ϕ) Y1 , ...,Ys, ξ, ξ, ..., ξ
1 µ r

r
+ ∑µ=1 t Y1 , ...,Ys, ξ, ..., ξ ◦ ∇X ϕ, ..., ξ .
(1.24)
Let now ∇ϕ = 0. Then from (1.24), we have

ψφX t = ∇ϕX t − (∇X t) ◦ ϕ. (1.25)

From (1.25) we easily see that ψϕt has components
i1 ...ir
i1 ...ir mi2 ...ir
ψϕt = ϕm j1
k ∇m t j1 ... js − ϕm ∇k t j1 ... js
k j1 ... js
with respect to the natural frame {∂i } .
The operator (1.25), first introduced by Vishnevskii [103] for an integrable
ϕ−structures.
In a manifold with an affinor field ϕ, a connection ∇ is called a
ϕ−connection if ∇ϕ = 0.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 13. If a torsion tensor of ϕ−connection ∇ is pure, then φϕX t = ψϕX t

∗
for any t ∈ ℑrs (M).

Proof. Let ∇ϕ = 0 and ϕT (X,Y ) = T (ϕX,Y ) = T (X, ϕY ), where T (X,Y ) =

∇X Y − ∇Y X − [X,Y ]. From (d) of Definition 2 and (d´), we obtain
φϕX Y = −(LY ϕ)X
= [ϕX,Y ] − ϕ[X,Y ]
= ∇ϕX Y − ∇Y ϕX − T (ϕX,Y ) − ϕ(∇X Y − ∇Y X − T [X,Y ])
= ∇ϕX Y − ϕ(∇X Y ) − (∇ϕ)(Y, X) + ϕT (X,Y ) − T (ϕX,Y )
= ∇ϕX Y − ϕ(∇X Y )
= ψϕX Y.

Remark 7. Since a zero torsion tensor field is pure, it follows that φϕX Y =
ψϕX Y for any torsion-free ϕ−connection.

1.4. ψϕ −operator Applied to a Pure Connection

Let ϕ ∈ ℑ11 (M), and ∇ be a torsion-free ϕ−connection, i.e. ∇ϕ
= 0. Then, it is
i
well known that ϕ is an integrable [92], i.e. the matrix ϕ = ϕ j is reduced to
constant form in a certain (holonomic) natural frame in a neighbourhood Ux of
every point x ∈ M.
If ∇ is a torsion-free ϕ−connection, then from ∇ϕ = 0 we have

Γkm j ϕm k m m k
i = Γim ϕ j = Γi j ϕm (1.26)
with respect to adapted local coordinates, where Γkij are components of ∇. We
call this connection a pure connection with respect to ϕ [36].
We denote the curvature tensor of the pure connection ∇ by R, which be-
longs to ℑ13 (M) :

R(X,Y)Z = ∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z, X,Y, Z ∈ ℑ10 (M).

Applying Ricci’s identity to ϕ:

∇X ((∇Y ϕ)Z) − ∇Y ((∇X ϕ)Z)

= R(X,Y)ϕZ − ϕ (R(X,Y )Z) + ∇[X,Y ] ϕ Z

+ (∇Y ϕ) (∇X Z) − (∇X ϕ)(∇Y Z) ,
we get

ϕR (X,Y ) Z = R (X,Y )ϕZ

by virtue of ∇ϕ = 0. Thus, we have

Theorem 14. Let ϕ ∈ ℑ11 (M). If ∇ is a pure connection with respect to ϕ, then
ϕR (X,Y ) Z = R (X,Y )ϕZ (1.27)
for any X,Y, Z ∈ ℑ10 (M).

Theorem 15. Let ϕ ∈ ℑ11 (M), and ∇ be a pure connection with respect to ϕ.
The curvature tensor field R of ∇ is a pure tensor field with respect to ϕ if and
only if
ψϕX (∇Y Z) = ∇ϕX ∇Y Z − ϕ (∇X ∇Y Z)
= 0

for any X,Y, Z ∈ Kerψϕ .

Proof. Since X,Y, Z ∈ Kerψϕ (see(d´)) and

ψϕY X = φϕY X = −(LX ϕ)Y = 0

(∇ϕ = 0, T (X,Y ) = ∇X Y − ∇Y X − [X,Y ] = 0)

we have

R(X, ϕY )Z = ∇X ∇ϕY Z − ∇ ϕY ∇X Z − ∇[X,ϕY ] Z

= ∇X ϕ(∇Y Z) − ∇ϕY ∇X Z − ∇(LX ϕ)Y +ϕLX Y Z
= ϕ(∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z)
+ϕ(∇Y ∇X Z) − ∇ ϕY ∇X Z
= ϕR(X,Y)Z − ψϕY (∇X Z).

Thus
R(X, ϕY )Z = ϕR(X,Y )Z − ψϕY (∇X Z). (1.28)
By virtue of R(X,Y )Z = −R(Y, X)Z, from (1.28) we have

R(ϕX,Y )Z = −R(Y, ϕX)Z

= −ϕR(Y, X)Z + ψϕX (∇Y Z). (1.29)

From (1.27)-(1.29), we see that the curvature tensor R of a pure connection ∇ is

pure in all arguments if and only if

ψϕX (∇Y Z) = 0

for any X,Y, Z ∈ Kerψϕ . Thus Theorem 15 is proved.

From (1.29) we see that

ψϕX (∇Y Z) ∈ ℑ13 (M).

Therefore, in the sequel we write ψϕ ∇ (X,Y, Z) for ψϕX (∇Y Z):

ψϕ ∇ (X,Y, Z) = ∇ϕX ∇Y Z − ϕ (∇X ∇Y Z) .
Thus ψϕ ∇ is a ψϕ −operator (or a φϕ −operator) applied to a pure connection ∇.
We note that, if a pure connection ∇ is a connection of Kahler-Norden man-
ifolds, then automatically it follows that ψϕ ∇ = 0 (see Chapter 3).

1.5. Tachibana Operators Applied to a Mixed Tensor

Field
1.5.1. Pure Tensor Fields of Mixed Kind on Submanifolds
Let M̃ be an m−dimensional submanifold of an n−dimensional manifold M
defined by immersion i : M̃ → M. The submanifold M̃ can be expressed locally
by
i
i i A
∂x
x = x u , i = 1, ..., n, A = 1, ..., m, rank = m.
∂uA
∂x i
We put BA = BiA ∂i = ∂u A ∂i , A = 1, ..., m, which are the m tangent vectors

to i M̃ in M. For an element X̃ ∈ ℑ10 (M̃) with local expression X̃ = X̃ A ∂A ,
we denote by BX̃ ∈ ℑ10 (M) the tangent vector field to i M̃ in M with local
expression

BX̃ = BiA X̃ A ∂i .
Thus the correspondence X̃ → BX̃ determines a mapping

B : ℑ10 (M̃) → ℑ10 i M̃ ,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

which is the differential of the immersion

i : M̃ → M, i.e. B = i∗ and so an
isomorphism of ℑ10 (M̃) onto ℑ10 i M̃ .
By a tensor field of mixed kind (or a tensor field of type (r, s; p, q)) we mean
a mapping

t : ℑ10 (M) × ... × ℑ10 (M) × ℑ10 (M̃) × ... × ℑ10 (M̃)×
×ℑ01 (M) × ... × ℑ01 (M) × ℑ01 (M̃) × ... × ℑ01 (M̃) → ℑ00 (M̃)

( ℑ10 (M): s times, ℑ01 (M): r times, ℑ10 (M̃): q times, ℑ01 (M̃): p times) defined
on M̃ that it is a tensor field of type (p, q) on M̃ for fixed arguments X1 , ..., Xs ∈
ℑ10 (M) and ξ1 , ..., ξr ∈ ℑ01 (M), and is a tensor field of type (r, s) on M for fixed
1 p
arguments X̃1 , ..., X̃q ∈ ℑ1 (M̃) and e
0 ξ , ..., e
ξ ∈ ℑ0 (M̃). 1

Definition 4. We assume the existence of ϕ ∈ ℑ11 (M) and ϕ̃ ∈ ℑ11 (M̃). If

1 r e1 ep
t X1 , ..., ϕXλ , ..., Xs, X̃1 , ..., X̃q, ξ , ..., ξ , ξ , ..., ξ

1 p
= t X1 , ..., Xs, X̃1 , ..., ϕ̃X̃µ , ..., X̃q, ξ1 , ..., ξr , e
ξ , ..., e
ξ ,

λ = 1, ..., s, µ = 1, ..., q,

then we say that t is pure with respect to the pair Xλ , X̃µ . If t is pure with
respect to all pairs of its vector and covector arguments, then it is called a pure
tensor field of type (r, s; p, q) with respect to (ϕ, ϕ̃).
∼
The vector fields X̃ ∈ ℑ10 (M̃) (or ξ ∈ ℑ01 (M̃)) and X ∈ ℑ10 (M) (or ξ ∈ ℑ01 (M))
restricted to M̃ is considered to be pure, by convention. In particular, the affinor
fields ϕ and ϕ̃ are pure tensor fields.
Let K and L be pure tensor fields of type (r1 , s1 ; p1 , q1 ) and (r2 , s2 ; p2 , q2 ),
respectively. The definition of pure product can be extended to a pure tensors of
mixed kind, as follows:

C

 pure product K(X̃ ,ξ̃) ⊗L(Ỹ ,η̃) on M,



 1 p1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

C
K ⊗L = for fixed X̃1 , ..., X̃q, e
ξ , ..., e
ξ , Ỹ1 , ...,Ỹq2 ,ηe1 , ...,ηe p2

 C

 pure product K(X,ξ) ⊗L(Y,η) on M̃,


for fixed X1 , ..., Xs1 , ξ1 , ..., ξr1 ,Y1 , ...,Ys2 , η1 , ..., ηr2
∗ ∗
r,p
Then as in Section 1.1, we can define an algebra ℑ(M̃) = ∑∞
r,s,p,q=0 ℑs,q (M̃)
C ∗
r,p
over R with respect to the pure product ⊗, where ℑs,q (M̃) is the module of all
pure tensor fields of type (r, s; p, q) with respect to (ϕ, ϕ̃) on M̃.

1.5.2. φϕ,ϕ̃ −operators

Let ϕ ∈ ℑ11 (M) and M̃ be a submanifold of M. If ϕTx M̃ ⊂ Tx (M) for each
x ∈ M̃, then M̃ is said to be invariant in M.
Let M̃ be an invariant submanifold. Then as m vectors BA = B∂˜ A =
i
BA ∂i , A = 1, ..., m in M span the tangent space of M̃, their images are linear
combinations of themselves. Hence there exists an affinor field ϕ̃ ∈ ℑ11 (M̃) such
that

ϕim Bm C i
A = ϕ̃A BC . (1.30)
Conversely if there exists an affinor field ϕ̃ ∈ ℑ11 (M̃) satisfying (1.30), then M̃
is invariant.
The equation (1.30) means that BiA is pure tensor field of type (1, 0, 0, 1) on
M̃ with respect to (ϕ, ϕ̃). The equation (1.30) may also be written as
0
B ϕ̃X̃ , ξ = B(X̃, ϕξ)
or

0
ξB ϕ̃X̃ = B(X̃, ϕξ) = B X̃, ξ ◦ ϕ = (ξ ◦ ϕ) BX̃ = ξ ϕ BX̃ ,
i.e.

ϕ BX̃ = B ϕ̃X̃ (1.31)
for any X̃ ∈ ℑ10 (M̃).
Now we introduce an operator on M̃ which is associated with ϕ and ϕ̃, and
applied to a pure tensors of mixed kind.
Definition 5. Let ϕ ∈ ℑ11 (M), and M̃ be an invariant submanifold with an
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

∗
induced affinor field ϕ̃ ∈ ℑ11 (M̃). A map φϕ,ϕ̃ |r+s+p+q>0 : ℑ(M̃) → ℑ(M̃) (
∗
r,p
ℑ(M̃) = ∑∞ r,s,p,q=0 ℑs,q (M̃) is an algebra of tensors of mixed kind ) is called
a Tachibana operator or φϕ,ϕ̃ −operator on M̃, if
(a) φϕ,ϕ̃ is linear with respect to constant coefficients,
∗ ∗
r,p r,p
(b) φϕ,ϕ̃ : ℑs,q (M̃) → ℑs,q+1 (M̃) for all r, s, p and q,
C C C ∗
(c) φϕ,ϕ̃ (K ⊗L) = φϕ,ϕ̃ K ⊗L + K ⊗ φϕ,ϕ̃ L for all K, L ∈ ℑ(M̃),

(d1 ) φϕ,ϕ̃Ỹ X̃ = φϕ̃Ỹ X̃ = − (LỸ ϕ̃) X̃ for any Ỹ ∈ ℑ10 (M̃),

(d2 ) φϕ,ϕ̃Y X̃ = φϕY BX̃ = − (LY ϕ) BX̃ for any Y ∈ ℑ10 (M),

(e1 ) φϕ,ϕ̃ ω̃ (X̃, Ye ) = φϕ̃ ω̃ X̃, Ye = d ((iỸ ω̃)) ϕ̃X̃

−d (iỸ ( ω̃ ◦ ϕ̃))
X̃ + ω̃((L
Ỹ ϕ̃)X̃) for any ω̃ ∈ ℑ01 (M̃),

(e2 ) φϕ,ϕ̃ ω (X̃,Y ) = φϕ ω BX̃,Y = d ((iY ω)) ϕ BX̃

−d (iY (ω ◦ ϕ)) BX̃ + ω((LY ϕ)(BX̃)) for any ω ∈ ℑ01 (M).

1.5.3. φϕ,ϕ̃ −operator Applied to Tensor Fields of Type (1, s, 0, q)

and (0, s, 1, q)

∗
1,0 e
Let t ∈ ℑs,q (M). Then t Y1 , ...,Ys, Ye1 , ..., Yeq ∈ ℑ10 (M) for any Y1 , ...,Ys ∈ ℑ10 (M)
and Ye1 , ..., Yeq ∈ ℑ1 (M).
0
e Using Definition 5 we have

(φϕ,eϕt Y1 , ...,Ys, Ye1 , ..., Yeq )Xe = (φϕ,eϕt)(X,Y
e 1 , ...,Ys, Ye1 , ..., Yeq)
s
e ...,Ys, Ye1 , ..., Yeq)
+ ∑ t(Y1, ..., (φϕ,eϕYλ )X,
λ=1
q
+ ∑ t(Y1 , ...,Ys, Ye1 , ..., (φϕ,eϕYeµ )X,
e ..., Yeq).
µ=1

From this we have following

e 1 , ...,Ys, Ye1 , ..., Yeq)

(φϕ,eϕt)(X,Y (1.32)

= (φϕ,eϕt Y1 , ...,Ys, Ye1 , ..., Yeq )Xe
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

s
e ...,Ys, Ye1 , ..., Yeq)
− ∑ t(Y1 , ..., (φϕ,eϕYλ )X,
λ=1
e
= −(Lt (Y1 ,...,Ys ,Ye1 ,...,Yeq ) ϕ)(BX)
s
e ...,Ys, Ye1 , ..., Yeq)
+ ∑ t(Y1 , ..., (LYλ ϕ)(BX),
λ=1
q
+ ∑ t(Y1 , ...,Ys, Ye1 , ..., (LYeµ ϕ e ..., Yeq).
e)X,
µ=1

By setting Xe = ∂C ,Yλ = ∂ jλ , Yeµ = ∂Aµ in the equation (1.32), we see that the
components (φϕ,eϕ t)Ci j1 ... js A1 ...Aq may be expressed as follows:

(φϕ,eϕ t)Ci j1 ... js A1 ...Aq (1.33)

= −BCh (t ij1 ... js A1 ...Aq ∂m ϕih − ϕm i h i m
h ∂m t j1 ... js A1 ...Aq ) − BC (ϕm ∂h t j1 ... js A1 ...Aq )

e 1 , ...,Ys, Ye1 , ..., Yeq) = −(L e e

(φϕ,eϕt)(X,Y t (Y1 ,...,Ys ,Ye1 ,...,Yeq ) ϕ)X

s
e ...,Ys, Ye1 , ..., Yeq)
+ ∑ t(Y1 , ..., (LYλ ϕ)(BX), (1.34)
λ=1
q
+ ∑ t(Y1, ...,Ys, Ye1 , ..., (LYeµ ϕ e ..., Yeq).
e)X,
µ=1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

e in (1.32), (1.34) respectively, then

In particular, if we put t = ϕ and t = ϕ
we have
e ) = Nϕ (BX,Y
(φϕ,eϕ ϕ)(X,Y e ), (φϕ,eϕ ϕ e Ye ) = Nϕe (X,
e)(X, e Ye),

e, respectively.
where Nϕ and Nϕe are Nijenhuis tensors of ϕ and ϕ
∗
Since B ∈ ℑ1,0 e
0,1 (M), we now consider the operator (1.32) acting on B:

e Ye) = −(L e ϕ)(BX)

(φϕ,eϕ B)(X, e + B((Le ϕ
e e
BY Y )X)
e + ϕ(L e BX)
= −LBYe (ϕ(BX)) e + B((Le ϕ
e e
BY Y )X).
h i h i
e BYe = B X,
By virtue of BX, e Ye [115, p.63] and (1.31), we have

e Ye ) = −L e (B(ϕ
(φϕ,eϕ B)(X, e + ϕ(L e BX)
eX)) e + B(Le ϕXe −ϕ e
e (LYe X))
BY Y
h i h BY i h i
e ϕ
= −B X, eYe + ϕB Ye , X
e + B(Le ϕX)
e − Bϕe Ye , Xe Y
= 0.

Thus we have
e be an invariant submanifold of M with an
Theorem 16. Let ϕ ∈ ℑ11 (M), and M
1 e
e ∈ ℑ1 (M). Then
induced ϕ
φϕ,eϕ B = 0
∗
for B ∈ ℑ1,0 e
0,1 (M),
e → M.
where B is the differential of the immersion i : M

C
e and B are pure tensor fields, a tensor fields ϕ ◦ B = ϕ ⊗ B and
Since ϕ, ϕ
C
B◦ϕe = B⊗ϕ e are also pure tensor fields of type (1, 0, 0, 1), where (ϕ ◦ B)Ye =
ϕ(BYe ) and (B ◦ ϕ
e)Ye = B(ϕ
eYe ) for any Ye ∈ ℑ10 (M).
e If we apply φϕ,eϕ to (1.31),
then by virtue of Theorem 2, Theorem 16 and (c) of Definition 5 we have
e Ye) = (φϕ,eϕ (B ◦ ϕ
(φϕ,eϕ (ϕ ◦ B))(X, e Ye ),
e))(X,

e BYe) = (B ◦ (φϕ,eϕ ϕ
(φϕ,eϕ ϕ)(X, e Ye) = (B ◦ Nϕe )(X,
e ))(X, e Ye) = B(Nϕe (X,
e Ye)),

Nϕ (BX, e Ye)),
e BYe ) = B(Nϕe (X, (1.35)
where Nϕ and Nϕe are Nijenhuis tensors of ϕ and ϕ e, respectively. Since B is
injective, from (1.35) we see that if Nϕ = 0, then Nϕe = 0.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Thus we have

Theorem 17. Let ϕ ∈ ℑ11 (M), and ϕ

e be the induced affinor field on invariant
e
submanifold M. Then if the Nijenhuis tensor Nϕ of ϕ vanishes, the Nijenhuis
e vanishes too.
tensor of ϕ

1.5.4. φϕ,eϕ −operator Applied to Tensor Fields of Type (0, s, 0, 0)

and (0, 0, 0, q)
∗
Let now ω ∈ ℑ0,0 e
s,0 (M). Using (e2 ), by similar devices, as in Section 1.3.2, we
have

e 1 , ...,Ys)
(φϕ,eϕ ω)(X,Y (1.36)
e
= (ϕ(BX))(ω(Y e
1 , ...,Ys)) − (BX)ω(ϕY1 , ...,Ys)
s
e − ϕ(LY BX),
+ ∑ ω(Y1 , ..., LYλ ϕ(BX) e ...,Ys)
λ
λ=1

(φϕ,eϕ ω e Ye1 , ..., Yeq) = (φϕe ω

e )(X, e Ye1 , ..., Yeq).
e )(X, (1.37)
∗
Let now g ∈ ℑ0,0 e
2,0 (M). We define a pure tensor field ge of type (0, 0, 0, 2) on
e
M by
ge(Ye1 , Ye2 ) = (i∗ g)(Ye1 , Ye2 ) = g(BYe1 , BYe2), (1.38)
for any Ye1 , Ye2 ∈ ℑ10 (M),
e where i∗ is the transpose of i∗ = B. In fact, by definition
C C C
of pure product ⊗, we easily see that a tensor field ge = (g ⊗ B) ⊗ B is a pure
∗
tensor field and ge ∈ ℑ0,0 e
2,0 (M). If we apply φϕ,e
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

ϕ to (1.38), then using (1.36) and

(1.37) we obtain

e Ye1 , Ye2 ) = (φϕ,eϕ ge)(X,

(φeϕ ge)(X, e Ye1 , Ye2 ) = (φϕ,eϕ g)(X,
e BYe1 , BYe2 )
e BYe1 , BYe2 ).
= (φϕ g)(BX,

i.e. φϕe ge = i∗ (φϕ g).

From here, it follows easily

Theorem 18. Let ϕ ∈ ℑ11 (M), and ϕ

e be the induced affinor field on an invariant
∗
submanifold M.e If φϕ g = 0 for g ∈ ℑ0 (M) then φϕe ge = 0, where ge is the induced
2
pure tensor field of type (0, 2) defined by (1.38).

The result in Theorem 18 can be specialized to the Norden geometry (see

Section 3.5.2).

1.6. Yano-Ako Operators

1.6.1. Definitions
Let S be a tensor field of type (1, 2).

Definition 6. A tensor field of type (r, s) is called pure tensor field with respect
to S if
1 r 1 r
t(SY1 X1 , ..., Xs; ξ, ..., ξ) = ... = t(X1, ..., SY1 Xs; ξ, ..., ξ) =
1 r 1 r
= t(X1 , ..., Xs;´ (SY1 )ξ, ..., ξ) = ... = t(X1 , ..., Xs; ξ, ...,´(SY1 )ξ)
1 r 1 r
t(SY2 X1 , ..., Xs; ξ, ..., ξ) = ... = t(X1, ..., SY2 Xs; ξ, ..., ξ) =
1 r 1 r
= t(X1 , ..., Xs;´ (SY2 )ξ, ..., ξ) = ... = t(X1 , ..., Xs; ξ, ...,´(SY2 )ξ)
1 r
for any X1 , ..., Xs,Y1 ,Y2 ∈ ℑ10 (M) and ξ, ..., ξ ∈ ℑ01 (M), where SY1 (SY2 ) denotes
an affinor field such that SY1 (Y2 ) =S(Y1 ,Y2 ), (SY2 (Y1 ) = S(Y1 ,Y2 )) , for any Y1 ∈
ℑ10 (M) Y2 ∈ ℑ10 (M) ,´ (SY1 ) ´(SY2 ) is the adjoint operator of SY1 , SY2 ∈ ℑ11 (M).

The condition of pure tensor fields with respect to S may be expressed in

terms of the local components as follows:

j1 ... jr m ... jr m ...m js
tmi S = ... = tij11...m
2 ...is i1 k
Sis k = tim... jr j1
S = ... = tij11...i
1 ...is mk s
Smk ,
j ... j j ... j m... j j j ...m j
tmi1 2 ...irs Sm 1 r m r 1 1 s
li1 = ... = ti1 ...m Slis = ti1 ...is Slm = ... = ti1 ...is Slm .

We consider for convenience’s sake the vector, kovector and scalar fields as
pure tensor fields with respect to S. Pure tensor fields with with respect to S
have been studied in [40], [86], [113].

Example 3. Let now t ∈ ℑ11 (M) be a pure tensor field with respect to S. Then,

S(Y1 ,tY2) = SY1 (tY2 ) = t(SY1 (Y2 )) = tS(Y1 ,Y2 ),

S(tY1,Y2 ) = SY2 (tY1 ) = t(SY2 (Y1 )) = tS(Y1 ,Y2 ),

which implies
S(tY1 ,Y2 ) = S(Y1 ,tY2 ) = tS(Y1 ,Y2 ),
i.e S is a pure tensor field with with respect to t ∈ ℑ11 (M) (see Section 1.2.2).

Definition 7. [71], [113] Let S ∈ ℑ12 (M), and ℑ(M) = ∑∞p,q=0 ℑrs (M) be a ten-
∗
sor algebra over R. A map φS |r+s>0 : S ℑ(M) → ℑ(M) is called a Yano-Ako
operator or φS -operator on M, if
(a) φS is a R- lineer, i.e. φS (at1 + bt2 ) = aφS (t1 ) + bφS (t2 ) for any a, b ∈
∗
R, t1 ,t2 ∈ S ℑrs (M),
∗
(b) φS : S ℑrs (M) → ℑrs+2 (M) for all r, s,
C C C ∗
(c) φS (K ⊗ L) = (φS K) ⊗ L + K ⊗ (φS L) for all K, L ∈ S ℑ(M),
(d) φS(X1 ,X2 )Y = −(LY S)(X1 , X2 ) for all X1 , X2 ,Y ∈ ℑ10 (M),
(e) (φS ω)(X1, X2 ,Y ) = (d(ιY ω))(S(X1, X2 )) − 2(d(ιY (ω ◦ S)))(X1, X2 )+
+ω(LY S)(X1 , X2 ) for all X1 , X2 ,Y ∈ ℑ10 (M), ω ∈ ℑ01 (M), where ιY ω =
C
ω(Y ) = ω ⊗ Y, S(X1 , X2 ) = −S(X2 , X1 ), d is the exterior differentiation of
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(ω ◦ S) ∈ ℑ01 (M).

In the sequel we write (φSY )(X1 , X2 ) = φS(X1 ,X2 )Y.

1.6.2. φS −operator Applied to a Tensor Field of Type (1, s)

∗
Let ϕ ∈ S ℑ11 (M). Then using Definition 7 we obtain

(φS ϕY )(X1, X2 ) = (φS ϕ)(X1, X2 ,Y ) + ϕ((φSY )(X1, X2 )),

which implies

(φS ϕ)(X1 , X2 ,Y ) = (φS ϕY )(X1 , X2 ) − ϕ((φSY )(X1, X2 )) (1.39)

= −(LϕY S)(X1 , X2 ) + ϕ((LY S)(X1, X2 ))
= −[ϕY, S(X1 , X2 )] + S([ϕY, X1 ], X2 ) + S(X1 , [ϕY, X2])
+ϕ([Y, S(X1, X2 )]) − S([Y, X1], X2) − S(X1 , [Y, X2]).

On the other hand, using Example 3, from (1.7) and (1.39) we have

(φS ϕ)(Y, X1, X2 ) = −(LS(X1 ,X2 ) ϕ)Y + S((LX1 ϕ)Y, X2) + S(X1 , (LX2 ϕ)Y )
= −[S(X1 , X2 ), ϕY ] + ϕ([S(X1, X2 ),Y ]) + S([X1 , ϕY ]
−ϕ[X1,Y ], X2) + S(X1 , [X2 , ϕY ] − ϕ[X2 ,Y ])
= −(φϕ S)(X1, X2 ,Y ).

Thus we have
∗
Theorem 19. Let ϕ ∈ S ℑ11 (M).Then

(φS ϕ)(Y, X1, X2 ) + (φϕ S)(X1, X2 ,Y ) = 0

for any X1 , X2 ,Y ∈ ℑ10 (M), where φS is the Yano-Ako operator applied to ϕ, φϕ

is the Tachibana operator applied to S .
∗
Let now t ∈ S ℑ1s (M), s > 1. Then using Definition 7 we have

(φS t)(X1, X2 ,Y1 , ...,Ys) = (φS t(Y1 , ...,Ys))(X1, X2 )

s
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

− ∑ t(Y1 , ..., (φSYλ )(X1 , X2 ), ...,Ys)

λ=1
= (−Lt(Y1 ,...,Ys ) S)(X1 , X2 )
s
+ ∑ t(Y1 , ..., (LYλ S)(X1, X2 ), ...,Ys).
λ=1

Thus

(φS t)(X1, X2 ,Y1 , ...,Ys) = (−Lt(Y1 ,...,Ys ) S)(X1 , X2 )

s
+ ∑ t(Y1 , ..., (LYλ S)(X1, X2 ), ...,Ys).
λ=1

1.6.3. φS −operator Applied to a Tensor Field of Type (0, s)

Let ω ∈ ℑ01 (M) and S be a vector-valued 2−form. Then using (e) of Definition
7 we have

(φS ω)(X1, X2 ,Y ) = (d(ιY ω))(S(X1, X2 )) − 2(d(ιY (ω ◦ S)))(X1, X2)

+ω(LY S)(X1 , X2 )
= (S(X1, X2 ))(ω(Y )) − X1 ω(S(Y, X2)) + X2 ω(S(Y, X1))
+(ω ◦ S)(Y, [X1, X2]) + ω((LY S)(X1 , X2 ))
= (S(X1, X2 ))(ω(Y )) − X1 ω(S(Y, X2)) + X2 ω(S(Y, X1))
−(ω ◦ S)([X1, X2 ],Y )ω([Y, S((X1, X2 ) − S(Y, X1 ), X2 ])
= (LS(X1 ,X2 ) ω)Y − (X1 ωS(Y, X2)) − (ω ◦ S)(Y, [X1, X2 ])
(ω ◦ S)([X1, X2 ],Y ) − (X2 ω(S(X1,Y ))
−(ω ◦ S)(X1 , [X2,Y ]) − (ω ◦ S)([X2, X1 ],Y )
−(ω ◦ S)([X1, X2 ],Y )
= (LS(X1 ,X2 ) ω)Y − (LX1 (ω ◦ S))(Y, X2)
−(LX2 (ω ◦ S))(X1,Y ) + (ω ◦ S)([X1 , X2 ],Y ).

Thus

(φS ω)(X1, X2 ,Y ) = (LS(X1 ,X2 ) ω)Y − (LX1 (ω ◦ S))(Y, X2)

−(LX2 (ω ◦ S))(X1,Y ) + (ω ◦ S)([X1 , X2 ],Y ).

∗
Let now ω ∈ S ℑ0s (M). Similarly we get (see Section 1.2.4)

(φS ω)(X1, X2 ,Y1 , ...,Ys) =

= S(X1 , X2)(ω(Y1 , ...,Ys) − X1 (ω(S(Y1 , X2 ),Y2 , ...,Ys))

−X2 (ω(S(X1 ,Y1 ),Y2, ...,Ys)) − (ω ◦ S)([X1 , X2 ],Y1,Y2 , ...,Ys)
s
+ ∑ ω(Y1 , ..., (LYλ S)(X1, X2 ), ...,Ys)
λ=1
= S(X1 , X2)(ω(Y1 , ...,Ys) − X1 (ω(S(Y1 , X2 ),Y2 , ...,Ys))
−X2 (ω(S(X1 ,Y1 ),Y2, ...,Ys)) − (ω ◦ S)([X1 , X2 ],Y1,Y2 , ...,Ys)
s
+ ∑ ω(Y1 , ..., LYλ S(X1 , X2 ) − S(LYλ X1 , X2 )
λ=1
−S(X1 , LYλ X2 ), ...,Ys)

= (LS(X1 ,X2 ) ω)(Y1, ...,Ys) − X1 ((ω ◦ S)(Y1 , X2 ,Y2 , ...,Ys))

−X2 ((ω ◦ S)(X1 ,Y1 ,Y2 , ...,Ys)) − (ω ◦ S)(LX1 X2 ,Y1 , ...,Ys)
s
+ ∑ (ω ◦ S)(Y1 , X2 ,Y2 , ..., LX1Yλ , ...,Ys)
λ=1
s
+ ∑ (ω ◦ S)(X1 ,Y1 , ..., LX2Yλ , ...,Ys)
λ=1
= (LS(X1 ,X2 ) ω)(Y1, ...,Ys) − (LX1 (ω ◦ S))(Y1, X2 ,Y2 , ...,Ys)
−(LX2 (ω ◦ S))(X1 ,Y1 , ...,Ys) + (ω ◦ S)([X1, X2 ],Y1 , ...,Ys).

Thus

(φS ω)(X1, X2 ,Y1 , ...,Ys) = (LS(X1 ,X2 ) ω)(Y1 , ...,Ys)

−(LX1 (ω ◦ S))(Y1, X2 ,Y2 , ...,Ys)
−(LX2 (ω ◦ S))(X1,Y1 , ...,Ys)
+(ω ◦ S)([X1, X2 ],Y1 , ...,Ys).

Local expressions of the Yano-Ako operators and their applications to the theory
of complete lifts we will consider in Chapter 4.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1.7. ψS −operators
Let ∇ be a linear connection on M, and let S ∈ ℑ12 (M). We can replace the
condition (d) of Definition 7 by

(d ´) ψS(X1 ,X2 )Y = ∇S(X1 ,X2 )Y − S(∇X1 Y, X2 ) − S(X1 , ∇X2 Y )

for any X1 , X2 ,Y ∈ ℑ10 (M). We can consider a new operator as follows.

∗
Definition 8. By a ψS −operator on M, we shall mean a map ψS : S ℑ(M) →
ℑ(M), which satisfies conditions (a), (b), (c), (e) of Definition 7 and the condi-
tion ( d ´).

Remark 8. We easily see that ψS(X1 ,X2 )Y is linear in X1 , and X2 but not Y . In
the sequal we write (ψSY )(X1 , X2 ) for ψS(X1 ,X2 )Y .

1.7.1. ψS −operator Applied to a Tensor Field of Type (1, s), s ≥ 0

∗
Let t ∈ S ℑ1s (M). Since t(Y1 ,Y2 , ...,Ys) ∈ ℑ10 (M), then using Definition 8, we have

(ψS(X1 ,X2 )t)(Y1,Y2 , ...,Ys) = (ψS t)(X1, X2 ,Y1 ,Y2 , ...Ys)

s
= ψS(X1 ,X2 )t(Y1 ,Y2 , ...,Ys) − ∑ t(Y1 , ..., (ψS(X1 ,X2 )Yλ ), ...,Ys)
λ=1
= ∇S(X1 ,X2 )t(Y1 ,Y2 , ...,Ys) − S(∇X1 (t(Y1,Y2 , ...,Ys), X2))
s
−S(X1 , ∇X2 (t(Y1,Y2 , ...,Ys))) − ∑ t(Y1 , ..., (ψS(X1 ,X2 )Yλ ), ...,Ys),
λ=1

(ψS(X1 ,X2 )t)(Y1,Y2 , ...,Ys)

s
= ∇S(X1 ,X2 )t(Y1 ,Y2 , ...,Ys) + ∑ t(Y1, ..., ∇S(X ,X )Yλ, ...,Ys)
1 2
λ=1
s
−S((∇X1 t)(Y1 , ...,Ys), X2 ) − S( ∑ t(Y1 , ..., ∇X1 Yλ , ...,Ys), X2 )
λ=1
s
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

−S(X1 , (∇X2 t)(Y1 , ...,Ys)) − S(X1 , ∑ t(Y1 , ..., ∇X2 Yλ , ...,Ys))

λ=1
s
− ∑ t(Y1 , ..., (ψS(X1 ,X2 )Yλ ), ...,Ys). (1.40)
λ=1

Since t is a pure tensor field with respect to S, we have

(1.41)
s s
S( ∑ t(Y1, ..., ∇X1 Yλ , ...,Ys), X2) = ∑ t(Y1, ..., S(∇X Yλ, X2), ...,Ys),
1
λ=1 λ=1
s s
S(X1 , ∑ t(Y1, ..., ∇X2 Yλ , ...,Ys)) = ∑ t(Y1, ..., S(X1, ∇X Yλ), ...,Ys).
2
λ=1 λ=1

Substituting (1.41) into (1.40), we obtain

(ψS t)(X1, X1 ,Y1 , ....,Ys) = (∇S(X1 ,X2 )t − SX1 (∇X2 t) − SX2 (∇X1 t))(Y1, ...,Ys)

1.7.2. ψS −operator Applied to a Tensor Field of Type (0, s)

Let ω ∈ ℑ01 (M). Using Definition 8, we have
(ψS ω)(X1, X2 ,Y ) = (ψS(X1 ,X2 ) ω)(Y )
= S(X1 , X2 )(ιY ω) − 2(d(ιY (ω ◦ S)))(X1, X2 )
−ω(∇S(X1 ,X2 )Y − S(∇X1 Y, X2) − S(X1 , ∇X2 Y ))
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

= (S(X1, X2 ))(ω(Y )) − X1 (ω(S(Y, X2)))

+X2 (ω(S(Y, X1))) + (ω ◦ S)(Y, [X1, X2 ])
−ω(∇S(X1 ,X2 )Y − S(∇X1 Y, X2) − S(X1 , ∇X2 Y ))
= (∇S(X1 ,X2 ) ω − ∇X1 (ω ◦ SX2 ) − ∇X2 (ω ◦ SX1 ))Y
−(ω ◦ S)([X1, X2 ],Y )
for any X,Y ∈ ℑ10 (M), where (ω ◦ SX1 )Y = ω(S(X1,Y )), (ω ◦ SX2 )Y =
ω(S(Y, X2)).
∗
Let now ω ∈ S ℑ0s (M), s > 1. As in Section 1.6.3, by similar devices we have
(ψS ω)(X1, X2 ,Y1 , ...,Ys) = (ψS(X1 ,X2 ) ω)(Y1 , ...,Ys)

= S(X1 , X2 )(ω(Y1 , ...,Ys)) − X1 (ω(S(Y1 , X2 ),Y2 , ...,Ys))

−X2 (ω(S(X1 ,Y1 ),Y2 , ...,Ys)) − (ω ◦ S)([X1, X2 ],Y1 , ...,Ys)
s
− ∑ ω(Y1 , ..., ∇S(X1 ,X2 )Yλ − S(∇X1 Yλ , X2 ) − S(X1 , ∇X2 Yλ ), ...,Ys)
λ=1

= (∇S(X1 ,X2 ) ω)(Y1 , ...,Ys) − X1 (ω(S(Y1 , X2 ),Y2 , ...,Ys))

−X2 (ω(S(X1 ,Y1 ),Y2 , ...,Ys)) − (ω ◦ S)([X1, X2 ],Y1 , ...,Ys)
s
+ ∑ ω(Y1 , ..., S(∇X1 Yλ , X2 ) + S(X1 , ∇X2 Yλ ), ...,Ys)
λ=1
= (∇S(X1 ,X2 ) ω − ∇X1 (ω ◦ SX2 ) − ∇X2 (ω ◦ SX1 ))(Y1, ...,Ys)
−(ω ◦ S)([X1 , X2 ],Y1, ...,Ys).

Thus

(ψS ω)(X1, X2 ,Y1 , ...,Ys) = (∇S(X1 ,X2 ) ω − ∇X1 (ω ◦ SX2 ) −

−∇X2 (ω ◦ SX1 ))(Y1, ...,Ys)
−(ω ◦ S)([X1, X2 ],Y1, ...,Ys).

The connection ∇ on M is a called a S−connection, if ∇S = 0.

Theorem 21. If ∇ is a torsion-free S−connection, then φS(X1 ,X2 )t = ψS(X1 ,X2 )t

∗ ∗
for any t ∈ S ℑ1s (M) and t ∈ S ℑ0s (M).

Proof. Let T (X,Y ) = ∇X Y − ∇Y X − [X,Y ] = 0. Then [X,Y ] = ∇X Y − ∇Y X.

On the other hand, from (d´´) of Definition 8 and (d´) of Definition 3, we obtain

φS(X1 ,X2 )Y = −(LY S(X1 , X2 ))

= [S(X1 , X2 ),Y ] − S([X1 ,Y ], X2 ) − S(X1 , [X2,Y ])
= ∇S(X1 ,X2 )Y − ∇Y S(X1 , X2 ) − S(∇X1 Y − ∇Y X1 , X2 )
−S(X1 , ∇X2 Y − ∇Y X2 )
= ∇S(X1 ,X2 )Y − S(∇X1 Y, X2 ) − S(X1 , ∇X2 Y )
= ψS(X1 ,X2 )Y.

1.8. Generalizations
The definition of the Yano-Ako operator ( φS −operator) can be extended to S of
type (1, q) as follows:

Let S be a tensor field of type (1, q), q > 0. As in Section 1.6.1, we denote
∗ ∞ ∗ C
by (S ℑ(M) = ∑ S ℑr (M), ⊗)
s an algebra over R of all pure tensor fields of type
r,s=0
(r, s) with respect to S.
∗
Definition 9. A map φS |r+s>0 : S ℑ(M) → ℑ(M) is called a generalized Yano-
Ako operator or a generalized φS -operator, if
(a) φS is a R- lineer, i.e. φS (at1 + bt2 ) = aφS (t1 ) + bφS (t2 ) for any a, b ∈ R
∗
and t1 ,t2 ∈ S ℑrs (M),
∗
(b) φS : S ℑrs (M) → ℑrs+q (M) for all r and s,
C C C ∗
(c) φS (K ⊗ L) = (φS K) ⊗ L + K ⊗ (φS L) for all K, L ∈ S ℑ(M),
(d) φS(X1 ,...,Xq )Y = −(LY S)(X1 , ..., Xq) for all Y, X1, ..., Xq ∈ ℑ10 (M),
(e) (φS(X1 ,...,Xq ) ω)Y = (d(ιY ω))(S(X1, ..., Xq)) − q(d(ιY (ω ◦ S)))(X1, ..., Xq)
+ω((LY S)(X1, ..., Xq)) for all X1 , ..., Xq,Y ∈ ℑ10 (M), ω ∈ ℑ01 (M), where
C
ιY ω = ω(Y ) = ω ⊗ Y, S is a vector-valued q−form, d is the exterior differ-
entiation of ιY (ω ◦ S) ∈ Λq−1 (M).
∗ ∗
In particular, if t ∈ S ℑ1s (M) and ω ∈ S ℑ0s (M), then as in Section 1.6.3, by
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

similar devices we have

(φS t)(X1, ..., Xq,Y1 , ...,Ys) = (−Lt(Y1 ,...,Ys ) S)(X1, ..., Xq)
s
+ ∑ t(Y1, ..., (LYλ S)(X1 , ..., Xq), ...,Ys)
λ=1

and
(φS ω)(X1 , ..., Xq,Y1 , ...,Ys) = (S(X1 , ..., Xq))(ω(Y1 , ...,Ys))
−q(d(ιY1 (ωY2 ,...,Ys ◦ S)))(X1, ..., Xq)
s
+ ∑ ω(Y1 , ..., (LYλ S)(X1 , ..., Xq), ...,Ys),
λ=1

where ωY2 ,...,Ys ∈ ℑ01 (M) for fixed Y2 , ...,Ys ∈ ℑ10 (M) and ωY2 ,...,Ys ◦ S ∈ Λq (M).
If we replace the condition (d) of Definition 9 by
q
ψS(X1 ,...,Xq )Y = ∇S(X1 ,...,Xq )Y − ∑ S(X1 , ..., ∇Xλ Y, ..., Xq),
λ=1

we have a generalized ψS −operator.

Let now q = 0, i.e. S = X ∈ ℑ10 (M). Then from (d) of Definition 9 we have
φX Y = −LY X = LX Y for all X,Y ∈ ℑ10 (M), i.e. φX is the Lie derivative with
respect to X, i.e. φX = LX .
Analogously, we easily see that ψX is the covariant derivative with respect
to X, i.e. ψX = ∇X .
Thus the operators φS and ψS , S ∈ ℑ1q (M), q ≥ 0 are generalizations of LX
and ∇X , respectively.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Algebraic Structures on
Manifolds

In this chapter we give the fundamental results and some theorems concerning
geometry of holomorphic hypercomplex manifolds which will be needed for the
later treatment of special types of hypercomplex manifolds. In Section 2.1, we
give the basic definitions and theorems on hypercomplex algebras. Section 2.2
is devoted to the study of hypercomplex structures on manifolds. We introduce
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

the regular hypercomplex structure on manifolds. Section 2.3 treats manifolds

with integrable regular hypercomplex structures. We show that such a manifold
is a real model of a holomorphic manifold over hypercomplex algebras. Section
2.4 is devoted to the study of pure tensor fields. We compute the components
for the pure tensor field with respect to the regular hupercomplex structures and
we show that the pure tensor fields are a real models of hypercomplex tensors.
In Section 2.5, we discuss holomorphic hypercomplex tensor fields. Using the
Tachibana operator we give the condition of holomorphic hypercomplex tensors
fields in real coordinate systems. In Section 2.6, we consider pure connections
which are real model of the hypercomplex connections. Section 2.7 is devoted
to the study of pure hypercomplex torsion tensors. In Section 2.8, we give a real
model of holomorphic hypercomplex connections by using the tensor operator
applied to pure torsion-free connection. In the last Section 2.9, we consider
some properties of pure curvature tensors. The main theorem of this section is
that the curvature tensor of holomorphic connection is holomorphic. Finally,
we also consider a holomorphic manifold of hypercomplex dimension 1 and we
show that the hypercomplex connection on such manifold is holomorphic if and

only if the real manifold is locally flat.

2.1. Algebraic Theory

In this section, we present some basic algebraic notions and theorems.

2.1.1. Associative Algebras

We consider an m−dimensional associative algebra Am over R ( hypercomplex
γ
algebra [36]) with basis {eα }, α = 1, ..., m and structural constants Cαβ :
γ
eα · eβ = Cαβ eγ .
γ
We note that Cαβ are components of the (1, 2)−tensor · : Am × Am → Am . In
this work we assume Am is a unitial algebra, i.e., it admits the principal unit
e1 = 1. We introduce the matrices
γ γ
Cα = (Cαβ ), Ceα = (Cβα ), (2.1)

where γ and β are the number of the row and column, respectively. Then the as-
τ Cσ = Cσ Cτ ) can be written
sociativity condition eα .(eβ .eγ) = (eα .eβ ).eγ (Cασ
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

βγ αβ σγ
in one of the following three equivalent forms:

σ
CαCβ = Cαβ Cσ , (2.2)

eβ = Cγ C´
eαC´ e
C´ αβ γ , (2.3)

CαCeβ = CeβCα , (2.4)

eα is the transpose of Ceα . From e1 = 1 = εσ eσ it follows that
where C´

εσCσ = εσCeσ = I = (δβα ), (2.5)

where δβα is the Kronecker delta.

e
Let C(A) and C´(A) eα , respectively
be the algebras are generated by Cα and C´
(see (2.2) and (2.3)). A mapping ρ1 : Am → C(A) :

Am 3 a = aσ eσ → aσCσ = C(a) ∈ C(A), a1 , ..., am ∈ R

is called a regular representation of type I. The representation ρ1 is an isomor-

e
phism. In a similar way, we introduce a regular representation ρ2 : Am → C´(A)
of type II:
eσ = C´(a)
Am 3 a = aσ eσ → aσC´ e ∈ C´(A),
e a1 , ..., am ∈ R

which is also isomorphism. In this paper the representations appearing in the

discussion will be supposed to be representations of type I. Notice that a regular
representation of type I is usually called simply a regular representation. From
(2.4) we see that all A = aσCeσ , a1 , ..., am ∈ R belongs to the commutator algebra
(center) of the algebra C(A). Conversely, using (2.5) we easily see that if CA =
AC for any C ∈ C(A), then A = aσCeσ . In fact, we put C = Cα , α = 1, ..., m , then
γ σ γ
we have AσγCαβ = Cαγ Aβ . Contracting this equation with εβ , we find

γ γ
AσγCαβ εβ = Cαγ
σ
A β εβ ,
Aσγδγα = Cαγ
σ γ
a,
A = (Aσα ) = aγ(Cαγ
σ
) = aγCeγ ,
γ e
where aγ = Aβ εβ , A ∈ C(A).Thus we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 22. Let A be a matrix of (m × m)-matrices. Then ACα = Cα A for all

α if and only if A = aαCeα.

By similar devices, we have

eα = C´
Theorem 23. Let A be a matrix of (m × m)-matrices. Then AC´ eα A for all
α eα .
α if and only if A = a C´

2.1.2. Commutative Algebras

We restrict ourselves to the consideration of commutative hypercomplex alge-
bras. In this and next sections, we always assume that the Am is a commutative
hypercomplex algebra. The commutativity condition eα · eβ = eβ · eα can be
written in one of the following equivalent forms:
γ γ
Cαβ = Cβα (Cα = Ceα ), (2.6)
γσ σ γ
CασCβδ = CβσCαδ (CαCβ = CβCα ). (2.7)

Since Cα = Ceα , then we have C´(A)

e = C´(A). Therefore, C´(A) is called a trans-
pose regular representations of commutative algebras. From Theorems 22 and
23 we have (see [36])

Theorem 24. Let Am be a commutative hypercomplex algebra. Then C(A) and

C´(A) are maximal commutative subalgebras of the algebra of (m×m)-matrices.

Among of commutative algebras, a special role is played by Frobenius al-

gebras for which there exist constants λγ such that
γ
ϕαβ = Cαβ λγ (2.8)

form a nonsingular symmetric matrix and can be taken as components of the

metric tensor of the Frobenius metric. Then, along with the basis {eα }, we can
introduce the dual basis {eα }, where eα = ϕαβ eβ . For the Frobenius metric, we
have relations
σ σ β
ϕασCγβ = ϕβσCγα , ϕασCγσ = ϕβσCγσ α
, (2.9)
β
α γ α β
eα · eβ = Cβγ e , e · e = ϕασCγσ eγ , ϕαβ εβ = λα , (2.10)

where εβ are components of e1 = 1.

An automorphism ψ : Am −→ Am is called a conjugation of Am, if ψ is a
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

involution, i.e ψ2 = id. From definition of the conjugation ψ : x −→ x ∈ Am we

have
β γ
eα = ψ(eα ) = ψα eα , ψαγ ψβ = δαβ , (2.11)
γ β γ
Cαβ ψασ ψτ = Cστ
τ
ψτ , (2.12)
β
ψα εα = εβ , (2.13)
where e1 = 1 = εβ eβ . A conjugate to a ∈ Am is given by

a = aα eα −→ a = aα eα = aα ψσα eσ .

From (2.11) follows that, there exists an adapted basis of Am such that ψαβ =
±δαβ . We set
( β
α
ψαβ 1 , i f ψα = δαβ ,
ψβ = 1
β
ψαβ 2 , i f ψα = −δαβ ,
2

where 1 ≤ αi < α, 1 ≤ βi < β, i = 1, 2 . Then from (2.11), we have

eα1 = eα1 , eα2 = −eα2 .

Since ψ(e1 ) = e1 (see (2.13)), we have e1 ∈ {eα1 }, where {eα1 } denotes the
plane spaned by eα1, 1 ≤ α1 < α. In particular, if A2 = C(m = 2) is the complex
1 0
algebra then, ψαβ = and e1 = e1 = 1, e2 = −e2 = −i, i2 = −1. Tak-
0 −1
ing account of (2.14) and writing expression (2.12), for the different indices, we
find
γ γ γ γ
Cα21 β = 0, Cα22 β = 0,Cα11 β = Cβ1 α1 = 0
1 2 2 2

with respect to the adapted basis.

2.1.3. Holomorphic Functions

Let z = xαeα be a variable in Am , where xα (α = 1, ..., m) are real variables.Using
a real-valued C∞ −functions yβ (x) = yβ (x1 , ..., xm), β = 1, ..., m, we introduce a
hypercomplex function
w = yβ (x)eβ
of variable z ∈ Am. Let dz = dxα eα and dw = dyα eα be respectively the differ-
entials of z and w(z). We shall say that the function w = w(z) is a holomorphic
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

function if there exists a functions w´(z) such that

dw = w´(z)dz. (2.14)

We shall call w´(z) the derivative of w(z).

Theorem 25. [89], [36], [107, p.87] The hypercomplex function w = w(z) is
holomorphic if and only if the Scheffers conditions hold:

Cα D = DCα . (2.15)
α
where D = ( ∂y
∂xβ
) is the Jacobian matrix of yα (x), Cα is the matrix defined by
(2.1).

e α eα . Then
Proof. Let w = w(z) be a holomorphic function. We put w´(z) = w
from (2.11) we have
∂yα β γ
dw =dyα eα = e α eα dxβ eβ = w
dx eα = w e α dxβCαβ eγ .
∂xβ

From here, we obtain

∂yα γ
=w e αCαβ . (2.16)
∂xβ
α w = w(z) is a holomorphic function if and
Thus, the hypercomplex function
only if the Jacobian matrix ∂y
∂xβ
have the form (2.16). Contracting (2.16) with
σ σ
ε (1 = ε eσ ) and using (2.5), we find
∂yγ
e γ = εβ
w ,
∂xβ
i.e.
∂yγ
w´(z) =εβ
eγ . (2.17)
∂xβ
Now applying Theorem 22 to (2.16), we see that the condition (2.16) is equiva-
lent to the Scheffers conditions. From (2.17)
follows
that the Jacobian matrix
γ of
∂yγ β ∂2 yγ ∂y
εβ ∂x β has components of the form D´= ε ∂xβ ∂xα
= ε β
∂β D, where D = ∂xβ
,
and the Jacobian matrix D´ = εβ ∂β D also satisfying the Scheffers conditions.
From here we easily see that exists the successive derivatives w´´(z), w´´(z), ... of
w(z). We note that, in particular, if A2 = C , where C is the complex algebra,
then the Scheffers conditions (2.15) reduces to the Cauchy-Riemann conditions.
In fact, by virtue of
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1 1
1 1

C11 C12 1 0 C21 C22 0 −1
C1 = 2 2 = , C2 = 2 2 =
C11 C12 0 1 C21 C22 1 0
from (2.15) we have
∂y1 ∂y2 ∂y2 ∂y1
= , = − ,
∂x1 ∂x2 ∂x1 ∂x2
where z = x1 + ix2 , w = y1 (x1 , x2 ) + iy2 (x1 , x2 ), i2 = −1.

Remark 9. It is well known that the concepts of holomorphic and analytic com-
plex functions are equivalent. The hypercomplex function w = w(z) is said to be
analytic if w(z) admits a convergent power series. In general, for hypercomplex
functions the concepts of holomorphic and analytic functions are not equivalent
(see [107, p.88], [11]).

The concept the holomorphic hypercomplex functions can be immediately

extended to the case of several algebraic variables. Let
zu = x(u−1)m+αeα , (u = 1, ..., r)

be an variables in Am. In fact the function w(z1 , ..., zr) = yβ (x1 , ..., xrm)eβ is a
holomorphic function of the variables
z1 , ..., zr if and only if Cα Du = DuCα for
∂y α
any u, where Du = ∂x(u−1)m+β , u = 1, ..., r.
The holomorphic hypercomplex functions possess the following properties:
(i) (w1 + w2 )´= w1´+ w2´if w1 and w2 are holomorphic functions,
(ii) (w1 w2 )´= w1´w2 + w1 w2´if w1 and w2 are holomorphic functions,
(iii) If F and w are holomorphic functions of w and z respectively then F =
F(w(z)) is a holomorphic function of z and Fz´= dF dF dw
dw w´, where Fz´= dz , w´= dz .
γ γ β
Proof. (ii) We put w1 w2 = wα1 w2 eα eγ = wα1 w2Cαγ eβ = w = wβ eβ . Then from
(2.17) we have
β γ
(w1 w2 )´ = εσ (∂σ wβ )eβ = εσCαγ (∂σ (wα1 w2 ))eβ
β γ γ
= εσCαγ ((∂σ wα1 )w2 + wα1 (∂σ w2 ))eβ
γ γ
= ((εσ ∂σ wα1 )w2 + wα1 (εσ ∂σ w2 ))eα eγ
= w1´w2 + w1 w2´
proof of (i) is analogous to (ii).
(iii) We put F = Fα eα , w = yβ (x)eβ , z = xγ eγ. If dF
dw = Feα eα , w´= w
eα eα , then
by virtue of (2.17), we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

dF ∂F β
Fz´ = = εα α eβ
dz ∂x
β γ
∂F ∂y β θ γ
= εα γ α eβ = εα FeσCσγ w e Cθα eβ
∂y ∂x
β θ γ
= FeσCσγ we δθ eβ = Feσ w
eγ eσ eγ
dF dw dF
= = w´.
dw dz dw

Example 4. Let A2 = R(ε) be a dual algebra with canonical bases {1, ε}, ε2 =
0. Using 1
C21 C22 1 0 0
C2 = 2 2 =
C21 C22 1 0
we see that, the condition (2.12) reduces to the following equations:
∂y1 ∂y2 ∂y1
= , = 0,
∂x1 ∂x2 ∂x2

where z = x1 +εx2 , w(z) = y1 (x1 , x2 )+εy2 (x1 , x2 ), ε2 = 0. From last equations

we have
w(z) = f(x1 ) + ε(x2 f´(x1 ) + g(x1 )).
We call w = w(z) in this form the synectic function [99], [16]. If g(x1 ) = 0,
then the function w(z) = f(x1 ) + εx2 f´(x1 ) is said to be natural extension of the
real C∞ - functions f(x1 ) to R(ε).

Example 5. Let A2 = A(e), e2 = 1 be an algebra of paracomplex numbers.

From
0 1
C2 =
1 0
we easily see that, the Scheffers conditions reduces to the para-Cauchy-Riemann
∂y1 ∂y2 ∂y2 ∂y1
conditions (see [11]) ∂x 1 = ∂x2 , ∂x1 = ∂x2 ,where

z = x1 + ex2 , w(z) = y1 (x1 , x2 ) + ey2 (x1 , x2 ), e2 = 1.

2.2. Algebraic Π−structures on Manifolds

If a collection of (1, 1)-tensor (affinor or endomorphism) fields ϕ, ϕ, ... are given
1 2
on a C∞ -manifold M, then one says that a polyaffinor structure (or Π−structure)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

is given on M : Π = {ϕ}. This structure is said to be rigid [36], if every ϕ ∈ Π

is reduced to the constant form in a certain (in general, nonholonomic) frame
{Xi }, i = 1, ..., n in a neigborhood Ux of every point x ∈ M. In this case, we shall
say that {Xi }, i = 1, ..., n is adapted frame with respect to the Π−structure. If
every ϕ ∈ Π is constant on a certain set of holonomic(natural) adapted frames
{Xi } = {∂i }, i = 1, ..., n of a smooth atlas, then the Π−structure is said to be
integrable. Obviously, the integrable Π−structure is always rigid: the converse
holds only under certain additional requirements on the Π−structure. For ex-
ample, if Π = ϕ, i.e. Π−structure consists of one affinor, and if there exists
a torsion-free ϕ-connection ∇ on M preserving the rigid ϕ−structure, i.e. the
realiton ∇ϕ = 0 holds, then the ϕ−structure is integrable [92], [36]. It is well
known that, for simplest rigid ϕ−structures (almost complex and almost para-
complex structures et al), the integrability is equivalent to the vanishing of the
Nijenhuis tensor (see Chapter 1).

Definition 10. Let ∇ be a linear connection on M. ∇ is called Π−connection

with respect to the Π−structure, if ∇ϕ = 0 for any ϕ ∈ Π.

Definition 11. Π−structure on M is called almost integrable, if there exists a

torsion-free Π−connection.

We note that for some simplest Π−structures (ϕ−structures, regular

Π−structures at al) the concepts of integrability and almost integrability are
equivalent. Let Am be a hypercomplex algebra. An almost hypercomplex struc-
ture on M is a polyaffinor Π−structure such that
γ
ϕ · ϕ = Cαβ ϕ, (2.18)
α β γ

i.e. if there exists an isomorphism Am ↔ Π, where ϕ are structural affinors

α
corresponding to the bases elements eα ∈ Am , α = 1, ..., m.

Definition 12. [36] An almost hypercomplex structure on M is said to be a

regular (or r-regular) Π−structure if matrices of ϕ of order n, α = 1, ..., m si-
α
multaneously reduced to the form
 
! Cα 0 · · · 0
 0 Cα · · · 0 
 
ϕij =  . .. ..  , α = 1, ..., m; i, j = 1, ..., n (2.19)
α  .. . ··· . 
0 0 ··· Cα
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

γ
with respect to the adapted frame {Xi }, where Cα = (Cαβ ) is the regular repre-
sentation of Am, r is a number of Cα −blocks.

Remark 10. Putting C´α in (2.19), we have a transpose regular Π−structure on

From definition immediately follows that the regular Π−structures are rigid
structures. In particular, for almost complex and paracomplex structures the
condition (2.19) immediately follows from (2.18), i.e. almost complex and para-
complex structures on M (dimM = 2r), automatically are regular structures. For
example, a Π−structure Π = {I, ϕ}, I = idM , ϕ2 = 0 on M is an isomorphic rep-
resentation of dual algebra R(ε), ε2 = 0 but it is nonregular Π−structure on M,
in general (see [102],
[103], [108]).
Let now Π = ϕ , α = 1, ..., m be a regular Π−structure on M. Then from
α
(2.12), we have
n = mr (n = dim M , m = dim Am), (2.20)

where r is a number of the bloks Cα . Thus, the condition (2.20) is a necessary

condition for exists of regular Π−structures on M and in this case we putting

i = (u − 1)m + α (i = 1, ..., n; u = 1, ..., r; α = 1, ..., m)

or
i = uα, j = vβ, k = wγ, ...
In other words, the structural affinors ϕ have the coordinates
α

u u γ
ϕij = ϕuα
vβ = δvCαβ (δv − Kronecker delta). (2.21)
α α

Let Xi´ = Sii´Xi (Det(Sii´) 6= 0) be a transformations of adapted frame {Xi } with

respect to the regular Π−structure. Then we have
! !
j
ϕi´j´ = (Sii´) ϕij (S j´), (2.22)
α α

where (Sii´) = (Sii´)−1 . If {Xi } is an adapted frame, then we call transformation

Xi → Xi´ the admissible transformation Xi → Xi´. Since in this case {Xi´} is adapted
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

! !
frame then ϕi´j´ = ϕij , and we have from (2.22)
α α

Sϕ = ϕS, (2.23)
α α
!
j
where S = (S j´) and ϕ = ϕij . Thus we have
α α

Theorem 26. Let Π = ϕ be a regular Π−structures on M. A transformation
α
S : Xi → Xi´ of adapted frames is an admissible if and only if it satisfies the
condition (2.23).

Using Theorem 22 and Theorem 23 we see that the matrix S has the special
structure
α
Sii´ = ∆uσ
u´ Cσα´ (i = uα, i´= u´α´). (2.24)

For inverse matrix, by similar devices, we have

α´
Sii´ = ∆u´σ
u Cσα (i = uα, i´= u´α´). (2.25)

We can associate matrices from the algebra Am :

∗ ∗ ∗ ∗
S = (Suu´) = (∆uσ u´ u´σ
u´ eσ ), S = (Su ) = (∆u eσ ). (2.26)

∗ ∗ −1 ∗
From here, we easily see that SS = I, where
   
e1 0 · · · 0 1 0 ··· 0
∗  0 e1 · · · 0   0 1 ··· 0 
    i
I = . .. . . .  =  .. .. . . ..  = (δ j ).
 .. . . ..   . . . . 
0 0 · · · e1 0 0 ··· 1

In fact, from (2.24) and (2.25) we obtain

α u´ε α´
δij = Sii´Si´j = ∆uσ
u´ Cσα´∆v Cεβ
γ
= ∆uσ u´ε uσ u´ε
u´ ∆v Cσε eγ = ∆u´ eσ ∆v eε
∗ ∗
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

= Suu´Su´v.

With each vector field ξ = ξi Xi = ξuα Xuα, where {Xi } is the adapted frame on
∗
M, we can associate r coordinates ξu (u = 1, ..., r) from the algebra Am :
∗
ξu = ξuαeα .
∗ ∗ ∗
i´
We easily see that, if ξ = Sii´ξi , u´
then ξ = Su´uξu . In fact, from (2.21) we obtain

ξu´α´ = Su´α´ uα u´σ α´ uα

uα ξ = ∆u Cσα ξ

or
∗
u´
ξ = ξu´α´eα´ = ∆u´σ α´ uα
u Cσα ξ eα´
∗ ∗
uα u´ u
= ∆u´σ
u eσ ξ eα = Su ξ .

Let now ϕ ∈ Π. Then ϕ = aα ϕ. The action of ϕ on a vector field ξ = ξi Xi , i.e the

α
equation ηi = ϕij ξ j reduces to

∗
ηu = ηuαeα = aσ ϕij ξ j eα = aσ δuvCσβ
α vβ
ξ eα
σ
∗
= aσ δuv ξvβ eσ eβ = aσ eσ ξuβ eβ = aξu ,

where i = uα, j = vβ, a ∈ Am. Thus we have

Theorem 27. If Π is a regular structure on M then, each tangent space

Tx (M), x ∈ M serves as a real model of the module Tr (Am) over algebra Am.

In particular, in the case ϕ = ϕ we have

∗ ∗
ηu = eα ξu .

2.3. Integrable Regular Π−structure

Let Π = ϕ be an integrable regular Π−structures on Mmr and let xi = xuα
α
and xi´ = xu´α´- adapted local coordinates in Ux (x ∈ Mmr ). It is well known that the
structural affinors ϕ have the constant form (2.19) with respect to the adapted
n o αn o
frames ∂x∂ i and ∂x∂ i´ . In this case the admissible transformation has the
∂ ∂xi
form = Sii´ ∂x∂ i , where Sii´ =
Sii´ = ∆uσ α
,i´
u´ Cσα´, Si = ∆
u´σ α´
C , i = uα, i´= u´α´(see
∂xi´ ∂xi´ u uασα uα
∂x ∂x
(2.21) and (2.24)). Then from Theorem 26 we have ∂x u´α´ Cσ = Cσ ∂xu´α´
for
fixed u and u´, i.e. zu´ = xu´α´eα´ is holomorphic function of zu = xuα eα . Using
(2.17) and (2.26), we have
∗
∂zu´ α ∂x
u´α´
α u´σ α´ u´σ α´ u´σ u´
= ε eα´ = ε ∆ C
u ασ eα´ = ∆ u σδ eα´ = ∆ u eσ = S u.
∂zu ∂xuα
By similar devices, we have
∂zu ∗
u´
= Suu´.
∂z

Thus, on the intersection of two coordinate neigbourhoods with local adapted

coordinates xuα and xu´α´, the transition functions zu´ = zu´(zu ) (zu = xuαeα , zu´ =
xu´α´eα´) are holomorphic, i.e. Mmr is holomorphic A-manifold of dimension
r : Xr (A). Conversely let Xr (A) be a holomorphic A-manifold. Then on the in-
tersection of two A-coordinate neigborhoods the transition functions zu´ = zu´(zu )
are holomorphic. Then by virtue of Theorem 25 we have
uα uα
∂x ∂x
u´α´
Cσ = Cσ
∂x ∂xu´α´

for fixed u and u´. From here, we have

uα uα
∂x ∂x
u´α´
ϕ=ϕ u´α´
∂x σ σ ∂x

u γ
for any u and u´, where ϕij = ϕuα
vβ = δ C
v σβ , from which, it follows that Π = ϕ
σ σ α
is integrable and regular Π−structure.Thus we have

Xr (A) is a real mani-

Theorem 28. A real model of a A-holomorphic manifold
fold Mmr with integrable regular Π−structure Π = ϕ .
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Remark 11. Let Mmr be a real manifold with regular Π−structure. Then the
following properties are equivalent [35]:
(i) Regular Π−structure is integrable,
(ii) Regular Π−structure is almost integrable,
(iii) NΠ = 0, where NΠ denotes the Nijenhuis-Shirokov tensor determined by
regular Π−structure.

Example 6. Let Xr (C) be a complex analytic (C-holomorphic) manifold [112,

p.50]. To show that every complex analytic manifold carries a natural almost
complex structure on the real model M2r of Xr (C), we consider the space Cr of
r−tuples of complex numbers (z1 , z2 , ..., zr) with zi = xi + iyi , i2 = −1, i =1, ..., r
. With respect to the real coordinate system (x1 , ..., xr, y1 , ..., yr) we define an
almost complex structure ϕ on R2r (real model of Cr )

∂ ∂ ∂ ∂
ϕ i
= i, ϕ i
= − i , i = 1, ..., r. (2.27)
∂x ∂y ∂y ∂x

Since the transition functions of two charts of Xr (C) are C-holomorphic, there-
fore to define an almost complex structures on M2r , we transfer the almost
complex structure of R2r in the form (2.27) to M2r by means of such charts.
From the construction of the almost complex structure ϕ given by (2.27), it is
clear that the components of ϕ with respect to the local (holonomic) coordinate
x1 , ..., xr, y1 , ..., yr system are constant and hence ϕ is integrable. On the other
hand, every almost complex structure n is regular. In fact, for ϕ there exist o ele-
∂ ∂ r ∂ ∂ ∂ ∂ ∂ ∂
ments ∂x1 , ..., ∂xr of C such that ∂x1 , ∂x2 , ..., ∂xr , ϕ( ∂x1 ), ϕ( ∂x2 ), ..., ϕ( ∂xr ) is
a basic for R2r (real model of Cr ) (see [34, p.141]). If we put

∂ ∂ ∂ ∂ ∂ ∂
, ϕ( 1 ), 2 , ϕ( 2 ), ..., r , ϕ( r ) ,
∂x1 ∂x ∂x ∂x ∂x ∂x
then we see that ϕ is given by the matrix
 
0 −1
 ... 0 
1 0
(ϕ) = 
,

0 −1
0...
1 0

i.e. ϕ is regular structure. Thus a real model of complex analytic manifold is a

real manifold with integrable regular natural almost complex structure given by
(2.27).

Example 7. By similar devices, we easily see that a real model of paracomplex

(paraholomorphic) manifold Xr (A(e)) (see [11] ) is a real manifold M2r with
integrable regular natural almost para-complex structure ψ given by

∂ ∂ ∂ ∂
ψ i
= i , ψ i
= i , i = 1, ..., r,
∂x ∂y ∂y ∂x

where A(e) = {zi = xi + eyi , e2 = 1}.

Example 8. Let T (Mn ) be a tangent bundle of Mn (see [114] for details). The
tangent bundle of Mn consist of pair (x, y), where x ∈ Mn and y ∈ Tx (Mn ). Let
π : T (Mn ) → Mn defined by π(x, y) = x be the natural projection of T (Mn ) onto
Mn . Let (U, x = (x1 , ..., xn)) be a coordinate chart on Mn . Then it induces
local coordinates (x1 , ..., xn, x1 , ..., xn) on π−1 (U), where x1 , ..., xn represent
n the
o
components of vector fields on Mn with respect to local frame {∂i } = ∂xii .

In the following we use the notation i = i + n for all i = 1, ..., n. If (U´, x´ =

(x1´, ..., xn´)) is another coordinate chart on Mn , then the induced coordinates
(x1´, ..., xn´, x1´, ..., xn´) with respect π−1 (U´) will be given by
(
xi´ = xi´(xi ), i = 1, ..., n,
i´ ∂xi´ i (2.28)
x = ∂xi x , i = n + 1, ...2n.

The Jacobian of (2.28) is given by matrix

!
∂xi´
∂xα´ ∂xi
0
S= = 2 xi´ ∂xi´
, α = 1, ..., 2n.
∂xα xs ∂x∂ i ∂x s ∂xi

It is clear that there exist a natural affinor field

0 0
ϕ = ϕαβ = (I − identity matrix of degree n)
I 0

such that
Sϕ = ϕS,
i.e. the transformation S : {∂α } → {∂α´} is an admissible with respect to the
structure ϕ. It is clear that ϕ is integrable and ϕ2 = 0. Also ϕ is a regular
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

persentation of the algebra of dual numbers R(ε), ε2 = 0. In fact, if we put

{∂1 , ∂1 , ∂2 , ∂2 , ..., ∂n, ∂n } instead of {∂1 , ∂2 , ..., ∂n, ∂1 , ∂2 , ..., ∂n}, then the affinor
field ϕ has the matrix
 
0 0
 ... 0 
1 0
ϕ= 


0 0
0...
1 0

with respect to the frame {∂1 , ∂1 , ∂2 , ∂2 , ..., ∂n , ∂n }, i.e. ϕ is a regular. Thus, on

the tangent bundle T (Mn ) there exist a natural integrable affinor ϕ-structure
[100], [94] which is a regular presentation of algebra of dual numbers R(ε)
(we note that an isomorphic representation of R(ε) is nonregular (see [106]), in
general). Therefore with each induced coordinates (xi , xi) in π−1 (U) ⊂ T (Mn ),
we associate the local dual coordinates X i = xi + εxi , ε2 = 0. Using (2.28) we
see that the local dual coordinates X i = xi + εxi transformed by

X i´ = xi´(xi) + εxs ∂s (xi´(xi)). (2.29)

The equation (2.29) show that the quantities X i are R(ε)-holomorphic functions
of X i = xi + εxi (see, Example 9). Thus the tangent bundle T (Mn ) with a natu-
ral integrable regular ϕ-structure is real model of R(ε)-holomorphic manifold
Xn (R(ε)).

2.4. Pure Tensors with Respect to the Regular Structure

In particular, being applied to a (1, 1)-tensor field t, the purity with respect to the
regular Π−structure means that in the local coordinates the following conditions
should hold:
t mj ϕim = tmi ϕmj for any ϕ ∈ Π. (2.30)
We putting i = uα, j = vβ, m = wσ and ϕ = (δuvCγβ
α
) instead of ϕ ∈ Π. Then
γ γ
from (2.30), we have
wσ u α uα w σ
tvβ δwCγσ = twσ δv Cγβ , (2.31)
uσ α uα σ
tvβ Cγσ = tvσ Cγβ .

Using contraction with εβ (1 = εβ eβ ) and (2.5), from (2.31) we have

uσ β α uα σ β uα σ uα
tvβ ε Cγσ = tvσ Cγβ ε = tvσ δγ = tvγ

or
σ
uα α
t ij = tvσ = ℑuvCσβ , (2.32)
σ ∗
uσ β
where ℑuv = tvβ ε . Thus, a pure tensor field t ∈ ℑ11 (M) has the form (2.32).
Conversely, from (2.32) it follows that the tensor field t of type (1, 1) is pure if
ε
ℑ are an arbitrary functions. In fact, substituting (2.32) into (2.30), we find
ε ε
σ u α u α w σ
ℑw
v Cεβ δwCγσ = ℑwCεσ δv Cγβ , (2.33)
ε
σ α α σ
ℑuv (CεβCγσ −CεσCγβ ) = 0.
σ α α σ
Since Am is a commutative algebra CεβCγσ = Cεσ Cγβ , we see that the equation
ε
(2.30) is satisfies for arbitrary functions ℑ . Thus, the tensor field t of type (0, 2)
is pure with respect to the regular Π− structure if and only if t has form (2.32)

∗
for arbitrary functions ℑ..σ . In the case G ∈ ℑ20 (M) the situation is more difficult.
The purity condition of G is given by

Gm j ϕim = Gim ϕmj for any ϕ ∈ Π.

By similar way, we have

α β
GuσvβCγσ = GuαvσCγσ . (2.34)

After contraction of (2.34) with λα , we have

σ σ
β
ϕγσ Guσvβ = ℑuvCγσ (ℑuv = Guαvσ λα ), (2.35)
γ
where ϕαβ = λγCαβ . If Det(ϕαβ ) 6= 0, i.e. if Am is a Frobenius algebra, then
from (2.35) we have following solition
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

σ
α µβ
Gi j = Guαvβ = ℑuvCσµ ϕ . (2.36)

Conversely, from (2.36) by virtue of (2.7) it follows that the tensor field G ∈
∗ σ ∗
ℑ20 (M) is pure if ℑ are an arbitrary functions. Thus, in the case G ∈ ℑ20 (M),
algebra Am must be Frobenius algebra. In general, the case for t ∈ ℑrs (M), in
the space of Am we consider the Kruchkovich tensors [36]:
α
Bαβ1 β2 ...βs = Cβα1 α1 Cβα1α2 ...Cβ s−2β (s > 2),
2 s−1 s

Bαβ1 β2 = Cβα1 β2 , Bαβ = δαβ .

If Am is the Frobenius algebra with metric ϕαβ , then we have

Bαβ 1...β
...α r
= Bαβ r ...β λ1 ...λr−1 ϕλ1 α1 ...ϕλr−1 αr−1 ,
1 s 1 s

Bα1 ...αr = Bαλ1r ...λr−1 ϕλ1 α1 ...ϕλr−1 αr−1 ,

Bβ1 ...βs = Bαβ β ϕαβs .
1 2 ...β s−1

Now we state some properties of Kruchkovich tensors Bαβ 1...β ...α r

:
1 s
α1 ...αr
(B1 ) Bβ ...β is a symmetric tensor with respect to the indices α1 , ..., αr and
1 s
β1 , ..., βs ,
λ σα1 ...αr
(B2 ) Cσµ Bβ ...β = Bλα 1 ...αr σ α1 ...αr α1 ...α r
µβ1 ...βs , Cλµ Bσβ1 ...βs = Bλµβ1 ...βs ,
1 s
(B3 ) Bσα 1 ...αr α1 ...αr σ α1 ...αr
β1 ...βs λσ = Bσβ1 ...βs ε = Bβ1 ...βs .
Proof of (B1 ), (B2 ) , (B3 ) immediately follows from (2.5), (2.7) and (2.9).
∗
Similarly, pure tensor field t ∈ ℑrs (M) of type (r, s) with respect to the regular
Π− structures has components of the form
σ
r α1 ...αr
t ij11...ir
... js = ℑ uv11 ...u
...vs Bβ ...β (2.37)
1 s

(ia = ua αa, jb = ub αb , a = 1, ..., r, b = 1, ..., s),

σ
where ℑ is an arbitrary functions in the adapted coordinate chart U. We note
that in the cases r = 0 and r = 1, the formulae (2.37) is true, if even the algebra
Am is non-Frobenius algebra.With each pure tensor field (2.37) we can associate
a hypercomplex values from Am :
∗ u ...u σ
t v11 ...vrs = t ij11...ir β1 βs u1 ...ur
... js ε ...ε λα1 ...λαr−1 eαr = ℑ v1 ...vs eσ . (2.38)
∗
u1 ...ur
We easily see that t are components of hypercomplex tensor field of type
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

v1 ...vs
(r, s), i.e.
∗ u´ ...u´ ∗ ∗ ∗ ∗ ∗ u ...u
t v´11 ...v´rs = S u´u11 ...S u´urr S vv´11 ...S vv´ss t v1 ...vs .
1 r

In fact, for simplicity we take r = s = 1 , then from (2.24) and (2.25) we have
vβ
u´σ α´ vε β
t ij´´ = tv´β´
u´α´
= Su´α´ uα uα
uα Sv´β´tvβ = ∆u Cσα ∆v´ Cεβ´tvβ

which implies
∗ u´
t v´ = t ij´´εβ´eα´ = tv´β
u´α´ β´
´ ε eα´
α´ vε uα β´ β
= ∆u´σ
u Cσα ∆v´ Cεβ´tvβ ε eα´
vβ
= ∆u´σ uα
u ∆v´ eσ eα tvβ
ε
u α vβ
= ∆u´σ
u ∆v´ eσ eα ℑv Bεβ
ε
vβ
= ∆u´σ u
u ∆v´ ℑv eσ eβ eε
∗ ∗ ∗u
= S u´uS vv´ t v

by virtue of (2.26).
Thus, we have

Theorem 29. Let Π be an regular integrable Π−structure on Mmr . Then pure

∗
tensor fields t of type (r, s) on Mmr are a real models of hypercomplex tensors t
in the A−holomorphic manifold Xr (Am).

In the next section, we will study a real model of the A−holomorphic con-
∗
ditions of t.

2.5. A-holomorphic Tensors in Real Coordinate

Systems
∗
Let Am be a Frobenius hypercomplex algebra and t ∈ ℑrs (Xr (Am)) be an hyper-
complex tensor field on Xr (Am). Then the real model of such a tensor field t is
∗
a pure tensor field of type (r, s) on Mmr . In general, t is not A−holomorphic.
Using the Tachibana operator we give the condition of A−holomorphic tensors
in real coordinate systems, i.e. the following theorem is true [36], [83]:
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 30. Let on Mmr be given the integrable regular Π−structure. The
∗
hypercomplex tensor field t ∈ ℑrs (Xr (Am)) is A−holomorphic tensor field if and
∗ ∗
only if the pure tensor field t ∈ ℑrs (Mmr ) ( the real model of t ) satisfies the
equation
φϕ t = 0, α = 1, ..., m,
α

where φϕ t is the Tachibana operator defined by (1.16).

Proof. It is well known that the components (φϕt)ij11...i

... js of φϕt with respect to
r

α α
local coordinate system x1 , ..., xmr may be expressed as (1.19). In the adapted
charts (∂k ϕij = 0), by virtue of (2.37), from (1.19) we have (ia = uaαa , jb = vb βb ,
k = wγ, a = 1, ..., r, b = 1, ..., s)

(φϕt)ij11...i r
... js = ϕm i1 ...ir i1 ...ir
k ∂m t j1 ... js − ∂k (t ◦ ϕ) j1 ... js
α α α
λ σ
µ µ α1 ...α r
= (Cαγ ∂wµ ℑ uv11 ...u u1 ...ur
...vs −Cαγ ∂wµ ℑ v1 ...vs )Bλβ ...β = 0.
r
1 s

From here and property B3 (see Section 2.4), it follows that the condition φϕt = 0
α
is equivalent to the condition
λ σ
µ µ
Cαγ ∂wµ ℑ uv11 ...u u1 ...ur
...vs = Cαγ ∂wµ ℑ v1 ...vs ,
r

∗ σ
which is the Scheffers condition (see (2.15)) of A−holomorphity of t uv11 ...u
...vs = ℑ
r

u1 ...ur e with respect to the local coordinates zu = xuα e from X (A ). Thus the
v1 ...vs σ α r m
proof is complete.
An infinitesimal automorphism of a regular Π− structure on Mmr is a vector
field X such that LX ϕ = 0, α = 1, ..., m, where LX denotes the Lie differentiation
α
with respect to X ∈ ℑ10 (Mmr ). From Theorem 30 and (d) of Definition 2, we
have

Corollary 31. Let on Mrm be given the integrable regular Π−structure. A vec-
tor field X ∈ ℑ10 (Mmr ) is an infinitesimal automorphism of Π−structure if and
only if X is A−holomorphic.

Remark 12. Let Mrm on be given the non-integrable regular Π−structure.

Then if t ∈ Kerφϕ , we say that t is an almost A−holomorphic tensor field.
α
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

2.6. Pure Connections

In this section, we always assume that the regular Π−structure is an integrable.
By a local coordinates we shall mean an adapted coordinates with respect to the
Π−structure.
Let ∇ be an Π−connection on Mmr , i.e. ∇ϕ = 0 for any ϕ ∈ Π. Since the
components of ϕ with respect to the local adapted coordinates x1 , ..., xmr are
constant, we have
∇ϕ = 0 ⇔ Γikm ϕmj = Γm i
k j ϕm . (2.39)
By the same arguments as developed in Section 2.5, we see that the Π− con-
nection has components of the form
α σ
Γik j = Γuα u
wγvβ = τ wγvCσβ (i = uα, j = vβ, k = wγ), (2.40)
σ
where τ− arbitrary functions in the adapted chart U. In fact, if ϕ = ϕ, then from
σ
(2.39), using contraction with εβ , we obtain

Γuα tε tε uα
wγtε ϕ vβ = Γwγvβ ϕ tε ,
σ σ
t ε u α
.

Γuα tε
wγtε δvCσβ = Γwγvβ δt Cσε ,
β α
Γuα uε
wγvσ = Γwγvβ ε Cσε
εα
= τ uwγvCεσ ,
ε β
where τ uwγv = Γuε
wγvβε , m = tε.
With each Π− connection (2.40) we can associate a hypercomplex values
from Am:
∗ σ
γ β γ
Γ uwv = Γuα u
wγvβ ε ε eα = τ wγv ε eσ . (2.41)
Definition 13. If the hypercomplex values (2.41) satisfies the connection con-
dition
∗ u´ ∂zu´ ∂zw ∂zv ∗ u ∂2 zu ∂zu´
Γw´v´ = u w´ v´ Γwv + v´ w´ u ,
∂z ∂z ∂z ∂z ∂z ∂z
∗ ∗
i.e. if Γ uwv are components of hypercomplex connection ∇ in Xr (Am), we say
that the Π−connection ∇ is a pure.
Theorem 32. Let Π be a regular integrable Π−structure on Mmr . The
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

α
Π−connection on Mmr is pure if and only if τ uwγv in (2.40) satisfies the con-
dition
αu σ α
τ wγv = τ uwvCσγ . (2.42)
Proof. Let Γik j = Γuαwγvβ be the components of the Π−connection ∇. Then,
taking account of the
admissible
u´α´transformation
{∂i } → {∂i´} of adapted frames
i´ ∂xi´ ∂x
with matrix (Si ) = ∂xi = ∂xuα , we have

∂zu´α´ ∂zwγ ∂zvβ uα ∂2 zuα ∂zu´α´

Γu´α´
w´γ´v´β´ = Γ + .
∂zuα ∂zw´γ´ ∂zv´β´ wγvβ ∂zw´γ´∂zv´β´ ∂zuα
After contraction with εγ´εβéα´, by virtue of (2.24), (2.25) and (2.26), we obtain
∗
Γ u´w´v´ = Γw´γ´v´β´
u´α´
εγ´εβéα´
σ ε θ
α´ γ β
γ´ β´
= ∆ uúCσα ∆w v uα
wĆεγ´∆ vĆθβ´Γwγvβ ε ε eα´
σ
α´ ∂ ε u α γ´ β´
+∆ uúCσα ( (∆ C ))ε ε eα´
∂zw´γ´ v´ εβ´

or
∗ σ ε θ
γ β u α ω γ´
σ ∂ αu
Γ u´w´v´ = ∆ uú∆ w v u´
w´∆ véσ eα δε δv τ wγvCωβ + ∆ u eσ ε ( (∆ )eα
∂zw´γ´ v´
∗ ∗ γ θ ∗ ∂ αu
u´ γ´
= S uúS vv´∆ w u
w´τ wγv eθ + S u ε ( (∆ )eα ,
∂zw´γ´ v´
where
∗ ∂zu´ u ∂zu u
S uú =, Su´ = u´ , z = xuα eα (see Section 2.3).
∂zu ∂z
Using (2.17), we see that

∂ αu ∂2 zu
εγ´( ( ∆ )eα = .
∂zw´γ´ v´ ∂zw´∂zv´
Thus, we have
∗ ∂zu´ γ w ∂zv θ u ∂2 zu ∂zu´
Γ u´w´v´ = ∆ w´ τ wγv eθ + . (2.43)
∂zu ∂zv´ ∂zv´∂zw´ ∂zu
∗ ∗
From (2.43) we easily see that ∇ with components Γ uwv is hypercomplex con-
θ
nection on Xr (Am ) if and only if τ uwγv satisfies the condition (2.42). The proof
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

is completed.
If we put (2.42), we get from (2.40) and (2.41), respectively

α σ α µ σ
Γik j = Γuα u u
wγvβ = τ wvCσγCµβ = τ wv Bσγβ (2.44)

and
∗ σ
Γ uwv = τ uwv eσ , (2.45)
where Bασγβ is the Kruchkovich tensor. Thus, we have

Corollary 33. A pure Π−connection ∇ has components (2.44) with respect to

the adapted coordinates.

Corollary 34. A pure Π−connection ∇ is a real model of the hypercomplex

∗
connection ∇ with components (2.45).

From Corollary 33 and (2.37) follows that the pure Π−connection as pure
tensor fields of type (1, 2) is defined by

Γikm ϕ mj = Γm i i m
k j ϕ m = Γm j ϕ k , α = 1, ..., m
α α α

with respect to the adapted charts (for pure torsion-free connection, see (1.26)).
But the pure tensor fields of type (1, 2) (see Section 2.4 ) is defined by similiar
equation with respect to the arbitrary charts.

2.7. Torsion Tensors of Pure Π−connections

Let S be a torsion tensor field of pure Π−connection ∇. Since Bασγβ = Bασβγ (see
Section 2.4), we have from (2.44):

Sik j = Γik j − Γijk (2.46)

σ σ
= (τ uwv − τ uvw )Bασγβ ,

i.e. S is a pure tensor (see (2.37)).

Conversely, we now assume that S is a pure tensor of Π−connection. Then,
by virtue of (2.37), we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

α λ
Sik j = Suα u
wγvβ = σ wv Bλγβ . (2.47)

On the other hand, from (2.40) we have

Sik j = Γik j − Γijk

λ
α α λ
= τ uwγvCλβ − τ uvβwCλγ . (2.48)

From (2.47) and (2.48), by virtue of contraction with εγ , we have

α λu γ λ u α
τvβw = τ wγv ε − σ vβw Cλβ ,

which shows the condition of type (2.42) is true, i.e. the Π−connection is a
pure.
Thus, we have

Theorem 35. The Π−connection ∇ is pure if and only if the torsion tensor of
∇ is pure.

For pure Π−connection by virtue of (2.38), (2.45) and (2.46), we obtain

∗ σ σ
S uwv = (τ uwv − τ uvw )eσ
∗ ∗
= Γ uwv − Γ uvw .

Thus, we have

Corollary 36. The pure torsion tensor field of the Π−connection ∇ is a real
∗
model of the hypercomplex torsion tensor of hypercomplex connection ∇.

Of course, a zero tensor field is pure, therefore we have

Corollary 37. A torsion-free Π−connection ∇ is always pure.

Also, from Corollary 36 and Corollary 37 we have

∗
Corollary 38. If ∇ is a torsion-free Π−connection, then ∇ with components
(2.45) is a torsion-free connection.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

2.8. A−holomorphic Hypercomplex Connection

In this and next sections, we give a real model of A−holomorphic hypercomplex
connections by using the ψϕ −operator applied to pure torsion-free connection
(see Section 1.4).
Let R be a curvature tensor of pure torsion-free connection ∇ with respect
to the integrable regular Π−structure.
Using (2.44) and properties of Kruchkovich tensors, we have

Rijkl = Ruα
vβwγtδ
= ∂vβ Γuα uα uα xε uα xε
wγtδ − ∂wγ Γvβtδ + Γvβxε Γwγtδ − Γwγxε Γvβtδ
σ σ
= ∂vβ (τ uwt Bασγδ ) − ∂wγ (τ uvt Bασβδ ) (2.49)
σ θ σ θ
+τ uvx Bασβε τ xwt Bεθγδ − τ uwx Bασγε τ xvt Bεθβδ
σ σ σ θ σ θ
= (∂vβ τ uwt )Bασγδ − (∂wγ τ uvt )Bασβδ + (τ uvx τ xwt − τ uwx τ xvt )Bασγθβδ .

We now assume that R is a pure tensor. Then by virtue of (2.37) we have

α λ
Rijkl = Ruα u
vβwγtδ = ρ vwt Bλβγδ (2.50)

From (2.49) and (2.50), we obtain

λ σ σ
ρ uvwt Bαλβγδ = (∂vβ τ uwt )Bασγδ − (∂wγ τ uvt )Bασβδ (2.51)
σ θ σ θ
+(τ uvx τ xwt − τ uwx τ xvt )Bασγθβδ .

After contraction with εβ εδ of (2.51) and by virtue of properties of Kruchkovich

tensors, we have
λ α σ σ σ θ σ θ
ρ uvwt Cλγ = εβ (∂vβ τ uwt )Cσγ
α λ α
− ∂wγ τ uvt + (τ uvx τ xwt − τ uwx τ xvt )CσθCλγ
α λ σ θ σ θ
= −∂wγ τ uvt + (εβ ∂vβ τ uwt + τ uvx τ xwt Cσθ
λ λ
− τ uwx τ xvt Cσθ α
)Cλγ ,

from which
α λ
α
∂wγ τ uvt = P uvwt Cλγ , (2.52)
where
λ λ σ θ σ θ λ
P uvwt = εβ ∂vβ τ uwt + τ uvx τ xwt Cσθ
λ λ
− τ uwx τ xvt Cσθ − ρ uvwt .
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

By virtue of (2.15), for fixed u, v, w and t (2.52) is the Scheffers condition

∗ α u
of A−holomorphity of Γ u =τ with respect to the local coordinates
wt vt eα
zu = xuα eα from Xr (Am ).
∗ α
Conversely, if Γ uwt = τ uvt eα is a A−holomorphic connection, then from
Theorem 22 (Ceα = Cα ) and (2.15) we obtain the condition (2.52). Using (2.52),
from (2.49) we have

Rijkl = Ruα
vβwγtδ
λ λ σ θ σ θ
σ
= (P uwvt Cλβ )Bασγδ − (P uvwt Cλγ
σ
)Bασβδ + (τ uvx τ xwt − τ uwx τ xvt )Bασγθβδ
λ λ σ θ σ θ
= P uwvt Bαλβγδ − P uvwt Bαλγβδ + (τ uvx τ xwt − τ uwx τ xvt )Cσθ
λ α
Bλγβδ (2.53)
λ
= ρ uvwt Bαλβγδ ,

where
λ λ λ σ θ σ θ
λ
ρ uvwt = P uwvt − P uvwt + (τ uvx τ xwt − τ uwx τ xvt )Cσθ .

Thus, from (2.37) and (2.53), we see that R is pure curvature tensor.
Summing up, we have
∗
Theorem 39. [36], [104] Let ∇ be a hypercomplex connection on Xr (Am) and
∇ its a real model (a pure connection) on Mmr . The curvature tensor R of ∇ is
∗
pure if and only if ∇ is A− holomorphic connection.

On the other hand, from Theorem 15 and Theorem 39 we have

Theorem 40. Let ∇ be a pure connection which satisfies the condition

(ψϕ ∇)(X,Y.Z) = ∇ϕX ∇Y Z − ϕ(∇X ∇Y Z) = 0

α α α

for any A−holomorphic vector fields X,Y, Z and α = 1, ..., m.Then such con-
∗
nection is a real model of A−holomorphic connection ∇ .

2.9. Some Properties of Pure Curvature Tensors

Let now R be a pure tensor field of pure connection ∇ which satisfies the condi-
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

tion ∇ ∈ Kerψϕ . Using (2.38), from (2.49) we have

∗
β γ δ
R uvwt = Ruα
vβwγtδ ε ε ε eα
σ σ σ θ σ θ
= εβ (∂vβ τ uwt )eα − εγ (∂wγ τ uvt )eα + (τ uvx τ xwt − τ uwx τ xvt )Cσθ
α
eα .
α
Since, Cσθ eα = eσ eθ , by virtue of (2.17) and (2.45), we have
∗u ∗u ∗u ∗u ∗x ∗u ∗x
Rvwt = ∂v Γwt − ∂w Γvt + Γvx Γwt − Γwx Γvt ,
∗ ∗
i.e. R is a curvature tensor of Γ.
Thus, we have
∗
Theorem 41. Let ∇ be a real model of A−holomorphic connection ∇. The
∗
hypercomplex components R uvwt of the pure curvature tensor R are components
∗
of curvature tensor of ∇.

(φϕ R)(X,Y1,Y2 ,Y3 ) = (φϕX R)(Y1,Y2 ,Y3 ) − ϕ(∇X R)(Y1,Y2 ,Y3 ), (2.54)
α α α
X,Y1 ,Y2 ,Y3 ∈ ℑ10 (Mmr ).

Using the purity of R and applying the Bianchi’s 2nd identity to (2.54), we get

(φϕ R)(X,Y1,Y2 ,Y3 ) = (φϕX R)(Y1,Y2 ,Y3 ) − ϕ(∇X R)(Y1 ,Y2 ,Y3 )
α α α
= −(∇Y1 R)(Y2 , ϕX,Y3 ) − (∇Y2 R)(ϕX,Y1 ,Y3 )
α α
−ϕ(∇X R)(Y1 ,Y2 ,Y3 ).
α

On the other hand, using ∇ϕ = 0, we find

α
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(∇Y2 R)(ϕX,Y1 ,Y3 ) = ∇Y2 (R(ϕX,Y1 ,Y3 )) − R(∇Y2 (ϕX),Y1 ,Y3 )

α α α
−R(ϕX, ∇Y2 Y1 ,Y3 ) − R(ϕX,Y1 , ∇Y2 Y3 )
α α
= (∇Y2 ϕ)(R(X,Y1,Y3 )) + ϕ(∇Y2 (R(X,Y1,Y3 ))
α α
−R((∇Y2 ϕ)X + ϕ((∇Y2 X),Y1 ,Y3 ))
α α
−R(ϕX, ∇Y2 Y1 ,Y3 ) − R(ϕX,Y1 , ∇Y2 Y3 )
α α
= ϕ(∇Y2 (R(X,Y1,Y3 )) − ϕ(R(∇Y2 X,Y1 ,Y3 )) (2.55)
α α
−ϕ(R(X, ∇Y2 Y1 ,Y3 )) − ϕ(R(X,Y1 , ∇Y2 Y3 ))
α α
= ϕ((∇Y2 R)(X,Y1,Y3 )).
α

Similarly
(∇Y1 R)(Y2 , ϕX,Y3 ) = ϕ((∇Y1 R)(Y2, X,Y3 )). (2.56)
α α

Substituting (2.55) and (2.56) into (2.54), and using again the Bianchi’s 2nd
identity, we obtain

(φϕ R)(X,Y1,Y2 ,Y3 ) = −ϕ((∇Y1 R)(Y2, X,Y3 )) − ϕ((∇Y2 R)(X,Y1,Y3 ))

α α α
−ϕ(∇X R)((Y1,Y2 ,Y3 ))
α
= −ϕ(σ{(∇X R)(Y1,Y2 },Y3 ) = 0,
α

where σ denotes the cyclic sum with respect to X,Y1 and Y2 . Therefore, by
virtue of Theorem 30 and Theorem 39, we have
∗ ∗
Theorem 42. The curvature tensor R of A− holomorphic connection ∇ is A−
holomorphic tensor.

Example 9. Let now r = 1, i.e. we consider a A−holomorphic manifold

X1 (Am ) of hypercomplex dimension 1. Since u = v = w = t = 1, we have from
(2.49)

Rijkl = R1α
1β1γ1δ
σ σ σ θ σ θ
= (∂1β τ 111 )Bασγδ − (∂1γ τ 111 )Bασβδ + (τ 111 τ 111 − τ 111 τ 111 )Bασγθβδ
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

σ σ
= (∂1β τ 111 )Bασγδ − (∂1γ τ 111 )Bασβδ (2.57)
σ
α ε ε α σ
= (∂β τ)CδεCσγ − (∂γ τ)CσβCεδ
ε εσ α σ
= (Cσγ ∂β τ −Cσβ ∂γ τ)Cδε ,

σ σ
where τ = τ 111 .

We now assume that

ε ε σ σ
Cσγ ∂β τ = Cσβ ∂γ τ. (2.58)
After contraction with εγ , we have from (2.58)
ε ε
∂β τ = εγ(∂γ τ)Cσβ
ε
, (2.59)

i.e. by virtue of (2.17), τ is A−holomorphic function of x = xαeα ∈ Am .

Conversely, let τ = τ(x) be a holomorphic function. Then, from (2.59) we

have
θ ε σ
Cετ ∂β τ = εγ (∂γ τ)Cσβ
ε θ
Cετ
σ
= εγ (∂γ τ)Cβε
θ ε
Cστ
σ
= εγ (∂γ τ)Cστ
ε θ
Cβε
ε
= εγ (∂σ τ)Cγτ
σ θ
Cβε
θ ε
= (∂τ τ)Cβε ,

i.e. the condition (2.58) is true and the condition (2.58) is equivalent to the
Scheffers condition (2.15). Thus, if τ = τ(x) is A−holomorphic, then Rijkl = 0.
Conversely, if R = 0, then from (2.57) we have
δ
0 = R1α
1β1γ1δ ε
ε ε σ α δ σ
= (Cσγ ∂β τ −Cσβ ∂γ τ)Cδε ε
α α σ σ
= Cσγ ∂β τ −Cσβ ∂γ τ,

i.e. the function τ = τ(x) is A−holomorphic.

Obviously, we have
∗
Theorem 43. Let Mm be a real model of X1 (Am). The connection ∇ with com-
∗ σ
ponents τ = τeσ on X1 (Am ) is A−holomorphic if and only if the real manifold
Mm is locally flat.
σ σ
Remark 13. In particular, if τ = τ 111 = εσ (1 = εσ eσ ∈ Am), then from (2.44)
we have Γαγβ = Cγβ
α
, i.e. Mm is the Vranceanu space [109], [63].

Applications to the Norden

Geometry

In this chapter, we study the various hypercomplex Norden manifolds. In Sec-

tion 3.1 we give the fundamental properties of almost hypercomplex Norden
manifolds. We give the condition for an almost hypercomplex Norden mani-
fold to be holomorphic (Kählerian). In Section 3.2 we discuss complex Norden
manifolds. We define the twin Norden metric, the main theorem of this section
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

is that the Levi-Civita connection of Kähler-Norden metric coincides with the

Levi-Civita connection of twin Norden metric. Section 3.3 devoted to the study
of almost product Riemannian manifolds. We give the definitions of decom-
posable Riemannian manifolds and para-Kähler-Norden manifolds. The final
theorem of this section is that there does not exist para-Kähler-Norden warped
metric. In Section 3.4, we give some examples of dual-Kähler-Norden man-
ifolds. In Section 3.5, we consider Norden-Hessian structures. We give the
condition for a Kähler (para-Kähler-Norden) manifold to be Norden-Hessian.
In Section 3.6, we shall focus our attention to Norden manifolds of dimen-
sion four. The main purpose of the present section is to study complex Norden
metrics on 4-dimensional Walker manifolds. We discuss the integrability and
Kahler (holomorphic) conditions for these structures. The curvature properties
for Norden-Walker metrics is also investigated. Examples of Norden-Walker
metrics are constructed from an arbitrary harmonic function of two variables.
We define the isotropic Kähler structures and moreover, show that a proper al-
most complex structure on almost Norden-Walker manifold is isotropic Kähler.
We also consider the quasi-Kähler-Norden metric and give the condition for an

almost Norden manifold to be quasi-Kähler-Nordenian. In this section we also

give progress to the conjecture of Goldberg under the additional restriction on
Norden-Walker metric. Section 3.7 is devoted to the analysis of opposite almost
complex structures on Norden-Walker 4-manifolds. In Section 3.8, we shall fo-
cus our attention to para-Norden manifolds of dimension four. Using a Walker
metric we construct para-Norden-Walker metrics together with a proper almost
paracomplex structures. In Section 3.9, we give some examples of Norden-
Walker metrics on 8-dimensional manifolds.

3.1. Hyper-Kähler-Norden Manifolds

Let Mmr be a Riemannian manifold with metric g , which is not necessarily pos-
itive
definite,
and let on Mmr be given the regular hypercomplex Π−structure
u α
Π = ϕ , ϕ ij = ϕ uα
vβ = δv Cσβ , i = uα, j = vβ; i, j = 1, ..., mr; α, β, σ = 1, ...., m;
σ σ σ
u, v = 1, ..., r (see Section 2.2). A Norden metric with respect to the hypercom-
plex structure is a Riemannian metric g such that

g(ϕX,Y ) = g(X, ϕY ), α = 1, ..., m (3.1)

α α
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

for any X,Y ∈ ℑ10 (Mmr ), i.e. g is pure with respect to the regular hypercomplex
Π−structure. Such Riemannian metrics were studied in [104], where they were
said to be B-metrics, since the metric tensor g with respect to the Π−structure is
B-tensor according to the terminology accepted by Norden [58]. If (Mmr , Π) is
an almost hypercomplex manifold with Norden metric, we say that (Mmr , Π, g)
is an almost hypercomplex Norden manifold. If Π−structure is integrable, we
say that (Mmr , Π, g) is a hypercomplex Norden manifold.
β
Remark 14. Let ψ be a conjugation of Am (see Section 2.1.2). We put ϕ = ψα ϕ.
α β
A Riemannian metric g which satisfies

g(ϕX,Y ) = g(X, ϕY ), α = 1, ..., m

α α

for any X,Y ∈ ℑ10 (Mmr ) is called a hybrid [37] with respect to the conjugation
ψ. In particular, if ψ = idAm , then the hybrid tensor g is pure. If Am = C (m = 2)
is complex algebra, then for the conjugation ψ 6= id , we have

1 0
β
ϕ = ψ(ϕ) = −ϕ, ψα = ,
0 −1
i.e. a hybrid tensor g with respect to the conjugation is characterized by

g (ϕX,Y ) = −g (X, ϕY )
for any X,Y ∈ ℑ10 (M2r ). It is well known that, if g hybrid with respect to the
conjugation ψ 6= id and positive-definite, then the triple (Mmr , Π, g) is an almost
Hermitian manifold.

A Norden metric g is called a A−holomorphic(see Theorem 30) if

(φϕ g) (X,Y, Z) = (ϕX) (g (Y, Z)) − X(g(ϕY, Z))
α α α
+g((LY ϕ)X, Z) + g(Y, (LZ ϕ)X)
α α
= (LϕX g − LX (g ◦ ϕ))(Y, Z) = 0, α = 1, ..., m
α α

for any X,Y, Z ∈ ℑ10 (Mmr ), where φϕ g, α = 1, ..., m are Tachibana operators
α
applied to a Norden metric. If (Mmr , Π, g) is a hypercomplex Norden man-
ifold with A−holomorphic Norden metric g, we say that (Mmr , Π, g) is a
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

A−holomorphic Norden manifold. Since the case where dimM = m, such a

manifold is flat (see Theorem 43 and Theorem 47), we assume in the sequel that
dim M ≥ 2m, i.e. r ≥ 2.
In some aspects, A−holomorphic Norden manifolds are similar to Kähler
manifolds. The following theorem is analogous to the next known result: An
almost Hermitian manifold is Kähler if and only if the almost complex structure
is parallel with respect to the Levi-Civita connection [34, Chapter 9].
The following theorem plays very important role in Norden geometry:

Theorem 44. An almost hypercomplex Norden manifold is A−holomorphic

Norden manifold if and only if structure affinors are parallel with respect to
the Levi-Civita connection ∇g .

Proof. Using (3.1) and [X,Y ] = ∇X Y − ∇Y X, we get from (1.13)

(φϕ g)(X, Z1 , Z2 ) = (LϕX g − LX (g ◦ ϕ))(Z1 , Z2 )

α α α
+g(Z1, ϕLX Z2 ) − g(ϕZ1 , LX Z2 )
α α

= (ϕX)g(Z1 , Z2 ) − Xg(ϕZ1 , Z2 ) − g(∇ϕX Z1 , Z2 )

α α α

+g(∇Z1 ϕX, Z2 ) − g(Z1 , ∇ϕX Z2 ) + g(Z1 , ∇Z2 ϕX)

α α α
+g(ϕ(∇X Z1 ), Z2 ) − g(ϕ(∇Z1 X), Z2 )
α α
+g(ϕZ1 , ∇X Z2 ) − g(Z1 , ϕ(∇Z2 X)).
α α

We find

g(∇Z1 ϕX, Z2 ) − g(ϕ(∇Z1 X), Z2 ) + g(Z1 , ∇Z2 ϕX) − g(Z1 , ϕ(∇Z2 X))
α α α α
= g((∇ϕ)(X, Z1 ), Z2 ) + g(Z1 , (∇ϕ)(X, Z2 )). (3.2)

Substituting (3.2) into (3.2), we have

(φϕ g)(X, Z1 , Z2 ) = (ϕX)g(Z1 , Z2 ) − Xg(ϕZ1 , Z2 ) + g((∇ϕ)(X, Z1 ), Z2 )

α α α α
+g(Z1 , (∇ϕ)(X, Z2 )) − g(∇ϕX Z1 , Z2 ) − g(Z1 , ∇ϕX Z2 )
α α α

+g(ϕ(∇X Z1 ), Z2 ) + g(ϕZ1 , ∇X Z2 ). (3.3)

α α

On the other hand, with respect to the Levi-Civita connection ∇, we have

(ϕX)g(Z1 , Z2 ) − g(∇ϕX Z1 , Z2 ) − g(Z1 , ∇ϕX Z2 ) = (∇ϕX g)(Z1 , Z2 ) = 0 (3.4)

α α α α

and

−Xg(ϕZ1 , Z2 ) + g(ϕ(∇X Z1 ), Z2 ) + g(ϕZ1 , ∇X Z2 ) = −g((∇X ϕ)Z1 ), Z2 ). (3.5)

α α α α

By virtue of (3.4) and (3.5), (3.3) reduces to

(φϕ g)(X, Z1, Z2 ) = −g((∇X ϕ)Z1 ), Z2 ) + g((∇Z1 ϕ)X, Z2 ) + g(Z1 , (∇X ϕ)Z2 ).
α α α α
(3.6)
Similarly, we have

(φϕ g)(Z2 , Z1 , X) = −g((∇Z2 ϕ)Z1 ), X) + g((∇Z1 ϕ)Z2 , X) + g(Z1 , (∇Z2 ϕ)X).

α α α α
(3.7)

The sufficiency follows easily from (3.6) or (3.7). By virtue of g(Z, (∇Y ϕ)X) =
α
g((∇Y ϕ)Z, X) we find
α

(φϕ g)(X, Z1, Z2 ) + (φϕ g)(Z2 , Z1 , X) = 2g(X, (∇Z1 ϕ)Z2 ). (3.8)

α α α

Now, putting φϕ g = 0 in (3.8), we find ∇ϕ = 0, α = 1, ..., m from which the

α α
necessity follows. Thus Theorem 44 is proved.
Using Remark 11, from Theorem 44 we have

Corollary 45. The Π−structure on almost hypercomplex Norden manifold is

integrable if φϕ g = 0, α = 1, ..., m.
α

Also, from Theorem 44 we have

Corollary 46. The Levi-Civita connection ∇g on A−holomorphic Norden man-

ifolds is pure connection.

Remark 15. Recall that a hyper-Kähler-Norden manifold can be defined as a

triple (M2n , Π, g) which consists of a manifold Mmr endowed with a hypercom-
plex structure Π and a pseudo-Riemannian metric g such that ∇ϕ = 0, α =
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

α
1, ..., m , where ∇ is the Levi-Civita connection of g and the metric is assumed
to be Nordenian: g(ϕX,Y ) = g(X, ϕY ), α = 1, ..., m. Therefore, there exists a
α α
one-to-one correspondence between hyper-Kähler-Norden manifolds and com-
plex Riemannian manifolds with a A−holomorphic metric.

From (3.1) we have

σ
gi j = guαvβ = GuvσCαβ (3.9)

for arbitrary functions Guvσ (see Section 2.4). The corresponding hypercomplex
∗
tensor guv is defined by
∗
guv = guαvβ εβ εα = Guvα eα , (3.10)
∗
where eα = ϕαβ eβ , ϕαβ is the Frobenius metric. Since is guv symmetric, non-
∗
∗ ∗
singular (Det(guv ) 6= 0) and A−holomorphic, it follows that ds2 = guvdzu dzv can

be regarded as hyper-Kähler-Norden metric in the A−holomorphic manifold

Xr (A).
Since the Levi-Civita connection ∇g of A−holomorphic Norden manifolds
is pure connection, by virtue of (2.44), we obtain
σ σ σ σ
α µ
Kki j = Kwγvβ
uα
= k uwvCσγCµβ = k uwv Bασγβ , k uwv = k uvw , (3.11)

where are Kki j components of ∇g and i = uα, j = vβ, k = wγ. Substituting (3.9)
and (3.11) into ∇k gi j = 0, we find
σ σ
σ τ µ τ µ
∂wγ (GuvσCαβ ) = k twuCσγ
v
Cvα GtvτCµβ + k twvCσγ
v
Cvβ GtuτCµα ,

After contraction with εβ εγεα , by virtue of (2.5) and (2.10), we obtain

σ σ
εγ ∂wγ Guvσeα = (k twu Gtvµ + k twv Gtuµ )eσ eµ .

By virtue of (2.17), (2.45) and (3.10), we have

∗
∂guv ∗ t ∗ ∗ ∗
w
= k wu gtv + k twv gut , (3.12)
∂z
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

∗
from which we easily see that the Christoffel symbols formed with g has com-
ponents
∗ ∗ ∗
∗ 1 ∗ ut ∂g ∂gwt ∂guv
k uwv = g ( tv + − ). (3.13)
2 ∂zw ∂zv ∂zw
∗
Since the hypercomplex Norden metric g is A−holomorphic, exists the suc-
∗ ∗
u
cessive derivatives of g (see Section 2.1.3), i.e. from (3.13) it follows that k wv
is A−holomorphic.
Thus we have

Theorem 47. The Levi-Civita connection of hyper-Kähler-Norden manifold is

A− holomorphic.

Using Theorem 39 we have

Theorem 48. The Riemannian curvature tensor field of hyper-Kähler-Norden

manifold is pure.

From Theorem 42 we have

Theorem 49. The Riemannian curvature tensor field of hyper-Kähler-Norden

manifold is A−holomorphic.

In particular, if Am = C (m = 2), an alternative proof of Theorem 48 will be

made in Section 3.2 by using the real model of hyper-Kähler-Norden manifolds.
From Theorem 6 we have: A necessary and sufficient condition for an exact
1-form d f , f ∈ ℑ00 (M2m) to be A−holomorphic is that associated 1-forms d f ◦
ϕ, α = 1, ..., m be closed 1-forms, i.e.
α

d(d f ◦ ϕ) = 0, α = 1, ..., m. (3.14)

If there exists functions g, α = 1, ..., m in a hyper-Kähler-Norden manifold

α
such that d f ◦ ϕ = dg, α = 1, ..., m for a function f , then we shall call f a
α α
A−holomorphic function and g, α = 1, ..., m associated functions. If such a
α
function f is defined locally, then we call it a locally A−holomorphic function.
We notice that the equation (3.14) is equivalent to d f ◦ ϕ = dg, α = 1, ..., m
α α
only locally. Hence the condition for f to be locally A−holomorphic (ϕm
i ∂m f =
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

α
∂i g) also is given by
α

(φϕ d f ) = ϕm m m
i ∂m ∂i f − ∂i (ϕi ∂m f ) + (∂i ϕi )∂m f = 0 , α = 1, ..., m.
α α α α

Let (Mmr , Π, g) be a A−holomorphic manifold with hyper-Kähler-Norden

metric g. Then using Theorem 13 and Theorem 47, by virtue of (1.23) we find
that in hyper-Kähler-Norden manifolds the covariant derivative of the curvature
tensor field ∇R is also pure. Therefore, the covariant derivative of the Ricci
tensor R ji = Rss ji = gts Rt jis is pure in all its indices and hence

ϕts ∇s R ji = ϕsj ∇t Rsi , α = 1, ..., m.

α α

Transvecting this equation with contravariant hyper-Kähler-Norden metric g ji ,

we find
∗
ϕts ∇s R = g ji ϕsj ∇t Rsi = ∇t (g ji ϕsj Rsi ) = ∇t R , α = 1, ..., m, (3.15)
α α
α α

where R = gi j Ri j are curvature scalars of hyper-Kähler-Norden metric g and

∗
R = g ji ϕsj Rsi . From (3.15) we have
α
α

Theorem 50. In a hyper-Kähler-Norden manifold, the curvature scalar R is a

locally A−holomorphic function.

Remark 16. In the next sections, we shall easily see that if Am(m = 2) is com-
∗
plex algebra (or paracomplex algebra), then the associated function R is a cur-
vature scalar of twin-Norden metric.

3.2. Complex Kähler-Norden Manifolds

Let Am = C (m = 2) be a complex algebra. In this section we shall only consider
C−holomorphic Norden (i.e. Kähler-Norden) manifolds.
Let M2n be a pseudo-Riemannian manifold with a neutral metric, i.e. with
a pseudo-Riemannian metric g of signature (n, n). We say (M2n , ϕ) is an almost
complex manifold if M2n can be endowed with an affinor field ϕ ∈ ℑ11 (M2n)
such that ϕ2 = −I, where I is a field of identity endomorphisms. If the Ni-
jenhuis tensor field Nϕ ∈ ℑ12 (M2n ) vanishes, then ϕ is a complex structure and
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

moreover M2n is a C−holomorphic manifold Xn (C) whose transition functions

are C−holomorphic mappings. A metric g is a Norden metric [19] if

g(ϕX,Y ) = g(X, ϕY ), (3.16)

for any X,Y ∈ ℑ10 (M2n), i.e. g is pure with respect to ϕ. Metrics of this
kind have been also studied under the names: anti-Hermitian and B-metrics
(see [5], [7], [14], [15], [25], [26], [31], [42], [50], [51], [52], [60], [61], [64],
[70], [71], [72], [80], [81], [82], [83], [101], [104], [105]). If (M2n , ϕ) is an
almost complex manifold with Norden metric, we say that (M2n , ϕ, g) is an al-
most Norden manifold. If ϕ is integrable, we say that (M2n , ϕ, g) is a Norden
manifold.
Let (M2n, ϕ, g) be an almost Norden manifold. The twin Norden metric of
almost Norden manifold is defined by

G(X,Y ) = (g ◦ ϕ)(X,Y), (3.17)

for all vector fields X and Y on M2n . One can easily prove that G is a metric,
which is also called the associated (or dual) metric of g and it plays a role similar

to the Kahler form in Hermitian Geometry. We shall now apply the Tachibana
operator to the pure Riemannian metric G :

(φϕ G)(X,Y, Z) = (LϕX G − LX (G ◦ ϕ))(Y, Z)

+G(Y, ϕLX Z) − G(ϕY, LX Z)
= (φϕ g)(X, ϕY, Z) + g(Nϕ (X,Y ), Z). (3.18)

Thus (3.18) implies the following

Theorem 51. In an almost Norden manifold, we have

φϕ G = (φϕ g) ◦ ϕ + g ◦ (Nϕ ).

Corollary 52. In a Norden manifold the following conditions are equivalent:

a) φϕ g = 0,
b) φϕ G = 0.

From Theorem 44 and Theorem 51 we have

Theorem 53. Almost Norden manifold with conditions φϕ G = 0 and Nϕ 6= 0,

i.e. analogues of the almost Kähler manifolds with closed Kähler form, do not
exist.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

We denote ∇g by the covariant differentiation of Levi-Civita connection of

Norden metric g. Then, we have

φϕ G = (φϕ g) ◦ ϕ + g ◦ (φg ϕ) = g ◦ (∇g ϕ),

which implies φϕ G = 0 by virtue of Theorem 44. Therefore we have

Theorem 54. Let (M2n , ϕ, g) be a Kähler-Norden manifold. Then the Levi-

Civita connection of Norden metric g coincides with the Levi-Civita connection
of twin Norden metric G.

We note that, in the case where Am = C (m = 2), we can prove the Theorem
48 by using Theorem 54. In fact, let R and S be the curvature tensors formed
by g and G respectively, then for the Kähler-Norden manifold we have R = S by
means of the Theorem 54. Applying the Ricci’s identity to ϕ, we get

ϕ(R(X,Y)Z) = R(X,Y )ϕZ (3.19)

by virtue of ∇ϕ = 0. Hence R(X1, X2 , X3 , X4 ) = g(R(X1, X2 )X3 , X4 ) is pure with

respect to X3 and X4 and also pure with respect to X1 and X2 :

R(X1 , X2 , ϕX3 , X4 ) = g(R(X1, X2 )ϕX3 , X4 )

= g(ϕ(R(X1, X2 )X3 ), X4 )
= g(R(X1, X2 )X3 , ϕX4 )
= R(X1 , X2 , X3 , ϕX4 ).

On the other hand, S being the curvature tensor formed by twin metric G, if we
put S(X1 , X2 , X3 , X4 ) = G(S(X1, X2 )X3 , X4 ) , then we have

S(X1 , X2 , X3 , X4 ) = S(X3 , X4, X1 , X2 ). (3.20)

Taking account of (3.16), (3.17), (3.19) and R = S , we find that

S(X1 , X2 , X3 , X4 ) = G(S(X1, X2 )X3 , X4 )

= g(ϕ(S(X1 , X2 )X3 ), X4 )
= g(S(X1, X2 )X3 , ϕX4 )
= g(R(X1, X2 )X3 , ϕX4 )
= R(X1, X2 , X3 , ϕX4 )
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

and

S(X3 , X4 , X1 , X2 ) = G(S(X1, X2 )X3 , X4 )

= g(ϕ(S(X3 , X4 )X1 ), X2 )
= g(S(X3, X4 )X1 , ϕX2 )
= g(R(X3, X4 )X1 , ϕX2 )
= R(X3, X4 , X1 , ϕX2 )
= R(X1, ϕX2 , X3 , X4 ).

Thus the equation (3.20) becomes

R(X1 , X2 , X3 , ϕX4 ) = R(X1 , ϕX2 , X3 , X4),

which shows that R(X1 , X2, X3 , X4 ) is pure with respect to X2 and X4 . Therefore
R(X1 , X2 , X3 , X4 ) is pure.

We now consider a Kähler-Norden manifold and its invariant submanifold.

Let (M2n , ϕ, g) be a Kähler-Norden manifold . From (1.31) we have

−BXe = ϕ2 (BX)
e = ϕB(ϕ e2 X)
e = B(ϕ
eX) e

or
e2 X
B(ϕ e + Xe ) = 0

for any Xe ∈ ℑ10 (M e 2 = −I , so an

e2m).Since B is injective, from here we can see ϕ
invariant submanifold M e2m becomes an almost complex manifold by virtue of
e.
the induced affinor ϕ
If we take account of Theorem 16, then we get

Theorem 55. An invariant submanifold in a Kähler-Norden manifold is itself

Kähler-Norden manifold with respect to the induced structure.

3.3. Almost Product Riemannian Manifolds

3.3.1. Decomposable Riemannian Manifolds
Let Mn be an almost product manifold with almost product structure F . If Mn
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

admits a Riemannian metric g such that

g(FX,Y ) = g(X, FY )

for X,Y ∈ ℑ10 (Mn ), i.e. if g is pure with respect to F, then Mn is called an almost
product Riemannian manifold [56], [115, p.423].
We define the operator φF : ℑ02 (Mn ) → ℑ03 (Mn ) associated with F and ap-
plied to the pure metric g :

(φF g)(X,Y1,Y2 ) = (FX)(g(Y1,Y2 )) − X(g(FY1 ,Y2 )

+g((LY1 F)X,Y2 ) + g(Y1 , (LY2 F)X).

By similar devices (see Section 3.1), we have

(φF g)(X, Z1 , Z2 ) = −g((∇X F)Z1 , Z2 ) + g((∇Z1 F)X, Z2 ) + g(Z1 , (∇Z2 F)X)

and
(φF g)(X, Z1, Z2 ) + (φF g)(Z2 , Z1 , X) = 2g(X, (∇Z1 F)Z2 ).

Putting φF g in the last equation, we find ∇F = 0.

On the other hand, we know that the integrability of the almost product
structure F is equivalent to the existing a torsion-free affine connection with
respect to which the equation ∇F = 0 holds. Since the Levi-Civita connection
∇ of g is a torsion-free affine connection, we have

Theorem 56. Let (Mn , F) be an almost product Riemannian manifold with pure
metric g. Then F is integrable if φF g = 0 .

We note that, if ∇F = 0, then the condition φF g = 0 follows from the above

expression of (φF g)(X, Z1, Z2 ). Thus we have

Theorem 57. [76] For an almost product Riemannian manifold with pure met-
ric g, the condition φF g = 0 is equivalent to ∇F = 0, where ∇ is the Levi-Civita
connection of g.

An integrable almost product Riemannian manifold with structure tensor

F is usually called a locally product Riemannian manifold. If the metric of a
locally product Riemannian manifold Mn has the form

ds2 = gab (xc )dxa dxb + gab (xc )dxa dxb ,

a, b, c, ... = 1, ..., m, a, b, c, ... = m + 1, ..., n

that gab is are functions of xc only, gab = 0 and gab are functions xc only, then
we call the manifold Mn a locally decomposable Riemannian manifold . On the
other hand, we know that the locally product Riemannian manifold with struc-
ture tensor F is locally decomposable if and only if F is covariantly constant
with respect to the Levi-Civita connection ∇ [115, p.420]. Thus, by Theorem
56 and Theorem 57 we have

Theorem 58. Let (Mn , F) be an almost product Riemannian manifold with pure
metric g. A necessary and sufficient condition for (Mn , F) to be a locally decom-
posable Riemannian manifold is that φF g = 0.

3.3.2. Para-Kähler-Norden Manifolds

Let Am = A(e),(m = 2) be a paracomplex algebra. We shall only consider A(e)-
holomorphic Norden (i.e. para-Kähler-Norden) manifolds.

An almost paracomplex manifold is an almost product manifold (Mn , ϕ),

ϕ2 = id, such that the two eigenbundles T + Mn and T − Mn associated to the
two eigenvalues +1 and −1 of ϕ, respectively, have the same rank. Note
that the dimension of an almost paracomplex manifold is necessarily even.
Considering the paracomplex structure ϕ, we obtain the following set of affi-
nors on Mn :{id, ϕ}, ϕ2 = id, which form a bases of a representation of
the algebra of order 2 over the field of real numbers R, which is called
algebra of
the paracomplex (or double) numbers and is denoted by A(e) =
2
a0 + a1 e | e = 1; a0, a1 ∈ R . Obviously, it is associative, commutative and
unitial, i.e., it admits principal unit 1. The canonical bases of this algebra has
the form {1, e}. Structural constants of an algebra are defined by the multipli-
cation law of the base units of this algebra: ei e j = Cikj ek . The components of Cikj
are given by C11 1 = C2 = C2 = C1 = 1, all the others being zero with respect
12 21 22
to the canonical bases of A(e).
Consider A(e) endowed with the usual topology of and a domain U of A(e).
Let
X = x1 + ex2
be a variable in A(e), where xi are real coordinates of a point of a certain domain
U for i = 1, 2. Using two real-valued functions f i (x1 , x2 ), i = 1, 2, we introduce
a paracomplex function
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

F = f1 +ef2
of variable X. It follows that F is paraholomorphic if and only if f 1 and f 2
satisfy the para-Cauchy-Riemann equations(see Example 5):

∂f1 ∂f2 ∂f1 ∂f2

= , = .
∂x1 ∂x2 ∂x2 ∂x1
By similar devices (see Theorem 54), we can prove a following: In a para-
Kähler-Norden case, the Levi-Civita connection of is also the Levi-Civita con-
nection of the twin metric G given by G(X,Y) = g(ϕX,Y ), moreover, in such a
manifold, the Riemannian curvature tensor is paraholomorphic.

Remark 17. The leaves of the foliations defined by the paracomplex structure
of a para-Kähler-Norden manifold are totally geoedesic submanifolds (see [56]
or the book Yano and Kon [115, p.420]). In the paper of Naveira [56] Rieman-
nian almost product manifolds were classified.

Example 10. Let now Mn be the locally product Riemannian manifold with in-
tegrable almost product structure
!
δij 0
ϕ= , i, j = 1, ..., k, i, j = k + 1, ...., n
0 −δij

and let n = 2k. Then the paracomplex manifold M2k , admit a metric of para-
Norden manifold:

gi j 0
g= , gi j = gi j (xt , xt ), gi j = gi j (xt , xt ).
0 gi j

Suppose that the metric of the locally product Riemannian manifold M2k has
the form

ds2 = gi j (xt )dxi dx j + gi j (xt )dxi dx j , i, j,t = 1, ..., k ; i, j,t = k + 1, ..., 2k,

that is gi j (x) are functions of xt only, gi j = 0, and gi j (x) are functions of xt only,
i.e. the manifold M2k is a locally decomposable Riemannian manifold. Since the
necessary and sufficient condition for a locally product Riemannian manifold to
be a locally decomposable Riemannian manifold is that ∇g ϕ = 0, we have from
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 44

Theorem 59. A locally decomposable Riemannian manifold M2k is a para-

Kähler-Norden manifold.

Example 11. [85] Let (M2k, ω) be a symplectic manifold and let D be a La-
grangian distribution, which is a k−dimensional distribution having ω/D = 0.
Then, M may be endowed with an almost para-Norden structure.First of all, we
shall prove that there exist a transversal Lagrangian distribution. Taking into
account that (M, ω) is an almost symplectic manifold one can find (see [1] or
[62]) an almost Hermitian structure (J, G) on M such that ω(X,Y ) = G(JX,Y ).
Let D⊥ the G−orthogonal distribution to D. Then one has:

(1) If X,Y ∈ D, then G(JX,Y ) = ω(X,Y ), thus proving that J(D) = D⊥ .

(2) D⊥ is a Lagrangian distribution, because ω(JX, JY ) = ω(X,Y ), for all
X,Y ∈ ℑ10 (Mn ).
Let F be the almost product structure defined by D and D⊥ , i.e.,F + = D
and F − = D⊥ . Then, one easily check that J ◦ F = −F ◦ J. Moreover, one can

prove that (M, F, G) is a Riemannian almost product manifold: If X ∈ ℑ10 (Mn ),

then X = X1 + X2 ,where X1 ∈ F + = D and X2 ∈ F − = D⊥ = J(D) one can write
X2 = J(X3 ), with X3 ∈ F + . Using this notation we obtain:

G(X,Y ) = G(X1 + JX3 ,Y1 + JY3 ) = G(X1 ,Y1 ) + G(JX3 , JY3 )

= G(X1 ,Y1 ) + G(X3 ,Y3 )

and

G(FX, FY ) = G(X1 − JX3 ,Y1 − JY3 ) = G(X1 ,Y1 ) + G(JX3 , JY3 )

= G(X1,Y1 ) + G(X3 ,Y3 )

thus proving G(X,Y ) = G(FX, FY ) .

3.3.3. Nonexistence of Para-Kähler-Norden Warped Metrics

Bishop and O’Neill [3] introduced warped product manifolds as follows: Sup-
pose (B, gB) and (F, gF ) are semi-Riemannian manifolds, and let f : B → (0, ∞)
be a smooth function. The warped product M = B × f F is the product manifold
B × F furnished with metric tensor
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

w
g = π∗ (gB ) + ( f ◦ π)2 σ∗ (gF ),

where π and σ are the projections of B × F onto B and F, respectively, and ∗

denotes the pull-back operator. Here, B is called the base of M, and F the fiber,
and f the warping function. Explicitly, if X and Y are tangent to B × F at (p, q),
then
w
g(X,Y ) = gB (π∗ X, π∗Y ) + f 2 (p)gF (σ∗ X, σ∗ X),
where π∗ and σ∗ are differentials of the projections π and σ respectively. If the
warping function f is constant , then the manifold B × f F is said to be trivial,
i.e. B × f F reduces to a semi-Riemannian product manifold with decomposable
metric g = π∗ (gB ) + σ∗ ( f 2 gF ).
On the other hand, the warped product manifold B × f F has a canonical
product structure ϕ ∈ ℑ11 (Mn ), ϕ2 = id. A Norden metric on B × f F is a pure
semi-Riemannian metric on B × f F with respect to the almost product structure
ϕ. The case, where ϕ is an almost paracomplex structure, we discuss the para-
Kähler-Norden conditions of warped metrics with respect to the structure ϕ, and
we prove the following theorem:

Theorem 60. Let dimB = dim F = k , and M2k = Bk × f Fk be a warped para-

complex manifold with metric w g. Then (Bk × f Fk ,w g, ϕ) is a para-Norden mani-
fold and there do not exist para-Kähler-Norden warped metric on this manifold.

Proof. Let dim B = dim F = k , and M2k = Bk × f Fk be a warped paracomplex

manifold with metric w g = π∗ (gB) + ( f ◦ π)2 σ∗ (gF ). The horizontal (vertical)
lift of X ∈ ℑ10 (Bk ) (of U ∈ ℑ10 (Fk )) to M2k = Bk × f Fk is the unique element
H X ∈ ℑ1 (B × F ) (V U ∈ ℑ1 (B × F ) ) that is π−related ( σ−related) to X (to
0 k f k 0 k f k
U ) and σ−related ( π−related) to the zero vector field on Fk (Bk ), i.e.

π∗ H X = X, σ∗ H X = 0 (π∗ V U = 0, σ∗ V U = U) (3.21)

Let X,Y ∈ ℑ10 (Bk) and U,W ∈ ℑ10 (Fk ). Then for Lie bracket of these vector
fields we have [59, p.25]

[H X,H Y ] =H [X,Y ], [V U,V W ] =V [U,W], [H X,V U] = 0. (3.22)

Putting, now
ϕ(H X) =H X, ϕ(V U) = −V U (3.23)
we have a canonical integrable paracomplex structure ϕ on Bk × f Fk .We put
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

e Ye ) =W g(ϕX,
G(X, e Ye) −W g(X,
e ϕYe )

e Ye ∈ ℑ1 (Bk × f Fk ). If G(X,
for any X, e Ye ) = 0 for all vector fields Xe and Ye which
0
H H V V
are of the form X, Y or U, W then G = 0. By virtue of (3.21) and (3.23),
we have

G(H X,H W ) = W
g(ϕH X,H Y ) −W g(H X, ϕH Y ) = 0,
G(H X,V W ) = W
g(ϕH X,V W ) −W g(H X, ϕV W ) = 2W g(H X,V W )
2
= gB (πH V H V
∗ X, π∗ W ) + f gF (σ∗ X, σ∗ W ) = 0,
G(V U,H Y ) = w
g(ϕV U,H Y ) −w g(V U, ϕH Y ) = 2w g(V U,H Y )
2
= gB (πV U, πH V H
∗ Y ) + f gF (σ∗ U, σ∗ Y ) = 0,
G(V U,V W ) = w
g(ϕV U,V W ) −w g(V U, ϕV W )
= −w g(V U,V W ) +w g(V U,V W ) = 0,

i.e. (Bk × f Fk ,w g, ϕ) is a para-Norden warped manifold.

We now define φϕ −operator applied to the para-Norden Walker metric w g

by
e Ye1 , Ye2 ) = (φX)(
(φϕ W g)(X, e W g(Ye1 , Ye2 )) − X
eW g(ϕYe1 , Ye2 )
+W g((Le ϕ)X,e Ye2 ) +W g(Ye1 , (Le ϕ)X) e
Y1 Y2

e Ye1 , Ye2 ∈ ℑ1 (Bk × f Fk ). Using (3.21), (3.22) and (3.23) we have

for any X, 0

(φϕ W g)(V U,H Y1 ,H Y2 )

= (ϕV U)W g(H Y1 ,H Y2 ) −V U W g(ϕH Y1 ,H Y2 ) +
+W g((LHY1 ϕ)V U,H Y2 ) +W g(H Y1 , (LHY2 V U)
= −2V U W g(HY1 ,H Y2 ) +W g(LHY1 (ϕV U)
−ϕ(LH Y1 V U),H Y2 ) +W g(HY1 , LH Y2 (ϕV U) − ϕ(LH Y2 V U))
= −2V U W g(HY1 ,H Y2 ) +W g(−LH Y1 V U − ϕ(LH Y1 V U),H Y2 ) +
+W g(H Y1 , −LH Y2 V U − ϕ(LH Y2 V U))
= −2V U W g(HY1 ,H Y2 ) = −2V UgB (πH H
∗ Y1 , π∗ Y2 ) = 0

for any U ∈ ℑ10 (Fk ) and Y1 , Y2 ∈ ℑ10 (Bk ).

By similar devices, we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(φϕ W g)(H X,H Y1 ,H Y2 ) = (φϕ W g)(H X,V U,H Y ) = (φϕ W g)(H X,H Y,V U)
= (φϕ W g)(V U,V W,H Y ) = (φϕ W g)(V U,H Y,V W )
= (φϕ W g)(V U,V W1 ,V W2 ) = 0.

In the case, where Xe =H X, Ye1 =V W1 , Ye2 =V W2 we obtain

(φϕW g)(H X,V W1 ,V W2 )

= 2H X W g(V W1 ,V W2 ) +W g((LV W1 ϕ)H X,V W2 )
+W g((V W1 , LV W2 ϕ)H X)
= 2H X W g(V W1 ,V W2 ) +W g(LV W1 H X − ϕ(LV W1 H X),V W2 )
+W g(V W2 , LV W2 H X − ϕ(LV W2 H X))
= 2H X(( f ◦ π)2 gF (σ∗V W1 , σ∗V W2 ))
= 2(H X( f ◦ π)2 )gF = 4 f (X f )gF ,

from which we see that the equations φϕW g = 0 and f =constant ( f > 0) are
equivalent. On the other hand, as stated in Theorem 44, a para-Norden metric

3.4. Dual-Kähler-Norden Manifolds

Let Am − C(ε)(m = 2) be a dual algebra. In this section we shall only consider
C(ε)-holomorphic Norden (i.e. dual-Kähler-Norden) manifolds.
We suppose that the manifold M2n is the tangent bundle π : T (Vn ) → Vn of
a Riemannian manifold Vn . If (u1 , u2 , ..., un) are local coordinates on Vn , then
xi = ui ◦ π together with the fibre coordinates xi = yi , i = n + 1, ..., 2n form local
coordinates on T (Vn). It is well known that there exists a tensor field of type
(1, 1) which has components of the form

0 0
γ= (3.24)
I 0

with respect to the induced coordinates (xi , xi ) in T (Vn), I being unit matrix
in Mn and γ satisfies γ2 = 0.Thus T (Vn ) has a natural integrable regular dual
γ−structure (see Example 8).
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

If X is a vector field on Vn , its vertical lift V X on T (Vn ) is the vector field

defined by V X(ιω) = ω(X) ◦ π =V (ω(X)), where ω is a 1-form on Vn, ιω is
regarded as a function on T (Vn), which has local expression

ιω = ωi xi .

For a vector field X on Vn , the complete lift C X of X is defined by C X C f =C (X f ),

where C f = ι(d f ) = xi ∂i f , f ∈ ℑ00 (Vn). V X and C X have, respectively local
expression of the form

V ∂ ∂ ∂
X = Xi , C
X = Xi i
+ xS ∂S X i (3.25)
∂xi ∂x ∂xi

with respect to the induced coordinates (xi , xi ) in T (Vn). From (3.24) and (3.25)
we have
γV X = 0, γC X =V X. (3.26)
Since a tensor field of type (0, q) (or (1, q) ) on T (Vn) completely determined
by its action on complete lifts of vector field [114, p.33], the complete lift C g of

the Riemennian metric g is defined by

C
g(C X,C Y ) =C (g(X,Y )),

which is a Riemannian metric on T (Vn ) too. C g satisfies the equation

C
g(V X,C Y ) =C g(C X,V Y ) =V (g(X,Y)). (3.27)

From (3.26) and (3.27), we have

C
g(γC X,C Y ) =C g(V X,C Y ) =V (g(X,Y )),
C
g(C X, γCY ) =C g(C X,V Y ) =V (g(X,Y )),

i.e. C g is a pue metric with respect to γ. Thus T (Vn ), γ,C g is a plural(dual)-
Norden manifold.
The complete lift C ∇ of the Levi – Civita connection ∇ is defined by

∇C X C Y =C (∇X Y ), (3.28)

which is a Levi Civita connection of C g and satisfies the equation

C
∇V X C Y =C ∇C X V Y =V (∇X Y ), ∀X,Y ∈ ℑ10 (Vn ). (3.29)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Using (3.26), (3.28) and (3.29), we have for any X,Y ∈ ℑ10 (Vn)

(C ∇γ)(CY,C X) = C
∇C X (γCY ) − γ(C ∇C X C Y )
C
= ∇C X V Y − γ C (∇X Y )
V
= (∇X Y ) −V (∇X Y ) = 0,

from which we see that C ∇γ = 0. Then φCγ g = 0 by virtue of Theorem 44. Thus
we have

Theorem 61. (T (Vn ), γ,C g) is a dual-Kähler-Norden manifold.

Remark 18. By virtue of Detγ = 0 and G(X,Y ) = g(γX,Y ) = (g ◦ γ)(X,Y), it

follows that a twin dual-Norden metric G is non-Riemannian, but it is symmet-
ric, i.e. G is a degenerate metric.

Let now G be a Riemannian metric I + II in the tangent bundle (see [114, p.

139]), i.e. G =C g +V g , where V g is a vertical lift of the Riemannian metric g.
By similar devices, we have from (3.26) and (3.27)

G(γC X,C Y ) = C
g(γC X,C Y ) +V g(γC X,C Y )
C V C
= g( X, Y ) +V g(V X,C Y )
V
= (g(X,Y )),

G(C X, γC Y ) = C
g(C X, γCY ) +V g(C X, γCY )
C C V
= g( X, Y ) +V g(C X,V Y )
V
= (g(X,Y )),

i.e. (T (Mn ), γ, G) is also a dual-Norden manifold.

Since the Levi-Civita connection of the metric C g and the metric G = C g +V
g coincide [14, p.149], from the equation C ∇γ = 0 it follows that φγ G = 0. From
this, we have

Theorem 62. (T (Mn ), γ, G) is a dual-Kähler-Norden manifold.

By similar devices, we can prove that is also (T (Mn ), γ,C g) is a dual-Kähler-

Norden manifold, where C g =S g +V a (V a− vertical lift of a symmetric tensor
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

field a ∈ T20 (Mn )) is a synectic lift of g [16], [99].

Remark 19. Let now T 2 (Mn ) be a tangent bundle of order 2 over Mn . It is also
well know there exists a affinor field bγ ∈ ℑ11 (T 2 (Mn )) which has components of
the form  
0 0 0
bγ =  I 0 0  , bγ3 = 0
0 I 0
with respect to the induced coordinates (xi , xn+i , x2n+i) in T 2 (M 2
n n ), i.e.oT (Mn )
2
has a natural integrable regular plural Π−structure: Π = I, bγ, bγ , which
is a isomorphic repersentation of the algebra of plural numbers R(1, ε, ε2 ),
ε3 = 0 [106], [108]. The 2−nd lift of g, i.e. CC g =H g (see [114, p.332])
is a plural-Norden metric with respect to bγ and CC ∇C g = 0, where CC ∇ de-
note the 2−nd lift of the Levi-Civita connection ∇ which is necessarily the
Levi-Civita connection determined by CC g. Thus, (T 2 (Mn ), Π,CC g) is a plural-
Kähler-Norden manifold.

3.5. Norden–Hessian Structures

Let (M2n , g) now be a Riemannian manifold with a metric tensor g. The gradient
grad f of a function f ∈ ℑ00 (M2n ) is the vector field metrically equivalent to the
differential d f ∈ ℑ01 (M2n). In terms of a coordinate system
grad f = (gi j ∂i f )∂ j .
Thus
g(grad f , X) = gi j (∂i f )X k g jk = X f = (d f )(X).
The Hessian of a function f ∈ ℑ00 (M2n ) is its second covariant differential h =
∇(∇ f ) = ∇2 f with respect to the Levi-Civita connection of g, i.e. since ∇Y f =
Y f = (d f )(Y ),
h(Y, X) = (∇(∇ f ))(Y, X) = (∇(d f ))(Y, X)
= X((d f )(Y )) − (d f )(∇X Y ) = XY f − (∇X Y ) f .
We easily see that h is symmetric tensor field. Also, we have
g(∇X (grad f ),Y ) = h(Y, X).
For the natural coordinates in Euclidean space, the components of h are just the
2
second partials ∂x∂i ∂xf j .
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

If the differentiable function f : M2n → R on a Riemannian manifold

(M2n , g) is convex (strictly), then the Hessian h = ∇2 f is indefinite (positive
definite) [24], [32]. Hence, h = ∇2 f defines a new metric on M2n if h is nonde-
generate, and is called a pseudo-Riemannian Hessian metric.
p
Let Γi j be the Christoffel symbols and Rm i jk be the components of the cur-
vature tensor fields produced by the Riemannian metric g. If h pk are the con-
travariant components of the pseudo-Riemannian Hessian metric h, then the
components of Levi-Civita connection h ∇ of h are given by the following for-
mula
e p = Γ p + 1 h pk [(∇i ∇ j ∇k f ) + (Rm + Rm )∇m f ].
Γij ij ik j jki
2
Let (M2n, g, ϕ) be a Kähler (para-Kähler-Norden) manifold. If there exist a func-
∗ ∗
tion f on a Kähler (para-Kähler-Norden) manifold such that d f ◦ ϕ = d f for a
function f , then we shall call f a holomorphic (para-holomorphic) function and
∗
f its associated function (see Section 3.1). If such a function f is defined locally,
then we call it a locally holomorphic (para-holomorphic) function.

Remark 20. If (M2n, ϕ) is a complex manifold, then in terms of a real coor-

∗
dinates (xi , xi ), i = 1, ..., n; i = n + 1, ..., 2n the equation d f ◦ ϕ = d f reduces
to 
 ∗
∂i f = ∂i f ,
 ∂ f = −∂ ∗f ,
i i
∗
which is the Cauchy-Riemann equations for the complex function F = f + i f
(see [34, p. 122]).

We notice that the condition for f to be locally holomorphic (para-

holomorphic) also is given by
(φϕ d f )i j = ϕm m m
i ∂m ∂ j f − ∂i (ϕ j ∂m f ) + (∂ j ϕi )∂m f = 0.

If we assume that f is holomorphic (para-holomorphic), then, from (1.13), we

have

(φϕ (d f ))(X,Y ) = (ϕX)((d f )(Y)) − X((d f )(ϕY )) + (d f )((LY ϕ)(X)) (3.30)

= ϕ(X)((d f )(Y)) − X((d f )(ϕY )) + (d f )([Y, ϕX] − ϕ([Y, X]))
= ϕ(X)((d f )(Y)) − X((d f )(ϕY ))
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

+(d f )(∇Y ϕX − ∇ϕX Y − ϕ(∇Y X − ∇X Y ))

= (∇ϕX d f )(Y ) − (∇X d f )(ϕY ) + (d f )(∇ϕX Y )
−(d f )(∇X ϕY ) + (d f )((∇ϕ)(X,Y ) − ∇ϕX Y + ϕ(∇X Y )
= (∇ϕX d f )(Y ) − (∇X d f )(ϕY ) − (∇ϕ)(Y, X) = 0.
We now consider a holomorphic (para-holomorphic) function f on a Kähler
(para-Kähler-Norden) manifold (M2n , g, ϕ) . On a Kähler (para-Kähler-Norden)
manifold (M2n , g, ϕ) (∇ϕ = 0), the equation (3.30) equivalent to the equation
(∇2 f )(Y, ϕX) = (∇2 f )(ϕY, X),
i.e. h = ∇2 f is pure and a manifold (M2n , ϕ, h = ∇2 f ) is a Norden manifold.
Thus, h naturally defines a Norden metric on the Kähler (para-Kähler-Norden)
manifold (M2n , g, ϕ) . We call it Norden-Hessian metric. Thus we have

Theorem 63. Let (M2n , g, ϕ) be a Kähler (para-Kähler-Norden) manifold.

Then, M2n admits a Norden-Hessian structure (ϕ, h = ∇2 f ), if f ∈ ℑ00 (M2n)
is holomorphic (para-holomorphic).

Let (M2n , g, J) be a locally decomposable Riemannian manifold with inte-

grable paracomplex structure

E 0
J= ,
0 −E

being (n × n)-unit matrix . In such manifolds, g is pure with respect to J, more-

over ∇J = 0, i.e. a triple (M2n , g, J) is a para-Kähler-Norden manifold. Also g
is para-holomorphic and the curvature tensor field R of g is pure with respect
to the structure J. Let (M2n , h = ∇2 f , J) be a Hessian-Norden structure, which
exists on a paracomplex decomposable Riemannian manifold. Then

(∇2 f )(JX,Y ) = (∇2 f )(X, JY ),

from which we have

(∇3 f )(JX,Y, Z) = (∇3 f )(X, JY, Z). (3.31)

Using the Ricci equation, from (3.31) we obtain

(∇3 f )(X, JY, Z) = ∇Z (∇JY (∇X f ) (3.32)

= ∇JY (∇Z (∇X f )) − (d f )(R(Z, JY )X)

= (∇3 f )(X, Z, JY ) − (d f )(R(Z, JY )X)

and

(∇3 f )(JX,Y, Z) = ∇Z (∇Y (∇JX f ) (3.33)

= ∇Y (∇Z (∇JX f )) − (d f )(R(Z,Y )JX)
= (∇3 f )(JX, Z,Y ) − (d f )(R(Z,Y )JX).

Since h is symmetric and the curvature tensor R of g is pure with respect to J,

from (3.32) and (3.33) we have

(∇3 f )(Z, JX,Y ) = (∇3 f )(Z, X, JY ), (3.34)

i.e. a tensor field ∇3 f is pure in all arguments. On the other hand

(φJ h)(X, Z1 , Z2 ) (3.35)

= J(X)(h(Z1 , Z2 )) − X(h(JZ1 , Z2 ))
−h(∇JX Z1 , Z2 ) + h((∇J)(X, Z1 ), Z2 ) + h(Z1 , (∇J)(X, Z2 ))
−h(Z1 , ∇JX Z2 ) + h(J(∇X Z1 ), Z2 ) + h(JZ1 , ∇X Z2 )
= (∇JX h)(Z1 , Z2 ) − (∇X h)(JZ1 , Z2 ) + h((∇J)(X, Z1 ), Z2 )
+h(Z1 , (∇J)(X, Z2 )).

Substituting h(Z1 , Z2 ) = ∇Z1 ∇Z2 f and ∇J = 0 in (3.35), by virtue of (3.34) we

have

(φJ h)(X, Z1, Z2 ) = (φJ h)(X, Z2 , Z1 )

= (∇JX (∇2 f ))(Z2 , Z1 ) − (∇X (∇2 f ))(Z2 , JZ1 )
= (∇3 f )(Z2 , Z1 , JX) − (∇3 f )(Z2 , JZ1 , X) = 0,

i.e. h is para-holomorphic. Then, using Theorem 44, we see that h ∇J = 0. Thus,

we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 64. Let (M2n , g, J) be a paracomplex decomposable Riemannian

manifold. If f is paraholomorphic, then a triple (M2n, h = ∇2 f , J) is a para-
Kähler-Norden-Hessian manifold.

Remark 21. For Kähler manifold (M2n, g, J), the curvature tensor R of the Her-
mitian metric g is not pure in all arguments. Therefore, Kähler manifolds may
not always locally admit any Kähler-Norden-Hessian metric.

3.6. Norden-Walker Manifolds with Proper Structures

In the present section, we shall focus our attention to Norden manifolds of di-
mension four. The main purpose of the present section is to study complex
Norden metrics on 4-dimensional Walker manifolds.
A neutral metric g on a 4-manifold M4 is said to be a Walker metric if there
exists a 2-dimensional null distribution D on M4 , which is parallel with respect

to g. From Walker’s theorem [110], there is a system of coordinates (x, y, z,t)

with respect to g which takes the following local canonical form
 
0 0 1 0
 0 0 0 1 
g = (gi j ) = 
 1 0 a c ,
 (3.36)
0 1 c b

where a, b, c are smooth functions of the coordinates (x, y, z,t). The paralel null
∂
2-plane D is spanned locally by {∂x , ∂y }, where ∂x = ∂x , ∂y = ∂∂y .

3.6.1. Almost Norden-Walker Metrics

In [47], a proper almost complex structure with respect to g is defined as a g-
orthogonal almost complex structure J so that J is a standard generator of a
positive π2 rotation on D, i.e., J∂x = ∂y and J∂y = −∂x . Then for the Walker
metric g, such a proper almost complex structure J is determined uniquely as
 1 
0 −1 −c 2 (a − b)
 1 0 1 (a − b) c 
 2 . (3.37)
 0 0 0 −1 
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

0 0 1 0

In [5], for such a proper almost complex structure J on Walker 4-manifold M,

an almost Norden structure (gN+, J) is constructed, where gN+ is a metric on
M, with properties gN+ (JX, JY ) = −gN+ (X,Y ). In fact, as one of these exam-
ples, such a metric takes the form (see Proposition 6 in [5], unfortunately, the
calculations of the component gN+
44 in [5] are erroneous):
 
0 −2 0 −b
 −2 0 −a −2c 
gN+ = 
 0 1
.

−a 0 2 (1 − ab)
1
−b −2c 2 (1 − ab) −2bc

We may call this an almost Norden-Walker metric. The construction of such a

structure in [5] is to find a Norden metric for a given almost complex structure,
which is different form the Walker metric.
The purpose of the present section is to find also an almost Norden-Walker
structure (g, F), where the metric is nothing but the Walker metric g, with an

appropriate almost complex structure F, to be determined. That is, for a fixed

metric g, we will find an almost complex structure F which satisfy g(FX, FY ) =
−g(X,Y ).
In [5], for a given almost complex structure, a metric is constructed. Our
method is, however, for a given metric, an almost complex structure is con-
structed.
Let F be an almost complex structure on a Walker manifold M4 , which
satisfies
(i) F 2 = −I,
(ii) g(FX,Y ) = g(X, FY ) (Nordenian property),
(iii) F∂x = ∂y , F∂y = −∂x ( F induces a positive π2 −rotation on D).
We easily see that these three properties define F non-uniquely, i.e.,


 F∂x = ∂y ,

F∂y = −∂x ,
1
 F∂z = α∂x + 2 (b + a)∂y − ∂t ,

 1
F∂t = − 2 (b + a)∂x + α∂y + ∂z

and has the local components

 
0 −1 α − 12 (a + b)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

 1
α 
F = (Fji ) = 
1 0 2 (a + b) 
 0 0 0 1 
0 0 −1 0

with respect to the natural frame {∂x , ∂y , ∂y , ∂t }, where α = α(x, y, z,t) is an

arbitrary function.
We must note that the proper almost complex structure J as in (3.37) is deter-
mined uniquely. In our case of the almost Norden-Walker structure, the almost
complex structure F just obtained contains an arbitrary function α(x, y, z,t).
Our purpose is to find a nontrivial almost Norden-Walker structure with the
Walker metric g explicitly (see Theorem 66 and Example 12). Therefore, we
now put α = c. Then g defines a unique almost complex structure
 
0 −1 c − 21 (a + b)
 1 0 1 (a + b) c 
ϕ = (ϕij ) = 
 0 0
2 
 (3.38)
0 1
0 0 −1 0

The triple (M4 , ϕ, g) is called almost Norden-Walker manifold. In conformity

with the terminology of [12], [13], [43], [44], [45], [46], [47], [48] we call ϕ the
proper almost complex structure.

Remark 22. From (3.38) we immediately see that in the case a = −b and c = 0,
ϕ is integrable.

3.6.2. Integrability of Proper Almost Complex Structures

We consider the general case for integrability.
The almost complex structure ϕ on almost Norden-Walker manifolds is in-
tegrable if and only if

(Nϕ )ijk = ϕmj ∂m ϕik − ϕm i i m i m

k ∂m ϕ j − ϕm ∂ j ϕk + ϕm ∂k ϕ j = 0. (3.39)

From (3.38) and (3.39) find the following integrability condition:

Theorem 65. The proper almost complex structure on almost Norden-Walker

manifolds is integrable if and only if the following PDEs hold:

ax + bx + 2cy = 0,
(3.40)
ay + by − 2cx = 0.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

From this theorem, we easily see that if a = −b and c = 0, then ϕ is inte-

grable (see Remark 22).
Let (M4 , ϕ, g) be a Norden-Walker manifolds (Nϕ = 0) and a = b. Then the
equation (3.40) reduces to
ax = −cy ,
(3.41)
ay = cx ,
from which follows

axx + ayy = 0, (3.42)

cxx + cyy = 0,

e.g. the functions a and c are harmonic with respect to the arguments x and y.
Thus we have

Theorem 66. If the triple (M4 , ϕ, g) is Norden-Walker and a = b, then a and c

are all harmonic with respect to the arguments x, y .

Example 12. We now apply the Theorem 66 to establish the existence of special
types of Norden-Walker metrics. In our arguments, the harmonic function plays
an important part. Let a = b and h(x, y) be a harmonic function of variables x
and y, for example h(x, y) = ex cosy. We put

a = a(x, y, z,t) = h(x, y) + α(z,t) = ex cos y + α(z,t)

where α is an arbitrary smooth function of z and t. Then, a is also hormonic

with respect to x and y. We have

ax = ex cos y,
ay = −ex siny.

From (3.41), we have PDEs for c to satisfy as

cx = ay = −ex sinx,
cy = −ax = −ex cosy.

For these PDEs, we have solutions

c = −ex siny + β(z,t),

where β is arbitrary smooth function of z and t. Thus the Norden-Walker metric

has components of the form
 
0 0 1 0
 0 0 0 1 
g = (gi j ) = 
 1 0 ex cos y + α(z,t) −ex siny + β(z,t)  .


0 1 −ex siny + β(z,t) ex cos y + α(z,t)

3.6.3. Holomorphic Norden-Walker (Kähler-Norden-Walker)

Metrics
Let (M4 , ϕ, g) be an almost Norden-Walker manifold. If

(Φϕ g)ki j = ϕm m m m m
k ∂m gi j − ϕi ∂k gm j − gm j (∂i ϕk − ∂k ϕi ) + gim ∂ j ϕk = 0, (3.43)

then by virtue of Theorem 44 ϕ is integrable and the triple (M4 , ϕ, g) is called

a holomorphic Norden-Walker or a Kähler-Norden-Walker manifold. Taking

account of Corollary 45, we see that almost Kähler-Norden-Walker manifold

with condition Φϕ g = 0 and Nϕ 6= 0 does not exist.
Substituting (3.36) and (3.38) in (3.43), we see that the non-vanishing com-
ponents of Φϕ g are

1
(Φϕg)xzz = ay, (Φϕ g)xzt = (Φϕg)xtz = (bx − ax ) + cy , (3.44)
2
(Φϕ g)xtt = by − 2cx , (Φϕ g)yzz = −ax ,
1
(Φϕ g)yzt = (Φϕg)ytz = (by − ay ) − cx , (Φϕ g)ytt = −bx − 2cy ,
2
(Φϕg)zxz = (Φϕg)zzx = (Φϕg)txt = (Φϕ g)ttx = cx ,
1
(Φϕ g)zxt = (Φϕg)ztx = −(Φϕg)txz = −(Φϕg)tzx = (ax + by ),
2
(Φϕg)zyz = (Φϕg)zzy = (Φϕg)tyt = (Φϕ g)tty = cy ,
1
(Φϕ g)zyt = (Φϕg)zty = −(Φϕ g)tyz = −(Φϕ g)tzy = (ax + by ),
2
1
(Φϕg)zzz = cax − at + 2cz + (a + b)ay ,
2
1
(Φϕg)zzt = (Φϕg)ztz = ccx + bz + (a + b)cy ,
2
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1
(Φϕ g)ztt = cbx + at − 2cz + (a + b)by ,
2
1
(Φϕg)tzz = cay − bz − (a + b)ax,
2
1
(Φϕ g)tzt = (Φϕg)ttz = ccy − at + 2cz − (a + b)cx ,
2
1
(Φϕg)ttt = cby + bz − (a + b)bx.
2
From these equations we have

Theorem 67. The triple (M4 , ϕ, g) is Kähler-Norden-Walker if and only if the

following PDEs hold:

ax = ay = cx = cy = bx = by = bz = 0, at − 2cz = 0. (3.45)

Corollary 68. The triple (M4 , ϕ, g) with metric

 
0 0 1 0
 0 0 0 1 
g = (gi j ) = 
 1

0 a(z) 0 
0 1 0 b(t)
is always Kähler-Norden-Walker.

3.6.4. Curvature Properties of Norden-Walker Manifolds

If R and r are respectively the curvature and the scalar curvature of the Walker
metric, then the non-vanishing components of R and r have, respectively, ex-
pressions (see [47], Appendix A and C)

1 1 1
Rxzxz = − axx, Rxzxt = − cxx , Rxzyz = − axy, (3.46)
2 2 2
1 1 1 1 1
Rxzyt = − cxy , Rxzzt = axt − cxz − ay bx + cx cy ,
2 2 2 4 4
1 1 1
Rxtxt = − bxx, Rxtyz = − cxy , Rxtyt = − bxy ,
2 2 2
1 1 1 2 1 1 1
Rxtzt = cxt − bxz − (cx ) + ax bx − bxcy + by cx ,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

2 2 4 4 4 4
1 1 1 1 1 1
Ryzyz = − ayy, Ryzyt = − cyy , Rytzt = cyt − byz − cx cy + ay bx ,
2 2 2 2 4 4
1 1 1 1 1 1 2 1
Ryzzt = ayt − cyz − ax cy + ay cx − ayby + (cy ) , Rytyt = − byy,
2 2 4 4 4 4 2
1 1 1 1 1 1
Rztzt = czt − att − bzz − a(cx )2 + aaxbx + caxby − ccx cy
2 2 4 4 4 2
1 1 1 1 1 1
− at cx + ax ct − ax bz + cay bx + bay by − b(cy)2
2 2 4 4 4 4
1 1 1 1
− bz cy + ay bt + by cz − at by
2 4 2 4
and
r = axx + 2cxy + byy. (3.47)
Suppose that the triple (M4 , ϕ, g) is Kähler-Norden-Walker. Then from the last
equation in (3.45) and (3.46), we see that
1 1
Rztzt = czt − att = − (at − 2cz )t = 0.
2 2

From (3.45) we easily we see that the another components of in (3.46) directly
all vanish. Thus we have

Theorem 69. If a Norden-Walker manifold (M4 , ϕ, g) is Kähler-Norden-Walker,

then M4 is flat.

Remark 23. In general, a Kähler-Norden manifold is non-flat and in such man-

ifold curvature tensor is pure and holomorphic. In particular, from Theorem 69
we see that a Kähler-Norden-Walker manifold is flat, and therefore g is Einstein.

Let (M4 , ϕ, g) be a Norden-Walker manifold with the integrable proper

structure ϕ, i.e. Nϕ = 0. If a = b, then from proof of the Theorem 66 we
see that the equation (3.41) hold. If c = c(y, z,t) and c = c(x, z,t) , then
cxy = (cx )y = (cy )x = 0. In these cases, by virtue of (3.41) we find a = a(x, z,t)
and a = a(y, z,t), respectively. Using of cxy = 0 and axx + byy = 0 (see (3.42)),
we from (3.47) obtain r = 0. Thus we have

Theorem 70. If (M4 , ϕ, g) is a Norden-Walker non-Kähler manifold with met-

rics
   
0 0 1 0 0 0 1 0
 0 0 0 1   0 0 0 1 
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

g=  
 1 0 a(x, z,t) c(y, z,t)  , ge =  1 0 a(y, z,t) c(x, z,t) 


0 1 c(y, z,t) a(x, z,t) 0 1 c(x, z,t) a(y, z,t)

then M4 is scalar flat.

3.6.5. Isotropic Kähler-Norden-Walker Structures

It is well known that the inner product in the vector space can be extended to an
inner product in the tensor space. In fact, if t and t are tensors of type (r, s) with
1 2
components t i1 ...ir and t k1 ...kr , then
1 j1 ... js 2l1 ...ls

g(t , t ) = gi1 k1 ...girkr g j1 l1 ...g jsls t i1 ...ir t k1 ...kr .

1 2 1 j1 ... js 2l1 ...ls

If t = t = ∇ϕ ∈ ℑ12 (M2n ), then the square norm k∇ϕk2 of ∇ϕ is defined by

1 2

k∇ϕk2 = gi j gkl gms (∇ϕ)m s

ik (∇ϕ) jl .

A proper almost complex structure ϕ on Norden-Walker manifold (M4 , ϕ, g) is

said to be isotropic Kähler if k∇ϕk2 = 0, but ∇ϕ 6= 0. Examples of isotropic
Kähler structures were given in [4], [20]. Our purpose in this section is to
show that a proper almost complex structure on almost Norden-Walker mani-
fold (M4 , ϕ, g) is isotropic Kähler as we will see Theorem 71.
The inverse of the metric tensor (3.36), g−1 = (gi j ), given by
 
−a −c 1 0
 −c −b 0 1 
g−1 =   1
. (3.48)
0 0 0 
0 1 0 0

For the covariant derivative ∇ϕ of the almost complex structure put (∇ϕ)kij =
∇i ϕkj . Then, after some calculations we see that the non-vanishing components
of ∇ϕ are
y y
∇x ϕxz = ∇x ϕt = cx , ∇y ϕxz = ∇y ϕt = cy , (3.49)
1 1
∇z ϕxx = −∇z ϕyy = ∇z ϕzz = −∇z ϕtt = ay + cx ,
2 2
1 1
∇z ϕyx = x t z
∇z ϕy = ∇z ϕz = ∇z ϕt = − ax + cy ,
2 2
1 1 1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

∇x ϕxz = 2cz + cax − at − ccy − acx + bcy ,

2 2 2
1 1 3 1
∇z ϕxz = az + acy − bcy + cay + aax + bax ,
4 4 4 4
1 1 3 1
∇z ϕtx = aax − bax + cay + bcy + ccx + acy ,
4 4 4 4
y 1 1 1 1
∇z ϕt = 2cz + ccy − at + bay + cax − acx ,
2 2 2 2
1 1
∇t ϕxx = y z t
−∇t ϕy = ∇t ϕz = −∇t ϕt = cy + bx ,
2 2
1 1
∇t ϕyx = ∇t ϕxy = ∇t ϕtz = ∇t ϕtz = − cx + by ,
2 2
3 1 1 1
∇t ϕxz = ccx + bz − cby − abx + bcy ,
2 2 2 2
1 1 1 1
∇t ϕyz = aby − bby − acx + bcx ,
4 4 4 4
1 1 1 1
∇t ϕtx = acx − bcx + ccx + bby + cbx − aby ,
4 4 4 4
1 1 1 1
∇t ϕty = cby + bz + bcy + ccx − abx .
2 2 2 2

Using (3.36), (3.48) and (3.49) we find

k∇ϕk2 = gi j gkl gms (∇ϕ)m s

ik (∇ϕ) jl = 0.

Thus we have

Theorem 71. A proper almost complex structure on almost Norden-Walker

manifold (M4 , ϕ, g) is isotropic Kähler.

3.6.6. Quasi-Kähler-Norden-Walker Structures

The basis class of non-integrable almost complex manifolds with Norden met-
ric is the class of the quasi-Kähler manifolds. An almost Norden manifold
(M2n , ϕ, g) is called a quasi- Kähler [42], if

σ g((∇X ϕ)Y, Z) = 0,
X,Y,Z

where σ is the cyclic sum by three arguments.

If we add (φϕ g)(X,Y, Z) and (φϕ g)(Z,Y, X) (see (3.6) or (3.7)), then by
virtue of g(Z, (∇Y ϕ)X) = g((∇Y ϕ)Z, X), we find
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(φϕ g)(X,Y, Z) + (φϕ g)(Z,Y, X) = 2g((∇Y ϕ)Z, X).

Since, from last equation we have

(φϕ g)(X,Y, Z) + (φϕ g)(Y, Z, X) + (φϕ g)(Z, X,Y ) = σ g((∇X ϕ)Y, Z).
X,Y,Z

Thus we have

Theorem 72. Let (M2n , ϕ, g) be an almost Norden manifold. Then the Norden
metric g is a quasi-Kähler-Norden if and only if

(Φϕg)(X,Y, Z) + (Φϕg)(Y, Z, X) + (Φϕg)(Z, X,Y ) = 0

for any X,Y, Z ∈ ℑ10 (M2n ).

The equation in Theorem 72 can also be written in the form

(Φϕ g)(X,Y, Z) + 2g((∇X ϕ)Y, Z) = 0.

From here we see that, if we take a local coordinate system, then a Norden-
Walker manifold (M4 , ϕ, g) satisfying the condition Φk gi j + 2∇k Gi j to be zero
is called a quasi-Kähler manifold, where G is defined by Gi j = ϕmi gm j . For the
covariant derivative ∇G of the twin metric G put (∇G)i jk = ∇i G jk . After some
calculations we see that the non-vanishing components of ∇G are

∇x Gzz = ∇x Gtt = cx , ∇y Gzz = ∇y Gtt = cy , (3.50)

1 1
∇z Gxz = ∇z Gzx = −∇z Gyt = −∇z Gty = ay + cx ,
2 2
1 1
∇z Gxt = ∇z Gtx = ∇z Gyz = ∇z Gzy = cy − ax ,
2 2
1
∇z Gzz = 2cz − at + ay(a + b) + cax ,
2
1 1 1 1
∇z Gzt = ∇z Gtz = cay + ccx − ax (a + b) + cy (a + b),
2 2 4 4
1
∇z Gtt = 2cz − at + ccy − cx (a + b),
2
1 1
∇t Gxz = ∇t Gzx = −∇t Gyt = −∇t Gty = bx + cy ,
2 2
1 1
∇t Gxt = ∇t Gtx = ∇t Gyz = ∇t Gzy = by − cx ,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

2 2
1 1 1 1
∇t Gzt = ∇t Gtz = ccy + cbx − cx (a + b) + by (a + b),
2 2 4 4
1
∇t Gzz = bz + ccx + cy (a + b),
2
1
∇t Gtt = bz + cby − bx (a + b).
2
From (3.44) and (3.50) we have

Theorem 73. A triple (M4 , ϕ, g) is a quasi-Kähler-Norden-Walker manifold if

and only if the following PDEs hold:

bx = by = bz = 0, ay − 2cx = 0, ax − 2cy = 0, cax − at + 2cz − (a + b)cx = 0.

3.6.7. On the Goldberg Conjecture

Let now (M4 , ϕ, g) be an almost Hermitian manifold. The Goldberg conjecture
(see [48]) states that an almost Hermitian manifold (M4 , ϕ, g) must be Kähler

(or ϕ must be integrable) if the following three conditions are imposed: (G1 )
if M is compact and (G2 ) g is Einstein, and (G3 ) if the fundamental 2-form is
closed.
It should be noted that no progress has been made on the Goldberg conjec-
ture, and the orginal conjecture is still an open problem. Despite many papers
by various authors concerning the Goldberg conjecture, there are only Sekigawa
papers (see for example [90]) which obtained substantial results to the orginal
Goldberg conjecture: Let be (M4 , ϕ, g) be an almost Hermitian manifold, which
satisfies the three conditions (G1 ),(G2 ) and (G3 ). If the scalar curvature of M is
nonnegative, then ϕ must be integrable.
Let (M4 , ϕ,w g) be an indefinite almost Kähler-Walker-Einstein compact
manifold with the proper almost complex structure (3.37). As noted before,
many examples of Norden-Walker metrics can be obtained by gN+ (JX, JY ) =
−gN+ (X,Y ) (see Section 3.7.1), and as one of these examples, such a metric
has components
 
0 −2 0 −b
 −2 0 −a −2c 
gN+ =  0 1
.

−a 0 2 (1 − ab)
1
−b −2c 2 (1 − ab) −2bc
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

with respect to the Walker coordinates. Using Corollary 45, we have

Theorem 74. The proper almost complex structure ϕ on indefinite almost
Kähler-Walker-Einstein compact manifold (M4 , ϕ,w g) is integrable if φϕ gN+ =
0, where gN+ is the induced Norden-Walker metric on M4 .
This resolves a conjecture of Goldberg under the additional restriction on
Norden-Walker metric( gN+ ∈ Kerφϕ ).
Example 13. (Counterexample of noncompact and neutral type to the Gold-
berg conjecture). Let (M4 , ϕ, g) be an almost Norden-Walker manifold. Con-
sider the metric
 
0 0 1 0
 0 0 0 1 
g = (gi j ) = 
 1 0 a(x, y, z,t)
.

0
0 1 0 a(x, y, z,t)
That is the metric is defined by putting a = b, c = 0 in the generic canonical
form (3.36).

Let Ri j and S denote the Ricci curvature and the scalar curvature of the
metric g in (3.36). The Einstein tensor is defined by Gi j = Ri j − 41 Sgi j and has
non zero components as follows (see [47], Appendix D):

1 1 1 1
Gxz = axx − byy, Gxt = cxx + bxy, (3.51)
4 4 2 2
1 1 1 1
Gyz = axy + cyy, Gyt = byy − axx ,
2 2 4 4
1 1 1 1
Gzz = aaxx + caxy + bayy − ayt + cyz − aycx + ax cy
4 2 2 2
1 1 1 1
+ ayby − (cy)2 − acxy − abyy,
2 2 2 4
1 1 1 1 1 1 1
Gzz = acxx + ccxy + axt − cxz − ay bx + cx cy + bcyy
2 2 2 2 2 2 2
1 1 1 1
− cyt + byz − caxx − cbyy,
2 2 4 4
1 1 1 1
Gtt = abxx + cbxy + cxt − bxz − (cx )2 + axbx − bxcy
2 2 2 2
1 1 1 1
+ bycx + bbyy − baxx − bcxy.
2 4 4 2
The metric g in (3.36) is almost Norden-Walker-Einstein if all the above
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

components Gi j vanish ( Gi j = 0 ). In this case, we see from (3.51) that the

Einstein condition consist of the following PDE’s :
aaxx − 2ayt + (ay )2 = 0, axx − ayy = 0, axy = 0,
axt − ax ay + azy = 0, aaxx − 2axz + (ax )2 = 0.
If a is independent of y and t, and if a contains x only linearly, the first four
PDE’s hold trivially, and the last one reduces to: 2axz − (ax)2 = 0. We see that
a = − 2xz is a solution to the PDE, and therefore the metric
 
0 0 1 0
 0 0 0 1 
g = (gi j ) = 
 1 0 − 2x
 (3.52)
z 0 
0 1 0 − 20x
z

is Einstein on the coordinate patch z > 0 (or z < 0). Thus, the second con-
dition (G2 ) of Goldberg conjecture holds. We know that this metric admits a
proper almost complex structure as follows:

ϕ∂x = ∂y , ϕ∂y = −∂x , ϕ∂z = a∂y − ∂t , ϕ∂t = −a∂x + ∂z . (3.53)

For the Einstein metric (3.52), the proper almost complex structure ϕ in (3.53)
becomes
2x 2x
ϕ∂x = ∂y , ϕ∂y = −∂x , ϕ∂z = − ∂y − ∂t , ϕ∂t = ∂x + ∂z .
z z
Then, the integrability of ϕ, given in Theorem 65, becomes
4
ax + bx + 2cy = 2ax = − =6 0, ay + by − 2cx = 2ay = 0.
z
Thus, ϕ cannot be integrable.

3.7. Opposite Almost Complex Structure

It is known that an oriented 4-manifold with a field of 2-planes, or equivalently
endowed with a neutral indefinite metric, admits a pair of almost comlex struc-
ture ϕ and an opposite almost complex structure ϕ´, which satisfy the following
properties ( [47]):
i) ϕ2 = ϕ´2 = −1,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

ii) g(ϕX, ϕY ) = g(ϕ´X, ϕ´Y ) = g(X,Y ),

iii) ϕϕ´= ϕ´ϕ,
iv) the preferred orientation of ϕ coincides with that of M4 ,
v) the preferred orientation of ϕ´is opposite to that of M4 .
We devote this section to the analysis of opposite almost complex structures
on Norden-Walker 4-manifolds. Let (M4 , ϕ, g) be an almost Norden-Walker
manifolds. For a Walker manifold M4 , with the proper almost complex structure
ϕ, the g-orthogonal opposite almost complex structure ϕ´takes the form
θ2 θ1
ϕ´∂1 = −(θ1 c + a)∂1 − b∂2 + θ2 ∂3 + θ1 ∂4 ,
2 2
θ1 θ2
ϕ´∂2 = (− a + θ2 c)∂1 + b∂2 + θ1 ∂3 − θ2 ∂4 ,
2 2
θ1 θ2 2 θ2 θ1 θ1
ϕ´∂3 = −( ac + a + 2 2
)∂1 − ( ab + 2 )∂2
2 4 θ1 + θ2 4 θ1 + θ22
θ2 θ1
+ a∂3 + a∂4 ,
2 2

Salimov, Arif. Tensor Operators and their Applications, Nova Science Publishers, Incorporated, 2012. ProQuest Ebook
Central, http://ebookcentral.proquest.com/lib/unilu-ebooks/detail.action?docID=3022757.
Created from unilu-ebooks on 2023-10-16 21:16:26.
106 Arif Salimov
θ1 θ2 θ1
ϕ´∂4 = −(θ1 c2 + ab + (ac − bc) + 2 )∂1
4 2 θ1 + θ22
θ1 θ2 θ2 θ1 θ2
+(− bc + b2 + 2 2
)∂2 + ( b + θ2 c)∂3 + (θ1 c − b)∂4 ,
2 4 θ1 + θ2 2 2
where θ1 and θ2 are two parameters.
We shall focus our attention to one of explicit forms of ϕ´, obtained by fixing
two parameters as θ1 = 1 and θ2 = 0 (only for simplicity), as follows:
1 1
ϕ´∂1 = −c∂1 − b∂2 + ∂4 , ϕ´∂2 = − a∂1 + ∂3 ,
2 2
1 1 1
ϕ´∂3 = − ac∂1 − ( ab + 1)∂2 + a∂4 ,
2 4 2
2 1 1 1
ϕ´∂4 = −(c + ab + 1)∂1 − bc∂2 + b∂3 + c∂4
4 2 2
and ϕ´has the local components
 
−c − 21 a − 12 ac −(c2 + 14 ab + 1)
 − b
1
0 1
−( 4 ab + 1) − 12 bc 
ϕ´= (ϕ´ij ) = 
 0
2
1
.

1 0 2b
1
1 0 2a c
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

By similar devices (as in the previous section for proper almost complex struc-
ture ϕ ), we can prove the following theorems for opposite almost complex
structure ϕ´(see [81]):

Theorem 75. The opposite almost complex structure of an almost Norden-

Walker manifold is integrable if and only if the following PDEs hold:
abx − 2bz = 0, by = 0, ax − 2cy = 0,
bay − 2at − 2acx + 4ccy + 4cz = 0.

Theorem 76. A triple (M4 , ϕ´, g) is a Kähler-Norden-Walker manifold if and

only if the following PDEs hold:
ax = ay = bx = by = bz = cx = cy = 0, at − 2cz = 0.

Theorem 77. The opposite almost complex structure of an almost Norden-

Walker manifold (M4 , ϕ´, g) is isotropic Kähler if and only if the following PDEs
hold:
cx (2bay − 2acx + 4cz − 2at + 2cax ) + cy (2bz − 2abx) = 0

Theorem 78. A triple (M4 , ϕ´, g) is a quasi-Kähler-Norden-Walker manifold if

and only if the following PDEs hold:
ax = ay = bx = by = bz = cx = cy = 0, at − 2cz = 0.

3.8. Para-Norden-Walker Metrics

In the present section, we shall focus our attention to para-Norden manifolds of
dimension four. Using a Walker metric we construct new para-Norden-Walker
metrics together with a proper almost paracomplex structures.
Let F be an almost paracomplex structure on a Walker manifold M4 , which
satisfies
i) F 2 = I,
ii) g(FX,Y ) = g(X, FY ) (Nordenian property),
iii) F∂x = ∂y , F∂y = ∂x .
We easily see that these three properties define F non-uniquely, i.e.,


 F∂x = ∂y ,

F∂y = ∂x ,
1

 F∂ z = −α∂x − 2 (b − a)∂y + ∂t ,

F∂t = 21 (b − a)∂x + α∂y + ∂z
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

and has the local components

 1 
0 1 −α 2 (b − a)
 1 0 − 12 (b − a) α 
F = (Fji ) = 
 0


0 0 1
0 0 1 0
with respect to the natural frame {∂x , ∂y , ∂z , ∂t }, where α = α(x, y, z,t) is an
arbitrary function.
Our propose is to find a nontrivial para-Norden structure with Walker metric
g explicitly. If α = c, then we have explicit examples of Walker metrics with
hyperbolic functions a, b and c (Theorem 80). Therefore, we put α = c. Then g
defines a unique almost paracomplex structure
 1 
0 1 −c 2 (b − a)
 1 0 − 1 (b − a) c 
ϕ = (ϕij ) = 
 0 0
2 .
 (3.54)
0 1
0 0 1 0

The triple (M4 , ϕ, g) is called almost para-Norden-Walker manifold.

Using the Nijenhuis tensor for paracomplex structure (3.54), we have

Theorem 79. The proper almost paracomplex structure ϕ of an almost para-

Norden-Walker manifolds is integrable if and only if the following PDEs hold:

ax − bx + 2cy = 0,
(3.55)
ay − by + 2cx = 0.

From this theorem, we easily see that if a = b and c = 0, then ϕ is integrable.

Let (M4 , ϕ, g) be a para-Norden-Walker manifolds (Nϕ = 0) and a = −b.

Then the equation (3.55) reduces to

ax = −cy ,
ay = −cx ,

from which follows

axx − ayy = 0,
cxx − cyy = 0,

e.g., the functions a and c are hyperbolic with respect to the arguments x and y.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Thus we have

Theorem 80. If the triple (M4 , ϕ, g) is para-Norden-Walker and a = −b, then

a, b and c are all hyperbolic with respect to the arguments x,y .

Let now (M4 , ϕ, g) be an almost para-Norden-Walker manifold. Next analy-

sis of para-Norden-Walker manifolds is very similar to the analysis of Norden-
Walker Manifolds (see, Section 3.6), i.e the situtaion is as folows [84]:

Theorem 81. A triple(M4, ϕ, g) is a para-Kähler-Norden-Walker manifold if

and only if the following PDEs hold:

ax = ay = bx = by = bz = cx = cy = 0, at − 2cz = 0.

Theorem 82. A triple (M4 , ϕ, g) is a quasi-para-Kähler-Norden-Walker mani-

fold if and only if the following PDEs hold:

bx = by = bz = 0, ay − 2cx = 0, ax − 2cy = 0, cax − at + 2cz − (a − b)cx = 0.

Theorem 83. If (M4 , ϕ, g) is an almost para-Norden-Walker manifold, then it

is isotropic para-Kählerian.
Theorem 84. If a para-Norden-Walker manifold (M4 , ϕ, g) is para-Kähler-
Norden-Walker, then M4 is flat.
Theorem 85. Let (M4 , ϕ, g) be a para-Norden-Walker manifold. If
ax = ay = bx = by = 0,
then g is para-Norden-Walker-Einstein.
We note that the almost para-Norden-Walker structure is a specialized al-
most product metric structure on a pseudo-Riemannian manifold (There is an
extensive literature on almost product metric structures, see, for examples [11],
[17], [18], [22], [23], [53], [56], [96]). In this context, we consider some prop-
erties of almost para-Norden-Walker manifolds.
An integrable almost product manifold is usually called a locally product
manifold. Thus a para-Norden-Walker manifold (M4 , ϕ, g) is a locally prod-
uct pseudo-Riemannian manifold such that dimT + M4 = dim T − M4 = 2. From
Theorem 79 we have
Corollary 86. An almost para-Norden-Walker manifold (M4 , ϕ, g) is a locally
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

product pseudo-Riemannian manifold if and only if ax − bx + 2cy = 0 and ay −

by + 2cx = 0.
Let now (M4 , ϕ, g) be a paraholomorphic Norden-Walker manifold (∇ϕ =
0). It is well known that a locally product (pseudo-)Riemannian manifold
(M4 , ϕ, g) is a decomposable if and only if ∇ϕ = 0, where ∇ is a Levi-Civita
connection of g. Then, from Theorem 44 and Theorem 81, we have
Corollary 87. A para-Norden-Walker manifold (M4 , ϕ, g) is a locally decom-
posable if and only if ax = ay = bx = by = bz = cx = cy = 0 and at − 2cz = 0.
Gil-Medrano and Naveira established in [24] that both distributions of the
almost product structures on the (pseudo-)Riemannian manifold (M, ϕ, g) are
totally geodesic if and only if
g((∇X P)Y, Z) + g((∇Y P)Z, X) + g((∇Z P)X,Y) = σ g((∇X P)Y, Z) = 0
X,Y,Z
(3.56)
for any X,Y, Z ∈ ℑ10 (M).
From Theorem 82 and (3.56), we have

Corollary 88. Let (M4 , ϕ, g) be an almost para-Norden-Walker manifold. Both

distributions of the manifold (M4 , ϕ, g) are a totally geodesic if and only if bx =
by = bz = 0, ay − 2cx = 0, ax − 2cy = 0, cax − at + 2cz − (a − b)cx = 0.

The minimal almost product structure on the (pseudo-)Riemannian man-

ifold (M, P, g) is characterized by the equality of trg J = (g jk J ijk ) to zero on
M [23], [56], [96], where is the Jordan tensor field, which in coordinate form
can be written as
j
J ijk = Psj (∇s Pk + ∇ k Psi ) + Pks (∇s Pij + ∇ j Psi ).

We can rewrite the condition of minimality as [96]

j
∇ j Pk = 0. (3.57)

Let now (M4 , ϕ, g) be an almost para-Norden-Walker manifold. Then using

(3.57), we see that ∇1 ϕ1k + ∇2 ϕ2k + ∇3 ϕ3k + ∇4 ϕ4k = 0, k = 1, 2, 3, 4 if and only if

ax − bx − 2cy = 0,
(3.58)
ay − by − 2cx = 0.

Thus, we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Corollary 89. An almost para-Norden-Walker manifold (M4 , ϕ, g) is a mini-

mal, if and only if g satisfies the condition (3.58).

Let ∗ T (Mn ) denote the cotangent bundle of a manifold Mn and let π :∗

T (Mn ) → Mn be the natural projection. A point ξ of the cotangent bundle is
represented by an ordered pair (p, ω), where p = π(ξ) is a point on Mn and ω
is a 1-form on cotangent spaces ∗ Tp (Mn ). For each coordinate neighborhood
(U, xi ) on Mn with p∈ U, denote by xi , i = n + 1, ..., 2n the components of ω in
the natural coframe dxi . Then, for any local coordinates (U, xi ) on Mn , (xi , xi )
are natural induced coordinates in π−1 (U) ⊂∗ T (Mn ).
For a given symmetric connection ∇ on Mn , the cotangent bundle ∗ T (Mn )
may be equipped with a pseudo-Riemann metric ∇ g of signature (n, n) [115,
p.268]: The Riemann extension of ∇, given by
 ∇ V C
 g( ω, X) =V (ω(X)),
∇ g(V ω,V θ) = 0,
 ∇ C C
g( X, Y ) = −γ(∇X Y, ∇Y X),

where V ω,V θ are the vertical lifts to ∗ T (Mn ) of 1-forms ω, θ and C X,C Y denote
the complete lifts to ∗ T (Mn ) of vector fields X,Y on Mn . Moreover, for any
vector field Z on Mn (Z = Z i ∂i ) γZ is the function on ∗ T (Mn ) defined by γZ =
∑xi Z i . In a system of induced coordinates (xi , xi ) on ∗ T (Mn ), the Riemann
i
extension is expressed by
!
∇
−2∑xk Γkij δij
g= k
j
δi 0

with respect to {∂i , ∂i }, where Γkıj are the components of ∇. Since inverse (∇g)−1
of the matrix ∇ g is given by
j !
0 δi
∇ −1
( g) = δi 2 xk Γk ,
j ∑ ij
k

the Walker metrics (3.36) can be viewed as Riemann extensions, i.e., they are
locally isometric to the cotangent bundle (∗ T (∑α ),∇ g) of a surface (∑2 , ∇)
equipped with the metric ∇g. Then we have

Corollary 90. Let ∗ T (∑2 ) a cotangent bundle of a surface (∑2 , g). Then
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(∗ T (∑2 ),∇ g) is an almost para-Norden-Walker manifold with an almost para-

complex structure (3.54), where a = 2(x3 Γ111 + x4 Γ211 ), b = 2(x3 Γ122 + x4 Γ222 ),
c = 2(x3 Γ112 + x4 Γ212 ) .

3.9. Some Notes Concerning Norden-Walker

8-manifolds
In the present section, we shall focus our attention to the Norden manifolds in di-
mension eight. Using the Walker metric we constructive a new Norden-Walker
metrics together with so called proper almost complex structures. Note that an
indefinite Kähler-Einstein metric on an eight-dimensional Walker manifolds has
been recently investigated in [49].
A neutral metric g on a 8-manifold M8 is said to be Walker metric if there
exists a 4-dimensional null distribution D on M8 , which is parallel with respect
to g. From Walker theorem [110], there is a system of coordinates (x1 , ..., x8)

with respect to which g takes the local canonical form

0 I4
g = (gi j ) = , (3.59)
I4 B
where I4 is the unit 4 × 4 matrix and B is a 4 × 4 symmetric matrix whose entries
are functions of the coordinates (x1 , ..., x8). Note that g is of neutral signature
(+ + + + − − −−), and that the paralel null 4-plane D is spanned locally by
{∂1 , ∂2 , ∂3 , ∂4 }, where ∂i = ∂x∂ i , i = 1, ..., 4 .
In this section, we consider the specific Walker metrics on M8 with B of the
form  
a 0 0 0
 0 0 0 0 
B=  0 0 b 0 ,
 (3.60)
0 0 0 0
where a, b are smooth functions of the coordinates (x1 , ..., x8).
We can construct various almost complex structures ϕ on a Walker 8-
manifold M8 with the metric g as in (3.59), (3.60) so that (M8 , ϕ, g) is almost
Nordenian. The following ϕ is one of the simplest examples of such an almost
complex structure:
ϕ∂1 = ∂3 , ϕ∂2 = ∂4 , ϕ∂3 = −∂1 , ϕ∂4 = −∂2 ,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

1
ϕ∂5 = (a + b)∂3 − ∂7 , ϕ∂6 = −∂8 ,
2
1
ϕ∂7 = − (a + b)∂1 + ∂5 , ϕ∂8 = ∂6 .
2
In conformity with the terminology of Matsushita (see, [47]- [49]) we call ϕ the
proper almost complex structure. The proper almost complex structure ϕ has
the local components
 
0 0 −1 0 0 0 − 21 (a + b) 0
 0 0 0 −1 0 0 0 0 
 
 1 0 0 1
0 2 (a + b) 0 0 0 
 
 0 1 0 0 0 0 0 0 
ϕ = (ϕ j ) = 
i
 0 0 0
 (3.61)
 0 0 0 1 0 

 0 0 0 0 0 0 0 1 
 
 0 0 0 0 −1 0 0 0 
0 0 0 0 0 −1 0 0
with respect to the natural frame {∂i } , i = 1, ..., 8 .

Remark 24. From (3.61) we see that in the case a = −b, ϕ is integrable.

The proper almost complex structure ϕ on almost Norden-Walker manifolds

is integrable if and only if (Nϕ )ijk = 0.Since N ijk = −Nki j , we need only consider
N ijk ( j < k). By explicit calculation, the nonzero components of the Nijenhuis
tensor are as follows:

1 1 3 3 1
N15 = N37 = N17 = −N35 = (a1 + b1 ), (3.62)
2
3 1
N57 = (a + b)(a1 + b1 ),
4
1 1 3 3 1
N25 = N47 = N27 = −N45 = (a2 + b2 ),
2
1 1 3 3 1
N17 = −N35 = −N15 = −N37 = − (a3 + b3 ),
2
1 1
N57 = − (a + b)(a3 + b3 ),
4
1 1 3 3 1
N27 = N45 = N25 = −N47 = (a4 + b4 ),
2
1 1 3 3 1
N56 = −N78 = N58 = −N67 = − (a6 + b6 ),
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

2
1 1 3 3 1
N58 = −N67 = −N56 = N78 = − (a8 + b8 ).
2

From (3.62) we have

Theorem 91. The proper almost complex structure ϕ on almost Norden-Walker

8-manifolds is integrable if and only if the following PDEs hold:

a1 + b1 = 0, a2 + b2 = 0, a3 + b3 = 0,
a4 + b4 = 0, a6 + b6 = 0, a8 + b8 = 0

Corollary 92. i) The proper almost complex structure on almost Norden-

Walker 8-manifolds is integrable if and only if

a = −b + ξ

where ξ is any function of x5 and x7 alone,

ii) The metric (3.59) with matrix

 
−b(x1 , ..., x8) + ξ(x5 , x7 ) 0 0 0
 0 0 0 0 
B= 

0 0 b(x1 , ..., x8) 0 
0 0 0 0
is always Norden-Walker with respect to the proper complex structure ϕ.
Let (M8 , ϕ, g) be an almost Norden-Walker manifold. If
(Φϕ g)ki j = ϕm m m m m
k ∂m gi j − ϕi ∂k gm j + gm j (∂i ϕk − ∂k ϕi ) + gim ∂ j ϕk = 0, (3.63)
then by virtue of Theorem 44, ϕ is integrable and the triple (M8 , ϕ, g) is called
a holomorphic Norden-Walker or a Kähler-Norden-Walker manifold. We will
write (3.59), (3.60) and (3.31) in (3.63). Since (Φϕg)i jk = (Φϕg)ik j , we need
only consider (Φϕ g)i jk ( j < k). By explicit calculation, the nonzero components
of Φϕ g the tensor are as follows:

1
(Φϕ g)155 = a3 , (Φϕ g)157 = (b1 − a1 ), (Φϕ g)177 = b3 , (3.64)
2
1
(Φϕ g)255 = a4 , (Φϕ g)257 = (b2 − a2 ), (Φϕ g)277 = b4 ,
2
1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

(Φϕ g)355 = −a1 , (Φϕ g)357 = (b3 − a3 ), (Φϕ g)377 = −b1 ,

2
1
(Φϕ g)455 = −a2 , (Φϕ g)457 = (b4 − a4 ), (Φϕ g)477 = −b2 ,
2
1 1
(Φϕ g)517 = −(Φϕ g)715 = (b1 + a1 ), (Φϕ g)527 = −(Φϕ g)725 = (b2 + a2 ),
2 2
1 1
(Φϕ g)537 = −(Φϕ g)735 = (b3 + a3 ), (Φϕ g)547 = −(Φϕ g)745 = (b4 + a4 ),
2 2
1
(Φϕ g)555 = (b + a)a3 − a7 , (Φϕ g)557 = −b5 ,
2
1 1
(Φϕ g)567 = −(Φϕ g)756 = (b6 + a6 ), (Φϕ g)577 = (b + a)b3 + a7 ,
2 2
1
(Φϕ g)578 = −(Φϕ g)758 = (b8 + a8 ), (Φϕ g)655 = −a8 ,
2
1 1
(Φϕ g)657 = (b6 − a6 ), (Φϕ g)677 = −b8 , (Φϕ g)755 = − (b + a)a1 − b5 ,
2 2
1
(Φϕ g)757 = −a7 , (Φϕ g)777 = − (b + a)b1 + b5 ,
2
1
(Φϕ g)855 = a6 , (Φϕ g)857 = (b8 − a8 ), (Φϕ g)877 = b6 .
2

From (3.64) we have

Theorem 93. The triple (M8 , ϕ, g) is Kähler-Norden-Walker if and only if the

following PDEs hold:

a1 = a2 = a3 = a4 = a6 = a7 = a8 = 0,

b1 = b2 = b3 = b4 = b5 = b8 = 0. (3.65)

Corollary 94. (M8 , ϕ, g) is Kähler-Norden-Walker if and only if the matrix B in

(3.60) has components
 
a(x5 ) 0 0 0
 0 0 0 0 
B=  0
.
0 b(x ) 0 
7

0 0 0 0

It is well known that, if (M8 , ϕ, g) be an almost Norden manifold, then the

Norden metric g is a quasi-Kähler-Norden if and only if

(Φϕg)(X,Y, Z) + (Φϕg)(Y, Z, X) + (Φϕg)(Z, X,Y ) = 0 (3.66)

for any X,Y, Z ∈ ℑ10 (M2n ) (see Theorem 72). From (3.66) we easily see that a
Kähler-Norden manifold is a quasi-Kähler-Norden. Conversely, quasi-Kähler-
Norden manifold is a non-Kähler-Norden, in general. In (M8 , ϕ, g) particular,
let be an almost Norden-Walker 8-manifold. Using (3.64) and (3.66), we have

(Φϕ g)155 + (Φϕ g)551 + (Φϕ g)515 = a3 = 0, (3.67)

1
(Φϕ g)157 + (Φϕ g)571 + (Φϕ g)715 = (b1 − a1 ) = 0,
2
(Φϕ g)177 + (Φϕ g)771 + (Φϕ g)717 = b3 = 0,
(Φϕ g)255 + (Φϕ g)552 + (Φϕ g)525 = a4 = 0,
1
(Φϕ g)257 + (Φϕ g)572 + (Φϕ g)725 = (b2 − a2 ) = 0,
2
(Φϕ g)277 + (Φϕ g)772 + (Φϕ g)727 = b4 = 0,
(Φϕ g)355 + (Φϕ g)553 + (Φϕ g)535 = −a1 = 0,
1
(Φϕ g)357 + (Φϕ g)573 + (Φϕ g)735 = (b3 − a3 ) = 0,
2

(Φϕ g)377 + (Φϕ g)773 + (Φϕ g)737 = −b1 = 0,

(Φϕ g)455 + (Φϕ g)554 + (Φϕ g)545 = −a2 = 0,
1
(Φϕ g)457 + (Φϕ g)574 + (Φϕ g)745 = (b4 − a4 ) = 0,
2
(Φϕ g)477 + (Φϕ g)774 + (Φϕ g)747 = −b2 = 0,
1
(Φϕ g)557 + (Φϕ g)575 + (Φϕ g)755 = − (a + b)a1 − 3b5 = 0,
2
1
(Φϕ g)555 = (a + b)a3 − a7 = 0,
2
1
(Φϕ g)567 + (Φϕ g)756 + (Φϕ g)675 = (b6 − a6 ) = 0,
2

From (3.64) and (3.67) we see that the triple (M8 , ϕ, g) is quasi-Kähler-
Norden-Walker if and only if the PDEs in the form (3.65) holds. On the other
hand, the equation (3.65) is a Kähler condition of almost Norden-Walker mani-
folds. Thus we have

Theorem 95. Let (M8 , ϕ, g) be an almost Norden-Walker 8−Manifold. Then

there does not exist a (non-Kähler) quasi-Kähler structure on this manifold.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Applications to the Theory

of Lifts

In this chapter we concentrate our attention to tensor bundles and we study the
various lifts from manifold to its tensor bundle by using tensor operators. In
section 4.1 some introductory materials concerning with the tensor bundle of
type (p, q) over differentiable manifold Mn are collected. Here some of our no-
tations are fixed too. In section 4.2 is devoted to the some particular types of
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p
vector fields on the tensor bundle Tq (Mn ). Explicit expressions of the vertical
lift of tensor fields of type (p, q) and the complete and horizontal lifts of vector
fields are presented. In this section we also discuss complete lifts of derivations.
These results generalize some of those already obtained for tangent and cotan-
gent bundles. Section 4.3 devoted to the analysis of lifts on the cross-section
of the tensor bundle. A tensor field of type (p, q) on the tensor bundle defines
a cross-section of the tensor bundle, which is an n−dimensional submanifold
p
of the tensor bundle Tq (Mn ). Also, the behavior along the cross-section of the
complete and horizontal lifts of vector fields was reviewed. In section 4.4 deals
p
with how affinor fields on Tq (Mn ) can be induced from affinor fields on Mn .
We note that the φϕ −operator is an extension of the operator of Lie derivation
LX to affinor fields ϕ. Models of the complete lifts of affinor fields along the
pure cross-section are found by using of the φϕ −operator, moreover we discuss
almost complex property of these lifts. Also we consider almost hyperholomor-
phic pure submanifolds of tensor bundles. By using of Vishnevskii operator, the
problem of the horizontal lifts of affinor fields along the pure cross-section is
solved in section 4.4.5. In section 4.4.6, a new φϕ −operator is defined and dis-

cussed relations between the φϕ −operator and the diagonal lifts of affinor fields
along the cross-section of Tqp (Mn ). In section 4.5, the adapted frame which al-
lows the tensor calculus to be done efficiently is inserted in the tensor bundle
Tqp (Mn ). The components of Levi-Civita and metric connections of Sasakian
metrics on tensor bundles with respect to the adapted frame are presented. This
having been done it is shown possible to study geodesics of Sasakian metrics
dealing with geodesics of the base manifold.Section 4.6 is devoted to the study
of para-Nordenian property of Sasaki metric on cotangent bundles. We also
study paraholomorphic property of Cheeger-Gromoll metric on tangent bundles
and almost complex structures along the holomorhic cross-section of the cotan-
gent bundle.In the last section 4.7, as applications of Yano-Ako operators we
construct model of complete lifts of skew-symmetric tensor field of type (1, 2)
on pure cross-sections in the tensor bundle.
Through the chapter we always suppose that all functions, vector fields and
tensor fields on tensor bundles are of class C∞ .

4.1. Tensor Bundles

Let Mn be a differentiable manifold of class C∞ and finite dimension n. Then
p p
the set Tq (Mn ) = ∪ Tq (P) is, by definition, the tensor bundle of type (p, q)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

P∈Mn
p
over Mn , where ∪ denotes the disjoint union of the tensor spaces Tq (Mn ) for
all P ∈ Mn . For any point Pe of Tqp (Mn ), the surjective correspondence Pe → P
p
determines the natural projection π : Tq (Mn ) → Mn . In order to introduce a
p
manifold structure in Tq (Mn ), we define local charts on it as follows: Let x j be
local coordinates in a neighborhood U of P ∈ Mn , then a tensor t at P which
i ...i i ...i
is an element of Tqp (Mn ) is expressible in the form (x j ,t j11 ... jpq ), where t j11 ... jpq are
i ...i
components of t with respect to natural base. We may consider (x j ,t j11... jpq ) =
(x j , x j ) = (xJ ), j = 1, ..., n , j = n + 1, ..., n + n p+q , J = 1, ..., n + n p+q as local
coordinates in a neighborhood π−1 (U) ⊂ Tqp (Mn ).
p
It is straightforward to see that Tq (Mn ) becomes an (n + n p+q )−manifold;
indeed if x j´ are local coordinates in a neighborhood V of P ∈ Mn , with U ∩V 6=
∅, then the change of coordinates is given by
(
x j´ = x j´(x j ),
i´ ...i´ i´ ...i´ j ... j i ...i (i´) ( j) (4.1)
x j´ = t j´11 ... j´pq = Ai11 ...i pp A j´11 ... j´qq t j11 ... jpq = A(i) A( j´) x j ,

where
(i´) ( j) i´ ...i´ j ... j ∂xi´ j1 ∂x j
A(i) A( j´) = Ai11 ...i pp A j´11 ... j´qq , Aii´11 = , A j´1 = j´ .
∂xi ∂x
The Jacobian of (4.1) is:
! !
∂x j´ ∂x j´ j´
∂xJ´ ∂x j ∂x j
Aj 0
= ∂x j´ ∂x j´
= (i) (i´) (k) (i´) ( j) , (4.2)
∂xJ t(k)∂ j A(i) A( j´) A(i) A( j´)
∂x j ∂x j

(i) i ...i
where J = ( j, j), J = 1, ..., n + n p+q, t(k) = tk11 ...kpq .
We denote by ℑrs (Mn ) the module over F(Mn ) ( F(Mn ) is the ring of
C∞ −functions on Mn ) all tensor fields of class C∞ and of type (r, s) on Mn .
p
If α ∈ ℑq (Mn ) , it is regarded, in a natural way (by contraction), as a function in
p
Tq (Mn ), which we denote by ια. If α has local expression
j ... j
α = αi11...i pq ∂ j1 ⊗ ... ⊗ ∂ jq ⊗ dxi1 ⊗ ... ⊗ dxi p

in a coordinate neighborhood U(x j ) ⊂ Mn , then ια = α(t) has the local expres-

j ... j i ...i
sion ια = αi11...i pq t j11 ... jpq with respect to the coordinates (x j , x j ) in π−1 (U).
For later use, we first shall state following Lemma.

Lemma 96. Let Xe and Ye be vector fields on Tqp (Mn ) such that X(ια)
e = Ye (ια),
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p p
e = Ye , i.e. Xe ∈ ℑ (Tq (Mn )) is completely deter-
1
for any α ∈ ℑq (Mn ). Then X 0
mined by its action on functions of type ια.
e
Proof. It is sufficient to prove that if Z(ια) = (Xe − Ye )(ια) = 0, then Ze is zero.
If Ze has components Ze with respect to the coordinates (x j , x j ) in π−1 (U), we
J

have
e j ... j i ...i j ... j i ...i
Z(ια) = Ze j ∂ j (αi11...i pq t j11 ... jpq ) + Ze j ∂ j (αi11...i pq t j11 ... jpq )
i ...i j ... j i ...i
= Ze j t j11 ... jpq ∂ j αi11...i pq + Ze j α j11 ... jpq = 0. (4.3)
j ... j
If this holds for any α ∈ ℑqp (M), ∂ j αi11...i pq taking any preassigned values at a
fixed point, from (4.3) (homogeneous system of linear equations), we have
i ...i
Ze j t j11 ... jpq = 0, Ze j = 0. (4.4)

From the first equation of (4.4), it follows that Ze j at all points of Tqp (Mn ) except
i ...i
possibly those at which all the components t j11 ... jpq are zero: that is, at points of

the base spaces. However, the components of Ze are continuous and so Ze j is zero
at points of the base space. Thus Ze j = 0 at all points of π−1 (U). Therefore,
taking account of the second equation of (4.4), we see that Ze is zero in π−1 (U).
This completes the proof of Lemma 96.

4.2. Horizontal and Complete Lifts of Vector Fields

4.2.1. Vertical Lifts of Tensor Fields and γ−operator
p p
Suppose that A ∈ ℑq (Mn ). Then there is a unique vector field V A ∈ ℑ10 (Tq (Mn ))
q
such that for α ∈ ℑ p (Mn ) [38]
V
A(ια) = α(A) ◦ π =V (α(A)), (4.5)
where V (α(A)) is the vertical lift of the function α(A) ∈ F(Mn ). We note that
the vertical lift V f = f ◦ π of the arbitrary function f ∈ F(Mn ) is constant along
each fibre π−1 (P).
If V A =V Ak ∂k +V Ak ∂k , then we have from (4.5)
V i ...i j ... j k ...k k ...k h ...h
Ak t j11 ... jpq ∂k αi11...i pq +V Ak αh11 ...hqp = αh11 ...hqp Ak11...kqp .
k ...k j ... j
But αh11 ...hqp and ∂k αi11...i pq can take any preassigned values at each point. Thus,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

we have from the equation above

V i ...i h ...h
Ak t j11 ... jpq = 0, V
Ak = Ak11...kqp .
p
Hence V Ak = 0 at all points of Tq (Mn ) except possibly those at which all the
i ...i
components x j = t j11 ... jpq are zero: that is, at points of the base space. Thus
we see that the components V Ak are zero a point such that x j 6= 0, that is, on
p p p
Tq (Mn ) − Mn . However, Tq (Mn ) − Mn is dense in Tq (Mn ) and the components
of V A are continuous at every point of Tqp (Mn ). Hence, we have V Ak = 0 at
p p
all points of Tq (Mn ). Consequently, the vertical lift V A of A to Tq (Mn ) has
components V j
V A 0
A= V j = i ...i (4.6)
A A j11 ... jpq

with respect to (x j , x j ) the coordinates in Tqp (Mn ).

We complete this section by giving the following definition which is needed
later on.

p
with respect to the coordinates (x j , x j ) in Tq (Mn ).

From (4.2) we easily see that the vector fields γϕ and eγϕ determine respec-
p
tively global vector fields on Tq (Mn ). The local expressions of the global vector
fields γϕ and eγϕ are as follows:

0 0
γϕ = p i1 ...m...i p iλ and eγϕ = q i1 ...i p .
∑ t j1 ...m... jq ϕmjµ
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

∑ t j1 ... jq ϕm
λ=1 µ=1

4.2.2. Complete Lifts of Vector Fields

Let LV be the Lie derivation with respect to V ∈ ℑ10 (Mn ). The complete lift
c p
V = LV of V to Tq (Mn ) is defined by [38]
c
V (ια) = ι(LV α) (4.7)
p
for α ∈ ℑq (M).
If cV =c V k ∂k +c V k ∂k , then we have from (4.7)

c i ...i j ... j k ...k

V kt j11 ... jpq ∂k αi11...i pq +c V k αh11 ...hqp (4.8)
q p
i ...i j ... j j ...k... jq j ... j
= t j11 ... jpq (V k ∂k αi11...i pq − ∑ (∂kV jµ )αi11...i p + ∑ (∂iλ V k )αi11...k...i
q
p
).
µ=1 λ=1

Thus, discussing in the same way as in the case of the vertical lift, from (4.8)
we see that, the complete lift cV has components
c
c Vj Vj
V= c j = p
i ...m...i p
1
q
i ...i (4.9)
V ∑ t j1 ... jq ∂mV iλ − ∑ t j11 ...m...
p
jq ∂ jµ V
m
λ=1 µ=1

p
with respect to the coordinates (x j , x j ) in Tq (Mn ) (see [67]).

4.2.3. Horizontal Lifts of Vector Fields

Let ∇ be a symmetric affine connection on Mn . We define the horizontal lift
H
∇=∇ e V ∈ ℑ1 (Tqp (Mn )) of V ∈ ℑ0 (Mn ) to Tqp (Mn ) by [38]
0 1

H
V (ια) = ι(∇V α), α ∈ ℑqp (Mn ).
p
The horizontal lift H V of V ∈ ℑ01 (Mn ) to Tq (Mn ) has components

H
Vj
V= q
i ...i
p
i i ...m...i p (4.10)
V s ( ∑ Γm 1 p λ 1
s jµ t j1 ...m... jq − ∑ Γsm t j1 ... jq )
µ=1 λ=1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

with respect to the coordinates (x j , x j ) in Tqp (Mn ) , where Γkij are local compo-
nents of ∇ (see [39]).

4.2.4. Complete Lifts of Derivations

∞
p
We now put F(Mn ) = ∑ ℑq (Mn ), which is the direct sum of all tensor mod-
p,q=0
ules on Mn . A map D : F(Mn ) → F(Mn ) is a derivation on Mn , if
a) D is linear with respect to constant coefficients,
b) For all p, q, Dℑqp (Mn ) ⊂ ℑqp (Mn ),
c) For all tensor fields T1 and T2 on Mn ,

D(T1 ⊗ T2 ) = (DT1 ) ⊗ T2 + T1 ⊗ (DT2 ),

d) D commutes with contraction.

For a derivation D on Mn , there exists a vector field P on Mn such that

P f = D f , f ∈ F(Mn ).

If we put
D(∂i ) = Qhi ∂h
in each coordinate neighborhood U of Mn , then the pair (Ph , Qhi ) is called the
components of the derivation D in U [114, p.26].
j ... j
Let α be an element of ℑqp (Mn ) with local expression α = αi11...i pq ∂ j1 ⊗ ... ⊗
∂ jq ⊗ dxi1 ⊗ ... ⊗ dxi p . Then we see that Dα has components of the form
q p
m j ... j j ...m... j j j1 ... jq
Dα = (P ∂m αi11...i pq +∑ αi11...i p q Qmµ − ∑ αi ...m...i
1 p
Qm
iλ )
µ=1 λ=1

in Mn , Ph being the components of P ∈ ℑ10 (Mn ) .

Let D be a derivation on Mn . Then there is a unique vector field C D ∈
p p
ℑ10 (Tq (Mn )) such that for α ∈ ℑq (Mn ) [8], [38]
C
D(ια) = ι(Dα).

We call C D the complete lift of D to Tqp (Mn ). The C D has components

C
Pj
D = q i1 ...i p p i ...m...i
p i
(4.11)
∑ t j1 ...m... jq ϕmjµ − ∑ t j1 ... jq ϕmλ
1

µ=1 λ=1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p
with respect to the coordinates (x j , x j ) in Tq (Mn ) .
j ... j
Let LV denote the Lie derivation with respect to V . From LV αi11...i pq =
q p
j ... j j ...m... jq j ... j
V m ∂m αi11...i pq − ∑ (∂mV jλ )αi11...i p + ∑ (∂iµ V m )αi11...m...i
q
p
, we see that the Lie
λ=1 µ=1
derivation LV is a derivation on Mn having components LV : (V h , −∂iV h ). Using
(4.9) and (4.11), we have
C
(LV ) =C V,
where C V is the complete lift of the vector field V to Tqp (Mn ).
Let now ∇ be an affine connection on Mn and ∇V denote the covariant
derivation with respect to V . By similar devices, we see that the covariant
derivation ∇V is a derivation on Mn having components

∇V : (V h ,V s Γhsi )

and
` `
C
(∇V ) =C V − γ(∇V ) + eγ(∇V ) ,

In particular, if ∇ is a symmetric affine connection, then

C
(∇V ) =H V , (4.12)
p
where H V is the horizontal lift of the vector field V ∈ ℑ10 (Mn ) to Tq (Mn ).

4.2.5. Derivations DKX Y and Formulas on Lie Derivations

When a derivation D on Mn satisfies the condition D f = 0 for any f ∈ F(Mn ), D
determines an element ϕ ∈ ℑ11 (Mn ) in such a way that DX = ϕX, ∀X ∈ ℑ10 (Mn )
. In such a case, D is denoted by Dϕ and called the derivation determined by ϕ.
From Dϕ f = 0 and Dϕ X = ϕX, we easily verify that the Dϕ has local compo-
nents
Dϕ : (0, ϕhi ), Ph = 0, Qhi = ϕhi , (4.13)
where ϕhi are local components of ϕ. From (4.11), (4.13) and Definition 4.2.1,
we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

C
0
(Dϕ) = q
i1 ...i p
p
i1 ...m...i p i = eγϕ − γϕ
∑ t j1 ...m... jq ϕmjµ − ∑ t j1 ... jq ϕmλ
µ=1 λ=1

or
C
(Dϕ ) = eγϕ − γϕ.
The Lie derivative LX ∇ of a symmetric affine connection ∇ with respect to
X ∈ ℑ10 (Mn ) is, by definition, an element of ℑ12 (Mn ) such that

(LX ∇)(Y, Z) = LX (∇Y Z) − ∇Y (LX Z) − ∇ [X,Y ] Z

for any Y, Z ∈ ℑ10 (Mn ). We now denote by KX Y the tensor field of type (1, 1),
defined by
(KX Y )Z = (LX ∇)(Y, Z) = [LX , ∇Y ]Z − ∇[X,Y ] Z, (4.14)
where
KX Y = [LX , ∇Y ] − ∇[X,Y ] , (4.15)

which is an equation in terms of derivations. If we take the complete lifts of

both sides in (4.15), we have

eγ(KX Y ) − γ(KX Y ) =C (DKX Y ) =C [LX , ∇Y ] −C (∇[X,Y ] ). (4.16)

Taking account of [38]

[C D1 ,C D2 ] =C [D1 , D2 ]
for any derivations D1 and D2 , from (4.16), we have

eγ(KX Y ) − γ(KX Y ) = C
(DKX Y )
= [ (LX ),C (∇Y )] −C (∇[X,Y ] ) = [C X,H Y ] −H [X,Y ].
C

Thus, we have

Theorem 97. [C X,H Y ] =H [X,Y ] + eγ(KX Y ) − γ(KX Y ) for any X,Y ∈ ℑ10 (Mn ) ,
where KX Y denotes the tensor field of type (1, 1) defined by (4.14).

An infinitesimal transformation defined by a vector field X ∈ ℑ10 (Mn ) is said

Theorem 98. Let X be an infinitesimal affine transformation on Mn . Then

[C X,H Y ] =H [X,Y ].

Let ∇ be a Riemannian connection on Mn and ∇X = 0. Then LX g = 0, i.e.

X is an infinitesimal isometry or a Killing vector field. We next have LX ∇ = 0
as a consequence of LX g = 0. Since C X =H X (∇X = 0), we have

Theorem 99. Let X be a vector field with vanishing Riemannian covariant

derivative. Then
[H X,H Y ] =H [X,Y ]
p
i.e. the operation of taking the horizontal lift H : ℑ10 (Mn ) → ℑ10 (Tq (Mn )) is a
homomorphism.

4.3. Cross-sections in the Tensor Bundle

Suppose that there is given a tensor field ξ ∈ ℑqp (Mn ) . Then the correspondence
x → ξx , ξx being the value of ξ at x ∈ Mn , determines a mapping σξ : Mn →
Tqp (Mn ), such that π ◦ σξ = idMn , and the n dimensional submanifold σξ (Mn ) of
p
Tq (Mn ) is called the cross-section determined by ξ. If the tensor field ξ has the
h ...h
local component ξk11...kqp (xk ), the cross-section σξ (Mn ) is locally expressed by
(
xk = xk ,
h ...h (4.17)
xk = ξk11...kqp (xk )

p
with respect to the coordinates (xk , xk ) in Tq (Mn ). Differentiating (4.17) by x j ,
we see that n tangent vector fields B j to σξ (Mn ) have components

∂xk δkj
(Bkj ) =( j)= h ...h (4.18)
∂x ∂ j ξk11...kqp
p
with respect to the natural frame {∂k , ∂k } in Tq (Mn ).
On the other hand, the fibre is locally expressed by
(
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

xk = const,
h1 ...h p h ...h
tk1 ...kq = tk11...kqp ,

h ...h
tk11...kqp being considered as parameters. Thus, on differentiating with respect to
i ...i
x j = t j11 ... jpq we see n p+q that tangent vector fields C j to the fibre have components

∂xk 0
(Ckj ) =( )= j h (4.19)
∂x j δkj11 ...δkqq δhi11 ...δi pp
p
with respect to the natural frame {∂k , ∂k } on Tq (Mn ), where δ is the Kronecker
symbol.
p
We consider in π−1 (U) ⊂ Tq (Mn ), n + n p+q local vector fields B j and C j
along σξ (Mn ) . They form a local family of frames {B j ,C j } along σξ (Mn ) ,
which is called the adapted (B,C)−frame of σξ (Mn ) in π−1 (U). From cV =c
V h ∂h +c V h ∂h and cV =c V j B j +c V jC j , We easily obtain cV k =c V j Bkj +c V jCkj ,
c
V k =c V j Bkj +c V jCkj . Now, taking account of (4.9) on the cross-section σξ (Mn ),

and also (4.18) and (4.19), we have cV e k = −LV ξh1 ...h p . Thus, the
e k = V k , cV
k1 ...kq
complete lift cV has along σξ (Mn ) components of the form
c
c Vek Vk
V = c ek = h ...h (4.20)
V −LV ξk11...kqp

with respect to the adapted (B,C)−frame.

From (4.6), (4.18) and (4.19), by using similar way the vertical lift V A also
has components V k
V A 0
A= V k = h ...h (4.21)
A Ak11...kqp
with respect to the adapted (B,C)−frame.
The lift H V has along σξ (Mn ) components of the form
H
H
ej
V Vj
V = Hej = i ...i (4.22)
V −(∇V ξ) j11 ... jpq

Let ϕ ∈ ℑ11 (M) now . We can easily verify that γϕ and eγϕ have along σξ (M)
components

J
0 J
0
e
γϕ = ((γϕ) ) = p
j1 ...m... j p i e e
, γϕ = ((γϕ) ) = q
i1 ...i p
∑ ξi1 ...iq ϕmλ ∑ ξ j1 ...m... jq ϕmjµ
λ=1 µ=1
(4.23)
i ...i
with respect to the adapted (B,C)−frame, where ξ j11 ... jpq are local components of
ξ.
Similarly, if S ∈ ℑ12 (M) , then γS and V (S ◦ ξ) are affinor fields along σξ (M)
with components
  !
0 0 0 0
γS = ((eγS)IJ ) =  p jλ j1 ...m... j p  , (V (Se oξ)IJ ) = m j ... j
∑ S jm ξi1 ...iq 0 S jjm
1
ξi1 ...i2 q p 0
λ=1
(4.24)
with respect to the adapted (B,C)−frame, where S jjm
1
are local components of S.

4.4. Lifts of Affinor Fields

4.4.1. Complete Lifts of Affinor Fields
The complete lift of tensor field of type (1, 1) on Mn is defined as follow.

Definition 15. [41], [86] We define a tensor field c ϕ ∈ ℑ11 (Tqp (Mn )) along the
ϕ
pure cross-section σξ (Mn ) by
c c
ϕ( V ) =c (ϕ(V )) − γ(LV ϕ) +V ((LV ϕ) ◦ ξ), (i)
c V (4.25)
ϕ( A) =V (ϕ(A)), (ii)
p p
for all V ∈ ℑ10 (Mn ) and A ∈ ℑq (Mn ), where ϕ(A) ∈ ℑq (Mn ),
((LV ϕ) ◦ ξ)(x1 , ..., xq; α1 , ..., αp ) = ξ(x1 , ..., xq; (LV ϕ)´α1 , ..., αp )
p
and call c ϕ the complete lift of ϕ ∈ ℑ11 (Mn ) to Tq (Mn ), p ≥ 1, q ≥ 0 along
ϕ
σξ (Mn ).

In particular, if we assume that p = 1, q > 0 then we get

c
ϕ(cV ) =c (ϕ(V )), cϕ(V A) =V (ϕ(A)).
∗
p p
Let ℑq (Mn ) denotes a module of all the tensor fields ξ ∈ ℑq (Mn ) which are
ϕ
pure with respect to ϕ. Now, we consider a pure cross-section σξ (Mn ) deter-
∗
p
mined by ξ ∈ ℑq (Mn ), p ≥ 1, q ≥ 0.
We observe that the local vector fields
h
c c∂ c h ∂ δj
X( j) = ( j ) = (δ j h ) =
∂x ∂x 0
and
V
X ( j) = V
(∂ j1 ⊗ ... ⊗ ∂ j p ⊗ dxi1 ⊗ ... ⊗ dxiq )
i k
= (δih11 ...δhqq δkj11 ...δ j pp ∂k1 ⊗ ... ⊗ ∂k p ⊗ dxh1 ⊗ ... ⊗ dxhq )
V

0
= i k
δih11 ...δhqq δkj11 ...δ j pp

j = 1, ..., n, j = n + 1, ..., n + n p+q span the module of vector fields in π−1 (U).
Hence any tensor field is determined in π−1 (U) by its action of cX( j) and V X ( j) .

Remark 25. The equation (4.25) is useful extension of the equation (see [38])
c
L(ια) = ι(LV α), α ∈ ℑqp (Mn )
ϕ
to affinor fields along the pure cross-section σξ (Mn ).

Now, let us compute components of the complete lift of tensor fields of type
(1, 1) by using Φϕ −operator.
ϕ
Theorem 100. Let ϕ ∈ ℑ11 (Mn ) and σξ be a pure cross-section of Tqp (Mn ) with
respect to ϕ. Then the complete lift c ϕ ∈ ℑ11 (Tqp (Mn )) of ϕ has along the pure
ϕ
cross-section σξ (Mn ) components

 Cϕ h ...h
ekl = ϕkl , C ϕ
elk = 0, C ϕ ekl = −(Φϕ ξ)lk11 ...kpq ,
r
 Cϕ ek = ϕh1 δhs 2 ...δhs p δr1 ...δ q
l s1 2 p k1 kq

ϕ
with respect to the adapted (B,C)− frame of σξ (Mn ), where Φϕ ξ is the
h ...h s ...s
Tachibana operator and xk = tk11...kqp , xl = tr11...rqp .
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Proof. Let cϕKL be components of c ϕ with respect to the adapted (B,C)−frame

ϕ
of the pure cross-section σξ (Mn ). Then from (4.25) we have
 K K
 c Kc eL c ∼ K ∼ ∼
eL V = (ϕ(V)) − (γ(LV ϕ)) +V ((LV ϕ) ◦ ξ) , (i)
ϕ
(4.26)
 c eK V eL V
∼
ϕL A = (ϕ(A))K , (ii)
where

!
V
∼
K
0
( (ϕ(A)) ) = mh ...h p ,
ϕhm1 Ak1 ...k
2
q
!
∼ K 0
V
((LV ϕ) ◦ ξ) = h h ...m...h ,
(LV ϕmλ )ξk11...kq p
!
∼ K 0
γ(LV ϕ) = h mh2 ...h p .
((LV ϕ)m )ξk1 ...k
1
q

First, consider the case where K = k. In this case, (i) of (4.26) reduces to
∼
C ek C e l k
e l =C (ϕ(V ))K = (ϕ(V))k = ϕklV l .
ϕl V +C ϕ
el C V (4.27)
Since the right-hand side of (4.27) are functions depending only on the base
coordinates xi , the left-hand sides of (4.27) are too. Then, since cV l depend on
fibre coordinates, from (4.27) we obtain
C ek
ϕl = 0. (4.28)
eklCV
From (4.27) and (4.28), we have C ϕ eklV l = ϕklV l , V i being arbitrary,
e l =C ϕ
which implies
C ek
ϕl = ϕkl .
When K = k, (ii) of (4.26) reduces to
∼
C e k V el ekl V A
el =V (ϕ(A))k
ϕl A +C ϕ
or
C e k s1 ...s p mh2 ...h p r s s ...s
ϕl Ar1 ...rq = ϕhm1 Ak1 ...k q
= δrk11 ...δkqq ϕhs11 δhs22 ...δhs pp Ar11 r22 ...rqp
p
for all A ∈ ℑq (Mn ), which implies
c ek r
ϕl = δrk11 ...δkqq ϕhs11 δhs22 ...δhs pp ,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

s ...s h ...h
where xl = tr11...rqp , xk = tk11...kqp .When K = k, (i) of (4.26) reduces to
∼ p
c ek c e l k h1 ...l...h p
ϕl V e l =c (ϕ(V ))k −
e l cV
+c ϕ ∑ (LV ϕhl )ξk ...k λ
1 q
λ=2
or
p ∼
c ek c e l r c el h1 h2 ...l...h p
ϕl V + ϕhs11 δhs22 ...δhs pp δrk11 ...δkqq V + ∑ (LV ϕhl )ξk ...k λ
1 q
=c (ϕ(V ))k . (4.29)
λ=2

ekl . The
Now, using the Tachibana operator we will investigate components cϕ
∗
p
Tachibana operator on the pure module ℑq (Mn ) is given by (see Section 1.2)
∗ q
h ...h h ...h h ...h h ...h
(Φϕξ)lk11 ...kpq = ϕm p 1 p m
l ∂m ξk1 ...kq − ∂l ξ k1 ...kq + ∑ (∂ka ϕl )ξk1 ...m...kq
1 1 p

a=1
p
h h ...m...h p
+2 ∑ ∂[l ϕm]λ ξk11...kq ,
λ=1

where
h ...h h ...h h ...h
ϕm 1 p m 1 p m 1 p
k1 ξmk2 ...kq = ϕk2 ξk1 m...kq = ... = ϕkq ξk1 k2 ...m =
∗
mh ...h h m...h h2 ...m h h h ...h
ϕhm1 ξk1 ...k
2
q
p
= ϕhm2 ξk11...kq p = ... = ϕmp ξhk11...k q
= ξ k11k22...kqp .

After some calculations we have

∗ p
h ...h h ...h h ...h h ...m...h p
V l (Φϕξ)lk11 ...kpq = (LϕV ξ)k11...kqp − (LV ξ)k11...kqp + ∑ (LV ϕhmλ )ξk11...kq
λ=1

or
h ...h mh ...h mh ...h
V l (Φϕ ξ)lk11 ...kpq + ϕhm1 (LV ξ)k1...k
2
q
p 2
+ ξk1 ...k q
p
(LV ϕ)hm1
p h ...m...h p h ...h (4.30)
h
− ∑ (LV ϕmλ )ξk11...kq = (LϕV ξ)k11...kqp
λ=1

for any V ∈ ℑ10 (Mn ).Using (4.20), from (4.30) we have

h ...h mh ...h mh ...h

h ...h r s ...s
= V (Φϕ ξ)lk11 ...kpq + ϕhs11 δhs22 ...δhs pp δrk11 ...δkqq (LV ξ)r11 ...rqp
l

p
h ...m...h p
− ∑ (LV ϕhmλ )ξk11...kq
λ=2
h ...h r
= c
V l (Φϕ ξ)lk11 ...kpq + ϕhs11 δhs22 ...δhs pp δrk11 ...δkqq cV l
p
h ...m...h p
− ∑ (LV ϕhmλ )ξk11...kq
λ=2

= − (ϕ(V ))k c

or
h ...h r
−(Φϕ ξ)lk11 ...kpq cV l + ϕhs11 δhs22 ...δhs pp δrk11 ...δkqq cV l
p (4.31)
h h ...m...h p
+ ∑ (LV ϕmλ )ξk11...kq =c (ϕ(V ))k.
λ=2
Comparing (4.29) and (4.31), we get
c ek h ...h
ϕl = −(Φϕ ξ)lk11 ...kpq .

Salimov, Arif. Tensor Operators and their Applications, Nova Science Publishers, Incorporated, 2012. ProQuest Ebook
Central, http://ebookcentral.proquest.com/lib/unilu-ebooks/detail.action?docID=3022757.
Created from unilu-ebooks on 2023-10-16 21:16:26.
132 Arif Salimov
p
Thus, the complete lift cϕ ∈ ℑ11 (Tq (Mn ) of ϕ has along the pure cross-section
ϕ
σξ (Mn ) components
 n
 cϕ h ...h
ekl = ϕkl , cϕ
e kl = 0 , c ϕ
e kl = −(Φϕξ)lk11 ...kpq ,
r
(4.32)
 c ek
ϕl = ϕhs11 δhs22 ...δhs pp δrk11 ...δkqq
ϕ
with respect to the adapted (B,C)−frame of σξ (Mn ), where Φϕξ is the
s ...s h ...h
Tachibana operator and xl = tr11...rqp , xk = tk11...kqp . This completes the proof.
Remark 26. c ϕ in the form (4.32) is a unique solution of (4.25). Therefore, if
∗ p ∗
ϕ is element of ℑ11 (Tq (Mn )), such that ϕ(cV ) =c ϕ(cV ) =c (ϕ(V )) − γ(LV ϕ) +V
∗ ∗
((LV ϕ) ◦ ξ), ϕ(V A) =c ϕ(V A) =V (ϕ (A)), then ϕ =c ϕ.
In particular, if we write p = 1, q = 0, then (4.32) is the formula of the
complete lift of affinor fields to tangent bundle along the cross-section σξ (Mn )
(for details, see [114, p. 126]).
ϕ
Now, on putting B j = C j , we write the adapted (B,C)−frame of σξ (M) as
ϕ
BJ = {B j , B j }. We define a coframe BeJ of σ (M) by BeI (BJ ) = δIJ . From (4.18),
ξ
(4.19) and BKJ BeIK = δIJ we see that covector fields BeI have components
(
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Bei = (BeiK ) = (δij , 0)

j ... j k j j (4.33)
Bei = (BeiK ) = (−∂k ξi11...iqp , δki11 ...δiqq δh11 ...δhpp )

with respect to the natural coframe (dxk , dxk ). Taking account of

c K
ϕL = c e JI BJ ⊗ BeI (dxk , ∂L )
ϕ(dxk , ∂L ) =c ϕ
c eJ K eJI dxK (BH
= ϕI dx (BJ )BeI (∂L ) =c ϕ eI
J ∂H )BL
= cϕeJI BH K eI c e J K eI
J δH BL = ϕ I BJ BL ,

and also (4.18), (4.19), (4.32) and (4.33), we see that c ϕ has along the pure
ϕ
cross-section σξ (Mn ) components of the form

c k

 ϕl = ϕkl , c ϕkl = 0,

 r

 ϕl = ϕhs11 δhs22 ...δhs pp δrk11 ...δkqq ,
c k

 q
c k mh2 ...h p h1 ...h p
ϕl = (∂l ϕhm1 )ξk1 ...k − ∑ (∂kµ ϕm l )ξk1 ...m...kq
(4.34)

 q
µ=1




p
h h h ...m...h

 − ∑ (∂l ϕmλ − ∂m ϕl λ )ξk11...kq p
λ=1

p
4.4.2. Almost Complex Structures on Tq (Mn )
Theorem 101. If ϕ is an integrable almost complex structure on Mn , then the
ϕ
complete lift C ϕ of ϕ to Tqp (Mn ) along the pure cross-section σξ (Mn ) is an al-
p
most complex structure on Tq (Mn ).

Proof. Let ϕ ∈ ℑ11 (Mn ) and S ∈ ℑ12 (Mn ) , using (4.20), (4.23), (4.24) and (4.34),
we have

γ(ϕ ± ψ) = γϕ ± γψ,
C
ϕ(γψ) = γ(ϕ ◦ ψ) = γ(ψ ◦ ϕ), (4.35)
(γS)CV = γSV ,

where SV is the tensor field of type (1, 1) on Mn defined by SV (W ) = S(V,W),

for any W ∈ ℑ10 (Mn ). If V ∈ ℑ10 (Mn ), then from (4.25) and (4.35), we have

(C ϕ)2 (CV ) = (C ϕ ◦C ϕ)CV =C ϕ(C ϕ(CV ))

C
= ϕ(C (ϕ(V ))) −C ϕ(γ(LV ϕ)) +C ϕ(V ((LV ϕ) ◦ ξ))
C
= (ϕ(ϕ(V ))) − γ(Lϕ(V ) ϕ) −C ϕ(γ(LV ϕ))
+C ϕ(V ((LV ϕ) ◦ ξ)) +V ((Lϕ(V ) ϕ) ◦ ξ)
C
= (ϕ(ϕ(V ))) − γ(Lϕ(V ) ϕ) − γ((LV ϕ) ◦ ϕ)
+V (ϕ((LV ϕ) ◦ ξ)) + (Lϕ(V ) ϕ) ◦ ξ) (4.36)
C
= (ϕ(ϕ(V ))) − γ(Lϕ(V ) ϕ + (LV ϕ) ◦ ϕ)
+V ((Lϕ(V ) ϕ) + (LV ϕ) ◦ ξ) +V ((LV (ϕ ◦ ϕ)) ◦ ξ)
C
= (ϕ ◦ ϕ)(CV ) + γ(LV (ϕ ◦ ϕ)) − γ(Lϕ(V ) ϕ + (LV ϕ) ◦ ϕ)
C
= (ϕ ◦ ϕ)(CV ) − γ(Lϕ(V ) ϕ − ϕ ◦ (LV ϕ))
+V ((Lϕ(V ) ϕ) − ϕ(LV ϕ) ◦ ξ)
C
= (ϕ ◦ ϕ)(CV ) − γNV +V (NV ◦ ξ)
C
= (ϕ ◦ ϕ)(CV ) − (γN)(CV ) +V (N ◦ ξ)(CV )
= (C (ϕ)2 − γN +V (N ◦ ξ)(CV ),

where NV = Lϕ(V ) ϕ − ϕ ◦ (LV ϕ) and (Φϕ ϕ)(V,W) = (Lϕ(V ) ϕ − ϕ ◦ (LV ϕ))W =

[ϕV, ϕW ]−ϕ[V, ϕW ]−ϕ[ϕV,W]+ϕ2 [V,W] = NV W is nothing but the Tachibana
operator or the Nijenhuis-Shirokov tensor N(V,W) ∈ ℑ12 (Mn ) constructed from
p
ϕ. Similarly, if A ∈ ℑq (Mn ), from (4.25), we have
(C ϕ)2 (C A) = (C ϕ ◦C ϕ)C A =C ϕ(C ϕ(C A)) =C ϕ(V ϕ(A)) =V (ϕ(ϕ(A))
= V
((ϕ ◦ ϕ)(A)) =C (ϕ ◦ ϕ)V A =C (ϕ2 )V A. (4.37)
If we take the integrability condition of ϕ (Nϕ = 0), then by the Remark 26,
(4.36), (4.37) and the linearity of the complete lift, we have
(C ϕ)2 =C (ϕ2 ) =C (−I) = −I.

Let n−dimensional Mn and m−dimensional Nm be two manifolds with com-

plex structures ϕ and ψ, respectively. A differentiable mapping f : Mn → Nm is
called a holomorphic (analytic) mapping, if at each point p ∈ Mn
d f p ◦ ϕ p = ψ f (p) ◦ d f p . (4.38)
ϕ
As the mapping f : Mn → Nm (m = n + n p+q ) we take a cross-section σξ :
p p
Mn → Tq (Mn ) determined by the pure tensor field ξ ∈ ℑq (Mn ) with respect
ϕ p
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

to ϕ−structure. The pure cross-section σξ : Mn → Tq (Mn ) can be locally ex-

pressed by (4.17). In (4.38), if ψ is the almost complex structure C ϕ (see The-
orem 101), the condition that the pure cross-section ξ be a holomorphic tensor
field is locally given by
ϕm k C K M
l ∂m x = ϕM ∂l x , (4.39)
ϕ
where C ϕKM are components of C ϕ along the pure cross-section σξ (Mn ) with
respect to the natural frame {∂k , ∂k }.
In the case K = k, by virtue of (4.17) and (4.34) we get the identity ϕkl ≡ ϕkl .
When K = k, by virtue of (4.17) and (4.34), (4.39) reduces to
∗ q
h ...h h ...h h ...h h ...h
(Φϕξ)lk11 ...kpq = ϕm p 1 p m
l ∂m ξk1 ...kq − ∂l ξ k1 ...kq + ∑ (∂ka ϕl )ξk1 ...m...kq
1 1 p

a=1
p
h h ...m...h
+2 ∑ ∂[l ϕm]λ ξk11...kq p , (4.40)
λ=1

where Φϕ ξ is the Tachibana operator. Thus, the holomorphic tensor field ξ is

given by (4.40) with respect to the natural frame {∂i , ∂i } (see Section 2.5).

4.4.3. Almost Hyperholomorphic Pure Submanifolds in the Tensor

Bundle

By using (4.34), in [65] (see also [69]) proved that if Π = ϕ is almost
α
algebraic hypercomplex Π−structure on Mn , then the complete lift
integrable
p
c
Π = cϕ of Π to Tq (Mn ) along the pure cross-section σΠ
ξ (M) is an algebraic
α
p
hypercomplex c Π−structure on Tq (Mn ).
For an element X ∈ ℑ10 (Mn ) with local coordinates X k , we denote by BX
and CX the vector fields with local components
!
Xk
BX = h ...h = X jB j,
X j ∂ j ξk11...kqp

! !
0 s 0
CX = s = δkr11 ...δkrqq X i1 δsh22 ...δhpp jq h1 h
δkr11 ...δkrqq X h1 δsh22 ...δhpp j1
δk1 ...δkq δi1 ...δi pp
s
= δkr11 ...δkrqq X i1 δsh22 ...δhpp C( j)

∗
along the pure cross-section σξ (Mn ) determined by ξ ∈ ℑqp (Mn ) that

C
ϕ(BX) = B(ϕX) −C((Φϕξ)(X;Y1, ...,Yq, ξ1 , ..., ξp )) (4.41)

for any X ∈ ℑ10 (Mn ). When C ϕ(BX) is tangent to σξ (Mn ) for any X ∈ ℑ10 (Mn ),
c
ϕ is said to leave σξ (Mn ) invariant. We have from (4.41)

Theorem 102. The complete lift c ϕ of an element ϕ ∈ ℑ11 (Mn ) leaves the pure
cross-section σξ (Mn ) invariant if and only if Φϕ ξ = 0.

Asubmanifold
in an almost algebraic hypercomplex manifold with structure
Π = ϕ , is said to be almost hyperholomorphic when ϕ, α = 1, ..., m leaves
α α
the submanifold invariant.
Thus, from Theorem 102, we have

Theorem 103. A necessary and sufficient condition for the pure cross-section
σξ (Mn ) in Tqp (Mn ) determined by a pure tensor field ξ ∈ ℑqp (Mn ) in an almost

hypercomplex manifold Mn with structure Π = ϕ to be almost hyperholo-
α
p
almost algebraic hypercomplex manifold Tq (Mn )
morphic submanifoldin the
with structure c Π = c ϕ is that the pure tensor field ξ ∈ ℑqp (Mn ) be almost
α
hyperholomorphic in Mn .

4.4.4. Horizontal Lifts of Affinor Fields

Now, let us define the horizontal lifts of tensor fields of type (1, 1)
p
Definition 16. [39] Let ϕ ∈ ℑ11 (Mn ). The tensor field H ϕ ∈ ℑ11 (Tq (Mn )) of
ϕ
ϕ ∈ ℑ11 (Mn ) along the cross-section σξ (Mn ) to tensor bundles of type (p, q) is
defined as following:
H H
ϕ( V ) =H (ϕ(V )), ∀V ∈ ℑ10 (Mn ),
H V p (4.42)
ϕ( A) =V (ϕ(A)), ∀A ∈ ℑq (Mn ),
where ϕ(A) = C(ϕ ⊗ A).
Using the Vishnevskii operator, we shall compute components of the hor-
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p
izontal lift H ϕ ∈ ℑ11 (Tq (Mn )) with respect to the adapted (B,C)−frame of
ϕ
σξ (Mn ). We can state following theorem

Theorem 104. The horizontal lift H ϕ of ϕ along the pure cross-section to tensor
bundles has the following components
H ek l ...l
ϕl ek = 0, H ϕ
= ϕkl , H ϕ ekl = −(Φϕξ)lk1 1 ...kp q ,
( ls s s r
H ek
ϕl11 δl22 ...δl pp δrk11 ...δkqq , p ≥ 1 (4.43)
ϕl = s r
δsl11 ...δl pp ϕrk11 δrk22 ...δkqq , q ≥ 1
ϕ
with respect to the adapted (B,C)−frame of σξ (Mn ), where Φϕ ξ is the Vish-
nevskii operator.
Proof. Let ϕeKL be components of H ϕ with respect to (B,C)−frame of the cross-
ϕ
section σξ (M). Then from (4.42) we have
(
H eK H e L H e
ϕL V = (ϕ(V ))K , (i)
(4.44)
HϕeKL V A
eL =H (ϕ
e(A))K , (ii)

with respect to (B,C)−frame, which is also projectable.

When K = k, (ii) of (4.44) reduces to
∼
H e k V el H e k V el V
ϕl A + ϕl A = (ϕ(A))k ,
s ...s ml2 ...l p s r s ...s
ekl Ar11 ...rqp = ϕlm1 Ak1 ...k
Hϕ
q
= ϕls11 δsl22 ...δl pp δrk11 ...δkqq Ar11 ...rqp
p
for all A ∈ ℑq (M). It immediately follows that
H ek s r
ϕl = ϕls11 δsl22 ...δl pp δrk11 ...δkqq , p ≥ 1, (4.46)
s ...s l ...l
where x j = tr11...rqp , xk = tk11 ...kpq .
By the similar way, we have
H ek s r
ϕl = δsl11 ...δl pp ϕrk11 δrk22 ...δkqq , q ≥ 1.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

When K = k, (i) of (4.44) reduces to

∼
H ek H e l k H el
ϕl V +H ϕ
el V =H (ϕ(V))k . (4.47)
k
For obtaining the component H ϕ e l , let us introduce Vishnevskii operator defined
by (see Section 1.3)
( ml2 ...l p
l1 ...l p m l1 ...l p
ϕlm1 ∇l ξk1 ...k , p≥1
(Φϕξ)lk1 ...kq = ϕl ∇m ξk1 ...kq − l ...l
q
(4.48)
ϕm 1 p
k1 ∇l ξmk2 ...kq , q ≥ 1

From (4.48), we have

l ...l ll ...l l ...l
V l (Φϕξ)lk1 1 ...kp q + ϕll1 ∇V ξk12...kqp = ∇ϕ(V ) ξk11 ...kpq , p ≥ 1 (4.49)

Using (4.22), from (4.49) we have

l ...l ll ...l l ...l s r
V l (Φϕ ξ)lk1 1 ...kp q + ϕll1 ∇V ξk12...kqp = V l (Φϕξ)lk1 1 ...kp q + ϕ ls11 δsl22 ...δl pp δrk11 ...δkqq
∼
l ...l s r
∇V ξsr11 ...s 1 p H el l1 s2 p r1 q H el H
...rq = (Φϕξ)lk1 ...kq V − ϕs1 δl2 ...δl p δk1 ...δkq V = − (ϕ(V))
p k

or ∼
l ...l s r
(Φϕξ)lk1 1 ...kp q H Ve l − ϕls11 δsl22 ...δl pp δrk11 ...δkqq H V
e l = −H (ϕ(V))k . (4.50)
Comparing (4.47) and (4.50), and using (4.46), we get
k l ...l
(H ϕ
el + (Φϕ ξ)lk1 1 ...kp q )V l = 0,

for all V ∈ ℑ10 (M), from which

H ek l ...l
ϕl = −(Φϕ ξ)lk1 1 ...kp q , p ≥ 1.
Similarly, we obtain
H ek l ...l
ϕl = −(Φϕξ)lk1 1 ...kp q , q ≥ 1.
This completes the proof.
Remark 27. Let ϕ be an integrable ϕ−structure in Mn and ∇ϕ = 0. If ξ ∈
p
ℑq (Mn ) is a pure tensor field with respect to ϕ−structure, then the equation
l ...l
(Φϕξ)lk1 1 ...kp q = 0 is the condition for ξ to be holomorphic.

Remark 28. The formula (4.43) is valid if and only if Φϕ ξ is the Vishnevskii
∗
operator, i.e. in the form (4.43) is a unique solution of (4.42). Therefore, if ϕ is
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p ∗ ∗
an element of ℑ11 (Tq (Mn )), such that ϕ(H V ) =H ϕ(H V ) =H (ϕ(V)), ϕ(V A) =H
∗
ϕ(V A) =V (ϕ(A)), then ϕ =H ϕ.

4.4.5. Diagonal Lifts of Affinor Fields

In this section we defined a new operator for obtaining components of the diag-
onal lift D ϕ along the cross-section σξ (Mn ) to tensor bundles with respect to the
adapted (B,C)−frame
Firstly, let us define the diagonal lift of tensor field of type (1, 1).
Definition 17. Let ϕ ∈ ℑ11 (Mn ). We define a tensor field D ϕ of type (1, 1) on
Tqp (Mn ) along the cross-section σξ (Mn ) by
D H
ϕ( V ) =H (ϕ(V )), ∀V ∈ ℑ10 (Mn ),
D ϕ(V A) = −V (ϕ(A)), ∀A ∈ ℑ p (M ), (4.51)
q n
p
where ϕ(A) = C(ϕ ⊗ A) ∈ ℑq (Mn ) and call D ϕ the diagonal lift of ϕ ∈ ℑ11 (Mn )
p
to Tq (Mn ) along σξ (Mn ).

with respect to the adapted (B,C)−frame of σξ (Mn ).

Proof. From (4.51) we have


 D eK H e L H ∼ K
ϕL V = (ϕ(V )) , (i)
∼ (4.53)
 Dϕ eK V A
eL = −V (ϕ(A))K , (ii)
L
! !
∼ 0 0
where V (ϕ(A)) = ∼ = ml2 ...l p . Substituting K = k in (i)
V (ϕ(A))k ϕhm1 Ak1 ...k q

of (4.53) and K = k, (ii) of (4.53), calculating in the same way as in Section

s r
e kl
Dϕ = −δsl11 ...δl pp ϕrk11 δrk22 ...δkqq , q≥1
s ...s l ...l
where xr = tr11...rqp , xk = tk11 ...kpq .
k
Now, we will study the component D ϕ
el of the diagonal lift D ϕ. In the case
when K = k, (i) of (4.53) reduces to
∼
D ek H e l k H el
ϕl V +D ϕ
el V =H (ϕ(V ))k . (4.55)
p
Let ξ ∈ ℑq (Mn ). We consider a new Φϕ −operator [21]
( ml2 ...l p
l1 ...l p m l1 ...l p ϕlm1 ∇l ξk1 ...k , p ≥ 1,
(Φϕξ)lk1...kq = ϕl ∇m ξk1 ...kq + l ...l
q
(4.56)
ϕm 1 p
k1 ∇l ξmk2 ...kq , q ≥ 1.

From (4.56), we have

l ...l l ...l ml ...l
V l (Φϕξ)lk1 1 ...kp q = (V l ϕm 1 p l1 l 2 p
l )∇m ξk1 ...kq + ϕmV (∇l ξk1 ...kq ), p ≥ 1,
l ...l ml ...l l ...l (4.57)
V l (Φϕξ)lk1 1...kp q − ϕlm1 (∇V ξk1 ...k
2
q
p
) = ∇ϕ(V )ξk11 ...kpq .

Using (4.22), from (4.57) we have

l ...l r
(Φϕ ξ)lk1 1 ...kp q H V l + ϕls11 δls22 ...δlspp δrk11 ...δkqq H V l = −H (ϕ(V ))k,
l ...l r (4.58)
−(Φϕ ξ)lk1 1 ...kp q H V l − ϕls11 δls22 ...δlspp δrk11 ...δkqq H V l =H (ϕ(V ))k.

Comparing (4.55) and (4.58) and making use of (4.54), we get

D ek l ...l
ϕl = −(Φϕξ)lk1 1 ...kp q , p ≥ 1.

Similarly, we obtain

D ek l ...l
ϕl = −(Φϕ ξ)lk1 1...kp q , q ≥ 1.

This completes the proof.

h ...h
Remark 29. In the case of ∂m ξk11...kqp , (B,C)−frame is considered as a natural
frame {∂h , ∂h } of σξ (Mn ). Then, from (4.56) we obtain components of D ϕ along
the cross-section
 ( r

 −ϕls11 δls22 ...δlspp δrk11 ...δkqq , p ≥ 1,


D k k D k
ϕl = ϕl , ϕl = 0, ϕl = D k
r

 −δls11 ...δlspp ϕrk11 δrk22 ...δkqq , q ≥ 1,
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.


 

 p
lλ l1 ...s...l p
q
l ...l p

 
 ϕm (− ∑ Γ ξ + ∑ Γsmkµ ξk11 ...s...k )+

 
 l ms k1 ...kq q

 
 λ=1 µ=1

 q p

 l1
Γ s
ξ
ml2 ...l p l ml2 ...s...l p
Γlsλ ξk1 ...k − Γm
s...l p
 D k  +ϕ m ( ∑ lk µ k 1 ...s...k q
− ∑ q ls ξk1 ...kq ), p ≥ 1,

 µ=1 λ=2

 ϕl = p q

 
 ϕ m
(− Γ
lλ
ξ
l1 ...s...l p
+
l ...l p
Γsmkµ ξk11 ...s...k )+

 
 l ∑ ms k1 ...kq ∑

 
 λ=1 µ=1
q

 
 q p

 
 s ξl1 ...l p l l1 ...s...l p l1 ...l p
  +ϕm k1 ( ∑ Γ lkµ mk2 ...s...kq − ∑ Γlsλ ξmk 2 ...kq
+ Γslm ξs...k q
), q ≥ 1
µ=2 λ=1

p
with respect to the natural frame {∂h , ∂h } of σξ (M) in π−1 (U) ⊂ Tq (Mn ) [21].

4.5. Lifts of Metrics

In this section, we first introduce adapted frame to overcome the difficulty,
which also allows the tensor calculation to do efficiently.

4.5.1. Adapted Frames

In each local chart U(xh ) of Mn , we put

X( j) = ∂x∂ j = δhj ∂x∂h ∈ ℑ10 (Mn ),

A( j) = ∂i1 ⊗ ... ⊗ ∂i p ⊗ dx j1 ⊗ ... ⊗ dx jq
k j j p
= δki11 ...δi pp δh11 ...δhqq ∂k1 ⊗ ... ⊗ ∂k p ⊗ dxh1 ⊗ ... ⊗ dxhq ∈ ℑq (Mn ),
j = n + 1, ..., n + n p+q.

Then from (4.6) and (4.10), we see that these vector fields have, respectively,
local expressions
p q
k k ...s...k p k ...k
H
X( j) = δhj ∂h + (− ∑ Γ jsλ th11...hq + ∑ Γsjhµ th11...s...h
p
q
)∂h , (4.59)
λ=1 µ=1

k j j
A( j) = δki11 ...δi pp δh11 ...δhqq ∂h
V
(4.60)
n o n o
p
with respect to the natural frame ∂x∂H = ∂x∂h , ∂h on Tq (Mn ), where xh =
∂x
k ...k j
th11...hqp , δi −Kronecker
delta. These n + n p+q
vector fields are linear indepen-
dent and generate, respectively, the horizontal
n distribution
o of ∇ and the vertical
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p H V ( j)
distribution of Tq (Mn ). We call the set X( j), A the frame adapted to the
affine connection ∇ in π−1 (U) ⊂ Tqp (Mn ). On putting

e( j) =H X( j) , e( j) =V A( j)
n o
we write the adapted frame as eβ = e( j) , e( j) . The indices α, β, γ, ... run
over the range {1, ..., n, n + 1, ...,n + n p+q} and indicate the indices with respect
to the adapted frame.
Using (4.59) and (4.60), we have
X j δhj
H
X = p
k ...s...k p
q
k ...k
k
−X( ∑ Γ jsλ th11...hq − ∑ Γsjhµ th11...s...h
p
q
)
λ=1 µ=1
δhj
j
= X p q = X j e( j) ,
k k ...s...k k ...k p
−( ∑ Γ jsλ th11...hq p − ∑ Γsjhµ th11...s...h q
)
λ=1 µ=1

i.e. the lifts H X and V A have respectively components

H X j X j
H
X = HXβ = H j = , (4.61)
X 0
V A j 0
Ve V β
A= A = V j = i ...i (4.62)
A A j11 ... jpq
i ...i
with respect to the adapted frame eβ , X j and A j11 ... jpq being local components
of X and A in Mn , respectively.

4.5.2. Sasakian Metrics on the Tensor Bundles

For each p ∈ Mn the extension of scalar product g (denoted also by g) is defined
p
on the tensor space π−1 (p) = Tq (P)
i ...i t ...t
g(A, B) = gi1t1 ...gi pt p g j1 l1 ...g jqlq A j11 ... jpq Bl11 ...lpq
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

for all A, B ∈ ℑqp (P).

Definition 18. A Sasakian metric s g (or a diagonal lift of g ) is defined on

p
Tq (Mn ) by the three equations
s
g(V A,V B) =V (g(A, B)), A, B ∈ ℑqp (Mn ), (4.63)
s
g(V A,H Y ) = 0, (4.64)
s
g(H X,H Y ) =V (g(X,Y)), X,Y ∈ ℑ10 (Mn ). (4.65)

These equations are easily seen to determine s g on Tqp (Mn ) with respect to
which the horizontal and vertical distributions are complementary and orthogo-
nal.
From (4.63)-(4.65) we see that the Sasakian metric s g has components
s
s g jl s g jl g jl 0 t ...t
geβγ = s = , xl = tl11...lqp , (4.66)
g jl s g jl 0 gi1 t1 ...gi pt p gej1 l1 ...e
g jq lq

with respect to the adapted frame, gi j and gi j being local covariant and con-
travariant components g of on Mn .
Sasakian metrics on tangent bundle were introduced in 1958 by the Japanese
geometer Sasaki [87]. Sasakian metrics on tangent and cotangent bundles were
also studied in [28], [73], [74], [114]. In a more general case of tensor bundles
of type (1, 1), (1, q) and (0, q), Sasakian lifting of metrics are considered in [78],
[79]. The Sasakian metrics on frame bundle was first considered by Mok [54]
(for details, see [10]).
We now consider local 1−forms ωα defined by
e αdxB
ωα = A B

in π−1 (U), where

!  
eh A
A eh δhj 0
eα = j
A j
= p
k k ...s...k p
q
(k) k j j 
B eh A
A j
eh 1
∑ Γlsλ t(h) − ∑ Γslhµ th1 ...s...hq δki11 ...δi pp δh11 ...δhqq
j λ=1 µ=1
(4.68)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

is the inverse matrix of the matrix

!  
A hj A hj δhj 0
AβA = = p
k k ...s...k
q
k ....k k j j 
A hj A hj − ∑ Γlsλ th11....hq p + ∑ Γslhµ th11...s...h
p
q
δki11 ...δi pp δh11 ...δhqq
λ=1 µ=1
(4.69)
of frames changes eβ = AβA ∂A . These n + n p+q 1−forms ωα are linearly inde-
p
pendent on Tq (Mn ). We call the set {ωα } the dual adapted coframe.
For various types of indices, we have
 !

 e = A A∂ = ∂ + −
p
k k ...s...k
q
(k)
j j A j ∑ Γjsλ t(h)
1 p
+ ∑ Γsjhµ th1 ...s...hq ∂h ,
λ=1 µ=1 (4.70)

 e j = A Aj ∂A = ∂ j

and (
ωj = Ae j dxB = dx j ,
B
e j dxB = δt i1 ...i p , (4.71)
ωj = A B j1 ... jq

from which we have

e α.
Ωαγβ = (eγAβA − eβ AγA )A A

Thus, according to (4.68), (4.69) and (4.70), the components of non-holonomic

object Ωαγβ are given by
 p q

 Ω r = −Ωr =
∑ Γ
vλ j1
δ ...δ jq v1
δ ...δs
...δ
vp
− ∑
v
Γslrµ δrj11 ...δsjµ ...δrjqq δvi11 ...δi pp ,
 lj jl ls r1 rq i1 iλ ip
λ=1 µ=1
q p

 Ωrl j = ∑
(v)
Rsl jrµ tr1 ...s...rq
v
− ∑ Rljsλ t(r)1
v ...s...v p

µ=1 λ=1
(4.72)
all the others being zero, with respect to the adapted frame, Rhijk being local
components of the curvature tensor R of Mn with metric g.
If s ∇ denote the Levi-Civita connection of s g, from

T (X,Y ) =s ∇X Y −s ∇Y X − [X,Y ] = 0, ∀X,Y ∈ ℑ10 (Tqp (Mn ))

we have
s α
Γγβ −s Γαβγ = Ωαγβ (4.73)
with respect to the adapted frame, where s Γαγβ are components of the Levi-Civita
connection s ∇.
The equation (s∇X s g)(Y, Z) = 0, ∀X,Y, Z ∈ ℑ10 (Tqp (Mn )) has form

eδ s gγβ −s Γεδγ s gεβ −s Γεδβ s gγε = 0 (4.74)

with respect to the adapted frame. Thus, we have from (4.73) and (4.74)

s α 1 s αε 1
Γγβ = g (eα s gεβ + eβ s gγε − eε s gγβ ) + (Ωγβ α + Ωα γβ + Ωα βγ), (4.75)
2 2

where Ωαγβ =s gαε s gδβ Ωδεγ . Taking account of (4.69), (4.72) and (4.75), for
various types of indices, we find

p
s r 1
Γl j = g j1 l1 ...e
[(− ∑ gt1 i1 ...∇X gtλ iλ ...gi pt p )e g jqlq
2 λ=1
q
+( ∑ gej1 l1 ...∇X gejµ lµ ...e
g jqlq )gi1t1 ...gi pt p ]
µ=1
= 0,
s r
Γl j = 0, s Γrl j = 0, s Γrl j = Γrl j ,
p q
v v v
s r
Γl j = ∑ Γlsλ δrj11 ...δrjqq δvi11 ...δsiλ ...δipp − ∑ Γslrµ δrj11 ...δsjµ ...δrjqq δvi11 ...δipp , (4.76)
λ=1 µ=1

s r 1 q (v)
p
v v ...s...v p
Γl j = ( ∑ Rsl jrµ tr1...s...rq − ∑ Rljsλ t(r)1 ),
2 µ=1 λ=1
q p
(k) k k ...s...k p
s r
Γl j = 12 gxr gi1 t1 ...gi pt p gej1 h1 ...e
g jq hq ( ∑ Rsxlhµ th1 ...s...hq − ∑ Rxls
λ 1
t(h) ),
µ=1 λ=1
q p
(k) k k ...s...k p
s r
Γl j = 12 gxr gt1 k1 ...gt pk p gel1 h1 ...e
glqhq ( ∑ Rsx jhµ th1 ...s...hq − ∑ Rxλjst(h)
1
)
µ=1 λ=1

tion components constructed with g on M with respect to the natural frame {∂i }.

4.5.3. Geodesics on Tensor Bundles

In this section, we shall characterize the geodesics on the tensor bundles with
respect to the Levi-Civita connections of Sasaki metric and the horizontal lift
H ∇ of a affine connection ∇.

Definition 19. Let eγ = eγ(t) be a curve in Tqp (Mn ) and suppose that eγ is locally
R R r r r v1...v p
expressed
n byo x n= x (t),oi.e. x = x (t), x = tr1...rq (t) with respect to the natural
∂ ∂
frame ∂xl
= , ∂
, t being a parameter an arc length of eγ. Then the curve
∂xi ∂xi
γ = π ◦ eγ on M is called the projection of the curve eγ and denoted by πeγ which is
expressed locally by xr = xr (t).

Let ∇ be a Riemannian connection on Mn . Then a curve eγ is, by definition,

p
a geodesic in Tq (Mn ) with respect to s ∇ if and only if it satisfies the differential
equations

δ2 xR d 2 xR s R dxC dxB
= + ΓCB = 0. (4.77)
dt 2 dt 2 dt dt
We find it more convenient to refer equations (4.77) to the adapted frame. Using
(4.71), we now put
v1 ...v p
ωr dxr ωr δtr1 ...rq
= , = , (4.78)
dt dt dt dt
along a curve eγ. Using (4.78) equations (4.77) can be transformed into

d ωε s α ωγ ωβ
( ) + Γγβ =0 (4.79)
dt dt dt dt
with respect to the adapted frame.
By means of (4.76), (4.79) reduces to

d ωr dxl dx j
( ) + Γrl j (4.80)
dt dt dt dt
p
1 k1 ...k p
+ gexr gt1 k1 ...gt p k p gel1 h1 ...e
glq hq (− ∑ Rx js kλ t(h)
2 λ=1
t ...t
q δl11 ...lqpdx j
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

s (k)
+∑ Rx jhµ th1...s...hq )
µ=1 dt dt
p
1 k1 ...k p
+ gexr gi1 k1 ...gi pk p gej1 h1 ...e
g jqhq (− ∑ Rxls kλ t(h)
2 λ=1
l ...l
q
(k) dxl δt j11 ... jpq
+∑ Rxlhµ sth1 ...s...hq )
µ=1 dt dt
= 0,

v1 ...v p
d δtr1 ...rq 1 q (k)
p l
k1 ...k p dx dx
j
( ) + ( ∑ Rl jhµ sth1 ...s...hq − ∑ Rl js kλ t(h) ) (4.81)
dt dt 2 µ=1 λ=1
dt dt
i ...i
p
v l
q
l dx
l δt j11 ... jpq
+( ∑ Γlsλ δrj11 ...δrjqq δli11 ...δsiλ ...δi pp −∑ Γslrµ δrj11 ...δsjµ ...δrjqq δli11 ...δi pp )
λ=1 µ=1 dt dt
= 0.

We now transform (4.80) as follows:

δ2 xr l1 h1 lqhq
q
(k)
dt 2
+ gt1 k1 ...gtpk p e
g ...e
g ( ∑ Rrjhµ sth1...s...hq (4.82)
µ=1
t ...t
p δ1 p dx j
r kλ k1 ...s...k p l1 ...lq
−∑ R js t(h) )
λ=1
dt dt
= 0.
!
p q
k k1 ...s...k p (k) dxl dx j
Using the identity − ∑ Rl λjst(h) + ∑ Rsl jhµ th1 ...s...hq dt dt = 0,
λ=1 µ=1
transform of (4.81) as follow:
v ...v
δ2 tr11...rqp
= 0.
dt 2

Thus we have
p
Theorem 106. Let eγ be a geodesic on Tq (Mn ) of s ∇. Then the tensor field
v1 ...v p
tr1 ...rq (t) defined along γ satisfies the differential equations (4.82) and has van-
ishing second covariant derivative.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p
Let now eγ be a geodesic of s ∇. If eγ lies on a fibre π−1 (P) = Tq (P), P = P(xh )
v ...v p
δ2tr11...rq
given by xh = ch = const, then dt 2
= 0 reduces to
v ...v
d 2tr11...rqp
= 0, (dxh = 0)
dt 2
from which we have
xr = ar t + br , r = n + 1, ..., n + n p+q ,
ar and br being constant. Hence we have
p
Theorem 107. If a geodesic e γ lies in a fibre of Tq (Mn ) with respect to the metric
s g, the geodesic e
γ is expressed by linear equations

xh = ch ,
x = ah t + bh ,
h

where ch , ah and bh are constant.

Since the parameter t in (4.77) is the arc length of eγ, we have

ωβ ωγ
gβγ( )( ) = 1.
dt dt
Using (4.66) and (4.71) we find
1 p i ...i t ...t
dx j dxl j1 l1
δt 1 p δt
jqlq j1 ... jq l1 ...lq
g jl + gi1 t1 ...gi pt p ge ...e g = 1. (4.83)
dt dt dt dt
On the other hand, (4.83) implies
i ...i t ...t
δ j1 l1
δt 1 p δt 1 p
jqlq j1 ... jq l1 ...lq
(gi t ...gi pt p ge ...e g ) = 0,
dt 1 1 dt dt
i ...i t ...tp
δt j1 ... jpq δtl 1 ...lq
i.e. gi1t1 ...gi pt p gej1 l1 ...e
g jq lq 1
dt
1
dt = const. Then we find that
2
ds dx j dxl
= g jl = const.
dt dt dt
and hence:
Theorem 108. If t and s are the arc lengths of the geodesic eγ of s ∇ and the
projection curve γ = πeγ, respectively, then s and t are related linearly: s =
αt + β, where α and β constants.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

i ...i
Next, let γ be a curve on M expressed locally by xh = xh (t) and S j11 ... jpq (t) be
p
a tensor field of type (p, q) along γ. Then, on the tensor bundle Tq (Mn ) over the
Riemannian manifold Mn , we define a curve γ by H
(
xh = xh (t),
k ...k
xh = Sh11...hpq (t).

If the curve H γ satisfies at all the points the relation

k ...k
δSh11 ...hpq
= 0, (4.84)
dt
i ...i
i.e. S j11 ... jpq (t) is a parallel tensor field along γ, then the curve H γ is said to be a
horizontal lift of γ. From (4.82) and (4.84), we obtain
δ2 xh
= 0.
dt 2
Thus we have

Theorem 109. The horizontal lift of a geodesic on M is always geodesic on

Tqp (M) with respect to the connection s ∇.

Definition 20. We define the horizontal lift H ∇ of Levi-Civita connection on M

to Tqp (M) by the conditions
H ∇V V B = 0, H ∇V H Y = 0,
A A
H (4.85)
∇H X V B =V (∇X B) , H ∇H X H Y =H (∇X B)
p
for any X,Y ∈ ℑ10 (M) and A, B ∈ ℑq (M).

Denoting by T and Te, respectively, the torsion tensors of ∇ and H ∇, we have

from (4.85)

Te V A,V B =H ∇V A V B −H ∇V B V A − [V A,V B] = 0,

Te V
A,H Y = −Te H Y,V A =H ∇V A H Y −H ∇H Y V A − [V A,H Y ]
= −V (∇Y A) +V (∇Y A) = 0,

= H (∇X Y ) −H (∇Y X) −H [X,Y ] + (eγ − γ) R(X,Y )

= H (T (X,Y )) + (eγ − γ) R(X,Y ).
p
Since the vector fields H X and V A for any X ∈ ℑ10 (Mn ) and A ∈ ℑq (Mn ) span
the module of vector fields on Tqp (M), then the tensor field Te is determined by
its action of H X and V A. Therefore, we have

Theorem 110. The connection H ∇ has nontrivial torsion even for the Levi-
Civita connection ∇ determined by g, unless g is locally flat.

We have, from (4.63)-(4.65) and (4.85)

(H ∇V C s g)(V A,V B) =V CV (g(A, B)) = 0,

(H ∇V C s g)(V A,V B) = H V
Z (g(A, B)) −s g(V (∇Z A),V B) −s g(V A,V (∇Z B))
V
= (Zg(A, B)) − g(∇Z A, B) − g(A, ∇Z B)
i ...i t ...t
= A j11 ... jpq Bl11 ...lpq (∇Z (gi1t1 ...gi pt p gej1 l1 ...e
g jq lq ) = 0,

(H ∇V C s g)(V A,H Y ) = 0,
(H ∇H Z s g)(V A,H Y ) = 0,
(H ∇V C s g)(H X,V B) = 0,
(H ∇H Z s g)(H X,V B) = 0,
(H ∇V C s g)(H X,H Y ) =V CV (g(X,Y)) = 0,
(H ∇H Z s g)(H X,H Y ) =V ((∇Z g)(X,Y)) = 0
for any A, B,C ∈ ℑqp (Mn ) and X,Y, Z ∈ ℑ10 (Mn ). Thus, from last equations we
have
Theorem 111. The horizontal lift H ∇ of the Levi-Civita connection ∇ to
p
Tq (Mn ) is a metric connection with respect to the Sasakian metric, i.e. H ∇ s g.
We now put
H
∇α =H ∇eα ,
where {eα } = {ei , ei }−adapted frame. Then
H γ
∇α eβ =H Γαβ eγ . (4.86)
Thus, taking account of (4.85) and writing expression (4.86) for the different
indices, we find

Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

H r
 Γl j =H Γrl j =H Γrl j =H Γrl j =H Γrl j =H Γrl j = 0,

 H r
Γl j = Γrl j ,
p q (4.87)

 H r vλ j1 jq l1 s lp s j1 jµ jq l1 lp
 Γ = ∑ Γ δ ...δ δ ...δ ...δ − ∑ Γ δ ...δ ...δ δ ...δ
lj ls r1 rq i1 iλ ip lrµ r1 s rq i1 ip
λ=1 µ=1

with respect to the adapted frame.

A curve eγ in Tqp (Mn ) is a geodesic with respect to the horizontal connection
H
∇ when it satisfies
d ωε H ε ωγ ωβ
( ) + Γγβ =0 (4.88)
dt dt dt dt
with respect to the adapted frame. As a consequence of (4.87), the equation
(4.88) is equivalent to

 d 2 xr l dx j

 dt 2
+ Γrl j dx
dt dt = 0,

 v ...v p
 d δtr11...rqp vλ j1 jq l1 s lp
dt ( dt ) + ( ∑ Γls δr1 ...δrq δi1 ...δiλ ...δi p − (4.89)
λ=1

 q i1 ...i p

 s j1 jµ jq l1 l δt
l p dx j1 ... jq

 − ∑ Γlrµ δr1 ...δs ...δrq δi1 ...δi p ) dt dt = 0,
µ=1

where second equation of (4.89) reduces to

v ...v
δ2 tr11...rqp
= 0.
dt 2
Thus we have
p
Theorem 112. Let eγ be a geodesic on Tq (Mn ) with respect to the metric con-
H s
nection ∇ of Sasakian metric g. Then the projection γ of eγ is a geodesic with
v ...v
respect to the Levi-Civita connection ∇ on Mn and the tensor field tr11...rqp (t) de-
fined along γ has vanishing second covariant derivative.

4.5.4. Jacobi Tensor Fields

We consider other possible connections on Tqp (Mn ). The complete lift c ∇ of
p
Levi-Civita connection ∇ to Tq (Mn ), defined by
 c
 ∇V AV B = 0, c ∇V A H Y = 0,
c ∇H V B =V (∇ B), (4.90)
X
 c XH
∇H X Y =H (∇X Y ) + (γ − eγ)(R( , X)Y )
p
for any A, B ∈ ℑq (Mn ) and X,Y ∈ ℑ10 (Mn ), where R( , X)Y denotes a tensor field
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

ϕ of type (1, 1) on M, such that ϕ(Z) = R(Z, X)Y for any Z ∈ ℑ10 (Mn ), R being
the curvature tensor of ∇.
If we put
e Ye ) =H ∇ eYe −C ∇ eYe
S(X, X X
p
e Ye ∈ ℑ1 (Tq (Mn )), then the tensor field S of type (1, 2) satisfies the
for any X, 0
conditions
V V
S( A, B) = 0, S(V A,H Y ) = 0,
(4.91)
S(H X,V B) = 0, S(H X,H Y ) = (γ − eγ)(R( , X)Y )

for any A, B ∈ ℑqp (Mn ) and X,Y ∈ ℑ10 (Mn ). Therefore S has components Sεγβ
such that
p q
v ...m...v p vλ v1 ...v p
Shlj = − ∑ tr11...rq Rml j + ∑ tr ...m...r Rmr l j ,
1 q µ
(4.92)
λ=1 µ=1

all the others being zero, with respect to the adapted frame.

Since the components of H ∇ are given by (4.87), it follows from (4.91) and
(4.92) that complete lift C ∇ has components
 c r c r c r c r c r c r
 Γ = Γl j = Γl j = Γl j = Γl j = Γl j = 0 ,

 lj c Γr = Γr ,
lj lj (4.93)
 p q
 c r v1 ...m...v p vλ v ...v p
 Γ = ∑ tr ...r
lj R − ∑ tr 1...m...r
1 q Rm ml j 1 q rµ l j
λ=1 µ=1

with respect to the adapted frame.

p
A curve eγ is, by definition, a geodesic in Tq (Mn ) with respect to C ∇ if and
only if it satisfies the differential equation

d ωα c α ωγ ωβ
( ) + Γγβ = 0. (4.94)
dt dt dt dt
By means of (4.93), (4.94) reduces to

 δ dxi
 dt ( dt ) = 0 ,
i ...i p
δ2 t j1 ... jq p i ...m...i p i
q i ...im dx l dx j (4.95)


1
dt 2
+ ( ∑ t j11 ... jq λ
Rml 1 p
j − ∑ t j1 ...m... jq R jµ l j ) dt dt = 0.
λ=1 µ=1

The second equation of (4.95) shows that the tensor field t j11 ... jpq (t) on Mn defined
along γ = πeγ is a Jacobi tensor field ( a Jacobi vector field for p = 1, q = 0) along
γ, where γ is a geodesic on Mn . Thus we have

Theorem 113. Let eγ be a geodesic on Tqp (Mn ) with respect to the C ∇ =H ∇ + S,

where H ∇ is a metric connection of Sasakian metric s g, S is a tensor field of type
(1, 2) defined by conditions (4.91). Then the projection γ = πeγ is a geodesic on
i ...i
M with respect to ∇ and the tensor field t j11 ... jpq (t) defined along γ is a Jacobi
tensor field along the geodesic γ.

4.6. Some Special Cases

In this section, we present some results concerning the study of φϕ −operators
in the geometry of tangent and cotangent bundles.

4.6.1. Para-Nordenian Structures in Cotangent Bundles

Let Mn be an n−dimensional differentiable Riemannian manifold of class
C∞ and with metric g, C T (Mn ) its cotangent bundle, and π the natural projec-
tion C T (Mn ) → Mn . A system of local coordinates (U, xi ), i = 1, ..., n in Mn in-
duces on C T (Mn ) a system of local coordinates (π−1 (U), xi , xi = pi ), i = n + i =
n + 1, ..., 2n, where xi = pi is the components of covectors p in each cotangent
space C Tx(Mn ), x ∈ U with respect to thenatural coframe {dxi }.
We denote by ℑrs (Mn ) ℑrs (C T (Mn )) the modul over F(Mn )(F(C T (Mn )))
of C∞ tensor fields of type (r, s), where F(Mn )(F(C T (Mn ))) is the ring of real-
valued C∞ functions on Mn (C T (Mn )) .
Let X = X i ∂x∂ i and ω = ωi dxi be the local expressions in U ⊂ Mn of a vector
and covector (1−form) fields X ∈ ℑ10 (Mn ) and ω ∈ ℑ01 (Mn ), respectively. Then
the complete and horizontal lifts C X,H X ∈ ℑ10 (C T (Mn )) of X ∈ ℑ10 (Mn ) and the
vertical lift V ω ∈ ℑ10 (C T (Mn )) of ω ∈ ℑ01 (Mn ) are given, respectively, by

C ∂ ∂
X = Xi − ph ∂i X h
∂xi ∑
, (4.96)
i ∂xi

V ∂
ω = ∑ ωi (4.98)
i ∂xi
with respect to the natural frame { ∂x∂ i , ∂x∂ i }, where Γhij are components of the
Levi-Civita connection ∇g on Mn (see [114] for more details).
For each x ∈ Mn the scalar product g−1 = (gi j ) is defined on the cotangent
space π−1 (x) =C T (Mn ) by

g−1 (ω, θ) = gi j ωi θ j

for all ω, θ ∈ ℑ01 (Mn ).

A Sasakian metric s g is defined on C T (Mn ) by the three equations
S
g(V ω,V θ) =V (g−1 (ω, θ)) = g−1 (ω, θ) ◦ π , (4.99)
S
g(V ω,H Y ) = 0, (4.100)
S
g(H X,H Y ) =V (g(X,Y )) = g(X,Y) ◦ π (4.101)

for any X,Y ∈ ℑ10 (Mn ) and ω, θ ∈ ℑ01 (Mn ). Since any tensor field of type (0, 2)
on C T (Mn ) is completely determined by its action on vector fields of type H X
and V ω (see [114, p.280]), it follows that S g is completely determined by equa-
tions (4.99), (4.100) and (4.101).
We now see, from (4.96) and (4.97), that the complete lift C X of X ∈ ℑ10 (Mn )
is expressed by
C
X =H Y −V (p(∇X)), (4.102)
where p(∇X) = pi (∇h X i )dxh .
Using (4.99), (4.100), (4.101) and (4.102), we have
S
g(C X,C Y ) =V (g(X,Y )) +V (g−1 (p(∇X), p(∇Y ))), (4.103)

where g−1 (p(∇X), p(∇Y )) = gi j (pl ∇i X i )(pk ∇ jY k ).

Since the tensor field S g ∈ ℑ02 (C T (Mn )) is completely determined also by
its action on vector fields of type C X and C Y (see [114, p.237]), we have an
alternative characterization of S g : Sasakian metric S g on C T (Mn ) is completely
determined by the condition (4.103).
In U ⊂ Mn , we put

Then from (4.97) and (4.98) we see that H X(i) and V θ(i) have respectively local
expressions of the form

∂ ∂
ee(i) =H X(i) = i
+ ∑ pa Γahi , (4.104)
∂x h ∂xh

∂
ee(i) =V θ(i) = . (4.105)
∂xi
n o n o
We call the set ee(α) = e e(i) , ee(i) = H X(i) ,V θ(i) the frame adapted to the
Levi-Civita connection ∇g . The indices α, β, ... = 1, ..., 2n indicate the indices
with respect to the adapted frame.
We now, from equations (4.97), (4.98), (4.104) and (4.105) see that H X and
V ω have respectively components

i
H i H H α X
X = X ee(i) , X =( X )= , (4.106)
0

Salimov, Arif. Tensor Operators and their Applications, Nova Science Publishers, Incorporated, 2012. ProQuest Ebook
Central, http://ebookcentral.proquest.com/lib/unilu-ebooks/detail.action?docID=3022757.
Created from unilu-ebooks on 2023-10-16 21:16:26.
Applications to the Theory of Lifts 155

0
V
ω = ∑ ωi ee(i), V
ω = (V ωα ) = (4.107)
i ωi

with respect to the adapted frame ee(α) , where X i and ωi being local compo-
nents of X ∈ ℑ10 (Mn ) and ω ∈ ℑ01 (Mn ), respectively.
Let S ∇ be the Levi - Civita connection determined by the Sasakian metric
S
g. The components of S ∇ are given by (see Section 4.5)
 S h

 Γ ji = Γhji , S Γhji =S Γhji =S Γhji = 0,
Γ ji = 12 pm Rh. j.im , S Γhji = 12 pm Rh. i.im ,
S h
(4.108)

 S Γh = 1 p R m , S Γh = −Γi
ji 2 m jih ji jh

with respect to the adapted frame ee(α) , where Rl jı α being local components
of the curvature tensor R of ∇g .
Let now X,e Ye ∈ ℑ1 (C T (Mn )) and Xe = Xeα eeα , Ye = Ye β eeβ . The covariant
0
derivative ∇Ye Xe along Ye has components
S

S e α = Ye γ eeγ X
∇Ye X e α +S Γα XeβYe γ (4.109)
γβ

with respect to the adapted frame ee(α) .
Using (4.106), (4.107), (4.108) and (4.109), we have
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Theorem 114. Let Mn be a Riemannian manifold with metric g and S ∇ be the

Levi – Civita connection of the cotangent bundle C T (Mn ) equipped with the
Sasakian metric S g. Then S ∇ satisfies

i) S ∇V ω V θ = 0,

1H
ii) S ∇V ω H Y = (P(g−1 ◦ R( ,Y )ω
e )),
2
1H
iii) S ∇H X V θ =V (∇X θ) + (P(g−1 ◦ R( , X)e
θ)),
2
1V
iv) S ∇H X H Y =V (∇X Y ) + (PR(X,Y))
2
for all X,Y ∈ ℑ10 (Mn ) and ω, θ ∈ ℑ01 (Mn ), where ω e = g−1 ◦ ω ∈
1 1 −1
e ∈ ℑ1 (Mn ), g ◦ R( , X)ω
ℑ0 (Mn ), R( , X)ω 2
e ∈ ℑ0 (Mn ),

Let (C T (Mn ),S g) be a cotangent bundle with the Sasakian metric S g. We

define a tensor field F of type (1, 1) on C T (Mn ) by
H
F X =V Xe ,
(4.110)
F V ω =H ωe

for any X ∈ ℑ10 (Mn ) and ω ∈ ℑ01 (Mn ), where Xe = g ◦ X ∈ ℑ01 (Mn ), ω
e = g−1 ◦ ω ∈
1
ℑ0 (Mn ). Then we obtain
F 2 = I.
In fact, we have by virtue of (4.110)

e =H e
e =H X,
F 2 (H X) = F(F H X) = F(V X) X
F 2 (V ω) = F(F V ω) = F(H ω
e ) =V e
e =V ω
ω

for any X ∈ ℑ10 (Mn ) and ω ∈ ℑ01 (Mn ), which implies F 2 = I.

Theorem 115. The triple (C T (Mn ),S g, F) is an almost para-Nordenian mani-

fold.

A(X,
e Ye ∈ ℑ1 (C T (Mn )). From (4.99), (4.100), (4.101) and (4.110), we have
for any X, 0

A(H X,H Y ) = S
g(F H X,H Y ) −S g(H X, F H Y )
e H Y ) −S g(H X,V Ye ) = 0,
= S g(V X,

A(H X,V ω) = S
g(F H X,V ω) −S g(H X, F V ω)
e V ω) −S g(H X,H ω
= S g(V X, e)
= g−1 (g ◦ X, ω) − g(X, g−1 ◦ ω) = 0,

A(V ω,H Y ) = −A(H Y,V ω) = 0,

A(V ω,V θ) = S
g(FV ω,V θ) −S g(V ω, FV θ)
e ,V θ) −S g(V ω,H e
= g−1 (H ω θ) = 0,

i.e. S g is pure with respect to F. Thus Theorem 115 is proved.

We now consider the covariant derivative of F. Taking account i)-iv) of

Theorem 114 and (4.110), we obtain
S

∇H X F H Y

= S ∇H X F H Y − F S ∇H X H Y =S ∇H X V Ye − F S ∇H X H Y
1H V
−1
1
= V
∇X Ye + p g ◦ R ( , X)Y − F H
(∇X Y ) + (pR(X,Y ))
2 2
1H
= pg−1 ◦ (R ( , X)Y − R(X,Y )) ,
2

S
H
S

∇V ωF Y = ∇V ω F H Y − F S ∇V ω H Y
1
= S ∇V ωV Ye − F H p g−1 ◦ R ( ,Y ) ω
e
2
1V
= − (pR( ,Y )ω e ),
2

1V
= H ∇X e θ + pR X, eθ −H g−1 ◦ (∇X θ)
2
V
1
− pg ◦ g−1 ◦ R( , X)eθ
2
1V
= pR(X, e
θ) − pR( , X)eθ ,
2

S
V
S

∇V ω F θ = ∇V ω F V θ − F S ∇V ωV θ
1 H −1
= S ∇V ω H e
θ= p g ◦ R ,e e .
θ ω
2
Using Theorem 44, from last equations we have

Theorem 116. The cotangent bundle of a Riemannian manifold is para-

Kahlerian (paraholomorphic Nordenian) with respect to the metric S g and al-
most paracomplex structure F defined by (4.102) if and only if the Riemannian
manifold is flat.

where (LX ∇)Y = ∇Y ∇X + R(X,Y ) and (LX ∇) (Y, Z) = LX (∇Y X) − ∇Y (LX Z) −

∇[X,Y ] Z .
A vector field X ∈ ℑ10 (Mn ) is called a Killing vector field (or an infinitesi-
mal isometry) if LX g = 0, and X is called an infinitesimal affine transformation
if LX ∇g = 0. A Killing vector field is necessarily an infinitesimal affine trans-
formation, i.e. we have LX ∇g = 0 as a consequence of LX g = 0.
We now consider the Lie derivative of F with respect to the complete lift
C
X. Taking account of (4.110) and (4.111), we obtain

= LC X H e
θ −H g−1 ◦ (LX θ)

= V [X, e
θ] +V p (LX ∇) e
θ −H g−1 ◦ (LX θ)

= H L g−1 ◦ θ − g−1 ◦ (LX θ) +V p (LX ∇) e θ ,

(LC X F)H Y = LC X F H Y − F LC X H Y (4.113)

= LC X V Ye − F H [X,Y ] +V (p (LX ∇)Y ) .

Let now be a Killing vector field (LX g = 0). Then by virtue of LX ∇ = 0

from (4.112) and (4.113) we have LC X F = 0, i.e. C X is a paraholomorphic with
to F. If we assume that LC X F = 0 and calculate the equation (4.113) at
respect
i
x , 0 , pi = 0 we get LX (g ◦Y ) = g ◦ LX Y from, which follows that LX g = 0.
We hence have

Theorem 117. An infinitesimal transformation X of Riemannian manifold

(Mn , g) is a Killing vector field if and only if its complete lift C X to cotangent
bundle C T (Mn ) is an almost paraholomorphic
vector field with respect to the
almost para-Nordenian structure F, g . S

Let now R ∇ ∈ ℑ02 C T (Mn ) be a Riemannian extension of the connection
∇g defined by [2], [114, p. 268]

R
∇ C X,C Y = −p (∇X Y + ∇Y X) , X,Y ∈ ℑ10 (Mn ).

The metric R ∇ has components

!
j
R −2pa Γaji δi
∇= (4.114)
δij 0

with respect to the natural frame {∂i , ∂i }. From (4.97), (4.98) and (4.118) we
easily see that
R

∇ H X,H Y = 0, R ∇ V ω,V θ = 0, R ∇ H X,V θ =V (θ (X)) , (4.115)

i.e. the metric R ∇ is completely determined also by conditions (4.115). Using

(4.101), (4.110) and (4.115), we have

R
∇ ◦ F H X,H Y = R ∇ F H X,H Y =R ∇ V X, e H Y =V X(Y e )

= V (g(X,Y )) =S g H X,H Y ,

R
∇ ◦ F H X,V θ =R ∇ F H X,V θ =R ∇ V X, e V θ =S g H X,V θ = 0,

R

∇ ◦ F V ω,H Y =R ∇ F V ω,H Y =R ∇ H ω e ,H Y =S g V ω,H Y = 0,

R
V

∇◦F ω,V θ = R
∇ F V ω,V θ =R ∇ H ωe ,V θ =V (θ (ω
e ))

= V g−1 (ω, θ) =S g V ω,V θ ,

i.e. R ∇ ◦ F =S g. Thus we have

Theorem 118. The almost para-Nordenian structure F determined by the con-

−1 S
dition (4.110) has an expression in the form F = R ∇ ◦ g.

4.6.2. Paraholomorphic Cheeger-Gromoll Metric in the Tangent

Bundle
In [9], Cheeger and Gromoll study complete manifolds of nonnegative curva-
ture and suggest a construction of Riemannian metrics useful in that context.
Inspired by a paper of Cheeger and Gromoll, in [55] Musso and Tricerri defined
a new Riemannian metric CG g on tangent bundle of a Riemannian manifold
which they called the Cheeger-Gromoll metric. The Levi-Civita connection of
CG g and its Riemannian curvature tensor are calculated by Sekizawa in [91]
(for more details see [27], [28]). The main purpose of this section is to inves-
tigate a paraholomorphic Cheeger-Gromoll metric with respect to the natural
paracompex structure on the tangent bundle.
Let Mn be a Riemannian manifold with metric g, and let T (Mn ) be a tan-
gent bundle of Mn , and π the projection π : T (Mn ) → Mn . Let the manifold
Mn be a covered by a system of coordinate neighbourhoods (U, xi ), where
(xi ), i = 1, ..., n is a local coordinate system defined in the neighbourhood U.
Let (yi ) be the cartesian coordinates in each tangent spaces Tp (Mn ) at p ∈ Mn
with respect to the natural base { ∂x∂ i }, P being an arbitrary point in U whose co-
ordinates are xi . Then we can introduce local coordinates (xi , yi ) in the open
set π−1 (U) of T (Mn ). We call them coordinates induced in π−1 (U) from
(U, xi ). The projection π is represented by (xi , yi ) → (xi ). We use the nota-
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

tions xl = (xi , xi ) and xi = yi . The indices I, J, ... run from 1 to 2n, the indices
i, j, ... from 1 to n and the indices i, j, ... from n + 1 to 2n.
Let X ∈ ℑ10 (Mn ), which locally are represented by

∂
X = X i ∂i (∂i = ).
∂xi
Then the vertical and horizontal lifts V X and H X of X are given by

V ∂
X = X i ∂i (∂i = )
∂xi
and
H
X = X i ∂i − Γijk x j X k ∂i ,
where Γijk are the coefficients of the Levi-Civita connection on Mn .
Suppose that we are given in U ⊂ Mn a vector field X ∈ ℑ10 (Mn ) and a covec-
tor field gX = (gi j X i dx j ). Then we define a function γgX in π−1 (U) ⊂ T (Mn ) by

γgX = x j gi j X i with respect to the induced coordinates (xi , xi ). The function γgX
defined in each π−1 (U) determine global function on T (Mn ), which is denoted
also by γgX . Now, let r be the norm a vector y = (yi ) = (xi ), i.e. r2 = gi j xi x j .
The Cheeger-Gromoll metric CG g on the tangent bundle T (Mn ) is given by

(i) CG g(H X,H Y ) =V (g(X,Y)),

(ii) CG g(H X,V Y ) = 0

1 V
(ii) CG g(V X,V Y ) = [ (g(X,Y )) + (γgX )(γgY )]
1 + r2
for all vector field X,Y ∈ ℑ10 (Mn ) , where V (g(X,Y)) = (g(X,Y)) ◦ π.
The diagonal lift D ϕ on T (Mn ) is defined by
D H
ϕ X =H (ϕX),
D ϕV X = −V (ϕX)

for X ∈ ℑ10 (Mn ) any and ϕ ∈ ℑ11 (Mn ). The diagonal lift D I of the identity tensor
field I ∈ ℑ11 (Mn ) has the components
!
D δij 0
I= j
−2yt Γti −δij
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

with respect to the induced coordinates and satisfies (DI)2 = IT (Mn ). Thus D I is
an almost paracomplex structure determining the horizontal distribution and the
distribution consisting of the tangent planes to fibres.
We put
e Ye ) =CG g(DI Xe , Ye) −CG g(X,
S(X, e D IYe ).
e Ye ) = 0 for all vector fields Xe and Ye which are of the form V X,V Y
If S(X,
or H X,H Y, then S = 0. By virtue of D IV X = −V X, D I H X =H X and (i)-(iii) we
have
S(V X,V Y ) =CG g(−V X,V Y ) −CG g(V X, −V Y ) = 0,
S(V X,H Y ) =CG g(−V X,H Y ) −CG g(V X,H Y ) = 0,
S(H X,V Y ) =CG g(H X,V Y ) −CG g(H X, −V Y ) = 0,
S(H X,H Y ) =CG g(H X,H Y ) −CG g(H X,H Y ) = 0,
i.e. CG g is pure metric with respect to D I.
We hence have:

Theorem 119. (T (Mn ),D I,CG g) is an almost paracomplex Riemannian mani-

fold.

Let CG ∇ be the Levi-Civita connection of CG g. Using properties of V X,H X

and γR(X,Y ) = ys Rkijs X iY j ∂k (see [28] and [114]), i.e.
∂x

V
X V (g(Y, Z)) = 0,

[V X,V Y ] = 0,
[V X,H Y ] =V [X,Y ] −V (∇X Y ) = −V (∇Y X),
H
X V (g(Y, Z)) =V (Xg(Y, Z)),
CG ∇H VY 1 H
X = 2(1+r 2)
(R(xi ,Y )X +V (∇X Y ) (R-curvature tensor field of)

[H X,H Y ] =H [X,Y ] − γR(X,Y ),

i)(ΦDI CG g)(V X,V Y,V Z)

= (DIV X)(CGg(V Y,V Z)) −V X(CGg(D IV Y,V Z))
+CG g((LV Y D I)V X,V Z) +CG g(V Y, (LV Z D I)V X)
= −V X(CG g(V Y,V Z)) +V X(CG g(V Y,V Z))
+CG g(LV Y (D IV X) −D ILV Y V X,V Z)
+CG g(V Y, LV Z (DIV X) −D ILV Z V X)
CG
= g([V X,V Y ] −D I[V Y,V X],V Z)
+CG g(V Y, [V X,V Z] −D I[V Z,V X])
CG
= g(0,V Z) +CG g(V Y, 0) = 0.

ii)(ΦDI CG g)(V X,V Y,H Z)

= (DIV X)(CGg(V Y,H Z)) −V X(CG g(DIV Y,H Z))
+CG g((LV Y D I)V X,H Z) +CG g(V Y, (LH Z D I)V X)

= −V X.0 +V X.0
+CG g(LV Y (D IV X) −D ILV Y V X,H Z)
+CG g(V Y, LH Z (D IV X) −D ILH Z V X)
CG
= g(−[V Y,V X]D I.0,H Z)
+CG g(V Y, −[H Z,V X] −D I[H Z,V X])
CG
= g(0,H Z)
+CG g(V Y,V [X, Z] −V (∇X Z) +D I(V [X, Z] −V (∇X Z))
CG
= g(V Y, 0) = 0.

iii)(ΦDICG g)(V X,H Y,V Z)

= (DIV X)(CGg(H Y,V Z)) −V X(CG g(DI H Y,V Z))
+CG g((LHY D I)V X,V Z) +CG g(HY, (LV Z D I)V X)
= −V X.0 +V X.0 +CG g(LHY (D IV X) −D ILH Y V X,V Z)
+CG g(H Y, LV Z (D IV X) −D ILV Z V X)
CG
= g(−[H Y,V X] −D I[H Y,V X],V Z)
+CG g(H Y, −[V Z,V X] −D I[V Z,V X])
= CG g(V [X,Y ] −V (∇X Y ) +D I(V [X,Y ] −V (∇X Y )),V Z)
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

+CG g(H Y, 0)
CG
= g(V [X,Y ] −V (∇X Y ) −V [X,Y ] +V (∇X Y )),V Z)
CG
= g(0,V Z) = 0.

iv)(ΦDI CG g)(H X,V Y,V Z)

= (D I H X)(CGg(V Y,V Z)) −H X(CG g(DIV Y,V Z))
+CG g((LV Y D I)H X,V Z) +CG g(V Y, (LV Z D I)H X)

= H X CG g(V Y,V Z) +H X CG g V Y,V Z

+CG g LV Y D I H X −D I LV Y H X ,V Z

+CG g V Y, LV Z D I H X −D ILV Z H X

= 2H X CG g V Y,V Z

+CG g LV Y H X −D I V [Y, X] −V (∇Y X) ,V Z

+CG g V Y, LV Z H X −D I V [Z, X] −V (∇Z X)

v)(ΦDI CG g)(H X,H Y,H Z)

= (DI H X)(CG g(HY,H Z)) −H X(CG g(D I H Y,H Z))

+CG g((LHY D I)H X,H Z) +CG g(H Y, (LH Z D I)H X)
H
= X V g(Y, Z) +H X V (g (Y, Z))
H
+CG g LH Y D I X −D I LH Y H X ,H Z

+CG g H Y, LH Z D I H X −D ILH Z H X

= CG g [H Y, H X] −D I[H Y,H X],H Z

+CG g H Y, [H Z,H X] −D I[H Z, H X]
CG H

= g [Y, X] − γR(Y, X) −D I H
[Z, X] − γR(Y, X) ,H Z

+CG g H Y,H [Z, X] − γR(Z, X) −D I H [Z, X] − γR(Z, X)

= CG g H [Y, X] − γR(Y, X) −H [Y, X] +D IγR(Y, X),H Z

+CG g H Y,H [Z, X] − γR(Z, X) −H [Z, X] +D IγR(Z, X)

= −2 CG g γR(Y, X),H Z −CG g H Y, γR(Z, X) = 0.

vi)(ΦDI CG g)(V X,H Y,H Z)

= (DIV X)(CGg(H Y,H Z)) −V X(CG g(D I H Y,H Z))
+CG g((LHY D I)V X,H Z) +CG g(H Y, (LH Z D I)V X)

= −2V X CG g H Y,H Z

+CG g LH Y D IV X −D I[H Y,V X],H Z

+CG g H Y, LH Z D IV X −D I[H Z,V X]

= −2V X CG g H Y,H Z

+CG g −[H Y,V X] −D I V [Y, X] −V (∇Y X) ,H Z

+CG g H Y, −[H Z,V X] −D I V [Z, X] −V (∇Z X)

= −2V X CG g H Y,H Z

+CG g −[H Y,V X] +V [Y, X] −V (∇Y X) ,H Z

+CG g H Y, −[H Z,V X] +V [Z, X] −V (∇Z X)
= −2V X V (g(Y, Z))

+CG g −V [Y, X] +V (∇Y X) +V [Y, X] −V (∇Y X),H Z

+CG g H Y, −V [Z, X] +V (∇Z X) +V [Z, X] −V (∇Z X)

= −2V X V (g(Y, Z)) +CG g 0,H Z +CG g H Y, 0 = 0.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

vii)(ΦDI CG g)(H X,H Y,V Z)

= (DI H X)(CG g(H Y,V Z)) −H X(CG g(DI H Y,V Z))
+CG g((LHY D I)H X,V Z) +CG g(H Y, (LV Z D I)H X)

= H X CG g(H Y,V Z) −H X CG g(HY,V Z)
+CG g((LHY H X −D I[H Y,H X]),V Z)
+CG g(H Y, (LV Z H X −D I[V Z,H X])
CG

= g([HY,H X] −D I H
[Y, X] − γR(Y, X ),V Z)
+CG g(H Y, ([V Z,H X] +D IV (∇Y Z))
CG
= g(H [Y, X] − γR(Y, X) −H [Y, X] − γR(Y, X),V Z)
+CG g(H Y, −2V (∇X Z))
= −2CG g(γR(Y, X),V Z).

viii) (ΦDI CG g)(H X,V Y,H Z) = −2CG g V Y, γR(Z, X) is analogoue to vii).
Therefore, by virtue of Theorem 44 we have

Theorem 120. [75] The almost paracomplex Riemannian manifold (T (Mn ),

D I,CG g) is paraholomorphic if and only if M is flat.
n

4.6.3. On Almost Complex Structures in Tangent Bundles

Let ϕ ∈ ℑ11 (Mn ) and ω ∈ ℑ01 (Mn ). It is well known that for an almost holomor-
phic 1−form ω on a manifold with an almost complex structure ϕ, we have the
following equation (see Section 1.2.3 and Capter 2):

ω ◦ Nϕ = 0. (4.116)

Let ϕ ∈ ℑ11 (Mn ). Then, the complete lift C ϕ of ϕ along the cross-section ω
to T ∗ (Mn ) has components of the form

C ϕhi 0
ϕ=
(∂i ϕai − ∂h ϕai ) − ϕti ∂t ωh + ϕth ∂i ωt ϕhi

with respect to the adapted (B,C)−frame [114, p. 308]. We consider that the
local vector fields
i
C C ∂ C h ∂ X
X(i) = i
= δi h =
∂x ∂x 0
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

and
(i) 0
V
X V i
= (dx ) = V
(δih dxh ) =
δih
i = 1, ..., n; i = n + 1, ..., 2n span the module of vector fields in π−1 (U). Hence,
any tensor fields is determined in π−1 (U) by their actions on C X and V θ for any
X ∈ ℑ10 (Mn ) and θ ∈ ℑ01 (Mn ). The complete lift C ϕ has the properties
C C
ϕ( X) =C (ϕ(X)) + γ(LX ϕ),
C V (4.117)
ϕ( θ) =V (ϕ(θ)),

which characterize C ϕ, where ϕ(θ) ∈ ℑ01 (Mn ).

Theorem 121. Let Mn be a manifold with an almost complex structure ϕ. Then

the complete lift C ϕ ∈ ℑ11 (T ∗ (Mn )) of ϕ, when restricted to the cross-section de-
termined by an almost analytic 1−form on Mn , is an almost complex structure.

Proof. Let ϕ, ψ ∈ ℑ11 (Mn ) and N ∈ ℑ12 (Mn ). Using (4.20)-(4.24) and (4.117),
we have
γ(ϕ ± ψ) = γ(ϕ) ± γ(ψ), (4.118)
C
ϕ(γψ) = γ(ψ ◦ ϕ),
(γN)(C X) = γNX ,
where NX is the tensor field of type (1, 1) on Mn defined by NX (Y ) = N(X,Y )
for any Y ∈ ℑ10 (Mn ).If X ∈ ℑ10 (Mn ), then from (4.117) and (4.118), we have

(C ϕ)2 (C X) = (C ϕ ◦C ϕ)(C X) =C ϕ(C ϕ(C X)) (4.119)

C C
= ϕ( ϕ(X)) + γ(LX ϕ)
C
= ϕ(C (ϕ(X))) +C ϕ(γ(LX ϕ))
C
= (ϕ(ϕ(X))) + γ(LϕX ϕ) + γ((LX ϕ) ◦ ϕ)
C
= (ϕ ◦ ϕ(X)) + γ(LϕX ϕ + (LX ϕ) ◦ ϕ)
= (C ϕ)2 (C X) − γ(LX (ϕ ◦ ϕ)) + γ(LϕX ϕ + (LX ϕ) ◦ ϕ)
= (C ϕ)2 (C X) + γ(LϕX ϕ − ϕ(LX ϕ))
= (C ϕ)2 (C X) + γ(Nϕ,X )
= (C ϕ)2 (C X) + (γNϕ )(C X),
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

where
Nϕ,X = (LϕX ϕ − ϕ(LX ϕ))(Y)
= [ϕX, ϕY ] − ϕ[X, ϕY ] − ϕ[ϕX,Y ] + ϕ2 [X,Y ] = Nϕ(X,Y )
is nothing but the Nijenhuis tensor constructed by and has local coordinates of
the form
0 0
γNϕ =
(ω ◦ Nϕ )i j 0
(see (4.24)).Similarly, if θ ∈ ℑ01 (Mn ), then by (4.117), we have
(C ϕ)2 (V θ) = (C ϕ ◦C ϕ)(V θ) (4.120)
C C V
= ϕ( ϕ( θ))
C
= ϕ(V (ϕ(θ))
V
= (ϕ(ϕ(θ)))
V
= ((ϕ ◦ ϕ)(θ))
= C
(ϕ2 )(V θ).

By virtue of (4.116), we can easily say that γNϕ = 0. From (4.119), (4.120) and
linearity of the complete lift, we have

(C ϕ)2 =C (ϕ2 ) =C (IMn ) = −IT ∗ (Mn ) .

This completes the proof.

4.7. Complete Lift of a Skew-Symmetric Tensor Field

Suppose now that S ∈ ℑ12 (Mn ) is a skew-symmetric tensor field of type (1, 2)
with local components Skij , that is, S(V,W) = −S(W,V), ∀V,W ∈ ℑ10 (Mn ). A
tensor field ξ ∈ ℑqp (Mn ) is called pure with respect to S ∈ ℑ12 (Mn ), if
( m...h h ...m h ...h h ...h
Shml1 2 ξk1 ...kpq = ... = Smlp 2 ξhk11...k q
= Sm 1 p m 1 p
k1 l2 ξm...kq = ... = Skq l2 ξk1 ...m ,
m...h h ...m h ...h h ...h
Shl11m ξk1 ...kpq = ... = Sl1pm ξhk11...k q
= Sm 1 p m 1 p
l1 k1 ξm...kq = ... = Sl1 kq ξk1 ...m .

∗
p p
Let ℑq (Mn ) denote a module of all the tensor fields ξ ∈ ℑq (Mn ) which are
pure with respect to S. We consider a pure cross-section σSξ (Mn ) determined by
∗
ξ ∈ ℑqp (Mn ). We define a tensor field c S ∈ ℑ12 (Tqp (Mn )) along the pure cross-
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

section σSξ (Mn ) by

 c c c c (S (V ,V )) + γ (L S)


 S( V 1 , V2 ) = 1 2 V2 V1
− γ (LV1 S)V 2 − γ S [V1 ,V2 ]


 −V ((LV2 S)V1 ◦ ξ) +V ((LV1 S)V2 ◦ ξ) +V ( S[V1 ,V2 ] ◦ ξ),
c S(V A,c V ) =V (S (A)),
2 V2

 c c

 S( V1 ,V B) =V (SV1 (B)),
 c S(V A,V B) = 0,

(4.121)
p
for any V1 ,V2 ∈ ℑ10 (Mn ), A, B ∈ ℑq (Mn ), where SV2 (A), SV1 (B), ((LV1 S)V2 ◦
ξ), ((LV2 S)V1 ◦ ξ), (S[V1 ,V2 ] ◦ ξ) ∈ ℑqp (Mn ) and call c S the complete lift of
p
S ∈ ℑ12 (Mn ) to Tq (Mn ), p ≥ 1, q ≥ 0 along σSξ (Mn ).
Let c SKL1 L2 be components of c S with respect to the adapted (B,C)−frame of
the pure cross-section σSξ (Mn ). Then, from (4.21), (4.23) and (4.121) we have

 ∼ ∼ ∼ ∼ ∼

 cSe K cV L1 cV L2 =c (S (V1 ,V2 ))K + γ((LV S)V )K − γ((LV S)V )K

 L1 L2 1 2 2 1 1 2


∼ ∼ ∼


K V K V
−γ(S[V1 ,V2 ] ) − ((LV2 S)V1 ◦ ξ) + ((LV1 S)V2 ◦ ξ) K

 ∼

+V ( S[V1 ,V2 ] ◦ ξ), (i)

 ∼ ∼
e K V AL1 cV L2 =V (SV (A))K , (ii)
cS

 L L

 1 2 2 2

 cS
∼ ∼
e K cV L1 V BL2 =V (SV (B))K , (iii)

 L L 1 1
 1 2
e K V AL1 V BL2 = 0, (iv)
cS
L1 L2
(4.122)
where

V
∼ 0 V
∼ 0
(SV1 (B)) = h1 j mh2 ...h p , (SV2 (A)) = h1 j mh2 ...h p ,
S jmV1 ξk1 ...kq Sm jV2 ξk1 ...kq

V
∼ 0
((LV1 S)V2 ◦ ξ) = h1 h2 ...l...h p j h ,
ξk1 ...kq V2 (LV1 Slmλ )

V
∼ 0
((LV2 S)V1 ◦ ξ) = h1 h2 ...l...h p j h ,
ξk1 ...kq V1 (LV2 Slmλ )
∼

V 0
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

( S[V1 ,V2 ] ◦ ξ) = h1 h2 ...l...h p hλ .

ξk1 ...kq Slm[V1 ,V2]m
Substituting K = k in of (4.122) and calculating in the same way as in Section
4.4 we obtain
cek
S l l =c Se kl l =c Se kl l = 0, c Se kl1 l2 = Skl1 l2 (4.123)
1 2 1 2 1 2

When K = k, of (4.122) reduces to

∼ ∼ ∼ ∼ ∼ ∼
c e k c l1 c l2
S l1 l2 V 1 V 2 +c Se kl l cV l 1 c l2
1 V 2 +c Se kl l cV l1 c l 2
1 V 2 (4.124)
1 2 1 2
lh ...h lh ...h lh ...h
−ξk12...kq p V1m (LV2 S)hlm1 + ξk12...kq p V2m (LV1 S)hlm1 + ξk12...kq p Shlm1 [V1 ,V2 ]m
p p
h h2 ...l...h p m hλ h h2 ...l...h p m h
+ ∑ ξk11...k q
V1 (LV2 S lm ) − ∑
ξk11...k q
V2 (LV1 Slmλ )
λ=2 λ=2
p
h h2 ...l...h p hλ
− ∑ ξk11...k q
Slm [V1,V2 ]m
λ=2
∼
c
= (S (V1 ,V2))k .

Now, we will study components c Se kl1 l2 ,c Se kl l ,c Se kl l and cSe kl l of the complete

1 2 1 2 1 2
lift c S. When K = k, (ii) and (iv) of (4.122) can be rewritten by virtue of (4.21)
and (4.123) as 0 = 0. For a case where K = k, of (4.122) we have c Se kl l = 0.
1 2
When K = k, from of (4.122) we write
c e k V l1 c e l2
S l1 l2 A V2 + c Se kl l V e l2
Al 1 cV2
1 2

+ c Se kl l V e l 2 + c Se k
Al1 cV2 l l
V e l2
Al 1 cV2
1 2 1 2
∼
V
= (SV2 (A))k

or ∼
c e k V l1 c e l2
S l l A V2 =V (SV2 (A))k ,
1 2

s ...s mh ...h r s ...s

cek
S l l Ar11 ...rqp V2l2 = Shm1jV2j Ak1 ...k
2
q
p
= δrk11 ...δkqq Shs11l2 δhs22 ...δhs pp Ar11 ...rqp V2l2
1 2

which implies
cek r
Sll = δrk11 ...δkqq Shs11l2 δhs22 ...δhs pp , (4.125)
1 2

s ...s
where xl 1 = tr11...rqp . We also have by of (4.122)

1 2

s ...s
where xl 2 = tr11...rqp .

Thus, by virtue of (4.125) and (4.126), (4.124) reduces to

c e k c e l1 c e l2 r
S l1 l2 V V2 + δrk11 ...δkqq Shs11l2 δhs22 ...δhs pp cV e l2
e l 1 cV
2 (4.127)
r
+δrk11 ...δkqq Shl11s1 δhs22 ...δhs pp cV e l 2 − ξ lh2 ...h p V1m (LV2 S)h1
e l1 cV
2 k1 ...kq lm
lh ...h lh ...h
+ξk12...kq p V2m (LV1 S)hlm1 + ξk12...kq p Shlm1 [V1 ,V2]m
p p
h h ...l...h p h h h ...l...h p h
+ ∑ ξk11...k
2
q
V1m (LV2 Slmλ ) − ∑ ξk11...k
2
q
V2m (LV1 Slmλ )
λ=2 λ=2
p
h h ...l...h p hλ
− ∑ ξk11...k
2
q
Slm [V1,V2 ]m
λ=2
∼
c
= (S (V1 ,V2))k .

Now, using the Yano-Ako operator we will investigate components c Se kl1 l2 . The
∗
Yano-Ako operator on the pure module ℑqp (Mn ) is given by (see Section 1.6)
h ...h
(ΨS ξ)l11l2 k1 ...k
p
q
h ...h h ...h h ...h
= Sm 1 p m 1 p m 1 p
l1 l2 ∂m ξk1 ...kq − ∂l1 (Sk1 l2 ξmk2 ...kq ) − ∂l2 (Sl1 k1 ξmk2 ...kq )
q q
h ...h hb hb h ...m...h p
+ ∑ (∂ka Sm
l1 l2 )ξk1 ...m...kq + ∑ (∂l1 Sml2 − ∂m Sl1 l2 )ξk1 ...kq
1 p 1

a=1 b=1
q
h ...m...h p
+ ∑ (∂l2 Shl1cm − ∂m Shl1cl2 )ξk11...kq .
c=1
After some calculations we have
h ...h mh ...h p
V1l1 V2l2 (ΨS ξ)l11l2 k1 p...kq +V1l1 Shl11m (LV2 ξ)k1 ...k
2
q
(4.128)
mh ...h mh ...h
+V2l2 Shml1 2 (LV1 ξ)k1 ...k
2
q
p 2
+ ξk1 ...k q
p l1
V1 (LV2 S)hl11m
mh ...hp l2 mh ...h
2
−ξk1 ...k q
V2 (LV1 S)hl21m − ξk1 ...k
2
q
p h1
Sl2 m [V1,V2 ]l2
p p
h h ...m...h p h h ...m...h p
− ∑ ξk11...k
2
q
V1l1 (LV2 Shl1bm ) + ∑ ξk11...k
2
q
V1l2 (LV1 Shl2bm )
b=1 b=1
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

p
h ...m...h p hb
+ ∑ ξk11...kq Sl2 m [V1 ,V2]l2
b=1
h ...h
= (LS(V1 ,V2 ) ξ)k11...kqp .
for any V1 ,V2 ∈ ℑ10 (Mn ). Using (4.20), (4.126) and (4.127), (4.128) reduces to

h ...h r
V1l1 V2l2 (ΨS ξ)l11l2 k1 p...kq − δrk11 ...δkqq Shs11l2 δhs22 ...δhs pp cV e l2
e l 1 cV
2 (4.129)
r
−δrk11 ...δkqq Shl11s1 δhs22 ...δhs pp cV e l 2 + ξmh2 ...h p V l1 (LV2 S)h1
e l1 cV
2 k1 ...kq 1 l1 m
mh ...hp l2 mh ...h
2
−ξk1 ...k q
V2 (LV1 S)hl21m − ξk1 ...k
2
q
p h1
Sl2 m [V1,V2 ]l2
p p
h h ...m...h p h h ...m...h p
− ∑ ξk11...k
2
q
V1l1 (LV2 Shl1bm ) + ∑ ξk11...k
2
q
V2l2 (LV1 Shl2bm )
b=2 b=2
p
h ...m...h p hb
+ ∑ ξk11...kq Sl2 m [V1 ,V2]l2
b=2
∼
= −c (S (V1 ,V2))k .

Comparing (4.127) and (4.129) we have

cek h ...h
S l1 l2 = − (ΨS ξ)l11l2 k1 p...kq .

Thus, the complete lift c S ∈ ℑ12 (Mn ) (S(V,W) = −S(W,V)) has along the pure
cross-section σSξ (Mn ) components [40], [86]

 e k =c S k , c Se k = − (ΨS ξ)h1 ...h p ,
cS

 l1 l2 l1 l2 l1 l2 l1 l2 k1 ...kq

 c Se k =c Sek =c Se k =c Se k = 0,
l 1 l2 l1 l 2 l1l2 l1 l 2 (4.130)
cek r1 rq h 1 h 2 hp

 S = δ ...δ S δ ...δ sp ,

 l 1 l2 k1 kq s1 l2 s2
 cS r
e k = δ 1 ...δ S 1 δs 2 ...δhs p
r q h h
l l
1 2 k1 kq l1 s1 2 p

h ...h
with respect to the adapted (B,C)− frame of σSξ (Mn ), where (ΨS ξ)l11l2 k1 ...k
p
q
is
the Yano-Ako operator.

Remark 30. c S in the form (4.130) is a unique solution of (4.121). Therefore,

∗
if S is element of ℑ12 (Tqp (Mn )), such that
 ∗
 S(cV1 ,c V2 ) =c (S(V1,V2 )) + γ((LV2 S)V1 ) − γ((L

 V1 S)V2 ) − γ(S[V1 ,V1 2 ] )
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.


 V V V

 − ((LV S) V1 ◦ ξ) + ((L V S)V2 ◦ ξ) + ( S [V ,V ] ◦ ξ), ∀V1 ,V2 ∈ ℑ0 (Mn ),
 2
∗
1 1 2
p
S(V A,c V2 ) =V (SV2 (A)), ∀A ∈ ℑq (Mn ),

 ∗

 S(cV1 ,V B) =V (SV1 (B)), ∀B ∈ ℑqp (Mn ),



 ∗
S(V A,V B) = 0,
∗
then S =c S.

Remark 31. The equation (4.121) is useful extension of the equation c L(ια) =
p
ι(LV α), α ∈ ℑq (Mn ) (see Section 4.2) to tensor fields of type (1,2) along the
pure cross-section σSξ (Mn ).

Remark 32. Let T and Nϕ be a torsion tensor field of connection ∇ and the
Nijenhuis tensor field of ϕ ∈ ℑ11 (Mn ), respectively. From (4.130) we easily see
that the complete lifts c T and c Nϕ to Tqp (Mn ) along σξ (Mn )are zero if T and Nϕ
are zero in the base manifold Mn .

Salimov, Arif. Tensor Operators and their Applications, Nova Science Publishers, Incorporated, 2012. ProQuest Ebook
Central, http://ebookcentral.proquest.com/lib/unilu-ebooks/detail.action?docID=3022757.
Created from unilu-ebooks on 2023-10-16 21:16:26.
Applications to the Theory of Lifts 173
h ...h
Remark 33. In the case of ∂m ξk11...kqp = 0, (B,C)−frame is considered as a nat-
ural frame {∂h , ∂eh } of σSξ (Mn ). Then, from (4.130) we obtain components of c S
along the pure-cross section



 Sl1 l2 = Skl1 l2 , Se lk l =c Se kl l =c Se kl l = 0,
c k


1 2 1 2 1 2

 cSe k = δr1 ...δrq Sh1 δh2 ...δh p ,c Se k = δr1 ...δrq Sh1 δh2 ...δh p ,
 l 1 l2 k1 kq s1 l2 s2 sp l1 l 2 k1 kq l1 s1 s2 sp
q
c k h ...h h ...h h ...h
 Sl1 l2 = ∂l1 (Sm 1 p m 1 p m 1 p
k1 l2 )ξmk2 ...kq + ∂l2 (Sl1 k1 )ξmk2 ...kq − ∑ (∂ka Sl1 l2 )ξk1 ...m...kq −

 a=1

 q q

 hb hb h1 ...m...h p hc h h1 ...m...h p
 − (∂
∑ l1 ml2
S − ∂ S )ξ
m l1 l2 k1 ...kq − ∑ l2 l1 m − ∂m Sl1cl2 )ξk1 ...kq .
(∂ S
b=1 c=1

[1] R. Abraham and J. Marsden, Foundations of Mechanics, Ben-

jamin/Cummings Publishing Co., Inc., Advanced Book Program, Read-
ing, Mass., 1978.

[2] S. Aslanci, S. Kazimova and A.A. Salimov, Some notes concerning Rie-
mannian extensions, Ukrainian Math. J. 62(2010), 661-675.

[3] R.L. Bishop and B. O’Neill, Manifolds of negative curvature, Trans. Am.
Math. Soc. 145 (1969), 1-49.

[4] D.E. Blair, J. Davidov and O. Muskarov, Isotropic Kähler hyperbolic

[5] A. Bonome, R. Castro, L.M. Hervella and Y. Matsushita, Construction of

Norden structures on neutral 4-manifolds, JP J. Geom. Topol., 5 (2005),
121-140.

[6] A. Borowiec, M. Ferraris, M. Francaviglia and I. Volovich, Almost-

complex and almost-product Einstein manifolds from a variational prin-
ciple, J. Math. Phys. 40 (1999), 3446-3464.

[7] A. Borowiec, M. Francaviglia and I. Volovich, Anti-Kählerian manifolds,

Dif. Geom. Appl. 12 (2000), 281-289.

[8] N. Cengiz and A.A. Salimov, Complete lifts of derivations to tensor bun-
dles, Bol. Soc. Mat. Mexicana. 8 (2002), 75-82.

[9] J. Cheeger and D. Gromoll, On the structure of complete manifolds of

nonnegative curvature, Ann. of Math. 96 (1972), 413-443.

[10] L.A. Cordero, C.T.J. Dodson and M. De Leon, Differential Geometry of

frame bundles, Kluwer Academic Publishers Group, Dordrecht, 1989.
[11] V. Cruceanu, P. Fortuny and P.M. Gadea, A survey on paracomplex ge-
ometry, Rocky Mountain J. Math. 26 (1996), 83-115.
[12] J. Davidov, J.C. Dı́az-Ramos, E. Garcı́a-Rı́o, Y. Matsushita, O. Muškarov
and R. Vázquez-Lorenzo, Almost Kähler Walker 4-manifolds, J. Geom.
Phys., 57 (2007), 1075-1088.
[13] J. Davidov, J.C. Dı́az-Ramos, E. Garcı́a-Rı́o, Y. Matsushita, O. Muškarov
and R. Vázquez-Lorenzo, Hermitian-Walker 4-manifolds, J. Geom.
Phys., 58 (2008), 307-323.
[14] S. Dragomir and M. Francaviglia, On Norden metrics which are locally
conformal to anti-Kählerian metrics, Acta Appl. Math. 60 (2000), 115-
135.
[15] F. Etayo and R. Santamarı́a, (J2 = ±1)−metric manifolds, Publ. Math.
Debrecen 57 (2000), 435-444.
[16] L.E. Evtushik, Ju.G. Lumiste, N.M. Ostianu and A.P. Shirokov,
Differential-geometric structures on manifolds. (Russian) Problems in
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

geometry, 9, Akad. Nauk SSSR, Vsesoyuz. Inst. Nauchn. i Tekhn. In-

formatsii, 1979.
[17] P.M. Gadea, J. Grifone and J. Munoz Masque, Manifolds modelled over
free modules over the double numbers, Acta Math. Hungar. 100 (2003),
187-203.
[18] P.M. Gadea, J. Grifone and J. Munoz Masque, Stein embedding theorem
for B-manifolds, Proc. Edinb. Math. Soc. 46 (2003), 489-499.
[19] G.T. Ganchev and A.V. Borisov, Note on the almost complex manifolds
with Norden metric, Compt. Rend. Acad. Bulg. Sci., 39 (1986), 31-34.
[20] E. Garcı́a-Rı́o and Y. Matsushita, Isotropic Kähler structures on Engel
4-manifolds, J. Geom. Phys., 33 (2000), 288-294.
[21] A. Gezer and A.A Salimov, Diagonal lifts of tensor fields of type (1,1)
on cross-sections in tensor bundles and its applications, J. Korean Math.
Soc. 45 (2008), 367-376.

[22] O. Gil-Medrano, Geometric properties of some classes of Riemannian

almost-product manifolds, Rend. Circ. Mat. Palermo, 32 (1983), 315-
329.

[23] O. Gil-Medrano and A.M. Naveira, Some remarks about the Riemannian
curvature operator of a Riemannian almost-product manifold, Rev. Roum.
Math. Pures Appl., 30 (1985), 647-658.

[24] W.B. Gordon, Convex functions and harmonic maps, Proc. Amer. Math.
Soc., 33 (1972), 433-437.

[25] K. Gribachev, D. Mekerov and G. Djelepov, Generalized B-manifolds,

C. R. Acad. Bulgare Sci. 38 (1985), 299-302.

[26] K. Gribachev, D. Mekerov and G. Djelepov, On the geometry of almost

B-manifolds, C. R. Acad. Bulgare Sci. 38 (1985), 563-566.

[27] S. Gudmundsson and E. Kappos, On the geometry of the Tangent Bundle

with the Cheeger-Gromoll metric, Tokyo J. Math. 25 (2002).

[28] S. Gudmundsson and E. Kappos, On the Geometry of the Tangent Bun-

[29] A. Frolicher and A. Nijenhuis, Some new cohomology invariants for

complex manifolds. I, II. Nederl. Akad. Wetensch. Proc. Ser. A. 59, Indag.
Math. 18 (1956), 540-564.

[30] A. Frolicher and A. Nijenhuis, Theory of vector-valued differential

forms, I. Derivations of the graded ring of differential forms, Nederl.
Akad. Wetensch. Proc. Ser. A. 59, Indag. Math. 18 (1956), 338-359.

[31] M. Iscan and A.A. Salimov, On Kahler-Norden manifolds, Proc. Indian

Acad. Sci.(Math. Sci.) 119 (2009), 71-80.

[32] A. Kasue, On Riemannian manifolds admitting certain strictly convex

functions, Osaka J. Math., 18 (1981), 577-582.

[33] S. Kobayashi and K. Nomizu, Foundations of differential geometry. Vol.

I, Interscience Publishers, New York-London, 1963.

[34] S. Kobayashi and K. Nomizu, Foundations of differential geometry. Vol.

II, Interscience Publishers, New York-London-Sydney, 1969.

[35] G. I. Kruchkovich, Conditions for the integrability of a regular hypercom-

plex structure on a manifold, (Russian) Ukrain. Geometr. Sb. 9 (1970),
67-75.

[36] G.I. Kruchkovich, Hypercomplex structures on manifolds I, (Russian)

Trudy Sem. Vektor. Tenzor. Anal. 16 (1972), 174-201.

[37] G.I. Kruchkovich, Hypercomplex structures on manifolds II, (Russian)

Trudy Sem. Vektor. Tenzor. Anal. 17 (1974), 218-227.

[38] A.J. Ledger and K. Yano, Almost complex structures on tensor bundles,
J. Dif. Geom. 1 (1967), 355-368.

[39] A. Magden and A.A. Salimov, Horizontal lifts of tensor fields to sections
of the tangent bundle, Russian Math. (Iz. VUZ) 45 (2001), 73-76.

[40] A. Magden and A.A. Salimov, On applications of the Yano-Ako operator,

Acta Univ. Palack. Olomuc. Fac. Rerum Natur. Math. 45 (2006), 135-141.

[41] A. Magden and A.A. Salimov, Complete lifts of tensor fields on a pure
cross-section in the tensor bundle, J. Geom. 93 (2009), 128-138.

[42] M. Manev and D. Mekerov, On Lie groups as quasi-Kähler manifolds

with Killing Norden metric, Adv. Geom., 8 (2008), 343-352.

[43] Y. Matsushita, Fields of 2-planes and two kinds of almost complex struc-
tures on compact four-dimensional manifolds, Math. Z., 207 (1991), 281-
291.

[44] Y. Matsushita, Some remarks on fields of 2-planes on compact smooth

4-manifolds, Adv. St. in Pure Math., vol. 22 (1993), 153-167.

[45] Y. Matsushita and P. Law, Hitchin-Thorpe-type inequalities for pseudo-

Riemannian 4-manifolds of metric signature, Geom. Ded., 87 (2001), 65-
89.

[46] Y. Matsushita, Four-dimensional Walker metrics and symplectic struc-

ture, J. Geom. Phys., 52 (2004), 89-99.

[47] Y. Matsushita, Walker 4-manifolds with proper almost complex structure,

J. Geom. Phys., 55 (2005), 385-398.

[48] Y. Matsushita, Counterexamples of compact type to the Goldberg con-

jecture and various version of the conjecture to appear in Proceedings
of The 8th International Workshop on Complex Structures and Vector
Fields, Sofia, Bulgaria, August 20 - 26, (2004), ed. S. Dimiev and K.
Sekigawa, World Scientific 2007.

[49] Y. Matsushita, S. Haze and P.R. Law, Almost Kähler-Einstein structure

on 8-dimensional walker manifolds, Monatsh. Math., 150 (2007), 41-48.

[50] D. Mekerov, On some classes of almost B-manifolds, C. R. Acad. Bulgare

Sci. 38 (1985), 559-561.

[51] D. Mekerov, A connection with skew symmetric torsion and Kähler cur-
vature tensor on quasi-Kähler manifolds with Norden metric, C. R. Acad.
Bulgare Sci. 61 (2008), 1249-1256.

[52] D. Mekerov, Connection with parallel totally skew-symmetric torsion on

almost complex manifolds with Norden metric, C. R. Acad. Bulgare Sci.
62 (2009), 1501-1508.

[53] V. Miquel, Some examples of Riemannian almost-product manifolds, Pa-

[54] K. P. Mok, On differential geometry of frame bundles of Riemannian

manifolds, J. reine angew. Math., 302 (1976), 16-31.

[55] E. Musso and F. Tricerri, Riemannian metric on Tangent Bundles, Ann.

Math. Pura. Appl. 150 (1988), 1-19.

[56] A.M. Naveira, A classification of Riemannian almost-product manifolds,

Rend. Mat., 3 (1983), 577-592.

[57] A. Nijenhuis, X -forming sets of eigenvectors. Nederl. Akad. Wetensch.

Proc. Ser. A. 54, Indag. Math. 13 (1951), 200-212.

[58] A.P. Norden, On a class of four-dimensional A-spaces. (Russian) Izv.

Vyssh. Uchebn. Zaved. Matematika. 17 (1960), 145-157.

[59] B. O’Neill, Semi-Riemannian geometry, Academic Press, New York,

1983.

[60] V. Oproiu and N. Papaghiuc, Some classes of almost anti-Hermitian

structures on the tangent bundle, Mediterr. J. Math. 1 (2004), 269-282.

[61] V. Oproiu and N. Papaghiuc, An anti-Kählerian Einstein structure on the

tangent bundle of a space form, Colloq. Math. 103 (2005), 41-46.

[62] W.A. Poor, Differential Geometric Structures, McGraw-Hill Book Co.,

New York, 1981.

[63] I. Popovici, Contributions à l’étude des espaces à connexion constante,

(French) Rev. Roumaine Math. Pures Appl. 23 (1978), 1211-1225.

[64] A.A. Salimov, Almost analyticity of a Riemannian metric and integrabil-

ity of a structure, (Russian) Trudy Geom. Sem. Kazan. Univ., 15 (1983),
72-78.

[65] A.A. Salimov, Quasiholomorphic mapping and a tensor bundle, Soviet

Math. (Iz. VUZ) 33 (1989), 89-92.

[66] A.A. Salimov, Almost -holomorphic tensors and their properties, Russian
Acad. Sci. Dokl. Math. 45 (1992), 602-605.

tensor bundle, translation in Russian Math. (Iz. VUZ) 38 (1994), 67-73.

[68] A.A. Salimov, Generalized Yano-Ako operator and the complete lift of
tensor fields, Tensor (N.S.) 55 (1994), 142-146.

[69] A.A. Salimov, Lifts of poly-affinor structures on pure sections of a tensor

bundle, Russian Math. (Iz. VUZ) 40 (1996), 52-59.

[70] A.A. Salimov, Nonexistence of Para-Kahler-Norden warped metrics, Int.

J. Geom. Methods Mod. Phys. 6 (2009), 1097-1102.

[71] A.A. Salimov, On operators associated with tensor fields, J. Geom. 99

(2010), 107-145.

[72] A.A. Salimov, A note on the Goldberg conjecture of Walker manifolds,

Int. J. Geom. Methods Mod. Phys. 8 (2011), 925-928.

[73] A.A. Salimov and F. Agca, On para-Nordenian structures, Ann. Polon.

Math. 99 (2010), 193-200.

[74] A.A. Salimov and F. Agca, Some properties of Sasakian metrics in cotan-
gent bundles, Mediterr. J. Math. 8 (2011), 243-255

[75] A.A. Salimov and K. Akbulut, A note on a paraholomorphic Cheeger-

Gromoll metric, Proc. Indian Acad. Sci.(Math. Sci.) 119 (2009), 187-
195.

[76] A.A. Salimov, K. Akbulut and S. Aslanci, A note on integrability of al-

most product riemannian structures, Arab. J. Sci. Eng. Sect. A Sci. 34
(2009), 153-157.

[77] A.A. Salimov and S. Aslanci, Applications of φ-operators to the hyper-

complex geometry, Adv. Appl. Clifford Algebr. 22(2011), 185-201.

[78] A.A. Salimov and A. Gezer, On the geometry of the (1,1)-tensor bundle
with Sasaki type metric, Chin. Ann. Math. Ser. B 32 (2011), 369-386.

[79] A.A. Salimov, A. Gezer and K. Akbulut, Geodesics of Sasakian metrics

on tensor bundles, Mediterr. J. Math. 6 (2009), 135-147.

[80] A.A. Salimov and M. Iscan, Some properties of Norden-Walker metrics,

[81] A.A. Salimov and M. Iscan, On Norden-Walker 4-manifolds, Note Mat.

30 (2010), 111-128.

[82] A.A. Salimov and M. Iscan, On the geometry of B-manifolds, Confer-

ence on Differential Geometry dedicated to 80-th anniversary of Prof.
V.V.Vishnevskii, Kazan. Gos. Univ. Ucen. Zap. 151 (2009), 231-239.

[83] A.A. Salimov, M. Iscan and K. Akbulut, Some remarks concerning hy-
perholomorphic B-manifolds, Chin. Ann. Math. Ser. B 29 (2008), 631-
340.

[84] A.A. Salimov, M. Iscan and K. Akbulut, Notes on para-Norden-Walker

4-manifolds, Int. J. Geom. Methods Mod. Phys. 7 (2010), 1331-1347.

[85] A.A. Salimov, M. Iscan and F. Etayo, Paraholomorphic B-manifold and

its properties, Topology Appl. 154 (2007), 925-933.

[86] A.A. Salimov and A. Magden, Complete lifts of tensor fields on a pure
cross-section in the tensor bundle, Note Mat. 18 (1998), 27-37.

[87] S. Sasaki, On the Differantial geometry of tangent bundles of Riemannian

manifols, Tohoku Math. J. 10 (1958), 338-358.

[88] I. Sato, Almost analytic tensor fields in almost complex manifolds, Ten-
sor (N.S.) 17 (1966), 105-119.

[89] G. Scheffers, Generalization of the foundations of ordinary complex

functions. I, II, Verallgemeinerung der Grundlagen der gewöhnlich com-
plexen Functionen. I. II.) (German) Leipz. Ber. XLV (1893), 828-848.

[90] K. Sekigawa, On some 4-dimensional compact Einstein almost Kähler

manifolds, Math. Ann. 271 (1985), 333-337.

[91] M. Sekizawa, Curvatures of Tangent Bundles with Cheeger-Gromoll

Metric, Tokyo J. Math., 14 (1991), 407-417.

[92] A.P. Shirokov, On a property of covariantly constant affinors. (Russian)

Dokl. Akad. Nauk SSSR (N.S.) 102 (1955), 461-464.

[93] A.P. Shirokov, On the question of pure tensors and invariant subspaces in
manifolds with almost algebraic structure. (Russian) Kazan. Gos. Univ.
Učen. Zap. 126 (1966), 81-89.
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

[94] A.P. Shirokov, Spaces over algebras and their applications, Geometry, 7.
J. Math. Sci. (New York) 108 (2002), 232-248.

[95] W. Slebodzinski, Contribution à la géométrie différentielle d’un tenseur

mixte de valence deux, Colloq. Math. 13 (1964), 49-54.

[96] S.E. Stepanov, Riemannian almost product manifolds and submersions,

Geometry, 5. J. Math. Sci. (New York), 99 (2000), 1788-1810.

[97] S. Tachibana, Analytic tensor and its generalization, Tohoku Math. J. 12

(1960), 208-221.

[98] S. Tachibana and S. Koto, On almost-analytic functions, tensors and in-

variant subspaces, Tohoku Math. J. 14 (1962), 177-186.

[99] N.V. Talantova and A.P. Shirokov, A remark on a certain metric in the tan-
gent bundle, (Russian) Izv. Vysš. Učebn. Zaved. Matematika. 157 (1975),
143-146.

[100] G. Thompson and U. Schwardmann, Almost tangent and cotangent struc-

tures in the large, Trans. Amer. Math. Soc. 327 (1991), 313-328.

[101] V.V. Vishnevskii, On the complex structure of B-spaces, (Russian)

Kazan. Gos. Univ. Uchen. Zap. 123 (1963), 24-48.

[102] V.V. Vishnevskii, A certain class of spaces over plural algebras, (Russian)
Izv. Vysš. Učebn. Zaved. Matematika. 81 (1969), 14-22.

[103] V.V. Vishnevskii, Affinor structures of spaces with affine connection,

(Russian) Izv. Vysš. Učebn. Zaved. Matematika. 92 (1970), 12-23.

[104] V.V. Vishnevskii, Certain properties of differential geometric structures

that are determined by algebras, (Russian) Visš Ped. Inst. Plovdiv Naučn.
Trud. 10 (1972), 23-30.

[105] V.V. Vishnevskii, Affinor structures of manifolds as structures defin-

able by algebras, (Survey article). (Russian) Commemoration volumes
for Prof. Dr. Akitsugu Kawaguchi’s seventieth birthday, Vol. III. Tensor
(N.S.) 26 (1972), 363-372.

[106] V.V. Vishnevskii, Structures of manifolds that are defined by non-regular

representations of algebras, (Russian) Differential geometry, 1 (1974),

3-14.

[107] V.V. Vishnevskii, A.P. Shirokov and V.V. Shurygin, (Russian) Spaces
over algebras, Kazanskii Gosudarstvennii Universitet, Kazan, 1985.

[108] V.V. Vishnevskii, Integrable affinor structures and their plural interpreta-
tions, Geometry, 7. J. Math. Sci. (New York), 108 (2002), 151-187.

[109] G. Vranceanu, Spazi a connessione affine e le algebre di numeri ipercom-

plessi, (Italian) Ann. Scuola Norm. Sup. Pisa. 12, (1958), 5-20.

[110] A.G. Walker, Canonical form for a Riemannian space with a paralel field
of null planes, Quart. J. Math. Oxford, 1 (1950), 69-79.

[111] T.J. Willmore, Note on the Slebodzinski tensor of an almost-complex

structure, J. London Math. Soc. 43 (1968), 321-322.

[112] K. Yano, Differential geometry on complex and almost complex space,

International Series of Monographs in Pure and Applied Mathematics,
Vol. 49 A Pergamon Press Book. The Macmillan Co., New York, 1965.

[113] K. Yano and M. Ako, On certain operators associated with tensor fields,
Kodai Math. Sem. Rep. 20 (1968), 414-436.

[114] K. Yano and S. Ishihara, Tangent and cotangent bundles Differential ge-
ometry, Pure and Applied Mathematics, No. 16. Marcel Dekker, Inc.,
New York, 1973.

[115] K. Yano and M. Kon, Structures on manifolds, Series in Pure Mathemat-

Salimov, Arif. Tensor Operators and their Applications, Nova Science Publishers, Incorporated, 2012. ProQuest Ebook
Central, http://ebookcentral.proquest.com/lib/unilu-ebooks/detail.action?docID=3022757.
Created from unilu-ebooks on 2023-10-16 21:16:26.
Index
B-metrics, 70 Exact 1-form, 9
B-tensor, 70 Exterior di¤erentiation, 29
C-holomorphic, 51
'-connections, 18 Frobenius algebra, 42
'-structure, 46
' -operator, 4 Generalized Yano-Ako operator, 36
' ;' -operator, 22 Geodesic, 145
S -operator, 29 Goldberg conjecture, 102
' -operator, 15
S -operator, 32 Harmonic, 95
-structure, 46 Hessian, 89
-operator, 120 Hessian metric, 89
Adjoint operator, 2 Hybrid tensor, 70
Adapted frames, 46, 128 Hyperbolic, 108
Adapted (B,C)-frame, 126 Hypercomplex algebra, 40
Admissible, 48 Hypercomplex connection, 59
A…nor …elds, 2 Hypercomlex function, 43
Algebra of pracomplex numbers, 46 Hypercomplex structure, 47
Algebraic -structures on manifolds, Hypercomplex tensor, 57
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

46 Hyperholomorphic submanifold, 136

Almost complex structure, 51 Hyper-Kähler-Norden manifolds, 73
Associative algebras, 40 Holomorphic manifold, 51
Almost product structure, 79 Holomorphic mapping, 134
Almost Norden-Walker metrics 93 Holomorphic Norden metric, 71
Anti-Hermitian, 76 Holomorphic Norden manifold, 71
Associated 1-form, 9 Holomorphic Norden-Walker metrics,
96
Cauchy-Riemann conditions, 44 Holomorphic tensor …eld, 57
Cheeger-Gromoll metric, 160 Horizontal lift, 122
Closed 1-form, 9
Complete lift, 128 In…nitesimal automorphism, 58
Complex analitic manifold, 52 In…nitesimal isometry, 125
Commutative Algebras, 42 Interior product, 4
Conjugation, 42 Integrable -structure, 46
Cotangent bundle, 110 Integrable regular - structure, 50
Cross-sections, 126 Invariant, 135
Invariant submanifold, 22
Decomposable, 79 Isotropic Kähler, 99
Derivation, 122
Diagonal lift, 138 Jacobi operator, 3
Dual algebra, 45 Jacobi tensor …elds, 152
Dual-Kähler-Norden, 86 Jordan tensor, 110

Kähler-Norden manifolds, 76 Pure torsion, 61

Kähler-Norden-Walker, 96
Killing vector …eld, 125 Quasi-Kähler, 101
Kruchkovich tensor, 55
Regular -structure, 47
Locally ‡at, 67 Regular representation, 41
Locally product, 80 Ricci operator, 3
Riemann extension, 110
Module, 50 Rigid '-structure, 46
Nijenhuis tensor, 6
Sasakian Metric, 142
Nijenhuis- Shirokov tensor, 6
Sche¤ers conditions, 43
Non-holonomic object, 144
Self-adjoint operators, 3
Nonregular, 53
Slebodzinski tensor, 8
Norden metric, 70
Submanifold, 21
Norden manifold, 70
Symplectic manifold, 82
Norden-Hessian, 89
Synectic function, 46
Norden-Walker metric, 93
Norden-Walker structure, 94
Norden-Walker 8-manifolds, 113 Tachibana operators, 4
Tangent bundle, 52
Opposite, 105 Tensor bundle, 118
Torsion tensors, 61
Para-Cauchy-Riemann conditions, 46 Totally geodesic, 109
Para-Kähler-Norden, 80 Twin-Norden metric, 76
Para-Norden-Walker manifold, 108
Vertical lift, 120
Copyright © 2012. Nova Science Publishers, Incorporated. All rights reserved.

Paraholomorphic Cheeger-Gromoll
metric, 160 Vishnevskii operator, 15
Paraholomorphic, 81 Vranceanu space, 67
Proper almost complex structure, 94
Pseudo-Riemannian metric, 3 Walker metric, 93
Pure connections, 19 Warped product, 83
Pure curvature tensors, 64 Warping function, 83
Pure product, 4 Warped metric, 83
Pure tensor …elds of mixed kind, 21
Pure tensor …elds, 2 Yano-Ako Operators, 29