Fetcu HT Final

UNIVERSITATEA TEHNICĂ ”GHEORGHE ASACHI” DIN IAŞI
DEPARTAMENTUL DE MATEMATICĂ ŞI INFORMATICĂ
Submanifolds with Parallel Mean

Curvature and Biharmonic Submanifolds in
Riemannian Manifolds
Habilitation Thesis
Dorel Fetcu
To Ilinca, Iuria, and Petronela
Abstract
A classical but still very dynamic topic in the field of Differential Geometry is
the study of surfaces with constant mean curvature (cmc surfaces) in 3-dimensional
spaces and, more generally, of submanifolds with parallel mean curvature vector
field (pmc submanifolds) in Riemannian manifolds with arbitrary dimension. Their
history is spread on more than six decades, in the case of cmc surfaces, and goes
back to the early 1970s, in the case of pmc submanifolds, to papers like [35] by B.-Y.
Chen and G. D. Ludden, [52, 53] by J. Erbacher, [58] by D. Ferus, [83] by D. A.
Hoffman, or [136] by S.-T. Yau.
In the following, we shall briefly recall only some of those results that represent
a source of inspiration for our work. Two very powerful tools were mainly used to
prove these results: holomorphic differentials defined on cmc (or pmc) surfaces and
Simons type equations.
H. Hopf [85] was the first to use a holomorphic differential to show that any cmc
surface homeomorphic to a sphere in Euclidean 3-space is actually a round sphere.
His result was extended to cmc surfaces in 3-dimensional space forms by S.-S. Chern
[41], and then to cmc surfaces in product spaces of type M 2 (c) × R, where M 2 (c)
is a complete simply-connected surface with constant curvature c, as well as Nil(3)
and PSL(2,
g R), by U. Abresch and H. Rosenberg [1, 2].
The next natural step was to study pmc surfaces in product spaces of type
M n (c) × R, where M n (c) is a space form with constant sectional curvature c, i.e.,
those surfaces satisfying ∇⊥ H = 0, where ∇⊥ is the connection in the normal bundle
and H is the mean curvature vector field. Two very important papers on this topic
are [5, 6] by H. Alencar, M. do Carmo, and R. Tribuzy. In these articles they intro-
duce a holomorphic differential (that generalizes the Abresch-Rosenberg differential
defined in [1] for cmc surfaces in M 2 (c) × R) and use it to study the geometry of
pmc surfaces. One of the main results in [6] is a reduction of codimension theorem,
showing that a pmc surface immersed in M n (c) × R either is a minimal surface in a
totally umbilical hypersurface of M n (c); or a cmc surface in a 3-dimensional totally
umbilical or totally geodesic submanifold of M n (c); or it lies in M 4 (c) × R.
Another very effective method developed in order to study minimal or, more
generally, cmc and pmc submanifolds in Riemannian manifolds, is to use Simons
type equations.
In 1968, J. Simons [128] discovered a fundamental formula for the Laplacian of
the second fundamental form of a minimal submanifold in a Riemannian manifold
and used it to characterize certain minimal submanifolds of a sphere and Euclidean
space. One year later, K. Nomizu and B. Smyth [112] generalized Simons’ equation
in the case of cmc hypersurfaces in a space form and their result was then extended,
in B. Smyth’s work [129], to the more general case of pmc submanifolds in a space
form. Over the years such formulas, nowadays called Simons type equations, were
i
ii
used more and more often in studies on cmc and pmc submanifolds (see, for example,
[4, 7, 8, 16, 18, 28, 39, 123]).
During the last three decades one could observe an ever growing interest in
the study of certain fourth order partial differential equations, which generalize the
notion of harmonic maps.
In their seminal paper [49], J. Eells and J. H. Sampson suggested the notion
of biharmonic maps ψ : M → N between two Riemannian manifolds, defined as
critical points of the bienergy functional
Z
E2 (ψ) = |τ (ψ)|2 dv,
M
where τ (ψ) = trace ∇ψ dψ is the tension field of ψ, ∇ψ being the connection in the
pull-back bundle ψ −1 T M . If M is not a compact manifold, then biharmonic maps
ψ : M → N are defined as solutions of the Euler-Lagrange equation τ2 (ψ) = 0,
where τ2 (ψ) = ∆τ (ψ) − trace R̄(dψ, τ (ψ))dψ is the bitension field of ψ. It is easy
to see that any harmonic map is biharmonic and that is why we are interested in
proper-biharmonic maps, i.e., those biharmonic maps which are not harmonic.
A special case is that of biharmonic Riemannian immersions, or biharmonic
submanifolds, i.e., those submanifolds for which the inclusion map is biharmonic.
This definition of biharmonic submanifolds coincides, when working in Euclidean
space (and only then), with that proposed by B.-Y. Chen [31], where a biharmonic
submanifold is characterized by the fact that its mean curvature vector field is har-
monic.
Although only non-existence results for proper-biharmonic submanifolds in Eu-
clidean space were obtained (see, for example, [34, 48, 91, 92, 106]), when the
ambient space is not flat, numerous examples and classification results for proper-
biharmonic submanifolds were found in papers like [12]-[14], [21]-[26], [44, 88, 100,
101, 104, 106], [117]-[119], [124]-[126], and [139].
Our thesis is organized in two parts, the first one, that contains two chapters,
being devoted to the study of pmc submanifolds, while in the second part, consisting
of three chapters, we consider biharmonic submanifolds.
In the first chapter, we present results from [62], [77], and [78] on pmc surfaces
in complex, cosymplectic, and Sasakian space forms, respectively. In all these situa-
tions, we prove reduction of codimension theorems and also introduce holomorphic
differentials that are then used to study the geometry of some of these surfaces.
The second chapter is devoted to the study of pmc submanifolds, this time of
arbitrary dimension, in M n (c) × R. We first prove two Simons type equations that
are then used to obtain gap theorems for such submanifolds. We also consider pmc
surfaces with finite total curvature and find a result on their compactness. The
chapter ends with a classification result for helix pmc surfaces. These results were
obtained in [17], [63], [72], [74], [75], and [76].
We begin dealing with biharmonic submanifolds in the third chapter, that is
based on [71] and [72]. Here, we first present a gap theorem for pmc proper-
biharmonic submanifolds in M n (c) × R and also classify pmc proper-biharmonic
surfaces in this space. We then turn our attention to biconservative surfaces in
M n (c) × R, i.e., those surfaces for which the tangent part of the bitension field
vanishes (we note that biconservative submanifolds have only very recently begun
to be studied in articles like [27, 80, 107, 108]). We completely determine such
iii
surfaces that have parallel mean curvature vector field and obtain explicit examples
of cmc biconservative surfaces, when n = 3. Also pmc biconservative surfaces with
finite total curvature in Hadamard manifolds are considered and a compactness
result is obtained, in the last part of the chapter.
In the fourth chapter, we study biharmonic submanifolds in Sasakian space forms
and obtain classification results, as well as explicit examples, for proper-biharmonic
curves, proper-biharmonic Hopf cylinders over homogeneous real hypersurfaces in
CP n , and 3-dimensional proper-biharmonic integral C-parallel submanifolds in a 7-
dimensional Sasakian space form. We also present a method to construct biharmonic
anti-invariant submanifolds from biharmonic integral submanifolds. The results in
this chapter first appeared in [59], [60], [61], and [65]-[70].
Biharmonic submanifolds in complex space forms are studied in the last chapter.
First, are presented some general results on the biharmonicity of certain classes
of submanifolds. We continue with a formula that relates the bitension fields of
a submanifold in CP n and its corresponding Hopf cylinder in S2n+1 . Next, we
prove a result on the biharmonicity of Clifford type submanifolds in CP n , while in
the last part of the chapter we classify proper-biharmonic curves and pmc proper-
biharmonic surfaces in CP n , and also 3-dimensional proper-biharmonic Lagrangian
parallel submanifolds in CP 3 . This chapter contains results from [64], [70], and
[73].
Whilst throughout the thesis we tried to offer the reader an image as complete
as possible of our work, due to the need of keeping the presentation at a reasonable
length, we were forced to skip many of the proofs and sometimes even not to mention
some of our results that otherwise we consider interesting.
Rezumat
Un subiect, ı̂n acelaşi timp clasic şi actual, ı̂n geometria diferenţială de astăzi ı̂l
reprezintă studiul suprafeţelor de curbură medie constantă (suprafeţe cmc) ı̂n spaţii
de dimensiune 3, şi deasemeni al cazului mai general al subvarietăţilor având câmpul
vectorial curbură medie paralel ı̂n fibratul normal (subvarietăţi pmc) ı̂n spaţii de
dimensiune oarecare. Primele studii de acest fel, consacrate suprafeţelor cmc, au
apărut acum mai bine de 60 de ani, ı̂n timp ce subvarietăţile pmc au ı̂nceput să
câştige interesul lumii matematice la ı̂nceputul anilor 1970 odată cu apariţia unor
articole ca [35] de B.-Y. Chen şi G. D. Ludden, [52, 53] ale lui J. Erbacher, [58] de
D. Ferus, [83] de D. A. Hoffman, sau [136] de S.-T. Yau.
În continuare, vom trece rapid ı̂n revistă doar unele din acele rezultate care
au reprezentat (şi ı̂ncă o fac) o sursă de inspiraţie ı̂n activitatea noastră de cerc-
etare. Două instrumente au fost folosite cu precădere ı̂n obţinerea acestor rezultate:
diferenţialele olomorfe definite pe suprafaţe cmc (sau pmc) precum şi formulele de
tip Simons.
H. Hopf [85] a folosit pentru prima oară o diferenţială olomorfă pentru a arăta că
orice suprafaţă cmc homeomorfă cu o sferă ı̂n spaţiul euclidian 3-dimensional este de
fapt o sferă euclidiană. Acest rezultat a fost extins de către S.-S. Chern [41] la cazul
suprafeţelor cmc ı̂n forme spaţiale reale, pentru ca apoi să fie demonstrat pentru
suprafeţe cmc ı̂n spaţii produs de tipul M 2 (c) × R, unde M 2 (c) este o suprafaţă
completă şi simplu conexă de curbură constantă c, ca şi pentru cele ı̂n Nil(3) şi
PSL(2,
g R), de către U. Abresch şi H. Rosenberg [1, 2].
În mod natural, următorul pas a fost trecerea la studiul suprafeţelor pmc ı̂n
M n (c) × R, unde M n (c) este o formă spaţială reală de curbură secţională con-
stantă c, adică, al acelor suprafeţe care satisfac ∇⊥ H = 0, unde ∇⊥ este conexiunea
din fibratul normal, iar H câmpul vectorial curbură medie. Două din articolele
importante dedicate acestui subiect sunt [5, 6] de H. Alencar, M. do Carmo, şi
R. Tribuzy. În aceste articole autorii introduc o diferenţială olomorfă (care gen-
eralizează diferenţiala Abresch-Rosenberg definită ı̂n [1] pentru suprafeţe cmc ı̂n
M 2 (c)×R) şi o folosesc apoi ı̂n studiul geometriei unora dintre aceste suprafeţe. Unul
dintre rezultatele principale din [6] este o teoremă de reducere a codimensiunii care
arată că o suprafaţă pmc ı̂n M n (c) × R sau este minimală ı̂ntr-o hipersuprafaţă total
umbilicală ı̂n M n (c); sau este o suprafaţă cmc ı̂ntr-o subvarietate 3-dimensională
total umbilicală sau total geodezică ı̂n M n (c); sau stă ı̂n M 4 (c) × R.
O altă metodă implicată cu real succes ı̂n studiul subvarietăţilor minimale şi, ı̂n
general, al subvarietăţilor cmc şi pmc, este folosirea ecuaţiilor de tip Simons.
În 1968, J. Simons [128] a obţinut expresia laplacianului normei formei a doua
fundamentale a unei subvarietăţi minimale ı̂ntr-o varietate riemanniană, pe care apoi
a folosit-o ı̂n caracterizarea unora dintre aceste subvarietăţi ı̂n sfera euclidiană şi ı̂n
spaţiul euclidian. Un an mai târziu, K. Nomizu şi B. Smyth [112] au generalizat
v
vi
această formulă ı̂n cazul hipersuprafeţelor cmc ı̂ntr-o formă spaţială reală, acest
rezultat fiind generalizat la rândul lui, de către B. Smyth’s [129], pentru subvarietăţi
pmc ı̂ntr-o formă spaţială. De-a lungul anilor astfel de formule, numite astă zi
ecuaţii de tip Simons, au fost folosite din ce ı̂n ce mai des ı̂n articolele dedicate
subvarietăţilor cmc şi pmc (vezi, de exemplu, [4, 7, 8, 16, 18, 28, 39, 123]).
În ultimii 30 de ani s-a putut observa un interes din ce ı̂n ce mai ridicat ı̂n
studierea unor ecuaţii cu derivate parţiale de ordin 4 care generalizează noţiunea de
aplicaţii armonice.
Astfel, ı̂n influentul lor articol [49], J. Eells şi J. H. Sampson au sugerat noţiunea
de aplicaţie biarmonică ψ : M → N ı̂ntre două varietăţi riemanniene, definită ca
fiind un punct critic al funcţionalei bienergiei
Z
E2 (ψ) = |τ (ψ)|2 dv,
M
unde τ (ψ) = trace ∇ψ dψ este câmpul tensiune al aplicaţiei ψ, ∇ψ fiind conexiunea

−1
ı̂n fibratul ψ T M . Dacă varietatea M nu este compactă, atunci aplicaţia ψ : M →
N este, prin definiţie, biarmonică dacă este o soluţie a ecuaţiei Euler-Lagrange
τ2 (ψ) = 0, unde τ2 (ψ) = ∆τ (ψ) − trace R̄(dψ, τ (ψ))dψ este câmpul bitensiune al
lui ψ. Este uşor de văzut că orice aplicaţie armonică este biarmonică, acesta fiind
motivul pentru care suntem interesaţi ı̂n studierea aplicaţiilor biarmonice care nu
sunt armonice, numite aplicaţii propriu-biarmonice.
Un caz special este cel al imersiilor biarmonice, sau, altfel spus, al subvarietăţilor
biarmonice, adică, al acelor subvarietăţi pentru care aplicaţia de incluziune este
biarmonică. În spaţiul euclidian (şi doar ı̂n acest spaţiu), această definiţie a sub-
varietăţilor biarmonice coincide cu cea propusă de către B.-Y. Chen [31], care ca-
racterizează aceste subvarietăţi prin faptul că au câmpul vectorial curbură medie
armonic.
Deşi ı̂n spaţiul euclidian au fost obţinute doar rezultate care sugerează că nu
există exemple de subvarietăţi propriu-biarmonice (vezi, de exemplu, [34, 48, 91,
92, 106]), ı̂n spaţii de curbură secţională diferită de zero, au fost găsite numeroase
exemple şi demonstrate multe rezultate de clasificare a unor astfel de subvarietăţi,
ı̂n articole cum ar fi [12]-[14], [21]-[26], [44, 88, 100, 101, 104, 106], [117]-[119],
[124]-[126] şi [139].
Această teză este structurată ı̂n două părţi, prima, constând din două capitole,
fiind dedicată subvarietăţilor pmc, ı̂n timp ce a doua, formată din trei capitole, are
drept scop prezentarea rezultatelor obţinute ı̂n studiul subvarietăţilor biarmonice.
În primul capitol, prezentăm rezultate din [62], [77] şi [78] privind suprafeţele
pmc ı̂n forme spaţiale complexe, cosimplectice şi respectiv sasakiene. În toate aceste
situaţii demonstrăm teoreme de reducere a codimensiunii şi introducem diferenţiale
olomorfe pe care le folosim ı̂n studierea geometriei unora dintre aceste suprafeţe.
Al doilea capitol este dedicat studiului subvarietăţilor pmc de dimensiune ar-
bitrară ı̂n M n (c) × R. Mai ı̂ntâi demonstrăm două ecuaţii de tip Simons pe care
apoi le folosim pentru a găsi rezultate de tip gap pentru astfel de subvarietăţi. Ne
ı̂ndreptăm atenţia şi spre suprafeţele pmc având curbura totală finită şi demonstrăm
o teoremă despre la compactitatea acestora. Capitolul se ı̂ncheie cu un rezultat de
clasificare a suprafeţelor pmc care fac un unghi constant cu ξ, câmpul vectorial uni-
tar tangent la dreapta reală. Aceste rezultate au fost obţinute ı̂n [17], [63], [72],
[74], [75] şi [76].
vii
Începem prezentarea subvarietăţilor biarmonice ı̂n capitolul al treilea, construit

pe articolele [71] şi [72]. Prezentăm mai ı̂ntâi o teoremă de tip gap pentru sub-
varietăţi pmc propriu-biarmonice ı̂n M n (c) × R şi deasemeni clasificăm suprafeţele
pmc propriu-biarmonice ı̂n acest spaţiu. Un alt subiect al acestui capitol ı̂l reprezintă
suprafeţele biconservative ı̂n M n (c) × R, adică, acele suprafeţe pentru care partea
tangentă a câmpului bitensiune se anulează (menţionăm că subvarietăţile biconserva-
tive au ı̂nceput să fie studiate foarte recent ı̂n articole cum ar fi [27, 80, 107, 108]).
Aici, determinăm ecuaţia explicită a suprafeţelor pmc biconservative şi găsim exem-
ple explicite de suprafeţe cmc biconservative ı̂n cazul când n = 3. Încheiem capitolul
cu un rezultat privind compactitatea suprafeţelor pmc biconservative de curbură to-
tală finită ı̂n varietăţi Hadamard.
În al patrulea capitol, studiem subvarietăţile biarmonice ı̂n forme spaţiale sasa-
kiene şi obţinem atât rezultate de clasificare, cât şi exemple explicite, pentru curbe
propriu-biarmonice, cilindri Hopf propriu-biarmonici peste hipersuprafeţe omogene
reale ı̂n CP n şi pentru subvarietăţi propriu-biarmonice integrale C-paralele de di-
mensiune 3 ı̂ntr-o formă spaţială sasakiană 7-dimensională. Deasemeni, prezentăm
o metodă prin care pot fi obţinute subvarietăţi biarmonice anti-invariante pornind de
la subvarietăţi biarmonice integrale. Rezultatele cuprinse ı̂n acest capitol au apărut
ı̂n [59], [60], [61] şi [65]-[70].
Subvarietăţile biarmonice ı̂n forme spaţiale complexe sunt studiate ı̂n ultimul
capitol al tezei. Mai ı̂ntâi prezentăm unele rezultate generale cu privire la biar-
monicitatea unor clase speciale de subvarietăţi ı̂n aceste spaţii. Continuăm cu o
formulă de legătură ı̂ntre câmpurile bitensiune ale unei subvarietăţi ı̂n CP n şi cilin-
drului Hopf corespunzător ı̂n S2n+1 . În continuare, caracterizăm subvarietăţile de
tip Clifford propriu-biarmonice ı̂n CP n , iar ı̂n ultima parte a capitolului clasificăm
curbele propriu-biarmonice şi suprafeţele pmc propriu-biarmonice ı̂n CP n şi dease-
meni subvarietăţile propriu-biarmonice lagrangiene paralele 3-dimensionale ı̂n CP 3 .
Acest capitol conţine rezultate obţinute ı̂n [64], [70] şi [73].
Contents
Abstract i
Rezumat v
Part 1. Submanifolds with Parallel Mean Curvature 1
Chapter 1. Reduction of Codimension Results and Holomorphic Differentials

for Surfaces with Parallel Mean Curvature 3
1. Introduction 3
2. Surfaces with parallel mean curvature in complex space forms 3
3. Surfaces with parallel mean curvature in CP n × R and CH n × R 12
4. Surfaces with parallel mean curvature in Sasakian space forms 24
Chapter 2. Simons Type Formulas and Applications. Surfaces with Parallel

Mean Curvature and Finite Total Curvature 39
1. Introduction 39
2. Preliminaries 39
3. Simons type formulas 41
4. Gap theorems for submanifolds with parallel mean curvature in M n (c)×R 50
5. Holomorphic differentials and Simons type formulas. Surfaces with
constant mean curvature and finite total curvature 59
6. Helix surfaces with parallel mean curvature in M n (c) × R 66
Part 2. Biharmonic Submanifolds 71
Chapter 3. Biharmonic and Biconservative Submanifolds in M n (c) × R 73

1. Introduction 73
2. Preliminaries 73
3. A gap theorem for biharmonic submanifolds with parallel mean curvature
in Sn × R 75
4. Biharmonic pmc surfaces in Sn (c) × R 79
5. Biconservative surfaces with parallel mean curvature in M n (c) × R 81
6. Biconservative surfaces with constant mean curvature in M 3 (c) × R 86
7. Biconservative surfaces with constant mean curvature in Hadamard
manifolds 90
Chapter 4. Biharmonic Submanifolds in Sasakian Space Forms 95

1. Introduction 95
2. Biharmonic Legendre curves in Sasakian space forms 95
3. Biharmonic non-Legendre curves in Sasakian space forms 102
4. A non-existence result for biharmonic curves in S7 109
ix
x
5. Biharmonic anti-invariant submanifolds in Sasakian space forms 111

6. Biharmonic hypersurfaces in Sasakian space forms 116
7. Biharmonic integral C-parallel submanifolds in 7-dimensional Sasakian
space forms 120
Chapter 5. Biharmonic Submanifolds in Complex Space Forms 129
1. Introduction 129
2. General biharmonicity results for submanifolds in complex space forms 129
3. The Hopf fibration and the biharmonic equation 132
4. Biharmonic submanifolds of Clifford type in CP n 135
5. Biharmonic curves in CP n 138
6. Biharmonic surfaces with parallel mean curvature in complex space forms 144
7. Biharmonic parallel Lagrangian submanifolds of CP 3 153
Further Developments 157
Index 159
Bibliography 161
Part 1
Submanifolds with Parallel Mean

Curvature
CHAPTER 1
Reduction of Codimension Results and Holomorphic

Differentials for Surfaces with Parallel Mean Curvature
1. Introduction
We consider surfaces with parallel mean curvature vector field (pmc surfaces)
in N , where N is either a complex space form, or a cosymplectic space form, or a
Sasakian space form. These three cases are treated in three separate sections.
In each of these sections, we prove reduction of codimension results (Theorems
1.15, 1.35, and 1.56) and study the geometry of some of these surfaces. The main
tool employed in these studies are some holomorphic differentials introduced by
Theorems 1.4, 1.19, and 1.57. Using these holomorphic differentials, we classify pmc
2-spheres in 2-dimensional complex space forms (Theorem 1.5) and anti-invariant
pmc 2-spheres in cosymplectic space forms (Theorems 1.40, 1.42, and 1.45), prove
non-existence results for pmc 2-spheres with constant Kähler angle in complex space
forms (Theorem 1.17) and anti-invariant pmc 2-spheres in Sasakian space forms
(Theorem 1.68), and determine all integral complete pmc surfaces that are pseudo-
umbilical and have non-negative Gaussian curvature in 7-dimensional Sasakian space
forms (Theorem 1.63).
2. Surfaces with parallel mean curvature in complex space forms

2.1. Preliminaries. Let Σm be an isometrically immersed submanifold in a
Riemannian manifold N . The second fundamental form σ of Σm is then defined by
the equation of Gauss
(1.1) ∇N
X Y = ∇X Y + σ(X, Y ),
while the shape operator A and the normal connection ∇⊥ are given by the equation
of Weingarten
⊥
(1.2) ∇N
X V = −AV X + ∇X V,
for any vector fields X and Y tangent to Σm and any normal vector field V , where
∇N and ∇ are the Levi-Civita connections on N and Σm , respectively. The mean
curvature vector field H of Σm is given by H = (1/m) trace σ.
We also have the Gauss equation of Σm in N
(1.3) hR(X, Y )Z, W i =hRN (X, Y )Z, W i + hσ(Y, Z), σ(X, W )i
− hσ(X, Z), σ(Y, W )i,
the Codazzi equation
(1.4) (RN (X, Y )Z)⊥ = (∇⊥ ⊥
X σ)(Y, Z) − (∇Y σ)(X, Z),
3
4 Reduction of Codimension and Holomorphic Differentials
and the equation of Ricci

(1.5) hR⊥ (X, Y )U, V i = h[AU , AV ]X, Y i + hRN (X, Y )U, V i,
for any vector fields X, Y , Z, and W tangent to Σm and any normal vector fields
U and V , where RN , R, and R⊥ are the curvature tensors corresponding to ∇N , ∇,
and ∇⊥ , respectively.
Definition 1.1. If the mean curvature vector field H of Σm is parallel in the
normal bundle, i.e., ∇⊥ H = 0, then Σm is called a pmc submanifold .
Definition 1.2. If the mean curvature |H| of Σm is constant, then Σm is called
a cmc submanifold .
Definition 1.3. If the second fundamental form σ of Σm satisfies ∇⊥ σ = 0,
then Σm is called a parallel submanifold .
Now, let N n (c) be a complex space form with complex dimension n, complex
structure (J, h, i), and constant holomorphic sectional curvature c. Then N n (c)
either is CP n (c), or Cn , or CH n (c), as c > 0, c = 0, or c < 0, respectively (see
[135]). The curvature tensor of N n (c) is given by
c
(1.6) RN (U, V )W = {hV, W iU − hU, W iU + hJV, W iJU − hJU, W iJV
4
+ 2hJV, U iJW }.
2.2. A holomorphic differential. We will consider pmc surfaces Σ2 in the
complex space form N n (c) and introduce a holomorphic differential defined on such
surfaces, that will be used to characterize the pmc 2-spheres in the 2-dimensional
complex space forms and also to prove the non-existence of non-pseudo-umbilical
pmc 2-spheres with constant Kähler angle in N n (c).
Theorem 1.4 ([62]). Let Σ2 be a pmc surface in a complex space form N n (c).
Then the (2, 0)-part of the quadratic form Q defined on Σ2 by
Q(X, Y ) = 8|H|2 hσ(X, Y ), Hi + 3chJX, HihJY, Hi,
is holomorphic.
Proof. First, let us consider isothermal coordinates (u, v) on Σ2 . Then we have
ds2 = λ2 (du2 + dv 2 ) and define
1 1
z = u + iv, z̄ = u − iv, dz = √ (du + idv), dz̄ = √ (du − idv)
2 2
and
1 ∂ ∂ 1 ∂ ∂
Z=√ −i , Z̄ = √ +i .
2 ∂u ∂v 2 ∂u ∂v
It is easy to see that hZ, Z̄i = h∂/∂u, ∂/∂ui = h∂/∂v, ∂/∂vi = λ2 .
In the following, we will show that Z̄(Q(Z, Z)) = 0.
First, since ∇Z̄ Z = 0, we have
(∇⊥ ⊥ ⊥
Z̄ σ)(Z, Z) = ∇Z̄ σ(Z, Z) − 2σ(∇Z̄ Z, Z) = ∇Z̄ σ(Z, Z),
and then, since H is parallel,
Z̄(hσ(Z, Z), Hi) = h∇N N
Z̄ σ(Z, Z), Hi + hσ(Z, Z), ∇Z̄ Hi
= h(∇⊥
Z̄ σ)(Z, Z), Hi.
Reduction of Codimension and Holomorphic Differentials 5
Now, from the Codazzi equation (1.4), again using ∇⊥ H = 0, we obtain

(1.7) Z̄(hσ(Z, Z), Hi) =h(∇⊥ N ⊥
Z σ)(Z̄, Z), Hi + h(R (Z̄, Z)Z) , Hi
+ hσ(Z, Z), ∇⊥
Z̄ Hi
=h(∇⊥ N
Z σ)(Z̄, Z), Hi + hR (Z̄, Z)Z, Hi.
Next, since ∇Z Z̄ = 0, we have

(∇⊥ ⊥
Z σ)(Z̄, Z) = ∇Z σ(Z̄, Z) − σ(Z̄, ∇Z Z)
and, therefore, using σ(Z̄, Z) = hZ̄, ZiH, ∇Z Z = (1/λ2 )h∇Z Z, Z̄iZ, and ∇⊥ H = 0,
we get
(1.8) h(∇⊥
Z σ)(Z̄, Z), Hi = 0.
We can also easily check that (JZ)> = (1/λ2 )hJZ, Z̄iZ and then, since ∇N J =
0, ∇N
Z̄
Z = σ(Z̄, Z) = hZ̄, ZiH, and ∇⊥ H = 0, we have
(1.9) Z̄(hJZ, Hi2 ) = −2|H|2 hZ̄, JZihJZ, Hi
Finally, from the expression (1.6) of the curvature tensor field RN , we see that
3c
(1.10) hRN (Z̄, Z)Z, Hi = hZ̄, JZihH, JZi.
4
We conclude replacing (1.8), (1.9), and (1.10) into (1.7).
2.3. Pmc surfaces in 2-dimensional complex space forms. In the follow-
ing, for pmc surfaces Σ2 in N 2 (c), with H 6= 0, we will introduce another quadratic
form Q0 with holomorphic (2, 0)-part. Then, using Q and Q0 , we will classify the
non-minimal pmc 2-spheres in N 2 (c).
Let us consider a pmc surface Σ2 in N 2 (c), a local orthonormal frame {E e2 }
e1 , E
2
on Σ , and denote by θ the Kähler angle function defined by
hJ E e2 i = cos θ.
e1 , E
The immersion x : Σ2 → N is said to be holomorphic if cos θ = 1, anti-holomorphic
if cos θ = −1, and totally real if cos θ = 0. In the following we shall assume that x
is neither holomorphic or anti-holomorphic.
Next, we take E3 = −(H/|H|) and let E4 be the unique unit normal vector field
orthogonal to E3 compatible with the orientation of Σ2 in N . Since E3 is parallel
in the normal bundle so is E4 , and, as the Kähler angle is independent of the choice
of the orthonormal frame on the surface (see [42]), we have hJE4 , E3 i = cos θ.
Now, we consider the vector fields
1 1
E1 = cot θE3 − JE4 and E2 = JE3 + cot θE4
sin θ sin θ
tangent to the surface and get an orthonormal frame field {E1 , E2 , E3 , E4 } along Σ2
in N .
We define a new quadratic form Q0 on Σ2 by
Q0 (X, Y ) = 8i|H|hσ(X, Y ), E4 i + 3chJX, E4 ihJY, E4 i.
In the same way as in the case of Q, also using
hZ̄, JZi = −ihZ̄, ZihE1 , JE2 i = ihZ̄, Zi cos θ,
it follows that also the (2, 0)-part of Q0 is holomorphic.
In order to classify pmc 2-spheres in N 2 (c), we will use a result of T. Ogata in

[113], that we will briefly recall in the following (see also [81, 94]). Using the lo-
cal orthonormal frame field {E1 , E2 , E3 , E4 } and considering isothermal coordinates
(u, v) on the surface, T. Ogata proved that there exist complex-valued functions a
and d on Σ2 such that θ, λ, a, and d satisfy

∂θ ∂λ 2
 ∂z = λ(a + b), ∂ z̄ = −|λ|2 (ā − b) cot θ


∂a 2 3c sin θ
(1.11) ∂ z̄ = λ̄ 2|a| − 2ab + 8 cot θ
 ∂d = 2λ(a − b)d cot θ, |d|2 = |a|2 + c(3 sin2 θ−2)


∂z 8
where z = u + iv and |H| = 2b; and also the converse: if c is a real constant, b a
positive constant, Σ2 a 2-dimensional Riemannian manifold, and there exist some
functions θ, a, and d on Σ2 satisfying (1.11), then there is an isometric immersion
of Σ2 into N 2 (c) with with Kähler angle θ and parallel mean curvature vector field
of length equal to 2b. The second fundamental form of Σ2 in N with respect to
{E1 , E2 , E3 , E4 } is given by
   
−2b − <(ā + d) −=(ā + d) =(ā − d) −<(ā − d)
σ3 =   , σ4 =  ,
−=(ā + d) −2b + <(ā + d) −<(ā − d) −=(ā − d)
where < and = denote the real and the imaginary parts, respectively, of a complex
number, and the Gaussian curvature of Σ2 is K = 4b2 − 4|d|2 + c/2 (see also [81]).
Assume now that the (2, 0)-part of Q and that of Q0 vanish on the surface Σ2 .
It follows, from the expression of the second fundamental form, that d¯ + a ∈ R,
d¯ − a ∈ R, and
32b(d¯ + a) − 3c sin2 θ = 0, 32b(d¯ − a) + 3c sin2 θ = 0.
Therefore d = 0 and a = 3c sin2 θ/(32b) and, from the fifth equation of (1.11), we
get
(1.12) 9c2 sin4 θ + 128cb2 (3 sin2 θ − 2) = 0.
We have to split the study of this equation in two cases. First, if c = 0, then
(1.12) holds and also a = 0. Next, if c 6= 0, it follows that θ is a constant function.
This, together with the first equation of (1.11), leads to a = 3c sin2 θ/(32b) = −b.
Replacing in equation (1.12), we obtain c = −12b2 and then sin2 θ = 8/9. We note
that in both cases the Gaussian curvature of Σ2 is given by K = 4b2 +c/2 = constant
(see [81]). Then, using [81, Theorem 1.1], we obtain the following classification
result.
Theorem 1.5 ([62]). If the (2, 0)-part of Q and the (2, 0)-part of Q0 vanish on
a pmc surface Σ2 in N 2 (c), with |H| = 2b > 0, then either
(1) N 2 (c) = CH 2 (−12b2 ) and Σ2 is the slant surface in [32, Theorem 3(2)];
(2) N 2 (c) = C2 and Σ2 is a part of a round sphere in a hyperplane in C2 .
Since the Gaussian curvature K is nonnegative only in the second case of The-
orem 1.5, we have also recovered a result in [81].
Corollary 1.6 ([81]). If Σ2 is a non-minimal pmc 2-sphere in a 2-dimensional
complex space form, then it is a round sphere in a hyperplane in C2 .
2.4. Reduction of codimension. Let Σ2 be a surface in a complex space form

N n (c),n ≥ 3 and c 6= 0, with parallel mean curvature vector field H 6= 0. We will
prove that either Σ2 is pseudo-umbilical or it lies in a complex space form with the
same holomorphic sectional curvature c and complex dimension 5.
First, since H is parallel, using the Ricci equation (1.5) of Σ2 in N and the
expression (1.6) of the curvature of the complex space form, we immediately have
the following lemma.
Lemma 1.7. For any vector V normal to Σ2 , which is also orthogonal to JT Σ2
and to JH, we have [AH , AV ] = 0, i.e., AH commutes with AV .
Corollary 1.8 ([62]). At any point p ∈ Σ2 either H is an umbilical direction,
or there exists a basis that diagonalizes simultaneously AH and AV , for all normal
vectors satisfying V ⊥ JH, if n = 3 and H ⊥ JT Σ2 , or the conditions in Lemma
1.7, otherwise.
Proposition 1.9 ([62]). Assume that H is nowhere an umbilical direction. Then
there exists a parallel subbundle of the normal bundle that contains the image of the
second fundamental form σ and has dimension less or equal to 8.
Proof. We consider a subbundle L of the normal bundle, given by
L = span{Im σ ∪ (J Im σ)⊥ ∪ (JT Σ2 )⊥ },
where (J(Im σ))⊥ = {(Jσ(X, Y ))⊥ : X, Y tangent to Σ2 }, (J(T Σ2 ))⊥ = {(JX)⊥ :
X tangent to Σ2 }, and we will show that L is parallel.
First, we will prove that, if V is orthogonal to L, then ∇⊥ Ei V is orthogonal to
JT Σ2 and to JH, where {E1 , E2 } is a local orthonormal tangent frame field with
respect to which we have hσ(E1 , E2 ), V i = hσ(E1 , E2 ), Hi = 0.
We get, indeed,
h(JH)⊥ , ∇⊥ ⊥ N N ⊥
Ei V i =h(JH) , ∇Ei V i = −h∇Ei (JH) , V i
>
= − h∇N N
Ei JH, V i + h∇Ei (JH) , V i
=hJAH Ei , V i + hσ(Ei , (JH)> ), V i

=0
and
h(JEj )⊥ , ∇⊥ N ⊥
Ei V i = − h∇Ei (JEj ) , V i
>
= − h∇N N
Ei JEj , V i + h∇Ei (JEj ) , V i
= − hJ∇Ei Ej , V i − hJσ(Ei , Ej ), V i + hσ(Ei , (JEj )> ), V i

=0.
Next, we shall prove that if a normal subbundle S is orthogonal to L, then so is
∇⊥ S, i.e.,
hσ(Ei , Ej ), ∇⊥
Ek V i = 0, hJσ(Ei , Ej ), ∇⊥
Ek V i = 0, and hJEi , ∇⊥
Ek V i = 0
for any V ∈ S and i, j, k ∈ {1, 2}. We only have to verify the first two of these
properties.
Denote by Aijk = h∇⊥ Ek σ(Ei , Ej ), V i and, since σ is symmetric, notice that
Aijk = Ajik . We also have Aijk = −hσ(Ei , Ej ), ∇⊥ Ek V i, since V is orthogonal to L.
Now, we have
h(∇⊥ ⊥
Ek σ)(Ei , Ej ), V i =h∇Ek σ(Ei , Ej ), V i − hσ(∇Ek Ei , Ej ), V i − hσ(Ei , ∇Ek Ej ), V i
=h∇⊥
Ek σ(Ei , Ej ), V i,
and, using the Codazzi equation (1.4),

h(∇⊥ ⊥ N ⊥
Ek σ)(Ei , Ej ), V i =h(∇Ei σ)(Ek , Ej ) + (R (Ek , Ei )Ej ) , V i
=h(∇⊥ N ⊥
Ej σ)(Ek , Ei ) + (R (Ek , Ej )Ei ) , V i
=h(∇⊥ ⊥
Ei σ)(Ek , Ej ), V i = h(∇Ej σ)(Ek , Ei ), V i,
that shows that Aijk = Akji = Aikj .

Next, since ∇⊥ 2
Ek V is orthogonal to JT Σ and to JH, it follows that {E1 , E2 }
diagonalizes A∇⊥ V and we get
Ek
Aijk = −hσ(Ei , Ej ), ∇⊥
Ek V i = −hEi , A∇⊥ Ej i = 0
Ek V
for any i 6= j. Hence, Aijk = 0 if two indices are different from each other.
Finally, we have
Aiii = − hσ(Ei , Ei ), ∇⊥ ⊥ ⊥
Ei V i = −h2H, ∇Ei V i + hσ(Ej , Ej ), ∇Ei V i
=h2∇⊥
Ei H, V i − Ajji = 0.
It is easy to see that, if V is orthogonal to L, then JV is normal and orthogonal

to L. It follows that
h(Jσ(Ei , Ej ))⊥ , ∇⊥ N ⊥
Ek V i = − h∇Ek (Jσ(Ei , Ej )) , V i
>
= − h∇N N
Ek Jσ(Ei , Ej ), V i + h∇Ek (Jσ(Ei , Ej )) , V i
=hJAσ(Ei ,Ej ) Ek , V i − hJ∇⊥

Ek σ(Ei , Ej ), V i
+ hσ(Ek , (Jσ(Ei , Ej ))> ), V i

=h∇⊥
Ek σ(Ei , Ej ), JV i = 0
and we conclude.
From Proposition 1.9, it is easy to see that, when H is not umbilical, T Σ2 ⊕ L
is parallel with respect to the Levi-Civita connection on N . Since we also have
J(T Σ2 ⊕ L) ⊂ T Σ2 ⊕ L along the surface, and then RN (X, Y )Z ∈ T Σ2 ⊕ L for any
X, Y, Z ∈ T Σ2 ⊕ L, we can apply [54, Theorem 2] to conclude that there exists a
totally geodesic submanifold N 5 (c) of N n (c) that contains our surface.
Proposition 1.10 ([62]). If Σ2 is not pseudo-umbilical, then it lies in a complex
space form N 5 (c).
When H is umbilical we use the quadratic form Q to prove the following propo-
sition.
Proposition 1.11 ([62]). Let Σ2 be a pmc surface in N n (c), with c 6= 0 and
H 6= 0. If H is umbilical, then Σ2 is totally real.
Proof. Since H is umbilical, it follows that hσ(Z, Z), Hi = 0, which implies
Q(Z, Z) = 3chJZ, Hi2 .
Next, since Z̄(Q(Z, Z)) = 0, we have

0 = Z̄(hJZ, Hi2 ) = −2|H|2 hJZ, HihJZ, Z̄i.
Hence, hJZ, Z̄i = 0 or hJZ, Hi = 0. Assume that the set of zeroes of equation
hJZ, Z̄i = 0 is not Σ2 . Then, it is a closed set without interior points and its
complement is an open dense set in Σ2 . In this last set we have hJZ, Hi = 0 and
then, since H is parallel and Σ2 is pseudo-umbilical,
0 = Z̄(hJZ, Hi) =hJ∇N N
Z̄ Z, Hi + hJZ, ∇Z̄ Hi
= − hZ̄, ZihJH, Hi − hJZ, AH Z̄i
= − |H|2 hJZ, Z̄i,
which means that Σ2 is totally real.
Remark 1.12. Some kind of a converse result was obtained by B.–Y. Chen and
K. Ogiue [36]. They proved that, if a unit normal vector field to a totally real
2-sphere in a complex space form is parallel and isoperimetric, then it is umbilical.
Remark 1.13. N. Sato [127] proved that, if Σm is a pseudo-umbilical subman-
ifold of a complex space form, with nonzero parallel mean curvature vector field,
then it is a totally real submanifold. Moreover, the mean curvature vector field H
is orthogonal to JT Σm . Therefore, the (2, 0)-part of Q, defined on Σ2 , vanishes.
Remark 1.14. Since the map p ∈ Σ2 → (AH − µ I)(p), where µ is a constant,
is analytic, it follows that if H is an umbilical direction, then this either holds on
Σ2 , or only for a closed set without interior points. In this second case H is not an
umbilical direction in an open dense set. Consequently, only the two above studied
cases can occur.
From Propositions 1.10 and 1.11, it follows the main result of this section.
Theorem 1.15 ([62]). Let Σ2 be a non-minimal pmc surface in a complex space
form N n (c), n ≥ 3, c 6= 0. Then, one of the following holds:
(1) Σ2 is a totally real pseudo-umbilical surface; or
(2) Σ2 is not pseudo-umbilical and lies in a complex space form N 5 (c).
Remark 1.16. The case when c = 0 was treated in [136, Theorem 4].
2.5. A non-existence result for pmc 2-spheres with constant Kähler
angle. We consider pmc surfaces Σ2 in a complex space form N n (c), n ≥ 3, c 6= 0,
with constant Kähler angle and H 6= 0, on which the (2, 0)-part of Q vanishes. We
will compute the Laplacian of |AH |2 and show that there are no 2-spheres with all
these properties.
Let {E1 , E2 } be a local orthonormal frame field on Σ2 such that H ⊥ JE1 . The
fact that the (2, 0)-part of the quadratic form Q vanishes can be written as
(
8|H|2 hσ(E1 , E1 ) − σ(E2 , E2 ), Hi = −3c(hJE1 , Hi2 − hJE2 , Hi2 )
(1.13)
8|H|2 hσ(E1 , E2 ), Hi = 3chJE1 , HihJE2 , Hi,
and, from the second equation, we see that hσ(E1 , E2 ), Hi = 0. It follows that
{E1 , E2 } diagonalizes simultaneously AH and AV , for all normal vectors V as in
Corollary 1.8, since we are in the second case of Theorem 1.15 (see Remark 1.12).
Next, since Σ2 is not holomorphic or anti-holomorphic, we have that cos θ 6= ±1

on an open dense set, where θ is the Kähler angle. We again consider the normal
vector fields
1 1
(1.14) E3 = − cot θE1 − JE2 and E4 = JE1 − cot θE2 ,
sin θ sin θ
with the property that {E1 , E2 , E3 , E4 } is a basis in span{E1 , E2 , JE1 , JE2 }.
It is easy to see that, if H ⊥ JT Σ2 , then our surface is pseudo-umbilical, which
is a contradiction.
On the other hand, if we assume that H ∈ span{E3 , E4 } it follows H = ±|H|E3 ,
since JE1 ⊥ H, and then E3 is parallel. Also, since all normal vectors but E4 verify
conditions in Corollary 1.8 we have σ(E1 , E2 ) k E4 . By using these facts and the
expression of E3 , we obtain σ(Ei , Ej ) ∈ span{E3 , E4 } for i, j ∈ {1, 2}, and then
dim L = 2, where L is the subbundle in Proposition 1.10. Therefore, again using
[54, Theorem 2], we get that Σ2 lies in a complex space form N 2 (c), a case that we
have already studied.
In the following, we will assume that H ∈ / span{E3 , E4 }, and, since we also know
that H is not orthogonal to JT Σ2 , one obtains that H can be written as
H = |H|(cos βE3 + sin βE5 ),
where β is a real-valued function defined locally on Σ2 and E5 is a unit normal
vector field such that E5 ⊥ JT Σ2 . We consider the local orthonormal frame field
{E1 , E2 , E3 , E4 , E5 , E6 = JE5 , . . . , E2n−1 , E2n = JE2n−1 } on N along Σ2 and its
dual frame {ωi }2n 2
i=1 . The vector fields Ei are well defined at the points of Σ where
sin(2β) 6= 0, which, due to our assumptions, form an open dense set in Σ2 . The
structure equations of the surface are
i
dφ = −iω12 ∧ φ and dω12 = − Kφ ∧ φ̄,
2
where φ = ω1 + iω2 , the real 1-form ω12 is the connection form of the Riemannian
metric on Σ2 , and K is the Gaussian curvature of the surface.
A result of T. Ogata [114], together with H ⊥ Ei , i ≥ 4, i 6= 5, implies that, with
respect to the above local orthonormal frame field, the components of the second
fundamental form are given by
 
|H| cos β − <(ā + c) −=(ā + c)
σ3 =  
−=(ā + c) |H| cos β + <(ā + c)
 
=(ā − d) −<(ā − d)
σ4 =  
−<(ā − d) −=(ā − d)
 
|H| sin β − <(ā3 + d3 ) −=(ā3 + d3 )
σ5 =  
−=(ā3 + d3 ) |H| sin β + <(ā3 + d3 )
 
−<(āα + dα ) −=(āα + dα )
σ 2α−1 =  
−=(āα + dα ) <(āα + dα )
 
=(āα − dα ) −<(āα − dα )
σ 2α =  
−<(āα − dα ) −=(āα − dα )
where a, d, aα , dα , with α ∈ {3, . . . , n}, are complex-valued functions defined lo-
cally on Σ2 . We note that, since σ(E1 , E2 ) ⊥ H and σ(E1 , E2 ) ⊥ E5 , it follows
σ(E1 , E2 ) ⊥ E3 . Moreover, since σ(E1 , E2 ) ⊥ Ei for any i ∈ {1, . . . , 2n} \ {4, 6}, we
have ā + d ∈ R, ā3 + d3 ∈ R, and aα = dα for any α ≥ 4.
In the same paper [114], the author computed the differential of the Kähler
angle function θ for a minimal surface. In the same way, this time for our surface,
we get
|H| |H|
0 = dθ = a − cos β φ + ā − cos β φ̄.
2 2
The next step is to determine the connection form ω12 and the differential of the
function β by using that H is parallel. We have
(1.15) ∇⊥ ⊥ ⊥
Ei H = (− sin βE3 + cos βE5 )Ei (β) + cos β∇Ei E3 + sin β∇Ei E5 = 0
for i ∈ {1, 2}, and then
cos βh∇N N
Ei E3 , E4 i + sin βh∇Ei E3 , E4 i = 0,
from where, using equation (1.14) and the expression of the second fundamental
form, we get
tan β
ω12 (E1 ) = cot θ=(ā − d) − =(ā3 − d3 )
sin θ
cot θ θ θ
ω12 (E2 ) = −|H| − 2 cot θ<a + tan β tan <a3 − cot <d3
cos β 2 2
¯
and, therefore, ω12 = f1 φ + f1 φ̄, where
i cot θ tan β ¯ ¯

(1.16) f1 = |H| + 2 cot θa − (a3 − d3 ) + cot θ tan β(a3 + d3 ) .
2 cos β sin θ
Now, from equation (1.15), we also obtain
Ei (β) + h∇N
Ei E3 , E5 i = 0, i ∈ {1, 2}
and then, using (1.14) and the expression of the second fundamental form, one
obtains θ θ
E1 (β) = |H| cot θ sin β + tan <a3 − cot <d3
2 2
and
1
dβ(E2 ) = =(ā3 − d3 ).
sin θ
Hence, the differential of β is given by dβ = f2 φ + f¯2 φ̄, where
1 1
(1.17) f2 = |H| cot θ sin β + (a3 − d¯3 ) − cot θ(a3 + d¯3 ) .
2 sin θ
We note that, since the Kähler angle θ is constant, we have a = ā = (|H|/2) cos β,
and then, from (1.16), it follows
in 1 tan β o
(1.18) f1 = |H| cot θ cos β + − (a3 − d¯3 ) + cot θ tan β(a3 + d¯3 ) .
2 cos β sin θ
Let us now return to the first equation of (1.13), which can be rewritten as
3
µ1 − µ2 = c sin2 θ cos2 β,
8
where AH Ei = µi Ei . Since µ1 + µ2 = 2|H|2 , we have

3 3
µ1 = |H|2 + c sin2 θ cos2 β and µ2 = |H|2 − c sin2 θ cos2 β.
16 16
Thus, we have
9 2 4
(1.19) |AH |2 = µ21 + µ22 = 2|H|4 + c sin θ cos4 β,
128
and then
9 2 4
∆|AH |2 = c sin θ∆(cos4 β).
128
After a straightforward computation, using (1.17) and (1.18), we get
cot θ 2
∆(cos4 β) = 4 cos4 β K + 4|f1 |2 + 12if1 + |H| ,

cos β
which means that
9 2 4 cot θ 2
∆|AH |2 = c sin θ cos4 β K + 4|f1 |2 + 12if1 + |H| .

32 cos β
Assume now that Σ2 is complete and has nonnegative Gaussian curvature. It
follows, from a result of A. Huber in [87], that Σ2 is parabolic. From the above
formula, we get that |AH |2 is a subharmonic function, and, since |AH |2 is bounded
by (1.19), it follows that K = 0, which leads to the following non-existence result.
Theorem 1.17 ([62]). There are no non-minimal pmc 2-spheres with constant
Kähler angle in a non-flat complex space form.
3. Surfaces with parallel mean curvature in CP n × R and CH n × R

3.1. Preliminaries. Let M n (c) be a complex space form with the complex
structure (J, h, iM ), consider the product manifold N 2n+1 = M n (c) × R and define
the following tensors on N 2n+1 :
∂
ϕ = J ◦ dπ, ξ= , η = dt, and h, iN = h, iM + dt ⊗ dt,
∂t
where π : M n (c)×R → M n (c) is the projection map and t is the standard coordinate
function on the real axis. Then (N 2n+1 , ϕ, ξ, η, h, iN ) is a cosymplectic space form
with constant ϕ-sectional curvature equal to c (see [3, 20]). We shall explain what
this means in the following.
An almost contact metric structure on an odd-dimensional manifold N 2n+1 is
given by (ϕ, ξ, η, h, i), where ϕ is a tensor field of type (1, 1) on N , ξ is a vector field,
η is its dual 1-form, and h, i is a Riemannian metric such that
ϕ2 U = −U + hU, ξiξ and hϕU, ϕV i = hU, V i − η(U )η(V ),
for all tangent vector fields U and V . An almost contact metric structure (ϕ, ξ, η, h, i)
is called normal if
Nϕ (U, V ) + 2dη(U, V )ξ = 0,
where
Nϕ (U, V ) = [ϕU, ϕV ] − ϕ[ϕU, V ] − ϕ[U, ϕV ] + ϕ2 [U, V ]
is the Nijenhuis tensor field of ϕ.
Definition 1.18. An almost contact metric manifold (N, ϕ, ξ, η, h, i) is a cosym-

plectic manifold if it is normal and both the 1-form η and the fundamental 2-form
Ω, defined by Ω(U, V ) = hU, ϕV i, are closed.
Equivalently, an almost contact metric manifold is cosymplectic if and only if ϕ
is parallel, i.e., ∇N ϕ = 0, where ∇N is the Levi-Civita connection. This implies that
the vector field ξ and the 1-form η are also parallel. We note that a cosymplectic
manifold has a natural local product structure as a product of a Kähler manifold
and a 1-dimensional manifold, but there exist compact cosymplectic manifolds which
are not global products (see [19, 43]). We also recall that a submanifold Σm of an
almost contact metric manifold is called invariant when ϕ(T Σm ) ⊂ T Σm and anti-
invariant when ϕ(T Σm ) ⊂ N Σm , where N Σm is the normal bundle of Σm .
Let (N, ϕ, ξ, η, h, i) be an almost contact metric manifold. The sectional curva-
ture of a 2-plane generated by U and ϕU , where U is a unit vector orthogonal to
ξ, is called ϕ-sectional curvature determined by U . A cosymplectic manifold with
constant ϕ-sectional curvature c is called a cosymplectic space form and is denoted
by N (c). The curvature tensor field of a cosymplectic space form N (c) is given by
c
(1.20) RN (U, V )W = {hV, W iU − hU, W iV + hU, ϕW iϕV − hV, ϕW iϕU
4
+ 2hU, ϕV iϕW + η(U )η(W )V − η(V )η(W )U
+ hU, W iη(V )ξ − hV, W iη(U )ξ}.
3.2. A holomorphic differential. Although our main interest is to study pmc

surfaces in product spaces of type M n (c) × R, where M n (c) is a complex space form,
it is more convenient to treat the more general case where the surfaces are immersed
in an arbitrary cosymplectic space form.
Let us consider N 2n+1 (c) a cosymplectic space form endowed with the cosym-
plectic structure (ϕ, ξ, η, h, i) and constant ϕ-sectional curvature c.
Working in the same way as in the proof of Theorem 1.4, one obtains the fol-
lowing result.
Theorem 1.19 ([77]). If Σ2 is a pmc surface in a cosymplectic space form
N 2n+1 (c), then the (2, 0)-part of the quadratic form Q, defined on Σ2 by
Q(X, Y ) = 8|H|2 hσ(X, Y ), Hi − c|H|2 η(X)η(Y ) + 3chϕX, HihϕY, Hi,
is holomorphic.
3.3. Some examples of pmc surfaces. Let us first recall the definition of
Frenet curves in a Riemannian manifold.
Definition 1.20. Let γ : I ⊂ R → M n be a curve parametrized by arc-length
in a Riemannian manifold. The curve γ is called a Frenet curve of osculating order
r, 1 ≤ r ≤ n, if there exist r orthonormal vector fields {E1 = γ 0 , . . . , Er }, along γ,
such that
∇M
E1 E1 = κ1 E2 , ∇M
E1 Ei = −κi−1 Ei−1 + κi Ei+1 , ∇M
E1 Er = −κr−1 Er−1 ,
for i ∈ {2, . . . , r − 1}, where {κ1 , . . . , κr−1 } are positive functions on I called the
curvatures of γ. A Frenet curve of osculating order r is called a helix of order r if
κi = constant > 0, 1 ≤ i ≤ r − 1. A helix of order 2 is called a circle, and a helix of
order 3 is simply called a helix.
When γ is a Frenet curve in a complex space form M n (c), then its complex
torsions are defined by τij = hEi , JEj i, 1 ≤ i < j ≤ r, where (J, h, i) is the complex
structure on M n (c). A helix of order r is called a holomorphic helix of order r if all
complex torsions are constant. It is easy to see that a circle is always a holomorphic
circle (see [103]).
Definition 1.21. Let M be a Riemannian manifold and consider the product
manifold M × R. A submanifold Σm in M × R is called a vertical cylinder over Σm−1
if Σm = π −1 (Σm−1 ), where π : M × R → M is the projection map and Σm−1 is a
submanifold in M .
It is easy to see that vertical cylinders Σm = π −1 (Σm−1 ) are characterized by
the fact that the unit vector field ξ tangent to R is also tangent to Σm .
In order to find examples of pmc surfaces we will focus our attention on vertical
cylinders Σ2 = π −1 (γ) in product spaces M n (c) × R, where π : M n (c) × R → M n (c)
is the projection map and γ : I → M n (c) is a Frenet curve of osculating order r in
M n (c). For any vector field X tangent to M n (c) we shall denote by X H its horizontal
lift to M n (c) × R. As for the Riemannian metrics on M n (c) and M n (c) × R, we will
use the same notation h, i.
Obviously, {E1H , ξ} is a local orthonormal frame on Σ2 and EiH , 1 < i ≤ r, are
normal vector fields. Then the mean curvature vector H is given by
1 1
H = (σ(E1H , E1H ) + σ(ξ, ξ)) = κ1 E2H ,
2 2
where κ1 = κ1 ◦ π and we used the first Frenet equation for γ and O’Neill’s equation
([110]) in the case of cosymplectic space forms, i.e., ∇N XH
Y H = (∇M H
X Y ) , for any
vector fields X and Y tangent to M n (c) (see also [3]).
Next, from the second Frenet equation, we have
1 1
(1.21) ∇N H = (∇M (κ1 E2 ))H = (κ01 E2 − κ21 E1 + κ1 κ2 E3 )H .
E1H 2 E1 2
It is easy to verify that ∇ξ E1H = ∇E H ξ = 0, where ∇ is the connection on the
1
surface, and then we get that [ξ, E1H ] = 0, which means ∇N H N
ξ E1 = ∇E H ξ = 0. Now,
1
since from (1.20) it follows that RN (ξ, E1H )E1H = 0, we obtain
1 N N H
(1.22) ∇Nξ H = ∇ξ ∇E1H E1 = 0.
2
From (1.21) and (1.22), we see that H is parallel if and only if either
• γ is a geodesic in M n (c); or
• γ is a circle in M n (c) with the curvature κ1 = 2|H| = constant > 0.
Obviously, in the first case, Σ2 is a minimal surface. In the second case, the (2, 0)-
part of Q vanishes if and only if
16|H|4 + c|H|2 + 3chϕE1H , Hi2 = 0,
that is equivalent to
4κ21 + c(1 + 3τ12
2
) = 0.
Proposition 1.22 ([77]). A vertical cylinder Σ2 = π −1 (γ) in M n (c) × R has
non-zero parallel mean curvature vector and the (2, 0)-part of the quadratic form
Q vanishes on Σ2 ifpand only if c < 0 and the curve γ is a circle in M n (c) with
curvature κ = (1/2) −c(1 + 3τ 2 ), where τ is the complex torsion of γ.
Remark 1.23. S. Maeda and T. Adachi proved in [102] that for any positive
number κ and for any number τ , such that |τ | < 1, there exists a circle with
curvature κ and complex torsion τ in any complex space form. Therefore, for any
c < 0, we know that circles
√ γ, as in√the previous proposition, do exist. Since
0 ≤ τ 2 ≤ 1, we get that −c/2 ≤ κ ≤ −c, which means that the mean curvature
of 2 −1
√ a non-minimal√pmc cylinder Σ = π (γ), with vanishing (2, 0)-part of Q, satisfies
−c/4 ≤ |H| ≤ −c/2.
3.4. Reduction of codimension. Let Σ2 be a non-minimal pmc surface in a
non-flat cosymplectic space form N 2n+1 (c), n ≥ 2.
From Ricci equation (1.5) and the expression (1.20) of the curvature tensor RN ,
we get the following lemma.
Lemma 1.24. For any vector field V normal to Σ2 , which is also orthogonal to
ϕ(T Σ2 ) and to ϕH, we have [AH , AV ] = 0, i.e., AH commutes with AV .
Corollary 1.25. At each point p ∈ Σ2 , either H is an umbilical direction or
there exists a basis that diagonalizes simultaneously AH and AV , for all normal
vectors V satisfying V ⊥ ϕ(T Σ2 ) and V ⊥ ϕH.
In the following, we will study the case when our surface Σ2 is pseudo-umbilical.
Since H is also parallel, we have
RN (X, Y )H = ∇N N N N N 2 N N
X ∇Y H − ∇Y ∇X H − ∇[X,Y ] H = −|H| (∇X Y − ∇Y X − [X, Y ])
= 0,
for any tangent vector fields X and Y .
We shall prove that, in this case, ξ ⊥ T Σ2 and ϕ(T Σ2 ) ⊂ N Σ2 , where N Σ2 is
the normal bundle of the surface. First, we have the following lemma.
Lemma 1.26 ([77]). If Σ2 is a pseudo-umbilical pmc surface, then the following
four relations are equivalent:
(i) ξ ⊥ T Σ2 ;
(ii) H ⊥ ξ;
(iii) ϕ(T Σ2 ) ⊂ N Σ2 ;
(iv) ϕH ⊥ T Σ2 .
Proof. Let us again consider isothermal coordinates (u, v) on the surface and
complex vector fields Z and Z̄ as in the proof of Theorem 1.4.
Since H is umbilical, we have hσ(Z, Z), Hi = 0 and then
Q(Z, Z) = −c|H|2 (η(Z))2 + 3chϕZ, Hi2 .
Since Q(Z, Z) is holomorphic and H is umbilical and parallel, it follows that
hZ, Z̄iη(Z)η(H) + 3hϕZ, HihϕZ, Z̄i = 0.
Now, it is easy to see that η(Z)η(H) = 0 is equivalent to hϕZ, HihϕZ, Z̄i = 0,
and then we only have to prove the equivalence between (i) and (ii) and between
(iii) and (iv), respectively.
First, we prove that (i) is equivalent to (ii). If η(Z) = 0 then 0 = η(∇NZ̄
Z) =
hZ, Z̄iη(H), since N 2n+1 (c) is a cosymplectic space form and ∇N Z̄
Z = hZ, Z̄iH.
Conversely, if η(H) = 0, we have
η(∇N 2
Z H) = −η(AH Z) = −|H| η(Z) = 0.
Next, we prove that (iii) is equivalent to (iv). Since RN (X, Y )H = 0, for any
tangent vector fields X and Y , we have
c
0 = RN (Z̄, Z)H = {hϕH, Z̄iϕZ − hϕH, ZiϕZ̄ + hϕZ, Z̄iϕH
4
+ η(Z̄)η(H)Z − η(Z)η(H)Z̄}.
Assume that relation (iii) holds, i.e., hϕZ, Z̄i = 0. As we have seen, this also implies
η(Z) = η(Z̄) = 0 and η(H) = 0. Then, using the definition of the cosymplectic
structure on N 2n+1 (c), we have
c
0 = hRN (Z̄, Z)H, ϕZi = − hZ, Z̄ihϕH, Zi.
4
Conversely, if (iv) holds, i.e., if hϕH, Zi = 0, we have
0 =h∇N N N
Z̄ ϕH, Zi = hϕ∇Z̄ H, Zi + hϕH, ∇Z̄ Zi = −hϕAH Z̄, Zi + hZ, Z̄ihϕH, Hi
=|H|2 hϕZ, Z̄i,
which completes the proof.
Now, let us consider a local orthonormal frame field {E1 , E2 } on Σ2 such that
E1 ⊥ ξ, i.e., η(E1 ) = 0.
Lemma 1.27 ([77]). If relations (i) − (iv) in Lemma 1.26 do not hold on Σ2 ,
then we have
(1) 2|H|2 hϕE2 , E1 i = hϕH, σ(E1 , E2 )i;
(2) hϕH, σ(E1 , E1 )i = hϕH, σ(E2 , E2 )i = 0;
(3) ∇E2 E2 = ∇E2 E1 = 0;
(4) η(σ(E1 , E2 )) = 0 and hϕE1 , σ(E1 , E2 )i = 0.
Proof. Let us assume that relations (i)−(iv) do not hold on our surface. Then,
from hRN (E1 , E2 )H, E2 i = 0, we obtain
hϕE2 , E1 ihϕH, E2 i = 0,
which means that hϕH, E2 i = 0, and then RN (E1 , E2 )H = 0 can be written as
(1.23) 2hϕE2 , E1 iϕH + hϕH, E1 iϕE2 − η(E2 )η(H)E1 = 0.
We take the inner product of this equation with ϕH, E1 , and ϕE2 and obtain
(1.24) hϕE2 , E1 ihϕH, ϕHi = η(E2 )η(H)hϕH, E1 i,
(1.25) 3hϕE2 , E1 ihϕH, E1 i = η(E2 )η(H),

and
(1.26) 3hϕE2 , E1 iη(E2 )η(H) = hϕH, E1 ihϕE2 , ϕE2 i,
respectively. Since hϕE2 , E1 i 6= 0 and hϕH, E1 i 6= 0, from the first two equations,
we get
(1.27) hϕH, ϕHi = |H|2 − (η(H))2 = 3hϕH, E1 i2
and, from the last two,
(1.28) hϕE2 , ϕE2 i = 1 − (η(E2 ))2 = 9hϕE2 , E1 i2 .
Now, from equation (1.27), it follows that
2hϕH, ϕ∇N N N
E2 Hi = 6hϕH, E1 i(hϕ∇E2 H, E1 i + hϕH, ∇E2 E1 i).
But we also know that ∇N 2

E2 H = −|H| E2 and hϕH, ∇E2 E1 i = 0, since hϕH, E2 i = 0.
Replacing in the above equation and using equation (1.25), one obtains
2|H|2 hϕE2 , E1 i = hϕH, σ(E1 , E2 )i.
In the same manner, from equation (1.27), we obtain hϕH, σ(E1 , E1 )i = 0, and
then hϕH, σ(E2 , E2 )i = 0.
As hϕH, E2 i = 0, we get hϕH, ∇N
E2 E2 i = 0, which implies
hϕH, ∇E2 E2 i = hϕH, E1 ih∇E2 E2 , E1 i = 0,

meaning that ∇E2 E2 = 0. Since E1 ⊥ E2 , we also have ∇E2 E1 = 0.
Finally, η(E1 ) = 0 and ∇N ξ = 0 imply η(∇N E2 E1 ) = 0. Since ∇E2 E1 = 0, it
follows that η(σ(E1 , E2 )) = 0. Then the last identity in our lemma follows easily by
taking the inner product of (1.23) with ϕσ(E1 , E2 ).
Proposition 1.28 ([77]). Let Σ2 be a non-minimal pmc surface in a non-flat
cosymplectic space form N 2n+1 (c), n ≥ 2. If the mean curvature vector H is an um-
bilical direction everywhere, then ξ ⊥ T Σ2 , ϕ(T Σ2 ) ⊂ N Σ2 , and n ≥ 3. Moreover,
H ⊥ ξ and ϕH ⊥ T Σ2 .
Proof. We will assume that relations (i) − (iv) in Lemma 1.26 do not hold on
Σ2 and we will prove that this leads to a contradiction.
Consider the same local frame field {E1 , E2 } on the surface as in Lemma 1.27.
From the expression of the curvature tensor RN , it can be easily checked that RN
is parallel, i.e., ∇N RN = 0. Therefore, we have (∇N N
E1 R )(E1 , E2 , H) = 0 and then,
since RN (X, Y )H = 0 for any tangent vectors X and Y , one obtains
|H|2 RN (E1 , E2 , E1 ) − RN (σ(E1 , E1 ), E2 , H) − RN (E1 , σ(E1 , E2 ), H) = 0.
By using (1.20), (1.23), and Lemma 1.27, the above equation becomes, after a
straightforward computation,
η(σ(E1 , E1 ))η(H)E2 − η(E2 )η(H)σ(E1 , E1 ) + hϕH, E1 iϕσ(E1 , E2 )
−5|H|2 hϕE2 , E1 iϕE1 = 0,
and, by taking the inner product with E2 , we obtain that
(1.29) η(σ(E1 , E1 ))η(H) + 9|H|2 hϕE2 , E1 i2 = 0.
Next, from equations (1.25), (1.27), and (1.28), it follows that
3|H|2 hϕE2 , E1 i2 = (1 − 6hϕE2 , E1 i2 )(η(H))2 .
Hence, replacing in (1.29), we get η(σ(E1 , E1 )) = 3η(H)(6hϕE2 , E1 i2 − 1) and then
η(σ(E2 , E2 )) = η(H)(5 − 18hϕE2 , E1 i2 ), which means that
(1.30) η(∇N 2
E2 E2 ) = η(H)(5 − 18hϕE2 , E1 i ),
since ∇E2 E2 = 0.
From equation (1.28), one obtains
2η(E2 )η(∇N
E2 E2 ) = −18hϕE2 , E1 iE2 (hϕE2 , E1 i),
and then, from (1.30) and (1.25), we have

(1.31) 3E2 (hϕE2 , E1 i) = (18hϕE2 , E1 i2 − 5)hϕH, E1 i.
Finally, we differentiate (1.25) and, using equations (1.30) and (1.31), the fact
that H is both umbilical and parallel, and Lemma 1.27, we obtain
|H|2 + (5 − 18hϕE2 , E1 i2 )(η(H))2 = 0.
But, from equation (1.28), we know that 9hϕE2 , E1 i2 < 1 and, therefore, the last
equation is a contradiction.
Thus, we have that ξ ⊥ T Σ2 , ϕ(T Σ2 ) ⊂ N Σ2 , H ⊥ ξ, and ϕH ⊥ T Σ2 . Now, it is
easy to see that, at any point on the surface, the system {E1 , E2 , ϕE1 , ϕE2 , H, ϕH, ξ}
is linearly independent, which means that n ≥ 3.
If N 2n+1 (c) = M n (c) × R, n ≥ 2 and c 6= 0, and Σ2 is a surface in N 2n+1 (c)
as in Proposition 1.28, then Σ2 is a totally real surface in M n (c). Moreover, since
N 2n+1 (c) is a product space, we have ∇N Z̄
Z = ∇M
Z̄
Z, ∇N M N
Z Z = ∇Z Z, and ∇X H =
∇M 2
X H, for any vector field X tangent to Σ , where we used that H ⊥ ξ. From
these identities, we see that the surface is pseudo-umbilical and with parallel mean
curvature vector in M n (c).
Corollary 1.29 ([77]). Let Σ2 be a non-minimal pmc surface in M n (c) × R,
n ≥ 2 and c 6= 0. If Σ2 is pseudo-umbilical then it is a non-minimal pseudo-umbilical
totally real pmc surface in M n (c) and n ≥ 3.
Remark 1.30. If the mean curvature vector field of the surface Σ2 is umbilical
everywhere, then the (2, 0)-part of the quadratic form Q, defined on Σ2 , vanishes.
The next step is to study the case when the mean curvature vector field of the
surface is not umbilical on an open dense set. We work on this set and then extend
our results throughout Σ2 by continuity.
To prove our next result, we can use the same method employed in the proof of
Proposition 1.9, considering a subbundle L of the normal bundle, defined by
L = span{Im σ ∪ (ϕ(Im σ))⊥ ∪ (ϕ(T Σ2 ))⊥ ∪ ξ ⊥ },
where (ϕ(T Σ2 ))⊥ = {(ϕX)⊥ : X tangent to Σ2 }, (ϕ(Im σ))⊥ = {(ϕσ(X, Y ))⊥ :
X, Y tangent to Σ2 }, and ξ ⊥ is the normal component of ξ along the surface.
Proposition 1.31 ([77]). If H is not umbilical on an open set, then there exists
a parallel subbundle of the normal bundle that contains the image of the second
fundamental form σ and has dimension less than or equal to 9.
Since ϕ(T Σ2 ⊕ L) ⊂ T Σ2 ⊕ L and ξ ∈ T Σ2 ⊕ L along the surface, it follows that
RN (X, Y)Z ∈ T Σ2 ⊕ L for any X, Y, Z ∈ T Σ2 ⊕ L. Therefore, using [54, Theorem 2]
and H. Endo’s result in [51], we obtain the following proposition.
Proposition 1.32 ([77]). Let Σ2 be a non-minimal pmc surface in a non-flat
cosymplectic space form N 2n+1 (c), n ≥ 2. If its mean curvature vector field is
nowhere an umbilical direction, then the surface lies in a cosymplectic space form
N 11 (c).
Consider the cosymplectic space form N 2n+1 (c) = M n (c) × R. Again using that
ϕ(T Σ2 ⊕ L) ⊂ T Σ2 ⊕ L and ξ ∈ T Σ2 ⊕ L, we get the following corollary.
Corollary 1.33 ([77]). Let Σ2 be a non-minimal pmc surface in M n (c) × R,
n ≥ 2 and c 6= 0. If its mean curvature vector is nowhere an umbilical direction,
then Σ2 lies in M 5 (c) × R.
Remark 1.34. As in the case of pmc surfaces in complex space forms (see Re-
mark 1.14), the map p ∈ Σ2 → (AH − µ I)(p), where µ = constant, is analytic, and
then either H is an umbilical direction everywhere, or only on a closed set without
interior points.
Summarizing, we can state
Theorem 1.35 ([77]). Let Σ2 be a non-minimal pmc surface in a non-flat cosym-
plectic space form N 2n+1 (c), n ≥ 2. Then, one of the following holds:
(1) Σ2 is pseudo-umbilical and then ξ ⊥ T Σ2 , ϕ(T Σ2 ) ⊂ N Σ2 , H ⊥ ξ, ϕH ⊥
T Σ2 , and n ≥ 3; or
(2) Σ2 is not pseudo-umbilical and lies in a cosymplectic space form N 11 (c).
Corollary 1.36 ([77]). Let Σ2 be a non-minimal pmc surface in N 2n+1 (c) =
M n (c) × R, where M n (c) is a non-flat complex space form with n ≥ 2. Then one of
the following holds:
(1) Σ2 is pseudo-umbilical in N 2n+1 (c) and then it is a pseudo-umbilical non-
minimal totally real pmc surface in M n (c) and n ≥ 3; or
(2) Σ2 is not pseudo-umbilical in N 2n+1 (c) and then it lies in M 5 (c) × R.
3.5. Anti-invariant pmc surfaces. Let Σ2 be a non-minimal anti-invariant
pmc surface in a non-flat cosymplectic space form N 2n+1 (c) and define a new qua-
dratic form Q0 on Σ2 by
Q0 (X, Y ) = 8hσ(X, Y ), Hi − cη(X)η(Y ).
In the same way as in the case of Q, it can be proved that the (2, 0)-part of Q0 is
holomorphic.
In the following, we shall assume that the (2, 0)-parts of Q and Q0 vanish on the
surface, i.e., the following equations hold on Σ2 :

2 2 2 2
8|H| hσ(E1 , E1 ) − σ(E2 , E2 ), Hi − c|H| ((η(E1 )) − (η(E2 )) )

(1.32) +3c(hϕE1 , Hi2 − hϕE2 , Hi2 ) = 0
8|H|2 hσ(E1 , E2 ), Hi − c|H|2 η(E1 )η(E2 ) + 3chϕE1 , HihϕE2 , Hi) = 0


and
(
8hσ(E1 , E1 ) − σ(E2 , E2 ), Hi − c((η(E1 ))2 − (η(E2 ))2 ) = 0
(1.33)
8hσ(E1 , E2 ), Hi − cη(E1 )η(E2 ) = 0,
where {E1 , E2 } is a local orthonormal frame field on the surface.
From (1.33), it follows that ξ is orthogonal to the surface at a point p if and only
if H is an umbilical direction at p. Therefore, using Remark 1.34, we obtain that
either ξ is orthogonal to the surface at every point or this holds only in a closed set
without interior points. From Theorem 1.35 we know that the first case is possible
only when n ≥ 3.
Next, if ξ is tangent to the surface at any point in an open, connected subset
of Σ2 , it follows that the Gaussian curvature K of Σ2 vanishes on this set, since
ξ is parallel. Therefore, K vanishes on the entire surface, and this cannot occur
for 2-spheres. However, we have already studied this case when N 2n+1 (c) is the
product between a non-flat complex space form and R (Proposition 1.22). In general,
for a surface in an arbitrary cosymplectic space form N 2n+1 (c), we can choose an
orthonormal frame field {E1 , ξ} on the surface, and easily see that σ(ξ, ξ) = 0,
σ(E1 , ξ) = 0, and σ(E1 , E1 ) = 2H. Moreover, from (1.32) and (1.33), we have that
H ⊥ ϕE1 and c = −16|H|2 .
Remark 1.37. Let f : Σ2 → L(N Σ2 , R) be the map that takes any point p ∈ Σ2
to the linear function fp on Np Σ2 , given by fp (Xp ) = ηp (Xp ), for any normal vector
Xp at p. Obviously, ξ is tangent to the surface at p ∈ Σ2 if and only if fp vanishes
identically. By analyticity, either f is identically zero on the surface or the set of its
zeroes is closed and without interior points.
Now, in order to treat the case where ξ has non-vanishing tangent and normal
components in an open dense set W ⊂ Σ2 , we shall split our study in two cases, as
n = 2 or n ≥ 3. We will work on the open dense set W and then extend our results
throughout Σ2 by continuity.
Case I: n = 2. Let us assume that Σ2 is a pmc surface in N 5 (c), and consider a
local orthonormal frame field {E1 , E2 } on W , where E2 = ξ > /|ξ > | is the unit vector
in the direction of the projection of ξ on the tangent space. Then, since η(E1 ) = 0
and, from (1.32) and (1.33), we have ϕE1 ⊥ H and ϕE2 ⊥ H, it follows that
( )
ϕE2 H
E1 , E2 , E3 = ϕE1 , E4 = , E5 =
|ϕE2 | |H|
is an orthonormal system. Observe that, at any point p ∈ W , the characteristic
vector field ξ can be written as
(1.34) ξ = µE2 + νE5 ,
where µ = η(E2 ) and ν = η(E5 ) = (η(H)/|H|). We will call ν the angle function.
Next, from the second equation of (1.33), we get that {E1 , E2 } diagonalizes AH .
Moreover, using the Ricci equation (1.5), one obtains that {E1 , E2 } also diagonalizes
AϕE1 and AϕE2 , since hRN (E1 , E2 )H, ϕE1 i = hRN (E1 , E2 )H, ϕE2 i = 0.
Finally, the first equation of (1.33) leads to
 
c 2
|H| 1 − 16|H|2 µ 0

λ1 0
(1.35) A5 = AE5 = = .
0 λ2 0 c
|H| 1 + 16|H| 2 µ 2
Lemma 1.38 ([77]). If ξ has non-vanishing tangent and normal components on

W , then we have
(1) E1 (µ) = E1 (ν) = 0;
(2) E2 (µ) = λ2 ν and E2 (ν) = −λ2 µ;
(3) ∇E1 E1 = −(λ1 ν/µ)E2 and ∇E2 E2 = 0;
(4) σ(Ei , Ei ) = (λi /|H|)H, i ∈ {1, 2}.
Proof. The fact that ξ is parallel, together with (1.34) and (1.35), implies that
0 =∇N N
E1 ξ = ∇E1 (µE2 + νE5 )
=E1 (µ)E2 + µ∇E1 E2 − νA5 E1 + µσ(E1 , E2 ) + E1 (ν)E5
=E1 (µ)E2 + µ∇E1 E2 − λ1 νE1 + E1 (ν)E5 .
The tangent and the normal part in the right hand side vanish and then, since
∇E1 E2 ⊥ E2 , it follows that E1 (µ) = E1 (ν) = 0 and ∇E1 E2 = (λ1 ν/µ)E1 . Since
h∇E1 E2 , E1 i + h∇E1 E1 , E2 i = 0 and ∇E1 E1 ⊥ E1 , the last identity is equivalent to
∇E1 E1 = −(λ1 ν/µ)E2 .
In the same way, we get

0 =∇N N
E2 ξ = ∇E2 (µE2 + νE5 )
=E2 (µ)E2 + µ∇E2 E2 − νA5 E2 + µσ(E2 , E2 ) + E2 (ν)E5
=E2 (µ)E2 + µ∇E2 E2 − λ2 νE2 + E2 (ν)E5 + µσ(E2 , E2 )
and then ∇E2 E2 = 0, E2 (µ) = λ2 ν and E2 (ν) = −λ2 µ, since µ2 + ν 2 = 1. We
also obtain that σ(E2 , E2 ) = −(E2 (ν)/µ)E5 = (λ2 /|H|)H and σ(E1 , E1 ) = 2H −
σ(E2 , E2 ) = (λ1 /|H|)H.
Remark 1.39. A direct consequence of the previous lemma is that AϕE1 and
AϕE2 vanish and then the only non-zero component of A is A5 .
Now, assume that the characteristic vector field ξ either is tangent to the surface
or it has non vanishing tangent and normal components in an open dense set W ⊂ Σ2
and consider the subbundle of the normal bundle L = Im σ. On the one hand, it
is easy to see that L is parallel, dim L = 1, ϕX ⊥ Y for any X, Y ∈ T Σ2 ⊕ L.
Also T Σ2 ⊕ L is invariant by RN , since ξ ∈ T Σ2 ⊕ L along the surface. On the
other hand, any non-minimal cmc surface in an anti-invariant totally geodesic 3-
dimensional submanifold of N 5 (c) is a non-minimal anti-invariant pmc surface in
N 5 (c). Moreover, if we assume that the (2, 0)-part of Q0 vanishes on such a surface,
it follows that also the (2, 0)-part of Q vanishes.
Therefore, from [54, Theorem 2], we get our next result.
Theorem 1.40 ([77]). A surface Σ2 can be immersed as a non-minimal anti-
invariant pmc surface in a non-flat cosymplectic space form N 5 (c), with vanishing
(2, 0)-parts of the quadratic forms Q and Q0 , if and only if Σ2 is an immersed non-
minimal cmc surface in a 3-dimensional totally geodesic anti-invariant submanifold
of N 5 (c), such that the (2, 0)-part of Q0 vanishes.
The 3-dimensional totally geodesic anti-invariant submanifolds of M 2 (c) × R,
where M 2 (c) is a non-flat complex space form, are M̄ 2 × R, where M̄ 2 is a totally
geodesic Lagrangian submanifold of M 2 (c). B.-Y. Chen and K. Ogiue proved ([36,
Proposition 3.2]) that a totally geodesic totally real submanifold M̄ m of a non-
flat complex space form M n (c) is a real space form with constant curvature c/4.
Moreover, it is known that S2 (c/4) and H2 (c/4) can be isometrically immersed as
totally geodesic Lagrangian submanifolds in CP 2 (c) and CH 2 (c), respectively (see
[29]).
Hence, a non-minimal anti-invariant surface pmc surface in M 2 (c) × R on which
the (2, 0)-parts of Q and Q0 vanish, is a non-minimal cmc surface in M̄ 2 (c/4)×R with
vanishing (2, 0)-part of Q0 , which, in this case, is the Abresch-Rosenberg differential,
where M̄ 2 (c/4) is a complete simply-connected surface with constant curvature c/4.
U. Abresch and H. Rosenberg proved that there are four classes of such surfaces,
the first three of them, namely the cmc spheres SH 2 ⊂ M̄ 2 (c/4) × R of Hsiang
2
and Pedrosa, their non-compact cousins DH , and surfaces of catenoidal type CH 2,
being embedded and rotationally invariant, and the fourth one being comprised of
parabolic surfaces PH2 (see [1, 2] for a detailed account on these surfaces).
Corollary 1.41 ([77]). A non-minimal anti-invariant pmc surface in M 2 (c) ×
R with vanishing (2, 0)-parts of the quadratic forms Q and Q0 is one of the surfaces
2 , D 2 , C 2 , or P 2 in the product space M̄ 2 (c/4) × R.
SH H H H
Therefore, we have the following theorem.

Theorem 1.42 ([77]). A non-minimal anti-invariant pmc 2-sphere in a non-flat
cosymplectic space form M 2 (c)×R is one of the embedded rotationally invariant cmc
2 ⊂ M̄ 2 (c/4) × R.
spheres SH
Remark 1.43. A surface Σ2 in a cosymplectic space form is called a slant surface

if for all vectors X tangent to Σ2 and orthogonal to ξ the angle θ between ϕX and
Tp Σ2 , p ∈ Σ2 , is constant, i.e., θ does not depend on X or the point p. Obviously,
the invariant and anti-invariant surfaces are slant surfaces. A slant surface which is
neither invariant nor anti-invariant is called a proper slant surface. If Σ2 is a proper
slant surface, then ξ is orthogonal to the surface (see [99]). It follows that, if Σ2 is
an immersed proper slant surface in M 2 (c) × R, then it lies in M 2 (c). On the other
hand, there are no non-minimal pmc 2-spheres in a non-flat complex space form
M 2 (c) (see [81]). Therefore, SH 2 ⊂ M̄ 2 (c/4) × R are the only non-minimal slant pmc
2-spheres in M 2 (c) × R.
In the following, we will see that Lemma 1.38 allows us to make some remarks
on the admissible range of the angle function ν.
Proposition 1.44 ([77]). Let Σ2 be a complete non-minimal cmc surface in
M̄ 2 (c/4) × R
with vanishing Abresch-Rosenberg differential and non-negative Gauss-
ian curvature. Then we have
(1) ν 2 ≥ 1/5 cannot occur on an open dense subset of Σ2 ;
(2) if ν 2 ≤ 1/3 on an open dense subset of Σ2 , then ν vanishes identically and
the surface is a vertical cylinder over a circle in H2 (−4|H|2 ), with curvature
κ = 2|H| and complex torsion equal to 0.
Proof. From Lemma 1.38, after some straightforward computations, one ob-
tains
(1.36) ∆ν 2 = 2λ22 (1 − 3ν 2 )
and
c2
(1.37) ∆|A|2 = λ2 µ2 (5ν 2 − 1).
32|H|2 2
If ν 2 ≥ 1/5 on an open dense subset of Σ2 , then, from (1.37), it follows that |A|2
is a subharmonic function on a parabolic space, and, since |A|2 is bounded by (1.35),
we get that either λ22 = 0, or µ2 = 0, or ν 2 = 1/5. J. M. Espinar and H. Rosenberg
[56] proved that, if the angle function ν is constant and the Gaussian curvature is
non-negative, then either ν 2 = 0, or ν 2 = 1, the second case being possible only
when the surface is minimal. Therefore, since we also know that µ2 cannot vanish
on an open dense subset of Σ2 , one obtains that λ22 = |H|2 (1 + (c/16|H|2 )µ2 )2 = 0,
and then that µ and ν are constant, which means that ν 2 = 0 and µ2 = 1. But this
is a contradiction, since we assumed that ν 2 ≥ 1/5.
If ν 2 ≤ 1/3 on an open dense subset of Σ2 , then, from (1.36), in the same way
as above, we obtain that ν 2 = 0, K = 0, and c = −16|H|2 . In this case, Σ2 is a
vertical cylinder over a circle in H2 (−4|H|2 ), with curvature κ = 2|H| and complex
torsion equal to 0 (see also [56]).
Case II: n ≥ 3. First, we note that, according to Theorem 1.35, the surface
cannot be pseudo-umbilical, since we have assumed that the tangent part of ξ does
not vanish in an open dense set.
Theorem 1.45 ([77]). A non-minimal non-pseudo-umbilical anti-invariant pmc
surface in a non-flat cosymplectic space form N 2n+1 (c), n ≥ 3, with vanishing (2, 0)-
parts of Q and Q0 , lies in a 5-dimensional totally geodesic anti-invariant submanifold
of N 2n+1 (c).
Proof. Let us again consider the local orthonormal frame field {E1 , E2 } on Σ2 ,
where E2 is the unit vector in the direction of the projection of ξ on the tangent space.
From (1.32) and (1.33), we can see that {E1 , E2 } diagonalizes AH in this case too.
Since the surface is anti-invariant, from the Ricci equation, we get [AH , AV ] = 0, for
any normal vector field V and, therefore, {E1 , E2 } diagonalizes AV for any normal
vector field V . We define the subbundle L = span{Im σ ∪ ξ ⊥ } in the normal bundle
and, in the same way as in Proposition 1.31, we can show that, for any normal
vector field V , orthogonal to L, we have hσ(Ei , Ej ), ∇⊥ ⊥ ⊥
Ek V i = 0 and hξ , ∇Ek V i = 0,
i, j, k ∈ {1, 2}, which means that L is parallel. It is also easy to see that T Σ2 ⊕ L is
invariant by RN .
We will prove that ϕX ⊥ Y , for any X, Y ∈ T Σ2 ⊕ L. Since the surface is
anti-invariant, we have ϕE1 ⊥ E2 and, moreover, hϕE1 , ξ ⊥ i = hϕE1 , ξ − ξ > i = 0.
Next, we obtain
hϕE1 , σ(E2 , E2 )i = hϕE1 , ∇N N
E2 E2 i = −hϕ∇E2 E1 , E2 i = −hϕ∇E2 E1 , E2 i = 0,
again using the fact that Σ2 is anti-invariant and σ(E1 , E2 ) = 0. From equations
(1.32) and (1.33), it follows that ϕEi ⊥ H, i ∈ {1, 2}, and then
hϕE1 , σ(E1 , E1 )i = hϕE1 , 2H − σ(E2 , E2 )i = 0.
Since T Σ2 ⊕ L = span{E1 , E2 , σ(E1 , E1 ), σ(E2 , E2 ), ξ ⊥ }, we have just proved that
ϕE1 is orthogonal to T Σ2 ⊕L. In the same way we get that ϕE2 and ϕξ ⊥ = −ϕξ > =
−|ξ > |ϕE2 are orthogonal to T Σ2 ⊕ L. Finally, since ϕH ⊥ Ei , i ∈ {1, 2}, it results
that ϕH is normal and one gets
hϕσ(E1 , E1 ), σ(E2 , E2 )i = hϕσ(E1 , E1 ), 2H − σ(E1 , E1 )i = 2hϕσ(E1 , E1 ), Hi
= −2h∇N N
E1 E1 , ϕHi = 2hE1 , ϕ∇E1 Hi = −2hE1 , ϕAH E1 i
= 0,
which means ϕσ(Ei , Ei ) ⊥ T Σ2 ⊕ L.
Hence, T Σ2 ⊕ L is parallel, invariant by RN , anti-invariant by ϕ, and its dimen-
sion is less than or equal to 5. Using [54, Theorem 2], we come to the conclusion.
When N 2n+1 (c) is of product type, we use [36, Proposition 3.2] to obtain the
last result of this section.
Corollary 1.46 ([77]). A non-minimal non-pseudo-umbilical anti-invariant
pmc surface in M n (c) × R, n ≥ 3 and c 6= 0, with vanishing (2, 0)-parts of Q and
Q0 , lies in a product space M̄ 4 (c/4) × R, where M̄ 4 (c/4) is a space form immersed
as a totally geodesic totally real submanifold in the complex space form M n (c).
Remark 1.47. The non-minimal non-pseudo-umbilical pmc 2-spheres immersed
in M̄ 4 (c/4) × R were characterized by H. Alencar, M. do Carmo, and R. Tribuzy
(see [6, Theorem 2(4)]). In the same paper, they also described non-minimal non-
pseudo-umbilical complete pmc surfaces with non-negative Gaussian curvature with
vanishing (2, 0)-part of Q0 (see [6, Theorem 3(4)]).
Remark 1.48. As we have seen, a proper slant surface Σ2 immersed in M n (c) ×
R, c 6= 0, lies in M n (c). Moreover, as a surface in M n (c), it has constant Kähler
angle. Since in the previous section we proved that there are no non-minimal non-
pseudo-umbilical pmc 2-spheres with constant Kähler angle in a non-flat complex
space form (see Theorem 1.17), it follows that there are no non-minimal non-pseudo-
umbilical proper slant pmc 2-spheres in M n (c) × R.
4. Surfaces with parallel mean curvature in Sasakian space forms

4.1. Preliminaries. Let us consider a contact metric structure (ϕ, ξ, η, h, i) on
an odd-dimensional manifold N 2n+1 , i.e., the structure is almost contact metric and
dη(U, V ) = hU, ϕV i for any vector fields U and V tangent to N .
Definition 1.49. A normal contact metric manifold (N, ϕ, ξ, η, h, i) is called a
Sasakian manifold .
Equivalently, a contact metric manifold is a Sasakian manifold if and only if
(∇N
U ϕ)V = hU, V iξ − η(V )U,
for all tangent vector fields U and V , where ∇N is the Levi-Civita connection. It
can be easily shown that on a Sasakian manifold we have ∇N U ξ = −ϕU (see [19]).
A submanifold Σm of a Sasakian manifold N 2n+1 is called anti-invariant when
ϕ(T Σm ) ⊂ N Σm , where N Σm is the normal bundle of Σm , and integral, if η(X) =
0, for all vector fields X tangent to Σm . The dimension m of an anti-invariant
submanifold satisfies m ≤ n + 1 and that of an integral submanifold m ≤ n. We
also note that any integral submanifold of a Sasakian manifold is anti-invariant. An
integral curve is called a Legendre curve.
A Sasakian manifold N 2n+1 with constant ϕ-sectional curvature c is called a
Sasakian space form and it is denoted by N 2n+1 (c) (see [19]).
The curvature tensor field of a Sasakian space form N 2n+1 (c) is given by
c+3 c−1
(1.38) RN (U, V )W = (hW, V iU − hW, U iV ) + (η(W )η(U )V
4 4
− η(W )η(V )U + hW, U iη(V )ξ − hW, V iη(U )ξ
+ hW, ϕV iϕU − hW, ϕU iϕV + 2hU, ϕV iϕW ).
In [115], it is proved that a Sasakian manifold is locally symmetric if and only if
it is of constant sectional curvature 1. However, all Sasakian space forms are locally
ϕ-symmetric spaces, i.e.,
ϕ2 (∇N N
U R )(X, Y )Z = 0,
for all tangent vector fields U , X, Y , and Z orthogonal to ξ. In order to characterize
locally φ-symmetric spaces, a very useful tool proved to be the affine connection ∇ ¯
introduced by M. Okumura, defined on a Sasakian manifold by
∇¯ U V = ∇N V + TU V,
U
where
TU V = hU, ϕV iξ − η(U )ϕV + η(V )ϕU
(see [115]). The torsion T̄ (U, V ) = 2TU V of this connection does not vanish, but
¯ and T. Takahashi showed that a Sasakian manifold
it is parallel with respect to ∇,
¯ is parallel, i.e., ∇
is locally ϕ-symmetric if and only if the curvature R̄ of ∇ ¯ R̄ = 0.
Here R̄ is given by
R̄(U, V )W =RN (U, V )W + η(W ){η(U )V − η(V )U } + hϕV, W iϕU
− hϕU, W iϕV + 2hϕU, V iϕW + {hU, W iη(V ) − hV, W iη(U )}ξ.
This is equivalent to
(1.39) (∇N N N N N
U R )(X, Y )Z = − TU R (X, Y )Z + R (TU X, Y )Z + R (X, TU Y )Z
+ RN (X, Y )TU Z,
for all tangent vector fields U , X, Y , and Z (see [131]). It is easy to verify that
equation (1.39) holds on any Sasakian space form.
Complete simply connected Sasakian space forms N 2n+1 (c) were classified by
S. Tanno [132], as follows:
• if c > −3, then either N 2n+1 (c) is isometric to the unit sphere S2n+1 en-
dowed with its canonical Sasakian structure, or N (c) is isometric to S2n+1
endowed with a deformed Sasakian structure, described in the following.
Let S2n+1 = {z ∈ Cn+1 : |z| = 1} be the unit 2n + 1-dimensional
sphere endowed with its standard metric field h, i0 . Consider the following
structure tensor fields on S2n+1 : ξ0 = −Iz for each z ∈ S2n+1 , where I is
the usual complex structure on Cn+1 , and ϕ0 = s ◦ J , where s : Tz Cn+1 →
Tz S2n+1 denotes the orthogonal projection. Equipped with these tensors,
S2n+1 becomes a Sasakian space form with ϕ0 -sectional curvature equal to
1.
Next, consider a deformed structure on S2n+1 , given by
1
η = aη0 , ξ= ξ0 , ϕ = ϕ0 ,
a
hU, V i = ahU, V i0 + a(a − 1)η0 (U )η0 (V ),
where a is a positive constant1. Then (S2n+1 , ϕ, ξ, η, h, i) is a Sasakian space
form with constant ϕ-sectional curvature c = (4/a) − 3 > −3.
• if c = −3, then N 2n+1 (c) is isometric to the generalized Heisenberg group
R2n+1 endowed with the following Sasakian structure. On R2n+1 , with
coordinates (x1 , . . . , xn , y 1 , . . . , P
y n , z), consider the vector field ξ = 2(∂/∂z),
its dual 1-form η = (1/2)(dz − ni=1 y i dxi ), the tensor field ϕ given by the
matrix  
0 δij 0
 −δij 0 0  ,
0 yj 0
and the Riemannian metric h, i = (1/4) ni=1 ((dxi )2 + (dy i )2 ) + η ⊗ η.
P
Then (R2n+1 , ϕ, ξ, η, h, i) is a Sasakian space form with constant ϕ-sectional
curvature c = −3.
1Every now and again, into the other chapters, we will denote the deformed metric by h, i ,
a
just to avoid possible confusions.
• if c < −3, then N 2n+1 (c) is isometric to B 2n × R, where B 2n is the unit

ball in Cn , with the Sasakian structure given by the 1-form η = π ∗ ω + dt
and the Riemannian metric h, i = π ∗ G + η ⊗ η, where (J, G) is a Kähler
structure with constant sectional holomorphic curvature k < 0 and exact
fundamental 2-form Ω = dω, π : B 2n ×R → B 2n is the canonical projection,
and t is the coordinate on R. Endowed with this structure, B 2n ×R becomes
a Sasakian space form with constant ϕ-sectional curvature c = k − 3.
4.2. Reduction of codimension. M. J. Ferreira and R. Tribuzy [57] proved

a reduction of codimension theorem for pmc surfaces in symmetric spaces that gen-
eralizes this kind of results in [6, 62, 77]. We will see, in this section, that we also
get such a result working in a non-symmetric space like a Sasakian space form.
Let Σ2 be a non-minimal pmc surface in a Sasakian space form N 2n+1 (c), en-
dowed with a Sasakian structure (ϕ, ξ, η, h, i) and with constant ϕ-sectional curva-
ture c. Since pmc surfaces in Euclidean sphere were studied by S.-T. Yau in [136],
henceforth we shall assume that our ambient space has constant ϕ-sectional curva-
ture c 6= 1.
As in the previous two cases, where we worked in complex and cosymplectic
space forms, either Σ2 is pseudo-umbilical, or its mean curvature vector field H is
umbilical only for a closed set without interior points, in which case it is not an
umbilical direction in an open dense set.
First, we need the following lemma, that, as usual, follows from the Ricci equa-
tion (1.5) and the expression (1.38) of the curvature tensor of N .
Lemma 1.50. For any vector V normal to Σ2 , that is also orthogonal to ϕT Σ2
and to ϕH, we have [AH , AV ] = 0, i.e., AH commutes with AV .
Corollary 1.51. Either H is an umbilical direction or there exists a basis
that diagonalizes simultaneously AH and AV , for all normal vectors V satisfying
V ⊥ ϕT Σ2 and V ⊥ ϕH.
Now, let us assume that the surface is pseudo-umbilical and, since H is also
parallel, we have
RN (X, Y )H = ∇N N N N N 2 N N
X ∇Y H − ∇Y ∇X H − ∇[X,Y ] H = −|H| (∇X Y − ∇Y X − [X, Y ])
= 0,
for any tangent vector fields X and Y .
As we have seen in Section 4.1, the Sasakian space form N 2n+1 (c) is a ϕ-
symmetric space, which means that equation (1.39) holds on N 2n+1 (c). This implies
that
(∇N N N N N
E2 R )(E1 , E2 )H = − TE2 R (E1 , E2 )H + R (TE2 E1 , E2 )H + R (E1 , TE2 E2 )H
+ RN (E1 , E2 )TE2 H.
Then, using these relations and working as in the cosymplectic case (Proposi-
tion 1.28), we get our next result.
Proposition 1.52 ([78]). Let Σ2 be a non-minimal pmc surface in a Sasakian
space form N 2n+1 (c), with c 6= 1. If the mean curvature vector field H is an umbilical
direction everywhere, then n ≥ 3 and Σ2 is an integral surface.
When H is not not umbilical on an open dense set, we work on this set and then
extend our results to the whole surface by continuity.
First we get the following result, working just as in the previous two sections
and using a subbundle L of the normal bundle, defined by
L = span{Im σ ∪ (ϕ(Im σ))⊥ ∪ (ϕ(T Σ2 ))⊥ ∪ ξ ⊥ },
where (ϕ(T Σ2 ))⊥ = {(ϕX)⊥ : X tangent to Σ2 }, (ϕ(Im σ))⊥ = {(ϕσ(X, Y ))⊥ :
X, Y tangent to Σ2 }, and ξ ⊥ is the normal component of ξ along the surface.
Proposition 1.53 ([78]). If H is not an umbilical direction, then there exists
a parallel subbundle of the normal bundle that contains the image of the second
fundamental form σ and has dimension less than or equal to 9.
Proposition 1.54 ([78]). Let L be the normal subbundle considered in Propo-
sition 1.53. Then T Σ2 ⊕ L is parallel with respect to Okumura’s connection ∇ ¯ and
invariant by T̄ and R̄, i.e., T̄ (X, Y ) ∈ T Σ2 ⊕ L and R̄(X, Y )Z ∈ T Σ2 ⊕ L, for any
X, Y, Z ∈ T Σ2 ⊕ L.
Proof. Since ϕ(T Σ2 ⊕ L) ⊂ T Σ2 ⊕ L and ξ ∈ T Σ2 ⊕ L, it is easy to see that
T Σ2 ¯ X Y ∈ T Σ2 ⊕ L, for
⊕ L is invariant by T̄ and R̄. It is also easy to see that ∇
any vector fields X and Y tangent to Σ .2
Next, let us consider a normal vector field V ∈ L. Then

¯ X V = ∇N ⊥
∇ X V + TX V = −AV X + ∇X V + hX, ϕV iξ − η(X)ϕV + η(V )ϕX,
which means that ∇¯ X V ∈ T Σ2 ⊕ L if and only if ∇⊥ V ∈ L. Since from Proposition

X
(1.53) we know that L is parallel, it follows that T Σ2 ⊕ L is parallel with respect to
Okumura’s connection ∇.¯
It is easy to see that, if γ : I → N 2n+1 (c) is a parametrized curve, then
¯
∇γ 0 γ 0 = ∇N 0 ¯ N have the same geodesics,
γ 0 γ , which means that connections ∇ and ∇
and therefore, that ∇ ¯ is a complete connection. Then, using that ∇
¯ T̄ = 0, ∇
¯ R̄ = 0
and Proposition 1.54, we can apply [54, Theorem 2] to prove that our surface
lies in an 11-dimensional totally geodesic submanifold of N 2n+1 (c). But, since
ϕ(T Σ2 ⊕ L) ⊂ T Σ2 ⊕ L and ξ ∈ T Σ2 ⊕ L, this totally geodesic submanifold actually
is a Sasakian space form with the same ϕ-sectional curvature c as the ambient space
(see [135]).
We conclude with the following proposition.
Proposition 1.55 ([78]). Let Σ2 be an isometrically immersed non-minimal
pmc surface in a Sasakian space form N 2n+1 (c). If its mean curvature vector field
is not an umbilical direction, then the surface lies in an 11-dimensional Sasakian
space form N 11 (c).
Now, from Propositions 1.52 and 1.55, we have the following reduction of codi-
mension theorem.
Theorem 1.56 ([78]). Let Σ2 be an isometrically immersed non-minimal pmc
surface in a Sasakian space form N 2n+1 (c) with constant ϕ-sectional curvature c 6= 1.
Then one of the following holds:
(1) Σ2 is an integral pseudo-umbilical surface and n ≥ 3; or
(2) Σ2 is not pseudo-umbilical and lies in a Sasakian space form N 11 (c).
4.3. Anti-invariant pmc surfaces.

4.3.1. Holomorphic differentials. Let Σ2 be a surface in a Sasakian space form
N 2n+1 (c). Let us consider the following two quadratic forms on Σ2 :
Q1 (X, Y ) = 8hσ(X, Y ), Hi − (c − 1)η(X)η(Y )
and
Q2 (X, Y ) = hϕX, HihϕY, Hi + η(X)η(Y ) − η(X)hϕY, Hi − η(Y )hϕX, Hi,
where σ is the second fundamental form of the surface and H its mean curvature
vector field.
Working as in Theorem 1.4, we can prove our next result.
Theorem 1.57 ([78]). If Σ2 is an anti-invariant pmc surface in a Sasakian space
(2,0) (2,0)
form N 2n+1 (c), then the (2, 0)-parts Q1 and Q2 of Q1 and Q2 , respectively, are
holomorphic.
4.3.2. Hopf cylinders. Let us now consider the orbit space N̄ = N/ξ of the
Sasakian space form N 2n+1 (c). Then N̄ is a complex space form with constant
holomorphic sectional curvature c + 3 (see [135]). The fibration π : N → N̄ is called
the Boothby-Wang fibration. An example of such a fibration is the well known Hopf
fibration π : S2n+1 → CP n .
In order to find examples of anti-invariant pmc surfaces in the Sasakian space
form N 2n+1 (c) we will first study Hopf cylinders, i.e., surfaces Σ2 = π −1 (γ), where
π : N → N̄ is the Boothby-Wang fibration and γ : I → N̄ n (c + 3) is a Frenet curve
of osculating order r in N̄ n (c + 3). For any vector field X tangent to N̄ n (c + 3) we
shall denote by X H its horizontal lift to N 2n+1 (c). For the Riemannian metrics on
N̄ n (c + 3) and N 2n+1 (c) we will use the same notation h, i.
Since {E1H , ξ} is a local orthonormal frame on Σ2 and EiH , 1 < i ≤ r, are normal
vector fields, the mean curvature vector H of Σ2 is given by
1 1
(1.40) H = (σ(E1H , E1H ) + σ(ξ, ξ)) = κ1 E2H ,
2 2
where κ1 = κ1 ◦ π and we used the first Frenet equation of γ and O’Neill’s equation
for Riemannian submersions in the case of Boothby-Wang fibration, i.e.,
∇N
XH Y
H
= (∇N̄ H H H
X Y ) − hX , ϕY iξ,
for any vector fields X and Y tangent to N̄ (c + 3) (see [110]). The first consequence
of this equation is that κ1 = 2|H| and then a Hopf cylinder Σ2 = π −1 (γ) is minimal
in N 2n+1 (c) if and only if the curve γ is a geodesic in N̄ n (c + 3).
In [90], it is proved that a non-minimal pmc Hopf cylinder Σ2 lies in a Sasakian
space form N 3 (c) of dimension 3. In this case, since ϕE1H is orthogonal to Σ2 , it
follows that H = ±|H|ϕE1H . Then equation (1.40) implies that τ12 = hE1 , JE2 i =
±1, where J is the complex structure on N̄ 1 (c + 3).
Next, from the second Frenet equation of γ, one obtains
1
∇N
E1H
H = {(∇N̄ H H H
E1 (κ1 E2 )) − κ1 hE1 , ϕE2 iξ}
2
1 1
= (κ01 E2 − κ21 E1 + κ1 κ2 E3 )H − κ1 hE1H , ϕE2H iξ.
2 2
It is easy to verify that ∇ξ E1H = ∇E H ξ = 0, where ∇ is the connection on the

1
surface, and then we get that [ξ, E1H ] = 0, which means ∇N H N
ξ E1 = ∇E H ξ = −ϕE1 ,
H
1
from where we obtain that ∇N H
ξ H = −ϕH = ∓|H|E1 .
Proposition 1.58 ([78]). A Hopf cylinder Σ2 = π −1 (γ) in a Sasakian space
form N 2n+1 (c) has parallel mean curvature vector field H if and only if either
(1) γ is a geodesic in the orbit space N̄ n (c + 3), i.e., Σ2 is a minimal surface
in N 2n+1 (c); or
(2) γ is a circle in N̄ 1 (c + 3) with curvature κ1 = 2|H| = constant and complex
torsion τ12 = ±1. In this case Σ2 lies in N 3 (c).
Remark 1.59. We note that non-minimal pmc Hopf cylinders do exist in N 3 (c)
(see Remark 1.23).
(2,0) (2,0)
Remark 1.60. It is easy to see that neither Q1 nor Q2 can vanish on a
non-minimal pmc Hopf cylinder, since this would imply that H is orthogonal to
ϕE1H , which is a contradiction.
4.3.3. Integral surfaces. Let Σ2 be an integral non-minimal pmc surface in a 7-
(2,0)
dimensional Sasakian space form N 7 (c) such that Q1 = 0 on Σ2 . It is easy to
see that, since our surface is integral, this condition is equivalent with that that
Σ2 is pseudo-umbilical. From Proposition 1.52, we know that, if {E1 , E2 } is an
orthonormal frame field tangent to the surface, then {E3 = ϕE1 , E4 = ϕE2 , E5 =
H/|H|, E6 = ϕE5 , E7 = ξ} is a normal orthonormal frame field, where H is the
(2,0)
mean curvature vector field of Σ2 . We also note that Q2 = 0 on such a surface.
Since Σ2 is integral, it is easy to verify that
(1.41) hσ(Ei , Ej ), ϕEk i = hσ(Ei , Ek ), ϕEj i, ∀i, j, k ∈ {1, 2}
and
(1.42) A7 = AE7 = 0
and, as the surface is pseudo-umbilical, we also have

|H| 0
(1.43) A5 = AE5 = .
0 |H|
In the following, we shall choose the tangent frame field {E1 , E2 } such that it
diagonalizes A3 = AE3 , and then we have

a 0
(1.44) A3 = ,
0 −a
where a : Σ2 → R is a function on the surface.
Next, using equation (1.41), we obtain
hσ(E1 , E1 ), ϕE2 i = hσ(E1 , E2 ), ϕE1 i = 0
and
hσ(E1 , E1 ), ϕE1 i = −hσ(E2 , E2 ), ϕE1 i = −hσ(E1 , E2 ), ϕE2 i = a,
which means that

0 −a
(1.45) A4 = AE4 = .
−a 0
Since for any tangent vector field X we have hH, ϕXi = 0 and ϕX is normal, it
follows that
h∇⊥ N
Y H, ϕXi + h∇Y ϕX, Hi = 0,
which gives, using that H is parallel and again that ϕH is a normal vector field,
hσ(X, Y ), ϕHi = 0,
for any vector field Y tangent to the surface. Therefore, we have
(1.46) A6 = AE6 = 0.
Next, using the equation of Gauss (1.3), the expression (1.38) of RN , and equa-
tions (1.42)–(1.46), one obtains the Gaussian curvature K of the surface as
(1.47) K = hR(E1 , E2 )E2 , E1 i
= hRN (E1 , E2 )E2 , E1 i + hσ(E1 , E1 ), σ(E2 , E2 )i − |σ(E1 , E2 )|2
c+3
= − 2a2 + |H|2 .
4
On the other hand, since ∇Ei E1 and ∇Ej E2 are orthogonal, we have
(1.48) K = hR(E1 , E2 )E2 , E1 i
= h∇E1 ∇E2 E2 , E1 i − h∇E2 ∇E1 E2 , E1 i − h∇[E1 ,E2 ] E2 , E1 i
= E1 (h∇E2 E2 , E1 i) + E2 (h∇E1 E1 , E2 i) − h∇E1 E1 , E2 i2 − h∇E2 E2 , E1 i2 .
Lemma 1.61 ([78]). The following equations hold on Σ2 :

(1) ∇⊥
Ei ϕEi = h∇Ei Ei , Ej iϕEj + ϕH + ξ, with i 6= j;
⊥
(2) ∇Ei ϕEj = h∇Ei Ej , Ei iϕEi , if i 6= j;
(3) ∇⊥ 2 ⊥
X ϕH = −|H| ϕX and ∇X ξ = −ϕX.
Proof. Since H is parallel and umbilical, we have h∇⊥

X ϕEi , Hi = 0 and also
h∇⊥ N N 2
X ϕEi , ϕHi = −hϕEi , ∇X ϕHi = −hϕEi , ϕ∇X Hi = hϕEi , ϕAH Xi = |H| hEi , Xi,
for any vector field X tangent to the surface.

Now, for i 6= j, one obtains
h∇⊥ N N N
X ϕEi , ϕEj i = h∇X ϕEi , ϕEj i = hϕ∇X Ei , ϕEj i = h∇X Ei , Ej i = h∇X Ei , Ej i.
Finally, we again use that H is parallel and umbilical, ∇N

X ξ = −ϕX and ϕX is
normal, to conclude.
Lemma 1.62. The derivatives of the function a : Σ2 → R are given by
E1 (a) = 3ah∇E2 E2 , E1 i and E2 (a) = 3ah∇E1 E1 , E2 i.
Proof. Using equations (1.42)–(1.46), Lemma 1.61, and the fact that H is
parallel, we can compute
(∇⊥ ⊥
E1 σ)(E2 , E1 ) =∇E1 σ(E2 , E1 ) − σ(∇E1 E2 , E1 ) − σ(E2 , ∇E1 E1 )
= − E1 (a)ϕE2 − a∇⊥
E1 ϕE2 − h∇E1 E2 , E1 iσ(E1 , E1 )
− h∇E1 E1 , E2 iσ(E2 , E2 )
= − E1 (a)ϕE2 + 3ah∇E1 E1 , E2 iϕE1
and, in the same way,

(∇⊥ ⊥
E2 σ)(E1 , E1 ) =∇E2 σ(E1 , E1 ) − 2σ(∇E2 E1 , E1 )
=E2 (a)ϕE1 − 3ah∇E2 E2 , E1 iϕE2 .
From the Codazzi equation (1.4), also using (1.38), one obtains
0 = (RN (E1 , E2 )E1 )⊥ = (∇⊥ ⊥
E1 σ)(E2 , E1 ) − (∇E2 σ)(E1 , E1 ),
which leads to the conclusion.
Now, we are ready to prove the following theorem.
Theorem 1.63 ([78]). Let Σ2 be an isometrically immersed integral complete
non-minimal surface in a Sasakian space form N 7 (c) with parallel mean curvature
(2,0)
vector field H and with non-negative Gaussian curvature K. If Q1 vanishes on
the surface or, equivalently, if Σ2 is pseudo-umbilical, then |H|2 ≥ −(c + 3)/4 and
one of the following holds:
(1) Σ2 is a cmc totally umbilical surface in a space form M 3 ((c + 3)/4), with
constant sectional curvature (c + 3)/4 and of dimension 3, immersed in
N 7 (c) as an integral submanifold; or
(2) Σ2 is flat and it is the standard product γ1 × γ2 , where γ1 : R → N 7 (c) is
a Legendre helix of osculating order 4 in N 7 (c) with curvatures
p p
p a 1 + |H| 2 |H| 1 + |H|2
κ1 = a2 + |H|2 , κ2 = p , and κ3 = p ,
a2 + |H|2 a2 + |H|2
and γ2 : R → N 7 (c) is a Legendre circle in N 7 (c) with curvature κ =
p
a2 + |H|2 , where a2 = (c + 3)/8 + |H|2 /2.
Proof. First, we shall prove that a2 : Σ2 → R is a subharmonic function.
Indeed, after a straightforward computation, using Lemma 1.62 and equation (1.48),
we have
X2
∆a2 = (Ei (Ei (a2 )) − (∇Ei Ei )(a2 ))
i=1
= 6a (E1 (h∇E2 E2 , E1 i) + E2 (h∇E1 E1 , E2 i) + 5h∇E2 E2 , E1 i2 + 5h∇E1 E1 , E2 i2 )
2
= 6a2 {K + 6(h∇E2 E2 , E1 i2 + h∇E1 E1 , E2 i2 )} ≥ 0.

Since K ≥ 0, it follows that Σ2 is a parabolic space. Moreover, from equation
(1.47), we get that 2a2 ≤ (c + 3)/4 + |H|2 , which means that a2 is a bounded
subharmonic function on a parabolic space and, therefore, due to a result in [81], a
constant. Then, either a vanishes or the surface is flat and ∇E1 = ∇E2 = 0.
Case I: a = 0. In this case, hσ(X, Y ), V i = 0 for any normal vector field
V orthogonal to H, which means that the normal subbundle L = span{Im σ} =
span{H} is parallel. Also, it is easy to see that T Σ2 ⊕ L is invariant by T̄ and
R̄, where T̄ and R̄ are the torsion and the curvature of Okumura’s connection,
respectively. Moreover, the characteristic vector field ξ is orthogonal to T Σ2 ⊕ L.
Therefore, we can use [54, Theorem 2] and [19, Proposition 8.1] to conclude that
Σ2 lies is a space form M 3 ((c + 3)/4) immersed in N 7 (c) as an integral submanifold.
Case II: a 6= 0. We have that K = 0, which gives a2 = (c + 3)/8 + |H|2 /2, and
∇E1 = ∇E2 = 0. Since E1 and E2 are parallel, they determine two distributions
which are mutually orthogonal, smooth, involutive and parallel. Therefore, from the
de Rham Decomposition Theorem, follows, also taking into account that the surface
is complete and using its universal cover if necessary, that Σ2 is a product γ1 × γ2 ,
where γi : R → N 7 (c), i ∈ {1, 2}, are integral curves of E1 and E2 , respectively,
parametrized by arc-length, i.e., γ10 = E1 and γ20 = E2 (see [95]). Moreover, since
the surface is integral, the two curves are Legendre curves. In the following, we shall
determine their curvatures.
Let κi , 1 ≤ i < 7, be the curvatures of γ1 and {Xj1 }, 1 ≤ j < 8, be its Frenet
frame field, where X11 = E1 . From equations (1.42)–(1.46), we have ∇N E1 E1 =
σ(E1 , E1 ) = aE3 + |H|E5 and then, from the first Frenet equation of γ1 , it follows
p 1
κ1 = a2 + |H|2 and X21 = − p (aE3 + |H|E5 ).
a + |H|2
2
Next, using Lemma 1.61, we get, after a straightforward computation,

a
∇N 1
E1 X2 = −κ1 E1 + (ϕH + ξ),
κ1
which shows that
p
a 1 + |H|2 a
κ2 = p and X31 = p (|H|E6 + E7 ).
a2 + |H|2 1 + |H|2
p
It follows that ∇N 1 2
E1 X3 = − 1 + |H| E3 and, from the third Frenet equation,
p
|H| 1 + |H|2 1
κ3 = p and X41 = − p (−|H|E3 + aE5 ).
a2 + |H|2 a2 + |H|2
Finally, we obtain ∇N 1 1
E1 X4 = −κ3 X3 , which shows that γ1 is a helix of osculating
order 4.
In the case of the curve γ2 , consider its Frenet frame field {Xj2 }, 1 ≤ j < 8, with
2
X1 = E2 , and, again using equations (1.42)–(1.46), we have
∇NE2 E2 = σ(E2 , E2 ) = −aE3 + |H|E5 ,
p
which means that its first curvature is κ = a2 + |H|2 and
1
X22 = p (−aE3 + |H|E5 ).
a + |H|2
2
Then, using Lemma 1.61, we obtain ∇N 2

E2 X2 = −κE2 . Therefore, the curve γ2 is a
7
circle in N (c).
Using a result in [41], we have the following corollary.
Corollary 1.64 ([78]). An integral pmc 2-sphere in N 7 (c), with H 6= 0, is a
round sphere in a space form M 3 ((c + 3)/4).
4.3.4. Anti-invariant surfaces. In the last part of this section, we consider anti-
invariant non-minimal pmc surfaces Σ2 in a Sasakian space form N 2n+1 (c), c 6= 1,
such that the mean curvature vector field H is not umbilical. We will prove that
there are no 2-spheres with these properties.
(2,0) (2,0)
First, let Σ2 be a surface as above and assume that Q1 and Q2 vanish on
2 2
Σ . If the characteristic vector field ξ is orthogonal to Σ at a point p, then, from
(2,0)
Q1 = 0, it follows that H is umbilical at p. Also, if we assume that ξ is tangent
(2,0)
to the surface at a point p, then Q2 6= 0 at p, which is a contradiction. Therefore,
2 2
since the map p ∈ Σ → (AH − |H| I)(p) is analytic, it follows that the tangent and
normal parts of the characteristic vector field ξ do not vanish on an open dense set.
We shall work on this set and then we will extend our results to the whole surface
by continuity.
Since the tangent part ξ > and the normal part ξ ⊥ of ξ do not vanish, we can
choose an orthonormal frame field {E1 , E2 } on Σ2 such that E2 = ξ > /|ξ > |. Then
(2,0)
we have that η(E1 ) = 0 and, from Q1 = 0, it follows that
c−1 c−1
(1.49) hAH E1 , E1 i = |H|2 − (η(E2 ))2 , hAH E2 , E2 i = |H|2 + (η(E2 ))2
16 16
and hAH E1 , E2 i = 0, which means that {E1 , E2 } diagonalizes AH .
(2,0)
From Q2 = 0, we also obtain
(1.50) hϕE1 , Hi = 0
and
(1.51) hϕE2 , Hi = η(E2 ).
Lemma 1.65 ([78]). The following identities hold on Σ2 :
(1.52) hϕH, σ(Ei , Ei )i = η(H) − η(σ(Ei , Ei )), for i = 1, 2,
(1.53) η(σ(E1 , E1 ) = −η(∇E1 E1 ),
(1.54) ∇E2 E1 = ∇E2 E2 = 0,
2
(1.55) σ(E1 , E2 ) = − ϕE1 ,
η(E2 )
2
(1.56) hσ(E1 , E1 ), ϕE2 i = η(E2 ) − ,
η(E2 )
2
(1.57) hσ(E2 , E2 ), ϕE2 i = η(E2 ) + .
η(E2 )
Proof. Since Σ2 is an anti-invariant pmc surface, from equation (1.50) and
using (1.51), one obtains
0 = E1 (hϕH, E1 i) = h∇N N
E1 ϕH, E1 i + hϕH, ∇E1 E1 i
= hϕ∇N
E1 H, E1 i − η(H) + hϕH, σ(E1 , E1 )i − h∇E1 E1 , E2 ihH, ϕE2 i
= −η(H) + hϕH, σ(E1 , E1 )i − h∇E1 E1 , E2 iη(E2 ),
which means that hϕH, σ(E1 , E1 )i = η(H) + η(∇E1 E1 ). On the other hand, from
η(E1 ) = 0, we easily get that η(∇N E1 E1 ) = 0, i.e., η(∇E1 E1 ) = −η(σ(E1 , E1 )).
Therefore, we have hϕH, σ(E1 , E1 )i = η(H) − η(σ(E1 , E1 )) and hϕH, σ(E2 , E2 )i =
η(H) − η(σ(E2 , E2 )).
In order to prove the third identity, let us first note that ϕE1 is orthogonal to
ϕH and, therefore, to its normal part (ϕH)⊥ . Then, from the Ricci equation (1.5)
and (1.38), we see that AH and A(ϕH)⊥ commute, which means, since {E1 , E2 }
diagonalizes AH and H is not umbilical, that hϕH, σ(E1 , E2 )i = 0.
Now, since E2 (hϕH, E1 i) = 0 and Σ2 is an anti-invariant pmc surface, using
equation (1.51), one obtains hϕH, σ(E1 , E2 )i = h∇E2 E1 , E2 iη(E2 ). Hence, we have
h∇E2 E1 , E2 iη(E2 ) = 0 and then ∇E2 E1 = ∇E2 E2 = 0.
Next, let V be a normal vector field orthogonal to ϕE1 . From the Ricci equation
(1.5) and (1.38), we get that AH and AV commute, which implies that {E1 , E2 }
diagonalizes AV .
Again using (1.5) and (1.38), it follows that
c−1
h[AH , AϕE1 ]E1 , E2 i = −hRN (E1 , E2 )H, ϕE1 i = − η(E2 ).
4
Then, from (1.49), one obtains that hσ(E1 , E2 ), ϕE1 i = −(2/η(E2 )). Thus, we have
σ(E1 , E2 ) = −(2/η(E2 ))ϕE1 .
Finally, since our surface is anti-invariant, we get
hσ(E1 , E1 ), ϕE2 i = h∇N N
E1 E1 , ϕE2 i = −hE1 , ∇E1 ϕE2 i
= −hE1 , ϕ∇NE1 E2 − η(E2 )E1 i = hϕE1 , σ(E1 , E2 )i + η(E2 )
2
= η(E2 ) − ,
η(E2 )
and then, from equation (1.51),
2
hσ(E2 , E2 ), ϕE2 i = h2H − σ(E1 , E1 ), ϕE2 i = η(E2 ) + ,
η(E2 )
which ends the proof.
Lemma 1.66. The second fundamental form σ of Σ2 satisfies
1 3
η(σ(E1 , E1 )) = η(H) and η(σ(E2 , E2 )) = η(H).
2 2
Proof. From equation (1.52), we have
(1.58) hϕσ(E2 , E2 ), σ(E1 , E1 )i = 2hϕσ(E2 , E2 ), Hi = 2η(σ(E2 , E2 )) − 2η(H).
Using Lemma 1.65, one obtains
(1.59) hϕσ(E2 , E2 ), σ(E1 , E1 )i =hϕ∇N N N
E2 E2 , ∇E1 E1 i − hϕ∇E2 E2 , ∇E1 E1 i
=h∇N N N
E2 ϕE2 , ∇E1 E1 i − η(∇E1 E1 )
+ η(E2 )hE2 , ∇N
E1 E1 i
+ h∇E1 E1 , E2 ih∇N
E2 E2 , ϕE2 i
=h∇N N
E2 ϕE2 , ∇E1 E1 i − η(σ(E1 , E1 ))
+ h∇E1 E1 , E2 ihσ(E2 , E2 ), ϕE2 i
=h∇N N
E2 ϕE2 , ∇E1 E1 i − η(σ(E1 , E1 )) + η(∇E1 E1 )
2h∇E1 E1 , E2 i
+
η(E2 )
=h∇N N
E2 ϕE2 , ∇E1 E1 i − 2η(σ(E1 , E1 ))
2h∇E1 E1 , E2 i
+ .
η(E2 )
Next, from equation (1.56), we have
2
(1.60) h∇N
E1 E1 , ϕE2 i = hσ(E1 , E1 ), ϕE2 i = η(E2 ) − ,
η(E2 )
which leads to
2
(1.61) h∇N E
E1 1 , ∇ N
E2 ϕE 2 i + h∇N
∇ N
E
E2 E1 1 , ϕE2 i = E2 η(E2 ) − .
η(E2 )
Since ∇E2 E1 = 0, we get, using (1.56) and (1.55),
2 2
∇N ∇ N
E1 E2 E 1 = ∇ N
E1 σ(E 1 , E2 ) = −E1 ϕE1 − ∇N ϕE1 ,
η(E2 ) η(E2 ) E1
and then, since η(∇N
E1 E1 ) = 0,
2 2h∇E1 E1 , E2 i
h∇N N
E1 ∇E2 E1 , ϕE2 i = − hϕ∇N
E1 E1 , ϕE2 i = − .
η(E2 ) η(E2 )
It follows that
h∇N N
E2 ∇E1 E1 , ϕE2 i =hRN (E2 , E1 )E1 , ϕE2 i + h∇N N N
E1 ∇E2 E1 , ϕE2 i + h∇[E2 ,E1 ] E1 , ϕE2 i
2h∇E1 E1 , E2 i
=hRN (E2 , E1 )E1 , ϕE2 i −
η(E2 )
− h∇E1 E2 , E1 ih∇N
E1 E1 , ϕE2 i.
From here, using the expression (1.38) of the curvature RN and equation (1.60), one
obtains that
4
h∇N ∇
E2 E1
N
E1 , ϕE2 i = η(E2 ) − h∇E1 E1 , E2 i.
η(E2 )
Replacing in (1.61) and using (1.54), we have
2η(σ(E2 , E2 )) 4
h∇N N
E1 E1 , ∇E2 ϕE2 i = η(σ(E2 , E2 )) + − η(E2 ) − h∇E1 E1 , E2 i,
(η(E2 ))2 η(E2 )
and then, from equation (1.59), also using
η(σ(E1 , E1 )) = −η(∇E1 E1 ) = −h∇E1 E1 , E2 iη(E2 ),
it follows, after a straightforward computation, that
hϕσ(E2 , E2 ), σ(E1 , E1 )i =2η(σ(E2 , E2 )) − 2η(H)
2
+ η(σ(E2 , E2 ) − 3σ(E1 , E1 )).
(η(E2 ))2
Finally, from equation (1.58), one sees that η(σ(E2 , E2 )) = 3η(σ(E1 , E1 )), i.e.,
η(σ(E1 , E1 )) = (1/2)η(H) and η(σ(E2 , E2 ) = (3/2)η(H).
Now, we can prove the following proposition.
Proposition 1.67 ([78]). Let f : Σ2 → R be a function on Σ2 given by
1 1 c−1
f = (η(H))2 + (1 + |H|2 )(η(E2 ))2 + (η(E2 ))4 .
2 3 96
Then f is a harmonic function.
Proof. First, since H is parallel and hσ(E1 , E2 ), Hi = 0, from equations (1.49)
and (1.50), we have
E1 (η(H)) = h∇N N
E1 H, ξi + hH, ∇E1 ξi = −hAH E1 , ξi − hH, ϕE1 i = 0
and
E2 (η(H)) = h∇N N
E2 H, ξi + hH, ∇E2 ξi = −hAH E2 , ξi − hH, ϕE2 i
c−1
= −η(E2 ) 1 + |H|2 + (η(E2 ))2 .
16
Next, since ∇E2 E2 = 0, from Lemma 1.66, one obtains
3 3
E2 (η(E2 )) = h∇N
E2 E2 , ξi + hE2 , ∇E2 ξi = η(H) − hE2 , ϕE2 i = η(H).
2 2
From Lemmas 1.65 and 1.66, we have that h∇E1 E1 , E2 i = −η(H)/(2η(E2 )).
Using all these equations, it is now straightforward to compute ∆f , where ∆ =
trace ∇2 = trace(∇∇ − ∇∇ ). We obtain
c−1
∆(η(H))2 =2(η(E2 ))2 1 + |H|2 + (η(E2 ))2
16
5(c − 1)
− 4(η(H))2 1 + |H|2 + (η(E2 ))2 ,
32
c−1
∆(η(E2 ))2 = 6(η(H))2 − 3(η(E2 ))2 1 + |H|2 + (η(E2 ))2 ,
16
c−1
∆(η(E2 ))4 = 30(η(E2 ))2 (η(H))2 − 6(η(E2 ))4 1 + |H|2 + (η(E2 ))2
16
and, therefore, ∆f = 0, which means that the function f is harmonic.
Now, let us assume that our surface Σ2 is complete and has non-negative Gauss-
ian curvature K. Then, since f is a bounded harmonic function on a parabolic
space, it follows that f is constant, which implies that η(E2 ) is constant. As we
have already seen, E2 (η(E2 )) = (3/2)η(H), and this leads to η(H) = 0. From here,
we get h∇E1 E1 , E2 i = −η(H)/(2η(E2 )) = 0, i.e., ∇E1 E1 = ∇E1 E2 = 0. Moreover,
since
c−1
E2 (η(H)) = −η(E2 )(1 + hAH E2 , E2 i) = −η(E2 ) 1 + |H|2 + (η(E2 ))2 ,
16
one obtains
c−1
(1.62) hAH E2 , E2 i) = |H|2 + (η(E2 ))2 = −1,
16
that implies that c < 1.
Theorem 1.68 ([78]). There are no anti-invariant complete non-minimal non-

pseudo-umbilical pmc surfaces with non-negative Gaussian curvature in a Sasakian
space form N 2n+1 (c), with c 6= 1. In particular, there are no anti-invariant non-
minimal pmc 2-spheres in N 2n+1 (c).
Proof. From expression (1.38) of RN , we get that

c+3 c−1
(1.63) hRN (E1 , E2 )E1 , E2 i = − + (η(E2 ))2 .
4 4
Using the fact that ∇E1 = ∇E2 = 0 and (1.55), we have

RN (E1 , E2 )E1 = ∇N N N N N
E1 ∇E2 E1 − ∇E2 ∇E1 E1 − ∇[E1 ,E2 ] E1
N
2
= ∇E1 − ϕE1 − ∇N N
E2 ∇E1 E1
η(E2 )
2
=− (ϕσ(E1 , E1 ) + ξ) − ∇N N
E2 ∇E1 E1
η(E2 )
and then, also using (1.56),
2
hRN (E1 , E2 )E1 , E2 i = −2 + hσ(E1 , E1 ), ϕE2 i − h∇N N
E2 ∇E1 E1 , E2 i
η(E2 )
2 2
= −2 + η(E2 ) − + hσ(E1 , E1 ), ∇NE2 E2 i
η(E2 ) η(E2 )
4
=− + hσ(E1 , E1 ), σ(E2 , E2 )i.
(η(E2 ))2
This, together with (1.63), gives
c+3 c−1 4
hσ(E1 , E1 ), σ(E2 , E2 )i = − + (η(E2 ))2 + ,
4 4 (η(E2 ))2
and, therefore, from (1.62), one obtains
|σ(E2 , E2 )|2 = 2hH, σ(E2 , E2 )i − hσ(E1 , E1 ), σ(E2 , E2 )i
c+3 c−1 4
= −2 + − (η(E2 ))2 −
4 4 (η(E2 ))2
1
= ((1 − c)(η(E2 ))4 + (c − 5)(η(E2 ))2 − 16).
4(η(E2 ))2
Finally, since c < 1 and (η(E2 ))2 < 1, it can be easily verified that (1 − c)(η(E2 ))4 +
(c − 5)(η(E2 ))2 − 16 < 0, which is a contradiction.
CHAPTER 2
Simons Type Formulas and Applications. Surfaces with

Parallel Mean Curvature and Finite Total Curvature
1. Introduction
In the first part of this chapter, we prove three Simons type formulas for pmc
submanifolds in M n (c) × R, where M n (c) is an n-dimensional space form, i.e., a
simply-connected manifold with constant sectional curvature c. Thus, we compute
the Laplacians of the squared norm of AV , where A is the shape operator and V is
a parallel (in the normal bundle) isoperimetric normal vector field (Theorem 2.6),
of the squared norm of the second fundamental form (Theorem 2.7), and also of
the squared norm of a certain operator S defined on pmc surfaces and related to a
holomorphic differential (the Abresch-Rosenberg differential, when n = 2) (Theorem
2.23). Then we use these equations to study the geometry of such submanifolds and
obtain gap theorems (Theorems 2.12, 2.15, and 2.17-2.20). In order to prove these
results we also often use a formula for the Laplacian of the norm of the tangent part
T of the unit vector field ξ tangent to R (Proposition 2.9).
Next, we consider pmc surfaces with finite total curvature in M n (c) × R and
find some compactness properties (Theorems 2.31 and 2.33, Corollaries 2.32 and
2.34). The main tool employed in this study is a result on the behavior of the
function u = |S|, where S is the operator mentioned above. Thus, we prove that,
under certain hypotheses (that in the case when the codimension is 1, reduce only
to the finiteness of total curvature), this function u goes to zero uniformly at infinity
(Theorem 2.28).
In the last section, we present a classification result for pmc helix surfaces in
M n (c) × R, i.e., those pmc surfaces whose tangent spaces make a constant angle
with ξ (Theorem 2.37).
2. Preliminaries
Let M n (c)
be an n-dimensional space form with constant sectional curvature c.
Thus, M (c) will be the sphere Sn (c), the Euclidean space, or the hyperbolic space
n
Hn (c), as c > 0, c = 0, or c < 0, respectively.

Now, let us consider the product manifold M̄ = M n (c) × R. Henceforth, ∇ ¯ will
denote the Levi-Civita connection on M̄ .
The expression of the curvature tensor R̄ of M̄ = M n (c) × R is given by
hR̄(X, Y )Z, W i = c{hdπY, dπZihdπX, dπW i − hdπX, dπZihdπY, dπW i},
where π : M̄ = M n (c) × R → M n (c) is the projection map (see [6]). One obtains
(2.1) R̄(X, Y )Z =c{hY, ZiX − hX, ZiY − hY, ξihZ, ξiX + hX, ξihZ, ξiY
+ hX, ZihY, ξiξ − hY, ZihX, ξiξ},
39
40 Simons Type Formulas and Applications
where ξ is the unit vector field tangent to R.

Let Σm be an m-dimensional submanifold of M̄ . As in the previous chapter,
we will denote by ∇, σ, and A the Levi-Civita connection, the second fundamental
form, and the shape operator of Σm , respectively, given by equations (1.1) and (1.2).
Also, ∇⊥ is the connection in the normal bundle.
From the equation of Gauss (1.3), we obtain the expression of the curvature
tensor of Σm
(2.2) R(X, Y )Z =c{hY, ZiX − hX, ZiY − hY, T ihZ, T iX + hX, T ihZ, T iY
+ hX, ZihY, T iT − hY, ZihX, T iT }
n+1
X
+ {hAα Y, ZiAα X − hAα X, ZiAα Y },
α=m+1
where T is the component of ξ tangent to Σm , Aα = AEα , and {Eα }n+1 α=m+1 is a local
orthonormal frame field in the normal bundle.
We end this section recalling the following results that will be used later on.
Lemma 2.1 ([37]). Let a1 , . . . , am , where m > 1, and b be real numbers such
that
X m 2 m
X
(2.3) ai ≥ (n − 1) a2i + b.
i=1 i=1
Then, for all i 6= j, we have
b
(2.4) 2ai aj ≥ .
m−1
Moreover, if the inequality (2.3) is strict, then so are the inequalities (2.4).
Lemma 2.2 ([97]). Let A1 , . . . , Ap , where p ≥ 2, be symmetric m × m matrices.
Then
p p
X 3 X 2
{N (Aα Aβ − Aβ Aα ) + (trace(Aα Aβ ))2 } ≤ N (Aα ) ,
2
α,β=1 α=1
where N (A) = trace(At A). Equality holds if and only if either

(1) A1 = . . . = Ap = 0; or
(2) only two matrices Aα0 and Aβ0 are different from the null m × m matrix.
Moreover, in this case, N (Aα0 ) = N (Aβ0 ) = L and there exists an orthog-
onal matrix T such that
   
1 0 0 ... 0 0 1 0 ... 0
r  0 −1 0 . . . 0  r  1 0 0 ... 0 
t L  0 0 0 ... 0 

t L 0 0 0 ... 0 

T Aα0 T =  , T Aβ0 T = .
2  .. .. .. .. ..  2  .. .. .. .. .. 
 
 . . . . .   . . . . . 
0 0 0 ... 0 0 0 0 ... 0
PmLemma 2.3 ([4,Pm116]). Let ai , where i ∈ {1, . . . , m}, be real numbers such that
a = 0 and a 2 = b2 , where b = constant ≥ 0. Then
i=1 i i=1 i
m
m−2 3
X m−2
−p b ≤ a3i ≤ p b3 ,
m(m − 1) i=1
m(m − 1)
Simons Type Formulas and Applications 41
and equality holds in the right-hand (left-hand) side if and only if (m − 1) of the
numbers ai are non-positive and equal ((m − 1) of the ai ’s are non-negative and
equal).
Theorem 2.4 (Omori-Yau Maximum Principle, [137]). Let Σm be a complete
Riemannian manifold with Ricci curvature bounded from below. Then for any smooth
function u ∈ C 2 (Σm ) with supΣm u < +∞ there exists a sequence of points {pk }k∈N
in Σm satisfying
1 1
lim u(pk ) = sup u, |∇u|(pk ) < and ∆u(pk ) < .
k→∞ Σm k k
3. Simons type formulas

Let Σm be an m-dimensional submanifold of M̄ = M n (c) × R, with mean cur-
vature vector field H. We shall compute the Laplacian of the squared norm of AV ,
where V is a normal vector field to the submanifold such that V is parallel in the
normal bundle, i.e., ∇⊥ V = 0, and trace AV = constant.
First, using Ricci equation (1.5) and (2.1), we get the following lemma.
Lemma 2.5. If U and V are normal vector fields to Σm and V is parallel in the
normal bundle, then [AV , AU ] = 0, i.e., AV commutes with AU .
Now, from the Codazzi equation (1.4), we get
hR̄(X, Y )Z, V i = h(∇X AV )Y − (∇Y AV )X, Zi,
since ∇⊥ V = 0. Therefore, using (2.1), we obtain
(2.5) (∇X AV )Y = (∇Y AV )X + chV, N i(hY, T iX − hX, T iY ),
where N is the normal part of ξ.
Next, we have the following Weitzenböck formula
1
(2.6) ∆|AU |2 = |∇AU |2 + htrace ∇2 AU , AU i,
2
for any normal vector field U , where we extended the metric h, i to the tensor space
in the standard way.
The second term in the right hand side of (2.6) can be calculated by using a
method introduced in [112].
Let us consider
C(X, Y ) = (∇2 AV )(X, Y ) = ∇X (∇Y AV ) − ∇∇X Y AV ,
and note that we have the following Ricci commutation formula
(2.7) C(X, Y ) = C(Y, X) + [R(X, Y ), AV ].
Next, consider an orthonormal basis {ei }m m m
i=1 in Tp Σ , p ∈ Σ , and extend vectors
ei to vector fields Ei in a neighborhood of p such that {Ei } is a geodesic frame field
around p, and let us denote X = Ek . We have
m
X
2
(trace ∇ AV )X = C(Ei , Ei )X.
i=1
Using equation (2.5), we get, at p,

C(Ei , X)Ei = ∇Ei ((∇X AV )Ei )
= ∇Ei ((∇Ei AV )X) + c∇Ei (hV, N i(hEi , T iX − hX, T iEi ))
and then
(2.8) C(Ei , X)Ei =C(Ei , Ei )X − chAV Ei , T i(hEi , T iX − hX, T iEi )
+ chV, N i(hAN Ei , Ei iX − hAN X, Ei iEi ),
where we used σ(Ei , T ) = −∇⊥ Ei N and ∇Ei T = AN Ei , which follow from the fact
¯
that ξ is parallel, i.e., ∇ξ = 0.
We also have, at p,
(2.9) C(X, Ei )Ei = ∇X ((∇Ei AV )Ei ),
and, from (2.7), (2.8), and (2.9), we get, also at p,
C(Ei , Ei )X =∇X ((∇Ei AV )Ei ) + [R(Ei , X), AV ]Ei
+ chAV Ei , T i(hEi , T iX − hX, T iEi )
− chV, N i(hAN Ei , Ei iX − hAN X, Ei iEi ).
Since ∇Ei AV is symmetric, from (2.5), one obtains
m
X m
X m
X
h (∇Ei AV )Ei , Zi = hEi , (∇Ei AV )Zi = hEi , (∇Z AV )Ei i
i=1 i=1 i=1
m
X
+ chV, N i hEi , hZ, T iEi − hEi , T iZi
i=1
= trace(∇Z AV ) + c(m − 1)hV, N ihT, Zi
=Z(trace AV ) + c(m − 1)hV, N ihT, Zi
=c(m − 1)hV, N ihT, Zi,
for any vector Z tangent to Σm , since trace AV = constant.
From the Gauss equation (2.2) of the submanifold Σm , and Lemma 2.5, we get,
after a straightforward computation,
m
X
R(Ei , X)AV Ei =c{AV X − (trace AV )X + (trace AV )hX, T iT
i=1
− hAV X, T iT − hX, T iAV T + hAV T, T iX}
n+1
X
+ {AV A2α X − (trace(AV Aα ))Aα X},
α=m+1
and
m
X
AV R(Ei , X)Ei = − c{(m − 1 − |T |2 )AV X − (m − 2)hX, T iAV T }
i=1
n+1
X
+ {AV A2α X − (trace Aα )AV Aα X}.
α=m+1
Therefore, we have
m
X
(trace ∇2 AV )X = C(Ei , Ei )X
i=1
Xm
= [R(Ei , X), AV ]Ei
i=1
+ c{mhV, N iAN X − (m − 1)hAV X, T iT + hAV T, T iX
− hX, T iAV T − mhH, N ihV, N iX}
=c{(m − |T |2 )AV X + 2hAV T, T iX − mhAV X, T iT
− mhX, T iAV T + mhV, N iAN X − mhH, N ihV, N iX
− (trace AV )X + (trace AV )hX, T iT }
n+1
X
+ {(trace Aα )AV Aα X − (trace(AV Aα ))Aα X}
α=m+1
and then
m
X
htrace ∇2 AV , AV i = h(trace ∇2 AV )Ei , AV Ei i
i=1
=c{(m − |T |2 )|AV |2 − 2m|AV T |2 + 3(trace AV )hAV T, T i
+ m(trace(AN AV ))hV, N i − (trace AV )2
− m(trace AV )hH, N ihV, N i}
n+1
X
+ {(trace Aα )(trace(A2V Aα )) − (trace(AV Aα ))2 }.
α=m+1
Thus, from (2.6), we obtain the first main result of this section.
Theorem 2.6 ([72]). Let Σm be a submanifold of M n (c) × R, with mean curva-

ture vector field H and shape operator A. If V is a normal vector field, parallel in
the normal bundle, with trace AV = constant, then
1
∆|AV |2 =|∇AV |2 + c{(m − |T |2 )|AV |2 − 2m|AV T |2
2
+ 3(trace AV )hAV T, T i + m(trace(AN AV ))hV, N i − (trace AV )2
− m(trace AV )hH, N ihV, N i}
n+1
X
+ {(trace Aα )(trace(A2V Aα )) − (trace(AV Aα ))2 },
α=m+1
where {Eα }n+1

α=m+1 is a local orthonormal frame field in the normal bundle.
In the second part of this section, let us consider pmc submanifolds Σm in

M̄ = M n (c) × R. We will compute the Laplacian of the squared norm of the second
fundamental form σ of Σm .
Consider a local orthonormal frame field {Eα }n+1

α=m+1 in the normal bundle and
then normal connection forms sαβ are determined by
n+1
X
∇⊥
X Eα = sαβ (X)Eβ
β=m+1
for any vector field X tangent to Σn and any α ∈ {m + 1, . . . , n + 1}. It is easy to

see that sαβ = −sβα and that
n+1
1 ⊥ 1 X
∇⊥
XH = ∇X (trace σ) = ∇⊥
X ((trace Aα )Eα )
m m
α=m+1
n+1 n+1
!
1 X X
= X(trace Aα ) − sαβ (X) trace Aβ Eα .
m
α=m+1 β=m+1
Therefore, the mean curvature vector field H is parallel if and only if
n+1
X
(2.10) X(trace Aα ) − sαβ (X) trace Aβ = 0, ∀α ∈ {m + 1, . . . , n + 1}.
β=m+1
In order to obtain ∆|σ|2 , we will compute ∆|Aα |2 for each index α, using the
Weitzenböck formula (2.6) and the same method as above.
First, from the Codazzi equation (1.4), we get
hR̄(X, Y )Z, Eα i =h(∇X Aα )Y − (∇Y Aα )X, Zi

n+1
X
− hsαβ (X)Aβ Y − sαβ (Y )Aβ X, Zi.
β=m+1
Therefore, using (2.1), one sees that
n+1
X
(2.11) (∇X Aα )Y =(∇Y Aα )X + (sαβ (X)Aβ Y − sαβ (Y )Aβ X)
β=m+1
+ chEα , N i(hY, T iX − hX, T iY ).
Let us again consider Cα (X, Y ) = (∇2 Aα )(X, Y ) = ∇X (∇Y Aα ) − ∇∇X Y Aα ,

a geodesic frame field {Ei } around a point p ∈ Σm , and denote X = Ek . Using
equation (2.11), we get, at p,
Cα (Ei , X)Ei =∇Ei ((∇X Aα )Ei )

n+1
!
X
=∇Ei ((∇Ei Aα )X) + ∇Ei (sαβ (X)Aβ Ei − sαβ (Ei )Aβ X)
β=m+1
+ c∇Ei (hEα , N i(hEi , T iX − hX, T iEi ))
and then, since σ(Ei , T ) = −∇⊥

Ei N and ∇Ei T = AN Ei ,
(2.12) Cα (Ei , X)Ei =Cα (Ei , Ei )X
n+1
!
X
+ ∇Ei (sαβ (X)Aβ Ei − sαβ (Ei )Aβ X)
β=m+1
n+1
X
+c hsαβ (Ei )Eβ , N i(hEi , T iX − hX, T iEi )
β=m+1
− chAα Ei , T i(hEi , T iX − hX, T iEi )
+ chEα , N i(hAN Ei , Ei iX − hAN X, Ei iEi ).
We also have, at p, Cα (X, Ei )Ei = ∇X ((∇Ei Aα )Ei ), and then, using (2.7) and
(2.12), for any vector Z tangent to Σm , we get
Cα (Ei , Ei )X =∇X ((∇Ei Aα )Ei ) + [R(Ei , X), Aα ]Ei
n+1
!
X
− ∇Ei (sαβ (X)Aβ Ei − sαβ (Ei )Aβ X)
β=m+1
n+1
X
−c hsαβ (Ei )Eβ , N i(hEi , T iX − hX, T iEi )
β=m+1
+ chAα Ei , T i(hEi , T iX − hX, T iEi )
− chEα , N i(hAN Ei , Ei iX − hAN X, Ei iEi ).
Since ∇Ei Aα is symmetric, from (2.11), one obtains
m
X m
X m
X
(2.13) h (∇Ei Aα )Ei , Zi = hEi , (∇Ei Aα )Zi = hEi , (∇Z Aα )Ei i
i=1 i=1 i=1
m
X n+1
X
− h sαβ (Z)Aβ Ei , Ei i
i=1 β=m+1
m
X n+1
X
+ h sαβ (Ei )Aβ Ei , Zi
i=1 β=m+1
Xm
+ chEα , N i hEi , hZ, T iEi − hEi , T iZi,
i=1
which, together with (2.10), leads to
m
X n+1
X
(2.14) h (∇Ei Aα )Ei , Zi =Z(trace Aα ) − sαβ (Z) trace Aβ
i=1 β=m+1
m
X n+1
X
+h sαβ (Ei )Aβ Ei + c(m − 1)hEα , N iT, Zi
i=1 β=m+1
m
X n+1
X
=h sαβ (Ei )Aβ Ei + c(m − 1)hEα , N iT, Zi,
i=1 β=m+1
for any vector field Z tangent to Σm . P

Therefore, since (trace ∇2 Aα )X = mi=1 Cα (Ei , Ei )X, we have
m
X n+1
X
(2.15) (trace ∇2 Aα )X = {X(sαβ (Ei ))Aβ Ei + sαβ (Ei )∇X Aβ Ei
i=1 β=m+1
− Ei (sαβ (X))Aβ Ei − sαβ (X)∇Ei Aβ Ei
+ Ei (sαβ (Ei ))Aβ X + sαβ (Ei )∇Ei Aβ X}
n+1
X
+ c(m − 1)h sαβ (X)Eβ , N iT
β=m+1
m
X n+1
X
−c h sαβ (Ei )Eβ , N i(hEi , T iX − hX, T iEi )
i=1 β=m+1
+ chAα T, T iX − chX, T iAα T − cmhEα , N ihH, N iX
+ cmhEα , N iAN X − c(m − 1)hAα T, XiT
m
X
+ [R(Ei , X), Aα ]Ei .
i=1
Now, using the Ricci equation (1.5), we get, after a straightforward computation,
m
X n+1
X
(2.16) (X(sαβ (Ei ))Aβ Ei − Ei (sαβ (X))Aβ Ei )
i=1 β=m+1
n+1
X m
X n+1
X
= Aβ [Aα , Aβ ]X − sαγ (Ei )sγβ (X)Aβ Ei
β=m+1 i=1 β,γ=m+1
m
X n+1
X
+ sαγ (X)sγβ (Ei )Aβ Ei .
i=1 β,γ=m+1
From (2.11), we have
m
X n+1
X m
X n+1
X
(2.17) sαβ (Ei )∇X Aβ Ei = sαβ (Ei )(∇X Aβ )Ei
i=1 β=m+1 i=1 β=m+1
m
X n+1
X
= sαβ (Ei ){(∇Ei Aβ )X
i=1 β=m+1
− chEβ , N i(hX, T iEi − hEi , T iX)
n+1
X
− (sβγ (Ei )Aγ X − sβγ (X)Aγ Ei )}.
γ=m+1
We now use (2.14) to compute
m
X n+1
X m
X n+1
X
(2.18) sαβ (X)∇Ei Aβ Ei = sαβ (X)(∇Ei Aβ )Ei
i=1 β=m+1 i=1 β=m+1
m
X n+1
X
= sαβ (X)sβγ (Ei )Aγ Ei
i=1 β,γ=m+1
n+1
X
+ c(m − 1) hsαβ (X)Eβ , N iT.
β=m+1
From equation (2.2), we get
m
X
(2.19) [R(Ei , X), Aα ]Ei =c{(m − |T |2 )Aα X − (trace Aα )X
i=1
+ (trace Aα )hX, T iT − hAα X, T iT
− (m − 1)hX, T iAα T + hAα T, T iX}
n+1
X
+ {Aβ Aα Aβ X − (trace(Aα Aβ ))Aβ X
β=m+1
+ Aα A2β X − (trace Aβ )Aα Aβ X}.
Finally, taking into account that Ei (sαβ (Ei ))Aβ X = (∇Ei sαβ )(Ei )Aβ X and
then replacing (2.16), (2.17), (2.18), and (2.19) in (2.15), we obtain, after a long but
straightforward computation,
m n+1
(
X X
2
(2.20) htrace ∇ Aα , Aα i = 2sαβ (Ei ) trace((∇Ei Aβ )Aα )
i=1 β=m+1
n+1
X
− sαγ (Ei )sγβ (Ei ) trace(Aα Aβ )
β,γ=m+1
n+1
)
X
+ (∇Ei sαβ )(Ei ) trace(Aα Aβ )
β=m+1
+ c{(m − |T |2 )|Aα |2 − 2m|Aα T |2

+ m(trace(AN Aα ))hEα , N i
+ 3(trace Aα )hAα T, T i − (trace Aα )2
− m(trace Aα )hH, N ihEα , N i}
n+1
X
+ {(trace Aβ )(trace(A2α Aβ )) + trace[Aα , Aβ ]2
β=m+1
− (trace(Aα Aβ ))2 }.
From equation (2.6), we know that

n+1 n+1
1 1 X X
(2.21) ∆|σ|2 = ∆|Aα |2 = {|∇Aα |2 + htrace ∇2 Aα , Aα i}.
2 2
α=m+1 α=m+1
In order to estimate this Laplacian using (2.20), we note that

n+1
X
(trace Aα )hAα T, T i = mhσ(T, T ), Hi,
α=m+1
n+1
X
(trace Aα )hH, N ihEα , N i = mhH, N i2 ,
α=m+1
n+1
X n+1
X
(trace(AN Aα ))hEα , N i = |AN |2 , (trace Aα )2 = m2 |H|2 ,
α=m+1 α=m+1
and, since sαβ = −sβα , that

n+1
X
(∇Ei sαβ )(Ei ) trace(Aα Aβ ) = 0.
α,β=m+1
Next, we easily get
(∇⊥ σ)(X, Y, Z) =∇⊥

X σ(Y, Z) − σ(∇X Y, Z) − σ(Y, ∇X Z)
n+1
X n+1
X
= h(∇X Aα )Y − sαβ (X)Aβ Y, ZiEα ,
α=m+1 β=m+1
for all tangent vector fields X, Y , and Z, and then

m
X
|∇⊥ σ|2 = |(∇⊥ σ)(Ei , Ej , Ek )|2
i,j,k=1
n+1
X m
X n+1
X
= h(∇Ei Aα )Ej − sαβ (Ei )Aβ Ej ,
α=m+1 i,j=1 β=m+1
n+1
X
(∇Ei Aα )Ej − sαγ (Ei )Aγ Ej i,
γ=m+1
which means that

n+1
X n m n+1
X X
⊥ 2 2
|∇ σ| = |∇Aα | + 2sαβ (Ei ) trace((∇Ei Aβ )Aα )
α=m+1 i=1 β=m+1
n+1
X o
− sαγ (Ei )sγβ (Ei ) trace(Aα Aβ ) .
β,γ=m+1
Using (2.20) and (2.21), we can state the following theorem.

Theorem 2.7 ([76]). Let Σm be a pmc submanifold of M n (c) × R, with mean

curvature vector field H, shape operator A, and second fundamental form σ. Then
we have
n+1
1 2 ⊥ 2
n
2 2
X
∆|σ| =|∇ σ| + c (m − |T | )|σ| − 2m |Aα T |2
2
α=m+1
o
+ 3mhσ(T, T ), Hi + m|AN | − m2 hH, N i2 − m2 |H|2
2
n+1
X
+ {(trace Aβ )(trace(A2α Aβ )) + trace[Aα , Aβ ]2 − (trace(Aα Aβ ))2 },
α,β=m+1
where {Eα }n+1

α=m+1 is a local orthonormal frame field in the normal bundle.
Corollary 2.8 ([76]). If Σm is a minimal submanifold of M n (c) × R, then we

have
n+1
1 n X o
∆|σ|2 =|∇⊥ σ|2 + c (m − |T |2 )|σ|2 − 2m |Aα T |2 + m|AN |2
2
α=m+1
n+1
X
+ {trace[Aα , Aβ ]2 − (trace(Aα Aβ ))2 }.
α,β=m+1
We will end this section computing the Laplacian of the squared norm of the
tangent part T of ξ.
As above, consider a geodesic frame field {Ei } around an arbitrary point p ∈ Σm
and then, at p, we have
m m
1 X X
∆|T |2 = (h∇Ei T, ∇Ei T i + h∇Ei ∇Ei T, T i) = |AN |2 + h∇Ei AN Ei , T i
2
i=1 i=1
and, after a straightforward computation, using that ∇X AN is symmetric, ∇X T =

AN X, and ∇⊥ X N = −σ(X, T ),
m
X m
X
h∇Ei AN Ei , T i = (h∇T AN Ei , Ei i − hR(T, Ei )T, Ei i).
i=1 i=1
Since ∇Ei = 0 at p, we have

m
X m
X
h∇Ei AN Ei , T i =T (trace AN ) − hR(T, Ei )T, Ei i
i=1 i=1
m
X
= − mhσ(T, T ), Hi) − hR(T, Ei )T, Ei i,
i=1
where T (trace AN ) = mT (hH, N i) = −mhσ(T, T ), Hi.

From the Gauss equation (2.2), it follows that
m
X n+1
X
2 2
hR(T, Ei )T, Ei i = c(1 − m)|T | (1 − |T | ) + {|Aα T |2 − (trace Aα )hAα T, T i}
i=1 α=m+1
and then we get

1
∆|T |2 =|AN |2 − mhσ(T, T ), Hi + c(m − 1)|T |2 (1 − |T |2 )
2
n+1
X
− {|Aα T |2 − (trace Aα )hAα T, T i}.
α=m+1
Proposition 2.9 ([76]). Let Σm be an m-dimensional pmc submanifold in
M n (c)× R, with shape operator A. Then we have
n+1
1 X
∆|T |2 = |AN |2 + c(m − 1)|T |2 (1 − |T |2 ) − |Aα T |2 .
2
α=m+1
4. Gap theorems for submanifolds with parallel mean curvature in

M n (c) × R
In this section, we will present some applications of the Simons type formulas
obtained in Theorems 2.6 and 2.7 and of Proposition 2.9. We note that other
applications of these results can be found in the second part of this thesis, the one
devoted to the study of biharmonic submanifolds.
First, let Σ2 be a pmc surface in a product space M 3 (c)×R, with mean curvature
vector field H 6= 0 and Gaussian curvature K.
Let us consider the orthonormal frame field {E3 = H/|H|, E4 } in the normal
bundle and denote by φ3 = A3 − |H| I and φ4 = A4 . The normal part N of ξ can be
written as N = ν3 E3 + ν4 E4 , where ν3 = hξ, E3 i and ν4 = hξ, E4 i.
Since H is parallel, it follows that also E4 is parallel in the normal bundle. From
Theorem 2.6, we get
1
(2.22) ∆|φH |2 =|∇φH |2 + {c(2 − 3|T |2 ) + 4|H|2 − |σ|2 }|φH |2
2
2c
− 2c|H|hφH T, T i + hH, N i trace(AN φH ).
|H|
and
1
(2.23) ∆|φ4 |2 = |∇φ4 |2 + {c(2 − 3|T |2 ) + 4|H|2 − |σ|2 }|φ4 |2 + 2cν4 trace(AN φ4 ).
2
Next, let φ be the traceless part of the second fundamental form σ of the surface.
We have |φ|2 = |φ3 |2 + |φ4 |2 = |σ|2 − 2|H|2 and then (2.22) and (2.23) lead to the
following result.
Proposition 2.10 ([75]). If Σ2 is a non-minimal pmc surface in M 3 (c) × R
and φ is the traceless part of its second fundamental form, then
1
(2.24) ∆|φ|2 =|∇φ3 |2 + |∇φ4 |2 − |φ|4 + {c(2 − 3|T |2 ) + 2|H|2 }|φ|2
2
− 2chφ(T, T ), Hi + 2c|ν3 φ3 + ν4 φ4 |2 ,
or, equivalently,
1
∆|φ|2 =|∇φ3 |2 + |∇φ4 |2 − |φ|4 + {c(2 − 3|T |2 ) + 2|H|2 }|φ|2
2
− 2chφ(T, T ), Hi + 2c|AN |2 − 4chH, N i2 .
Remark 2.11. Proposition 2.10 can also be directly obtained from Theorem 2.7.
Now, from (2.24) and Proposition 2.9, one obtains
1 c
(2.25) ∆(|φ|2 − c|T |2 ) =|∇φ3 |2 + |∇φ4 |2 + {−|φ|2 + (4 − 5|T |2 ) + 2|H|2 }|φ|2
2 2
+ c|AN |2 − 4chH, N i2 + c|T |2 |H|2 − c2 |T |2 (1 − |T |2 )
and, we will use this equation, to prove our next result.
Theorem 2.12 ([75]). Let Σ2 be a complete non-minimal pmc surface in M̄ =
M 3 (c) × R, with c > 0. Assume that
(i) |φ|2 ≤ 2|H|2 + 2c − (5c/2)|T |2 , and
(ii) (a) |T | = 0; or
(b) |T |2 > 2/3 and |H|2 ≥ c|T |2 (1 − |T |2 )/(3|T |2 − 2).
Then either
(1) |φ|2 = 0 and Σ2 is a round sphere in M 3 (c); or p
(2) |φ|2 = 2|H|2 +2c and Σ2 is a torus S1r1 ×S1r2 in M 3 (c), where r2 = 1/c − r12
and r12 6= 1/(2c).
Proof. When |T |2 = 0 on the surface, it is easy to see that (2.25) implies
1
∆(|φ|2 − c|T |2 ) ≥ {−|φ|2 + 2c + 2|H|2 }|φ|2 ≥ 0.
2
Next, let us assume that |T |2 > 2/3 and |H|2 ≥ c|T |2 (1−|T |2 )/(3|T |2 −2). Since
|AN |2 − 2hH, N i2 = |ν3 φ3 + ν4 φ4 |2 ≥ 0
and, from the Schwarz inequality, we have
(2.26) hH, N i2 ≤ |N |2 |H|2 = (1 − |T |2 )|H|2 ,
then, from (2.25), it follows that
1 c
∆(|φ|2 − c|T |2 ) ≥{−|φ|2 + (4 − 5|T |2 ) + 2|H|2 }|φ|2
2 2
+ c(3|T | − 2)|H|2 − c2 |T |2 (1 − |T |2 )
2
≥0.
The Gaussian curvature of the surface satisfies
1
2K = 2c(1 − |T |2 ) + 2|H|2 − |φ|2 ≥ c|T |2 ≥ 0,
2
which means that Σ is a parabolic space. Now, since |φ|2 − c|T |2 is a bounded
2
subharmonic function, it follows that it is constant. Therefore, either |φ|2 = 0 or

|φ|2 = 2|H|2 + 2c − (5c/2)|T |2 .
If |φ| = 0, the mean curvature vector field H is an umbilical direction and this
implies that Σ2 is a totally umbilical surface in M 3 (c) ([6, Lemma 3]). Therefore,
Σ2 is either a horosphere or a round sphere. Since K = c + |H|2 in this case, Σ2
cannot be flat, and, therefore, it is a round sphere.
If |φ|2 = 2|H|2 +2c−(5c/2)|T |2 , from (2.25), we have that ν3 φ3 +ν4 φ4 = 0, which
means that AN = hH, N i I, and the equality holds in (2.26), i.e., either N = ν3 H or
ξ is tangent to the surface. If ξ is tangent to Σ2 , from (2.25) and the hypothesis, it
follows that Σ2 is a minimal surface, which is a contradiction. Hence, N = ν3 H and
we get AH = |H|2 I, which, also using [6, Lemma 3.1], implies that the surface lies
in M 3 (c) and then it is flat. We use a similar argument to that in [4, Theorem 1.5]
to conclude (see also [123]).
In [5] the authors introduced a holomorphic differential, defined on pmc surfaces

in M n (c) × R, that is just the Abresch-Rosenberg differential when n = 2 (see [1]).
This holomorphic differential is the (2, 0)-part of the quadratic form Q given by
Q(X, Y ) = 2hσ(X, Y ), Hi − chX, ξihY, ξi.
Using this result and Proposition 2.9, we can characterize certain pmc 2-spheres Σ2
immersed in M n (c)×R, whose second fundamental form satisfies a certain condition.
Theorem 2.13 ([75]). Let Σ2 be a pmc 2-sphere in M n (c) × R, such that
(1) |T |2 = 0 or |T |2 ≥ 2/3 and |σ|2 ≤ c(2 − 3|T |2 ), if c < 0; or
(2) |T |2 ≤ 2/3 and |σ|2 ≤ c(2 − 3|T |2 ), if c > 0.
Then, Σ2 is either a minimal surface in a totally umbilical hypersurface of M n (c),
or a standard sphere in M 3 (c).
Proof. If ξ is orthogonal to the surface in an open connected subset, then this
subset lies in M n (c), and by analyticity, it follows that Σ2 lies in M n (c). In this
case we use [136, Theorem 4] to conclude.
Next, let us assume that we are not in the previous case. Then, we can consider a
local orthonormal frame field {E1 = T /|T |, E2 } on the surface. Since Σ2 is a sphere,
then the (2, 0)-part of Q vanishes. This means that Q(E1 , E1 ) = Q(E2 , E2 ) and
Q(E1 , E2 ) = 0. From Q(E1 , E1 ) = Q(E2 , E2 ), we get 2hσ(E1 , E1 ) − σ(E2 , E2 ), Hi =
c|T |2 . We then have
1 1
hφ(T, T ), Hi = hσ(T, T ), Hi − |T |2 |H|2 = |T |2 hσ(E1 , E1 ) − σ(E2 , E2 ), Hi = c|T |4 ,
2 4
and Proposition 2.9 gives
1 1
∆|T |2 = |AN |2 + |T |2 (−|σ|2 + c(2 − 3|T |2 )) ≥ 0.
2 2
Since Σ2 is a sphere and |T |2 is a bounded subharmonic function, it follows that
|T |2is constant, and then that |AN |2 = 0 and |T |2 (−|σ|2 + c(2 − 3|T |2 )) = 0. Since
AN = 0 and ξ is parallel, we obtain that ∇T = 0, which implies K = 0. Since our
surface is a sphere, this is a contradiction. Therefore, Σ2 lies in M n (c) and, again
using [136, Theorem 4], we come to the conclusion (see also [6, Theorem 2]).
Next, we consider pmc submanifolds Σm in a M n (c) × R and present other

applications of the Simons type formulas in this general case.
Proposition 2.14 ([76]). Let Σm be a complete pmc submanifold in M n (c) × R
with second fundamental form σ. If
sup{|σ|2 + c(m − 1)|T |2 } < max{0, c(m − 1)},
Σm
then either
(1) Σm lies in M n (c), if c > 0; or
(2) Σm is a vertical cylinder π −1 (Σm−1 ) over a pmc submanifold Σm−1 in the
space form M n (c), if c < 0.
Proof. Let us consider first the case when c > 0. Then, from Theorem 2.9, we
have that
n+1
1 2 2 2 2
X
∆|T | = |AN | + c(m − 1)|T | (1 − |T | ) − |Aα T |2
2
α=m+1
2 2 2
≥ |T | (c(m − 1)(1 − |T | ) − |σ| )
≥ 0.
Next, let us consider a local orthonormal frame field {Ei }m m
i=1 on Σ , X a unit
n+1
tangent vector field, and {Eα }α=m+1 an orthonormal frame field in the normal
bundle. From equation (2.2), we get the expression of the Ricci curvature of our
submanifold
Xm
Ricci X = hR(Ei , X)X, Ei i
i=1
m n
X
= c(|X|2 − hX, Ei i2 − hX, T i2 + 2hX, T ihT, Ei ihX, Ei i
i=1
n+1
X o
2 2 2
− hT, Ei i |X| ) + (hAα Ei , Ei ihAα X, Xi − hAα X, Ei i )
α=m+1
n+1
X
=c(m − 1 − |T |2 − (m − 2)hX, T i2 ) + mhAH X, Xi − |Aα X|2 .
α=m+1
It follows that
n+1
X
Ricci X ≥ c(m − 1)(1 − |T |2 ) − m|AH X| − |Aα |2
α=m+1
2
≥ −m|AH | − |σ| .
Since |σ| is bounded by hypothesis, we can see that the Ricci curvature is bounded
from below and then that the Omori-Yau Maximum Principle holds on Σm .
Therefore, we can use Theorem 2.4 with u = |T |2 . It follows that there exists a
sequence of points {pk }k∈N ⊂ Σm satisfying
1
lim |T |2 (pk ) = sup |T |2 and ∆|T |2 (pk ) < .
k→∞ Σ m k
Since supΣm {|σ|2 + c(m − 1)|T |2 } < c(m − 1), it follows that 0 = limk→∞ |T |2 (pk ) =
supΣm |T |2 , which means that T = 0, i.e., Σm lies in M n (c).
When c < 0, we come to the conclusion in the same way as above, using the
facts that
n+1
1 1 X
∆|N |2 = − ∆|T |2 = −|AN |2 − c(m − 1)|T |2 (1 − |T |2 ) + |Aα T |2
2 2
α=m+1
2 2 2
≥ |N | (−|σ| − c(m − 1)|T | )
≥0
and
Ricci X ≥ c(m − 1) − m|AH | − |σ|2 ,
and then applying Theorem 2.4 to function u = |N |2 .

For minimal submanifolds in M n (c)×R, with c > 0, we have the following result.
Theorem 2.15 ([76]). Let Σm be a complete minimal submanifold in M n (c)×R,
with c > 0. If
sup{3|σ|2 + 2c(2m + 1)|T |2 } < 2cm,
Σm
then Σm is a totally geodesic submanifold in M n (c).
Proof. From Corollary 2.8, and since Schwarz inequality implies |Aα T |2 ≤
|T |2 |A 2 2
α | , using |AN | ≥ 0 and Lemma 2.2, we obtain
∆|σ|2 ≥ −(3|σ|2 + 2c((2m + 1)|T |2 − m))|σ|2 ≥ 0.

As we have seen, since |σ| is bounded, the Ricci curvature of Σm is bounded from
below, and then we can apply the Omori-Yau Maximum Principle to function u =
|σ|2 . One obtains that there exists a sequence of points {pk }k∈N ⊂ Σm satisfying
1
lim |σ|2 (pk ) = sup |σ|2 and ∆|σ|2 (pk ) < ,
k→∞ Σ m k
from where it follows that 0 = limk→∞ |σ|2 (pk ) = supΣm |σ|2 , which means that
σ = 0. Moreover, AN = 0 and then the hypothesis imply that |T |2 = constant < 1.
From Proposition 2.9, it follows that T = 0, which means that our submanifold is
totally geodesic in M n (c).
We now recall the following lemma, that will be used later on.
Lemma 2.16 ([76]). Let Σm be a non-minimal pmc submanifold in M n (c) × R
with mean curvature vector field H. Then we have
∆hH, N i = −c(m − 1)|T |2 hH, N i − trace(AH AN ).
Our next results are similar to those obtained in [8, 39] in the case of pmc
submanifolds in spheres and Euclidean space and, again as in the above cited papers,
their proofs rely on the use of Theorem 2.7 and Lemmas 2.1 and 2.2.
Theorem 2.17 ([76]). Let Σm be a complete non-minimal pmc submanifold in
M n (c)
× R, n > m ≥ 3, c > 0, with mean curvature vector field H and second
fundamental form σ. If the angle between H and ξ is constant and
2c(2m + 1) 2 m2
(2.27) |σ|2 + |T | ≤ 2c + |H|2 ,
m m−1
then Σm is a totally umbilical cmc hypersurface in M m+1 (c).
Proof. We shall prove first that Σm actually lies in a space form M m+1 (c). In
order to do that, we will show that, if {Em+1 , . . . , En+1 } is a local orthonormal frame
field in the normal bundle with Em+1 = H/|H|, then Aα = 0, for any α > m + 1.
From Theorem 2.6, we have
(2.28)
1
∆|Am+1 |2 =|∇Am+1 |2 + c{(m − |T |2 )|Am+1 |2 − 2m|Am+1 T |2 + 3mhσ(T, T ), Hi
2
+ m(trace(AN Am+1 ))hEm+1 , N i − m2 hH, N i2 − m2 |H|2 }
n+1
X
+ {(trace Aα )(trace(A2m+1 Aα )) − (trace(Am+1 Aα ))2 }.
α=m+1
Next, we define a function |A|2 on Σm by |A|2 = 2

P
α>m+1 |Aα | and, using
(2.28), we obtain, from Theorem 2.7, that
(
1 2
X
∗
X
(2.29) ∆|A| = |∇ Aα | + c (m − |T |2 )|A|2 − 2m
2
|Aα T |2
2
α>m+1 α>m+1
)
+ m|AN |2 − m(trace(AN Am+1 ))hEm+1 , N i
X
+ {(trace Am+1 )(trace(A2α Am+1 )) − (trace(Aα Am+1 ))2 }
α>m+1
X
+ {trace[Aα , Aβ ]2 − (trace(Aα Aβ ))2 },
α,β>m+1
where ∇∗ is the sum of the tangent and normal connections and

X
∇∗X Aα = ∇X Aα − sαβ (X)Aβ .
β>m+1
The Schwarz inequality implies that

X X
(2.30) − |Aα T |2 ≥ −|T |2 |Aα |2 = −|T |2 |A|2 .
α>m+1 α>m+1
From Lemma 2.16, since hH, N i = constant, we have

(2.31) |AN |2 −(trace(AN Am+1 ))hEm+1 , N i = |AN |2 +c(m−1)|T |2 hEm+1 , N i2 ≥ 0.
Since trace[Aα , Aβ ]2 = −N (Aα Aβ − Aβ Aα ), using Lemma 2.2, we get
(2.32)
!2
X
2 2 3 X
2 3
{trace[Aα , Aβ ] − (trace(Aα Aβ )) } ≥ − |Aα | = − |A|4 .
2 2
α,β>m+1 α>m+1
Next, we shall evaluate the term

X
{(trace Am+1 )(trace(A2α Am+1 )) − (trace(Aα Am+1 ))2 }
α>m+1
in (2.29). In order to do that, we first note that, since [Am+1 , Aα ] = 0, the matrices
Am+1 and Aα can be diagonalized simultaneously, for each α > m + 1. Let λi and
λαi , i ∈ {1, . . . , m}, be the eigenvalues of Am+1 and Aα , respectively. Then, for each
α > m + 1, we have
(2.33) (trace Am+1 )(trace(A2α Am+1 )) − (trace(Aα Am+1 ))2
m
! m ! m
! m !
X X X X
= λi λj (λαj )2 − λi λαi λj λαj
i=1 j=1 i=1 j=1
m
1 X
= λi λj (λαi − λαj )2 .
2
i,j=1
Our hypothesis (2.27) can be written as

2c(2m + 1) 2
(m|H|)2 ≥ (m − 1)|Am+1 |2 + (m − 1) |A|2 + |T | − 2c
m
which means that

m
!2 m
X X 2c(2m + 1) 2
(2.34) λi ≥ (m − 1) (λi )2 + (m − 1) |A|2 + |T | − 2c .
m
i=1 i=1
Thus, from Lemma 2.1, it follows that

1 c(2m + 1) 2
(2.35) λi λj ≥ |A|2 + |T | − c,
2 m
for i 6= j, and then
m m
1 X 1 1 2 c(2m + 1) 2 X
(2.36) λi λj (λαi − λαj )2 ≥ |A| + |T | − c (λαi − λαj )2
2 2 2 m
i,j=1 i,j=1
1 m
c(2m + 1) 2 X
= |A|2 + |T | − c ((λαi )2 − λαi λαj )
2 m
i,j=1
m
= |A|2 + c(2m + 1)|T |2 − cm |Aα |2
2
m
!2
1 c(2m + 1) X
− |A|2 + |T |2 − c λαi
2 m
i=1
m
= |A|2 + c(2m + 1)|T |2 − cm |Aα |2 .
2
Replacing in (2.33), we get
X
(2.37) {(trace Am+1 )(trace(A2α Am+1 )) − (trace(Aα Am+1 ))2 }
α>m+1
m
≥ |A|2 + c(2m + 1)|T |2 − cm |A|2 .
2
Now, from (2.29), (2.30), (2.31), (2.32), and (2.37), one obtains
1 m−3 4
(2.38) ∆|A|2 ≥ |A| .
2 2
As we have seen in Proposition 2.14, the fact that |σ| is bounded implies that the
Ricci curvature of Σm is bounded from below. Therefore we can apply Theorem 2.4
to function u = |A|2 , and we get that there exists a sequence of points {pk }k∈N ⊂ Σm
satisfying
1
lim |A|2 (pk ) = sup |A|2 and ∆|A|2 (pk ) < .
k→∞ Σm k
From the inequality (2.38), it follows that
0 = lim ∆|A|2 (pk ) ≥ (m − 3) sup |A|2 ≥ 0,
k→∞ Σm
i.e., (m − 3) supΣm |A|2

= 0. Therefore, we get that m = 3 or |A|2 = 0.
Next, we shall split our study in two cases as m ≥ 4 or m = 3.
Case I: m ≥ 4. In this case, we have |A|2 = 0, and then Aα = 0 for all α > m+1.
Moreover, all inequalities (2.30), (2.31), (2.32), and (2.37) become equalities. Since
AN = 0, we get that |T |2 is constant and that hH, N i = 0. We also have
0 = X(hH, N i) = hH, ∇⊥
X N i = −|H|hEm+1 , σ(T, X)i = −|H|hAm+1 T, Xi,
for any tangent vector field X. Therefore, from Proposition 2.9, it follows that
0 = c(m − 1)|T |2 (1 − |T |2 ),
i.e., either T = 0, or T = ±ξ.
If T = ±ξ, then Σm is a vertical cylinder π −1 (Σm−1 ) over a pmc submanifold
Σ m−1 in M n (c) with second fundamental form σ0 , satisfying |σ0 | = |σ|, and mean
curvature vector field H0 = (m/(m − 1))H. Then, condition (2.27) can be rewritten
as
2c(m + 1)
|σ0 |2 ≤ (m − 1)|H0 |2 − < (m − 1)|H0 |2 ,
m
which is a contradiction, since the squared norm of the traceless part φ0 of σ0 satisfies
0 ≤ |φ0 |2 = |σ0 |2 − (m − 1)|H0 |2 .
Hence, we have T = 0, i.e. ξ is normal to Σm . Since Aα = 0 for all α > m + 1, it
follows that the subbundle L = span{σ} = span{H} of the normal bundle is parallel,
i.e., ∇⊥ V ∈ L for all V ∈ L. Now, one can see that T Σm ⊕L is parallel, orthogonal to
ξ, and invariant by the curvature tensor R̄. Using [54, Theorem 2], all these lead to
the conclusion that Σm lies in an m + 1-dimensional totally geodesic submanifold of
M n (c) × R, which is also orthogonal to ξ, i.e., Σm is a cmc hypersurface in M m+1 (c).
Case II: m = 3. We shall prove that |A|2 = 0 in this situation too, which
means, as we have seen above, that Σ3 is a cmc hypersurface in M 4 (c).
Our hypothesis (2.27) implies that the sequence {σij α (p )}
k k∈N , where
α
σij = hσ(Ei , Ej ), Eα i,
is bounded for all i, j, and α. We also know that the sequence {|T |2 (pk )}k∈N is
bounded. Therefore, there exits a subsequence {pkr }kr ∈N of {pk }k∈N such that the
following limits exit
α α
σ̄ij = lim σij (pkr ) < ∞ and |T̄ |2 = lim |T |2 (pkr ) < ∞,
kr →∞ kr →∞
and we denote by
Āα = lim Aα (pkr )
kr →∞
the matrix with entries σ̄ij α.
From limkr →∞ ∆|A|2 (pkr ) = 0, it follows that, when we take the limit after
kr → ∞, all inequalities (2.30), (2.31), (2.32), and (2.37) become equalities. Then,
from (2.32) and (2.37) we obtain
(2.39)
!2 !2
X 3 X 3
{trace[Āα , Āβ ]2 − (trace(Āα Āβ ))2 } = − |Āα |2 =− sup |A|2
2 2 Σ3
α,β>4 α>4
and
X
(2.40) {(trace Ā4 )(trace(Ā2α Ā4 )) − (trace(Āα Ā4 ))2 }
α>4
3 X X
= |Āα |2 + 7c|T̄ |2 − 3c |Āα |2
2
α>4 α>4
3
= sup |A|2 + 7c|T̄ |2 − 3c sup |A|2 ,
2 Σ3 Σ3
respectively. From (2.39) and Lemma 2.2, it follows that either
(1) Ā5 = . . . = Ān+1 = 0; or

(2) only two matrices Āα0 and Āβ0 are different from the null m × m matrix,
|Āα0 |2 = |Āβ0 |2 = L, and there exists an orthogonal matrix T such that
r  1 0 0  r  0 1 0 
L L
(2.41) T t Āα0 T = 0 −1 0  , T t Āβ0 T = 1 0 0 .
2 0 0 0 2 0 0 0
In the first case, one obtains
X
0= |Āα |2 = sup |A|2 ,
α>4 Σ3
which means that |A|2 = 0 or, equivalently, that Aα = 0 for all α > 4.
In the following, we shall assume that the second case occurs, and we will come
to a contradiction.
Restricting (2.36) to the sequence of points {pkr }kr ∈N and then taking the limit,
we get, also using (2.40), that
3 1 3
X 7c 2 X
λ̄i λ̄j (λ̄αi − λ̄αj )2 = 2
sup |A| + |T̄ | − c (λ̄αi − λ̄αj )2 ,
2 Σ3 3
i,j=1 i,j=1
where λ̄i = limkr →∞ λi and λ̄αi = limkr →∞ λαi . From (2.41) we have λ̄αi 6= λ̄αj , if
i 6= j, and then, from (2.35), one obtains
1 7c
(2.42) λ̄i λ̄j = sup |A|2 + |T̄ |2 − c, if i 6= j.
2 Σ3 3
Now, on the one hand, taking the limit in (2.34) and applying Lemma 2.1, we
get
3
!2 3
!
X X 14c
λ̄i =2 (λ̄i )2 + 2 sup |A|2 + |T̄ |2 − 2c ,
Σ3 3
i=1 i=1
or, equivalently,
3 14c 2
(2.43) |H|2 = |φ̄4 |2 + sup |A|2 + |T̄ | − 2c,
2 Σ3 3
where φ4 = A4 − |H| I is the traceless part of A4 and φ̄4 = limkr →∞ φ4 (pkr ).
On the other hand, we have
3
!2 3
X X X
λi λj = λi − (λi )2 = 9|H|2 − (|φ4 |2 + 3|H|2 ) = 6|H|2 − |φ4 |2 ,
i6=j i=1 i=1
which, taking the limit and using (2.42), gives

(2.44) |φ̄4 |2 = 6|H|2 − 3 sup |A|2 − 14c|T̄ |2 + 6c.
Σ3
Summarizing, from (2.43) and (2.44), one obtains
3
|φ̄4 |2 = − |H|2 ,
4
which is a contradiction and, therefore, this case cannot occur.
We have just proved that our submanifold Σm actually is a cmc hypersurface in
M m+1 (c) for any m ≥ 3.
Now, from (2.27), it is easy to see that

m
|φ|2 ≤ 2c + |H|2 < r2 ,
m−1
where φ is the traceless part of σ and r is the positive root of the polynomial
m(m − 2)
P (t) = t2 + p |H|t − m(c + |H|2 ).
m(m − 1)
We then use [4, Theorem 1.5] (see also [123]) to conclude that φ = 0, i.e., Σm is a
totally umbilical cmc hypersurface in M m+1 (c).
In the same way as above, one can prove the next three results.
Theorem 2.18 ([76]). Let Σm be a complete non-minimal pmc submanifold in
M n (c)
× R, n > m ≥ 3, c < 0, with mean curvature vector field H and second
fundamental form σ. If H is orthogonal to ξ and
2c(m + 1) 2 m2
|σ|2 + |T | ≤ 4c + |H|2 ,
m m−1
then Σm is a totally umbilical cmc hypersurface in M m+1 (c).
Theorem 2.19 ([76]). Let Σ2 be a complete non-minimal pmc surface in M n (c)×
R, n > 2, c > 0, such that the angle between H and ξ is constant and
|σ|2 + 3c|T |2 ≤ 4|H|2 + 2c.
Then, either
(1) Σ2 is pseudo-umbilical and lies in M n (c); or p
(2) Σ2 is a torus S1r1 × S1r2 in M 3 (c), where r2 = 1/c − r12 and r2 6= 1/2c.
R, n > 2, c < 0, such that H is orthogonal to ξ and
|σ|2 + 5c|T |2 ≤ 4|H|2 + 4c.
Then Σ2 is pseudo-umbilical and lies in M n (c).
5. Holomorphic differentials and Simons type formulas. Surfaces with

constant mean curvature and finite total curvature
Let Σ2 be a pmc surface in M̄ = M n (c) × R, with mean curvature vector field
H 6= 0, and again consider the holomorphic differential introduced in [1, 5], that we
used in the previous section, i.e., the (2, 0)-part of the quadratic form Q defined on
Σ2 by
(2.45) Q(X, Y ) = 2hAH X, Y i − chX, ξihY, ξi.
We will need a Simons type equation found by S.-Y. Cheng and S.-T. Yau
(equation 2.8 in [40]), which generalizes some previous results in [112, 128, 129].
Let N be an n-dimensional Riemannian manifold, and consider a symmetric operator
S on N , that satisfies the Codazzi equation (∇X S)Y = (∇Y S)X, where ∇ is the
Levi-Civita connection on the manifold. Then, we have
n n
1 X 1 X
(2.46) ∆|S|2 = |∇S|2 + λi (trace S)ii + Rijij (λi − λj )2 ,
2 2
i=1 i,j=1
where λi , 1 ≤ i ≤ n, are the eigenvalues of S, and Rijkl are the components of the
Riemannian curvature of N .
Now, let us consider an operator S, defined on Σ2 by
1 c c
(2.47) S= AH − hT, ·iT + |T |2 − |H| I,
|H| 2|H| 4|H|
where T is the component of ξ tangent to the surface. When the ambient space is
3-dimensional, this operator was introduced in [19]. We shall prove that |S|2 is a
bounded subharmonic function on the surface.
First, it is easy to see that
trace Q
(2.48) 2|H|hSX, Y i = Q(X, Y ) − hX, Y i,
2
where Q is given by (2.45), which implies that S is symmetric and traceless. Another
direct consequence of (2.48) is the following lemma.
Lemma 2.21. The (2, 0)-part of Q vanishes on Σ2 if and only if S = 0 on the
surface.
Lemma 2.22 ([19]). The operator S satisfies the Codazzi equation (∇X S)Y =
(∇Y S)X, where ∇ is the Levi-Civita connection on the surface.
From Lemma 2.22, equation (2.46), and the fact that trace S = 0, we easily get
the following proposition.
Theorem 2.23 ([19, 74]). If Σ2 is a pmc surface in M n (c) × R and S is the
operator given by (2.47), then we have
1
∆|S|2 = 2K|S|2 + |∇S|2 ,
2
where K is the Gaussian curvature of the surface.
Corollary 2.24 ([17]). Consider a non-minimal pmc surface Σ2 in M n (c) × R
such that µ = supΣ (|σ|2 − (1/|H|)2 |AH |2 ) < +∞ and let u be the function u = |S|.
Then
−∆u ≤ au3 + bu,
where a and b are constants depending on c, |H|, and µ.
Proof. Let {E3 = H/|H|, E4 , . . . , En+1 } be a local orthonormal frame field in
the normal bundle, and denote Aα = AEα .
From the definition (2.47) of S, we have, after a straightforward computation,
1 c2 c
det A3 = |H|2 − |S|2 − 2
|T |4 − hST, T i
2 16|H| 2|H|
and then, using (2.2), the Gaussian curvature K of our surface can be written as
1 c2 c X
(2.49) K = c(1 − |T |2 ) + |H|2 − |S|2 − |T |4
− hST, T i + det Aα .
2 16|H|2 2|H|
α>3
Next, from Theorem 2.23, we obtain
1 1 c2
(2.50) ∆|S|2 =|∇S|2 + 2 c(1 − |T |2 ) + |H|2 − |S|2 − |T |4
2 2 16|H|2
c X
− hST, T i + det Aα |S|2 .
2|H|
α>3
Since Aα is traceless for any α > 3, we have

X 1
−2 det Aα = |σ|2 − |AH |2 ≤ µ.
|H|2
α>3
1
We note that |∇|S|| ≤ |∇S| and, since S is traceless, |ST | = √ |T ||S|. Then,
2
from (2.50), also using that the Schwarz inequality implies |hST, T i| ≤ |T ||ST |, it is
easy to see that
c2

3 2 2 |c|
∆|S| ≥ −|S| + |S| 2c(1 − |T | ) + 2|H| − 2
−µ − √ |S|2
8|H| 2|H|
c2

3 2 |c|
≥ −|S| + |S| 2 min{c, 0} + 2|H| − 2
−µ − √ |S|2
8|H| 2|H|
c2

|c| 3 |c| 2
≥− 1+ √ |S| − √ − 2|H| + − 2 min{c, 0} + µ |S|,
2 2|H| 2 2|H| 8|H|2
Now, assume that K ≥ 0. Since trace Aα = 0 implies det Aα ≤ 0, it follows,
from (2.49), that
1 c c2
− |S|2 − hST, T i − |T |4 + c(1 − |T |2 ) + |H|2 ≥ 0.
2 2|H| 16|H|2
√
From |hST, T i| ≤ (1/ 2)|T ||S|, one obtains that
c |c|
− hST, T i ≤ √ |S|,
2|H| 2 2|H|
which implies
1 |c|
− |S|2 + √ |S| + c(1 − |T |2 ) + |H|2 ≥ 0.
2 2 2|H|
Next, we will consider two cases as c < 0 or c > 0, and prove that, in both
situations, the function |S| is bounded.
If c < 0, we have
1 c
− |S|2 − √ |S| + |H|2 ≥ 0
2 2 2|H|
p √
and then |S| ≤ ( c2 + |H|2 − c)/(2 2|H|), while, when c > 0, one obtains
1 c
− |S|2 + √ |S| + c + |H|2 ≥ 0,
2 2|H|
p √
which is equivalent to |S| ≤ ( c2 + 16c|H|2 + 16|H|2 + c)/(2 2|H|).
Theorem 2.25 ([74]). Let Σ2 be a complete non-minimal pmc surface with non-
negative Gaussian curvature K in M n (c)×R, c 6= 0. Then one of the following holds:
(1) K = 0;
(2) Σ2 is a minimal surface of a totally umbilical hypersurface of M n (c);
(3) Σ2 is a cmc surface in a 3-dimensional totally umbilical submanifold of
M n (c);
(4) Σ2 lies in M 4 (c)×R ⊂ R5 ×R, and there exists a plane π such that the level
lines of the height function p ∈ Σ2 → hx(p), ξi are curves lying in planes
parallel to π.
Proof. As Σ2 is complete and has non-negative Gaussian curvature, it follows
that it is a parabolic space (see [81]). Therefore, since the above calculation and
Theorem 2.23 show that |S|2 is a bounded subharmonic function, we have that |S| is
a constant. Thus, from Theorem 2.23, we can see that K = 0 or S = 0. Then, from
Lemma 2.21, we have that, when Σ2 is not flat, the (2, 0)-part of the quadratic form
Q vanishes and then the last three items of our theorem can be obtained exactly as
in the proofs of Theorems 2 and 3 in [6].
Remark 2.26. The same result as in Theorem 2.25 was previously obtained by
H. Alencar, M. do Carmo, and R. Tribuzy in the particular case when c < 0 ([6,
Theorem 3]).
The Simons type equation given in Theorem 2.23 can be also used in the study
of pmc surfaces with finite total curvature, as we will see in the following.
Definition 2.27. A surface Σ2 in M n (c) × R has finite total curvature if it
satisfies Z
|S|2 dv < +∞,
Σ2
where S is given by (2.47).
We will use the following notations. Let Σ2 be a surface in M n (c) × R and
x0 ∈ Σ2 a fixed a point. Consider the Riemannian distance function d(x0 , x) to x0
and the following open domains
B(R) = {x ∈ Σ2 |d(x0 , x) < R} and E(R) = {x ∈ Σ2 |d(x0 , x) > R}.
Before stating our next result, let us recall that a pmc surface Σ2 in M n (c) × R
satisfies a Sobolev inequality of the form
(2.51) ∀f ∈ C0∞ (Σ2 ), ||f ||2 ≤ A||∇f ||1 + B||f ||1 ,
where ||f ||p = ( Σ2 |f |p dv)1/p is the Lp -norm of the function f and A and B are
R
constants that depends only on the mean curvature |H| of the surface (see [84]).
R such that the norm of its second fundamental form σ is bounded and
Z
(2.52) |S|2 dv < +∞.
Σ2
Then the function u = |S| goes to zero uniformly at infinity. More precisely,
there exist positive constants C0 and C1 depending on c, |H|, and µ, where µ =
supZΣ2 (|σ|2 − (1/|H|)2 |AH |2 ), and a positive radius RΣ2 determined by the condition
C1 u2 dv ≤ 1 such that, for all R ≥ RΣ2 ,
E(RΣ2 )
Z
||u||∞,E(2R) ≤ C0 u2 dv.
Σ2
Moreover, there exist someZ positive constants D0 and E0 depending on c, |H|, and
µ such that the inequality u2 dv ≤ D0 implies
Σ2
Z
||u||∞ ≤ E0 u2 dv.
Σ2
Proof. Since the function u satisfies the Sobolev inequality (2.51) and the
inequality in Corollary 2.24, we can work as in the proof of [18, Theorem 4.1]
to conclude.
Remark 2.29. When n = 2, we have µ = 0 and it is easy to see that (2.52)
implies that |σ| is bounded, since a straightforward computation, using (2.47), shows
that
1 c c2 |T |4
|σ|2 = 2
|AH |2 = |S|2 + hST, T i + + 2|H|2 .
|H| |H| 8|H|2
Corollary 2.30 ([17]). Let Σ2 be a complete non-minimal cmc surface in
M 2 (c)
× R with Z
|S|2 dv < +∞.
Σ2
Then the function u = |S| goes to zero uniformly at infinity. More precisely, there
exist positive constants C0 and C1 depending
Z on c and |H|, and a positive radius
RΣ2 determined by the condition C1 u2 dv ≤ 1 such that, for all R ≥ RΣ2 ,
E(RΣ2 )
Z
||u||∞,E(2R) ≤ C0 u2 dv.
Σ2
Z positive constants D0 and E0 depending on c and |H|

Moreover, there exist some
such that the inequality u2 dv ≤ D0 implies
Σ2
Z
||u||∞ ≤ E0 u2 dv.
Σ2
In the following, we will use Theorem 2.28 to prove some results on the com-
pactness of pmc surfaces with finite total curvature.
R with mean curvature vector field H and such that the norm of its second funda-
mental form σ is bounded and
Z
|S|2 dv < +∞.
Σ2
Then we have
p
(1) If c > 0 and |H|2 > (µ + µ2 p + c2 )/4, then Σ2 is compact;
(2) If c < 0 and |H|2 > (µ − 2c + µ2 − 4cµ + 5c2 )/4, then Σ2 is compact,
where µ = supΣ2 (|σ|2 − (1/|H|2 )|AH |2 ).
Proof. As we have seen in the proof of Corollary 2.24, the Gaussian curvature
K of Σ2 can be written as
1 c2 c X
K = c(1 − |T |2 ) + |H|2 − |S|2 − |T | 4
− hST, T i + det Aα
2 16|H|2 2|H|
α>3
and then we have

1 c2 |c| X
(2.53) K ≥ c(1 − |T |2 ) + |H|2 − |S|2 − − √ |S| + det Aα ,
2 16|H|2 2 2|H|
α>3
√
since |hST, T i| ≤ |T ||ST |Pand |ST | = (1/ 2)|T ||S|.
Next, if c > 0, since α>3 det Aα = −(1/2)(|σ|2 − (1/|H|2 )|AH |2 ) ≥ −µ/2, we
get
1 c c2 µ
K ≥ − |S|2 − √ |S| + |H|2 − 2
− .
2 2 2|H| 16|H| 2
When c < 0, from (2.53), we have
1 c c2 µ
K ≥ c − |S|2 − √ |S| + |H|2 − 2
− .
2 2 2|H| 16|H| 2
In both cases, the hypotheses and Theorem 2.28 imply that the superior limit of
K at infinity is positive. This means that the negative part K − of K has compact
support and, therefore, satisfies
Z
|K − | dv < +∞.
Σ2
It follows, from Huber’s Theorem (see [138, Theorem 1]), that the positive part K +
of K also satisfies Z
K + dv < +∞.
Σ2
Next, outside a compact set Ω we have K + ≥ k/2 > 0, where
16|H|4 − 8(µ − 2c)|H|2 − c2

, if c < 0


16|H|2

k=
16|H|4 − 8µ|H|2 − c2
, if c > 0


16|H|2

and then Vol(Σ2 \Ω) < +∞. Since the volume of a complete non-compact surface
in M n (c) × R is infinite (see [79]), it follows that Σ2 is compact.
When n = 2, we use Remark 2.29 to prove the following result.
Corollary 2.32 ([17]). Let Σ2 be a complete cmc surface in M 2 (c) × R such
that Z
|S|2 dv < +∞.
Σ2
Then we have
√
(1) If c > 0 and |H| > p c/2, then Σ2 is compact.
√ √
(2) If c < 0 and |H| > ( 5 + 2/2) −c, then Σ2 is compact.
R, c < 0, with mean curvature vector field H and such that the norm of its second
fundamental form σ is bounded and
Z
(|S|2 + |N |2 ) dv < +∞.
Σ2
p
If |H|2 > (µ + µ2 + c2 )/4, where µ = supΣ2 (|σ|2 − (1/|H|2 )|AH |2 ), then Σ2 is
compact.
Proof. From Proposition 2.9, as |N |2 = 1 − |T |2 , we have

n+1
1 X
(2.54) − ∆|N |2 = |AN |2 + c|N |2 (1 − |N |2 ) − |Aα T |2 .
2
α=3
Next, since ∇⊥
X N = −σ(X, T ), one obtains
2|N |X(|N |) = X(|N |2 ) = 2h∇⊥

X N, N i = −2hAN T, Xi
and then
(2.55) |N |2 |∇|N ||2 = |AN T |2 .
Replacing (2.55) in (2.54), we get that

n+1
X
3 2 2 2 2
−|N | ∆|N | = (|AN | + c|N | (1 − |N | ))|N | − |Aα T |2 |N |2 + |AN T |2 ,
α=3
which gives
−|N |3 ∆|N | ≤ (|σ|2 + c(1 − |N |2 ))|N |4 ,
since, using the Schwarz inequality, we can see that n+1 2 2 2
P
α=3 |Aα T | |N | ≥ |AN T | . It
follows that there exists a constant d such that
−∆|N | ≤ −c|N |3 + d|N |.
Since the function w = |N | also satisfies the Sobolev inequality (2.51) and, by
hypothesis,
Z
w2 dv ≤ +∞,
Σ2
we can again work as in the proof of [18, Theorem 4.1] to obtain that w goes to zero
uniformly at infinity.
Now, from (2.53), we get
1 c2 |c| µ
K ≥ c|N |2 + |H|2 − |S|2 − 2
− √ |S| − ,
2 16|H| 2 2|H| 2
which, together with Theorem 2.28, shows that the superior limit of K is positive.
This means that we can use the same arguments as in the proof of Theorem 2.31 to
conclude.
Again using Remark 2.29, we have the following corollary.
Corollary 2.34 ([17]). Let Σ2 be a complete non-minimal cmc surface in

M 2 (c)
× R, c < 0, such that
√
−c
Z
|H| > and (|S|2 + |N |2 ) dv < +∞,
2 Σ2
Then Σ2 is compact.
6. Helix surfaces with parallel mean curvature in M n (c) × R

As we have seen in the previous sections, pmc surfaces in M n (c) × R with the
property that the tangent part T of the unit vector field ξ has constant length often
appears. We will see that such surfaces will also play a role in the second part
of this thesis. Therefore, we will end this chapter with a classification result for
non-minimal pmc helix surfaces in M n (c) × R.
Definition 2.35. A surface Σ2 in M n (c) × R is called a helix surface (or a
constant angle surface) if the angle function θ ∈ [0, π) between its tangent spaces
and the unit vector field ξ tangent to R is constant on Σ2 .
Remark 2.36. A helix surface is characterized by the fact that the tangent part
T of ξ has constant length.
Since we will use their result, we recall that U. Abresch and H. Rosenberg de-
termined all complete cmc surfaces in M 2 (c) × R on which the Abresch-Rosenberg
2 ,
differential vanishes. Thus, they found four classes of such surfaces, denoted by SH
2 2 2
DH , CH , and PH , all of them described in detail in [1].
We also have to mention that complete cmc helix surfaces in M 2 (c) × R, and
actually in all 3-dimensional homogeneous spaces, were determined in [56, Theo-
rem 2.2], while [33, Theorem 1] shows that there are no non-minimal pmc helix
surfaces in S3 (1) × R with 0 < |T | < 1.
In the general case, we have the following classification theorem.
Theorem 2.37 ([63]). Let Σ2 be a non-minimal pmc helix surface with Gaussian
curvature K and mean curvature vector field H in M n (c) × R, with c 6= 0, and let T
be the tangent part of the unit vector field ξ tangent to R. Then one of the following
holds:
(1) Σ2 is a minimal surface in a non-minimal totally umbilical hypersurface of
M n (c);
(2) Σ2 is a surface with constant mean curvature in a 3-dimensional totally
umbilical or totally geodesic submanifold of M n (c);
(3) Σ2 is a vertical cylinder over a circle in M n (c) with curvature κ = 2|H|;
(4) c < 0 and Σ2 lies in M 2 (c) × R. Moreover, 4|H|2 + c|T |2 = 0, K = c(1 −
|T |2 ) < 0, and the Abresch-Rosenberg differential vanishes on Σ2 . If the
surface is complete, then it is a cmc surface of type PH2 in M 2 (c) × R (see
[1] for a detailed description of these surfaces);
(5) c > 0 and locally Σ2 is the standard product γ1 × γ2 , wherep γ1 : I ⊂ R →
M 4 (c) × R is a helix in M 4 (c) × R with curvatures κ1 = c(1 − |T |2 ) and
√
κ2 = |T | c, and γ2 p : I ⊂ R → M 4 (c) ⊂ M 4 (c) × R is a circle in M 4 (c)
with curvature κ = 4|H|2 + c(1 − |T |2 ). If the surface is complete, then
the above decomposition holds globally.
Proof. As we have seen before, either Σ2 is pseudo-umbilical, i.e., H is an
umbilical direction everywhere, or, at any point in an open dense set W ⊂ Σ2 , there
exists a local orthonormal frame field that diagonalizes AU for any normal vector
field U defined on W (see [6, Lemma 1]). If Σ2 is a pseudo-umbilical pmc surface in
M n (c) × R, then it lies in M n (c), i.e., |T | = 0 (see [6, Lemma 3]).
From [6, Remark 1], we also know that, since H is parallel, the immersion of Σ2
in M n (c) × R is analytic, i.e., the functions of two variables that locally define the
immersion are real analytic. Therefore, it satisfies a principle of unique continuation

and, as a consequence, it cannot vanish on an open connected subset of Σ2 unless it
vanishes identically.
Now, when |T | = 0, our surface lies in M n (c) and we obtain the first two items
of the theorem using [136, Theorem 4].
When |T | = 1, the vector field ξ is tangent to the surface and this means that Σ2
is a pmc vertical cylinder over a curve γ in M n (c). Since ξ is parallel in M n (c) × R,
it follows that σ(ξ, ξ) = 0 and then we easily get that γ is a circle in M n (c) with
curvature κ = 2|H|.
Henceforth, let us assume that 0 < |T | < 1. It follows that H is not an umbilical
direction on an open dense set W . We will work on this set and then we will extend
our results throughout Σ2 by continuity.
Consider a global orthonormal frame field {E1 = T /|T |, E2 } on Σ2 , and let N
be the normal part of ξ. Then, since Σ2 is a helix surface, it follows that ∇E1 E1 =
∇E1 E2 = 0 and, as ∇ ¯ X ξ = 0 implies ∇X T = AN X and σ(T, X) = −∇⊥ N , that
X
AN E1 = 0 (see [121, Proposition 2.1]). We also have
1 1
h∇E2 E2 , E1 i = −hE2 , ∇E2 E1 i = − hE2 , AN E2 i = − hσ(E2 , E2 ), N i
|T | |T |
1 2hH, N i
= − h2H − σ(E1 , E1 ), N i = − ,
|T | |T |
which means that
2hH, N i 2hH, N i
(2.56) ∇E2 E2 = − E1 and ∇E2 E1 = E2 .
|T | |T |
Since AN E1 = 0, we also get AN E2 = 2hH, N iE2 . From the Ricci equation
(1.5), we obtain [AH , AN ] = 0, which means that hH, N ihAH E1 , E2 i = 0. Since we
also have
E2 (hH, N i) = hH, ∇⊥
E2 N i = −|T |hH, σ(E1 , E2 )i = −|T |hAH E1 , E2 i,
one sees that hAH E1 , E2 i = 0. Moreover, again using (1.5), we have [AH , AU ] = 0
for any normal vector field U and then, since H is not umbilical, σ(E1 , E2 ) = 0 and
∇⊥E2 N = 0.
Next, let us denote λi = hAH Ei , Ei i, i ∈ {1, 2}, the eigenfunctions of AH and in
the following we will compute Ej (λi ), i, j ∈ {1, 2}.
Using (2.1), (2.56), and σ(E1 , E2 ) = 0, we get
∇⊥ ⊥
E2 σ(E1 , E1 ) = (R̄(E2 , E1 )E1 ) + 2σ(E1 , ∇E2 E1 ) = 0
and then, from the Codazzi equation (1.4),

(2.57) ∇⊥ ⊥ ⊥
E1 σ(E2 , E2 ) =(R̄(E1 , E2 )E2 ) + 2σ(E2 , ∇E1 E2 ) + ∇E2 σ(E1 , E2 )
− σ(∇E2 E1 , E2 ) − σ(E1 , ∇E2 E2 )
2hH, N i
= − c|T |N + (σ(E1 , E1 ) − σ(E2 , E2 )).
|T |
Therefore, since H is parallel, we have
hH, N i
(2.58) E1 (λ1 ) = −E1 (λ2 ) = −h∇⊥
E1 σ(E2 , E2 ), Hi = (4|H|2 + c|T |2 − 4λ1 )
|T |
and
(2.59) E2 (λ1 ) = −E2 (λ2 ) = h∇⊥
E2 σ(E1 , E1 ), Hi = 0.
From Proposition 2.9 (or from [75, Proposition 1.4]), we know that
1
∆|T |2 = |AN |2 + K|T |2 − 2hAH T, T i
2
and then, since |T | = constant 6= 0, the Gaussian curvature K of Σ2 is given by
4hH, N i2
(2.60) K = 2λ1 − .
|T |2
Since ∇⊥
X N = −σ(X, T ) implies that
(2.61) E1 (hH, N i) = −|T |λ1 and E2 (hH, N i) = 0,

from (2.58) and (2.59), we obtain
2hH, N i
(2.62) E1 (K) = (4|H|2 + c|T |2 ) and E2 (K) = 0.
|T |
The fact that Σ2 is not pseudo-umbilical implies that it lies in M 4 (c) × R (see
[6, Theorem 1]).
In order to describe our surface by taking advantage of the above formulas, we
will first consider the case when H k N on an open connected subset W0 of Σ2 .
This means that hH, N i = ±|H||N | = constant and, from (2.61), one obtains that
λ1 = 0. Then, from (2.58) and (2.60), we get 4|H|2 + c|T |2 = 0 and K = c(1 − |T |2 )
on W0 .
Let {E3 = H/|H|, E4 , E5 } be a global orthonormal frame field in the normal
bundle. Then, since σ(E1 , E2 ) = 0, on W0 we have

0 0 a 0 b 0
A3 = , A4 = , A5 = ,
0 2|H| 0 −a 0 −b
where Aα = AEα , and, from the Gauss equation (1.3),
K = c(1 − |T |2 ) − a2 − b2 .
But, as we have seen, K = c(1 − |T |2 ) on W0 , which implies that a = b = 0 on
W0 . Consider the subbundle L = span{Im σ} = span{H} in the normal bundle. It
follows that ξ ∈ T Σ ⊕ L and T Σ2 ⊕ L is parallel with respect to ∇ ¯ and invariant by
2
R̄. Therefore, we use [54, Theorem 2] to show that W0 lies in M (c) × R. From the
analyticity of the immersion of Σ2 in M 4 (c) × R, it follows that Σ2 lies in M 2 (c) × R
(see [6, Remark 1] for more details). As a direct consequence, we have H k N on Σ2
and then λ1 = 0, 4|H|2 + c|T |2 = 0, and K = c(1 − |T |2 ) < 0 on Σ2 . Now, it is easy
to verify that the Abresch-Rosenberg differential vanishes on the surface. Moreover,
if Σ2 is complete, all these properties of Σ2 lead to the conclusion that our surface
is a cmc surface of type PH2 (see [1]).
Next, we will consider the remaining case, when H k N only at isolated points.
We can then define a local orthonormal frame field
( )
1 cot β N
E3 = H− N, E4 = , E5
|H| sin β |N | |N |
in the normal bundle, where β ∈ (0, π) is the angle between H and N . With respect
to {E1 , E2 }, we can write
! !
λ1
0 0 0

|H| sin β λ 0
A3 = , A4 = i , A5 =
0 2|H| sin β − |H|λsin
1
β
0 2hH,N
|N |
0 −λ
and then, from the Gauss equation (1.3), one obtains
λ21
(2.63) K = c(1 − |T |2 ) + 2λ1 − − λ2 .
|H|2 sin2 β
Using equation (2.60), we have
λ21 4hH, N i2
(2.64) λ2 = c(1 − |T |2 ) − +
|H|2 sin2 β |T |2
and then
λ21 4hH, N i2
(2.65) 2λE1 (λ) = E1 − + .
|H|2 sin2 β |T |2
Next, we shall compute E1 (λ). Using equation (2.57), the fact that H is parallel,
and ∇⊥
X N = −σ(X, T ), we obtain
E1 (λ) = −E1 (hσ(E2 , E2 ), E5 i) = −h∇⊥ ⊥
E1 σ(E2 , E2 ), E5 i − hσ(E2 , E2 ), ∇E1 E5 i
4hH, N i
=− λ − hσ(E2 , E2 ), E3 ih∇⊥ ⊥
E1 E5 , E3 i − hσ(E2 , E2 ), E4 ih∇E1 E5 , E4 i
|T |
4hH, N i λ1 2hH, N i
=− λ + 2|H| sin β − hE5 , ∇⊥ E1 E3 i + hE5 , ∇⊥
E1 E4 i
|T | |H| sin β |N |
4hH, N i |T | cos β
=− λ− λλ1 .
|T | |H||N | sin2 β
Replacing in (2.65) and using (2.58) and (2.61), we get, after a straightforward
computation
λ21 2

(2.66) hH, N i 2λ1 − − λ = 0,
|H|2 sin2 β
which, together with (2.63), leads to hH, N i(K − c(1 − |T |2 )) = 0, that, taking (2.62)
into account, can be written as E1 ((K − c(1 − |T |2 ))2 ) = 0. Again from (2.62), we
have E2 ((K − c(1 − |T |2 ))2 ) = 0. It follows that K − c(1 − |T |2 ) = constant and
then, using (2.60), one obtains
4hH, N i2
E1 2λ1 − = 0.
|T |2
Now, from equations (2.58) and (2.61), we have (4|H|2 + c|T |2 )hH, N i = 0. We
will consider two cases as 4|H|2 + c|T |2 = 0 or 4|H|2 + c|T |2 6= 0.
Case I: 4|H|2 +c|T |2 = 0. Let us assume that hH, N i = 0 on an open connected
set W0 . From (2.61) it follows that λ1 = 0 and then, from (2.60), that K = 0 on W0 .
But, from (2.63), we have K = c(1 − |T |2 ) − λ2 , which means that λ2 = c(1 − |T |2 ).
This implies c > 0 and that is a contradiction, since 4|H|2 + c|T |2 = 0. Therefore,
hH, N i = 0 on a closed set without interior points and then, from (2.66), we have
λ21
2λ1 − − λ2 = 0
|H|2 sin2 β
and K = c(1 − |T |2 ) on an open dense set. Since 4|H|2 + c|T |2 = 0, from (2.60),
one obtains sin2 β = 2λ1 /(c|N |2 ) and then λ2 = 2λ1 /|T |2 , which means that λ1 = 0,
λ = 0, and sin2 β = 0 on an open dense set. The last identity shows that H k N on
an open dense set, which is a contradiction.
Case II: 4|H|2 + c|T |2 6= 0. In this case, hH, N i = 0 on an open dense set and,
from (2.61), it follows that λ1 = 0 and then, from (2.60), we have K = 0 and, from
(2.64), λ2 = c(1 − |T |2 ), which implies c > 0. The shape operator is given by

0 0 0 0 λ 0
A3 = , A4 = , A5 = .
0 2|H| 0 0 0 −λ
From equations (2.56), we see that ∇E1 = ∇E2 = 0 and we then can apply the de
Rham Decomposition Theorem (see [95]) to show that locally Σ2 is the standard
product γ1 × γ2 of two Frenet curves γ1 : I ⊂ R → M 4 (c) × R and γ2 : I ⊂ R →
M 4 (c) × R such that γ10 = E1 and γ20 = E2 . We note that γ2 actually lies in M 4 (c).
If the surface is complete, then this decomposition holds globally.
Next, we will characterize γ1 and γ2 by using their Frenet equations. Let {X11 =
E1 , X21 , . . . , Xr1 }, 1 ≤ r ≤ n + 1, be the Frenet frame field of γ1 . We have that
¯ E E1 = σ(E1 , E1 ) = λE5 ,
∇ 1
p
and then, from the first Frenet equation, it follows that κ1 = |λ| = c(1 − |T |2 )
and X21 = ±E5 .
Next, we have h∇⊥ ⊥
E1 E5 , E3 i = 0, since E3 = H/|H| is parallel, and h∇E1 E5 , E4 i =
⊥
−hE5 , ∇E1 E4 i = (|T |/|N |)hE5 , σ(E1 , E1 )i = (|T |/|N |)λ. Therefore, one obtains
¯ E E5 = ∓A5 E1 ± ∇⊥ E5 = ∓λE1 ± |T | λE4 .

¯ E X 1 = ±∇
∇ 1 2 1 E1
|N |
√
From the second Frenet equation, we get κ2 = (|T |/|N |)|λ| = |T | c and X31 = ±E4 .
It follows that
∇¯ E X31 = ±∇ ¯ E E4 = ±∇⊥E E4 = ∓(|T |/|N |)σ(E1 , E1 ) = ∓(|T |/|N |)λE5 = −κ2 X
1
1 1 1 2
and we have just proved that γ1 is a helix.
Now, let {X12 = E2 , X22 , . . . , Xr2 }, 1 ≤ r ≤ n + 1, be the Frenet frame field of γ2 .
Then, from
¯ E E2 = σ(E2 , E2 ) = 2|H|E3 − λE5 ,
∇ 2
and the first Frenet equation of γ2 , one obtains

p p
κ = 4|H|2 + λ2 = 4|H|2 + c(1 − |T |2 )
and X22 = (1/κ)(2|H|E3 − λE5 ).
¯ = 0, that ∇⊥ E5 = 0. Then, since E3 is parallel,
It is easy to verify, using ∇ξ E2
we have ∇¯ E X = −(1/κ)(2|H|A3 E2 + λA3 E2 ) = −κE2 and we conclude.
2
2 2
Remark 2.38. From the proof of Theorem 2.37 it is easy to see that the only
non-minimal pmc helix surfaces in M 3 (c) × R are those given by the first four cases
of our theorem.
Part 2
Biharmonic Submanifolds
CHAPTER 3
Biharmonic and Biconservative Submanifolds in

M n (c) × R
1. Introduction
In this chapter, we continue the study of pmc and cmc submanifolds in product
spaces of type M n (c) × R under the supplementary biharmonicity (or biconservativ-
ity) hypothesis.
First, we use Theorem 2.6 to prove a gap theorem for pmc biharmonic sub-
manifolds (Theorem 3.14) and determine all pmc proper-biharmonic surfaces and,
moreover, all pmc biharmonic submanifolds in M n (c) × R satisfying ∇AH = 0 (The-
orems 3.21 and 3.22).
Next, we consider biconservative surfaces and explicitly find these surfaces that
have parallel mean curvature in M n (c) × R (Theorem 3.28) and classify cmc bicon-
servative surfaces in M 3 (c) × R with the property that the mean curvature vector
field is orthogonal to ξ (Theorem 3.32). Then, using a result similar to Theorem 2.28
for cmc biconservative surfaces in Hadamard manifolds (Theorem 3.38), we prove
compactness results for such surfaces (Theorem 3.40, Corollary 3.41).
2. Preliminaries
The biharmonic maps were suggested in 1964 by J. Eells and J. H. Sampson,
as a generalization of harmonic maps (see [49]). Thus, whereas a harmonic map
ψ : (M, g) → (N, h) between two Riemannian manifolds is a critical point of the
energy functional
Z
1
E(ψ) = |dψ|2 dv,
2 M
a biharmonic map is a critical point of the bienergy functional
Z
1
E2 (ψ) = |τ (ψ)|2 dv,
2 M
where τ (ψ) = trace ∇dψ is the tension field that vanishes for harmonic maps. The
Euler-Lagrange equation corresponding to the bienergy functional was obtained by
G. Y. Jiang [91]:
τ2 (ψ) = ∆τ (ψ) − trace R̄(dψ, τ (ψ))dψ
=0
where τ2 (ψ) is the bitension field of ψ, ∆ = trace(∇ψ )2 = trace(∇ψ ∇ψ − ∇ψ ∇ ) is

the rough Laplacian defined on sections of ψ −1 (T N ) and R̄ is the curvature tensor
of N . Since any harmonic map is biharmonic, we are interested in non-harmonic
biharmonic maps, which are called proper-biharmonic.
73
74 Biharmonic and Biconservative Submanifolds in M n (c) × R
A biharmonic submanifold in a Riemannian manifold is a submanifold for which

the inclusion map is biharmonic. In Euclidean space the biharmonic submanifolds
are the same as those defined by Chen [31], characterized by ∆H = 0, where H is
the mean curvature vector field and ∆ is the rough Laplacian.
In general, biharmonic submanifolds in Riemannian manifolds are characterized
by the following theorem.
Theorem 3.1 ([14, 118]). A submanifold Σm in a Riemannian manifold N ,
with second fundamental form σ, mean curvature vector field H, and shape operator
A, is biharmonic if and only if
(
−∆⊥ H + trace σ(·, AH ·) + trace(R̄(·, H)·)⊥ = 0
(3.1) m 2
2 grad |H| + 2 trace A∇⊥· H
(·) + 2 trace(R̄(·, H)·)> = 0,
where ∆⊥ is the Laplacian in the normal bundle and R̄ is the curvature tensor of
N.
The stress-energy tensor associated to a variational problem, which was de-
scribed by D. Hilbert [82], is a symmetric 2-covariant tensor S conservative at
critical points, i.e., it satisfies div S = 0.
Such a tensor S, given by S = (1/2)|dψ|2 g − ψ ∗ h, was employed in the study
of harmonic maps by P. Baird and J. Eells [10] and A. Sanini [122]. It satisfies
div S = −hτ (ψ), dψi and, therefore, div S vanishes when ψ is harmonic. Since for
isometric immersions τ (ψ) is normal, it follows that div S = 0 is always satisfied in
this case.
The stress-energy tensor S2 of the bienergy is given by
1
S2 (X, Y ) = |τ (ψ)|2 hX, Y i + hdψ, ∇τ (ψ)ihX, Y i − hdψ(X), ∇Y τ (ψ)i
2
− hdψ(Y ), ∇X τ (ψ)i
and it satisfies
div S2 = −hτ2 (ψ), dψi.
If ψ :(Σm , g)→ (M̄ , h) is an isometric immersion, then we have div S2 = −τ2 (ψ)>
and thus div S2 does not always vanish.
Definition 3.2. A submanifold ψ : Σm → M̄ of a Riemannian manifold M̄ is
called a biconservative submanifold if div S2 = 0, i.e., τ2 (ψ)> = 0.
Biconservative submanifolds are characterized by the following corollary, that
follows from Theorem 3.1.
Corollary 3.3. A submanifold Σm in a Riemannian manifold M̄ is biconser-
vative if and only if
m
grad |H|2 + 2 trace A∇⊥
· H
(·) + 2 trace(R̄(·, H)·)> = 0.
2
We also recall the following theorem that will be used later on.
Theorem 3.4 ([101]). Let Σ2 be a biconservative oriented surface in a Rie-
mannian manifold M̄ . Then the (2, 0)-part of the Hopf quadratic form Q, defined
on Σ2 by
Q(X, Y ) = hσ(X, Y ), Hi,
is holomorphic if and only if the mean curvature |H| of the surface is constant.
Biharmonic and Biconservative Submanifolds in M n (c) × R 75
3. A gap theorem for biharmonic submanifolds with parallel mean

curvature in Sn × R
Whereas complete pmc biharmonic submanifolds of Sn × R are the subject of
our first main theorem, we have the following result for compact submanifolds.
Proposition 3.5 ([72]). If Σm is a compact biharmonic submanifold in Sn (c) ×
R, then Σm lies in Sn (c).
Proof. The height function of a submanifold Σm in Sn (c) × R is defined by
h = t ◦ i : Σm → R,
where t : Sn (c) × R → R is the projection map and i : Σm → Sn (c) × R is the
inclusion map. It is easy to verify that
τ (h) = dt(τ (i)) and τ2 (h) = dt(τ2 (i)),
and we see that, if Σm is biharmonic, then h is also a biharmonic function.
Since Σm is a compact biharmonic submanifold, it follows that h is a real valued
biharmonic function defined on a compact manifold, which, according to a result in
[91] (see also the translation in English [93, Proposition7]), leads to the fact that
h is actually a harmonic function, but then, using the maximum principle, we get
that h is constant, i.e., Σm lies in Sn (c).
We recall the following results which we shall use later on.
Theorem 3.6 ([11]). Let Σm be a cmc proper-biharmonic
√ submanifold in S√n (c)
with mean curvature vector field H. Then |H| ∈ (0, c] and, moreover, |H| = c if
and only if Σm is minimal in a small hypersphere Sn−1 (2c) ⊂ Sn (c).
Theorem 3.7 ([15]). If Σm is a pmc proper-biharmonic submanifold in √
Sn (c),

with mean curvature vector field H and m > 2, then |H| ∈ 0, ((m − 2)/m) c ∪
√ √
{ c}. Moreover, |H| = ((m − 2)/m) c if and only if Σm is (an open part of) a
standard product
Σm−1
1 × S1 (2c) ⊂ Sn (c),
where Σm−1
1 is a minimal submanifold in Sn−2 (2c).
¯ = 0, we get the following corollary.
Now, from Theorem 3.1, using (2.1) and ∇ξ
Corollary 3.8 ([72]). A pmc submanifold Σm in M n (c) × R, with m ≥ 2, is
biharmonic if and only if
(
H ⊥ ξ, |AH |2 = c(m − |T |2 )|H|2
(3.2)
trace(AH AU ) = 0, for any normal vector U ⊥ H.
Remark 3.9. A direct consequence of Corollary 3.8 is that there are no pmc
proper-biharmonic submanifolds in a product space M n (c) × R with c ≤ 0.
Remark 3.10. It is straightforward to verify that Σm = π −1 (Σm−1 ) is a pmc
vertical cylinder in M n (c) × R if and only if Σm−1 is a pmc submanifold in M n (c).
Moreover, the mean curvature vector field of Σm is H = ((m − 1)/m)H0 , where H0
is the mean curvature vector field of Σm−1 . It is also easy to prove that the vertical
cylinder Σm = π −1 (Σm−1 ) is proper-biharmonic in M n (c) × R if and only if Σm−1
is proper-biharmonic in M n (c).
The next corollary follows from Corollary 3.8, Theorem 3.7, and a result in [23].
Corollary 3.11 ([72]). If Σn is a cmc proper-biharmonic hypersurface in
then it is (an open part of) a vertical cylinder π −1 (Σn−1 ), where Σn−1 is
Sn (c) × R,
a cmc proper-biharmonic hypersurface in Sn (c). Moreover, if
1 2 √
(1) n = 2,√ then Σ is a circle in S (c) with curvature equal to c, and |H| =
(1/2) c;
(2) n = 3, then √Σ2 is an open part of a small hypersphere S2 (2c) ⊂ S3 (c), and
|H| = (2/3) c; √ √
(3) n > 3, then |H| ∈ (0, ((n
√ − 3)/n) c] ∪ {((n − 1)/n) c}. Furthermore,
(a) |H| = ((n−3)/n) c if and only if Σn−1 is an open part of the standard
product Sn−2 (2c) ×√S1 (2c) ⊂ Sn (c);
(b) |H| = ((n − 1)/n) c if and only if Σn−1 is an open part of a small
hypersphere Sn−1 (2c) ⊂ Sn (c).
Proposition 3.12 ([72]). Let Σm be a pmc proper-biharmonic submanifold in
Sn (c)×R, with m ≥ 2. Then its second fundamental form σ satisfies |σ|2 ≥ c(m−1),
and the equality holds if and only if Σm is a vertical cylinder π −1 (Σm−1 ) in Sm (c)×R,
where Σm−1 is a cmc proper-biharmonic hypersurface in Sm (c).
Proof. From the first equation of (3.2), we have
|σ|2 ≥ |A H |2 = c(m − |T |2 ) ≥ c(m − 1).
|H|
Thus, |σ|2 = c(m − 1) if and only if |T | = 1 at every point on Σm , i.e., Σm is a

vertical cylinder π −1 (Σm−1 ), and AU = 0 for any normal vector field U orthogonal
to H.
Next, let us consider the subbundle L = span{Im σ} of the normal bundle.
Since L is actually spanned by H, we see that it is parallel in the normal bundle
and dim L = 1. Thus the bundle T Σm ⊕ L is parallel with respect to the Levi-Civita
connection ∇ ¯ on Sn (c) × R and dim(T Σm ⊕ L) = m + 1.
From equation (2.1) we get that R̄(X, Y )Z ∈ T Σm ⊕L for any X, Y, Z ∈ T Σm ⊕L,
i.e., T Σm ⊕ L is invariant by R̄.
Therefore, since ∇¯ R̄ = 0, we can use [54, Theorem 2] to prove that there exists
an m + 1-dimensional totally geodesic submanifold of Sn (c) × R such that Σm lies
in this submanifold.
Finally, since Σm is a vertical cylinder, i.e., ξ is tangent to Σm , one obtains that
Σ lies in Sm (c) × R.
m
Proposition 3.13 ([72]). Let Σm be a pmc proper-biharmonic submanifold in
Sn (c) × R,with m ≥ 2. Then its mean curvature satisfies |H|2 ≤ c, and the equality
holds if and only if Σm is minimal in a small hypersphere Sn−1 (2c) ⊂ Sn (c).
Proof. Since |AH |2 ≥ m|H|4 , from the first equation of (3.2), we get that
c(m − |T |2 ) ≥ m|H|2 ,
and then |H|2 ≤ c. The equality holds if and only if T = 0, which means that Σm
lies in Sn (c). Thus, using Theorem 3.6, we conclude.
Now, for the sake of simplicity, we shall consider only the case c = 1, and we are
ready to prove the main result of this section.
Theorem 3.14 ([72]). Let Σm be a complete pmc proper-biharmonic submanifold

in Sn × R, with m ≥ 2, such that its mean curvature satisfies
p
2 (m − 1)(m2 + 4) + (m − 2) (m − 1)(m − 2)(m2 + m + 2)
(3.3) |H| > C(m) =
2m3
and the norm of its second fundamental form σ is bounded. Then m < n, |H| = 1,
and Σm is a minimal submanifold of a small hypersphere Sn−1 (2) ⊂ Sn .
Proof. From Corollary 3.8, we have that hH, ξi = 0, which implies
¯ X H, ξi = −hAH X, T i = −hAH T, Xi,
0 = h∇
for any tangent vector field X, and then AH T = 0. Therefore, if we consider a local
orthonormal frame field {Em+1 = H/|H|, . . . , En+1 } in the normal bundle, using
Theorem 2.6 and equation (3.2), we get
1
(3.4) ∆|AH |2 = |∇AH |2 + m(trace A3H ) − m2 |H|4 .
2
Let us consider φH = AH − |H|2 I, the traceless part of AH . We have
trace A3H = trace φ3H + 3|H|2 |φH |2 + m|H|6
and, using the first equation of (3.2),
|φH |2 = |AH |2 − m|H|4 = (m − |T |2 )|H|2 − m|H|4 .
Replacing in equation (3.4), one obtains
1
∆|φH |2 = |∇φH |2 + m(trace φ3H ) + 3m|H|2 |φH |2 − m2 |H|4 (1 − |H|2 ).
2
Using Lemma 2.3, we get
m−2
trace φ3H ≥ − p |φH |3
m(m − 1)
and then, since |T |2 |H|4 = |φH T |2 ≤ |T |2 |φH |2 ,
1 m(m − 2)
(3.5) ∆|φH |2 ≥ − p |φH |3 + 3m|H|2 |φH |2 − m2 |H|4 (1 − |H|2 )
2 m(m − 1)
m(m − 2)
= −p |φH |3 + 2m|H|2 |φH |2 − m|T |2 |H|4
m(m − 1)
m(m − 2)
≥ −p |φH |3 + 2m|H|2 |φH |2 − m|T |2 |φH |2
m(m − 1)
m−2
= m|φH |2 − p |φH | + 2|H|2 − |T |2 .
m(m − 1)
Now, we shall split our study in two cases, as m ≥ 3, or m = 2.
Case I: m ≥ 3. If 2|H|2 − |T |2 > 0, then we can write
m−2 1 P (|T |2 )
−p |φH | + 2|H|2 − |T |2 = ,
m(m − 1) m(m − 1) √ m−2 |φH | + 2|H|2 − |T |2
m(m−1)
where P (t) is a polynomial with constant coefficients, given by

P (t) = m(m − 1)t2 − (3m2 − 4)|H|2 t + m|H|2 (m2 |H|2 − (m − 2)2 ).
By using elementary arguments, we obtain that, if |H|2 > C(m), then P (t) ≥
P (1) > 0, for any t ∈ (−∞, 1].
Since C(m) > 12 , for any m ≥ 3, our hypothesis |H|2 > C(m) implies that
2|H|2 − |T |2 > 0, and then, from (3.5), we get
1 mP (|T |2 )
(3.6) ∆|φH |2 ≥ p p |φH |2
2 2 2
m(m − 1)((m − 2)|φH | + m(m − 1)(2|H| − |T | ))
P (|T |2 )
≥√ p √ |φH |2
m − 1|H|((m − 2) 1 − |H|2 + 2 m − 1|H|)
P (1)
≥√ p √ |φH |2
m − 1|H|((m − 2) 1 − |H|2 + 2 m − 1|H|)
≥ 0.
Next, let us consider a local orthonormal frame field {Ei }m m
i=1 on Σ , X a unit
tangent vector field, and {Em+1 = H/|H|, . . . , En+1 } an orthonormal frame field in
the normal bundle. Using equation (2.2), we can compute the Ricci curvature of our
submanifold
Xm
Ricci X = hR(Ei , X)X, Ei i
i=1
n+1
X
=m − 1 − |T |2 − (m − 2)hX, T i2 + mhAH X, Xi − |Aα X|2 ,
α=m+1
and it follows that

n+1
X
Ricci X ≥ (m − 1)(1 − |T |2 ) − m|AH X| − |Aα |2
α=m+1
2
≥ −m|AH | − |σ| .
Since |σ| is bounded by hypothesis, we can see that the Ricci curvature of Σm
is bounded from below, and then the Omori-Yau Maximum Principle holds on our
submanifold.
Therefore, we can use Theorem 2.4 with u = |φH |2 . It follows that there exists
a sequence of points {pk }k∈N ⊂ Σm satisfying
1
lim |φH |2 (pk ) = sup |φH |2 and ∆|φH |2 (pk ) < .
k→∞ Σm k
Since P (1) > 0, from (3.6), we get that 0 = limk→∞ |φH |2 (pk ) = supΣm |φH |2 , which
means that φH = 0, i.e., Σm is pseudo-umbilical.
Now, since AH T = 0, we have 0 = AH T = |H|2 T , i.e., T = 0 on Σm , and
therefore Σm lies in Sn , which also implies that m < n. Since |H|2 > C(m) >
((m − 1)/m)2 > ((m − 2)/m)2 , using Theorems 3.6 and 3.7, we conclude.
Case II: m = 2. In this case, from equation (3.5), we have
1 2|φH |2
∆|φH |2 ≥ 2|φH |2 (2|H|2 − |T |2 ) = (|φH |2 + 2|H|2 (2|H|2 − 1)).
2 |H|2
Now, since |H|2 > C(2) = 1/2, we have (1/2)∆|φH |2 ≥ 2|φH |4 /|H|2 ≥ 0 and,
working as in the first case, we conclude.
Remark 3.15. We note that, in the case of pmc proper-biharmonic surfaces

in Sn × R, if we take |H|2 ≥ C(2), then the conclusion of Theorem 3.14 remains
unchanged.
4. Biharmonic pmc surfaces in Sn (c) × R

A result that we have used before is that the map p ∈ Σ2 → (AH −µ I)(p), where
Σ2 is an oriented surface and µ is a constant, is analytic, and, therefore, either Σ2
is a pseudo-umbilical surface, or H is an umbilical direction on a closed set without
interior points. We shall denote by W the open dense set of points where H is not
an umbilical direction.
If Σ2 is a pseudo-umbilical pmc surface in Sn (c) × R, then it lies in Sn (c), as it
was proved in [6, Lemma 3], and, therefore, Σ2 is minimal in a small hypersphere
of Sn (c) (see [38]).
Proposition 3.16 ([72]). Let Σ2 be a pmc surface in Sn (c) × R. Then Σ2 is
proper-biharmonic if and only if either
(1) Σ2 is pseudo-umbilical and, therefore, it is a minimal surface of a small
hypersphere Sn−1 (2c) ⊂ Sn (c); or
(2) the mean curvature vector field H is orthogonal to ξ, |AH |2 = c(2−|T |2 )|H|2
and AU = 0 for any normal vector field U orthogonal to H.
Proof. As we have seen, in the first case, Σ2 is a minimal surface in a small
hypersphere of Sn (c), and then the conclusion follows from [22, Theorem 3.4].
Assume now that Σ2 is not pseudo-umbilical. From Corollary 3.8, we get that
H ⊥ ξ and |AH |2 = c(2 − |T |2 )|H|2 .
In the following, we shall work on the set W defined above. Let p be an arbitrary
point in W and consider {e1 , e2 } an orthonormal basis at p that diagonalizes AH
and AU for any normal vector U orthogonal to H. Since H ⊥ U , it follows that
trace AU = 2hH, U i = 0. The matrices of AH and AU , with respect to {e1 , e2 }, are
a + |H|2
   
0 b 0
AH =   and AU =  
0 −a + |H|2 0 −b
and then, from the last biharmonic condition (3.2), we get 0 = trace(AH AU ) = 2ab.
Since a 6= 0, we get b = 0, i.e., AU = 0.
Finally, we extend the result by continuity throughout Σ2 and conclude.
¯ = 0.
We then easily obtain our next result, using ∇ξ
Corollary 3.17 ([72]). If Σ2 is a pmc proper-biharmonic surface in Sn (c) × R
then the tangent part T of ξ has constant length.
Remark 3.18. We note that, if Σ2 is a pmc proper-biharmonic surface in Sn (c)×
R with T = 0, then it lies in Sn (c) as a minimal surface in a hypersphere Sn−1 (2c) ⊂
Sn (c) and, therefore, it is pseudo-umbilical (see [11, Theorem 5.6]).
Since for a proper-biharmonic surface in Sn (c) × R we have |T | = constant,
H ⊥ N , and AN = 0, we get the following result which is a direct consequence of
Proposition 2.9.
Proposition 3.19 ([72]). If Σ2 is a non-pseudo-umbilical pmc proper-biharmo-
nic surface in Sn (c) × R, then it is flat.
Now, we need to recall a lemma proved in [6].

Lemma 3.20 ([6]). Let Σ2 be a non-pseudo-umbilical pmc surface in M n (c) × R,
with second fundamental form σ, and define on W the subbundle L = span{Im σ∪N }
of the normal bundle. Then L is parallel, i.e., if U is a smooth section on L, then
∇⊥ U ∈ L.
In the following, let Σ2 be a non-pseudo-umbilical pmc proper-biharmonic surface
in Sn (c) × R. It follows that |T | = constant 6= 0, i.e., |N | = constant ∈ [0, 1). Since
AU = 0 for any normal vector field U orthogonal to H, we obtain dim span{Im σ} = 1
and then dim L = 1, if N = 0, and dim L = 2, if N 6= 0. Now, since the bundle
T Σ2 ⊕ L is parallel with respect to ∇, ¯ invariant by R̄, and dim(T Σ2 ⊕ L) ≤ 4, we
use [54, Theorem 2] to show that Σ2 lies in S2 (c) × R, if N = 0, or in S3 (c) × R,
otherwise.
Further, we shall prove that |T | = 1 on Σ2 .
Assume that |N | > 0. Then there is a global orthonormal frame field {E3 =
H/|H|, E4 = N/|N |}, and, from Proposition 3.16, we have A4 = 0 and |σ|2 =
|A3 |2 = c(2 − |T |2 ).
On the other hand, since the surface is flat, from (2.2), it follows that
0 = 2K = 2c(1 − |T |2 ) + 4|H|2 − |σ|2 = −c|T |2 + 4|H|2
and then, that
(3.7) 4|H|2 = c|T |2 .
From Theorem 2.6 and Corollary 3.8, in the same way as in the proof of Theo-
rem 3.5, we have
1
(3.8) ∆|AH |2 = |∇AH |2 + 2(trace A3H ) − 4c|H|4 .
2
Next, since Corollary 3.17 implies that |AH |2 = c(2 − |T |2 )|H|2 is constant, and,
from relation (3.7), it follows
trace A3H = trace φ3H + 3|H|2 |φH |2 + 2|H|6 = 3c(2 − |T |2 )|H|4 − 4|H|6
= 6c|H|4 − 16|H|6 ,
equation (3.8) leads to
0 = |∇AH |2 + 8c|H|4 − 32|H|6 = |∇AH |2 + 8|H|4 (c − 4|H|2 )
= |∇AH |2 + 8c|H|4 |N |2 ,
and we get that |∇AH |2 = 0 and N = 0, which is a contradiction.
Therefore, the surface is a vertical cylinder Σ2 = π −1 (γ) over a pmc proper-bi-
harmonic√curve γ in S2 (c), i.e., Σ2 is a vertical cylinder over a circle with curvature
equal to c in S2 (c).
Summarizing, we have the following rigidity result.
Theorem 3.21 ([72]). Let Σ2 be a pmc proper-biharmonic surface in Sn (c) × R.
Then either
(1) Σ2 is a minimal surface of a small hypersphere Sn−1 (2c) ⊂ Sn (c); or
(2) Σ2 is (an open part of) a vertical
√ cylinder π −1 (γ), where γ is a circle in
S (c) with curvature equal to c, i.e., γ is biharmonic in S2 (c).
2
From Theorem 3.21, we can see that equation ∇AH = 0 holds for all proper-
biharmonic surfaces. From this point of view, the following result can be seen as a
generalization of this theorem for higher dimensional submanifolds. Before stating
the theorem, we have to mention that pmc proper-biharmonic submanifolds in Sn (c),
with ∇AH = 0, were classified in [15, Theorem 3.16].
Theorem 3.22 ([72]). If Σm is a pmc proper-biharmonic submanifold in Sn (c)×
R, with m ≥ 3, such that ∇AH = 0, then either
(1) Σm is a pmc proper-biharmonic submanifold in Sn (c), with ∇AH = 0; or
(2) Σm is (an open part of) a vertical cylinder π −1 (Σm−1 ), where Σm−1 is a
pmc proper-biharmonic submanifold in Sn (c) such that the shape operator
corresponding to its mean curvature vector field in Sn (c) is parallel.
Proof. Since ∇AH = 0, we have that [R(X, Y ), AH ] = 0 for any vector fields
X and Y tangent to Σm . On the other hand, since Σm is a pmc proper-biharmonic
submanifold, as we have seen in the proof of Theorem 3.14, we also have AH T = 0,
and then it follows that AH R(X, T )T = 0.
Now, let us consider a local orthonormal frame field {Ei }m m
i=1 on Σ , and {Em+1 =
H/|H|, . . . , En+1 } an orthonormal frame field in the normal bundle. We obtain, us-
ing (2.2), Lemma 2.5, and again AH T = 0,
m
X
0= hAH R(Ei , T )T, Ei i
i=1
n+1
X
= c(trace AH )|T |2 (1 − |T |2 ) + (trace(AH Aα ))hAα T, T i.
α=m+2
From the last equation of (3.2), we know that trace(AH Aα ) = 0, for any α ∈
{m + 2, . . . , n + 1}, and, therefore, we have
0 = cm|H|2 |T |2 (1 − |T |2 ),
i.e., either |T | = 0 or |T | = 1, which completes the proof.
5. Biconservative surfaces with parallel mean curvature in M n (c) × R

Let Σ2 be a non-minimal pmc surface in M̄ = M n (c) × R. For the sake of
simplicity we will only consider the cases c = ±1, i.e., M n (c) is either Sn or Hn .
From Corollary 3.3, we have that Σ2 is biconservative if and only if
(trace R̄(·, H)·))> = 0,
where H is the mean curvature vector field of our surface, which, using (2.1), is
equivalent to
(3.9) chH, N iT = 0,
where T and N are the tangent and the normal components of ξ, respectively.
As our result is of local nature, in the following, we will split our study as |T | = 0,
or |T | = 1, or |T | ∈ (0, 1) on Σ2 .
Case I. Let us assume that |T | = 0 at any point of Σ2 . This means that ξ is
orthogonal to our surface or, equivalently, that Σ2 lies in M n (c). Obviously, equation
(3.9) holds automatically in this case and Σ2 is biconservative. Moreover, Σ2 is a
pmc surface in a space form and these surfaces were classified in [136].
Case II. If |T | = 1 on the surface, then ξ is tangent to Σ2 at any point,

which means that Σ2 is a vertical cylinder over a circle with curvature κ = 2|H| in
M 2 (c) (see [6]). Moreover, H is orthogonal to ξ and then (3.9) implies that Σ2 is
biconservative in this case too.
Case III. Henceforth we shall assume that |T | ∈ (0, 1) at any point of the
surface Σ2 . Also assume that Σ2 is biconservative and orientable. We will see that,
in this case, our surface has no pseudo-umbilical points.
First, from Theorem 3.4, it follows that either H is umbilical everywhere and
then Σ2 lies in M n (c) (or equivalently |T | = 0), which is a contradiction, or H is
not umbilical on the surface, which implies that Σ2 lies in M 4 × R (see [6]).
6 0 on Σ2 , from (3.9), we have that H is orthogonal to ξ, that
Next, since |T | =
implies
X(hH, ξi) = 0,
⊥
or, equivalently, as ∇ H = 0 and ∇ξ ¯ = 0,
hAH T, Xi = 0,
for any vector field X tangent to the surface, so AH T = 0.
Now, let us consider the global, positive oriented orthonormal frame field {E1 =
T /|T |, E2 } on the surface and, since AH E1 = 0, we note that this frame field diag-
onalizes AH . From the equation of Ricci (1.5), we see, using the expression (2.1) of
the curvature tensor R̄ and the fact that H is parallel, that AH and AU commute for
any vector field U normal to Σ2 , which shows that {E1 , E2 } diagonalizes the second
fundamental form σ of our surface.
Next, consider the following decomposition of ξ
(3.10) ξ = cos θE1 + sin θE3 ,
where θ ∈ (0, π/2) is a local angle function, and {E3 = N/|N |, E4 , E5 = H/|H|} is
a global orthonormal frame field in the normal bundle.
First, we will prove the following lemma.
Lemma 3.23 ([71]). The following equations hold on the surface Σ2 :
(1) ∇E1 = ∇E2 = 0;
(2) θ = constant;
(3) ∇⊥E1 E3 = − cot θσ(E1 , E1 );
(4) ∇⊥E2 E3 = 0;
(5) A3 = AE3 = 0;
(6) ∇⊥E1 (σ(E2 , E2 )) = −c cos θ sin θE3 ;
(7) ∇⊥E2 (σ(E2 , E2 )) = 0.
Moreover, we have c = 1, i.e., Σ2 lies in S4 × R, and |σ(E1 , E1 )| = sin θ.
Proof. From (3.10), since ∇ ¯ E ξ = 0, we get
1
¯ E E1 + E1 (θ) cos θE3 + sin θ∇
−E1 (θ) sin θE1 + cos θ∇ ¯ E E3 = 0
1 1
and then, since {E1 , E2 } diagonalizes σ,

(3.11) ∇E1 E1 = ∇E1 E2 = 0,
(3.12) A3 E1 = AE3 E1 = −E1 (θ)E1

and
∇⊥
E1 E3 = − cot θ(σ(E1 , E1 ) + E1 (θ)E3 ).
¯ E ξ = 0, one obtains ∇⊥ E3 = 0,
In the same way, from ∇ 2 E2
(3.13) E2 (θ) = 0
and
(3.14) cos θ∇E2 E1 − sin θA3 E2 = 0.
Next, we will compute ∇⊥ ⊥
E2 (σ(E2 , E2 )) and ∇E1 (σ(E2 , E2 )). Using the Codazzi
equation (1.4) of Σ2 in M̄ , the expression (2.1) of the curvature tensor R̄, and
equations (3.11), since ∇⊥ H = 0 and {E1 , E2 } diagonalizes σ, we have
∇⊥ ⊥ ⊥
E2 (σ(E2 , E2 )) = −∇E2 (σ(E1 , E1 )) = −((∇E2 σ)(E1 , E1 ) + 2σ(E1 , ∇E2 E1 ))
= −((∇⊥
E1 σ)(E1 , E2 ) + 2σ(E1 , ∇E2 E1 )) + (R̄(E2 , E1 )E1 )
⊥
= −2σ(E1 , ∇E2 E1 )) = −2h∇E2 E1 , E2 )σ(E1 , E2 )

=0
and
(3.15) ∇⊥ ⊥
E1 (σ(E2 , E2 )) = (∇E1 σ)(E2 , E2 ) − 2σ(E2 , ∇E1 E2 )
= (∇⊥
E2 σ)(E1 , E2 ) + (R̄(E1 , E2 )E2 )
⊥
= −σ(E2 , ∇E2 E1 ) − σ(E1 , ∇E2 E2 ) − c cos θ sin θE3 .

Now, since AH E1 = 0, we know that hσ(E2 , E2 ), Hi = 2|H|2 and then
E1 (hσ(E2 , E2 ), Hi) = 0.
The facts that H is parallel and orthogonal to E3 , using (3.15), lead to
hσ(E2 , ∇E2 E1 ) + σ(E1 , ∇E2 E2 ), Hi = 0,
or, equivalently, since AH E1 = 0,
h∇E2 E1 , E2 i|H|2 = 0,
which means that
(3.16) ∇E2 E1 = ∇E2 E1 = 0.
From equations (3.11) and (3.16), we see that ∇E1 = ∇E2 = 0. Moreover, since
trace A3 = 0, from (3.12) and (3.14), we have A3 E2 = E1 (θ)E2 = 0 and then A3 = 0
and E1 (θ) = 0. From (3.13), it follows that the function θ is constant.
Finally, the shape operator A of the surface is given, with respect to {E1 , E2 },
by

λ 0 0 0
A3 = A N = 0, A4 = , A5 = A H = ,
|N | 0 −λ |H| 0 2|H|
where λ is a smooth function on Σ2 , and we have σ(E1 , E1 ) = λE4 . Then, from the
Gauss equation (1.3) of Σ2 in M̄ , we obtain the Gaussian curvature K of Σ2 as
K = c sin2 θ − λ2 .
Since the equations ∇E1 = ∇E2 = 0 imply that Σ2 is flat, it follows that
λ2 = c sin2 θ, which means that c > 0 and, therefore, c = 1, which completes the
proof.
Remark 3.24. We note that if Σ2 is a pmc biconservative surface in M 4 (c) × R

with |T | ∈ (0, 1), then it lies in a totally geodesic submanifold M 3 (c) × R if and only
if A4 = 0. But, from the Gauss equation of the surface, we get that there are no
pmc biconservative surfaces in M 3 (c) × R with |T | ∈ (0, 1).
Next, we consider the immersion of S4 × R in R5 × R and denote by ∇ e the Levi-
5
Civita connection on R × R. Then the integral curves of E1 and E2 , thought as
curves in R5 × R, are characterized by the following two lemmas, that can be proved
by using Lemma 3.23.
Lemma 3.25 ([71]). The integral curves δ of E1 are helices in R5 × R with
curvatures p p
κ1 = sin θ 1 + sin2 θ and κ2 = cos θ 1 + sin2 θ,
where θ = constant ∈ (0, π/2).
Remark 3.26. We note that two equations that appear in the proof of the
previous lemma lead to ∇
eE ∇
1
e E E2 = 0, when c = 1.
2
5
Lemma 3.27p ([71]). The integral curves γ of E2 are plane circles in R × R with
2
curvature κ = 1 + 4|H|2 + sin θ, where θ = constant ∈ (0, π/2).
Now, we can state the main result of this section.
Theorem 3.28 ([71]). Let Σ2 be a pmc biconservative surface with mean cur-
vature vector field H in M̄ = M n (c) × R, c = ±1 and H 6= 0. Then either
(1) Σ2 either is a minimal surface of an umbilical hypersurface of M n (c), or a
cmc surface in a 3-dimensional umbilical submanifold of M n (c); or
(2) Σ2 is a vertical cylinder over a circle in M 2 (c) with curvature κ = 2|H|; or
(3) Σ2 lies in S4 × R ⊂ R5 × R and, as a surface in R5 × R, is locally given by
1
X(u, v) = {C3 + sin θ(D1 cos(au) + D2 sin(au))} + (u cos θ + b)ξ
a
1
+ (C1 (cos v − 1) + C2 sin v),
κ
p
wherep θ ∈ (0, π/2) is a constant, a = 1 + sin2 θ, b is a real constant,
κ = 1 + 4|H|2 + sin2 θ, C1 and C2 are two constant orthonormal vectors
in R5 × R such that C1 ⊥ ξ and C2 ⊥ ξ, C3 is a unit constant vector
such that hC3 , C1 i = a/κ ∈ (0, 1), C3 ⊥ C2 , and C3 ⊥ ξ, and D1 and
D2 are two constant orthonormal vectors in the orthogonal complement of
span{C1 , C2 , C3 , ξ} in R5 × R.
Proof. We only have to study the case when the surface Σ2 is not pseudo-
umbilical and |T | ∈ (0, 1). In order to do that, we will use the same method employed
in [27] to study biconservative surfaces in space forms.
We consider again the local orthonormal frame field {E1 = T /|T |, E2 } and let
γ be an integral curve of E2 parametrized by arc-length.
p Then, from Lemma 3.27,
we know that γ is a circle with curvature κ = 1 + 4|H|2 + sin2 θ in R5 × R and,
therefore, it can be written as
(3.17) γ(s) = c0 + c1 cos(κs) + c2 sin(κs), c0 , c1 , c2 ∈ R5 × R,
where |c1 | = |c2 | = 1/κ and hc1 , c2 i = 0.
At an arbitrary point p0 ∈ Σ2 we consider δ(u) an integral curve of E1 , with

δ(0) = p0 , and the flow φ of E1 near p0 . We note that δ(u) is a helix characterized
in Lemma 3.25. Now, for all u ∈ (−ω, ω) and s ∈ (−, ), we have
φδ(u) (s) = c0 (u) + c1 (u) cos(κs) + c2 (u) sin(κs),
with
1
δ(u) = c0 (u) + c1 (u), |c1 (u)| = |c2 (u)| =
, hc1 (u), c2 (u)i = 0,
κ
and, therefore, the surface can be parametrized locally by
X(u, s) = φδ(u) (s).
Next, X(u, s) can be reparametrized using u and v = κs as the new parameters,
with u ∈ (−ω, ω) and v ∈ (−κ, κ), and we have
1
X(u, v) = c0 (u) + (C1 (u) cos v + C2 (u) sin v),
κ
where C1 (u) = κc1 (u) and C2 (u) = κc2 (u).
Since at v = 0 the integral curves of E2 start from δ, we have
1
δ(u) = X(u, 0) = c0 (u) + C1 (u)
κ
and then
1
(3.18) X(u, v) = δ(u) + (C1 (u)(cos v − 1) + C2 (u) sin v).
κ
From (3.17) it follows that C2 = κc2 = γ 0 (0) = E2 (γ(0)), that is
C2 (u) = E2 (δ(u)),
and also −κ2 c1 = γ 00 (0) = (∇
e E E2 )(γ(0)), which gives
2
1 e
C1 (u) = κc1 (u) = − (∇ E2 E2 )(δ(u)).
κ
Now, using Lemma 3.23 and Remark 3.26, we have
dC1 1e e dC2
=− ∇ E1 ∇E2 E2 = 0 and =∇e E E2 = 0,
1
du κ du
which means that C1 and C2 are constant orthonormal vectors and that the image
of parametrization (3.18) is given by a 1-parameter family of circles centered in
δ(u) − (1/κ)C1 and passing through the points of δ(u) lying in planes parallel to
the one spanned by C1 and C2 . Moreover, from Lemma 3.23, one also obtains that
C1 ⊥ ξ and C2 ⊥ ξ.
Next, we will determine the explicit equation of δ(u). In order to do that, let us
consider the vector field
C(u) = δ 00 (u) + (1 + sin2 θ)η(δ(u))
along δ(u). It is then easy to verify, using Remark 3.26, that C 0 (u) = 0, which
means that C(u) = C is a constant vector. p From Lemmas 3.23, 3.25, and 3.27, we
also get that hC, C1 i = a /κ, where a = 1 + sin2 θ, C ⊥ C2 , C ⊥ ξ, and |C| = a.
2
Moreover, C, C1 , and C2 are linearly independent.

Next, consider δ1 (u) = δ(u) − hδ(u), ξiξ. Since ∇
e E ξ = 0, it follows that
1
δ10 (u) = E1 − cos θξ

e E E1 = δ 00 (u) =
and then |δ10 (u)| = sin θ. Differentiating δ10 (u) along δ(u), since ∇ 1
2 2
C − a η(δ(u)) = C − a δ1 (u), we can see that δ1 (u) satisfies
δ100 (u) + a2 δ1 (u) = C,
that shows that
1 1
δ1 (u) = 2
C + (F1 cos(au) + F2 sin(au)),
a a
where F1 and F2 are two constant vectors in R5 × R. Since δ10 (u) is orthogonal to
ξ, we have that F1 ⊥ ξ and F2 ⊥ ξ. Also, from |δ10 (u)| = sin θ, one obtains that
F1 ⊥ F2 and |F1 | = |F2 | = sin θ. Then, considering C3 = (1/a)C, D1 = (1/ sin θ)F1 ,
and D2 = (1/ sin θ)F2 , we can write
1
δ1 (u) =(C3 + sin θ(D1 cos(au) + D2 sin(au))),
a
where C3 , D1 and D2 are unit constant vectors such that D1 ⊥ D2 . It follows that
δ10 (u) = −D1 sin(au) + D2 cos(au),
which, taking into account that δ10 = E1 −cos θξ is orthogonal to C1 , C2 , and C3 , im-
plies that D1 and D2 are vectors in the orthogonal complement of span{C1 , C2 , C3 }
in R5 × R.
Finally, since (d/du)(hδ(u), ξi) = hE1 , ξi = cos θ along δ(u), we have hδ(u), ξi =
u cos θ + b, where b is a real constant. Hence, we conclude that δ(u) is given by
1
δ(u) = {C3 + sin θ(D1 cos(au) + D2 sin(au))} + (u cos θ + b)ξ,
a
Remark 3.29. We note that surfaces given by the third case of Theorem 3.28
lie in the Riemannian product of a small hypersphere of S4 with R. In order to see
this, let us consider X1 (u, v) = X(u, v) − hX(u, v), ξiξ and the constant vector C e
in R5 , orthogonal to ξ, given by C e = (1/a)C3 − (1/κ)C1 . Then, is easy to verify
that hX1 (u, v) − C, e = 0 and that X1 (u, v) lies in S4 , which shows that X1 (u, v)
e Ci
actually lies in S4 ∩π, where π is a hyperplane of R5 that passes through C
e such that
e is orthogonal to π. Moreover, since |X1 (u, v) − C|2 2 2 2
e = (a + κ sin θ)/a2 κ2 , we get
C p p
e a2 + κ2 sin2 θ/aκ) × R, where S3 (C,
that X(u, v) lies in S3 (C, e a2 + κ2 sin2 θ/aκ)
is
p the 3-dimensional sphere in the hyperplane π, centered in C and with radius
e
2
a2 + κ2 sin θ/aκ.
Remark 3.30. We note that, since all non-minimal pmc biharmonic surfaces
in Sn × R that do not lie in Sn are vertical cylinders (Theorem 3.21), the surfaces
described in the third case of Theorem 3.28 are not biharmonic.
6. Biconservative surfaces with constant mean curvature in M 3 (c) × R

As we have seen in Theorem 3.28, there are no pmc biconservative surfaces in
M 3 (c) × R, where c = ±1, that do not lie in M 3 (c) nor are vertical cylinders. It
is then interesting to see if there is possible to find examples of cmc biconservative
surfaces Σ2 with |T | ∈ (0, 1) in these spaces.
We first note that it can be easily verified that a cmc surface in M 3 (c) × R
with |T | = 0 is biconservative since it actually lies in M 3 (c), while, in the case
when |T | = 1, we easily get the following proposition. In both situations, the mean
curvature vector field is orthogonal to ξ.
Proposition 3.31 ([71]). Let Σ2 be a cmc surface with |T | = 1 and constant
mean curvature |H| in M 3 (c) × R. Then Σ2 is a vertical cylinder over a curve in
M 3 (c) with constant first curvature κ1 = 2|H|. Moreover, Σ2 is a biconservative
surface.
When |T | ∈ (0, 1), we have the following characterization of cmc biconservative
surfaces in M 3 (c) × R whose mean curvature vector field H is orthogonal to ξ.
Theorem 3.32 ([71]). Let Σ2 be a cmc biconservative surface in M 3 (c) × R,
c 6= 0, with mean curvature vector field H 6= 0 orthogonal to ξ and |T | ∈ (0, 1). Then
Σ2 is flat and it is locally given by X = X(u, v), where X : D ⊂ R2 → M 3 (c) × R is
an isometric immersion, D is an open set in R2 , and either
(1) Σ2 is pseudo-umbilical, c < 0, |H|2 = −c(1 − |T |2 ), the integral curve
of Xu√is a helix such that hXu , ξi = |T |, with curvatures κ11 = |H| and
κ12 = −c|T |, and the integral curve of Xv is a circle such that hXv , ξi = 0,
with curvature κ21 = |H|; or
(2) |H|2 > −c(1 − |T |2 ) and the integral curves of Xu and Xv are helices in
M 3 (c) × R satisfying
hXu , ξi = a and hXv , ξi = b,
where a and b are two real constants such that
0 < a2 + b2 = |T |2 < 1 and |H|2 + c(1 − a2 − b2 ) > 0,
and with curvatures
p |a|
κ11 = |H| + |H|2 + c(1 − a2 − b2 ), κ12 = √ κ11
2
1−a −b2
and
p |b|
κ21 = |H| − |H|2 + c(1 − a2 − b2 ), κ22 = √ κ21 ,

1 − a2 − b2
respectively.
Proof. Since our surface is biconservative, it follows, from Theorem 3.28, that
it cannot have parallel mean curvature vector field. Therefore, ∇⊥ H 6= 0, which
means that there exists an open subset U ⊂ Σ2 such that ∇⊥ H 6= 0 at any point
p ∈ U . Let us now consider a local orthonormal frame field {E1 , E2 } on U and an
orthonormal frame field {E3 = H/|H|, E4 = N/|N |} in the normal bundle of Σ2 .
Then {E1 , E2 , E3 , E4 } can be extended to local orthonormal frame field on an open
subset of M 3 (c) × R. Denote by ωB A the corresponding connection 1-forms on this
subset given by
∇¯ X EA = ωAB
(X)EB .
Then, from Corollary 3.3, we get that the biconservative equation becomes
trace A∇⊥
· H
(·) = |H|(ω34 (E1 )A4 E1 + ω34 (E2 )A4 E2 ) = 0,
that is equivalent to
(
ω34 (E1 )hA4 E1 , E1 i + ω34 (E2 )hA4 E2 , E1 i = 0
(3.19)
ω34 (E1 )hA4 E1 , E2 i + ω34 (E2 )hA4 E2 , E2 i = 0.
Since ∇⊥ H 6= 0, we have (ω34 (E1 ))2 + (ω34 (E2 ))2 6= 0, which, using (3.19), implies
that
hA4 E1 , E1 ihA4 E2 , E2 i − hA4 E2 , E1 ihA4 E1 , E2 i = 0.
Next, the fact that E4 is orthogonal to H shows that trace A4 = 0 and then,
since A4 is symmetric, one obtains
|A4 |2 = −2hA4 E1 , E1 ihA4 E2 , E2 i + 2hA4 E2 , E1 ihA4 E1 , E2 i = 0,
i.e., A4 = 0. We note that, since ∇ξ ¯ = 0 and N = |N |E4 , we also have ∇T = AN = 0
and |T |, |N | ∈ (0, 1) are constant.
We know, from Theorem 3.4, that H either is umbilical at any point of Σ2 , and
then Σ2 is pseudo-umbilical, or H is umbilical only on a closed set without interior
points. In the second case, H is not umbilical on an open dense connected set W .
Let us first treat the case when our surface is pseudo-umbilical. Then, {E1 , E2 }
diagonalizes A3 and, moreover, since |T | ∈ (0, 1), we can choose E1 = T /|T |, which
implies that ∇E1 = ∇E2 = 0.
Using the Codazzi equation (1.4) of Σ2 in M 3 (c) × R, first with X = E1 , Y =
Z = E2 and then with X = Z = E1 , Y = E2 , and taking the inner product with
E4 , one obtains
c|T ||N |
(3.20) ω34 (E1 ) = − = constant and ω34 (E2 ) = 0.
|H|
Next, since hE3 , ξi = 0, we have h∇ ¯ E E3 , ξi = 0, which gives
1
|T ||H|
(3.21) ω34 (E1 ) = .
|N |
From (3.20) and (3.21), one sees that |H|2 = −c|N |2 , that implies c < 0 and,
using the Gauss equation (1.3), that the surface is flat. Moreover, one obtains
√
(3.22) ω34 (E1 ) = ± −c|T |.
As we have seen, we have ∇E1 = ∇E2 = 0 and then [E1 , E2 ] = 0, which means
that there exists a local parametrization X = X(u, v) of Σ2 such that Xu = E1 and
Xv = E2 .
In the following, we shall determine the curvatures of the integral curves γ1 and
γ2 of Xu and Xv , respectively.
From the first Frenet equation ∇ ¯ E E1 = κ1 X 1 of γ1 , since ∇
¯ E E1 = σ(E1 , E1 ) =
1 1 2 1
|H|E3 , it follows that the first curvature of γ1 is κ1 = |H| and X21 = E3 . The second
1
Frenet equation ∇ ¯ E X 1 = −κ1 E1 + κ1 X 1 , together with

1 2 1 2 3
∇ ¯ E E3 = −A3 E1 + ∇⊥
¯ E X21 = ∇ 4
E1 E3 = −|H|E1 + ω3 (E1 )E4
1 1
√
and (3.22), leads to κ12 = |ω34 (E1 )| = −c|T | and X31 = (ω34 (E1 )/|ω34 (E1 )|)E4 .
Finally, the third Frenet equation of γ1 is
4
¯ E X31 = ω3 (E1 ) ∇
∇ ¯ E E4 = −|ω34 (E1 )|E3 = −κ12 X21 .
1
|ω34 (E1 )| 1
¯ E E2 = κ2 X 2 of γ2 and ∇
The first Frenet equation ∇ ¯ E E2 = σ(E2 , E2 ) = |H|E3
2 1 2 2
2 2
give κ1 = |H| and X2 = E3 . Then, the second Frenet equation ∇ ¯ E X 2 = −κ2 E2 +
2 2 1
2 2 2
κ2 X3 and (3.20) imply that κ2 = 0, which shows that γ2 is a circle.
Let us now consider the case when Σ2 is not pseudo-umbilical.
First, we choose E1 and E2 such that A3 Ei = λi Ei , i ∈ 1, 2, and λ1 > λ2 . Since

A4 = 0, we have
σ(E1 , E1 ) = λ1 E3 , σ(E1 , E2 ) = 0, σ(E2 , E2 ) = λ2 E3
and also (
∇E1 E1 = ω12 (E1 )E2 , ∇E1 E2 = −ω12 (E1 )E1
∇E2 E1 = ω12 (E2 )E1 , ∇E2 E2 = −ω12 (E2 )E1 .
Next, we again use the Codazzi equation (1.4) with X = Z = E1 and Y = E2 ,
to obtain, taking the inner product first with E3 and then with E4 ,
(3.23) E2 (λ1 ) = (λ1 − λ2 )ω12 (E1 )
and
(3.24) λ1 ω34 (E2 ) + chT, E2 i|N | = 0.
In the same way, this time taking X = E1 and Y = Z = E2 , we get
(3.25) E1 (λ2 ) = (λ1 − λ2 )ω12 (E2 )
and
(3.26) λ2 ω34 (E1 ) + chT, E1 i|N | = 0.
¯ E E3 , ξi = 0 and h∇
From hE3 , ξi = 0 we have h∇ ¯ E E3 , ξi = 0, that are
1 2
(3.27) λ1 hT, E1 i − ω34 (E1 )|N | = 0

and
(3.28) λ2 hT, E2 i − ω34 (E2 )|N | = 0.
Now, from (3.24), (3.26), (3.27), and (3.28) it follows
λ1 λ2 + c|N |2 = 0,
which, using the Gauss equation (1.3), shows that Σ2 is flat. Moreover, since λ1 +
λ2 = 2|H|, we get
p
(3.29) λi = |H| ± |H|2 + c|N |2 = constant, i ∈ {1, 2}.
From (3.24), (3.26), and (3.29), one sees that
 √
4 ca 1 − a2 − b2
ω3 (E1 ) = −

 p
|H| − |H| 2 2 2
√ + c(1 − a − b )

4 cb 1 − a2 − b2
ω3 (E2 ) = −


 p ,
|H| + |H|2 + c(1 − a2 − b2 )
where a = hT, E1 i and b = hT, E2 i.
The fact that λ1 and λ2 are constants, together with (3.23) and (3.25), leads to
ω12 (E1 ) = ω12 (E2 ) = 0, i.e., ∇E1 = ∇E2 = 0. Since ∇T = 0, we can also see that
ω34 (Ei ), i ∈ {1, 2}, are constant and then that the Ricci equation does not provide
any other supplementary information about Σ2 .
Finally, since [E1 , E2 ] = 0, there exists a local parametrization X = X(u, v) of
Σ2 such that Xu = E1 and Xv = E2 . We conclude by computing the curvatures of
the integral curves of Xu and Xv , in the same way as in the case when the surface
is pseudo-umbilical.
Remark 3.33. We note that surfaces described in Theorem 3.32 do exist, as it

can be easily verified by using [98, Theorem 2.2].
Computing the normal part of the bitension field τ2 of surfaces in Theorem 3.32,
we get the following result.
Theorem 3.34 ([71]). If Σ2 is a cmc biharmonic surface in M 3 (c) × R, c 6= 0,
with mean curvature vector field H 6= 0 orthogonal to ξ and |T | ∈ (0, 1), then c > 0,
b2 > a2 , and Σ2 is one of the non-pseudo-umbilical cmc biconservative surfaces in
Theorem 3.32, with
c(1 − a2 − b2 )(b2 − a2 )2
|H|2 = .
4(1 − a2 )(1 − b2 )
7. Biconservative surfaces with constant mean curvature in Hadamard

manifolds
In order to prove some compactness results for cmc biconservative surfaces in
M n (c) × R, with c < 0, we will work in the more general setting where the ambient
space is a Hadamard manifold, i.e., a Riemannian manifold that is complete simply-
connected and has non-positive sectional curvature everywhere.
We will begin by showing that a cmc biconservative surface in a Riemannian
manifold satisfies a Simons type equation.
Theorem 3.35 ([71]). Let Σ2 be a non-minimal cmc biconservative surface in
a Riemannian manifold M̄ with mean curvature vector field H and shape operator
A. Then
1
∆|φH |2 = 2K|φH |2 + |∇φH |2 ,
2
where φH = AH − |H|2 I is the traceless part of AH and K is the Gaussian curvature
of the surface.
Proof. In our case, where Σ2 is a biconservative surface, using isothermal co-
ordinates (u, v) on the surface, we get, by a straightforward computation,
∂ ∂ 3 ∂ ∂ ∂ ∂
∇ ∂ AH − ∇ ∂ AH = − (|H|2 ) + (|H|2 ) ,
∂u ∂v ∂v ∂u 2 ∂v ∂u ∂u ∂v
which, since Σ2 has constant mean curvature, shows that AH , and then φH , satisfies
the Codazzi equation. Since φH is symmetric and traceless, we conclude using
equation (2.46) with S = φH .
Corollary 3.36 ([71]). Let Σ2 be a cmc biconservative surface in a Riemannian
manifold M̄ and assume that Σ2 is compact and K ≥ 0. Then ∇AH = 0 and the
surface is pseudo-umbilical or flat.
Proof. From Theorem 3.35, we get that Σ2 (2K|φH |2 + |∇φH |2 ) dv = 0 and,
R
since 2K|φH |2 + |∇φH |2 ≥ 0, one obtains that ∇AH = ∇φH = 0 and, at any point
on the surface, K = 0 or φH = 0. We conclude using Theorem 3.4, that shows
that φH either vanishes at any point of Σ2 , or only on a closed set without interior
points. Hence, if the surface is not pseudo-umbilical, it follows that K = 0 on an
open dense set in Σ2 , and then the Gaussian curvature vanishes everywhere.
Corollary 3.37 ([71]). Let Σ2 be a non-minimal cmc biconservative surface

in a Riemannian manifold M̄ , with sectional curvature bounded from below by a
constant K0 , such that µ = supΣ2 (|σ|2 − (1/|H|)2 |AH |2 ) < +∞. Then
−∆|φH | ≤ a|φH |3 + b|φH |,
where a and b are constants depending on K0 , |H|, and µ.
Proof. Let {E3 = H/H, E4 , . . . , En } be a local orthonormal frame field in
the normal bundle, where n is the dimension of the ambient space M̄ , and denote
Aα = AEα . Then, from the Gauss equation (1.3) of Σ2 in M̄ , we obtain the following
expression of the Gaussian curvature of Σ2
n
X
K = hR̄(E1 , E2 )E2 , E1 i + det Aα
α=3
1 1
= hR̄(E1 , E2 )E2 , E1 i + |H|2 − |φH |2 − (|σ|2 − |A3 |2 ),
2|H|2 2
where {E1 , E2 } is a local orthonormal frame field on the surface. Since by hypothesis
we have hR̄(E1 , E2 )E2 , E1 i ≥ K0 , we get that
1 µ
(3.30) K ≥ K0 + |H|2 − |φH |2 − ,
2|H|2 2
and then, from Theorem 3.35, one obtains

1 1 µ
∆|φH |2 ≥ 2 K0 + |H|2 − |φ H | 2
− |φH |2 + |∇φH |2 .
2 2|H|2 2
Since |∇|φH || ≤ |∇φH |, we easily get that
1
|φH |3 − 2K0 + 2|H|2 − µ |φH |,

−∆|φH | ≤ 2
|H|
Now, let us consider a cmc surface Σ2 in a Hadamard manifold. Such a surface

satisfies the same Sobolev inequality mentioned in the previous chapter, i.e.,
(3.31) ∀f ∈ C0∞ (Σ), ||f ||2 ≤ A||∇f ||1 + B||f ||1 ,
where ||f ||p = ( Σ2 |f |p dv)1/p is the Lp -norm of the function f and A and B are
R
constants that depends only on the mean curvature |H| of the surface (see [84]).
In the following, we will use the same notations as in the section on pmc surfaces
with finite total curvature in the previous chapter.
Theorem 3.38 ([71]). Let Σ2 be a complete non-minimal cmc biconservative
surface in a Hadamard manifold M̄ , with sectional curvature bounded from below by
a constant K0 < 0, such that the norm of its second fundamental form σ is bounded
and
Z
(3.32) |φH |2 dv < +∞.
Σ2
Then the function u = |φH | goes to zero uniformly at infinity. More exactly, there
exist positive constants C0 and C1 , depending on K0 , |H|, and µ = supΣ2 (|σ|2 −
Z
(1/|H|)2 |AH |2 ), and a positive radius RΣ2 , determined by C1 u2 dv ≤ 1,
E(RΣ2 )
such that Z
||u||∞,E(2R) ≤ C0 u2 dv,
Σ2
for all R ≥ RΣ2 . Moreover, there exist some positive
Z constants D0 and E0 , depending
on K0 , |H|, and µ, such that the inequality u2 dv ≤ D0 implies
Σ2
Z
||u||∞ ≤ E0 u2 dv.
Σ2
Proof. Since the function u = |φH | satisfies the Sobolev inequality (3.31) and
the Simons type inequality in Corollary 3.37, we again work as in the proof of [18,
Theorem 4.1] and come to the conclusion.
We note that, when n = 2, we have µ = 0 and then it is easy to see that (3.32)
implies that |σ| is bounded. Therefore, we have the following corollary.
Corollary 3.39 ([71]). Let Σ2 be a complete non-minimal cmc biconservative
surface in a 3-dimensional Hadamard manifold M̄ , with sectional curvature bounded
from below by a constant K0 < 0, such that
Z
|φH |2 dv < +∞.
Σ2
Then the function u = |φH | goes to zero uniformly at infinity. More exactly, there
exist positive constants ZC0 and C1 , depending on K0 and |H|, and a positive radius
RΣ2 , determined by C1 u2 dv ≤ 1, such that
E(RΣ2 )
Z
||u||∞,E(2R) ≤ C0 u2 dv,
Σ2
for all R ≥ RΣ2 . Moreover, there exist some

Z positive constants D0 and E0 , depending
on K0 and |H|, such that the inequality u2 dv ≤ D0 implies
Σ2
Z
||u||∞ ≤ E0 u2 dv.
Σ2
Next, we will use Theorem 3.38 to prove a compactness result for cmc biconser-
vative surfaces in Hadamard manifolds.
Theorem 3.40 ([71]). Let Σ2 be a complete non-minimal cmc biconservative
surface in a Hadamard manifold M̄ , with sectional curvature bounded from below by
a constant K0 < 0, such that the norm of its second fundamental form σ is bounded,
Z
|φH |2 dv < +∞,
Σ2
and |H|2 > (µ − 2K0 )/2, where µ = supΣ2 (|σ|2 − (1/|H|2 )|AH |2 ). Then Σ2 is
compact.
Proof. Using inequality (3.30) and Theorem 3.38, we have that the superior
limit at infinity of the Gaussian curvature K of Σ2 is positive. It follows that the
negative part K − of K has compact support and, therefore,
Z
|K − | dv < +∞,
Σ2
which implies, using [138, Theorem 1], that also the positive part K + of K satisfies
Z
K + dv < +∞.
Σ2
Next, since outside a compact set Ω we have K + ≥ k/2 > 0, where
µ
k = K0 + |H|2 − ,
2
2
it follows that Vol(Σ \Ω) < +∞. Since the volume of a complete non-compact
surface in a Hadamard manifold is infinite (see [79]), we conclude that Σ2 is compact.

When n = 2, we use Theorem 3.38 and Corollary 3.39 to prove the last result of
this section.
Corollary 3.41 ([71]). Let Σ2 be a complete non-minimal cmc biconservative
surface in a 3-dimensional Hadamard manifold M̄ , with sectional curvature bounded
from below by a constant K0 < 0, such that
Z
|φH |2 dv < +∞,
Σ2
and |H|2 > −K0 . Then Σ2 is compact.
CHAPTER 4
Biharmonic Submanifolds in Sasakian Space Forms
1. Introduction
We begin this chapter with a study on proper-biharmonic curves in Sasakian
space forms. First, we classify proper-biharmonic Legendre curves (Theorems 4.1,
4.4, 4.5, and 4.7), prove, amongst other properties of these curves, that, when the
ambient space is 5-dimensional, they must be helices (Proposition 4.11) and then find
the explicit equations of some of these curves in odd-dimensional spheres endowed
with their usual and deformed Sasakian structures (Theorems 4.14, 4.17, and 4.18).
We then consider non-Legendre curves and, under some additional hypotheses, study
their biharmonicity. We obtain some classification results (Theorems 4.19, 4.20, 4.22,
and 4.24), as well as the explicit equations of some of these curves in R2n+1 (−3)
(Theorems 4.25 and 4.26). Finally, we prove a non-existence result for biharmonic
Legendre curves in S7 (1) endowed with its 3-Sasakian structure (Theorem 4.27).
Our next result provides a method to obtain proper-biharmonic anti-invariant
submanifolds of dimension m + 1, from proper-biharmonic integral submanifolds
of dimension m in a Sasakian space form (Theorem 4.28). We also present some
applications of this theorem (Theorems 4.29 and 4.35).
Using the classification of homogeneous real hypersurfaces in CP n ([130]), we
also classify proper-biharmonic Hopf cylinders over them, obtained via the Boothby-
Wang fibration, in a Sasakian space form. Thus, we prove that, while there are five
different classes of homogeneous real hypersurfaces in CP n , only A1 and A2 type
hypersurfaces produce proper-biharmonic Hopf cylinders (Theorems 4.45 and 4.47).
In the last section, we consider 3-dimensional integral C-parallel submanifolds
in a 7-dimensional Sasakian space form. First, we find their explicit equation, when
they are flat and lie in S7 (c) (Theorem 4.50). Next, we study their biharmonicity
and classify those that are proper-biharmonic (Theorem 4.59). Then we specialize
this result to the case when the ambient space is S7 (c) (Theorem 4.60, Corollary
4.61).
2. Biharmonic Legendre curves in Sasakian space forms

Let (N 2n+1 , ϕ, ξ, η, h, i) be a Sasakian space form with constant ϕ-sectional cur-
vature c and γ : I → N a Legendre Frenet curve of osculating order r with Frenet
frame field {E1 = T = γ 0 , E2 , . . . , Er }. Since
∇3T T = (−3κ1 κ01 )E1 + (κ001 − κ31 − κ1 κ22 )E2 + (2κ01 κ2 + κ1 κ02 )E3 + κ1 κ2 κ3 E4 ,
where ∇ is the Levi-Civita on N , and

(c + 3)κ1 3(c − 1)κ1
RN (T, ∇T T )T = − E2 − hE2 , ϕ, T iϕT,
4 4
95
96 Biharmonic Submanifolds in Sasakian Space Forms
where RN is the curvature tensor field of N , we obtain the expression of the bitension
vector field
(4.1) τ2 (γ) =∇3T T − RN (T, ∇T T )T
(c + 3)κ1
=(−3κ1 κ01 )E1 + κ001 − κ31 − κ1 κ22 + E2
4
3(c − 1)κ1
+ (2κ01 κ2 + κ1 κ02 )E3 + κ1 κ2 κ3 E4 + hE2 , ϕT iϕT.
4
In the following, we shall solve the biharmonic equation τ2 (γ) = 0. As suggested
by the form of the last term of τ2 (γ) we must do a case by case analysis.
Case I: c = 1. In this case, from (4.1), it follows that γ is proper-biharmonic if
and only if
κ1 = constant > 0, κ2 = constant, κ21 + κ22 = 1, and κ2 κ3 = 0.
One obtains the following result.
Theorem 4.1 ([68]). Let γ : I → N be a Legendre Frenet curve of osculating
order r in N 2n+1 (1). If n ≥ 2, then γ is proper-biharmonic if and only if it is either
a circle with κ1 = 1, or a helix with κ21 + κ22 = 1.
Remark 4.2. If n = 1 and γ is a non-geodesic Legendre curve, we have ∇T T =
±κ1 ϕT and then E2 = ±ϕT and ∇T E2 = ±∇T ϕT = ±(ξ ∓ κ1 T ) = −κ1 T ± ξ.
Therefore κ2 = 1 and γ cannot be biharmonic.
Case II: c 6= 1, E2 ⊥ ϕT . From (4.1) we obtain that γ is proper-biharmonic if
and only if
c+3
κ1 = constant > 0, κ2 = constant, κ21 + κ22 = , and κ2 κ3 = 0.
4
Before stating our next theorem, we need the following lemma which imposes a
restriction on the dimension of the ambient space N 2n+1 (c).
Lemma 4.3 ([68]). Let γ be a Legendre Frenet curve of osculating order 3 such
that E2 ⊥ ϕT . Then {T = E1 , E2 , E3 , ϕT, ξ, ∇T ϕT } is linearly independent at any
point, and, therefore, n ≥ 3.
order r in N 2n+1 (c), c 6= 1, such that E2 ⊥ ϕT . We have
(1) If c ≤ −3, then γ is biharmonic if and only if it is a geodesic.
(2) If c > −3, then γ is proper-biharmonic if and only if either
(a) n ≥ 2 and γ is a circle with κ21 = c+3
4 . In this case {E1 , E2 , ϕT, ξ} are
linearly independent; or
(b) n ≥ 3 and γ is a helix with κ21 +κ22 = c+3
4 . In this case {E1 , E2 , E3 , ϕT,
ξ, ∇T ϕT } are linearly independent.
Case III: c 6= 1, E2 k ϕT . In this case, from (4.1), γ is proper-biharmonic if
and only if
κ1 = constant > 0, κ2 = constant, κ21 + κ22 = c, and κ2 κ3 = 0.
We can assume that E2 = ϕT . Then we have ∇T T = κ1 E2 = κ1 ϕT and
∇T E2 = ∇T ϕT = ξ − κ1 T . Therefore, E3 = ξ and κ2 = 1, and then one obtains
∇T E3 = ∇T ξ = −ϕT = −E2 .
Biharmonic Submanifolds in Sasakian Space Forms 97
Theorem 4.5 ([68]). Let γ : I → N a Legendre Frenet curve of osculating order

r in N 2n+1 (c), c 6= 1, such that E2 k ϕT . Then {T, ϕT, ξ} is the Frenet frame field
of γ and we have
(1) If c < 1, then γ is biharmonic if and only if it is a geodesic.
(2) If
√ c > 1, then γ is proper-biharmonic if and only if it is a helix with κ1 =
c − 1 and κ2 = 1.
Remark 4.6. If n = 1, for any Legendre curve, we have E2 k ϕT , and obtain
J. Inoguchi’s result in [89].
Case IV: c 6= 1 and hE2 , ϕT i is not constant 0, 1, or −1. One can check
that, in this case, 4 ≤ r ≤ 2n + 1, n ≥ 2, and ϕT ∈ span{E2 , E3 , E4 }.
Now, we denote f (s) = hE2 , ϕT i and, differentiating, we obtain
f 0 (s) = h∇T E2 , ϕT i + hE2 , ∇T ϕT i = h∇T E2 , ϕT i + hE2 , ξ + κ1 ϕE2 i
= h∇T E2 , ϕT i = h−κ1 T + κ2 E3 , ϕT i
= κ2 hE3 , ϕT i
Since ϕT = hϕT, E2 iE2 + hϕT, E3 iE3 + hϕT, E4 iE4 , the curve γ is proper-
(
3(c−1) 2
κ1 = constant > 0, κ21 + κ22 = c+3
4 + 4 f
0 3(c−1) 3(c−1)
κ2 = − 4 f hϕT, E3 i, κ2 κ3 = − 4 f hϕT, E4 i.
From the expression of f 0 (s), we can see that the third equation of the above
system is equivalent to
3(c − 1) 2
κ22 = − f + ω0 ,
4
where ω0 = constant. Replacing in the second equation, it follows that
c+3 3(c − 1) 2
κ21 = − ω0 + f ,
4 2
which implies f = constant. Thus, we have κ2 = constant > 0 and hE3 , ϕT i = 0.
Then, one obtains ϕT = f E2 + hϕT, E4 iE4 . It follows that there exists a unique
constant α0 ∈ (0, 2π) \ {π/2, π, 3π/2} such that f = cos α0 and hϕT, E4 i = sin α0
(see [60, 68]).
order r in N 2n+1 (c), c 6= 1, n ≥ 2, such that hE2 , ϕT i is not constant 0, 1, or −1.
We have
(1) If c ≤ −3, then γ is biharmonic if and only if it is a geodesic.
(2) If c > −3, then γ is proper-biharmonic if and only if ϕT = cos α0 E2 +
sin α0 E4 and

κ1 , κ2 , κ3 = constant > 0

3(c−1)
κ21 + κ22 = c+3 4 + 4 cos2 α0
 3(c−1)
κ2 κ3 = − 8 sin 2α0 ,

where α0 ∈ (0, 2π) \ {π/2, π, 3π/2} is a constant such that

c + 3 + 3(c − 1) cos2 α0 > 0 and 3(c − 1) sin 2α0 < 0.
Remark 4.8. We note that, in this case, we may obtain biharmonic curves that
are not helices.
Proposition 4.9 ([68]). Assume that c > −3, c 6= 1, and n = 2, and let γ be a
proper-biharmonic Legendre Frenet curve of osculating order r, such that hE2 , ϕT i
is not constant 0, 1, or −1. Then γ is a helix of order 4 or 5.
Proof. We know that r ∈ {4, 5}. If r = 4, then the result follows immediately
from Theorem 4.7.
Assume now that r = 5. Since ϕT = cos α0 E2 + sin α0 E4 , and ξ ⊥ ϕT , ξ ⊥ E2 ,
we get ξ ⊥ E4 , and then, along γ, we have ξ ∈ span{E3 , E5 }.
From the Frenet equations of γ, it follows that
h∇T E3 , ξi = h−κ2 E2 + κ3 E4 , ξi = 0 and h∇T E5 , ξi = h−κ4 E4 , ξi = 0.
Then, we obtain T (hE3 , ξi) = T (hE5 , ξi) = 0, i.e., a = hE3 , ξi = constant and
b = hE5 , ξi = constant along γ.
Now, we have
h∇T E4 , ξi = −κ3 hE3 , ξi + κ4 hE5 , ξi = −κ3 a + κ4 b
and, since h∇T E4 , ξi = hE4 , ϕT i = sin α0 , we get
(4.2) sin α0 = −κ3 a + κ4 b
which implies that b = 0 or κ4 = constant.
Case b = 0. Since ξ ∈ span{E3 , E5 }, we have E3 = ±ξ and, therefore,
∇T E3 = ∓ϕT = ∓ cos α0 E2 ∓ sin α0 E4 .
From the third Frenet equation, we get κ2 = ± cos α0 and κ3 = ∓ sin α0 , and then,
from Theorem 4.7, κ2 κ3 = −(1/2) sin 2α0 = −(3(c
√ − 1)/8) sin 2α0 . Thus, we have
c = 7/3 and, again using Theorem 4.7, κ1 = 2/ 3.
We shall prove now that κ4 = κ1 , and so γ is a helix of order 5. From the last
Frenet equation, we obtain
(4.3) h∇T E5 , ϕT i = −κ4 hE4 , ϕT i = −κ4 sin α0 .
Since hE5 , ϕT i = 0 we have h∇T E5 , ϕT i + hE5 , ∇T ϕT i = 0. We can check that
hE5 , ∇T ϕT i = κ1 hE5 , ϕE2 i, and, therefore, using (4.3), we get
(4.4) κ1 hE5 , ϕE2 i = κ4 sin α0 .
Next, from the fourth Frenet equation and (4.4), we have
κ24
(4.5) h∇T E4 , ϕE2 i = κ4 hE5 , ϕE2 i = sin α0 .
κ1
Since ϕT = cos α0 E2 + sin α0 E4 , one obtains that hE4 , ϕE2 i = 0. It follows
(4.6) h∇T E4 , ϕE2 i = −hE4 , ∇T ϕE2 i = −hE4 , ϕ∇T E2 i = κ1 hE4 , ϕT i
= κ1 sin α0 .
√
From (4.5) and (4.6) we obtain κ4 = κ1 = 2/ 3.
Case b 6= 0. In this situation, from (4.2), we have κ4 = constant and, therefore,
γ is a helix. Moreover, we can prove an additional relation between the curvatures.
Indeed, since ξ ∈ span{E3 , E5 }, it follows that a2 + b2 = 1. On the other hand

h∇T E2 , ξi = hE2 , ϕT i = cos α0
= h−κ1 T + κ2 E3 , ξi = κ2 a
and, as −κ3 a + κ4 b = sin α0 , replacing in a2 + b2 = 1, we get
(κ2 sin α0 + κ3 cos α0 )2 + κ24 (cos α0 )2 = κ22 κ24
and conclude.
From Theorems 4.1, 4.4, 4.5, and the above proposition, one obtains our next
result.
Theorem 4.10 ([68]). Let γ be a proper-biharmonic Legendre curve in N 5 (c).
Then c > −3 and γ is a helix of order r with 2 ≤ r ≤ 5.
In the same way as in the case of Frenet curves in complex manifolds (see [103]),
for Legendre Frenet curves γ : I → (N 2n+1 , ϕ, ξ, η, h, i) of osculating order r in a
Sasakian manifold, we define the ϕ-torsions τij by τij = hEi , ϕEj i = −hϕEi , Ej i,
i, j ∈ {1, . . . , r}, i < j.
From our previous results, it is easy to see that we have the following proposition.
Proposition 4.11 ([66]). Let γ : I → N 2n+1 (c) be a proper-biharmonic Le-
gendre Frenet curve in a Sasakian space form N 2n+1 (c), c 6= 1. Then c > −3 and
τ12 is constant.
We can also prove the two theorems that characterize proper-biharmonic Le-
gendre curves of osculating order r ≤ 4.
Theorem 4.12 ([66]). If γ is a proper-biharmonic Legendre Frenet curve in a
Sasakian space form N 2n+1 (c), c > −3, c 6= 1, of osculating order r < 4, then it is
a circle or a helix with constant ϕ-torsions.
Theorem 4.13 ([66]). If γ is a proper-biharmonic Legendre Frenet curve in a
Sasakian space form N 2n+1 (c) of osculating order r = 4, then c ∈ (7/3, 5) and the
curvatures of γ are
√ r r
c+3 1 6(c − 1)(5 − c) 1 3(c − 1)(3c − 7)
κ1 = , κ2 = , and κ3 = .
2 2 c+3 2 c+3
Moreover, the ϕ-torsions of γ are given by
 q q
τ12 = ∓ 2(5−c) , τ13 = 0, τ14 = ± 3c−7

c+3 c+3
q
τ23 = ∓ √
 3c−7
, τ24 = 0, τ34 = ± 2(5−c)(3c−7)
3(c−1)(c+3) .
3(c−1)(c+3)
In the following, we will work in the unit sphere S2n+1 with its canonical or de-
formed Sasakian structures, thought as a complete simply connected Sasakian space
form with constant ϕ-sectional curvature c > −3, and find the explicit equations of
biharmonic Legendre curves described in the first three cases as curves in R2n+2 .
Theorem 4.14 ([68]). Let γ : I → (S2n+1 , ϕ0 , ξ0 , η0 , h, i0 ), n ≥ 2, be a proper-
biharmonic Legendre curve parametrized by arc length. Then the equation of γ in
the Euclidean space E2n+2 = (R2n+2 , h, i), is either
1 √ 1 √ 1
γ(s) = √ cos 2s e1 + √ sin 2s e2 + √ e3 ,
2 2 2
where {ei , Iej }3i,j=1 are constant unit vectors orthogonal to one another, or
1 1 1 1
γ(s) = √ cos(As)e1 + √ sin(As)e2 + √ cos(Bs)e3 + √ sin(Bs)e4 ,
2 2 2 2
where
√ √
(4.7) A= 1 + κ1 , B= 1 − κ1 , κ1 ∈ (0, 1),
and {ei }4i=1 are constant unit vectors orthogonal to one another, satisfying
he1 , Ie3 i = he1 , Ie4 i = he2 , Ie3 i = he2 , Ie4 i = 0, Ahe1 , Ie2 i + Bhe3 , Ie4 i = 0,
where I is the usual complex structure on R2n+2 .
Proof. Let us denote by ∇ ˙ and ∇
e the Levi-Civita connections on (S2n+1 , h, i0 )
and (R2n+2 , h, i), respectively.
First, assume that γ is the biharmonic circle, that is κ1 = 1. From the Gauss
and Frenet equations we get
∇ ˙ T T − hT, T iγ = κ1 E2 − γ
eTT = ∇
and
∇ e T T = (−κ21 − 1)T = −2T,
eT∇
which imply
000
γ + 2γ 0 = 0.
The general solution of the above equation is
√ √
γ(s) = cos( 2s)c1 + sin( 2s)c2 + c3 ,
where {ci } are constant vectors in E2n+2 .
Now, since γ satisfies
hγ, γi = 1, hγ 0 , γ 0 i = 1, hγ, γ 0 i = 0,
hγ 0 , γ 00 i = 0, hγ 00 , γ 00 i = 2, hγ, γ 00 i = −1,
√
and, since at s = 0 we have γ = c1 + c3 , γ 0 = 2c2 , γ 00 = −2c1 , we obtain
1 1 1
c11 + 2c13 + c33 = 1, c22 = , c12 + c23 = 0, c12 = 0, c11 = , c11 + c13 = ,
2 2 2
where cij denotes hci , cj i. The above√relations imply that {ci } are orthogonal vectors
in E2n+2 with |c1 | = |c2 | = |c3 | = 1/ 2.
Finally, using the fact that γ is a Legendre curve, one √ easily obtains that
hci , Icj i = 0, for any i, j ∈ {1, 2, 3}. If we denote ei = 2ci , one obtains the
first part of the theorem.
Next, assume that γ is a biharmonic helix, that is κ21 + κ22 = 1, κ1 ∈ (0, 1). From
the Gauss and Frenet equations we get
∇eTT = ∇˙ T T − hT, T iγ = κ1 E2 − γ,

∇
eT∇ e T E2 − T = κ1 − κ1 T + κ2 E3 − T = − κ21 + 1 T + κ1 κ2 E3 ,
e T T = κ1 ∇
and

∇
eT∇ e T T = − κ2 +1 ∇
eT∇ e T E3 = − κ2 +1 ∇
e T T +κ1 κ2 ∇ e T T −κ1 κ2 E2 = −2γ 00 −κ2 γ.
1 1 2 2
Hence
γ iv + 2γ 00 + κ22 γ = 0,
with general solution

γ(s) = cos(As)c1 + sin(As)c2 + cos(Bs)c3 + sin(Bs)c4 ,
where A and B are given by (4.7), while {ci } are constant vectors in E2n+2 .
Since γ satisfies
hγ, γi = 1, hγ 0 , γ 0 i = 1, hγ, γ 0 i = 0, hγ 0 , γ 00 i = 0, hγ 00 , γ 00 i = 1 + κ21 ,
hγ, γ 00 i = −1, hγ 0 , γ 000 i = −(1 + κ21 ), hγ 00 , γ 000 i = 0,
hγ, γ 000 i = 0, hγ 000 , γ 000 i = 3κ21 + 1
and, since at s = 0 we have γ = c1 + c3 , γ 0 = Ac2 + Bc4 , γ 00 = −A2 c1 − B 2 c3 ,
γ 000 = −A3 c2 − B 3 c4 , one obtains
(4.8) c11 + 2c13 + c33 = 1
(4.9) A2 c22 + 2ABc24 + B 2 c44 = 1
(4.10) Ac12 + Ac23 + Bc14 + Bc34 = 0
(4.11) A3 c12 + AB 2 c23 + A2 Bc14 + B 3 c34 = 0
(4.12) A4 c11 + 2A2 B 2 c13 + B 4 c33 = 1 + κ21
(4.13) A2 c11 + (A2 + B 2 )c13 + B 2 c33 = 1
(4.14) A4 c22 + (AB 3 + A3 B)c24 + B 4 c44 = 1 + κ21
(4.15) A5 c12 + A3 B 2 c23 + A2 B 3 c14 + B 5 c34 = 0
(4.16) A3 c12 + A3 c23 + B 3 c14 + B 3 c34 = 0
(4.17) A6 c22 + 2A3 B 3 c24 + B 6 c44 = 3κ21 + 1

where cij = hci , cj i. Since the determinant of the system given by (4.10), (4.11),
(4.15), and (4.16) is equal to −A2 B 2 (A2 − B 2 )4 6= 0, it follows that
c12 = c23 = c14 = c34 = 0.
From (4.8), (4.12), and (4.13), we get
1 1
c11 = , c13 = 0, c33 =
2 2
and, from (4.9), (4.14), and (4.17), it follows that
1 1
c22 = , c24 = 0, c44 = .
2 2
2n+2 with |c | = |c | =
√ obtain that {ci } are orthogonal vectors in E
Therefore, we 1 2
|c3 | = |c4 | = 1/ 2.
Finally, since γ is a Legendre curve, the second part of our result follows.
Remark 4.15. We note that vectors {ei }4i=1 as in Theorem 4.14 do exist (see
[68]).
Before going further with our study, we need the following result.
Lemma 4.16 ([69]). Let Σm be an integral submanifold of S2n+1 . If X and Y

are vector fields tangent to Σm then
∇˙ X Y = ∇X Y and ∇
˙ X ϕY = ∇X ϕY,
where ∇ ˙ and ∇ are the Levi-Civita connections on (S2n+1 , h, i0 ) and (S2n+1 , h, ia ),
respectively.
Using this result and then the same method as in the proof of Theorem 4.14, we
get the following two theorems.
Theorem 4.17 ([68]). Let γ : I → (S2n+1 , ϕ, ξ, η, h, ia ), n ≥ 2, a > 0, a 6= 1
(c = 4/a − 3 > −3 and c 6= 1), be a proper-biharmonic Legendre curve parametrized
by arc length such that E2 ⊥ ϕT . Then the equation of γ in the Euclidean space
E2n+2 is either
1 r 2 1 r 2 1
γ(s) = √ cos s e1 + √ sin s e2 + √ e3 ,
2 a 2 a 2
for n ≥ 2, where {ei , Iej }3i,j=1 are constant unit vectors orthogonal to one another,
or
1 1 1 1
γ(s) = √ cos(As)e1 + √ sin(As)e2 + √ cos(Bs)e3 + √ sin(Bs)e4 ,
2 2 2 2
for n ≥ 3, where
r √ r √
1 + κ1 a 1 − κ1 a 1
A= , B= , κ1 ∈ 0, ,
a a a
and {ei , Iej }3i,j=1 are constant unit vectors orthogonal to one another.
Theorem 4.18 ([65, 68]). Let γ : I → (S2n+1 , ϕ, ξ, η, h, ia ), 0 < a < 1 (c > 1),
be a proper-biharmonic Legendre curve parametrized by arc length such that E2 k ϕT .
Then the equation of γ in the Euclidean space E2n+2 is
r r
B B
γ(s) = cos(As)e1 − sin(As)Ie1
A+B A+B
r r
A A
+ cos(Bs)e3 + sin(Bs)Ie3 ,
A+B A+B
where {e1 , e3 } are constant unit orthogonal vectors in E2n+2 with e3 orthogonal to
Ie1 , and
s p s p
3 − 2a − 2 (a − 1)(a − 2) 3 − 2a + 2 (a − 1)(a − 2)
A= , B= .
a a
3. Biharmonic non-Legendre curves in Sasakian space forms

Let γ : I → (N 2n+1 , ϕ, ξ, η, h, i) be a non-Legendre Frenet curve of osculating
order r with η(T ) = f , where f is a function defined along γ and f 6= 0. Since
(c + 3)κ
1 (c − 1)κ1 2 (c − 1) 0 (c − 1) 0
RN (T, ∇T T )T = − + f E2 − ff T + fξ
4 4 4 4
3(c − 1)κ1
− hE2 , ϕT iϕT,
4
we get the biharmonic equation of our curve

(4.18)
τ2 (γ) =∇3T T − RN (T, ∇T T )T
c−1 0 (c + 3)κ1 (c − 1)κ1 2
=(−3κ1 κ01 + f f )E1 + κ001 − κ31 − κ1 κ22 + − f E2
4 4 4
c−1 0 3(c − 1)κ1
+ (2κ01 κ2 + κ1 κ02 )E3 + κ1 κ2 κ3 E4 − f ξ+ hE2 , ϕT iϕT
4 4
=0.
Now, if c = 1, the curve γ is proper-biharmonic if and only if
κ1 = constant > 0, κ2 = constant, κ21 + κ22 = 1, and κ2 κ3 = 0
and then we can state the following theorem.
Theorem 4.19 ([61]). Let γ : I → N 2n+1 (1) be a non-Legendre Frenet curve of
osculating order r. Then γ is proper-biharmonic if and only if either it is a circle
with κ1 = 1, or a helix with κ21 + κ22 = 1.
Next, let us assume that c 6= 1. From the biharmonic equation (4.18), it follows
that the curve γ is proper-biharmonic if and only if κ1 > 0 and
(1) f 0 = 0 or ξ ∈ span{E1 , E2 , E3 , E4 } at any point of γ;
(2) ϕT ⊥ E2 or ϕT ∈ span{E2 , E3 , E4 };
(3) g(τ2 (γ), Ei ) = 0, for any i ∈ {1, 4}.
Computing hτ2 (γ), Ei i = 0 for all indices i ∈ {1, 4}, we obtain that γ is proper-
(2) ϕT ⊥ E2 or ϕT ∈ span{E2 , E3 , E4 };
and
(3)


 κ1 = constant > 0
κ2 + κ2 = c+3 − c−1 f 2 − 12 c−1 (f 0 )2 + 3(c−1) hE2 , ϕT i2


1 2 4 4 κ1 4 4
(4.19) 0 1 c−1 0 3(c−1)
κ2 −

 κ1 4 f η(E3 ) + 4 hE2 , ϕT ihE3 , ϕT i = 0
κ κ − 1 c−1 f 0 η(E ) + 3(c−1) hE , ϕT ihE , ϕT i = 0.

2 3 κ1 4 4 4 2 4
We note that, to obtain the second equation, we used η(E2 ) = hE2 , ξi = f 0 /κ1 ,
which follows from η(T ) = hT, ξi = f and the first Frenet equation of γ.
In order to solve these equations and find examples of non-Legendre proper-
biharmonic curves, in the following, we will consider some particular cases.
3.1. Biharmonic non-Legendre curves with η(T ) = constant. When the
angle β1 ∈ (0, π) \ {π/2} between the tangent vector field T and the characteristic
vector field ξ is constant, which means that f = η(T ) = cos β1 is also a constant,
the equations (4.19) become significantly easier to deal with. Therefore, we shall
first study this special case.
Theorem 4.20 ([61]). Let γ : I → N be a non-Legendre Frenet curve in
N 2n+1 (c), c 6= 1, of osculating order r such that f = η(T ) = cos β1 = constant ∈
/
{−1, 0, 1}. Then γ is proper-biharmonic if and only if either
(1) γ is a circle with ϕT = ± sin β1 E2 and κ21 = 1 + (c − 1) sin2 β1 > 0; or

(2) γ is a helix with ϕT = ± sin β1 E2 and κ21 + κ22 = 1 + (c − 1) sin2 β1 > 0; or
(3) n ≥ 2 and γ is a Frenet curve of osculating order r, where r ≥ 4, with
ϕT = sin β1 cos β2 E2 + sin β1 sin β2 E4
and

κ1 = constant > 0, κ2 = constant

3(c−1)
κ21 + κ22 = c+3 c−1
4 − 4 cos β1 +
2
4 sin2 β1 cos2 β2
κ2 κ3 = − 3(c−1)

sin2 β1 sin(2β2 ),

8
where β2 ∈ (0, 2π) is a constant such that

c + 3 − (c − 1) cos2 β1 + 3(c − 1) sin2 β1 cos2 β2 > 0 and 3(c − 1) sin(2β2 ) < 0.
Proof. First, we can see that η(E2 ) = hE2 , ξi = (1/κ1 )f 0 = 0. Next, let us
consider α = hE2 , ϕT i a function defined along γ. Then, using the second Frenet
equation of γ, one obtains
α0 = h∇T E2 , ϕT i + hE2 , ∇T ϕT i = κ2 hE3 , ϕT i + hE2 , κ1 ϕE2 + ξ − f T i
and, since the second term in the right hand side vanishes, it follows that
κ2 hE3 , ϕT i = α0 .
Now, let us assume that the curve γ is proper-biharmonic. Replacing the term
hE3 , ϕT i into the third equation of (4.19), we obtain
3(c − 1) 0
κ2 κ02 + αα = 0
4
and then
3(c − 1) 2
κ22 + α + ω0 = 0,
4
where ω0 is a constant. The second equation of (4.19) becomes
c+3 c−1 2
κ21 + κ22 =− f − κ22 − ω0 ,
4 4
which means that κ2 = constant and then α = constant. We also get
(4.20) κ2 hE3 , ϕT i = 0.
If κ2 = 0, then, from the biharmonic equation (4.18), we get E2 k ϕT and, since
hϕT, ϕT i = 1 − f 2 = sin2 β1 , it follows that ϕT = ± sin β1 E2 . Hence γ is a circle
with
c+3 c−1 3(c − 1)
κ21 = − cos2 β1 + sin2 β1 = 1 + (c − 1) sin2 β1 .
4 4 4
Let us now assume that κ2 6= 0. Then, from (4.20), it follows that hE3 , ϕT i = 0
and then, if κ3 = 0, the curve γ is a helix with
c+3 c−1 3(c − 1)
κ21 + κ22 = − cos2 β1 + sin2 β1 = 1 + (c − 1) sin2 β1 ,
4 4 4
since, again using (4.18), one obtains E2 k ϕT and then ϕT = ± sin β1 E2 , in this
case too.
Next, assume that n ≥ 2, κ2 6= 0, and κ3 6= 0. Then the osculating order of the

curve γ is r ≥ 4 and, from (4.20), we have hE3 , ϕT i = 0. From equation (4.18), it
follows that ϕT ∈ span{E2 , E4 }. Since
hϕT, ϕT i = 1 − f 2 = sin2 β1 ,
we get that
ϕT = sin β1 cos β2 E2 + sin β1 sin β2 E4 ,
where
hE2 , ϕT i = α = sin β1 cos β2 and hE4 , ϕT i = sin β1 sin β2 ,
with β2 = constant ∈ (0, 2π), and the conclusion follows from (4.19).
Remark 4.21. If γ is a Frenet curve of osculating order r > 1, not necessarily
biharmonic, with η(T ) = f = constant, in a 3-dimensional Sasakian space form,
then we can consider an orthogonal system of vector fields {E = T − f ξ, ϕT, ξ}
along γ and, using it, we easily get E2 k ϕT , in this case.
3.2. Biharmonic non-Legendre curves with E2 ⊥ ϕT or E2 k ϕT . Let
γ : I → N 2n+1 (c), c 6= 1, n ≥ 2, be a Frenet curve parametrized by arc length in a
Sasakian space form.
As a special role in the biharmonic equation (4.18) is played by the relation
between E2 and ϕT , it seems interesting to study the particular cases when E2 ⊥ ϕT
or E2 k ϕT and η(T ) = f (s) = cos(β(s)) 6= 0 is not necessarily constant. However,
we will prove that any of these hypotheses, together with the biharmonicity, also
implies η(T ) = f = constant.
Case I: c 6= 1, E2 ⊥ ϕT . In this case, γ is proper-biharmonic if and only if
and
(2)



 κ1 = constant > 0
κ2 + κ2 = c+3 − c−1 f 2 − 1 c−1 (f 0 )2

1 2 4 4 κ21 4
(4.21) 0 1 c−1 0


 κ2 − κ1 4 f η(E3 ) = 0
κ κ − 1 c−1 f 0 η(E ) = 0.

2 3 κ1 4 4
From hE2 , ξi = (1/κ1 )f 0 , one obtains h∇T E2 , ξi − hE2 , ϕT i = (1/κ1 )f 00 and then
κ2 η(E3 ) = (1/κ1 )f 00 + κ1 f . Replacing into the third equation of (4.21), we have
1 c − 1 0 00
κ2 κ02 − 2 (f f + κ21 f f 0 ) = 0
κ1 4
and then
1 c−1 0 2
κ22 − ((f ) + κ21 f 2 ) + ω1 = 0,
κ21 4
where ω1 is a constant. Now, from the second equation of (4.21), we have
c+3
κ21 + κ22 = − κ22 − ω1 .
4
Hence, κ2 = constant and (f 00 + κ21 f )f 0 = 0.
Next, using the Frenet equations, from hE2 , ϕT i = 0, one obtains
κ2 hE3 , ϕT i = −(1/κ1 )f 0
and then, from κ2 hE3 , ξi = (1/κ1 )f 00 + κ1 f , we get

1 000
κ2 κ3 hE4 , ξi = (f + (κ21 + κ22 )f 0 ).
κ1
Since τ2 (γ) = 0 implies η(τ2 (γ)) = 0, one obtains, after a straightforward com-
putation, that f 0 f 000 = 0. Using this result and differentiating (f 00 +κ21 f )f 0 = 0 along
γ, we have
κ21 (f 0 )2 + (f 00 + κ21 f )f 00 = 0.
Thus, we have just proved that η(T ) = f = constant. Then, from equation (4.18),
also using Remark 4.21, we can state our next result.
Theorem 4.22 ([61]). Let γ : I → N be a non-Legendre Frenet curve in
N 2n+1 (c), c 6= 1, n ≥ 2, of osculating order r such that E2 ⊥ ϕT . Then γ is
(1) γ is a circle with η(T ) = cos β0 and κ21 = (c + 3)/4 − ((c − 1)/4) cos2 β0 ; or
(2) γ is a helix with η(T ) = cos β0 and κ21 + κ22 = (c + 3)/4 − ((c − 1)/4) cos2 β0 ,
where β0 ∈ (0, 2π) \ {π/2, π, 3π/2} is a constant such that (c + 3)/4 − ((c −
1)/4) cos2 β0 > 0.
Case II: c 6= 1, E2 k ϕT . In this case hE2 , ξi = (1/κ1 )f 0 = 0 and then f =
cos β0 = constant. Since hϕT, ϕT i = 1 − |T |2 = sin2 β0 , it follows ϕT = ± sin β0 E2
and we have, using equation (4.18), the following proposition.
Proposition 4.23 ([61]). Let γ : I → N be a non-Legendre Frenet curve in
N 2n+1 (c),c 6= 1, of osculating order r such that E2 k ϕT . Then γ is proper-
biharmonic if and only if either
(1) γ is a circle with η(T ) = cos β0 and κ21 = c − (c − 1) cos2 β0 ; or
(2) γ is a helix with η(T ) = cos β0 and κ21 + κ22 = c − (c − 1) cos2 β0 , where
β0 ∈ (0, 2π) \ {π/2, π, 3π/2} is a constant such that c − (c − 1) cos2 β0 > 0.
Next, let γ be a proper-biharmonic non-Legendre curve with E2 k ϕT . As
ϕT = ± sin β0 E2 , one obtains, after a straightforward computation, that
1 κ1 1 κ1 cos β0
∇T E2 = − ± cos β0 T + ± 1 ξ.
sin β0 sin β0 sin β0 sin β0
Using the second Frenet equation of γ, we have
(κ1 cos β0 ± sin β0 )2
κ22 = .
sin2 β0
Thus, γ is a circle if and only if κ1 = ∓ tan β0 > 0. From Proposition 4.23, we easily
get that γ is a proper-biharmonic circle if and only if
√ √
2 c − 1 + c2 − 2c + 5 2 c + 1 − c2 − 2c + 5
κ1 = and cos β0 = .
2 2(c − 1)
If κ2 6= 0, from the expression of κ2 and the third Frenet equation of γ, it follows
that κ3 = 0 and then γ is a helix. Now, γ is proper-biharmonic if and only if κ1
satisfies
κ21 ± cos(2β0 )κ1 + (1 − c) sin4 β0 = 0,
where β0 ∈ (0, 2π)p \ {π/2, π, 3π/2},p if c > 1, or β0 ∈ (0, 2π) \ {π/2, π, 3π/2} such
that cos β0 ∈ (− (c − 1)/(c − 2), (c − 1)/(c − 2)), if c < 1.
We conclude with the following theorem.
Theorem 4.24 ([61]). Let γ : I → N a non-Legendre Frenet curve in N 2n+1 (c),

c 6= 1, of osculating order r such that E2 k ϕT . Then γ is proper-biharmonic if and
only if either
q √
(1) γ is a circle with η(T ) = ± (c + 1 − c2 − 2c + 5)/(2(c − 1)) and κ21 =
√
(c − 1 + c2 − 2c + 5)/2; or
(2) γ is a helix with η(T ) = cos β0 and κ1 satisfies
κ21 ± cos(2β0 )κ1 + (1 − c) sin4 β0 = 0,
where β0 = constant ∈ (0, 2π) \ {π/2, π, 3π/2}, if c > 1, or β0 = constant ∈
(0, 2π) \ {π/2, π, 3π/2} such that
rc − 1 rc − 1
cos β0 ∈ − , ,
c−2 c−2
if c < 1. In this case κ22 = (κ1 cot β0 ± 1)2 .
3.3. Biharmonic curves in R2n+1 (−3). As we have seen, whilst proper-bihar-
monic Legendre curves exist only in Sasakian space forms N 2n+1 (c) with constant
ϕ-sectional curvature c > 1, if n = 1, or c > −3, if n > 1, proper-biharmonic
non-Legendre curves can be found in Sasakian space forms with any ϕ-sectional
curvature.
In the following, we will obtain the explicit equations of proper-biharmonic circles
with E2 ⊥ ϕT and of all proper-biharmonic curves with E2 k ϕT in R2n+1 (−3).
Let us consider the Sasakian space form R2n+1 (−3) of constant ϕ-sectional curva-
ture −3. On R2n+1 , with coordinates (x1 , . . . , xn , y 1 , . . . , y n , z), consider the vector
fields Xi = 2(∂/∂y i ), Xn+i = ϕXi = 2(∂/∂xi ) + y i (∂/∂z), i ∈ {1, . . . , n}, and
ξ = 2(∂/∂z). Then, {Xi , Yi , ξ} form an orthonormal frame field in R2n+1 (−3) and,
after some straightforward computations, one obtains
[Xi , Xj ] = [Xn+i , Xn+j ] = [Xi , ξ] = [Xn+i , ξ] = 0, [Xi , Xn+j ] = 2δij ξ
and
∇Xi Xj = ∇Xn+i Xn+j = 0, ∇Xi Xn+j = δij ξ, ∇Xn+i Xj = −δij ξ,
∇Xi ξ = ∇ξ Xi = −Xn+i , ∇Xn+i ξ = ∇ξ Xn+i = Xi ,
for any i, j ∈ {1, . . . , n}.
Now, let γ : I → R2n+1 (−3) be a Frenet curve of osculating order r > 1,
parametrized by arc length, with the tangent vector field T = γ 0 , given by
n
X
(4.22) T = (Ti Xi + Tn+i Xn+i ) + cos β0 ξ,
i=1
where β0 is a constant. Using the above formulas of the Levi-Civita connection, we

have
Xn
(4.23) ∇T T = ((Ti0 + 2 cos β0 Tn+i )Xi + (Tn+i
0
− 2 cos β0 Ti )Xn+i ).
i=1
From Theorems 4.22 and 4.24, using the same techniques as in [24, 25, 44], we
get the following result.
Theorem 4.25 ([61]). The parametric equations of proper-biharmonic circles

parametrized by arc length in R2n+1 (−3), n ≥ 2, with E2 ⊥ ϕT , are
(4.24)


 xi (s) = ± κ11 {2 sin(κ1 s)ci1 ∓ 2 cos(κ1 s)ci2 − cos(2κ1 s)di1 − sin(2κ1 s)di2 } + ai

y i (s) = κ11 {2 cos(κ1 s)ci1 ± 2 sin(κ1 s)ci2 + sin(2κ1 s)di1 − cos(2κ1 s)di2 } + bi




 2 Pn i 2 i 2
z(s) = ± κ1 {1P+ i=1 ((c1 ) + (c2 ) )}s



+ 2κ1 2 ni=1 {± cos(4κ1 s)di1 di2 − 2 cos(2κ1 s)ci1 ci2
 1
+4 cos(3κ1 s)ci2 di2 − 4 sin(3κ1 s)ci1 di2 }




∓ κ11 ni=1 bi {−2 sin(κ1 s)ci1 ± 2 cos(κ1 s)ci2

 P



+ cos(2κ1 s)di1 + sin(2κ1 s)di2 } + e,


where κ21 = cos2 β0 , β0 ∈ (0, 2π) \ {π/2, π, 3π/2} is a constant, and ai , bi , ci1 , ci2 , di1 ,
di2 , and e are constants such that the n-dimensional constant vectors cj = (c1j , . . . , cnj )
and dj = (d1j , . . . , dnj ), j ∈ {1, 2}, satisfy
(
|c1 |2 + |c2 |2 + |d1 |2 + |d2 |2 = sin2 β0
hc1 , d1 i ± hc2 , d2 i = 0, hc1 , d2 i ∓ hc2 , d1 i = 0.
Proof. Let γ : I → R2n+1 (−3) be a circle parametrized by arc length, with the
tangent vector field T = γ 0 given by (4.22) and E2 ⊥ ϕT . From equation (4.23), one
obtains
n
1 X
E2 = ((Ti0 + 2 cos β0 Tn+i )Xi + (Tn+i
0
− 2 cos β0 Ti )Xn+i )
κ1
i=1
and, using hE2 , ϕT i = 0, a direct computation shows that
n
1 X
∇T E2 = ( ((Ti0 + 2 cos β0 Tn+i )0 + (Tn+i0
− 2 cos β0 Ti ) cos β0 )Xi
κ1
i=1
0
+ ((Tn+i − 2 cos β0 Ti )0 − (Ti0 + 2 cos β0 Tn+i ) cos β0 )Xn+i )
and, since γ is a circle, it follows that
(4.25) A0i + Bi cos β0 = 0 and Bi0 − Ai cos β0 = 0,
where Ai = (1/κ1 )(Ti0 + 2 cos β0 Tn+i ) and Bi = (1/κ1 )(Tn+i
0 − 2 cos β0 Ti ).
Solving equations (4.25) and using that γ is proper-biharmonic (according to
Theorem 4.22, this means that κ1 = ± cos β0 > 0), we get the following equations
(
Ti0 ± 2κ1 Tn+i = κ1 cos(κ1 s)ci1 ± κ1 sin(κ1 s)ci2
0
Tn+i ∓ 2κ1 Ti = ±κ1 sin(κ1 s)ci1 − κ1 cos(κ1 s)ci2 ,
whose general solutions are
(
Ti = − sin(κ1 s)ci1 ± cos(κ1 s)ci2 + cos(2κ1 s)di1 + sin(2κ1 s)di2
Tn+i = ± cos(κ1 s)ci1 + sin(κ1 s)ci2 ± sin(2κ1 s)di1 ∓ cos(2κ1 s)di2 ,
where ci1 , ci2 , di1 and di2 are constants, such that
Pn
i 2 i 2 i 2 i 2 2
Pi=1 ((c1 ) + (c2 ) + (d1 ) + (d2 ) ) = sin β0

n i i i i
i=1 ((c1 )(d1 ) ± (c2 )(d2 )) = 0
 n
P i i i i
i=1 ((c1 )(d2 ) ∓ (c2 )(d1 )) = 0,
since |T | = 1. Finally, replacing into expression of γ 0 and integrating, we get equa-

tions (4.24).
Using the same method, we can prove the following result.
Theorem 4.26 ([61]). Proper-biharmonic curves parametrized by arc length in
R2n+1 (−3) with E2 k ϕT , are either
(1) proper-biharmonic circles given by
 √
i (s) = ( 5 + 1) cos
√
5−1

i + sin
√
5−1

 x s c1 s ci + ai
√ 2 √ 2 2

√


5−1 5−1
 i i i i
y (s) = ( 5 + 1) sin 2 s c1 − cos 2 s c2 + b


√ √

 √ √ P √
z(s) = 1− 5±22 1+ 5 s + 3+2 5 ni=1 {((ci1 )2 − (ci2 )2n) sin(( 5 − 1)s)

 √ i i
√ Pn √
5−1

−2 cos(( 5 − 1)s)c c } + (1 + 5) b sin s ci2

i=1 i

 1 2 2
√

 o
5−1
 i
+ cos 2 s c1 + d,


where ai , bi , ci1 , ci2 , and d are constants and the n-dimensional √ constant
vectors cj = (c1j , . . . , cnj ), j ∈ {1, 2}, satisfy |c1 |2 + |c2 |2 = (3 − 5)/4; or
(2) proper-biharmonic helices given by

i (s) = − 2κ1 κ1 ±sin(2β0 ) i + sin κ1 ±sin(2β0 ) s ci + ai
x cos s c
κ1 ±sin(2β0) κ1 1 κ1 2



 i 2κ κ ±sin(2β ) i κ ±sin(2β )
y (s) = κ1 ±sin(2β0 ) sin
1 1 0
s c1 − cos 1 0
s ci2 + bi




 κ1 κ1
κ21

z(s) = 2 cos β0 + κ1 sin2 β0 s +


κ ±sin(2β ) (κ ±sin(2β 2
n 1 0P 1 0 ))
2(κ1 ±sin(2β0 )) n i 2 i 2
sin s i=1 ((c1 ) − (c2 ) )



 κ1 o
+ cos 2(κ1 ±sin(2β 0 ))
 Pn i i
s i=1 (c1 c2 )




 κ1 n o
 2κ1 Pn i κ1 ±sin(2β0 ) i + sin κ1 ±sin(2β0 ) s ci + d,
− κ1 ±sin(2β b cos s c


0) i=1 κ1 1 κ1 2
where β0 ∈ (0, 2π) \ {π/2, π, 3π/2} is a constant such that

√ √
2 5 2 5
cos β0 ∈ − 1, − ∪ ,1 ,
5 5
κ1 is a positive solution of the equation
κ21 ± sin(2β0 )κ1 + 4 sin4 β0 = 0,
ai , bi , ci1 , ci2 , and d are constants, and the n-dimensional constant vectors
cj = (c1j , . . . , cnj ), j ∈ {1, 2}, satisfy |c1 |2 + |c2 |2 = sin2 β0 .
4. A non-existence result for biharmonic curves in S7

We begin with a short account on 3-Sasakian manifolds as presented in [19].
If a manifold N admits three almost contact structures (ϕa , ξa , ηa ), a ∈ {1, 2, 3},
satisfying
ϕc = ϕa ϕb − ηb ⊗ ξa = −ϕb ϕa + ηa ⊗ ξb
ξc = ϕa ξb = −ϕb ξa , ηc = ηa ◦ ϕb = −ηb ◦ ϕa ,
for any even permutation (a, b, c) of (1, 2, 3), then the manifold is said to have an
almost contact 3-structure. The dimension of such a manifold is of the form 4n + 3.
We know that there exists an associated metric to all of these three structures. If
all structures are Sasakian, then N is called a 3-Sasakian manifold . We note that
every 3-contact structure is 3-Sasakian.
The canonical example of a 3-Sasakian manifold is the unit 7-sphere S7 = {z ∈
C4 : |z| = 1} endowed with a 3-Sasakian structure obtained as follows (see also [9]).
Consider the Euclidean space R8 with three complex structures
 
0 0 0 I2
0 −I4  0 0 −I2 0 
I= , J =  , K = −IJ ,
I4 0  0 I2 0 0 
−I2 0 0 0
where In denotes the n × n identity matrix. Let z denote the position vector of the
unit sphere in R8 , define three vector fields on S7 by
ξ1 = ξ0 = −Iz, ξ2 = −J z, and ξ3 = −Kz
and consider their dual 1-forms η1 = η0 , η2 , and η3 , respectively. Also consider
(1, 1)-tensors ϕa defined by
ϕ1 = ϕ0 = s ◦ I, ϕ2 = s ◦ J , ϕ3 = s ◦ K,
where s : Tz R8 → Tz S7 . Then (ϕa , ξa , ηa , h, i0 ), a ∈ {1, 2, 3}, determine a 3-Sasakian
structure on S7 .
Theorem 4.27 ([59]). There are no proper-biharmonic Legendre curves with re-
spect to all three Sasakian structures in the unit sphere S7 endowed with the canonical
3-Sasakian structure.
Proof. Let γ : I → S7 be a Frenet curve of osculating order r. Using the Frenet
equations and the expression of the curvature vector field of S7 (1), one obtains the
biharmonic equation of γ
(4.26) τ2 (γ) = − 3κ1 κ01 T + (κ001 − κ31 − κ1 κ22 + κ1 )E2 + (κ1 κ02 + 2κ01 κ2 )E3
+ κ1 κ2 κ3 E4
=0.
Thus the curve γ is proper-biharmonic if and only if
κ1 = constant 6= 0, κ2 = constant, κ21 + κ22 = 1, and κ2 κ3 = 0,
which means that γ is either a circle with κ1 = 1, or a helix with κ21 + κ22 = 1.
Next, let γ : I → S7 be a Legendre curve of osculating order r with respect to
all three Sasakian structures on S7 .
Differentiating hT, ξa i = ηa (T ) = 0, a ∈ {1, 2, 3}, along γ, we obtain that ∇T T =
κ1 E2 is orthogonal to ξa , for any a ∈ {1, 2, 3}. Now, it is easy to see that, at any
point of γ, {T, ϕ1 T, ϕ2 T, ϕ3 T, ξ1 , ξ2 , ξ3 } is an orthonormal basis and we then can
write ∇T T = κ1 E2 = 3a=1 λa ϕa T where λa = λa (s), a ∈ {1, 2, 3}, are real valued
P
functions alongqthe curve, called the relative curvatures of γ (see also [9]). Hence
P3 2
P3
κ1 = |∇T T | = a=1 λa and E2 = a=1 (λa /κ1 )ϕa T .
7
Since S is a 3-Sasakian manifold, we easily get

∇T ϕ1 T = ξ1 − λ1 T + λ2 ϕ3 T − λ3 ϕ2 T

(4.27) ∇T ϕ2 T = ξ2 − λ1 ϕ3 T − λ2 T + λ3 ϕ1 T

∇T ϕ3 T = ξ3 + λ1 ϕ2 T − λ2 ϕ1 T − λ3 T.

After a straightforward computation, from (4.27) and the second Frenet equation,
one obtains ∇T E2 = −κ1 T + 3a=1 {((λa /κ1 )0 ϕa T + (λa /κ1 )ξa } = −κ1 T + κ2 E3 and
P
then κ22 = 3a=1 {((λa /κ1 )0 )2 + (λa /κ1 )2 }.

P
Assume now that γ is proper-biharmonic. If γ is a circle, using the expression
of κ2 , we get λa = 0, for any a ∈ {1, 2, 3}, and therefore κ1 = 0, which is a
contradiction. If γ is a helix with κ21 + κ22 = 1, again using (4.27) and the third
Frenet equation, we have
3
X 2λ0 κ1 κ2 − λa (2κ0 κ2 + κ1 κ0 )
a 1 2
0 = κ3 E4 =κ2 E2 + ∇T E3 = ξa
a=1
κ21 κ22
n 1 λ 0 0 λ0 λ − λ0 λ − λ (1 − κ2 ) o
1 3 3 2 1
+ + 2 2
ϕ1 T
κ2 κ1 κ1 κ2
n 1 λ 0 0 λ0 λ − λ0 λ − λ (1 − κ2 ) o
2 1 1 3 2
+ + 3 2
ϕ2 T
κ2 κ1 κ1 κ2
n 1 λ 0 0 λ0 λ − λ0 λ − λ (1 − κ2 ) o
3 2 2 1 3
+ + 1 2
ϕ3 T
κ2 κ1 κ1 κ2
and then
3
1 X 0
0= 2λa ξa + (λ001 + λ02 λ3 − λ03 λ2 − λ1 κ21 )ϕ1 T
κ1 κ2
a=1

+ (λ002 + λ03 λ1 − λ01 λ3 − λ2 κ21 )ϕ2 T + (λ003 + λ01 λ2 − λ02 λ1 − λ3 κ21 )ϕ3 T ,
Hence, we get λ0a = 0, for any a ∈ {1, 2, 3}, and, from the expressions of the
curvatures κ1 and κ2 and since κ21 + κ22 = 1, one obtains κ2 = 1 and κ1 = 0, which
is also a contradiction.
5. Biharmonic anti-invariant submanifolds in Sasakian space forms

Our main result in this section provides a method to obtain biharmonic anti-
invariant submanifolds in a Sasakian space form.
Theorem 4.28 ([68]). Let (N 2n+1 , ϕ, ξ, η, h, i) be a Sasakian space form with
constant ϕ-sectional curvature c and let f : Σr → N be an r-dimensional integral
submanifold of N , 1 ≤ r ≤ n. Consider
e = I × Σr → N,
F :Σ F (t, p) = φt (p) = φp (t),
where I = S1 or I = R and {φt }t∈I is the flow of the vector field ξ. Then, we
e h, i e = dt2 + f ∗ h, i) → N is a Riemannian immersion which is
have that F : (Σ, Σ
proper-biharmonic if and only if Σr is a proper-biharmonic submanifold of N .
Proof. From the definition of the flow of ξ, we have
∂ d
dF (t, p) = |s=t {φp (s)} = φ̇p (t) = ξ(φp (t)) = ξ(F (t, p)),
∂t ds
i.e., ∂/∂t is F -correlated to ξ and
∂ ∂
dF (t, p) = |ξ(F (t, p))| = 1 = .

∂t ∂t
The vector Xp ∈ Tp Σr can be identified with (0, Xp ) ∈ T(t,p) (I × Σr ) and we

have
d
dF(t,p) (Xp ) = (dF )(t,p) (γ̇(0)) = |s=0 {φt (γ(s))} = (dφt )p (Xp ).
ds
Since φt is isometric, we also have |dF(t,p) (Xp )| = |(dφt )p (Xp )| = |Xp |. Moreover,
∂
hdF(t,p) , dF(t,p) (Xp )i = hξ(φp (t)), (dφt )p (Xp )i
∂t
= h(dφt )p (ξp ), (dφt )p (Xp )i = hξp , Xp i = 0
∂
= h , Xp iΣe ,
∂t
and therefore F : (I × Σr , h, iΣe ) → N is a Riemannian immersion.
Let F −1 (T N ) be the pull-back bundle over Σ e and ∇F the pull-back connection
determined by the Levi-Civita connection on N . We shall prove that
τ (F )(t,p) = (dφt )p (τ (f )) and τ2 (F )(t,p) = (dφt )p (τ2 (f )),
so, from the point of view of harmonicity and biharmonicity, Σ e and Σr have the
same behaviour.
We begin with two remarks. First, let ν ∈ C(F −1 (T N )) be a section in F −1 (T N )
defined by ν(t,p) = (dφt )p (Zp ), where Z is a vector field along Σr , i.e., Zp ∈ Tp N ,
∀p ∈ Σr . One can easily check that
(4.28) (∇FX ν)(t,p) = (dφt )p (∇N
X Z), ∀X ∈ C(T Σr ).
Then, if ν ∈ C(F −1 (T N )), it follows that ϕν given by (ϕν)(t,p) = ϕφp (t) (ν(t,p) ) is a
section in F −1 (T N ) and
(4.29) ∇F∂ ϕν = ϕ∇F∂ ν.
∂t ∂t
Now, we consider {X1 , ..., Xr } a local orthonormal frame field on U , where U is

an open subset of Σr . The tension field of F is given by
∂ r
e ∂
X
(4.30) τ (F ) = ∇F∂ dF − dF ∇Σ∂ + {∇FXi dF (Xi ) − dF (∇Σ
Xi Xi )}.
e
∂t ∂t ∂t ∂t
i=1
Since
∂ ∂ ∂
∇F∂ dF = ∇N Σ
ξ ξ = 0, ∇ ∂ = ∇I∂ = 0,
e
∂t ∂t ∂t ∂t ∂t ∂t
r
(∇FXa dF (Xa ))(t,p) = (dφt )p (∇N
Xa Xa ), dF(t,p) (∇Σ Σ
Xi Xi ) = (dφt )p (∇Xi Xi ),
e
replacing in (4.30), we get

τ (F )(t,p) = (dφt )p (τ (f )).
To show that τ2 (F )(t,p) = (dφt )p (τ2 (f )), we shall prove first that ∇F∂/∂t τ (F ) =
−ϕ(τ (F )).
Since [∂/∂t, Xi ] = 0, i ∈ {1, ..., r}, it follows that
∂
∇F∂ dF (Xi ) = ∇FXi dF .
∂t ∂t
But we also have

∂
∇FXi dF = ∇N N
dF(t,p) Xi ξ = ∇(dφt )p Xi ξ = −ϕ((dφt )p (Xi ))
∂t (t,p)
= −(dφt )p (ϕXi )
and then

(4.31) ∇F∂ dF (Xi ) = −(dφt )p (ϕXi ).
∂t (t,p)
We note that
∂
RF , Xi dF (Xi ) = ∇F∂ ∇FXi dF (Xi ) − ∇FXi ∇F∂ dF (Xi )
∂t ∂t ∂t
and, on the other hand, since N is a Sasakian space form,

∂
RF , Xi dF (Xi ) = RφNt (p) (ξ, (dφt )p (Xi ))(dφt )p (Xi ) = ξ.
∂t (t,p)
Therefore
(4.32) ∇F∂ ∇FXi dF (Xi ) − ∇FXi ∇F∂ dF (Xi ) = ξ.
∂t ∂t
Using (4.28) and (4.31), ∇FXa ∇F∂ dF (Xa ) can be written as

∂t

(4.33) ∇FXi ∇F∂ dF (Xi ) = −(dφt )p (∇N
Xi ϕXi )
∂t (t,p)
= −(dφt )p (ξ + ϕ∇N
Xi Xi ).
Moreover, from (4.31), we have

r

(4.34) ∇F∂ dF (∇Σ X ) = ∇F∂ dF (∇Σ X )
e
Xi i Xi i
∂t (t,p) ∂t (t,p)
Σr
= −(dφt )p (ϕ∇Xi Xi ).
Replacing (4.33) in (4.32) and using (4.34), we obtain
ξ =∇F∂ ∇FXi dF (Xi ) − ∇F∂ dF (∇Σ F Σ F F
Xi Xi ) + ∇ ∂ dF (∇Xi Xi ) − ∇Xi ∇ ∂ dF (Xi )
e e
∂t ∂t ∂t ∂t
F r
=∇ ∂ ∇dF (Xi , Xi ) − (dφt )p (ϕ∇Σ
Xi Xi ) + (dφt )p (ξ + ϕ∇N
Xi Xi )
∂t
r
=∇F∂ ∇dF (Xi , Xi ) + ϕ(dφt )p (∇N Σ
Xi Xi − ∇Xi Xi ) + ξ
∂t
and then

(4.35) ∇F∂ ∇dF (Xi , Xi ) = −ϕ(dφt )p (∇df (Xi , Xi )).
∂t (t,p)
Since ∇dF (∂/∂t, ∂/∂t) = 0, summing up in (4.35), we obtain

(4.36) ∇F∂ τ (F ) = −ϕ(τ (F )).
∂t
From (4.29) and (4.36), we have

(4.37) ∇F∂ ∇F∂ τ (F ) = −∇F∂ ϕ(τ (F )) = −ϕ∇F∂ τ (F ) = ϕ2 τ (F )
∂t ∂t ∂t ∂t
= −τ (F ),
and, from (4.28),
(4.38) (∇FXi ∇FXi τ (F ))(t,p) = (dφt )p (∇N N
Xi ∇Xi τ (f ))
and

(4.39) ∇F Σe τ (F ) = (dφt )p ∇N Σ
∇ X
r τ (f ) .
∇X Xi (t,p) Xi i
i
From (4.37), (4.38), and (4.39), one obtains

r
X
(4.40) −(∆F τ (F ))(t,p) =∇F∂ ∇F∂ τ (F ) + {∇FXi ∇FXi τ (F ) − ∇F Σe τ (F )}
∂t ∂t ∇X Xi
i=1 i
= − τ (F )(t,p) − (dφt )p (∆f τ (f )).

Using the expression of the curvature tensor field RN , after a straightforward
computation, we get
F
(4.41) trace R(t,p) (dF, τ (F ))dF = −τ (F ) + (dφt )p (trace RpN (df, τ (f ))df ).
Finally, from (4.40) and (4.41), we obtain
τ2 (F )(t,p) = (dφt )p (τ2 (f ))
and conclude.
Theorem 4.29 ([68]). Let Σ e 2 be a surface of N 2n+1 (c) invariant under the flow-
action of the characteristic vector field ξ. Then Σ e 2 is proper-biharmonic if and only
if it is locally given by x(t, s) = φt (γ(s)), where γ is a proper-biharmonic Legendre
curve.
Proof. A surface Σ e 2 of N 2n+1 (c) invariant under the flow-action of the char-
acteristic vector field ξ, that is φt (p) ∈ Σ e 2 , for any t and any p ∈ Σ e 2 , can be
written (locally) x(t, s) = φt (γ(s)), where γ is a Legendre curve in N . Then,
from Theorem 4.28, such a surface is proper-biharmonic if and only if γ is proper-
biharmonic.
Corollary 4.30 ([68]). Let Σ e 2 be a surface in S2n+1 (1) which is invariant under
the flow-action of the characteristic vector field ξ. Then Σ e 2 is proper-biharmonic if
and only if it is (locally) given by x(t, s) = φt (γ(s)), where γ is a proper-biharmonic
Legendre curve given by Theorem 4.14.
Remark 4.31. When the ambient space is S2n+1 endowed with its canonical
or deformed Sasakian structure, the flow of ξ is φt (z) = exp(−i(t/a))z, and from
Theorems 4.17, 4.18, and 4.28, we can obtain explicit examples of proper-biharmonic
surfaces with constant mean curvature in (S2n+1 , ϕ, ξ, η, h, ia ), a ≥ 0. Such a result,
that gives the explicit equation of proper-biharmonic Hopf cylinders in S3 endowed
with its deformed Sasakian structure, was first obtained, by a different method, in
[65].
Next, consider Σe 2 a surface of N 2n+1 (c) invariant under the flow-action of the
characteristic vector field ξ and let T = γ 0 and E2 be the first two vector fields
defined by the Frenet equations of the above Legendre curve γ. Since ∇F∂/∂t τ (F ) =
−ϕ(τ (F )), we have the following proposition.
Proposition 4.32 ([68]). Let Σ e 2 be a proper-biharmonic surface of N 2n+1 (c)
invariant under the flow-action of the characteristic vector field ξ. Then Σ e 2 has
parallel mean curvature vector field if and only if c > 1 and ϕT = ±E2 .
Corollary 4.33 ([68]). The proper-biharmonic surfaces of S2n+1 (1) invariant

under the flow-action of the characteristic vector field ξ0 are not of parallel mean
curvature vector field.
In the following, we will look for biharmonic surfaces in S7 that are also integral
surfaces with respect to two of the Sasakian structures on the sphere and invariant
under the flow-action of the characteristic vector field of the third one and prove
that there are no such surfaces. This result was obtained for the first time, by a
different method, in [59] and then recovered when we were preparing our article
[68], by using the next proposition. Eventually, these results did not make it into
that paper, but we think that they are interesting enough to be presented here.
Proposition 4.34. A surface Σ2 of S7 is invariant under the flow-action of the
characteristic vector field of one of the three Sasakian structures on the sphere, that
is, for example, φ1t (p) ∈ Σ2 , for any t and any p ∈ Σ2 , where φ1 is the flow of ξ1 ,
and integral with respect to the other two, if and only if it can be locally written
as x(t, s) = φ1t (γ(s)), where γ is a Legendre curve in S7 , with respect to all three
Sasakian structures.
Proof. Let Σ2 be a surface of S7 invariant under the flow-action of the charac-
teristic vector field ξ1 . Then it can be locally written as x(t, s) = φ1t (γ(s)), where γ
is a Legendre curve in S7 with respect to the first Sasakian structure. Moreover, if
Σ2 is integral with respect to the other two Sasakian structures of S7 , then so is the
curve γ.
Conversely, let γ be a Legendre curve in S7 , with respect to the three Sasakian
structures, parametrized by its arc-length and consider the surface Σ2 locally given
by x(t, s) = φ1t (γ(s)), and, therefore, invariant under the flow-action of ξ1 .
Since the flow of the characteristic vector field ξ1 is given by
φ1t (z) = (cos t)z − (sin t)Iz,
we get
x(t, s) = φ1t (γ(s)) = (cos t)γ(s) − (sin t)Iγ(s).
Next, we have
∂x
(t, s) = ξ1 (x(t, s)) = −Ix(t, s) = −(sin t)γ(s) − (cos t)Iγ(s),
∂t
∂x
(t, s) = (cos t)γ 0 (s) − (sin t)Iγ 0 (s)
∂s
and it is easy to see that h∂x/∂t, ∂x/∂si0 = 0 at any point on the surface, since γ is
a Legendre curve. Using the same argument, we obtain
∂x ∂x
hξ2 (x(t, s)), (t, s)i0 = h−J (x(t, s)), (t, s)i0 = 0
∂s ∂s
and
∂x ∂x
hξ3 (x(t, s)), (t, s)i0 = h−K(x(t, s)), (t, s)i0 = 0
∂s ∂s
at any point on the surface and, as ξ2 and ξ3 are orthogonal to ξ1 , it follows that Σ2
is integral with respect to the second and the third Sasakian structures.
From Proposition 4.34 and Theorems 4.27 and 4.29, one obtains the following
result.
Theorem 4.35 ([59]). There are no proper-biharmonic integral surfaces with

respect to two of the Sasakian structures and invariant under the flow-action of the
characteristic vector field of the third one in the unit sphere S7 endowed with the
canonical 3-Sasakian structure.
6. Biharmonic hypersurfaces in Sasakian space forms

Let (N 2n+1 , ϕ, ξ, η, h, i) be a Sasakian space form with constant ϕ-sectional cur-
vature c, and π : N → N̄ = N/ξ the Boothby-Wang fibration. If X̄ is a vector field
tangent to N̄ , then we will denote by X̄ H its horizontal lift to N . Let ̄ : Σ̄m−1 ,→ N̄
be a submanifold and consider the associated Hopf cylinder  : Σm = π −1 (Σ̄m−1 ) ,→
N , of dimension m.
From Theorem 3.1 and (1.38), we obtain our first result.
Theorem 4.36 ([67]). The Hopf cylinder  : Σm = π −1 (Σ̄m−1 ) ,→ N is bihar-
monic if and only if
(
∆⊥ H − trace σ(·, AH ·) + c(m+2)+3m−2
4 H + 3(c−1) ⊥ ⊥
4 (ϕ(ϕH) ) = 0
4 trace A∇⊥ 2 ⊥ >
(·) + m grad(|H| ) − 3(c − 1)(ϕ(ϕH) ) = 0,
· H
where σ, A, and H are the second fundamental form of Σm in N , the shape operator,
and the mean curvature vector field, respectively.
Corollary 4.37 ([67]). If Σ̄2n−1 is a hypersurface of N̄ , then Σ2n = π −1 (Σ̄2n−1 )
is biharmonic if and only if
(
∆⊥ H = |σ|2 − c(n+1)+3n−12 H
(4.42)
2 trace A∇⊥· H
(·) + n grad(|H|2 ) = 0.
Proof. The result follows easily, since, in codimension 1, we have (ϕH)⊥ = 0

and trace σ(·, AH ·) = |σ|2 H.
Now, since τ () = 2nH = (τ (̄))H = (2n−1)H̄ H , we have the following corollary.
Corollary 4.38 ([67]). If Σ̄2n−1 is a hypersurface with |H̄| = constant 6= 0,
then Σ2n = π −1 (Σ̄2n−1 ) is proper-biharmonic if and only if
c(n + 1) + 3n − 1
|σ|2 = .
2
Next, we shall prove a Lawson type formula which relates |σ|2 to |σ̄|2 (see [55,
86, 96]). Denote by ϕ⊥ Σ the restriction of ϕ to the normal bundle of Σ
m in N
composed with the projection on the same normal bundle, that is ϕΣ ν = (ϕν)⊥ , for
⊥
any section ν in the normal bundle of Σm in N .

Proposition 4.39 ([67]). Let Σ̄m−1 be a submanifold of N̄ , and denote by
σ̄ its second fundamental form. Then, the second fundamental form σ of Σm =
π −1 (Σ̄m−1 ) in N and σ̄ are related by the following formula
|σ|2 = |σ̄|2 + 2(2n + 1 − m) − 2|ϕ⊥ 2
Σ| .
Proof. Let us consider X̄, Ȳ ∈ C(T Σ̄m−1 ). We have

1
∇N
X̄ H Ȳ
H
= (∇N̄ H H H
X̄ Ȳ ) + 2 V [X̄ , Ȳ ]
and
1
∇X̄ H Ȳ H = (∇Σ̄ H H H
X̄ Ȳ ) + 2 V [X̄ , Ȳ ].
Thus σ(X̄ H , Ȳ H ) = (σ̄(X̄, Ȳ ))H .

We also have
2n+1
X 2n+1
X
σ(X̄ H , ξ) = hσ(X̄ H , ξ), να iνα = h∇N
X̄ H ξ − ∇X̄ H ξ, να iνα
α=m+1 α=m+1
2n+1
X 2n+1
X
=− hϕX̄ H , να iνα = hϕνα , X̄ H iνα ,
α=m+1 α=m+1
where {να }2n+1 m

α=m+1 is a local orthonormal frame in the normal bundle of Σ in N .
m−1 m−1
Next, let {X̄i }i=1 be a local orthonormal frame field on Σ̄ . It follows that
H m−1 m
{X̄i }i=1 ∪ {ξ} is a local orthonormal frame field on Σ and one obtains
m−1
X m−1
X
2 2
|σ| = |σ(ξ, ξ)| + 2 |σ(X̄iH , ξ)|2 + |σ(X̄iH , X̄jH )|2
i=1 i,j=1
m−1
X 2n+1
X
=2 hϕνα , X̄iH i2 + |σ̄|2
i=1 α=m+1
2n+1
X
= |σ̄|2 + 2 2n + 1 − m − |(ϕνα )⊥ |2 ,
α=m+1

Corollary 4.40 ([67]). If Σ̄2n−1 is a hypersurface, then |σ|2 = |σ̄|2 + 2.
Our next result follows from Corollaries 4.38 and 4.40.
Proposition 4.41 ([67]). If |H̄| = constant 6= 0, then Σ2n = π −1 (Σ̄2n−1 ) is
proper-biharmonic if and only if
c(n + 1) + 3n − 5
|σ̄|2 = .
2
Remark 4.42. From Proposition 4.41, we can see that there are no proper-
biharmonic hypersurfaces of type Σ2n = π −1 (Σ̄2n−1 ) in a Sasakian space form
N 2n+1 (c) with c ≤ (5 − 3n)/(n + 1), which implies that such hypersurfaces do
not exist when c ≤ −3.
Proposition 4.43 ([67]). If Σ2n = π −1 (Σ̄2n−1 ) is a proper-biharmonic hyper-
surface with constant mean curvature, then
(2n − 1)(c(n + 1) + 3n − 5)
|H|2 ∈ 0, .
8n2
Proof. Let Σ2n = π −1 (Σ̄2n−1 ) be a proper-biharmonic hypersurface with con-
stant mean curvature. Then, from Corollary 4.38 and Proposition 4.41, it follows
that
c(n + 1) + 3n − 1 c(n + 1) + 3n − 5
|σ|2 = and |σ̄|2 = .
2 2
On the other hand, the following inequalities hold

|σ|2 ≥ 2n|H|2 and |σ̄|2 ≥ (2n − 1)|H̄|2 .
It can be easily proved that there are no non-minimal umbilical hypersurfaces
of type Σ2n = π −1 (Σ̄2n−1 ) and it is also known that Σ̄2n−1 cannot be umbilical.
Therefore, in the above inequalities we cannot have equality and then
c(n + 1) + 3n − 1 c(n + 1) + 3n − 5
> 2n|H|2 and > (2n − 1)|H̄|2 .
2 2
But, we also have
(2n − 1)2 (2n − 1)(c(n + 1) + 3n − 5)
|H|2 = 2
|H̄|2 <
(2n) 8n2
and, since
(2n − 1)(c(n + 1) + 3n − 5) c(n + 1) + 3n − 1
2
< ,
8n 4n
we conclude.
Using Corollary 4.38, we get our next result.
Proposition 4.44 ([67]). If Σ2n = π −1 (Σ̄2n−1 ) is a proper-biharmonic hyper-
surface with constant mean curvature, then its scalar curvature s is constant and
c−1
s = (c + 3)(n2 − n) + (n − 3) + 4n2 |H|2 .
2
In the following, we will prove some classification results for biharmonic hyper-
surfaces in Sasakian space forms N 2n+1 (c), with c > −3.
Homogeneous real hypersurfaces Σ̄2n−1 in the complex projective space CP n ,
n > 1, were classified by R. Takagi, who proved that there are five different types
of such hypersurfaces (see [111, 130]). We shall use this classification to study
proper-biharmonic Hopf cylinders Σ2n = π −1 (Σ̄2n−1 ) in N 2n+1 (c).
6.1. Types A1 and A2. Let us consider u ∈ (0, π/2) and r a positive constant
given by 1/r2 = (c + 3)/4. A hypersurface of type A1 in CP n (c + 3) is a geodesic
sphere and it has two distinct principal curvatures: λ2 = (1/r) cot u of multiplicity
2n−2 and a = (2/r) cot 2u of multiplicity 1, while a hypersurface of type A2 has three
distinct principal curvatures: λ1 = −(1/r) tan u of multiplicity 2p, λ2 = (1/r) cot u
of multiplicity 2q, and a = (2/r) cot 2u of multiplicity 1, where p > 0, q > 0, and
p + q = n − 1.
We note that, if c = 1 and Σ̄2n−1 is a hypersurface of type A1 or A2, then
π (Σ̄2n−1 ) = S1 (cos u) × S2n−1 (sin u) ⊂ S2n+1 or π −1 (Σ̄2n−1 ) = S2p+1 (cos u) ×
−1
S2q+1 (sin u), respectively.

Theorem 4.45 ([67]). Let Σ2n = π −1 (Σ̄2n−1 ) be the Hopf cylinder over Σ̄2n−1 .
Then we have
(1) If Σ̄2n−1 is of type A1, then Σ2n is proper-biharmonic if and only if either
(a) c = h1 and (tan u)√2 = 1; or

−3n2 +2n+1+8 2n−1
(b) c ∈ n2 +2n+5
, +∞ \ {1} and
p
2c − 2 ± c2 (n2 + 2n + 5) + 2c(3n2 − 2n − 1) + 9n2 − 30n + 13
(tan u)2 = n + .
c+3
(2) If Σ̄2n−1 is of type A2, then Σ2n is proper-biharmonic if and only if either
(a) c = 1, (tan u)2 = 1, and
√ p 6= q; or
−3(p−q)2 −4n+4+8 (2p+1)(2q+1)
h
(b) c ∈ 2
(p−q) +4n+4
, +∞ \ {1} and
n 2c − 2
(tan u)2 = + ±
2p + 1 (c + 3)(2p + 1)
p
c2 ((p − q)2 + 4n + 4) + 2c(3(p − q)2 + 4n − 4) + 9(p − q)2 − 12n + 4
.
(c + 3)(2p + 1)
Proof. First, let us consider Σ̄2n−1 a hypersurface of type A1. Then, from
Proposition 4.41, we have that Σ2n = π −1 (Σ̄2n−1 ) is biharmonic if and only if
1 4
|σ̄|2 = (2n − 2)λ22 + a2 = (2n − 2) 2
(cot u)2 + 2 (cot 2u)2
r r
c(n + 1) + 3n − 5
= .
2
Denote tan u = t and then, after a straightforward computation, we obtain the
following equation
(4.43) (c + 3)t4 − 2(c(n + 2) + 3n − 2)t2 + (2n − 1)(c + 3) = 0,
which admits real solutions if and only if
c2 (n2 + 2n + 5) + 2c(3n2 − 2n − 1) + 9n2 − 30n + 13 ≥ 0.
But c > (5 − 3n)/(n + 1), which means that (4.43) has real solutions if and only if
h −3n2 + 2n + 1 + 8√2n − 1
c∈ , +∞ ,
n2 + 2n + 5
and these solutions are given by
p
2 2c − 2 ± c2 (n2 + 2n + 5) + 2c(3n2 − 2n − 1) + 9n2 − 30n + 13
t1,2 = n + > 0.
c+3
Now, we have that Σ2n is minimal if and only if Σ̄2n−1 is minimal, that is
(2n − 2)λ2 + a = 0, which leads to (tan u)2 = 2n − 1.
It is easy to see that, if one of the solutions t21 or t22 is equal to 2n − 1, then
c = 1. If c = 1, then Σ2n is proper-biharmonic if and only if (tan u)2 = 1, and if
c 6= 1, then t21 6= 2n − 1 and t22 6= 2n − 1.
Next, let Σ̄2n−1 be a hypersurface of type A2. Then, according to Proposi-
tion 4.41, Σ2n is biharmonic if and only if
1 1 4
|σ̄|2 = 2pλ21 + 2qλ22 + a2 = 2p 2 (tan u)2 + 2q 2 (cot u)2 + 2 (cot 2u)2
r r r
c(n + 1) + 3n − 5
= .
2
After a straightforward computation, this equation becomes
(4.44) (c + 3)(2p + 1)t4 − 2(c(n + 2) + 3n − 2)t2 + (c + 3)(2q + 1) = 0,
where t = tan u.
Equation (4.44) has real solutions if and only if
c2 ((p − q)2 + 4n + 4) + 2c(3(p − q)2 + 4n − 4) + 9(p − q)2 − 12n + 4 ≥ 0,
which, together with c > (5 − 3n)/(n + 1), leads to

h −3(p − q)2 − 4n + 4 + 8p(2p + 1)(2q + 1)
c∈ , +∞ \ {1}.
(p − q)2 + 4n + 4
Then the solutions of equation (4.44) are
(4.45)
n 2c − 2
t21,2 = + ±
2p + 1 (c + 3)(2p + 1)
p
c2 ((p − q)2 + 4n + 4) + 2c(3(p − q)2 + 4n − 4) + 9(p − q)2 − 12n + 4
> 0.
(c + 3)(2p + 1)
The hypersurface Σ̄2n−1 is minimal if and only if 2pλ1 + 2qλ2 + a = 0, or,
equivalently, (tan u)2 = (2q + 1)/(2p + 1). It follows that Σ2n is proper-biharmonic if
and only if c = 1, (tan u)2 = 1, and p 6= q, or c 6= 1 and tan u is given by (4.45).
Remark 4.46. If c 6= 1, in the A1 case, we can obtain two proper-biharmonic
Hopf cylinders, not only one. The same thing happens in the A2 case when p 6= q;
for p = q we have a proper-biharmonic Hopf cylinder if and only if c ∈ (1, +∞) and,
in this case, we get
p
2 2(c − 1) + 2 c2 (n + 1) + 2c(n − 1) − 3n + 1
(tan u) = 1 + .
n(c + 3)
6.2. Types B, C, D, and E. Consider u ∈ (0, π4 ) and r a positive constant
as above. The B type hypersurfaces in complex projective space CP n (c + 3) have
three distinct principal curvatures: −(1/r) cot u and (1/r) tan u of multiplicity n−1,
and (2/r) tan 2u of multiplicity 1. The hypersurfaces of types C, D, or E have five
distinct principal curvatures: λ1 = −(1/r) cot u, λ2 = (1/r) cot(π/4 − u), λ3 =
(1/r) cot(π/2 − u), λ4 = (1/r) cot(3π/4 − u), and a = −(2/r) cot 2u, having, in each
of the three cases, specific multiplicities (see [111, 130]).
We have the following non-existence result for biharmonic Hopf cylinders Σ2n =
π −1 (Σ̄2n−1 ), that can be proved in the same way as Theorem 4.45.
Theorem 4.47. There are no proper-biharmonic Hopf cylinders π −1 (Σ̄2n−1 ),
where Σ̄2n−1 is a hypersurface of type B, C, D, or E in CP n (c + 3).
7. Biharmonic integral C-parallel submanifolds in 7-dimensional

Sasakian space forms
Let Σm be an integral submanifold of a Sasakian manifold (N 2n+1 , ϕ, ξ, η, h, i).
We recall that, in this case, ϕ(T Σm ) ⊂ N Σm and m ≤ n. Moreover, when m = n,
it follows that ϕ(N Σm ) = T Σm . If σ is the second fundamental form of Σm , then,
after a straightforward computation, one obtains
hϕZ, σ(X, Y )i = hϕY, σ(X, Z)i
for any vector fields X, Y , and Z tangent to Σm (see also [9]). We also note that
Aξ = 0, where A is the shape operator of Σm (see [19]).
An integral submanifold Σm is said to be integral C-parallel if ∇⊥ σ is parallel
to the characteristic vector field ξ. Thus, Σm is an integral C-parallel submanifold
if (∇⊥ σ)(X, Y, Z) = S(X, Y, Z)ξ, for any vector fields X, Y , and Z tangent to Σm ,
where S(X, Y, Z) = hϕX, σ(Y, Z)i is a totally symmetric tensor field of type (0, 3)
on Σm . It is not difficult to check that, when m = n, ∇⊥ σ = 0 if and only if σ = 0,

i.e., Σn is totally geodesic.
Now, let Σm be an integral submanifold and denote by H its mean curvature
vector field. We say that H is C-parallel if ∇⊥ H is parallel to ξ, i.e., ∇⊥
X H = θ(X)ξ,
where θ is a 1-form on Σm . As we shall see, θ(X) = hH, ϕXi, for any vector field X
tangent to Σm .
The following two results will be used later on.
Proposition 4.48 ([70]). If the mean curvature vector field H of an integral
submanifold Σn of a Sasakian manifold (N 2n+1 , ϕ, ξ, η, h, i) is parallel, then Σn is
minimal.
Proposition 4.49 ([70]). Let (N 2n+1 , ϕ, ξ, η, h, i) be a Sasakian manifold and
Σm an integral C-parallel submanifold with mean curvature vector field H. Then, we
have
(1) ∇⊥ m
X H = hH, ϕXiξ, for any vector field X tangent to Σ , i.e., H is C-
parallel;
(2) the mean curvature |H| is constant;
(3) if m = n, then ∆⊥ H = −H.
In the case when the ambient space is a 7-dimensional Sasakian space form,
C. Baikoussis, D. E. Blair, and T. Koufogiorgios classified 3-dimensional integral
C-parallel submanifolds (see [9]). To obtain this classification, they worked with a
special local orthonormal basis. In the following, we will briefly recall how this basis
is constructed (see also [47]).
Let  : Σ3 → N 7 (c) be a non-minimal cmc integral submanifold. Let p be an
arbitrary point of Σ3 and consider a function fp : Up Σ3 → R given by
fp (u) = hσ(u, u), ϕui,
where Up Σ3 = {u ∈ Tp Σ3 : |u| = 1} is the unit sphere in the tangent space Tp Σ3 . If
fp (u) = 0, for all u ∈ Up Σ3 , then, for any v1 , v2 ∈ Up Σ3 such that hv1 , v2 i = 0, we
have that
hσ(v1 , v1 ), ϕv1 i = 0 and hσ(v1 , v1 ), ϕv2 i = 0.
One obtains σ(v1 , v1 ) = 0 and it follows that σ vanishes at p.
Next, assume that fp does not vanish identically. Since Up Σ3 is compact, fp
attains an absolute maximum at a unit vector X1 . It follows that
(
hσ(X1 , X1 ), ϕX1 i > 0, hσ(X1 , X1 ), ϕX1 i ≥ |hσ(w, w), ϕwi|
hσ(X1 , X1 ), ϕwi = 0, hσ(X1 , X1 ), ϕX1 i ≥ 2hσ(w, w), ϕX1 i,
where w is a unit vector tangent to Σ3 at p and orthogonal to X1 . It is easy to see

that X1 is an eigenvector of the shape operator A1 = AϕX1 with the corresponding
eigenvalue λ1 . Then, since A1 is symmetric, we can consider X2 and X3 to be unit
eigenvectors of A1 , orthogonal to each other and to X1 , with the corresponding
eigenvalues λ2 and λ3 . Further, we distinguish two cases.
If λ2 6= λ3 , we can choose X2 and X3 such that
(
hσ(X2 , X2 ), ϕX2 i ≥ 0, hσ(X3 , X3 ), ϕX3 i ≥ 0
hσ(X2 , X2 ), ϕX2 i ≥ hσ(X3 , X3 ), ϕX3 i.
If λ2 = λ3 , we consider f1,p the restriction of fp to {w ∈ Up Σ3 : hw, X1 i = 0},

and we have two subcases:
(1) the function f1,p is identically zero. In this case, we have
(
hσ(X2 , X2 ), ϕX2 i = 0, hσ(X2 , X2 ), ϕX3 i = 0
hσ(X2 , X3 ), ϕX3 i = 0, hσ(X3 , X3 ), ϕX3 i = 0.
(2) the function f1,p does not vanish identically. Then we choose X2 such that
f1,p (X2 ) is an absolute maximum. We have that
(
hσ(X2 , X2 ), ϕX2 i > 0, hσ(X2 , X2 ), ϕX2 i ≥ hσ(X3 , X3 ), ϕX3 i ≥ 0
hσ(X2 , X2 ), ϕX3 i = 0, hσ(X2 , X2 ), ϕX2 i ≥ 2hσ(X3 , X3 ), ϕX2 i.
Now, with respect to the orthonormal basis {X1 , X2 , X3 }, the shape operators
A1 , A2 = AϕX2 , and A3 = AϕX3 , at p, can be written as follows
(4.46)      
λ1 0 0 0 λ2 0 0 0 λ3
A1 =  0 λ2 0  , A2 =  λ2 α β , A3 =  0 β µ  .
0 0 λ3 0 β µ λ3 µ δ
We also have A0 = Aξ = 0. With these notations, we have
λ1 > 0, λ1 ≥ |α|, λ1 ≥ |δ|, λ1 ≥ 2λ2 , λ1 ≥ 2λ3 .
When λ2 6= λ3 we get
α ≥ 0, δ ≥ 0, and α ≥ δ
and, when λ2 = λ3 , we obtain that either
α=β=µ=δ=0
or
α > 0, δ ≥ 0, α ≥ δ, β = 0, and α ≥ 2µ.
We can extend X1 on a neighbourhood Vp of p such that X1 (q) is a maximal
point of fq : Uq Σ3 → R, for any point q of Vp .
If the eigenvalues of A1 have constant multiplicities, then the basis {X1 , X2 , X3 },
defined at p, can be smoothly extended and we can work on the open dense subset
of Σ3 defined by this property.
Using this basis, in [9], the authors proved that, when Σ3 is an integral C-
parallel submanifold, the functions λi , i ∈ {1, 2, 3}, and α, β, µ, δ are constant
on Vp , and then classified 3-dimensional integral C-parallel submanifolds in a 7-
dimensional Sasakian space form. According to this classification, when c > −3, Σ3
is a non-minimal integral C-parallel submanifold if and only if either:
Case I. Σ3 is flat and it is locally a product of three curves which are helices of
osculating orders r ≤ 4, and λ1 = (λ2 − (c + 3)/4)/λ, λ2 = λ3 = λ =
constant
√ 6= 0, α = constant, β = 0, µ = constant, δ = constant, such that
− c + 3/2 < λ < 0, 0 < α ≤ λ1 , α > 2µ, α ≥ δ ≥ 0, (c + 3)/4 + λ2 + αµ −
µ2 = 0 and ((3λ2 − (c + 3)/4)/λ)2 + (α + µ)2 + δ 2 > 0; or
Case II. Σ3 is locally isometric to a product γ × Σ̄2 , where γ is a curve and Σ̄2 is a
C-parallel surface, and either
√ √
(1) λ1p= 2λ2 = −λ √ 3 = c + 3/(2 2), α = µ = δ = 0, and β =
± 3(c + √ 3)/(4 2). In this 2case γ is a helix in N with curvatures
κ1 = 1/ 2 and κ2 = 1, and Σ̄ isplocally isometric to the 2-dimensional
Euclidean sphere of radius ρ = 8/(3(c + 3)); or
(2) λ1 = (λ2 − (c√+ 3)/4)/λ, λ2 = λ3 = λ = constant, α = β = µ = δ = 0,
such that − c + 3/2 < λ < 0 and λ2 6= (c + 3)/12. In this case, γ
is a helix in N with curvatures κ1 = λ1 and κp 2
2 = 1, and Σ̄ is the
2-dimensional Euclidean sphere of radius ρ = 1/ (c + 3)/4 + λ2 .
In the same paper [9], explicit parametric equation of flat 3-dimensional integral
C-parallel submanifolds in S7 (1) were also found. Using the same techniques and
Lemma 4.16, we can extend this result to S7 (c), c > −3.
Theorem 4.50 ([9, 69]). The position vector in the Euclidean space (R8 , h, i) of
a flat 3-dimensional integral C-parallel submanifold in S7 (c), c = 4/a − 3 > −3, is
λ 1
x(u, v, w) = q exp i u E1
λ2 + a1 aλ
1
+p exp(−i(λu − (µ − α)v))E2
a(µ − α)(2µ − α)
1
+p exp(−i(λu + µv + ρ1 w))E3
aρ1 (ρ1 + ρ2 )
1
+p exp(−i(λu + µv − ρ2 w))E4 ,
aρ2 (ρ1 + ρ2 )
where {Ei }4i=1 is an orthonormal basis in C4 with respect to the usual Hermitian
inner product.
In the following, we will study the biharmonicity of maximum dimensional inte-
gral submanifolds in a Sasakian space form N 2n+1 (c).
Our first result follows from Theorem 3.1 and the expression (1.38) of the cur-
vature tensor field RN .
Theorem 4.51 ([70]). An integral submanifold  : Σn → N 2n+1 (c) is biharmonic
if and only if
−∆⊥ H + trace σ(·, AH ·) − c(n+3)+3n−3
(
4 H=0
2
4 trace A∇⊥ H (·) + n grad(|H| ) = 0.
(·)
Corollary 4.52 ([70]). Let N 2n+1 (c) be a Sasakian space form with constant
ϕ-sectional curvature c ≤ (3 − 3n)/(n + 3). Then a cmc integral submanifold Σn in
N 2n+1 (c) is biharmonic if and only if it is minimal.
Proof. Assume that Σn is a biharmonic integral submanifold with constant
mean curvature |H| in N 2n+1 (c). It follows, from Theorem 4.51, that
c(n + 3) + 3n − 3
h∆⊥ H, Hi = htrace σ(·, AH ·), Hi − |H|2
4
c(n + 3) + 3n − 3
= |AH |2 − |H|2 .
4
Thus, from the Weitzenböck formula

1
∆|H|2 = h∆⊥ H, Hi + |∇⊥ H|2 ,
2
one obtains
c(n + 3) + 3n − 3
|H|2 − |AH |2 − |∇⊥ H|2 = 0.
4
If c < (3 − 3n)/(n + 3), this equation is equivalent to H = 0. Now, assume that
c = (3 − 3n)/(n + 3). Since, for integral submanifolds, ∇⊥ H = 0 is equivalent to
H = 0, the above relation is again equivalent to H = 0.
Corollary 4.53 ([70]). Let N 2n+1 (c) be a Sasakian space form with constant
ϕ-sectional curvature c ≤ (3 − 3n)/(n + 3). Then a compact integral submanifold Σn
is biharmonic if and only if it is minimal.
Proof. Assume that Σn is a biharmonic compact integral submanifold. As in
the proof of Corollary 4.52, we have
c(n + 3) + 3n − 3
h∆⊥ H, Hi = − |H|2 + |AH |2
4
and so ∆|H|2 ≥ 0, which implies that |H|2 is constant. Therefore, we obtain that
Σn is minimal in this case too.
Remark 4.54. From Corollaries 4.52 and 4.53, it is easy to see that in a Sasakian
space form N 2n+1 (c) with constant ϕ-sectional curvature c ≤ −3 a biharmonic com-
pact integral submanifold, or a biharmonic integral submanifold Σn with constant
mean curvature, is minimal.
Proposition 4.55 ([70]). Let  : Σn → N 2n+1 (c) be an integral C-parallel sub-
manifold. Then (τ2 ())> = 0.
Proof. Indeed, from Proposition 4.49, we have that |H| is constant and ∇⊥ H
is parallel to ξ, which implies that A∇⊥ H = 0 for any vector field X tangent to Σn ,
X
since Aξ = 0.
Proposition 4.56 ([70]). A non-minimal integral C-parallel submanifold Σn in
N 2n+1 (c)is proper-biharmonic if and only if c > (7 − 3n)/(n + 3) and
c(n + 3) + 3n − 7
trace σ(·, AH ·) = H.
4
Proof. We know, from Proposition 4.49, that ∆⊥ H = −H. Hence, from The-
orem 4.51 and the above proposition, it follows that Σn is biharmonic if and only
if
c(n + 3) + 3n − 7
trace σ(·, AH ·) = H.
4
Next, if Σn verifies the above condition, we contract with H and get
c(n + 3) + 3n − 7
|AH |2 = |H|2 .
4
Since AH and H do not vanish, it follows that c > (7 − 3n)/(n + 3).
Now, let {Xi }ni=1 be an arbitrary orthonormal local frame field on the integral
C-parallel submanifold Σn of a Sasakian space form N 2n+1 (c), and let Ai = AϕXi ,
i ∈ {1, . . . , n}, be the corresponding shape operators. Then, from Proposition 4.56,
we obtain our next result.
Proposition 4.57 ([70]). An integral C-parallel submanifold Σn in N 2n+1 (c),

c > (7 − 3n)/(n + 3), is proper-biharmonic if and only if
    
hA1 , A1 i . . . hA1 , An i trace A1 trace A1
.. .. .. .. ..
 = k ,
    
 . . .  . .
hAn , A1 i . . . hAn , An i trace An trace An
where k = (c(n + 3) + 3n − 7)/4.
We will now turn our attention to the case when the ambient space is 7-dimen-
sional. To treat this case we shall again consider the special basis {X1 , X2 , X3 } at
each point on the submanifold Σ3 .
First, by identifying the shape operators Ai with their corresponding matrices,
from Proposition 4.57, we get the following result.
Proposition 4.58 ([70]). An integral C-parallel submanifold Σ3 in N 7 (c), c >
−1/3, is proper-biharmonic if and only if
!   
3 trace A1 trace A1
X 3c + 1 
(4.47) A2i  trace A2  = trace A2  ,
trace A3 2 trace A3
i=1
where matrices Ai are given by (4.46).

Now, we can state our next theorem.
Theorem 4.59 ([70]). A 3-dimensional integral C-parallel submanifold Σ3 in
N 7 (c)is proper-biharmonic if and only if either:
(1) c > −1/3 and Σ3 is flat and it is locally a product of three curves, as follows:
• a helix with curvatures κ1 = (λ2 − (c + 3)/4)/λ and κ2 = 1,
• a helix of order 4 with curvatures
p p p
κ1 = λ2 + α2 , κ2 = (α/κ1 ) λ2 + 1, κ3 = −(λ/κ1 ) λ2 + 1,
p p p
κ1 = λ2 + µ2 + δ 2 , κ2 = (δ/κ1 ) λ2 + µ2 + 1, κ3 = (κ2 /δ) λ 2 + µ2 ,
p
if δ 6= 0, or a circle with curvature κ1 = λ2 + µ2 , if δ = 0,
where λ, α, µ, δ are constants given by
(c + 3)2

2 c + 3 4 2
3λ − 3λ − 2(c + 1)λ + + λ4 ((α + µ)2 + δ 2 ) = 0,





 4 16
(α + µ) 5λ2 + α2 + µ2 − 7c+5 2
4 + µδ = 0

(4.48)
δ 5λ2 + δ 2 + 3µ2 + αµ − 7c+5 =0


4



 c+3 + λ2 + αµ − µ2 = 0

4
√
such that − c + 3/2 < λ < 0, 0 < α ≤ (λ2 − (c + 3)/4)/λ, α ≥ δ ≥ 0,
α > 2µ, and λ2 6= (c + 3)/12; or
(2) Σ3 is locally isometric to a product γ × Σ̄2 between a curve and a C-parallel
surface of N , and either √
(a) c = 5/9, γ is a helix in N 7 (5/9), with curvatures κ1 = 1/ 2 and
κ2 = 1, and Σ̄2 is√locally isometric to the 2-dimensional Euclidean
sphere with radius 3/2; or
√
(b) c ∈ [(−7 + 8 3)/13, +∞) \ {1}, γ is a helix in N 7 (c) with curvatures
κ1 = (λ2 − (c + 3)/4)/λ and κ2 = 1, and Σ̄2 is √
locally isometric to the
2-dimensional Euclidean sphere with radius 2/ 4λ2 + c + 3, where
( √
4c+4± 13c2 +14c−11
, if c < 1
(4.49) λ < 0 and λ2 = 4c+4−√13c 12
2 +14c−11
12 , if c > 1.
Proof. First, from Proposition 4.58, we can see that c > −1/3. Also, we easily
get that equation (4.47) is equivalent to
(4.50)
 P
3 2 − 3c+1 + (α + µ)(αλ + µλ ) + (β + δ)(βλ + δλ ) = 0


 t λ
i=1 i 2 2 3 2 3

2 2 2 2 3c+1
t(αλ2 + µλ3 ) + (α + µ) 2λ2 + α + 3β + µ + βδ − 2 + µ(β + δ)2 = 0

t(βλ2 + δλ3 ) + β(α + µ)2 + (β + δ) 2λ2 + δ 2 + 3µ2 + β 2 + αµ − 3c+1 = 0,


3 2
where t = trace A1 = 3i=1 λi .

P
In the following, we shall split the study of this system, as Σ3 is given by one case
or the other in the classification of non-minimal integral C-parallel submanifolds.
Case I. The system (4.50) is equivalent to the system given by the first three
equations of (4.48). Now, Σ3 is non-minimal if and only if at least one of the
components of the mean curvature vector field H does not vanish and, from the first
equation of (4.48), it follows that λ2 must be different from (c + 3)/12. Thus, again
using [9] for the expressions of the curvatures of the three curves, we obtain the first
case of our theorem.
Case II. (1) In this case, the second equation of (4.50) is identically satisfied and
the other two are equivalent to c = 5/9. Thus, from the classification of the integral
C-parallel submanifolds, we get the first part of the second case of the theorem.
(2) The second and the third equation of (4.50) are satisfied, in this case, and
the first equation is equivalent to
(c + 3)2
3λ4 − 2(c + 1)λ2 + = 0.
16
This equation has solutions if and only if
√ √
−7 − 8 3 i h −7 + 8 3
c ∈ − ∞, ∪ , +∞ ,
13 13
and these solutions are given by
√
4c + 4 ± 13c2 + 14c − 11
λ2 = .
12
√
Since c > −1/3, it follows that c ∈ [(−7 + 8 3)/13, +∞). Moreover, if c = 1,
from the above relation, it follows that λ2 must √ be equal to 1 or 1/3, which is
a contradiction, and, therefore, c √ ∈ [(−7 + 8 3)/13, +∞) \ {1}. Further, it is
easy to check 2 4 + 13c2 +√14c − 11)/12 < (c + 3)/4 if and only if
√ that λ = (4c +2
c ∈ [(−7 + 8 3)/13,√1) and λ = (4c + 4 − 13c2 + 14c − 11)/12 < (c + 3)/4 if and
only if c ∈ [(−7 + 8 3)/13, +∞) \ {1}.
Next, we consider the standard model for simply connected Sasakian space forms
N 7 (c) with c + 3 > 0, which is the sphere S7 endowed either with its canonical or
deformed Sasakian structure. Using Theorems 4.50 and 4.59, we easily get the
following theorem.
Theorem 4.60 ([70]). A 3-dimensional integral C-parallel submanifold Σ3 in
S7 (c),c = 4/a − 3 > −3, is proper-biharmonic if and only if either:
(1) c > −1/3 and Σ3 is flat, locally is a product of three curves, and its position
vector in C4 is
λ 1
x(u, v, w) = q exp i u E1
λ2 + 1 aλ
a
1
a(µ − α)(2µ − α)
1
aρ1 (ρ1 + ρ2 )
1
aρ2 (ρ1 + ρ2 )
p
where ρ1,2 = ( 4µ(2µ − α) + δ 2 ± δ)/2 and λ, α, µ, δ are real constants
√
given by (4.48) such that −1/ a < λ < 0, 0 < α ≤ (λ2 −1/a)/λ, α ≥ δ ≥ 0,
α > 2µ, λ2 6= 1/3a, and {Ei }4i=1 is an orthonormal basis of C4 with respect
to the usual Hermitian inner product; or
(2) Σ3 is locally isometric to a product γ × Σ̄2 between a curve and an integral
C-parallel surface of N and either √
(a) c = 5/9, γ is a helix in S7 (5/9) with curvatures κ1 = 1/ 2 and κ2 = 1,
and Σ̄2√is locally isometric to the 2-dimensional Euclidean sphere with
radius 3/2;√or
(b) c ∈ [(−7 + 8 3)/13, +∞) \ {1}, γ is a helix in S7 (c) with curvatures
κ1 = (λ2 − (c + 3)/4)/λ and κ2 = 1, and Σ̄2 is √ locally isometric to the
2-dimensional Euclidean sphere with radius 2/ 4λ2 + c + 3, where
( √
4c+4± 13c2 +14c−11
2 12 , if c < 1
λ < 0 and λ = 4c+4−√13c 2 +14c−11
12 , if c > 1.
When c = 1, we can completely solve (4.48) and, applying the previous theorem,
we get the following corollary.
Corollary 4.61 ([70]). A 3-dimensional integral C-parallel submanifold M 3 of
S7 (1)is proper-biharmonic if and only if it is flat, locally is a product of three curves,
and its position vector in C4 is
√
1 √ 1 1 4 3
x(u, v, w) = − √ exp(−i 5u)E1 + √ exp i √ u − √ v E2
6 6 5 10
√ √
1 1 3 3 2
+ √ exp i √ u + √ v − w E3
6 5 10 2
√ √
1 1 3 2
+ √ exp i √ u + √ v + w E4 ,
2 5 10 2
where {Ei }4i=1 is an orthonormal basis of C4 with respect to the usual√Hermitian inner
product. Moreover, the xu -curve is a helix with curvatures κ1 = 4 5/5 and κ2 = 1,
√ √ √ √
the xv -curve√is √
a helix of order 4 with curvatures κ1 = 29/ 10, κ2 = 9 2/ 145,
and
√ √ κ3 = 2 3/√ 145, √ and the xw -curve
√ √is a helix of order 4 with curvatures κ1 =
5/ 2, κ2 = 2 3/ 10, and κ3 = 3/ 10.
Remark 4.62. By a straightforward computation, we can deduce that the map
3 3 8
√ √T =√R /Λ into
x factorizes to a map from the torus √ R , where Λ√ is the√lattice gen-
√
erated by the vectors√a1 = (6π/ 5, 3π/ 10, π/ 2), a2 = (0, −3 5π/ 6, −π/ 2)
and a3 = (0, 0, −4π/ 2), and the quotient map is a Riemannian immersion.
By the meaning of Theorem 4.28, we know that the cylinder over x, given by
y(t, u, v, w) = φt (x(u, v, w)),
is a proper-biharmonic map into S7 (1). Moreover, we have the following proposition.
Proposition 4.63 ([70]). The cylinder over x determines a proper-biharmonic
4 4 7
Riemannian embedding from √ the torus T = R /Λ √ into S , where the lattice
√ Λ
is generated by a1 =√(2π/ 6, 0, 0, 0), a2 = (0, 2π/ 6, 0, 0), a3 = (0, 0, 2π/ 6, 0),
and a4 = (0, 0, 0, 2π/ 2). The image of√ this embedding is the Riemannian product
between a√Euclidean circle of radius 1/ 2 and three other Euclidean circles, each of
radius 1/ 6.
Proof. Since the flow of the characteristic vector field ξ is given by φt (z) =
exp(−it)z, we get
√
1 √ 1 1 4 3
y(t, u, v, w) = − √ exp(−i(t + 5u))E1 + √ exp i − t + √ u − √ v E2
6 6 5 10
√ √
1 1 3 3 2
+ √ exp i − t + √ u + √ v − w E3
6 5 10 2
√ √
1 1 3 2
+ √ exp i − t + √ u + √ v + w E4 ,
2 5 10 2
where {Ei }4i=1 is an orthonormal basis of C4 with respect to the usual Hermitian
inner product.
Now, we consider the following two orthogonal transformations of R4 :
 √  √2
 √1 t + √1 u + √3 v + 1 w = t0 √ t0 + √2 u0 = e t
 2 √10 2√ 5 2  6√ 6

 
 √
 √2 u − √6 v − 2 w = u0
 − √2 t0 + √1 u0 − √3 v 0 = u

√5 4 √5 4 e
and √6 6 √6
5
√ v− √ w =v3 0 2 0 1 0 3 0

 2 2 2 2

− √6 t + √6 u + √6 v = ve

 √ 

 √1 1 3 1 0
t − √10 u − 2√5 v − 2 w = w w0 = w.
 
2
e
Then we obtain
1 √ 1 √ 1 √
ye(e
t, u e = − √ exp(−i( 6e
e, ve, w) t))E1 + √ exp(i( 6e
u))E2 + √ exp(i( 6e
v ))E3
6 6 6
1 √
+ √ exp(i( 2w))E
e 4,
2
which ends the proof.
CHAPTER 5
Biharmonic Submanifolds in Complex Space Forms
1. Introduction
We first consider submanifolds in a complex space such that the image of their
mean curvature vector field through the structure tensor of the ambient space is
either tangent or normal. We find the conditions for such submanifolds to be bi-
harmonic (Propositions 5.1 and 5.7), the admissible range for their mean curvature
when they are proper-biharmonic and cmc (Propositions 5.4 and 5.8), and the value
of the scalar curvature of submanifolds in the first considered case that also are
proper-biharmonic (Proposition 5.6).
Next, we consider the Hopf fibration and find the relation between the bitension
field of a submanifold in CP n and the bitension field of its corresponding Hopf
cylinder in S2n+1 (Theorem 5.12). Then, we use this result to show that a cmc
hypersurface in CP n or a Lagrangian pmc submanifold in CP n are biharmonic if and
only if the Hopf cylinders over them are −4-biharmonic in S2n+1 (Propositions 5.14
and 5.15). When, for a submanifold in CP n , the image of its mean curvature vector
field through the structure tensor is normal, we prove that such a submanifold is
biharmonic if and only if so it is the corresponding Hopf cylinder (Proposition 5.16).
We continue with a result on the biharmonicity of some Clifford type sub-
manifolds in CP n (Theorem 5.18) and then use it to produce examples of proper-
biharmonic submanifolds (Examples 5.20-5.23).
The next section is devoted to the study of proper-biharmonic curves in CP n .
We present classification results (Theorems 5.26, 5.30, and 5.34) and also character-
izations in terms of horizontal lifts to S2n+1 via Hopf fibration (Propositions 5.27,
5.28, and 5.31). Also, the complete classification of proper-biharmonic curves in
CP 2 is given here (Theorem 5.39).
Using, among other tools, a Simons type equation, we classify all complete
pmc proper-biharmonic surfaces with non-negative Gaussian curvature in a com-
plex space form (Theorem 5.49).
We end the chapter with a characterization of 3-dimensional proper-biharmonic
parallel Lagrangian submanifolds in CP 3 in terms of their horizontal lifts to S2n+1
(Theorem 5.53, Corollary 5.54).
2. General biharmonicity results for submanifolds in complex space

forms
Let N n (c) be a complex space form of constant holomorphic sectional curvature
c. Let us denote by J¯ the complex structure, by h, i the Riemannian metric, and by
R̄ the curvature tensor on N n (c) (the overbar notation will be justified in the next
section). We recall that R̄ is given by equation (1.6).
129
130 Biharmonic Submanifolds in Complex Space Forms
Now, let ̄ : Σ̄ → N n (c) be the canonical inclusion of a submanifold Σ̄ in N n (c)

of real dimension m̄.
We use Theorem 3.1 and equation (1.6), to easily obtain the following proposi-
tion.
Proposition 5.1 ([64]). Let Σ̄ be a real submanifold of N n (c) of dimension m̄
such that J¯H̄ is tangent to Σ̄. Then Σ̄ is biharmonic if and only if
−∆⊥ H̄ + trace σ̄(·, ĀH̄ (·)) − c(m̄+3)
(
4 H̄ = 0
(5.1) 2
4 trace Ā∇⊥ H̄ (·) + m̄ grad(|H̄| ) = 0,
(·)
where Ā denotes the shape operator and σ̄ the second fundamental form.
If Σ̄ is a hypersurface, then J¯H̄ is tangent to Σ̄, and, using the previous propo-
sition, we recover the following result in [88].
Corollary 5.2 ([88]). Let Σ̄ be a real hypersurface of N n (c) of non-zero con-
stant mean curvature. Then it is proper-biharmonic if and only if
c(n + 1)
|σ̄|2 = .
2
We can also use Proposition 5.1 to study the biharmonicity of Lagrangian sub-
manifolds, i.e., submanifolds Σ̄ of dimension n in N n (c) such that ̄∗ Ω = 0, where Ω
is the fundamental 2-form on N n (c) defined by Ω(X, Y ) = hX, JY ¯ i, for any vector
n
fields X and Y tangent to N (c).
Corollary 5.3 ([64]). Let Σ̄ be a pmc Lagrangian submanifold in N n (c). Then
it is biharmonic if and only if
c(n + 3)
trace σ̄(·, ĀH̄ (·)) = H̄.
4
In the sequel, we shall consider only the case of complex space forms with posi-
tive holomorphic sectional curvature. A partial motivation of this fact is that Corol-
lary 5.2 rules out the case c ≤ 0. As usual, we consider the complex projective space
CP n endowed with the Fubini-Study metric, as a model for the complex space form
of positive constant holomorphic sectional curvature 4.
Proposition 5.4 ([64]). Let Σ̄ be a non-minimal cmc real submanifold of CP n
of dimension m̄ such that J¯H̄ is tangent to Σ̄. Then the following hold:
(1) If Σ̄ is proper-biharmonic, then |H̄|2 ∈ (0, (m̄ + 3)/m̄].
(2) If |H̄|2 = (m̄ + 3)/m̄, then Σ̄ is proper-biharmonic if and only if it is
pseudo-umbilical and ∇⊥ H̄ = 0.
Proof. Let Σ̄ be a cmc proper-biharmonic real submanifold of CP n of dimen-
sion m̄ such that J¯H̄ is tangent to Σ̄. Then, we have
∆⊥ H̄ = −(m̄ + 3)H̄ + trace σ̄(·, ĀH̄ (·)),
and, therefore,
m̄
X
⊥ 2
h∆ H̄, H̄i = −(m̄ + 3)|H̄| + hσ̄(Xi , ĀH̄ (Xi )), H̄i = −(m̄ + 3)|H̄|2 + |ĀH̄ |2 .
i=1
Biharmonic Submanifolds in Complex Space Forms 131
Replacing in the Weitzenböck formula

1
∆|H̄|2 = h∆⊥ H̄, H̄i + |∇⊥ H̄|2 ,
2
and since |H̄| = constant, we obtain
(5.2) (m̄ + 3)|H̄|2 = |ĀH̄ |2 + |∇⊥ H̄|2 .
Now, let p be an arbitrary point of Σ̄ and let {Xi }m̄
i=1 be an orthonormal basis
of Tp Σ̄ such that ĀH̄ (Xi ) = λi Xi . We have
λi = hĀH̄ (Xi ), Xi i = hσ̄(Xi , Xi ), H̄i,
Pm̄
which implies = m̄|H̄|2 , or, equivalently,
i=1 λi
Pm̄
λi
|H̄|2 = i=1 .
m̄
Then the squared norm of ĀH̄ is given by
m̄
X m̄
X
|ĀH̄ |2 = hĀH̄ (Xi ), ĀH̄ (Xi )i = (λi )2
i=1 i=1
and, replacing in (5.2), we get
m̄ m̄
( m̄ 2
P
m̄ + 3 X i=1 λi )
X
⊥
λi = 2 2
(λi ) + |∇ H̄| ≥ + |∇⊥ H̄|2 .
m̄ m̄
i=1 i=1
Hence, one obtains
(m̄ + 3)|H̄|2 ≥ m̄|H̄|4 + |∇⊥ H̄|2 ≥ m̄|H̄|4
and, therefore, we have |H̄|2 ∈ (0, (m̄ + 3)/m̄].
Next, if |H̄|2 = (m̄ + 3)/m̄, since Σ̄ is biharmonic, the above inequalities become
equalities, and then λ1 = · · · = λm̄ and ∇⊥ H̄ = 0, i.e., Σ̄ is pseudo-umbilical and
pmc.
Conversely, it is clear that, if |H̄|2 = (m̄ + 3)/m̄ and Σ̄ is pseudo-umbilical with
∇⊥ H̄ = 0, then Σ̄ is proper-biharmonic.
Remark 5.5. Later in this chapter, we shall see that the upper bound of |H̄|2
is reached in the case of curves.
From Corollary 5.2, we obtain our next result.
Proposition 5.6 ([64]). Let Σ̄ be a cmc proper-biharmonic real hypersurface of
CP n . Then its scalar curvature s is given by
s = 4n2 − 2n − 4 + (2n − 1)2 |H̄|2 = constant .
Another important family of submanifolds in CP n is that of those submanifolds
Σ̄ for which J¯H̄ is normal to Σ̄. In this case, again using Theorem 3.1 and formula
(1.6), we have the following result.
Proposition 5.7 ([64]). Let Σ̄ be a real submanifold of CP n of dimension m̄
such that J¯H̄ is normal to Σ̄. Then Σ̄ is biharmonic if and only if
(
−∆⊥ H̄ + trace σ̄(·, ĀH̄ (·)) − m̄H̄ = 0
(5.3)
4 trace Ā∇⊥ H̄ (·) + m̄ grad(|H̄|2 ) = 0.
(·)
Moreover, if J¯H̄ is normal to Σ̄ and the mean curvature vector field is parallel, then
Σ̄ is biharmonic if and only if
trace σ̄(·, ĀH̄ (·)) = m̄H̄.
If the mean curvature is constant, then we can prove that it is bounded, in this
case too, as shown by the following proposition.
Proposition 5.8 ([64]). Let Σ̄ be a cmc real submanifold of CP n of dimension
m̄ such that J¯H̄ is normal to Σ̄. We have
(1) If Σ̄ is proper-biharmonic, then |H̄|2 ∈ (0, 1].
(2) If |H̄|2 = 1, then Σ̄ is proper-biharmonic if and only if it is pseudo-umbilical
and ∇⊥ H̄ = 0.
Remark 5.9. As we will see later on, the upper bound is reached in the case of
curves.
3. The Hopf fibration and the biharmonic equation

Let π : Cn+1 \ {0} → CP n be the natural projection. Then π restricted to the
sphere S2n+1 of Cn+1 gives rise to the Hopf fibration π : S2n+1 → CP n and, if c = 4,
then π : S2n+1 → CP n defines a Riemannian submersion. In the sequel, we shall
look at S2n+1 as a hypersurface of R2n+2 and denote by Jˆ the complex structure of
R2n+2 .
Let Σ̄ be a real submanifold of CP n of dimension m̄ and denote by Σ := π −1 (Σ̄)
the Hopf cylinder over Σ̄ and by ̄ : Σ̄ → CP n and  : Σ → S2n+1 the respective
inclusions.
We shall now find the relation between the bitension field ̄ and the one of
. First, let us consider a local orthonormal frame field {X̄k }m̄ k=1 tangent to Σ̄,
2n
1 ≤ m̄ ≤ 2n − 1, and a local orthonormal frame field {η̄α }α=m̄+1 normal to Σ̄.
Denote by Xk = X̄kH and ηα = η̄αH the horizontal lifts with respect to the Hopf map
and by ξ the Hopf vector field on S2n+1 which is tangent to the fibres of the Hopf
ˆ for any p ∈ S2n+1 . Then {Xk , ξ} is a local orthonormal
fibration, i.e., ξ(p) = −Jp,
frame field tangent to Σ and {ηα } is a local orthonormal frame field normal to Σ.
Lemma 5.10 ([64]). Let X = X̄ H ∈ C(T Σ), where X̄ ∈ C(T Σ̄) and V = V̄ H ∈
C(−1 (T S2n+1 )),
with V̄ ∈ C((̄)−1 (T CP n )). Then
∇X V = (∇̄X̄ V̄ )H + hV, JXiξ
ˆ = (∇̄X̄ V̄ )H + (hV̄ , J¯X̄i ◦ π)ξ,
where ∇ and ∇̄ denote the pull-back connections on −1 (T S2n+1 ) and (̄)−1 (T CP n ),
respectively.
Proof. Decomposing ∇X V in its horizontal and vertical components, we have
∇X V = ∇X̄ H V̄ H = (∇̄X̄ V̄ )H + h∇X V, ξiξ
and then, since
h∇X V, ξi = − hV, ∇X ξi = −hV, ∇
ˆ X ξ + hX, ξipi
ˆ X Jpi
=hV, ∇ ˆ = hV, JXi
ˆ = hV̄ , J¯X̄i ◦ π,
ˆ is the Levi-Civita connection on E2n+2 , we conclude.
where ∇
Lemma 5.11 ([64]). If V = V̄ H ∈ C(−1 (T S2n+1 )), V̄ ∈ C((̄)−1 (T CP n )), then

ˆ )> )ξ + hV, Jτ
∆ V = (∆̄ V̄ )H + 2 div((JV ˆ ()iξ + V − J( ˆ )> ,
ˆ JV
where ∆ and ∆̄ are the rough Laplacians acting on sections of −1 (T S2n+1 ) and
(̄)−1 (T CP n ), respectively, whilst V > denotes the component of V tangent to Σ.
Proof. The Laplacian ∆ is given by
m̄
{∇Xi ∇Xi V − ∇∇M X V } + ∇ξ ∇ξ V − ∇∇M ξ V.
X

−∆ V =
Xi i ξ
i=1
In the following, we will compute each term in the above expression.
From Lemma 5.10, we have
∇ ∇ V =(∇̄ ∇̄ V̄ )H + h∇ V, JX
Xi XiX̄i X̄i
ˆ i iξ + ∇ (hV, JX
Xi
ˆ i iξ)
Xi
̄ ̄ H  ˆ i iξ + hV, ∇ JX
ˆ i iξ + hJV,
ˆ Xi iJX
ˆ i.
=(∇X̄ ∇X̄ V̄ ) + 2h∇Xi V, JX Xi
i i
Using the fact that ∇Xi JX

ˆ i = J∇ ˆ  Xi + ξ, we get
Xi
(5.4) ∇Xi ∇Xi V = (∇X̄ ∇X̄ V̄ ) + 2h∇Xi V, JX

  ̄ ̄ H ˆ  Xi iξ
ˆ i iξ + hV, J∇
Xi
i i
ˆ JV,
+J(h ˆ Xi iXi ).
We also have
(5.5) ∇∇M X V = (∇̄ ˆ M
V̄ )H + hV, J∇ Xi Xi iξ.
Xi i ∇M̄
X̄
X̄i
i
Summing (5.4) and (5.5) up, we find

m̄ m̄
h∇Xi V, JX (∇Xi Xi − ∇M
X X
−∆ V  ̄
= −(∆ V̄ ) + 2 H ˆ i iξ + hV, Jˆ Xi Xi )iξ
i=1 i=1
m̄
ˆ Xi iXi ) + ∇ ∇ V
X
+ ˆ JV,
J(h ξ ξ
i=1
m̄
h∇Xi V, JX
X
̄
= −(∆ V̄ ) + 2 H ˆ i iξ + hV, Jτ
ˆ ()iξ
i=1
ˆ )> + ∇ ∇ V.
ˆ JV
+J( ξ ξ
Next, we get the following
m̄ m̄
h∇Xi V, JX ˆ ∇ Xi i}
X X
ˆ ii = ˆ Xi i + hJV,
{−Xi hJV, Xi
i=1 i=1
m̄
X
ˆ τ ()i −
=hJV, ˆ Xi i − hJV,
{Xi hJV, ˆ ∇MXi Xi i}
i=1
ˆ )> ).
ˆ τ ()i − div((JV
=hJV,
Finally, one obtains
∇ξ V =H(∇ξ V ) + h∇ξ V, ξiξ = H(∇ξ V )
=H(∇V ξ) = H(∇
ˆ V ξ + hV, ξip) = H(−JV
ˆ ) = −JV,
ˆ
which gives ∇ξ ∇ξ V = −V and we conclude using all the above equations.
Before computing the relation between the bitension fields we need to find the
trace of the curvature operators. After a straightforward computation, one obtains
2n+1
(5.6) trace RS (d, τ ())d = −(m̄ + 1)τ ()
and
n
(5.7) trace RCP (d̄, τ (̄))d̄ = −m̄τ (̄) + 3J( ¯ (̄))> .
¯ Jτ
We can now state the main result of this section.
Theorem 5.12 ([64]). Let ̄ : M̄ → CP n be a real submanifold of dimension m̄
and denote by  : Σ = π −1 (Σ̄) → S2n+1 the corresponding Hopf cylinder. Then, we
have that
(5.8) ˆ ())> + 2 div((Jτ
ˆ Jτ
(τ2 (̄))H = τ2 () − 4J( ˆ ())> )ξ.
Proof. From (5.6) we have τ2 () = ∆ τ () + (m̄ + 1)τ () and then, since τ () =
(τ (̄))H , using Lemma 5.11 and formula (5.7), we come to the conclusion.
Remark 5.13. We have the following immediate applications of Theorem 5.12:
• Using the horizontal lift, it is straightforward to check that (5.8) can be
written as
¯ (̄))> )H + 2(div ((Jτ
¯ Jτ
(τ2 (̄))H = τ2 () − 4(J( ¯ (̄))> ) ◦ π)ξ.
Σ̄
ˆ () is normal to Σ, then τ2 (̄) = 0 if and only if τ2 () = 0.
• If Jτ
ˆ () is tangent to Σ, then τ2 (̄) = 0 and div ((Jτ
• If Jτ ¯ (̄))> ) = 0 if and only
M̄
if τ2 () + 4τ () = 0.
• Let us assume that locally Σ = π −1 (Σ̄) = S1 × Σ, e where Σ e is an integral
submanifold of S 2n+1 , i.e., hXpe, ξ(e
e p)i = 0 for any vector Xpe tangent to Σ.
e e
Denote by e :Σ e → S2n+1 the canonical inclusion and {φt } the flow of ξ.
We know, from Theorem 4.28, that τ2 ()(t,ep) = (dφt )pe(τ2 (e )), and we can
check that, at pe,
(τ2 (̄))H = τ2 (e
) − 4J(ˆ Jτ
ˆ (e))> + 2 div e ((Jτ ))> )ξ.
ˆ (e
Σ
To state our next results, we first need to recall that a smooth map ψ : M → N
between two Riemannian manifolds is called λ-biharmonic if it is a critical point of
the λ-bienergy
E2 (ψ) + λE(ψ),
where λ is a real constant. The critical points of the λ-bienergy satisfy the equation
τ2 (ψ) − λτ (ψ) = 0.
Proposition 5.14 ([64]). Let Σ̄ be a cmc real hypersurface of CP n and Σ =
π −1 (Σ̄) the Hopf cylinder over Σ̄. Then τ2 (̄) = 0 if and only if τ2 () + 4τ () = 0,
i.e.,  is (−4)-biharmonic.
Proof. We have (Jτ ¯ (̄))> = Jτ ¯ (̄) and, therefore, it only remains to prove that
divM̄ (Jτ¯ (̄)) = 0.
Let η̄ be a local unit section in the normal bundle of Σ̄ in CP n and consider
{X̄1 , J¯X̄1 , . . . , X̄n−1 , J¯X̄n−1 , J¯η̄} a local orthonormal frame field tangent to Σ̄. Since
Σ̄ is a cmc hypersurface, it is enough to prove that divM̄ (J¯η̄) = 0. But, denoting by
Āη̄ the shape operator of Σ̄, we have
h∇Σ̄ ¯ ¯ h∇Σ̄ ¯ ¯ ¯
X̄a J η̄, X̄a i = hĀη̄ (X̄a ), J X̄a i, J¯X̄b J η̄, J X̄b i = −hĀη̄ (X̄b ), J X̄b i,
for any 1 ≤ a, b ≤ n − 1, and h∇Σ̄

J¯η̄
J¯η̄, J¯η̄i = 0, that concludes the proof.
Proposition 5.15 ([64]). Let Σ̄ be a Lagrangian pmc submanifold of CP n and

Σ = π −1 (Σ̄) the Hopf cylinder over Σ̄. Then ̄ is biharmonic if and only if  is
(−4)-biharmonic.
Proof. Since Σ̄ is Lagrangian, we have that dim Σ̄ = n and J(T ¯ Σ̄) = N Σ̄ (and,
¯ ¯
therefore, J(N Σ̄) = T Σ̄). We get that Jτ (̄) ∈ C(T Σ̄) and we shall prove that
¯ (̄) = 0 which implies div (Jτ
∇Σ̄ Jτ ¯ (̄)) = 0. Indeed, for any vector fields X̄ and Ȳ
Σ̄
tangent to Σ̄, we have
¯ ̄ ¯ ¯ ̄ ¯
h∇Σ̄
X̄ Jτ (̄), Ȳ i =h∇X̄ Jτ (̄), Ȳ i = hJ∇X̄ τ (̄), Ȳ i = h−J Āτ (̄) (X̄), Ȳ i
=0
and conclude.
Our last result in this section is the following proposition.

¯ (̄)
Proposition 5.16 ([64]). Let Σ̄ be a real submanifold of CP n , such that Jτ
−1
is normal to Σ̄, and Σ = π (Σ̄) the Hopf cylinder over Σ̄. Then ̄ is biharmonic if
and only if  is biharmonic.
4. Biharmonic submanifolds of Clifford type in CP n

For a fixed n > 1, consider the spheres Sa2p+1 ⊂ R2p+2 = Cp+1 and S2q+1
b ⊂
R2q+2 = Cq+1 (here a and b are the radii of the two spheres), with a2 + b2 = 1
p,q
and p + q = n − 1. Denote by Ta,b = Sa2p+1 × S2q+1
b ⊂ S2n+1 the Clifford torus.
Let now Σ1 be a minimal submanifold of Sa2p+1 of dimension m1 and Σ2 a minimal
submanifold of S2q+1
b of dimension m2 . The submanifold Σ1 × Σ2 is then√minimal
p,q
in Ta,b and, therefore, proper-biharmonic in S2n+1 if and only if a = b = 2/2 and
m1 6= m2 (see [22]). If Σ1 × Σ2 is invariant under the action of the one-parameter
group of isometries generated by the Hopf vector field ξ on S2n+1 , then it projects
onto a submanifold of CP n . We will find the conditions that a, b, m1 , and m2 have
to satisfy such that this submanifold to be proper-biharmonic.
We begin with the following lemma.
p,q
Lemma 5.17 ([64]). Let us denote by 1 : Σ1 ×Σ2 → Ta,b the inclusion of Σ1 ×Σ2
p,q 2n+1
in the Clifford torus and by  : Ta,b → S the inclusion of the Clifford torus into
the sphere. Then, we have
(
τ ( ◦ 1 ) = ( ab m2 − ab m1 )η
(5.9) 2 2
τ2 ( ◦ 1 ) = ( ab m2 − ab m1 ) m1 + m2 − ab 2 m1 − ab2 m2 η,

p,q
where η is the unit normal section in the normal bundle of Ta,b in S2n+1 given by
η(x, y) = ((b/a)x, −(a/b)y), x ∈ S2p+1
a , y ∈ Sb2q+1 .
p,q p,q
Proof. We identify X = (X, 0) ∈ T(x,y) Ta,b , Y = (0, Y ) ∈ T(x,y) Ta,b and then,
after a straightforward computation, one obtains
b a
∇X η = −A (X) = X and ∇Y η = −A (Y ) = − Y.
a b
Let {Xk = (Xk , 0)} be a local orthonormal frame field tangent to Sa2p+1 and {Yl =
(0, Yl )} a local orthonormal frame field tangent to S2q+1
b . Then, applying the com-
position law for the tension field and using that 1 is harmonic, we have
τ ( ◦ 1 ) =d(τ (1 )) + trace ∇d(d1 , d1 )
m1 m2 a
X X b
= hA (Xk ), Xk iη + hA (Yl ), Yl iη = m2 − m1 η.
b a
k=1 l=1
To compute τ2 ( ◦ 1 ), let us choose, around (x, y) ∈ Σ1 × Σ2 a frame field {(Xk , Yl )}

such that {Xk }m m2
k=1 is a geodesic frame field around x and {Yl }l=1 is a geodesic frame
1
field around y. Then, at (x, y), we have

m1 m2 1 m 2 m
bX aX
∇◦ ◦1
∇◦ ◦1
∇◦ ∇◦
X X
(5.10) ∆◦1 η = 1
Xk ∇Xk η +
1
Yl ∇ Yl η = Xk
1
Xk − 1
Yl Yl
a b
k=1 l=1 k=1 l=1
m1 p,q m2
b Ta,b aX T p,q
X
= σ  (Xk , Xk ) + ∇Xk Xk − σ  (Yl , Yl ) + ∇Yla,b Yl
a b
k=1 l=1
m1 m2
b X a X
= σ  (Xk , Xk ) − σ  (Yl , Yl )
a b
k=1 l=1
b2 a 2
= − 2 m1 − 2 m2 η.
a b
Finally, using the standard formula for the curvature of S2n+1 , we get
2n+1
trace RS (d( ◦ 1 ), τ ( ◦ 1 ))d( ◦ 1 ) = − (m1 + m2 )τ ( ◦ 1 )
a b
= − (m1 + m2 ) m2 − m1 η,
b a
that, together with (5.10), leads to the conclusion.
Theorem 5.18 ([64]). Let π : S2n+1 → CP n be the Hopf map and Σ = Σ1 × Σ2
the product of two minimal submanifolds of Sa2p+1 and S2q+1
b , respectively. If Σ is
invariant under the action of the one-parameter group of isometries generated by the
Hopf vector field ξ on S2n+1 , then π(Σ) is a proper-biharmonic submanifold of CP n
if and only if Σ is (−4)-biharmonic, that is
a b b2 a2
a2 + b2 = 1,m2 − m1 6= 0, and m1 + m2 = 4 + m1 + m2 ,
b a a2 b2
where m1 and m2 are the dimensions of Σ1 and Σ2 , respectively.
Proof. The Hopf vector field ξ is a Killing vector field on S2n+1 that, at a point
(x, y), is given by
ξ = −(−x2 , x1 , . . . , −x2p+2 , x2p+1 , −y 2 , y 1 , . . . , −y 2q+2 , y 2q+1 ) = (ξ1 , ξ2 ).
Since Σ1 × Σ2 is invariant under the action of the one-parameter group of isometries
generated by ξ, it remains Killing when restricted to Σ1 × Σ2 . Since
ˆ = − b ξ1 , a ξ2 ,

Jη
a b
ˆ is a Killing vector field on Σ1 × Σ2 .
it follows that Jη
ˆ ( ◦ 1 )) = div(cJη)
Since div(Jτ ˆ = 0, using Remark 5.13, it follows that π(Σ1 ×
Σ2 ) is a biharmonic submanifold of CP n if and only if
τ2 ( ◦ 1 ) + 4τ ( ◦ 1 ) = 0.
Finally, from Lemma 5.17, we get
a b b2 a2
τ2 ( ◦ 1 ) + 4τ ( ◦ 1 ) = m2 − m1 4 + m1 + m2 − 2 m1 − 2 m2 η,
b a a b
that ends the proof.
Remark 5.19. If Σ1 = S2p+1
a and Σ2 = Sb2q+1 , we recover a result in [88] on
proper-biharmonic homogeneous real hypersurfaces of type A in CP n .
Example 5.20. Let e1 and e3 be two constant unit vectors in E2n+2 , with e3
orthogonal to e1 and Jeˆ 1 . We consider two circles S1a and S1 lying in the 2-planes
b
ˆ 1 } and {e3 , Je
spanned by {e1 , Je ˆ 3 }, respectively. Then Σ = S1a × S1 is invariant
b
under the
p flow-action of ξ and π(Σ) is a proper-biharmonic curve of CP n if and only
√
if a = ( 2 ± 2)/2.
Example 5.21. For p = 0 and q = n − 1, we get√that π(S1a × S2n−1
b ) is proper-
n 2 2
biharmonic in CP if and only if a = (n + 3 ± n + 2n + 5)/(4(n + 1)). In
particular, π(S1 × S3 ) is a proper-biharmonic real hypersurface in CP 2 if and only
√ a b
if a2 = (5 ± 13)/12.
p,p
Example 5.22. If p = q, then Σ = Ta,b is never a proper-biharmonic hypersur-
face of S2n+1 and it is easy to check that
p π(Σ) is a proper-biharmonic hypersurface
n 2
of CP if and only if a = (2p + 2 − 2(p + 1))/(4(p + 1)).
Example 5.23. Consider Σ = Sa2p+1 × Spb/√2 × Spb/√2 , with p odd. Then Σ is

p,p √
minimal in Ta,b and proper-biharmonic in S2n+1 if and only if a = b = 1/ 2. By a
straightforward computation,√we can check that π(Σ) is proper-biharmonic in CP n
if and only if a2 = (8p + 7 ± 32p + 25)/(16p + 12).
We will end this section recovering a result of W. Zhang [139]. We denote by T
the n + 1-dimensional Clifford torus
 : T = S1a1 × · · · × S1an+1 → S2n+1 ,
where a21 +· · ·+a2n+1 = 1. The projection T̄ = π(T ) is a Lagrangian pmc submanifold
in CP n .
Theorem 5.24 ([139]). The Lagrangian submanifold T̄ = π(T ) of CP n is
proper-biharmonic if and only if T is (−4)-biharmonic, that is
(
1
a2k0 6= n+1 , for some k0 ∈ {1, . . . , n + 1}
(5.11)
ak d − a3 = a2k (n + 3)((n + 1)a2k − 1), k ∈ {1, . . . , n + 1},
1
k
Pn+1 2
where d = j=1 (1/aj ).
Proof. We denote a point x ∈ T by

x = (x1 , . . . , xn+1 ) = (x11 , x21 , . . . , x1n+1 , x2n+1 ),
where we identify
xk = (x1k , x2k ) = (0, 0, . . . , 0, 0, x1k , x2k , 0, 0, . . . , 0, 0), k ∈ {1, . . . , n + 1}.
We define ηk (x) = (1/ak )xk and Xk = Jη ˆ k , k ∈ {1, . . . , n + 1}, where
ˆ 11 , x21 , . . . , x1n+1 , x2n+1 ) = (−x21 , x11 , . . . , −x2n+1 , x1n+1 ).
J(x
The vector fields {Xk } form an orthonormal frame field of C(T T ). It is easy to
check that, at a point x,
1
σ(Xk , Xk ) = − ηk + x and σ(Xk , Xj ) = 0, if k 6= j.
ak
Pn+1 ˆ ())> = Jτ
ˆ ()
Therefore τ () = k=1 ((n+1)ak −1/ak )ηk , which implies that (Jτ
ˆ
and div(Jτ ()) = 0.
Since ∇⊥ τ () = 0 and Aτ () (Xk ) = −((n + 1) − 1/a2k )Xk , after a straightforward
computation, we get τ2 () + 4τ () = 0 if and only if (5.11) holds.
√ T̄ is a proper-
Remark 5.25. Following [139], when n = 2, one obtains that
2 2
biharmonic√Lagrangian surface in CP if and only if a1 = (9 ± 41)/20 and a22 =
a23 = (11 ∓ 41)/40 (see also [126]).
5. Biharmonic curves in CP n
Let γ̄ : I ⊂ R → CP n be a Frenet curve of osculating order r. In this section, we
¯ the Levi-Civita connection
will denote by {k̄1 , . . . , k̄r−1 } the curvatures of γ̄ and by ∇
on CP .n
We recall that the complex torsions of such a curve γ̄ are τ̄ij = hĒi , J¯Ēj i,
1 ≤ i < j ≤ r, and that a helix of order r is called a holomorphic helix of order r if
its complex torsions are constant (see [103]).
Using the Frenet equations, the bitension field of γ̄ becomes
(5.12) τ2 (γ̄) = − 3k̄1 k̄10 Ē1 + (k̄100 − k̄13 − k̄1 k̄22 + k̄1 )Ē2 + (2k̄10 k̄2 + k̄1 k̄20 )Ē3
+ k̄1 k̄2 k̄3 Ē4 − 3k̄1 τ̄12 J¯Ē1 .
In order to solve the biharmonic equation τ2 (γ̄) = 0, we will split our study in three
cases, as suggested by the form of the last term in (5.12).
5.1. Biharmonic curves with τ̄12 = ±1. In this case, we have J¯Ē2 = ±E1
and, using the Frenet equations of γ̄, one obtains
¯∇
J( ¯ Ē1 ) = ±k̄1 Ē1 = ∇
¯ (∓Ē2 ) = ∓∇
¯ Ē2
Ē1 Ē1 Ē1
and then
¯ Ē2 = −k̄1 Ē1 .
∇ Ē1
Consequently, k̄i = 0, i ≥ 2, and, from (5.12), our first result follows.
Theorem 5.26 ([64]). A Frenet curve γ̄ : I ⊂ R → CP n with τ̄12 = ±1 is
proper-biharmonic if and only if it is a circle with k̄1 = 2.
Next, let us consider a Frenet curve γ̄ : I ⊂ R → CP n with τ̄12 = ±1 and
denote by γ : I ⊂ R → S2n+1 one of its horizontal lifts. We will characterize the
biharmonicity of γ̄ in terms of γ.
First, we have γ 0 = E1 = (Ē1 )H and then
∇˙ E E1 = (∇¯ Ē1 )H = k̄1 Ē H = k1 E2 ,
1 Ē1 2
where ∇ ˙ is the Levi-Civita connection on S2n+1 , which means that k1 = k̄1 and
E2 = Ē2H = ∓(J¯Ē1 )H = ∓JE
ˆ 1 . It follows that
∇ ¯ Ē2 )H + h∇
˙ E E2 =(∇ ˙ E ξiξ = −k1 E1 ∓ hE2 , E2 iξ
˙ E E2 , ξiξ = −k1 E1 − hE2 , ∇
1 Ē1 1 1
= − k1 E1 ∓ ξ,
that gives k2 = 1 and E3 = ∓ξ. Then we have ∇ ˙ E E3 = ∓∇ ˙ E ξ = −E2 . Therefore
1 1
γ is a helix with k1 = k̄1 and k2 = 1.
Now, we have Jτ ˆ 2 = ±k1 E1 , which is tangent to γ, and then
ˆ (γ) = k1 JE
ˆ Jτ
J{( ˆ (γ))> } = Jˆ2 τ (γ) = −τ (γ).
Since we have
¯ (γ̄))> } = div{k̄1 hJ¯Ē2 , Ē1 iĒ1 } = h∇
div{(Jτ ¯ (k̄1 hJ¯Ē2 , Ē1 i)Ē1 , Ē1 i
Ē1
=k̄10 hJ¯Ē2 , Ē1 i + k̄1 hJ¯∇
¯ Ē2 , Ē1 i
Ē1
=± k̄10 = 0,
applying Remark 5.13, one obtains the following proposition.
Proposition 5.27 ([64]). A Frenet curve γ̄ : I ⊂ R → CP n with τ̄12 = ±1
is proper-biharmonic if and only if its horizontal lift γ : I ⊂ R → S2n+1 is (−4)-
biharmonic, i.e., γ is a helix with k1 = 2 and k2 = 1.
Moreover, we can determine the explicit parametric equations of the horizontal
lifts of a proper-biharmonic Frenet curve γ̄ : I → CP n , working as in Theorem 4.14
in the case of biharmonic curves.
Proposition 5.28 ([64]). Let γ̄ : I ⊂ R → CP n be a proper-biharmonic Frenet
curve with τ̄12 = ±1. Then its horizontal lift γ : I ⊂ R → S2n+1 can be parametrized
in the Euclidean space E2n+2 = (R2n+2 , h, i) by
p √ p √
2− 2 √ 2− 2 √
γ(s) = cos(( 2 + 1)s)e1 − ˆ1
sin(( 2 + 1)s)Je
p2 √ p2
√
2+ 2 √ 2+ 2 √
+ cos(( 2 − 1)s)e3 + ˆ 3,
sin(( 2 − 1)s)Je
2 2
ˆ 1.
where e1 and e3 are constant unit vectors in E2n+2 with e3 orthogonal to e1 and Je
Remark 5.29. Under the flow-action of ξ, the (−4)-biharmonic curves γ induce
the (−4)-biharmonic surfaces obtained in Example 5.20.
5.2. Biharmonic curves with τ̄12 = 0. From the expression (5.12) of the
bitension field of γ̄, we get that γ̄ is proper-biharmonic if and only if
k̄1 = constant > 0, k̄2 = constant, k̄12 + k̄22 = 1, and k̄2 k̄3 = 0.
Theorem 5.30 ([64]). A Frenet curve γ̄ : I ⊂ R → CP n with τ̄12 = 0 is
(1) n = 2 and γ̄ is a circle with k̄1 = 1; or
(2) n ≥ 3 and γ̄ is a circle with k̄1 = 1 or a helix with k̄12 + k̄22 = 1.
Proof. We only have to prove the statements concerning the values of n.
First, since {Ē1 , Ē2 , J¯Ē2 } are linearly independent, it follows that n > 1.
Now, assume that γ̄ is a Frenet curve of osculating order 3 such that J¯Ē2 ⊥ Ē1 .
We have
Ē1 = γ̄ 0 , ∇ Ē1
¯ Ē2 = −k̄1 Ē1 + k̄2 Ē3 , and ∇
¯ Ē1 = k̄1 Ē2 , ∇
Ē1
¯ Ē3 = −k̄2 Ē2 .
Ē1
It is easy to see that, at any point, S1 = {Ē1 , Ē2 , Ē3 , J¯Ē1 , J¯Ē2 } consists of non-zero
vectors orthogonal to one another and, therefore, n ≥ 3.
Next, consider the horizontal lift γ : I ⊂ R → S2n+1 of a Frenet curve γ̄ : I ⊂
R → CP n arc-length with τ̄12 = 0. As in the previous case, we have γ 0 = E1 =
Ē1H , E2 = Ē2H and then JEˆ 2 ⊥ E1 . This means J(τ ˆ (γ)) ⊥ E1 , which implies
ˆ (γ)))> = 0. We then use Theorem 5.12 to obtain the following proposition.
(J(τ
Proposition 5.31 ([64]). A Frenet curve γ̄ : I ⊂ R → CP n with τ̄12 = 0 is
proper-biharmonic if and only if its horizontal lift γ : I ⊂ R → S2n+1 is proper-
biharmonic.
Remark 5.32. We recall that the parametric equations of the proper-biharmonic
ˆ 2 ⊥ E1 are given by Theorem 4.17.
Frenet curves in S2n+1 with JE
5.3. Biharmonic curves with τ̄12 different from 0, 1 or −1. Consider a
proper-biharmonic Frenet curve γ̄ of osculating order r such that τ̄12 is different
from 0, 1, or −1. We will first prove that r ≥ 4.
Let us assume that r = 2. From the biharmonic equation τ2 (γ̄) = 0, we have
k̄1 = constant > 0 and then (−k̄13 + k̄1 )Ē2 − 3k̄1 τ̄12 J¯Ē1 = 0. It follows that Ē2 is
parallel to J¯Ē1 , i.e., τ̄12
2 = 1.
Next, if d = 3, from the biharmonic equation of γ̄, one obtains k̄1 = constant > 0
and then
(5.13) (−k̄12 − k̄22 + 1)Ē2 + k̄20 Ē3 − 3τ̄12 J¯Ē1 = 0.
Differentiating −τ̄12 (s) = hĒ2 , J¯Ē1 i, we obtain
0 ¯ Ē2 , J¯Ē1 i + hĒ2 , ∇
¯ J¯Ē1 i = h∇ ¯ Ē2 , J¯Ē1 ) + hĒ2 , k̄1 J¯Ē2 )
−τ̄12 (s) =h∇ Ē1 Ē1 Ē1
¯ ¯ ¯
=h∇ Ē2 , J Ē1 i = h−k̄1 Ē1 + k̄2 Ē3 , J Ē1 i
Ē1
=k̄2 hĒ3 , J¯Ē1 i.
Therefore, taking the inner product with k̄2 Ē3 in (5.13), we get k̄20 k̄2 +3τ̄12 τ̄12
0 =0
2 2
and then k̄2 = −3τ̄12 + ω0 , where ω0 = constant. Using (5.13), it follows that
k̄12 = 1 − ω0 + 6τ̄12
2 . Hence, we get that f = constant and k̄ = constant. Finally,
2
(5.13) becomes (−k̄12 − k̄22 + 1)Ē2 − 3τ̄12 J¯Ē1 = 0, which means that Ē2 is parallel to
J¯Ē1 .
Proposition 5.33 ([64]). Let γ̄ be a proper-biharmonic Frenet curve in CP n of
osculating order r, 1 ≤ r ≤ 2n, with τ̄12 different from 0, 1, or −1. Then we have
r ≥ 4.
Next, we will show that, for a proper-biharmonic Frenet curve in CP n , τ̄12 and
k̄1 are constant functions.
First, as we have already seen, we have −τ̄12 0 (s) = k̄ hĒ , J¯Ē i. If τ (γ̄) = 0,
2 3 1 2
then one obtains J Ē1 = hJ Ē1 , Ē2 iĒ2 + hJ Ē1 , Ē3 iĒ3 + hJ¯Ē1 , Ē4 iĒ4 and
¯ ¯ ¯
(
k̄1 = constant > 0, k̄12 + k̄22 = 1 + 3τ̄12 2
(5.14)
k̄2 k̄20 = −3τ̄12 τ̄12
0 , k̄2 k̄3 = 3τ̄12 hJ¯Ē1 , Ē4 i.
From the third equation of (5.14), we get

k̄22 = −3τ̄12
2
+ ω0 ,
where ω0 = constant. Replacing in the second equation, it follows that
k̄12 = 1 + 6τ̄12 − ω0 ,
which implies τ̄12 = constant and, therefore, k̄2 = constant > 0. From −τ̄12 0 (s) =
k̄2 hĒ3 , J¯Ē1 i, we have hJ¯Ē1 , Ē3 i = 0 and then J¯Ē1 = f Ē2 + hJ¯Ē1 , Ē4 iĒ4 . It follows
that there exists a unique constant α0 ∈ (0, 2π) \ {π/2, π, 3π/2} such that −τ̄12 =
cos α0 and hJ¯Ē1 , Ē4 i = sin α0 = k̄2 k̄3 /3τ̄12 .
Theorem 5.34 ([64]). A Frenet curve γ̄ : I ⊂ R → CP n , n ≥ 2, with τ̄12
different from 0, 1, or −1, is proper-biharmonic if and only if J¯Ē1 = cos α0 Ē2 +
sin α0 Ē4 and
(
k̄1 , k̄2 , k̄3 = constant > 0, k̄12 + k̄22 = 1 + 3 cos2 α0
k̄2 k̄3 = − 23 sin(2α0 ), τ̄12 = − cos α0 ,
where α0 ∈ (π/2, π) ∪ (3π/2, 2π) is a constant.
We now consider proper-biharmonic curves in CP n of osculating order r ≤ 4. In
order to classify these curves, we will first prove the following proposition.
osculating order r < 4. Then γ̄ is either a holomorphic circle of curvature k̄1 = 2,
or a holomorphic circle of curvature k̄1 = 1, or a holomorphic helix with k̄12 + k̄22 = 1.
Proof. Let γ̄ be a proper-biharmonic Frenet curve of osculating order r < 4.
Then, from Proposition 5.33, τ̄12 = ±1 or τ̄12 = 0. If τ̄12 = ±1, from Proposi-
tion 5.26, we have that γ̄ is a circle of curvature k̄1 = 2. If τ̄12 = 0, then we know
that γ̄ is either a holomorphic circle of curvature k̄1 = 1 or a helix. It only remains
to be shown that, in the later case, our curve is a holomorphic helix. We will prove
that complex torsions τ̄13 and τ̄23 are constant. First, we have
1 ¯ 1 ¯ J¯Ē1 i = k̄1 hĒ2 , J¯Ē2 i
τ̄13 =hĒ1 , J¯Ē3 i = − h∇ ¯
Ē1 Ē2 , J Ē1 i = hĒ2 , ∇ Ē1
k̄2 k̄2 k̄2
=0.
0 + k̄ τ̄ , one
Now, since for a Frenet curve of osculating order 3 we have k̄1 τ̄23 = τ̄13 2 12
can see that also τ̄23 is constant.
When the osculating order of a biharmonic curve is 4, (5.34) has four solutions,
as shown by the following result.
osculating order r = 4. Then γ̄ is a holomorphic helix and, moreover, depending on
the value of τ̄12 = − cos α0 , we have
(1) If τ̄12 > 0, then the curvatures of γ̄ are given by
sin
p √
√α0


 k̄2 = 2
1 − 3 cos2 α0 ± 9 cos4 α0 − 42 cos2 α0 + 1
(5.15) k̄3 = − 2k̄3 sin(2α0 )
 2
k̄1 = − sin1α0 (k̄2 cos α0 − k̄3 sin α0 )

and
τ̄34 = −τ̄12 = cos α0 , τ̄14 = −τ̄23 = − sin α0 , τ̄13 = τ̄24 = 0,
√ √
where α0 ∈ (π/2, arccos(−(2 − 3)/ 2)).
(2) If τ̄12 < 0, then the curvatures of γ̄ are given by
 sin α0
p √
k̄
 2
 = − √
2
1 − 3 cos2 α0 ± 9 cos4 α0 − 42 cos2 α0 + 1
(5.16) k̄3 = − 2k̄3 sin(2α0 )
 2
k̄1 = − sin1α0 (k̄2 cos α0 − k̄3 sin α0 ),

and
τ̄34 = −τ̄12 = cos α0 ,
τ̄14 = −τ̄23 = − sin α0 , τ̄13 = τ̄24 = 0,
√ √
where α0 ∈ (3π/2, π + arccos(−(2 − 3)/ 2)).
Proof. Let γ̄ be a proper-biharmonic Frenet curve in CP n of osculating order
r = 4. Then τ̄12 = − cos α0 is different from 0, 1, or −1 and J¯Ē1 = cos α0 Ē2 +
sin α0 Ē4 . It follows that
τ̄12 = − cos α0 , τ̄13 = 0, τ̄14 = − sin α0 ,
and τ̄24 = 0.
To prove that τ̄23 is constant, we differentiate the expression of J¯Ē1 and, using
the Frenet equations, one obtains
¯ J¯Ē1 = cos α0 ∇
∇ ¯ Ē2 + sin α0 ∇ ¯ Ē4
Ē1 Ē1 Ē1
= − k̄1 cos α0 Ē1 + (k̄2 cos α0 − k̄3 sin α0 )Ē3 .
¯ J¯Ē1 = k̄1 J¯Ē2 and then
On the other hand, we have ∇ Ē1
(5.17) k̄1 J¯Ē2 = −k̄1 cos α0 Ē1 + (k̄2 cos α0 − k̄3 sin α0 )Ē3 .
Taking the inner product of (5.17) with Ē3 , J¯Ē2 , and J¯Ē4 , we get
(5.18) k̄1 τ̄23 = −(k̄2 cos α0 − k̄3 sin α0 ),
(5.19) k̄1 sin2 α0 = −(k̄2 cos α0 − k̄3 sin α0 )τ̄23 ,

and
(5.20) 0 = k̄1 cos α0 sin α0 + (k̄2 cos α0 − k̄3 sin α0 )τ̄34 ,
respectively.
From (5.18) and (5.19), one obtains
(5.21) k̄12 sin2 α0 = (k̄2 cos α0 − k̄3 sin α0 )2
2 = sin2 α . Since τ 2 = sin2 α and α ∈ (π/2, π) ∪ (3π/2, 2π), using (5.18),
and τ23 0 23 0 0
we have
τ̄23 = sin α0 .
From τ̄23 = sin α0 , (5.18), and (5.20) we get
τ̄34 = cos α0 .
Finally, from Theorem 5.34 and (5.21), it follows that
k̄24 + k̄22 sin2 α0 (3 cos2 α0 − 1) + 9 sin4 α0 cos2 α0 = 0.
The solution of this equation is either

q
sin α0 p
k̄2 = √ 1 − 3 cos2 α0 ± 9 cos4 α0 − 42 cos2 α0 + 1,
2
√ √
provided that α0 ∈ (π/2, arccos(−(2 − 3)/ 2)), or
q
sin α0 p
k̄2 = − √ 1 − 3 cos2 α0 ± 9 cos4 α0 − 42 cos2 α0 + 1,
2
√ √
if α0 ∈ (3π/2, π + arccos(−(2 − 3)/ 2)). We note that, in both cases, k̄22 ∈ (0, 4)
and, therefore, both solutions are compatible with k̄12 + k̄22 = 1 + 3 cos2 α0 .
Corollary 5.37 ([64]). Any proper-biharmonic Frenet curve in CP 2 is a holo-
morphic circle or a holomorphic helix of order 4.
Remark 5.38. The existence of biharmonic curves of osculating order r ≥ 4 is
an open problem. We note that there is no curve (not necessarily biharmonic) of
order r = 5 in CP n such that J¯Ē1 = cos α0 Ē2 + sin α0 Ē4 , where α0 ∈ (0, 2π) \ {π}.
We will end this section with the complete classification of proper-biharmonic
Frenet curves in CP 2 . To do that, from the above results, we only have to classify
proper-biharmonic Frenet curves of osculating order 4.
In the proof of Proposition 5.36 we have seen that
τ̄34 = −τ̄12 = cos α0 , τ̄14 = −τ̄23 = − sin α0 , τ̄13 = τ̄24 = 0,
and
k̄1 sin α0 = −(k̄2 cos α0 − k̄3 sin α0 ),
√ √implies that k̄1 − k̄3 = −k̄2 cot α0 . Moreover, if α0 ∈ (π/2, arccos(−(2 −
which
3)/ 2)), then we have
k̄1 − k̄3 k̄2
q = − cos α0 = τ̄12 , q = sin α0 = τ̄23
k̄22 + (k̄1 − k̄3 )2 2
k̄2 + (k̄1 − k̄3 )2
√ √
and, if α0 ∈ (3π/2, π + arccos(−(2 − 3)/ 2)), then
k̄1 − k̄3 k̄2
q = cos α0 = −τ̄12 , q = − sin α0 = −τ̄23 .
k̄22 + (k̄1 − k̄3 )2 k̄22 + (k̄1 − k̄3 )2
In order to conclude, we briefly recall a result of S. Maeda and T. Adachi (see
[102]). They showed that, for any given positive constants k̄1 , k̄2 , and k̄3 , there exist
four equivalence classes of holomorphic helices of order 4 in CP 2 with curvatures
k̄1 , k̄2 , and k̄3 with respect to holomorphic isometries of CP 2 . These four classes are
defined by certain relations on the complex torsions as follows: when k̄1 6= k̄3
k̄1 6= k̄3
I1 τ̄12 = τ̄34 = µ τ̄23 = τ̄14 = k̄2 µ/(k̄1 + k̄3 ) τ̄13 = τ̄24 =0
I2 τ̄12 = τ̄34 = −µ τ̄23 = τ̄14 = −k̄2 µ/(k̄1 + k̄3 ) τ̄13 = τ̄24 =0
I3 τ̄12 = −τ̄34 = ν τ̄23 = −τ̄14 = k̄2 ν/(k̄1 − k̄3 ) τ̄13 = τ̄24 =0
I4 τ̄12 = −τ̄34 = −ν τ̄23 = −τ̄14 = −k̄2 ν/(k̄1 − k̄3 ) τ̄13 = τ̄24 =0
where
k̄1 + k̄3 k̄1 − k̄3
µ= q and ν = q ,
k̄22 + (k̄1 + k̄3 )2 k̄22 + (k̄1 − k̄3 )2
and when k̄1 = k̄3 the classes I3 and I4 are substituted by

k̄1 = k̄3
I30 τ̄12 = τ̄34 = τ̄13 = τ̄24 = 0 τ̄23 = −τ̄14 = 1
I40 τ̄12 = τ̄34 = τ̄13 = τ̄24 = 0 τ̄23 = −τ̄14 = −1
Using Maeda-Adachi classification, we have the following theorem.
Theorem 5.39 ([64]). Let γ̄ be a proper-biharmonic Frenet curve in CP 2 of
osculating order 4. Then γ̄ is a holomorphic helix of order 4 of class I3 , if τ̄12 < 0
and τ̄23 < 0, or I4 , if τ̄12 > 0 and τ̄23 > 0. Conversely,
√ √
(1) For any α0 ∈ (π/2, arccos(−(2 − 3)/ 2)), there exist two proper-bihar-
monic holomorphic helices of order 4 of class I3 with
sin
p √
√α0


 k̄ 2 = 2
1 − 3 cos2 α0 ± 9 cos4 α0 − 42 cos2 α0 + 1
k̄3 = − 2k̄3 sin(2α0 )
 2
k̄1 = − sin1α0 (k̄2 cos α0 − k̄3 sin α0 ).

√ √
(2) For any α0 ∈ (3π/2, π + arccos(−(2 − 3)/ 2)), there exist two proper-
biharmonic holomorphic helices of order 4 of class I4 with
sin
p √
√α0


k̄ 2 = − 2
1 − 3 cos2 α0 ± 9 cos4 α0 − 42 cos2 α0 + 1
k̄3 = − 2k̄3 sin(2α0 )
 2
k̄1 = − sin1α0 (k̄2 cos α0 − k̄3 sin α0 ).

6. Biharmonic surfaces with parallel mean curvature in complex space

forms
6.1. A Simons type formula for pmc surfaces in complex space forms.
Let (N n (c), J, h, i) be a complex space form, with constant holomorphic sectional
curvature c and complex dimension n, and Σ̄2 a pmc surface in N n (c).
We recall that the (2, 0)-part Q(2,0) of the quadratic form Q defined on Σ̄2 by
Q(X, Y ) = 8|H|2 hAH X, Y i + 3chX, T ihY, T i,
where T is the tangent part (JH)> of JH, is holomorphic (Theorem 1.4).
Using this holomorphic differential, we can define on Σ̄2 an operator S by
3c
(5.22) S = 8|H|2 AH + 3chT, ·iT − |T |2 + 8|H|4 I,
2
that satisfies
trace Q
(5.23) hSX, Y i = Q(X, Y ) − hX, Y i,
2
which implies that S is symmetric and traceless. It is also easy to see that Q(2,0)
vanishes on Σ̄2 if and only if S = 0 on the surface. Working in the same way as in
Theorems 2.23 and 2.25, we can first prove that
1
(5.24) ∆|S|2 = 2K|S|2 + |∇S|2
2
and then the following theorem.
Theorem 5.40 ([73]). Let Σ̄2 be a complete non-minimal pmc surface with non-
negative Gaussian curvature K isometrically immersed in a complex space form
N n (c), c 6= 0. Then one of the following holds:
(1) the surface is flat;
(2) there exists a point p ∈ Σ̄2 such that K(p) > 0 and Q(2,0) vanishes on Σ̄2 .
Remark 5.41 ([73]). For a surface Σ̄2 as in Theorem 5.40 we have |S| = constant
and ∇S = 0.
6.2. Biharmonic surfaces with parallel mean curvature in CP n (c). From
Theorem 3.1, also using formula (1.6) of the curvature tensor of N n (c), we get the
following result.
Theorem 5.42 ([73]). Let Σ̄2 be a pmc surface in a complex space form N n (c).
Then Σ̄2 is biharmonic if and only if
c
(5.25) trace σ(·, AH ·) = {2H − 3(JT )⊥ } and (JT )> = 0,
4
where σ is the second fundamental form of Σ̄ in N n (c), A is the shape operator, T
is the tangent part of JH, and (JT )⊥ and (JT )> are the normal and the tangent
components of JT , respectively.
Remark 5.43 ([73]). It is easy to see, from the first equation of (5.25), that,
for a proper-biharmonic pmc surface, we have
c
0 < |AH |2 = {2|H|2 + 3|T |2 }
4
which implies that c > 0, and, therefore, such surfaces exist only in CP n (c).
Proposition 5.44 ([73]). If Σ̄2 is a pmc proper-biharmonic surface in CP n (c),
then T has constant length.
Proof. As we have already seen, either Σ̄2 is a pseudo-umbilical surface, or H
is an umbilical direction on a closed set without interior points. As usual, we shall
denote by W the set of points where H is not an umbilical direction. Since in the
second case this set is open and dense in Σ̄2 , when the surface is not pseudo-umbilical
we shall work on W and then extend our results throughout Σ̄2 by continuity.
If Σ̄2 is pseudo-umbilical, then JH is normal to the surface, i.e., T = 0 on the
surface (see [127]).
Let us now assume that Σ̄2 is not pseudo-umbilical and let N be the normal part
of JH. Then, for any vector field X tangent to the surface, we have
∇¯ X JH = −J ∇¯ X H = −JAH X
= ∇X T + σ(X, T ) − AN X + ∇⊥
XN
¯ is the Levi-Civita connection on N n (c), and, therefore,
where ∇
(5.26) h∇X T, T i = hAN X, T i + hAH X, JT i = hAN X, T i,
since, from the second equation of (5.25), we know that JT is normal.
It easy to see that
hN, Hi = 0 and hN, JT i = 0
and, again using (5.25), that
hN, JXi = 0, ∀X ∈ C(T Σ̄2 ).
Then, from the first equation of (5.25), we get that

(5.27) trace(AH AN ) = 0.
Moreover, using the Ricci equation (1.5), we obtain
(5.28) [AH , AN ]T = 0,
since R⊥ (X, Y )H = 0 and hR̄(X, T )H, N i = 0, for tangent vector fields X and Y ,
where R̄ is the curvature tensor field of N n (c).
Next, consider a point p ∈ W and an orthonormal basis {e1 , e2 } in Tp Σ̄2 such
that AH ei = λi ei , i ∈ {1, 2}. Obviously, we have λ1 6= λ2 and we can write AH and
AN , with respect to {e1 , e2 }, as

λ1 0 a b
AH = and AN = ,
0 λ2 b −a
since N ⊥ H, i.e., trace AN = 0. From (5.27), we get a = 0 and then (5.28) becomes
(λ2 − λ1 )b(hT, e2 ie1 − hT, e1 ie2 ) = 0.
Therefore, at p, we have that either T = 0 or b = 0. We can see that, in both cases,
hAN X, T i = 0, which implies that equation (5.26) reduces to
X(|T |2 ) = 2h∇X T, T i = 0,
for any tangent vector field X. It follows that X(|T |2 ) = 0, which means that |T | is
constant on the surface.
Remark 5.45. If |T | = constant 6= 0, we have ∇X T = AN X = 0 for any tangent
vector field X. Indeed, if T 6= 0 everywhere, since JT is a normal vector field, it
follows that Σ̄2 is a totally real surface. Then we get
∇¯ X JH = −J ∇ ¯ X H = −JAH X ∈ C(N Σ̄2 ),
which means that ∇X T = AN X. On the other hand, we have hR̄(X, Y )H, N i = 0,
for any tangent vector fields X and Y , and then, from the Ricci equation (1.5), one
sees that [AH , AN ] = 0. Using this equation and (5.27) in the same way as in the
proof of Proposition 5.44 and since T 6= 0 implies that H is not umbilical on an
open dense set, we obtain AN = 0 on this set and, therefore, on the whole surface.
Proposition 5.46 ([73]). If Σ̄2 is a complete proper-biharmonic pmc surface
in CP n (c) with non-negative Gaussian curvature K and T = 0, then n ≥ 3 and
Σ̄2 is √
pseudo-umbilical and totally real. Moreover, the mean curvature of Σ̄2 is
|H| = c/2.
Proof. From Theorem 5.42, we see that a pmc surface Σ̄2 with T = 0 is proper-
c
(5.29) trace σ(·, AH ·) = H.
2
Now, from Theorem 5.40, we know that either the Gaussian curvature K vanishes
identically on the surface, or there exists a point p ∈ Σ̄2 such that K(p) > 0 and
Q(2,0) = 0 on Σ̄2 .
In the second case, since T = 0 and Q(2,0) = 0, it is easy to see that Σ̄2 is
pseudo-umbilical and then totally real (see [127]). From (5.29), we get that |AH |2 =
(c/2)|H|2 , but since 2 2 4
√ Σ̄ is pseudo-umbilical, we also have |AH | = 2|H| , which
means that |H| = c/2.
If the surface is flat, we shall first prove that it is also totally real. Since JH is
a normal vector field to Σ̄2 , we have
¯ X JH = J ∇
∇ ¯ X H = −JAH X
= −AJH X + ∇⊥
X JH.
Let us now consider an orthonormal basis {e1 , e2 } in Tp Σ̄2 , where p ∈ Σ̄2 , such
that AH ei = λi ei , i ∈ {1, 2}. It follows that JAH ei = λi Jei and, for i 6= j, we have
hAJH ei , ej i = hJAH ei , ej i = λi hJei , ej i.
Thus, we obtained λ1 hJe1 , e2 i = λ2 hJe2 , e1 i, which means that
0 = (λ1 + λ2 )hJe1 , e2 i = 2|H|2 hJe1 , e2 i.
Therefore, we have hJe1 , e2 i = 0, i.e., Σ̄2 is totally real.
In the following, we will prove that Σ̄2 is also pseudo-umbilical. Assume that it
is not so and we will work on the set W defined in the proof of Proposition 5.44.
Let p be a point in W , consider a basis {e1 , e2 } in Tp Σ̄2 such that AH ei = λi ei , and
extend the ei to vector fields Ei in a neighborhood of p. First, using the expression
(1.6) of the curvature tensor of CP n (c), we obtain, hR̄(E2 , E1 )H, JE1 i = 0 and then,
from the Ricci equation (1.5), h[AH , AJE1 ]E1 , E2 i = 0, which can be written at p as
(λ2 − λ1 )hAJE1 E1 , E2 i = 0.
In the same way, we can also show that (λ1 − λ2 )hAJE2 E2 , E1 i = 0.
First, since λ1 6= λ2 , we get that hAJE1 E1 , E2 i = hAJE2 E2 , E1 i = 0. Using the
fact that Σ̄2 is totally real, it is easy to verify that
hσ(X, Y ), JZi = hσ(X, Z), JY i, ∀X, Y, Z ∈ C(T Σ̄2 ),
and then, at p, we obtain
hAJE2 E1 , E1 i = hσ(E1 , E1 ), JE2 i = hσ(E1 , E2 ), JE1 i = hAJE1 E1 , E2 i = 0
and
hAJE1 E2 , E2 i = hAJE2 E2 , E1 i = 0.
Since JH is normal to Σ̄2
is equivalent to trace AJE1 = trace AJE2 = 0, we have just
proved that AJE1 = AJE2 = 0 at p.
Next, for any normal vector field U , which is also orthogonal to H, JE1 , and JE2 ,
we have hR̄(X, Y )H, U i = 0 and then, from the Ricci equation (1.5), [AH , AU ] = 0.
Since H is not umbilical on W , this implies that, with respect to {E1 , E2 }, we have,
at p,
a + |H|2

0 b 0
AH = and AU = ,
0 −a + |H|2 0 −b
with a 6= 0. From (5.29), we have that trace(AH AU ) = 0, which implies, using the
above expressions, that AU = 0.
Now, we consider a local orthonormal frame field in the normal bundle of Σ̄2 ,
as follows {E3 = H/|H|, E4 = JE1 , E5 = JE2 , E6 , . . . , E2n } and, since the surface
is flat, from the Gauss equation (1.3) of Σ̄2 in CP n (c), at p, we get
2n
c X c c a2
0=K= + det Aα = + det A3 = + |H|2 − .
4 4 4 |H|2
α=3
From (5.29), we have |AH |2 = 2a2 + 2|H|4 = (c/2)|H|2 and, therefore, K = 2|H|2 at
p, which means that |H| = 0. This is a contradiction, since Σ̄2 is proper-biharmonic.
Hence, the surface is pseudo-umbilical in this case too.
Finally, we have that, for any vector field X tangent to the surface, the vector
field JX is normal and orthogonal to both H and JH, which are also normal vector
fields. Therefore, one obtains n ≥ 3 and we conclude.
Proposition 5.47 ([73]). If Σ̄2 is a complete pmc proper-biharmonic surface in
CP n (c) with non-negative Gaussian curvature K and T 6= 0, then the surface is flat
and ∇AH = 0.
Proof. Since |T | = constant 6= 0 on Σ̄2 , from the second equation of (5.25), we
know that our surface is totally real. In [36], it is proved that the (2, 0)-part Q(2,0)
of the quadratic form
Q(X, Y ) = hAH X, Y i,
defined on a pmc totally real surface, is holomorphic. Consider the traceless part
φH = AH − |H|2 I of AH . Since Q(2,0) is holomorphic, working in the same way
as in [19, Proposition 3.3], we can prove that φH satisfies the Codazzi equation
(∇X φH )Y = (∇Y φH )X. Hence, from equation (2.46), we have
1
∆|φH |2 = 2K|φH |2 + |∇φH |2 .
2
Let us now assume that there exists a point p ∈ Σ̄2 such that K(p) > 0. Then,
from Theorem 5.40, we have that S = 0, which implies
9c2
|φH |2 = |AH |2 − 2|H|4 = |T |4 = constant 6= 0,
128|H|4
which means that K = 0 on Σ̄2 and this is a contradiction.
Hence the surface is flat. Since Σ̄2 is proper-biharmonic, it follows, from the first
equation of (5.25), that |φH |2 is bounded. Thus, |φH |2 is a bounded subharmonic
function on a parabolic space and, therefore, constant, which implies ∇AH = ∇φH =
0.
Remark 5.48. In the proof of Proposition 5.47 we used the fact that Q(2,0) is
holomorphic when Q is defined on a totally real pmc surface in a complex space form.
We recall that, if Σ̄2 is a proper-biharmonic surface with constant mean curvature
in a Riemannian manifold, then Q(2,0) is holomorphic (see [101]).
Before proving our main result, let us briefly recall a property of the Hopf fibra-
tion (see [120]). Let π : Cn+1 \ {0} → CP n (c) be the natural projection and
S2n+1 (c/4) = {z ∈ Cn+1 : hz, zi = 4/c}. The restriction of π to the sphere
S2n+1 (c/4) ⊂ Cn+1 is the Hopf fibration π : S2n+1 (c/4) → CP n (c), that is a
Riemannian submersion. Now, let  : Σ̄m → CP n (c) be a totally real isometric
immersion. Then this immersion can be lifted locally (or globally, if Σ̄m is sim-
ply connected) to a horizontal immersion e  : Σe m → S2n+1 (c/4). Conversely, if
:Σ
e e m → S2n+1 (c/4) is a horizontal isometric immersion, then π(e ) : Σ̄m → CP n (c)
is a totally real isometric immersion. Moreover, we have π∗ σ e = σ, where σ e and σ
are the second fundamental forms of the immersions e  and , respectively.
We are now ready to prove the main result of this section.
Theorem 5.49 ([73]). Let Σ̄2 be a complete pmc proper-biharmonic surface with
non-negative Gaussian curvature in CP n (c). Then Σ̄2 is totally real and either
√
(1) Σ̄2 is pseudo-umbilical and its mean curvature is equal to c/2. Moreover,
Σ̄2 = π(Σ
e 2 ) ⊂ CP n (c), n ≥ 3,
where π : S2n+1 (c/4) → CP n (c) is the Hopf fibration, and the horizontal lift
Σ of Σ̄ is a complete minimal surface in a small hypersphere S2n (c/2) ⊂
e 2 2
S2n+1 (c/4); or
(2) Σ̄2 lies in CP 2 (c) as a complete Lagrangian pmc proper-biharmonic surface.
Moreover, if c = 4, then
s √ ! s √ ! s √ !!
9 ± 41 11 ∓ 41 11 ∓ 41
Σ̄2 = π S1 × S1 × S1 ⊂ CP 2 (4),
20 40 40
where π : S5 (1) → CP 2 (4) is the Hopf fibration; or

(3) Σ̄2 lies in CP 3 (c) and
Σ̄2 = γ̄1 × γ̄2 ⊂ CP 3 (c),
where γ̄1 : R → CP 2 (c) ⊂ CP 3 (c) is a holomorphic helix of order 4 with
curvatures
r r r
7c 1 5c 3 c
κ̄1 = , κ̄2 = , κ̄3 =
6 2 42 2 42
and complex torsions
√ √
11 14 70
τ̄12 = −τ̄34 = , τ̄23 = −τ̄14 = , τ̄13 = τ̄24 = 0
42 42
p
and γ̄2 : R → CP 3 (c) is a holomorphic circle with curvature κ̄ = c/2 and
complex torsion τ̄12 = 0. Moreover, the curves γ̄1 and γ̄2 always exist and
are unique up to holomorphic isometries.
Proof. Let Σ̄2 be a complete pmc proper-biharmonic surface with non-negative
Gaussian curvature K and mean curvature vector field H in CP n (c). Let T and
N be the tangent and the normal parts of JH, respectively. As we have seen in
Proposition 5.44, the length of T is constant along the surface. We shall consider
two cases, as T = 0 or T 6= 0 on Σ̄2 .
Case I: T = 0. From Proposition 5.46, we know that n ≥ 3 √ and the surface
is pseudo-umbilical and totally real with mean curvature |H| = c/2. Consider
the Hopf fibration π : S2n+1 (c/4) → CP n (c) and the horizontal lift Σ e 2 of Σ̄2 to
S2n+1 (c/4). Then, from [120, Theorem 1], we have that Σ e 2 is pseudo-umbilical
in S 2n+1 (c/4) and has parallel mean√curvature vector field. Moreover, its mean
curvature is constant and equal to c/2. Next, using the relation between the
bitension fields of the immersions  : Σ̄2 → CP n (c) and e
:Σe 2 → S2n+1 (c/4), given
>
in Remark 5.13, together with (JH) = T = 0, we get that Σ̄2 is proper-biharmonic
e 2 is proper-biharmonic. We apply Theorem 3.6 to conclude that Σ
if and only if Σ e2
2n
is a complete minimal surface in a small hypersphere S (c/2) ⊂ S 2n+1 (c/4).
Case II: T 6= 0. In this case, Σ̄2 is totally real and, by Proposition 5.47, flat.
From the same Proposition 5.47, we also know that ∇AH = 0, which means that
the eigenfunctions of AH are actually real constants. Since Σ̄2 is pseudo-umbilical

implies that T = 0, it follows that the surface does not have umbilical points.
Now, let U be a normal vector field orthogonal to H and to J(T Σ̄2 ). Then, it
is easy to see that hR̄(X, Y )H, U i = 0 and, from the Ricci equation (1.5), we get
that [AH , AU ] = 0. Since H is not umbilical, this implies that AH and AU can be
simultaneously diagonalized.
On the other hand, the first equation of (5.25) shows that trace(AH AU ) = 0
and, therefore, AU = 0.
Let us consider a global orthonormal frame field {E1 = T /|T |, E2 } on the surface.
We know, from Remark 5.45, that ∇E1 = 0 and then ∇E2 = 0.
Next, if |T | = |H|, i.e., if JH is a tangent vector field, we consider the subbundle
L = span{JE1 , JE2 } of the normal bundle. Since JH is tangent, we get that H ∈ L
and, therefore, for any normal vector field U ⊥ L, we have AU = 0, which means that
Im σ ⊂ L. It is also easy to see that dim(T Σ̄2 ⊕ L) = 4 and J(T Σ̄2 ⊕ L) = T Σ̄2 ⊕ L,
which implies that R̄(X, Y )Z ∈ L, for all vector fields X, Y, Z ∈ T Σ̄2 ⊕ L, i.e.,
T Σ̄2 ⊕ L is invariant by R̄. In the following, we shall prove that L is parallel, i.e.,
if U is a normal vector field orthogonal to L, then U is also orthogonal to ∇⊥ L.
Indeed, for all tangent vector fields X and Y , we obtain
h∇⊥ JY, U i = h∇
X
¯ X JY, U i = hJ ∇ ¯ X Y, U i = hJ∇X Y + Jσ(X, Y ), U i = 0,
¯ R̄ = 0, we can use [54,
since Im σ ⊂ L and J(T Σ̄2 ⊕L) = T Σ̄2 ⊕L. Therefore, since ∇
Theorem 2] to show that there exists a 4-dimensional totally geodesic submanifold
of CP n (c) such that Σ̄2 lies in this submanifold. Since J(T Σ̄2 ⊕ L) = T Σ̄2 ⊕ L,
we get that Σ̄2 is a complete Lagrangian pmc proper-biharmonic surface in CP 2 (c).
When c = 4, these surfaces were determined in [126], as follows
s √ ! s √ ! s √ !!
2 1 9 ± 41 1 11 ∓ 41 1 11 ∓ 41
Σ̄ = π S ×S ×S ⊂ CP 2 (4).
20 40 40
Now, assume that |T | < |H|. Since Σ̄2 is totally real, we can consider the
following local normal orthonormal frame field
( )
1 1
E3 = JE1 , E4 = JE2 , E5 = JN, E6 = N, E7 , . . . , E2n ,
|N | |N |
where E3 , E4 , E5 , and E6 are globally defined. It can be easily verified that H is
orthogonal to E4 , E6 , and Eα , where α ∈ {7, . . . , 2n}, and, therefore, that
(5.30) H = −|T |E3 − |N |E5 .
All vector fields Eα , α ≥ 7, are orthogonal to H and to J(T Σ̄2 ), which means
that Aα = 0, α ≥ 7 and, therefore, Im σ ⊂ L = span{E3 , E4 , E5 , E6 }. Moreover,
the bundle T Σ̄2 ⊕ L is invariant by J and R̄. Let U be a normal vector field,
orthogonal to L. Using the facts that ∇T = 0, Im σ ⊂ L = span{E3 , E4 , E5 , E6 },
and J(T Σ̄2 ⊕ L) = T Σ̄2 ⊕ L, one obtains
h∇⊥ ¯ ¯
X JY, U i = h∇X JY, U i = hJ ∇X Y, U i = hJ∇X Y + Jσ(X, Y ), U i = 0,
h∇⊥ ¯
X N, U i = h∇X (JH − T ), U i = h−JAH X − σ(X, T ), U i = 0,
and
h∇⊥ ¯
X JN, U i = h∇X (−H − JT ), U i = h−Jσ(X, T ), U i = 0,
that show that L is parallel. We again use [54, Theorem 2] to conclude that Σ̄2 lies
in CP 3 (c).
In the following, we shall determine the shape operators A3 , A4 , A5 , and A6 .
First, since N is orthogonal to H and to J(T Σ̄2 ), we have A6 = 0.
Next, since Σ̄2 is totally real, we have
hσ(X, Y ), JZi = hσ(X, Z), JY i, ∀X, Y, Z ∈ C(T Σ̄2 ),
and, using this property, together with
trace A3 = 2hE3 , Hi = −2|T | and trace A4 = 2hE4 , Hi = 0,
we see that A3 and A4 can be written as

a − |T | b b −a − |T |
A3 = and A4 = .
b −a − |T | −a − |T | −b
As for A5 , we have trace A5 = 2hE5 , Hi = −2|N | and then

e − |N | d
A5 = .
d −e − |N |
Taking into account that E5 is orthogonal to J(T Σ̄2 ), one obtains
hR̄(X, Y )H, E5 i = 0
and, from the Ricci equation (1.5), one sees that [AH , A5 ] = 0 and then, from (5.30),
that [A3 , A5 ] = 0. After a straightforward computation, we get
(5.31) ad = be.
Next, we have hR̄(E2 , E1 )H, JE2 i = −(c/4)|T | and then, again using the Ricci
formula (1.5),
c
h[AH , A4 ]E1 , E2 i = hR̄(E2 , E1 )H, JE2 i = − |T |,
4
which can be written as
c
(5.32) b(b|T | + d|N |) + (a + |T |)(a|T | + e|N |) = |T |.
8
Since Σ̄2 is proper-biharmonic, from the first equation of (5.25), taking into
account that E4 is orthogonal to H and to E3 , we see that trace(AH A4 ) = 0, which,
using (5.31), gives
(5.33) b|T | + d|N | = 0.
Assume now that there exists a point p ∈ Σ̄2 such that b 6= 0 or d 6= 0 at p. Then,
from (5.31), (5.32), and (5.33), we obtain that |T | = 0 at p, which is a contradiction.
Therefore, from (5.33), it follows that b = d = 0 on Σ̄2 , and then equation (5.32)
becomes
c
(5.34) (a + |T |)(a|T | + e|N |) = |T |.
8
Finally, again using the first equation of (5.25), we have
5c c
trace(AH A3 ) = − |T | and trace(AH A5 ) = − |N |,
4 2
or, equivalently,
(5c − 8|H|2 )|T |
(5.35) a(a|T | + e|N |) =
8
and
(c − 4|H|2 )|N |
(5.36) e(a|T | + e|N |) = ,
4
respectively. From (5.34), (5.35), and (5.36) one obtains
(5c − 8|H|2 )|T | (c − 4|H|2 )|N |
(5.37) a= , e=
4(2|H|2 − c) 2(2|H|2 − c)
and
(5.38) 16|H|4 − 10c|H|2 − 3c|T |2 + 2c2 = 0.
The surface Σ̄2 is flat and, therefore, from the Gauss equation (1.3), it follows
5
c X
0=K= + det Aα ,
4
α=3
which, together with (5.37), gives

(5.39) 16|H|4 + 4c|H|2 − 48|T |2 |H|2 + 22c|T |2 − 4c2 = 0.
From (5.38) and (5.39), we obtain |H|2 = c/3, |T |2 = 4c/27, and |N |2 = 5c/27.
Hence, the shape operator A is given by
(5.40) r
r 11 r
1 5c − 31 0

1 c −3 0 1 c 0 1
A3 = , A4 = , A5 = − .
2 3 0 1 2 3 1 0 2 3 0 1
Now, since E1 and E2 are parallel, they determine two distributions which are
mutually orthogonal, smooth, involutive, and parallel. Therefore, from the de Rham
Decomposition Theorem follows, also taking into account that the surface is complete
and using its universal cover if necessary, that Σ̄2 is the standard product γ̄1 × γ̄2 ,
where γ̄k : R → CP 3 (c), k ∈ {1, 2}, are integral curves of E1 and E2 , respectively,
parametrized by arc-length, i.e., γ̄10 = E1 and γ̄20 = E2 (see [95]). In the following,
we shall determine these curves in terms of their curvatures and complex torsions.
Let us denote by κ̄i , 1 ≤ i < 6, the curvatures of γ̄1 and by {Xj1 }, 1 ≤ j < 7, its
Frenet frame field. Using (5.40), we first have
r r
¯ 11 c 1 5c
∇E1 E1 = σ(E1 , E1 ) = − E3 − E5
6 3 6 3
and then, the first Frenet equation of γ̄1 gives
r √ √
7c 1 11 14 70
κ̄1 = and X2 = − E3 − E5 .
6 42 42
Next, since ∇E1 = ∇E2 = 0 and 0 = ∇⊥ H = −|T |∇⊥ E3 − |N |∇⊥ E5 , we get
h∇⊥
E1 E3 , E4 i = 0 and h∇⊥
E1 E3 , E5 i = 0
and
1 ¯ 1
h∇⊥
E1 E3 , E6 i = h∇E1 JE1 , JH − T i = ¯ E E1 , E3 i)
(hAH E1 , E1 i + |T |h∇ 1
|N | |N |
r
1 5c
= −hA5 E1 , E1 i = .
6 3
It follows that ∇⊥ E3 = (1/6) 5c/3E6 . In the same way, one obtains ∇⊥

p
p E 1 E1 E5 =
(−1/3) c/3E6 . Thus, after a straightforward computation, we have
r
¯ 1 1 5c
∇E1 X2 = −κ̄1 E1 − E6 ,
2 42
which means that r
1 5c
κ̄2 = and X31 = −E6 .
2 42
It follows that r r
¯E X = 1 1 5c 1 c
∇ 1 3 E3 − E5
6 3 3 3
and then r √ √
3 c 1 70 11 14
κ̄3 = and X4 = E3 − E5 .
2 42 42 42
Finally, we get ∇ ¯ E X 1 = −κ̄3 X 1 and, therefore, γ̄1 is a helix of osculating order 4.
1 4 3
A simple computation gives its complex torsions
√ √
11 14 70
τ̄12 = −τ̄34 = , τ̄23 = −τ̄14 = , τ̄13 = τ̄24 = 0.
42 42
Hence γ̄1 is a holomorphic helix of order 4.
Consider now the subbundle L = span{E3 , E5 , E6 } in the normal bundle of γ̄1 .
It is easy to see that L is parallel and T γ̄1 ⊕ L is invariant by J and R̄. Then, since
X21 ∈ L, we apply [54, Theorem 2] to conclude that γ̄1 lies in CP 2 (c). Moreover,
it can be easily verified that γ̄1 is of class I3 in the Maeda-Adachi classification of
Frenet curves in CP 2 of osculating order 4 used in the previous section.
For the curve γ̄2 we have
r r
¯ 1 c 1 5c
∇E2 E2 = σ(E2 , E2 ) = E3 − E5 ,
2 3 2 3
p √ √
and then its first curvature is κ̄ = c/2 and X22 = ( 6/6)E3 − ( 30/6)E5 . It can
be easily verified that ∇⊥ ⊥ ¯ 2
E2 E3 = ∇E2 E5 = 0 and then one obtains ∇E2 X2 = −κ̄E p 2
.
3
Therefore, the curve γ̄2 is a holomorphic circle in CP (c) with curvature κ̄ = c/2
and complex torsion τ̄12 = 0.
We then conclude using Remark 1.23.
Remark 5.50. Working in the same way as in the case when c = 4, considered
in [126], the result in Theorem 5.49(2) can be extended to surfaces in CP n (c).
However, for the sake of simplicity we only present this particular case.
7. Biharmonic parallel Lagrangian submanifolds of CP 3

We consider the Hopf fibration π : S2n+1 (1) → CP n (4), and Σ̄ a Lagrangian
submanifold of CP n . Then Σ = π −1 (Σ̄) is an n + 1-dimensional anti-invariant
submanifold of S2n+1 invariant under the flow-action of the characteristic vector field
ξ0 and, locally, Σ is isometric to S1 × Σ
e n . The submanifold Σ̄ is a parallel Lagrangian
submanifold if and only if Σe is an integral C-parallel submanifold (see [109]), and, as
we have seen in Proposition 5.15, a parallel Lagrangian submanifold Σ̄ is biharmonic
if and only if Σ
e is (−4)-biharmonic.
Thus, in order to determine all proper-biharmonic parallel Lagrangian subman-

ifolds of CP 3 , we shall determine the (−4)-biharmonic integral C-parallel submani-
folds of S7 (1).
Just as in the case of Theorems 4.51 and 4.59 we can prove the following results.
Theorem 5.51 ([70]). An integral submanifold  : Σe 3 → S7 (1) is (−4)-biharmo-
nic if and only if
(
∆⊥ H + trace σ(·, AH ·) − 7H = 0
4 trace A∇⊥ H (·) + 3 grad(|H|2 ) = 0,
(·)
where σ is the second fundamental form of Σ

e and A is the shape operator.
e 3 of
Proposition 5.52 ([70]). A non-minimal integral C-parallel submanifold Σ
S7 (1) is (−4)-biharmonic if and only if
(5.41) trace σ(·, AH ·) = 6H.
Theorem 5.53 ([70]). A 3-dimensional integral C-parallel submanifold Σ e 3 of
S7 (1) is (−4)-biharmonic if and only if either:
(1) Σ e 3 is flat and locally is a product of three curves:
• a helix with curvatures κ1 = (λ2 − 1)/λ√and κ2 = 1, √
• a helix of order 4 with 2 2 2
√ curvatures κ1 = λ + α , κ2 = (α/κ1 ) λ + 1
and κ3 = −(λ/κ1 ) λ + 1,2

p p p
κ1 = λ2 + µ2 + δ 2 , κ2 = (δ/κ1 ) λ2 + µ2 + 1, κ3 = (κ2 /δ) λ2 + µ2 ,
p
if δ 6= 0, or a circle with curvature κ1 = λ2 + µ2 , if δ = 0,
where λ, α, µ, δ are real constants given by
 2

 (3λ − 1)(3λ4 − 8λ2 + 1) + λ4 ((α + µ)2 + δ 2 ) = 0,
(α + µ)(5λ2 + α2 + µ2 − 7) + µδ 2 = 0,

(5.42)

 δ(5λ2 + δ 2 + 3µ2 + αµ − 7) = 0,

1 + λ2 + αµ − µ2 = 0

such that −1 < λ < 0, 0 < α ≤ (λ2 − 1)/λ, α ≥ δ ≥ 0, α > 2µ and

λ2 6= 1/3; or
e 3 is locally isometric to a product γ × Σ2 between a helix with curvatures
(2) Σ
√ p √
κ1 = ( 13 − 1)/ 12 − 3 13 and κ2 = 1 and an integral C-parallel surface
of S7 (1) which
q is locally isometric to the 2-dimensional Euclidean sphere
√
with radius 3/(7 − 13).
Proof. It is easy to see, using (4.46), that equation (5.41) is equivalent to the
system
(5.43)
 P3
t λ 2 − 6 + (α + µ)(αλ + µλ ) + (β + δ)(βλ + δλ ) = 0,

 i=1 i 2 3 2 3
2 2 2 2 2
t(αλ2 + µλ3 ) + (α + µ)(2λ2 + α + 3β + µ + βδ − 6) + µ(β + δ) = 0,
t(βλ2 + δλ3 ) + β(α + µ)2 + (β + δ)(2λ23 + δ 2 + 3µ2 + β 2 + αµ − 6) = 0,

where t = 3i=1 λi .
P
We shall split the study of this system as Σ e is given by the first or the second
case of the classification of non-minimal integral C-parallel submanifolds in Section
7 of Chapter 4.
Case I. The system (5.43) is equivalent to the system given by the first three
equations of (5.42) and, just like in the proof of Theorem 4.59, we conclude.
Case II. (1) It is easy to verify that this case cannot occur in this setting.
(2) The second and the third equation of system (5.43) are satisfied and the
4 2 2
√ equation is 2equivalent to 3λ − 8λ2 + 1 =√0, whose solutions are λ = (4 ±
first
13)/3. Since λ < 1, it follows that λ = (4 − 13)/3 and this, together with the
classification of integral C-submanifolds, leads to the conclusion.
When c = 1, solving (5.43) and using the explicit equation of the 3-dimensional
integral C-parallel flat submanifolds in S7 (1) given in [9], we obtain the following
corollary.
Corollary 5.54 ([70]). Any 3-dimensional flat (−4)-biharmonic integral C-
parallel submanifold Σe 3 of S7 (1) is locally given by
λ 1
x(u, v, w) = √ exp i u E1
λ2 + 1 λ
1
(µ − α)(2µ − α)
1
ρ1 (ρ1 + ρ2 )
1
ρ2 (ρ1 + ρ2 )
p
where ρ1,2 = ( 4µ(2µ − α) + δ 2 ±δ)/2, −1 < λ < 0, 0 < α ≤ (λ2 −1)/λ, α ≥ δ ≥ 0,
α > 2µ, λ2 6= 1/3, (λ, α, µ, δ) being one of the following
s √ s √ s √ !
4 − 13 7 − 13 7 − 13
− , , − , 0 ,
3 6 6
s s √ s !
1 45 + 21 3 6
− √ , , − √ , 0 ,
5+2 3 13 21 + 11 3
or s s √ s √ s √ !
1 523 + 139 13 79 − 17 13 14 + 2 13
− √ , , − ,
6 + 13 138 138 3
and {Ei }4i=1 is an orthonormal basis of C4 with respect to the usual Hermitian inner
product.
Remark 5.55 ([70]). By a straightforward computations we can check that
the images of the cylinders over
p the√above√x are, respectively: the Riemannian
product of a circle of radius 5 − 13/(2 3) and three circles,peach of radius
p √ √ √
7 + 13/6; the Riemannian p product of two circles each of radius 3 + 3/(2 3)
√ √
and two circles each of radius 3 − 3/(2 3); the Riemannian p product of a circle
p √ √ √
of radius 5 + 13/(2 3) and three circles, each of radius 7 − 13/6.
Further Developments
In [72] and [76], we considered pmc submanifolds in M n (c) × R and found the
expressions of the Laplacians of |AH |2 and |σ|2 , where A, σ, and H are the shape
operator, the second fundamental form, and the mean curvature vector field, respec-
tively. We then used these Simons type equations mainly to prove gap theorems.
We will continue our study on such submanifolds and we expect to be able to find
other applications of these formulas. One possible application is suggested by the
work of T. H. Colding and W. P. Minicozzi [45], where, using the original Simons’
inequality [128], they found curvature estimates for minimal surfaces in R3 (for an
excellent presentation of other applications of Simons’ inequality in the study of
minimal surfaces, see also [46]).
P. Bérard, M. do Carmo, and W. Santos [18] considered, for the first time,
submanifolds with finite total curvature in space forms, i.e., submanifolds Σm sat-
isfying Σm |φ|m dv < +∞, where φ is the traceless part of the second fundamental
R
form. Since φ measures by how much a submanifold fails to be totally umbilical,
the above condition is quite a natural one. We studied such submanifolds in [71],
where we proved compactness results for cmc biconservative surfaces with finite total
curvature in Hadamard manifolds, while in [17] we used a different definition, when
working in M 2 (c) × R, to obtain the same type of results for cmc surfaces. Thus, a
2 2 2
R
surface Σ in M (c)×R has finite total curvature if Σ2 |S| dv < +∞, where S is the
traceless part of the Abresch-Rosenberg differential [1]. We were motivated in our
choise by the fact that, in this case, Abresch-Rosenberg differential plays the same
role as the classical Hopf differential in space forms. Since the Abresch-Rosenberg
differential is also defined (meaning that it is holomorphic) for cmc surfaces in some
other homogeneous 3-manifolds (see [2]), we think that it would be interesting to
also study surfaces that have finite total curvature in these spaces too.
Y.-L. Ou and Z.-P. Wang [119] studied the biharmonicity of cmc surfaces in
3-dimensional Thurston geometries. One of our prospective research’s goal is to do
the same thing without the constant mean curvature hypothesis. We expect to ob-
tain classification results and explicit examples, especially when working in product
spaces M 2 (c) × R. We will also continue to study biharmonic and biconservative
submanifolds in M n (c) × R, n ≥ 3. We expect that we will be able to extend
our results in [71] from cmc surfaces to cmc (or pmc) submanifolds with arbitrary
dimension.
Surfaces with constant mean curvature in 3-dimensional metric semidirect prod-
ucts were studied by W. H. Meeks and J. Pérez [105]. It seems interesting to also
study their biharmonicity, as well as the biharmonicity of curves in these spaces,
and this could also be a direction of our research.
In [73], we determined all complete pmc proper-biharmonic surfaces with non-
negative Gaussian curvature in complex space forms. We are interested in improving
157
these results and we will study the biharmonicity of surfaces in these spaces without
any other supplementary hypotheses. We will also consider biconservative surfaces
in this context.
The theory of minimal and cmc surfaces in 3-dimensional Riemannian manifolds
is a beautiful and classical one and I am sure that a course devoted to this subject
would appeal to a large number of graduate students in our universities willing to
study Differential Geometry. As I had the opportunity to witness myself, during
a two year postdoctoral fellowship at The National Institute of Pure and Applied
Mathematics in Rio de Janeiro, such a course, taught by Harold Rosenberg, was one
of the best liked by the students of the leading mathematical institution of South
America. Obviously, this is not the only possibility offered by such a rich topic, as it
is proved by the fact that many of the most interesting mathematicians all over the
world obtained their PhD degrees with theses on cmc or pmc submanifolds. A course
on biharmonic submanifolds would also be very interesting for Romanian graduate
students and there also are a lot of open problems and new research directions in
this field, that could lead to excellent PhD theses. That is why I would like to teach
such courses in a Romanian university and also to direct PhD theses devoted to
this subjects. I even have gained some experience teaching an invited mini course
on biharmonic submanifolds as part of the summer postgraduate program of the
Institute of Mathematics of the Federal University of Bahia in Salvador and I have
certainly enjoyed doing this.
During the period of more than ten years since I have received my PhD de-
gree I had the opportunity to collaborate with very good mathematicians not only
from Romania but also from Brazil, France, and Italy. I intend to continue these
collaborations and also contribute, if possible, to the development of institutional
cooperation with these countries. In particular, I would like to be part in a fu-
ture possible cooperation with universities from Brazil, where I spent four years as
a postdoctoral researcher and Visiting Professor. It should be mentioned that (at
least as far as I know) there is not even a student exchange agreement between our
countries yet, even though, in the last years, Brazil developed an intensive coopera-
tion with almost all the countries in the European Union with great benefits for all
those implied.
Index
3-Sasakian manifold, 110 parallel submanifold, 4

λ-bienergy, 134 pmc submanifold, 4
λ-biharmonic submanifold, 134 proper-biharmonic map, 73
ϕ-sectional curvature, 13
ϕ-symmetric space, 24 Sasakian manifold, 24
ϕ-torsion, 99 Sasakian space form, 24
slant surface, 22
almost contact 3-structure, 109 stress-energy tensor, 74
almost contact metric structure, 12 submanifold with finite total curvature, 62
anti-invariant submanifold, 13
vertical cylinder, 14
biconservative submanifold, 74
bienergy functional, 73
biharmonic map, 73
biharmonic submanifold, 74
bitension field, 73
Boothby-Wang fibration, 28
circle, 13
cmc submanifold, 4
complex space form, 4
complex torsion, 14
constant angle surface, 66
cosymplectic manifold, 13
cosymplectic space form, 13
energy functional, 73
Frenet curve, 13
harmonic map, 73
height function, 75
helix, 13
helix of order r, 13
helix surface, 66
holomorphic helix, 14
Hopf cylinder, 28
integral C-parallel submanifold, 120

integral submanifold, 24
invariant submanifold, 13
Lagrangian submanifold, 130

Legendre curve, 24
normal almost contact metric structure, 12
159
Bibliography
[1] U. Abresch and H. Rosenberg, A Hopf differential for constant mean curvature surfaces in
S2 × R and H2 × R, Acta Math. 193 (2004), 141–174.
[2] U. Abresch and H. Rosenberg, Generalized Hopf differentials, Mat. Contemp. 28 (2005), 1–28.
[3] P. Alegre, D. E. Blair, and A. Carriazo, Generalized Sasakian space forms, Israel J. Math. 141
(2004), 157–183.
[4] H. Alencar and M. do Carmo, Hypersurfaces with constant mean curvature in spheres, Proc.
Amer. Math. Soc. 120 (1994), 1223–1229.
[5] H. Alencar, M. do Carmo, and R. Tribuzy, A theorem of Hopf and the Cauchy-Riemann in-
equality, Comm. Anal. Geom. 15 (2007), 283–298.
[6] H. Alencar, M. do Carmo, and R. Tribuzy, A Hopf Theorem for ambient spaces of dimensions
higher than three, J. Differential Geometry 84 (2010), 1–17.
[7] L. J. Alı́as and S. C. Garcı́a-Martı́nez, On the scalar curvature of constant mean curvature
hypersurfaces in space forms, J. Math. Anal. Appl. 363 (2010), 579–587.
[8] K. O. Araújo and K. Tenenblat, On submanifolds with parallel mean curvature vector, Kodai
Math. J. 32 (2009), 59–76.
[9] C. Baikoussis, D. E. Blair, and T. Koufogiorgos, Integral submanifolds of Sasakian space forms
M̄ 7 , Results Math. 27 (1995), 207–226.
[10] P. Baird and J. Eells, A conservation law for harmonic maps, Geometry Symposium, Utrecht
1980, 1–25, Lecture Notes in Math. 894, Springer, Berlin-New York, 1981.
[11] A. Balmuş, S. Montaldo, and C. Oniciuc, Classification results for biharmonic submanifolds in
spheres, Israel J. Math. 168 (2008), 201–220.
[12] A. Balmuş, S. Montaldo, and C. Oniciuc, Properties of biharmonic submanifolds in spheres, J.
Geom. Symmetry Phys. 17 (2010), 87–102.
[13] A. Balmuş, S. Montaldo, and C. Oniciuc, Biharmonic hypersurfaces in 4-dimensional space
forms, Math. Nachr. 283 (2010), 1696–1705.
[14] A. Balmuş, S. Montaldo, and C. Oniciuc, Biharmonic PNMC submanifolds in spheres, Ark.
Mat. 51 (2013), 197–221.
[15] A. Balmuş and C. Oniciuc, Biharmonic submanifolds with parallel mean curvature vector field
in spheres, J. Math. Anal. Appl. 386 (2012), 619–630.
[16] M. H. Batista da Silva, Simons type equation in S2 × R and H2 × R and applications, Ann.
Inst. Fourier 61 (2011), 1299–1322.
[17] M. Batista, M. P. Cavalcante, and D. Fetcu, Constant mean curvature surfaces in M2 (c) × R
and finite total curvature, arXiv:1402.1231, preprint 2014.
[18] P. Bérard, M. do Carmo, and W. Santos, Complete hypersurfaces with constant mean curvature
and finite total curvature, Ann. Global Anal. Geom. 16 (1998), 273–290.
[19] D. E. Blair, Riemannian Geometry of Contact and Symplectic Manifolds, Birkhäuser Boston,
Progress in Mathematics, 203, 2002.
[20] D. E. Blair and S. I. Goldberg, Topology of almost contact manifolds, J. Differential Geometry
1 (1967), 347–354.
[21] R. Caddeo, S. Montaldo, and C. Oniciuc, Biharmonic submanifolds of S3 , Internat. J. Math.
12 (2001), 867–876.
[22] R. Caddeo, S. Montaldo, and C. Oniciuc, Biharmonic submanifolds in spheres, Israel J. Math.
130 (2002), 109–123.
[23] R. Caddeo, S. Montaldo, and P. Piu, Biharmonic curves in a surface, Rend. Mat. Appl. 21
(2001), 143–157.
161
162 Bibliography
[24] R. Caddeo, C. Oniciuc, and P. Piu, Explicit formulas for non-geodesic biharmonic curves of
the Heisenberg group, Rend. Sem. Mat. Univ. Politec. Torino 62 (2004), 265–278.
[25] R. Caddeo, S. Montaldo, C. Oniciuc, and P. Piu, The clasification of biharmonic curves of
Cartan-Vrânceanu 3-dimensional spaces, Modern trends in geometry and topology, 121–131,
Cluj Univ. Press, Cluj-Napoca, 2006.
[26] R. Caddeo, S. Montaldo, C. Oniciuc, and P. Piu, The Euler-Lagrange method for biharmonic
curves, Mediterr. J. Math. 3 (2006), 449–465.
[27] R. Caddeo, S. Montaldo, C. Oniciuc, and P. Piu, Surfaces in three-dimensional space forms
with divergence-free stress-bienergy tensor, Ann. Mat. Pura Appl. (4) 193 (2014), 529–550.
[28] M. do Carmo, L.-F. Cheung, and W. Santos, On the compactness of constant mean curvature
hypersurfaces with finite total curvature, Arch. Math. (Basel) 73 (1999), 216–222.
[29] I. Castro, C. R. Montealegre, and F. Urbano, Minimal Lagrangian submanifolds in the complex
hyperbolic space, Illinois J. Math. 46 (2002), 695–721.
[30] B.-Y. Chen, Minimal hypersurfaces of an m-sphere, Proc. Amer. Math. Soc. 29 (1971), 375–380.
[31] B.-Y. Chen, A report on submanifolds of finite type, Soochow J. Math. 22 (1996), 117–337.
[32] B.-Y. Chen, Special slant surfaces and a basic inequality, Results Math. 33 (1998), 65–78.
[33] D. Chen, G. Chen, H. Chen, and F. Dillen, Constant angle surfaces in S3 (1) × R, Bull. Belg.
Math. Soc. Simon Stevin 19 (2012), 289–304.
[34] B.-Y. Chen and S. Ishikawa, Biharmonic pseudo-Riemannian submanifolds in pseudo-Euclidian
spaces, Kyushu J. Math. 52 (1998), 167–185.
[35] B.-Y. Chen and G. D. Ludden, Surfaces with mean curvature vector parallel in the normal
bundle, Nagoya Math. J. 47 (1972), 161–167.
[36] B.-Y. Chen and K. Ogiue, On totally real submanifolds, Trans. Amer. Math. Soc. 193 (1974),
257–266.
[37] B.-Y. Chen and M. Okumura, Scalar curvature, inequality and submanifold, Proc. Amer. Math.
Soc. 38 (1973), 605–608.
[38] B.-Y. Chen and K. Yano, Pseudo-umbilical submanifolds in a Riemannian manifold of constant
curvature, Differential geometry (in honor of Kentaro Yano), Kinokuniya, Tokyo, 1972, 61–71.
[39] Q.-M. Cheng and K. Nonaka, Complete submanifolds in Euclidean spaces with parallel mean
curvature vector, Manuscripta Math. 105 (2001), 353–366.
[40] S.-Y. Cheng and S.-T. Yau, Hypersurfaces with constant scalar curvature, Math. Ann. 225
(1977), 195–204.
[41] S.-S. Chern, On surfaces of constant mean curvature in a three-dimensional space of constant
curvature, Geometric dynamics (Rio de Janeiro, 1981), Lecture Notes in Math. 1007, Springer,
Berlin, 1983, 104–108.
[42] S.-S. Chern and J. Wolfson, Minimal surfaces by moving frames, Amer. J. Math 105 (1983),
59–83.
[43] D. Chinea, M. de León, and J. C. Marrero, Topology of cosymplectic manifolds, J. Math. Pures
Appl. 72 (1993), 567–591.
[44] J. T. Cho, J. Inoguchi, and J.-E. Lee, Biharmonic curves in 3-dimensional Sasakian space
form, Ann. Math. Pura Appl. 186 (2007), 685–701.
[45] T. H. Colding and W.P. Minicozzi II, Estimates for parametric elliptic integrands, Int. Math.
Res. Not. 2002 (6), 291–297.
[46] T. H. Colding and W. P. Minicozzi II, A course in minimal surfaces, American Mathematical
Society, Providence, RI, Graduate Studies in Mathematics 121, 2011.
[47] F. Dillen and L. Vrancken, C-totally real submanifolds of S7 (1) with non-negative sectional
curvature, Math. J. Okayama Univ. 31 (1989), 227–242.
[48] I. Dimitric, Submanifolds of Em with harmonic mean curvature vector, Bull. Inst. Math. Acad.
Sinica 20 (1992), 53–65.
[49] J. Eells and J. H. Sampson, Harmonic mappings of Riemannian manifolds, Amer. J. Math. 86
(1964), 109–160.
[50] J. Eells and L. Lemaire, Selected topics in harmonic maps, Conf. Board. Math. Sci. 50 (1983).
[51] H. Endo, A note on invariant submanifolds in an almost cosymplectic manifold, Tensor (N.S.)
43 (1986), 75–78.
[52] J. Erbacher, Isometric immersions of constant mean curvature and triviality of the normal
connection, Nagoya Math. J. 45 (1971), 139–165.
Bibliography 163
[53] J. Erbacher, Reduction of the codimension of an isometric immersion, J. Differential Geometry

5 (1971), 333–340.
[54] J. H. Eschenburg and R. Tribuzy, Existence and uniqueness of maps into affine homogeneous
spaces, Rend. Sem. Mat. Univ. Padova 89 (1993), 11–18.
[55] R. H. Escobales, Riemannian submersions from complex projective space, J. Differential Geom.
13 (1978), 93–107.
[56] J. M. Espinar and H. Rosenberg, Complete constant mean curvature surfaces in homogeneous
spaces, Comment. Math. Helv. 86 (2011), 659–674.
[57] M. J. Ferreira and R. Tribuzy, Parallel mean curvature surfaces in symmetric spaces, Ark. Mat.
52 (2014), 93–98.
[58] D. Ferus, The torsion form of submanifolds in E N , Math. Ann. 193 (1971), 114–120.
[59] D. Fetcu, Integral submanifolds in three-Sasakian manifolds whose mean curvature vector fields
are eigenvectors of the Laplace operator, Geometry, integrability and quantization, 210–223,
Softex, Sofia, 2008.
[60] D. Fetcu, Biharmonic Legendre curves in Sasakian space forms, J. Korean Math. Soc. 45 (2008),
393–404.
[61] D. Fetcu, A note on biharmonic curves in Sasakian space forms, Ann. Mat. Pura Appl. (4)
189 (2010), 591–603.
[62] D. Fetcu, Surfaces with parallel mean curvature vector in complex space forms, J. Differential
Geom. 91 (2012), 215–232.
[63] D. Fetcu, A classification result for helix surfaces with parallel mean curvature in product spaces,
Ark. Mat. 53 (2015), 249–258.
[64] D. Fetcu, E. Loubeau, S. Montaldo, and C. Oniciuc, Biharmonic submanifolds of CP n , Math.
Z. 266 (2010), 505–531.
[65] D. Fetcu and C. Oniciuc, Explicit formulas for biharmonic submanifolds in non-Euclidean 3-
spheres, Abh. Math. Sem. Univ. Hamburg 77 (2007), 179–190.
[66] D. Fetcu and C. Oniciuc, On the geometry of biharmonic submanifolds in Sasakian space forms,
J. Geom. Symmetry Phys. 14 (2009), 21–34.
[67] D. Fetcu and C. Oniciuc, Biharmonic hypersurfaces in Sasakian space forms, Differential Geom.
Appl. 27 (2009), 713–722.
[68] D. Fetcu and C. Oniciuc, Explicit formulas for biharmonic submanifolds in Sasakian space
forms, Pacific J. Math. 240 (2009), 85–107.
[69] D. Fetcu and C. Oniciuc, A note on integral C-parallel submanifolds in S7 (c), Rev. Un. Mat.
Argentina 52 (2011), 33–45.
[70] D. Fetcu and C. Oniciuc, Biharmonic integral C-parallel submanifolds in 7-dimensional
Sasakian space forms, Tohoku Math. J. 64 (2012), 195–222.
[71] D. Fetcu, C. Oniciuc, and A. L. Pinheiro, CMC biconservative surfaces in Sn × R and Hn × R,
J. Math. Anal. Appl. 425 (2015), 588–609.
[72] D. Fetcu, C. Oniciuc, and H. Rosenberg, Biharmonic submanifolds with parallel mean curvature
in Sn × R, J. Geom. Anal. 23 (2013), 2158–2176.
[73] D. Fetcu and A. L. Pinheiro, Biharmonic surfaces with parallel mean curvature in complex
space forms, Kyoto J. Math. 55 (2015), 837–855.
[74] D. Fetcu and H. Rosenberg, A note on surfaces with parallel mean curvature, C. R. Math.
Acad. Sci. Paris 349 (2011), 1195–1197.
[75] D. Fetcu and H. Rosenberg, Surfaces with parallel mean curvature in S3 × R and H3 × R,
Michigan Math. J. 61 (2012), 715–729.
[76] D. Fetcu and H. Rosenberg, On complete submanifolds with parallel mean curvature in product
spaces, Rev. Mat. Iberoam. 29 (2013), 1283–1306.
[77] D. Fetcu and H. Rosenberg, Surfaces with parallel mean curvature in CP n × R and CH n × R,
Trans. Amer. Math. Soc. 366 (2014), 75–94.
[78] D. Fetcu and H. Rosenberg, Surfaces with parallel mean curvature in Sasakian space forms,
Math. Ann. 362 (2015), 501–528.
[79] K. R. Frensel, Stable complete surfaces with constant mean curvature, Bol. Soc. Bras. Mat. 27
(1996), 129–144.
[80] Y. Fu, Explicit classification of biconservative surfaces in Lorentz 3-space forms, Ann. Mat.
Pura Appl. (4) 194 (2015), 805–822.
164 Bibliography
[81] S. Hirakawa, Constant Gaussian curvature surfaces with parallel mean curvature vector in two-
dimensional complex space forms, Geom. Dedicata 118 (2006), 229–244.
[82] D. Hilbert, Die grundlagen der physik, Math. Ann. 92 (1924), 1–32.
[83] D. A. Hoffman, Surfaces of constant mean curvature in manifolds of constant curvature, J.
Differential Geom. 8 (1973), 161–176.
[84] D. Hoffman and J. Spruck, Sobolev and isoperimetric inequalities for Riemannian submanifolds,
Comm. Pure. Appl. Math. 27 (1974), 715–727.
[85] H. Hopf, Differential geometry in the large, Lecture Notes in Math. 1000, Springer-Verlag,
1983.
[86] W. Y. Hsiang and H. B. Lawson, Minimal submanifolds of low cohomogeneity, J. Differential
Geometry 5 (1971), 1–38.
[87] A. Huber, On subharmonic functions and differential geometry in the large, Comment. Math.
Helv. 32 (1957), 13–71.
[88] T. Ichiyama, J. Inoguchi, and H. Urakawa, Bi-harmonic maps and bi-Yang-Mills fields, Note
Mat. 28 (2009), suppl. 1, 233–275.
[89] J. Inoguchi, Submanifolds with harmonic mean curvature in contact 3-manifolds, Colloq. Math.
100 (2004), 163–179.
[90] I. Ishihara, Anti-invariant submanifolds of a Sasakian space form, Kodai Math. J. 2 (1979),
171–186.
[91] G. Y. Jiang, 2-harmonic maps and their first and second variational formulas, Chinese Ann.
Math. Ser. A7(4) (1986), 389–402.
[92] G. Y. Jiang, The conservation law for 2-harmonic maps between Riemannian manifolds, Acta
Math. Sinica 30 (1987), 220–225.
[93] G. Y. Jiang, 2-harmonic maps and their first and second variational formulas. Translated from
the Chinese by Hajime Urakawa, Note Mat. 28 (2008), suppl. 1, 209–232.
[94] K. Kenmotsu and D. Zhou, The classification of the surfaces with parallel mean curvature
vector in two-dimensional complex space forms, Amer. J. Math. 122 (2000), 295–317.
[95] S. Kobayashi and K. Nomizu, Foundations of Differential Geometry, vol. I, Interscience Pub-
lishers, New York, London, 1963.
[96] H. B. Lawson, Rigidity theorems in rank-1 symmetric spaces, J. Differential Geometry 4 (1970),
349–357.
[97] A.-M. Li and J. M. Li, An intrinsic rigidity theorem for minimal submanifolds in a sphere,
Arch. Math. 58 (1992), 582–594.
[98] J. H. Lira, R. Tojeiro, and F. Vitório, A Bonnet theorem for isometric immersions into products
of space forms, Arch. Math. (Basel) 95 (2010), 469–479.
[99] A. Lotta, Slant submanifolds in contact geometry, Bull. Math. Soc. Sci. Math. Roumanie (N.S.)
39 (1996), 183-198.
[100] E. Loubeau, S. Montaldo, and C. Oniciuc, The stress-energy tensor for biharmonic maps,
Math. Z. 259 (2008), 503–524.
[101] E. Loubeau and C. Oniciuc, Biharmonic surfaces of constant mean curvature, Pacific J. Math.
271 (2014), 213–230.
[102] S. Maeda and T. Adachi, Holomorphic helices in a complex space form, Proc. Amer. Math.
Soc. 125 (1997), 1197–1202.
[103] S. Maeda and Y. Ohnita, Helical geodesic immersions into complex space forms, Geom. Ded-
icata 30 (1989), 93–114.
[104] S. Maeta and H. Urakawa, Biharmonic Lagrangian submanifolds in Kähler manifolds, Glasg.
Math. J. 55 (2013), 465–480.
[105] W. H. Meeks and J. Pérez, Constant mean curvature surfaces in metric Lie groups, Geometric
analysis: partial differential equations and surfaces, 25–110, Contemp. Math., 570, Amer. Math.
Soc., Providence, RI, 2012.
[106] S. Montaldo and C. Oniciuc, A short survey on biharmonic maps between Riemannian man-
ifolds, Rev. Un. Mat. Argentina 47 (2006), 1–22.
[107] S. Montaldo, C. Oniciuc, and A. Ratto, Proper biconservative immersions into the Euclidean
space, Ann. Mat. Pura Appl. 195 (2016), 403–422.
[108] S. Montaldo, C. Oniciuc, and A. Ratto, Biconservative surfaces, J. Geom. Anal. 26 (2016),
313–329.
Bibliography 165
[109] H. Naitoh, Parallel submanifolds of complex space forms I, Nagoya Math. J. 90 (1983), 85–117.
[110] B. O’Neill, Semi-Riemannian Geometry with Applications to Relativity, Pure and Applied
Mathematics 103, Academic Press, New York, 1983.
[111] R. Niebergall and P. J. Ryan, Real hypersurfaces in complex space forms, Tight and Taut
Submanifolds, MSRI Publications 32 (1997), 233–305.
[112] K. Nomizu and B. Smyth, A formula of Simons’ type and hypersurfaces with constant mean
curvature, J. Differential Geometry 3 (1969), 367–377.
[113] T. Ogata, Surfaces with parallel mean curvature vector in P 2 (C), Kodai Math. J. 18 (1995),
397–407.
[114] T. Ogata, Curvature pinching theorem for minimal surfaces with constant Kähler angle in
complex projective spaces, Tôhoku Math. J. 43 (1991), 361–374.
[115] M. Okumura, Some remarks on space with a certain contact structure, Tôhoku Math. J. 14
(1962), 135–145.
[116] M. Okumura, Hypersurfaces and a pinching problem on the second fundamental tensor, Amer.
J. Math. 96 (1974), 207–213.
[117] Y. L. Ou, On conformal biharmonic immersions, Ann. Global Anal. Geom. 36 (2009), 133–
142.
[118] Y.-L. Ou, Biharmonic hypersurfaces in Riemannian manifolds, Pacific J. Math. 248 (2010),
217–232.
[119] Y.-L. Ou and Z.-P. Wang, Constant mean curvature and totally umbilical biharmonic surfaces
in 3-dimensional geometries, J. Geom. Phys. 61 (2011), 1845–1853.
[120] H. Reckziegel, Horizontal lifts of isometric immersions into the bundle space of a pseudo-
Riemannian submersion, Global differential geometry and global analysis 1984 (Berlin, 1984),
264–279, Lecture Notes in Math., 1156, Springer, Berlin, 1985.
[121] G. Ruiz-Hernández, Minimal helix surfaces in N n × R, Abh. Math. Semin. Univ. Hambg. 81
(2011), 55–67.
[122] A. Sanini, Applicazioni tra varietà riemanniane con energia critica rispetto a deformazioni di
metriche, Rend. Mat. 3 (1983), 53–63.
[123] W. Santos, Submanifolds with parallel mean curvature vector in spheres, Tôhoku Math. J. 46
(1994), 403–415.
[124] T. Sasahara, Submanifolds in a Sasakian manifold R2n+1 (−3) whose φ-mean curvature vectors
are eigenvectors, J. Geom. 75 (2002), 166–178.
[125] T. Sasahara, Legendre surfaces in Sasakian space forms whose mean curvature vectors are
eigenvectors, Publ. Math. Debrecen 67 (2005), 285–303.
[126] T. Sasahara, Biharmonic Lagrangian surfaces of constant mean curvature in complex space
forms, Glasg. Math. J. 49 (2007), 497–507.
[127] N. Sato, Totally real submanifolds of a complex space form with nonzero parallel mean curva-
ture vector, Yokohama Math. J. 44 (1997), 1–4.
[128] J. Simons, Minimal varieties in Riemannian manifolds, Ann. of Math. (2) 88 (1968), 62–105.
[129] B. Smyth, Submanifolds of constant mean curvature, Math. Ann 205 (1973), 265–280.
[130] R. Takagi, On homogeneous real hypersurfaces in a complex projective space, Osaka J. Math.
10 (1973), 495–506.
[131] T. Takahashi, Sasakian φ-symmetric spaces, Tôhoku Math. J. 29 (1977), 91–113.
[132] S. Tanno, Sasakian manifolds with constant ϕ-holomorphic sectional curvature, Tôhoku Math.
J. 21 (1969), 501–507.
[133] S. Tanno, The topology of contact Riemannian manifolds, Ill. J. Math. 12 (1968), 700–717.
[134] R. Tribuzy, Hopf ’s method and deformations of surfaces preserving mean curvature, An. Acad.
Brasil. Cienc. 50 (1978), 447–450.
[135] K. Yano and M. Kon, Structures on Manifolds, Series in Pure Mathematics 3, World Scientific
Publishing Co., Singapore, 1984.
[136] S.-T. Yau, Submanifolds with constant mean curvature. I, Amer. J. Math. 96 (1974), 346–366.
[137] S.-T. Yau, Harmonic functions on complete Riemannian manifolds, Commun. Pure. Appl.
Math. 28 (1975), 201–228.
[138] B. White, Complete surfaces of finite total curvature, J. Differential Geom. 26 (1987), 315–326.
[139] W. Zhang, New examples of biharmonic submanifolds in CP n and S2n+1 , An. Ştiinţ. Univ.
Al. I. Cuza Iaşi. Mat. (N.S.) 57 (2011), 207–218.

Fetcu HT Final

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fetcu HT Final

Uploaded by

Copyright:

Available Formats

UNIVERSITATEA TEHNICĂ ”GHEORGHE ASACHI” DIN IAŞI

DEPARTAMENTUL DE MATEMATICĂ ŞI INFORMATICĂ

Submanifolds with Parallel Mean

unde τ (ψ) = trace ∇ψ dψ este câmpul tensiune al aplicaţiei ψ, ∇ψ fiind conexiunea

Începem prezentarea subvarietăţilor biarmonice ı̂n capitolul al treilea, construit

Part 1. Submanifolds with Parallel Mean Curvature 1

Chapter 1. Reduction of Codimension Results and Holomorphic Differentials

Chapter 2. Simons Type Formulas and Applications. Surfaces with Parallel

Part 2. Biharmonic Submanifolds 71

Chapter 3. Biharmonic and Biconservative Submanifolds in M n (c) × R 73

Chapter 4. Biharmonic Submanifolds in Sasakian Space Forms 95

5. Biharmonic anti-invariant submanifolds in Sasakian space forms 111

Submanifolds with Parallel Mean

Reduction of Codimension Results and Holomorphic

2. Surfaces with parallel mean curvature in complex space forms

and the equation of Ricci

Now, from the Codazzi equation (1.4), again using ∇⊥ H = 0, we obtain

Next, since ∇Z Z̄ = 0, we have

In order to classify pmc 2-spheres in N 2 (c), we will use a result of T. Ogata in

2.4. Reduction of codimension. Let Σ2 be a surface in a complex space form

=hJAH Ei , V i + hσ(Ei , (JH)> ), V i

= − hJ∇Ei Ej , V i − hJσ(Ei , Ej ), V i + hσ(Ei , (JEj )> ), V i

and, using the Codazzi equation (1.4),

that shows that Aijk = Akji = Aikj .

It is easy to see that, if V is orthogonal to L, then JV is normal and orthogonal

=hJAσ(Ei ,Ej ) Ek , V i − hJ∇⊥

+ hσ(Ek , (Jσ(Ei , Ej ))> ), V i

Next, since Z̄(Q(Z, Z)) = 0, we have

Next, since Σ2 is not holomorphic or anti-holomorphic, we have that cos θ 6= ±1

where AH Ei = µi Ei . Since µ1 + µ2 = 2|H|2 , we have

3. Surfaces with parallel mean curvature in CP n × R and CH n × R

Definition 1.18. An almost contact metric manifold (N, ϕ, ξ, η, h, i) is a cosym-

3.2. A holomorphic differential. Although our main interest is to study pmc

(1.25) 3hϕE2 , E1 ihϕH, E1 i = η(E2 )η(H),

But we also know that ∇N 2

hϕH, ∇E2 E2 i = hϕH, E1 ih∇E2 E2 , E1 i = 0,

and then, from (1.30) and (1.25), we have

Lemma 1.38 ([77]). If ξ has non-vanishing tangent and normal components on

In the same way, we get

Therefore, we have the following theorem.

Remark 1.43. A surface Σ2 in a cosymplectic space form is called a slant surface

4. Surfaces with parallel mean curvature in Sasakian space forms

• if c < −3, then N 2n+1 (c) is isometric to B 2n × R, where B 2n is the unit

4.2. Reduction of codimension. M. J. Ferreira and R. Tribuzy [57] proved

Next, let us consider a normal vector field V ∈ L. Then

which means that ∇¯ X V ∈ T Σ2 ⊕ L if and only if ∇⊥ V ∈ L. Since from Proposition

4.3. Anti-invariant pmc surfaces.

It is easy to verify that ∇ξ E1H = ∇E H ξ = 0, where ∇ is the connection on the

Lemma 1.61 ([78]). The following equations hold on Σ2 :

Proof. Since H is parallel and umbilical, we have h∇⊥

for any vector field X tangent to the surface.

Finally, we again use that H is parallel and umbilical, ∇N

and, in the same way,

= 6a2 {K + 6(h∇E2 E2 , E1 i2 + h∇E1 E1 , E2 i2 )} ≥ 0.

Next, using Lemma 1.61, we get, after a straightforward computation,

Then, using Lemma 1.61, we obtain ∇N 2

Theorem 1.68 ([78]). There are no anti-invariant complete non-minimal non-

Proof. From expression (1.38) of RN , we get that

Using the fact that ∇E1 = ∇E2 = 0 and (1.55), we have

Simons Type Formulas and Applications. Surfaces with

Hn (c), as c > 0, c = 0, or c < 0, respectively.

where ξ is the unit vector field tangent to R.