Totally geodesic property of the Hopf vector field. *   *   Acta Math. Hungar. 101, 1-2 (2003), 73-92.

Yampolsky A.

Abstract
We prove that the Hopf vector field is a unique one among geodesic unit vector fields on spheres such that the submanifold generated by the field is totally geodesic in the unit tangent bundle with Sasaki metric. As application, we give a new proof of stability (instability) of the Hopf vector field with respect to volume variation using standard approach from the theory of submanifolds and find exact boundaries for the sectional curvature of the Hopf vector field. 1   Introduction Let ( M , g )   be ( n + 1 )   -dimensional Riemannian manifold with metric g   and ξ   a fixed unit vector field on M   . Consider ξ   as a local mapping ξ : M → T 1 M   .
Then the image ξ ( M )   is a submanifold in the unit tangent sphere bundle T 1 M   .
The Sasaki metric on the tangent bundle T M   induces the Riemannian metric on T 1 M   and on ξ ( M )   as well. So, one may use the geometrical properties of this submanifold to determine geometrical characteristics of a unit vector field.
Namely, a unit vector field ξ   is said to be minimal if ξ ( M )   is of minimal volume with respect to the induced metric [6, 4] ; ξ   is totally geodesic if ξ ( M )   is totally geodesic submanifold in T 1 M   . A complete description of totally geodesic vector fields on 2-dimensional manifolds of constant curvature has been given in [14] . It was proved that only unit sphere (among non-flat space forms) admits totally geodesic unit vector fields (which are non-geodesic ones). In contrast to this case, on a 3-dimensional sphere the Hopf (and therefore, geodesic) unit vector field is a unique one with globally minimal volume [6] . Since totally geodesic submanifold is always minimal, it was natural to suspict that the Hopf vector field is totally geodesic in fact. In addition, for the spheres of greater dimensions the Hopf vector fields stays minimal but become unstable [10, 6] . Nevertheless, something should distinguish the Hopf vector fields. And this something, as we prove here, is their totally geodesic property. Namely, ( Theorem  2.1 ) Let ξ   be a unit geodesic vector field on a unit sphere S n + 1   . Then ξ   is totally geodesic vector field if and only if n = 2 m   and ξ   is a Hopf vector field. 2   Using this result we give another proof of stability or instability of the Hopf vector fields with respect to volume functional (cf. [5] )using a standard formula for the second volume variation from the theory of the submanifolds ( Theorems  3.1 and  3.2 ).
The Hopf vector field on S 2 m + 1   is stable for m = 1   and unstable for m > 1   .
We also find an exact boundaries for the sectional cutvature of this field ( Theorem  4.1  ).
Sectional curvature of ξ ( S 2 m + 1 )   varies between 1 4   and 5 4   . Minimal bound the curvature achieves on ξ   -tangential lift ( 3 ) of ξ   -section and maximal on ξ   -tangential lift of φ   -section in terms of contact metric geometry.

1 Some preliminaries

Let ( M , g )   be ( n + 1 )   -dimensional Riemannian manifold with metric g   .
Denote by < ⋅ , ⋅ >   a scalar product with respect to g   and by ∇   the Levi-Civita connection on M   .
The Sasaki metric on T M   is defined by the following scalar product: if X ~ , Y ~ ∈ T T M   , then
< < X ~ , Y ~ > > = < π * X ~ , π * Y ~ > + < K X ~ , K Y ~ > (1)
where π * : T T M → T M   is a differential of projection π : T M → M   and K : T T M → T M   is the connection map.
Let ξ   be a unit vector field on M   . A vector field X ~ ∈ T T M   is tangent to ξ ( M )   if and only if [12]  X ~ = ( π * X ~ ) h + ( ∇ π * X ~ ξ ) v ,   where ( ⋅ ) h   and ( ⋅ ) v   mean horizontal and vertical lifts of fields into the tangent bundle.
It is well known that ξ v   is a unit normal vector field on T 1 M   . Let X   be tangent to M   , then X t = X v − < X , ξ > ξ v ,   is always tangent to T 1 M   and is called the tangential lift of X   [1] . It is easy to see that < < X t , Y t > > = < X , Y > − < X , ξ > < Y , ξ > .   Introduce, now, a notion of ξ   -tangential and ξ   -normal lifts with respect to given field ξ .   We proceed in the following way.
Introduce a shape operator A ξ   for the field ξ   by A ξ X = − ∇ X ξ ,   where X   is arbitrary vector field on M   .
Define a conjugate shape operator A ξ *   by
< A ξ * X , Y > = < X , A ξ Y > . (2)
Let X , Y   be the vector fields on M   . Define ξ   -tangential lift X ξ τ   and ξ   -normal lift Y ξ ν   of X   and Y   respectively by
X ξ τ = X h − ( A ξ X ) t , Y ξ ν = ( A ξ * Y ) h + Y t . (3)
Then, X ξ τ   is evidently tangent to ξ ( M )   . Moreover,
< < X ξ τ , Y ξ ν > > = < < X h , ( A ξ * Y ) h > > − < < ( A ξ X ) t , Y t > > =
< X , A ξ * Y > − < A ξ X , Y > + < A ξ X , ξ > < Y , ξ > = < A ξ X , Y > − < A ξ X , Y > ≡ 0
and therefore the linear space of all ξ   -normal lifts coincides with the normal space of ξ ( M )   at each point.
To construct a natural tangent and normal orthonormal frames for ξ ( M )   , one can use a singular decomposition of the shape operator A ξ   , based on the following linear algebra result ( [9] , Theorem 7.3.5 and Exercise 7.3.5) which we present here in a slightly modified form.
Theorem 1.1 A matrix A ∈ M m , n   of rank k   may be represented in the form A = F Σ E * ,   where F ∈ M m   and E ∈ M n   are unitary matrices.
The matrix Σ = [ σ i j ] ∈ M m , n   is such that σ i j = 0 , i ≠ j , σ 11 ≥ σ 22 ≥ ⋅ ⋅ ⋅ ≥ σ k k > σ k + 1 , k + 1 = ⋅ ⋅ ⋅ = σ q q = 0 , q = min { m , n } .   The values { σ i i } ≡ { λ i }   are non-negative square roots of eigenvalues of the matrix A A *   and hence are uniquely defined. The columns of the matrix F   are the eigenvectors of the matrix A A *   and the columns of the matrix E   are the eigenvalues of the matrix A * A   .
Moreover, A * f i = λ i e i   and A e i = λ i f i   for i = 1 , . . . , k   . If the matrix A   is real, then F , Σ   and E   can be real.
The columns of the matrices F   and E   are called respectively left and right singular vectors of matrix A   . The values λ i   are called singular values of the matrix A   .
Set A = A ξ   and apply Theorem  1.1 . Since A ξ * ξ = 0   for any unit vector field ξ   , there exist an orthonormal local frames e 0 , e 1 , . . . , e n   and f 0 = ξ , f 1 , . . . , f n   on M   such that A ξ e 0 = 0 , A ξ e α = λ α f α , A ξ * f 0 = 0 , A ξ * f α = λ α e α , α = 1 , . . . , n ,   where λ 1 ≥ λ 2 ≥ . . . λ n ≥ 0   are the real-valued functions.
It is natural to call the functions λ i   ( i = 1 , . . . , n   ) the singular principal curvatures of the field ξ   with respect to chosen singular frame. Remark, that if necessary one may use the signed singular values fixing the directions of the vectors of the singular frame. Setting λ 0 = 0   , we may rewrite the relations on singular frames in a unified form
A ξ e i = λ i f i , A ξ * f i = λ i e i , i = 0 , 1 , . . . , n , λ 0 = 0 , λ 1 , . . . , λ n ≥ 0 . (4)
The following lemma is easy to prove using ( 2 ) and ( 4 ).
Lemma 1.1 [12] At each point of ξ ( M ) ⊂ T M   the orthonormal frames
e ~ i = 1 1 + λ i 2 ( e i h − λ i f i v ) , i = 0 , 1 , . . . , n , n ~ σ | = 1 1 + λ σ 2 ( λ σ e σ h + f σ v ) , σ = 1 , . . . , n (5)
form the orthonormal frames in the tangent space of ξ ( M )   and in the normal space of ξ ( M )   respectively.
Set ( ∇ X A ξ ) Y = ∇ X ( A ξ Y ) − A ξ ∇ X Y .   If we introduce a half tensor of Riemannian curvature as
r ( X , Y ) ξ = ∇ X ∇ Y ξ − ∇ ∇ X Y ξ , (6)
we can easily see that
− ( ∇ X A ξ ) Y = r ( X , Y ) ξ . (7)
Now, we are able to formulate a basic lemma for our considerations.
Lemma 1.2 [12] The components of second fundamental form of ξ ( M ) ⊂ T 1 M   with respect to the frame ( 5 ) are given by
Ω ~ σ | i j = 1 2 Λ σ i j { < r ( e i , e j ) ξ + r ( e j , e i ) ξ , f σ > +
λ σ [ λ j < R ( e σ , e i ) ξ , f j > + λ i < R ( e σ , e j ) ξ , f i > ] } ,
where Λ σ i j = [ ( 1 + λ σ 2 ) ( 1 + λ i 2 ) ( 1 + λ j 2 ) ] − 1 / 2   ( i , j = 0 , 1 , . . . , n ; σ = 1 , . . . , n )   .
Remark. We say that the given unit vector field is holonomic if ξ   is a field of unit normals for a family of hypersurfaces in M   and non-holonomic otherwise.
If the integral trajectories of ξ   are geodesics in M   then ξ   is called geodesic vector field. Evidently, in the case of holonomic geodesic unit vector field, A ξ   becomes a usual shape operator for each hypersurface. In this case A ξ   is self-adjoint (symmetric), i.e.
< A ξ X , Y > = < X , A ξ Y > and thus A ξ * = A ξ   with respect to some orthonormal frame. Let R ( X , Y ) ξ = [ ∇ X , ∇ Y ] ξ − ∇ [ X , Y ] ξ   be a curvature tensor of M   . The non-holonomic shape operator satisfies the non-holonomic Codazzi equation
R ( X , Y ) ξ = ( ∇ Y A ξ ) X − ( ∇ X A ξ ) Y = r ( X , Y ) ξ − r ( Y , X ) ξ . (8)
These facts justify the terminology for the operator A ξ   and the tensor r   .

2 Proof of the main result.

In this section we will prove of the following theorem.
Theorem 2.1 Let ξ   be a unit geodesic vector field on a unit sphere S n + 1   . Then ξ   is totally geodesic vector field if and only if n = 2 m   and ξ   is a Hopf vector field 3   .
Sufficient part of the proof.
The sufficiency is a consequence of more general considerations involving the properties of Killing and Sasakian structure vector fields. Remind that the Hopf vector field is both Killing and Sasakian structure characteristic vector field.
We begin with Killing vector fields. Let M   be ( n + 1 )   -dimensional a Riemannian manifold admitting a Killing vector field ξ   . Then A ξ   is skew-symmetric
< A ξ X , Y > = − < X , A ξ Y > and thus A ξ * = − A ξ (9)
with respect to some orthonormal frame.
The field ξ   is necessarily geodesic and evidently A ξ ξ = 0   . For the singular frames we have e 0 = f 0 = ξ   and e α , f α ∈ ξ ⊥   for all α = 1 , . . . , n   . Moreover, set 2 m = r a n k A ξ   . There exists an orthonormal frame e 0 = ξ , e 1 , . . . , e m , e m + 1 , . . . , e 2 m , e 2 m + 1 , . . . , n   such that
A ξ e α = λ α e m + α , A ξ e m + α = − λ α e α for α = 1 , . . . , m A ξ e 0 = 0 , A ξ e α = 0 for α = 2 m + 1 , . . . , n . (10)
From the definition of the singular frame ( 4 ), we see that one may set
f α = e m + α , f m + α = − e α for α = 1 , . . . , m f 0 = e 0 , f α = e α for α = 2 m + 1 , . . . , n . (11)
A unit vector field ξ   is called normal if < R ( X , Y ) Z , ξ > = 0   and strongly normal if < ( ∇ X A ) Y , Z > = 0   for all X , Y , Z ∈ ξ ⊥   [7] . It is evident that each strongly normal vector field is always normal. If ξ   is a unit Killing vector field the converse is also true.
Now we can essentially simplify the result of Lemma  1.2 .
Lemma 2.1 Let ξ   be a Killing vector field on a Riemannian manifold M n + 1   .
Suppose that r a n k A ξ = 2 m   . Denote e α   ( α = 1 , . . . , n   ) a singular frame for the field ξ   satisfying ( 10 ) and ( 11 ). If ξ   is (strongly) normal then the non-zero components of a second fundamental form of ξ ( M ) ⊂ T 1 M   are given by Ω ~ σ | m + σ 0 = − Ω ~ m + σ | σ 0 = 1 2 K σ ( 1 − K σ ) 1 + K σ σ = 1 , . . . , m ,   where K α   are the sectional curvatures of M n + 1   along the planes ξ ∧ e α   .
Proof. Since ξ   is a Killing vector field then [4] 
A ξ * A ξ X = R ( X , ξ ) ξ (12)
and hence, the right singular vectors for Killing vector field are the Jacobi fields along ξ   geodesics. Let e 1 , . . . , e n   be an orthonormal frame of Jacobi fields. Then R ( e α , ξ ) ξ = K α e α α = 1 , . . . , n ,   where K α   are the sectional curvatures of M   along ξ ∧ e α   planes.
On the other hand, by definition of singular vectors A ξ * A ξ e α = λ α 2 e α .   Thus K α = λ α 2 ≥ 0   .
Any Killing vector field ξ   satisfies
r ( X , Y ) ξ = d e f − ( ∇ X A ξ ) Y = R ( X , ξ ) Y . (13)
Therefore, Ω ~ σ | 00 = < r ( e 0 , e 0 ) ξ , f σ > = 0 .   Next, from the relation λ α < e σ , f α > = < e σ , A ξ e α > = − < A ξ e σ , e α > = − λ σ < f σ , e α >   we find λ α λ σ < e σ , f α > = − λ σ 2 < f σ , e α >   and therefore
Ω ~ σ | α 0 = 1 2 Λ σ α 0 { < r ( e α , e 0 ) ξ + r ( e 0 , e α ) ξ , f σ > + λ σ λ α < R ( e σ , e 0 ) ξ , f α > } =
1 2 Λ σ α 0 { < R ( e α , ξ ) ξ , f σ > + λ σ λ α < R ( e σ , ξ ) ξ , f α > } =
1 2 Λ σ α 0 { λ α 2 < e α , f σ > + λ σ 3 λ α < e σ , f α > } =
1 2 Λ σ α 0 { λ α 2 < e α , f σ > − λ σ 4 < e α , f σ > } =
1 2 λ α 2 − λ σ 4 ( 1 + λ σ 2 ) ( 1 + λ α 2 ) < e α , f σ > .
Taking into account ( 11 ), we get
Ω ~ σ | m + σ 0 = − Ω ~ m + σ | σ 0 = 1 2 K σ ( 1 − K σ ) 1 + K σ σ = 1 , . . . , m . (14)
Since ξ   is strongly normal, r ( e α , e β ) ξ ∼ ξ , R ( e α , e β ) ξ = 0   for all α , β = 1 , . . . , n   . Therefore for the remain components we have Ω ~ σ | α β = 0 .  
One of the most important examples of Killing vector fields is a characteristic vector field of Sasakian structure. Using Lemma  2.1 we can prove the following.
Lemma 2.2 Let M 2 m + 1   be Sasakian manifold and ξ   be a characteristic vector field. Then ξ ( M )   is totally geodesic in T 1 M   .
Proof. Let M 2 m + 1   be an odd dimensional manifold admitting a unit vector field ξ   , linear operator φ   and 1-form η   such that φ 2 X = − X + η ( X ) ξ , φ ξ = 0 , η ( φ X ) = 0 , η ( ξ ) = 1   for any vector field X   on M   . A triple ( φ , ξ , η )   is called an almost contact structure on M   and the manifold is called an almost contact manifold.
If the almost contact manifold is Riemannian with metric g ( ⋅ , ⋅ ) = < ⋅ , ⋅ >   and < φ X , φ Y > = < X , Y > − η ( X ) η ( Y ) , η ( X ) = < X , ξ >   for any vector fields X   and Y   on M   then a tetrad ( φ , ξ , η , g )   is called an almost contact metric structure and the manifold is called an almost contact metric manifold. The almost contact metric structure is said to be contact if d η ( X , Y ) = < φ X , Y >   for all X , Y   .
The vector field ξ   is called characteristic vector field of almost contact metric structure. This field is always geodesic. If, in addition, ξ   is a Killing vector field, then the almost contact metric manifold is called K-contact. If, moreover, the Riemannian curvature tensor R   satisfies
R ( X , Y ) ξ = < ξ , Y > X − < ξ , X > Y . (15)
for all vector fields X , Y   on M   , then the K-contact manifold is called Sasakian. In Sasakian manifold
∇ X ξ = φ X , ( ∇ X φ ) Y = < ξ , Y > X − < X , Y > ξ = R ( X , ξ ) Y . (16)
The property ( 15 ) implies that the sectional curvature of M   along the planes involving characteristic vector field ξ   is equal to 1   and ξ   is a normal unit vector field while ( 16 ) means that ξ   is strongly normal. So, we may apply Lemma  2.1 and immediately get Ω σ | α β ≡ 0   .
Now the sufficient part of the Theorem  2.1 follows immediately from Lemma  2.2 . Remark, that for the 3-dimensional manifold the we can prove a stronger result.
Theorem 2.2 Let ξ   be a unit Killing vector field on 3-dimensional Riemannian manifold M 3   . If ξ ( M 3 )   is totally geodesic in T 1 M 3   then M 3   is either Sasakian and ξ   is the characteristic vector field or M 3 = M 2 × E 1   metrically and ξ   is the unit vector field of Euclidean factor.
Proof. Indeed, ( 14 ) is true for any unit Killing vector field. If ξ ( M 3 )   is totally geodesic, then there exist a Jacobi frame e 1 , e 2   in ξ ⊥   such that the sectional curvatures K ξ ∧ e 1 = K ξ ∧ e 2 = K   and K   satisfies K ( 1 − K ) = 0   . Since e 1   and e 2   are Jacobi fields along ξ   -geodesics, < R ( e 1 , ξ ) ξ , e 2 > = 0   and therefore, for any unit X   in ξ ⊥   it is easy to find that K ξ ∧ X ≡ K   . Thus, all the sectional curvatures along 2-planes involving ξ   is equal either 0 or 1. The first case means that ξ   is a parallel vector field on M 3   .
So, M 3 = M 2 × E 1   metrically and ξ   is a unit vector field of Euclidean factor.
The second case means that M 3   is K-contact which means in dimension 3 that M 3   is Sasakian and ξ   is the characteristic vector field.
The necessary part of the proof. Suppose now that ξ   is geodesic vector field generating a totally geodesic submanifold in T 1 S n + 1   .
Since ξ   is geodesic vector field, r ( e α , ξ ) ξ = ∇ e α ∇ ξ ξ − ∇ ∇ e α ξ ξ = − A ξ 2 e α .   Since the manifold is of constant curvature 1,
r ( ξ , e α ) ξ = R ( ξ , e α ) ξ + r ( e α , ξ ) ξ = − e α − A ξ 2 e α ,
λ α λ σ < R ( e σ , ξ ) ξ , f α > = λ α λ σ < f α , e σ > = < A ξ e α , A ξ * f σ > = < A ξ 2 e α , f σ > .
So we have
Ω ~ σ | α 0 = 1 2 Λ σ α 0 { − 2 < A ξ 2 e α , f σ > − < e α , f σ > + < A ξ 2 e α , f σ > } =
− 1 2 Λ σ α 0 < A ξ 2 e α + e α , f σ > .
If ξ   is a totally geodesic vector field, then < A ξ 2 e α + e α , f σ > = 0   for all α , σ = 1 , . . . , n   . This means that A ξ 2 e α + e α = 0   for all α   and therefore each e α   is the eigenvector for the operator A ξ 2   corresponding to a common eigenvalue − 1   . Since A ξ   is a real operator, its eigenvalues are ± i   . In this case, with respect to some orthonormal frame, the matrix of A ξ | ξ ⊥   takes a box-diagonal form with boxes of type 4   ( 0 ± 1 ∓ 1 0 ) .   This means that A ξ + A ξ * = 0   and therefore, ξ   is a Killing vector field on a sphere. Evidently, the dimension has to be be odd and the field has to be Hopf 's one.
The proof is complete.
Remark. The Hopf vector field is totally geodesic on the unit sphere only.
If S 2 m + 1 ( r )   is a sphere of radius r   , then ξ   is not Sasakian structure vector field but still unit Killing normal vector field satisfying K σ = 1 r   and therefore
Ω ~ σ | 00 ≡ 0 , Ω ~ σ | α β ≡ 0 ,
Ω ~ σ | m + σ 0 = − Ω ~ m + σ | σ 0 = 1 2 ( 1 + 1 r 2 ) 3 / 2 1 r 2 ( 1 − 1 r 2 ) ( α , β , σ = 1 , . . . , m ) .
Thus, if r ≠ 1   then ξ   does not generate a totally geodesic submanifold.
Nevertheless, it stays minimal.

4   With additional condition of covariant normality (see the footnote on page )

3 Stability of the Hopf vector field.

3.1 General formula and preparations.

Let M n   be a submanifold in a Riemannian space R m   . Denote by ∇ ⊥   a covariant derivative in normal bundle connection. Let e i   ( i = 1 , . . . , n   ) be an orthonormal frame on M n   and η   be a normal variation field along M n   . Denote by k i ( η )   an i   -th principal normal curvature of M n   with respect to η   and K ( e i , η )   the sectional curvature of R m   along a plane e i ∧ η   . Then the volume second variation of M n   with respect to η   is given by [3] 
δ 2 V o l ( η ) = ∫ M n { ∑ i = 1 n < ∇ e i ⊥ η , ∇ e i ⊥ η > − | η | 2 ( − ∑ i , j ; i ≠ j k i ( η ) k j ( η ) + ∑ i = 1 n K ( e i , η ) ) } d V . (17)
To apply the formula ( 17 ) to our case we should find the normal bundle connection for the submanifold ξ ( S 2 m + 1 ) ⊂ T 1 S 2 m + 1   .
Lemma 3.1 Let X   and Y   be arbitrary vector fields on S n + 1   ( n = 2 m   ).
Denote Y ⊥ = Y − < ξ , Y > ξ   . Then, with respect to the Hopf unit vector field ξ   , ∇ ¯ X ξ τ Y ξ ν = ( ∇ X Y ⊥ ) ξ ν − 1 2 < ξ , X > ( A ξ Y ⊥ ) ξ ν ,   where ∇ ¯   means a covariant derivative with respect to the Levi-Civita connection on T 1 S n   .
Proof. With respect to the Levi-Civita connection ∇ ¯   on T 1 M   we have [1] 
∇ ¯ X h Y h = ( ∇ X Y ) h − 1 2 ( R ( X , Y ) ξ ) t , ∇ ¯ X h Y t = ( ∇ X Y ) t + 1 2 ( R ( ξ , Y ) X ) h ,
∇ ¯ X t Y h = 1 2 ( R ( ξ , X ) Y ) h , ∇ ¯ X t Y t = − < Y , ξ > X t .
Using this formulas, we get
∇ ¯ X ξ τ Y ξ ν = ∇ ¯ X h ( A ξ * Y ) h + ∇ ¯ X h Y t − ∇ ¯ ( A ξ * X ) t ( A ξ * Y ) h − ∇ ¯ ( A ξ * X ) t Y t =
[ ∇ X ( A ξ Y ) + 1 2 R ( ξ , Y ) X − 1 2 R ( ξ , A ξ X ) A ξ * Y ] h + [ ∇ X Y − 1 2 R ( X , A ξ * Y ) ξ + < Y , ξ > A ξ X ] t .
Since ξ   is a Killing vector field, ( 9 ) and ( 13 ) are fulfilled and we have
∇ ¯ X ξ τ Y ξ ν = [ − ( ∇ X A ξ ) Y − A ξ ( ∇ X Y ) + 1 2 R ( ξ , Y ) X + 1 2 R ( ξ , A ξ X ) A ξ Y ] h +
[ ∇ X Y + 1 2 R ( X , A ξ Y ) ξ + < Y , ξ > A ξ X ] t =
[ R ( X , ξ ) Y + 1 2 R ( ξ , Y ) X + 1 2 R ( ξ , A ξ X ) A ξ Y ] h + [ 1 2 R ( X , A ξ Y ) ξ + < Y , ξ > A ξ X ] t + ( ∇ X Y ) ξ ν .
Since R   is a curvature tensor of a unit sphere and keeping in mind ( 9 ) and ( 12 ), we continue
∇ ¯ X ξ τ Y ξ ν = [ < ξ , Y > X − 1 2 < X , Y > ξ − 1 2 < ξ , X > Y + 1 2 < A ξ X , A ξ Y > ξ ] h +
[ − 1 2 < ξ , X > A ξ Y + < ξ , Y > A ξ X ] t + ( ∇ X Y ) ξ ν =
[ < ξ , Y > X − 1 2 < X , Y > ξ − 1 2 < ξ , X > Y + 1 2 < X , Y > ξ −
1 2 < ξ , X > < ξ , Y > ξ ] h + [ − 1 2 < ξ , X > A ξ Y + < ξ , Y > A ξ X ] t + ( ∇ X Y ) ξ ν =
[ < ξ , Y > ( X − < ξ , X > ξ ) − 1 2 < ξ , X > ( Y − < ξ , Y > ξ ) ] h +
[ − 1 2 < ξ , X > A ξ Y + < ξ , Y > A ξ X ] t + ( ∇ X Y ) ξ ν =
[ < ξ , Y > ( A ξ * A ξ X ) − 1 2 < ξ , X > ( A ξ * A ξ Y ) ] h +
[ − 1 2 < ξ , X > A ξ Y + < ξ , Y > A ξ X ] t + ( ∇ X Y ) ξ ν =
( ∇ X Y ) ξ ν + < ξ , Y > ( A ξ X ) ξ ν − 1 2 < ξ , X > ( A ξ Y ) ξ ν .
Consider, now, ( ∇ X Y ⊥ ) ξ ν   .
( ∇ X Y ⊥ ) ξ ν = [ A ξ * ∇ X Y ⊥ ] h + [ ∇ X Y ⊥ ] t =
[ A ξ * ∇ X ( Y − < ξ , Y > ξ ) ] h + [ ∇ X ( Y − < ξ , Y > ξ ) ] t =
[ A ξ * ∇ X Y + < ξ , Y > A ξ * A ξ X ] h + [ ∇ X Y + < ξ , Y > A ξ X ] t =
( ∇ X Y ) ξ ν + < ξ , Y > ( A ξ X ) ξ ν .
Since A ξ ξ = 0   , we see that A ξ Y = A ξ Y ⊥   . Comparing the results, we get ∇ ¯ X ξ τ Y ξ ν = ( ∇ X Y ⊥ ) ξ ν − 1 2 < ξ , X > ( A ξ Y ⊥ ) ξ ν   what was claimed.
Remark. As we see, the tangent component of the latter derivative is zero, which gives another proof of totally geodesic property of the Hopf vector field.
Since Y ξ ν = ( Y ⊥ ) ξ ν   , we may consider only the vectors from ξ ⊥   in all ξ   -normal lifting operations. So, keep this in mind in what follows.
Corollary 3.1 Let ∇ ¯ ⊥   denote a covariant derivative in normal bundle of ξ ( M )   . If ξ   is a Hopf vector field on the unit sphere S n + 1   ( n = 2 m )   then
∇ ¯ X ξ τ ⊥ Y ξ ν = ( ∇ X Y ) ξ ν − 1 2 < ξ , X > ( A ξ Y ) ξ ν = − < ξ , X > Y h + 2 ( ∇ X Y ) t , (18)
for any X   and Y ∈ ξ ⊥   .
Proof. Indeed, if ξ   is a Hopf vector field for any Y , Z   we have
< < Y ξ ν , Z ξ ν > > = < A ξ * Y , A ξ * Z > + < Y , Z > − < ξ , Z > < ξ , Y > =
2 < Y , Z > − 2 < ξ , Z > < ξ , Y > .
Let Y , Z ∈ ξ ⊥   . Then by Lemma  3.1 and ( 9 )
< < ∇ ¯ X ξ τ Y ξ ν , Z ξ ν > > = < < ( ∇ X Y ) ξ ν − 1 2 < ξ , X > ( A ξ Y ) ξ ν , Z ξ ν > > =
2 < ∇ X Y − 1 2 < ξ , X > A ξ Y , Z > − 2 < ∇ X Y , ξ > < ξ , Z > =
2 < ∇ X Y − < ∇ X Y , ξ > ξ , Z > + < ξ , X > < Y , A ξ Z > =
< < 2 ( ∇ X Y ) t , Z t > > − < < < ξ , X > Y h , − ( A ξ Z ) h > >
< < − < ξ , X > Y h + 2 ( ∇ X Y ) t , ( A ξ * Z ) h + Z t > > =
< < − < ξ , X > Y h + 2 ( ∇ X Y ) t , Z ξ ν > > .
Since Z   is arbitrary, we get the result.

3.2 Second variation of the volume for the Hopf vector field.

Let ξ   be a given unit vector field on M n + 1   . Then the vector field η ~ = ( η ) ξ ν   can be considered as a field of volume variation for the submanifold ξ ( M n + 1 )   , where η   is an arbitrary vector field in ξ ⊥   . For the case of the Hopf vector field on S n + 1   ( n = 2 m )   , we may choose
e ~ i = 1 | | ( e i ) ξ τ | | ( e i ) ξ τ = { e 0 h for i = 0 , 1 2 ( e α h + f α t ) for α = 1 , . . . , n , (19)
where e 0 , e 1 , . . . , e n   , f 0 , f 1 , . . . , f n   form the singular bases for the A ξ   operator, as an orthonormal tangent frame of ξ ( S n + 1 )   . Since ξ ( S n + 1 )   is totally geodesic, the Duschek formula ( 17 ) obtains the form
δ 2 V o l ( η ~ ) = ∫ S n + 1 { ∑ i = 0 n ∥ ∇ e ~ i ⊥ η ~ ∥ − ∥ η ~ ∥ 2 ∑ i = 0 n K ~ ( e ~ i , η ~ ) } d V , (20)
where ∇ ¯ e ~ i ⊥   is given by ( 18 ) and K ~   is given by [2] 
K ~ ( X ~ , Y ~ ) = < R ( X 1 , Y 1 ) Y 1 , X 1 > − 3 4 | R ( X 1 , Y 1 ) ξ | 2 + 1 4 | R ( ξ , Y 2 ) X 1 + R ( ξ , X 2 ) Y 1 | 2 + | X 2 | 2 | Y 2 | 2 − < X 2 , Y 2 > 2 + 3 < R ( X 1 , Y 1 ) Y 2 , X 2 > − < R ( ξ , X 2 ) X 1 , R ( ξ , Y 2 ) Y 1 > + < ( ∇ X 1 R ) ( ξ , Y 2 ) Y 1 , X 1 > + < ( ∇ Y 1 R ) ( ξ , X 2 ) X 1 , Y 1 > , (21)
for an orthonormal basis X ~ = X 1 h + X 2 v   , Y ~ = Y 1 h + Y 2 v   ( X 2 , Y 2 ∈ ξ ⊥   ) of a 2-plane tangent to T 1 S n + 1   .
Lemma 3.2 Let ξ   be the Hopf vector field on S n + 1   ( n = 2 m   ). Let η ∈ ξ ⊥   generates a normal volume variation η ~ = ( η ) ξ ν   for the submanifold ξ ( S n + 1 ) ⊂ T 1 ( S n + 1 )   . Let e 0 = ξ , e 1 , . . . , e n   is a right singular frame for the operator A ξ .   Then, the Duschek formula ( 17 ) can be reduced to the form
δ 2 V o l ( η ~ ) = ∫ S n + 1 { 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n | ∇ e α η | 2 − 2 n − 1 2 | η | 2 } d V , (22)
Proof. Denote ∇ ¯ ⊥   the normal bundle connection for ξ ( S n + 1 )   . Prove, first, that ∑ i = 0 n ∥ ∇ ¯ e ~ i ⊥ η ~ ∥ = 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n | ∇ e α η | 2 − | η | 2 .   Applying ( 18 ) and keeping in mind ( 19 ), we find
∇ ¯ e ~ 0 ⊥ η ~ = − ( η ) h + 2 ( ∇ e 0 η ) t
∇ ¯ e ~ α ⊥ η ~ = 2 ( ∇ e α η ) t
Therefore
∑ i = 0 n ∥ ∇ ¯ e ~ i ⊥ η ~ ∥ = | η | 2 + 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n ( | ∇ e α η | 2 − < ∇ e α η , ξ > 2 ) =
| η | 2 + 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n ( | ∇ e α η | 2 − < η , ∇ e α ξ > 2 ) =
| η | 2 + 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n | ∇ e α η | 2 − 2 ∑ α = 1 n < η , f α > 2 =
| η | 2 + 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n | ∇ e α η | 2 − 2 | η | 2 =
4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n | ∇ e α η | 2 − | η | 2
and the proof is complete.
Now prove that ∥ η ~ ∥ 2 ∑ i = 0 n K ~ ( e ~ i , η ~ ) = 1 2 | η | 2 + ( n − 2 ) | η | 2 .   The sectional curvature of T 1 M n   is given by ( 21 ) in presumption that X ~   and Y ~   are orthonormal. To find ∥ η ~ ∥ 2 K ~ ( e ~ i , η ~ )   we can use ( 21 ) setting Y ~ = η ~   and keeping in mind that η ~   is of arbitrary length.
Now set ζ = A ξ η   and then η ~ = ζ h + η t .   Evidently,
< η , ζ > = 0 , < η , ξ > = < ζ , ξ > = 0 , | η | = | ζ | . (23)
Set X ~ = e ~ 0 = e 0 h   and Y ~ = ζ h + η t   . Then
∥ η ~ ∥ 2 K ~ ( e ~ 0 , η ~ ) = < R ( e 0 , ζ ) ζ , e 0 > − 3 4 | R ( e 0 , ζ ) ξ | 2 + 1 4 | R ( ξ , η ) e 0 | 2 =
| ζ | 2 − 3 4 | ζ | 2 + 1 4 | η | 2 = 1 2 | η | 2
To find K ~ ( e ~ α , η ~ )   for α = 1 , . . . , n   we set X ~ = 1 2 ( e α h + f α t )   and therefore X 1 = 1 2 e α , X 2 = 1 2 f α , Y 1 = ζ , Y 2 = η   in application of ( 21 ). Thus we have
∥ η ~ ∥ 2 K ~ ( e ~ α , η ~ ) = 1 2 < R ( e α , ζ ) ζ , e α > − 3 8 | R ( e α , ζ ) ξ | 2 +
1 8 | R ( ξ , η ) e α + R ( ξ , f α ) ζ | 2 + 1 2 | η | 2 − 1 2 < f α , η > 2 +
3 2 < R ( e α , ζ ) η , f α > − 1 2 < R ( ξ , f α ) e α , R ( ξ , η ) ζ > .
Now set η α = < η , f α >   . Then ζ α = < ζ , e α > = − < η , f α > = − η α .   Keeping in mind ( 23 ) we get
∥ η ~ ∥ 2 K ~ ( e ~ α , η ~ ) = 1 2 | η | 2 − 1 2 ( η α ) 2 + 1 8 | < e α , η > + < f α , ζ > | 2 + 1 2 | η | 2 −
1 2 ( η α ) 2 − 3 2 < e α , η > < f α , ζ > = | η | 2 − ( η α ) 2 − < η , e α > 2 .
It is also easy to see that ∑ α = 1 n ( η α ) 2 = ∑ α = 1 n < η , e α > 2 = | η | 2 .   Hence we have ∥ η ~ ∥ 2 ( K ~ ( e ~ 0 , η ~ ) + ∑ α = 0 n K ~ ( e ~ α , η ~ ) ) = 1 2 | η | 2 + ( n − 2 ) | η | 2 .   Combining the results we get what was claimed.
Proposition 3.1 The Hopf vector field on the unit 3-sphere is stable.
Proof. Since the subspace of right (and left) singular frames for the Hopf vector field coincides with ξ ⊥   , we may choose the other two Hopf vector fields on S 3   as e 1   and e 2   . Then
∇ e 1 e 1 = 0 , ∇ e 1 e 2 = − e 0 , ∇ e 2 e 1 = e 0 , ∇ e 2 e 2 = 0 . (24)
Set η = η 1 e 1 + η 2 e 2   . Set | g r a d η α | 2 = ( e 1 ( η α ) ) 2 + ( e 2 ( η α ) ) 2   ( α = 1 , 2   ).
Then, using ( 24 ) we find ∑ α = 1 2 | ∇ e α η | 2 = ∑ α = 1 n | g r a d η α | 2 + | η | 2 .   Therefore, δ 2 V o l ( η ~ ) = ∫ S n + 1 { 4 | ∇ e 0 η | 2 + 2 ∑ α = 1 n | g r a d η α | 2 + 1 2 | η | 2 } d V > 0   which means that ξ ( S 3 )   is stable.
Proposition 3.2 The Hopf vector field on the unit n   -sphere for n = 2 m + 1 > 3   is unstable.
Proof. Choose the vectors of the singular frame such that e 0 = ξ   while the other e 1 , . . . , e 2 m   are the horizontal lifts of vectors of orthonormal frame q k   of C P m   with respect to the Hopf fibration S 2 m + 1 ⟶ S 1 C P m   . Then ∇ q k q j = 0   and J q 2 k = q 2 k + 1   ( k , j = 1 , . . . , m   ) for complex structure of C P n   (see [8] ). Then along the fiber, i.e. along the integral curves of e 0 = ξ   , the following table of non-zero covariant derivatives can be achieved [8, 10] .
∇ e 0 e 2 k = e 2 k − 1 , ∇ e 0 e 2 k − 1 = − e 2 k ,
∇ e 2 k e 0 = e 2 k − 1 , ∇ e 2 k − 1 e 0 = − e 2 k
∇ e 2 k − 1 e 2 k = e 0 , ∇ e 2 k e 2 k − 1 = − e 0 ,
where k = 1 , . . . , m   .
Let η = η 2 k − 1 e 2 k − 1 + η 2 k e 2 k   be a variation field. Then, using a table of derivatives, we find
∇ e 0 η = ( e 0 ( η 2 k − 1 ) + η 2 k ) e 2 k − 1 + ( e 0 ( η 2 k ) − η 2 k − 1 ) e 2 k − 1
∇ e 2 t − 1 η = e 2 t − 1 ( η 2 k − 1 ) e 2 k − 1 + e 2 t − 1 ( η 2 k ) e 2 k + δ s t η 2 k e 0
∇ e 2 t η = e 2 t ( η 2 k − 1 ) e 2 k − 1 + e 2 t ( η 2 k ) e 2 k − δ k t η 2 k − 1 e 0 ,
where δ k t   is the Kronecker symbol.
Set | g r a d η σ | 2 = ∑ α = 1 n [ e α ( η σ ) ] 2 .   Then ∑ α = 1 n | ∇ e α η | 2 = ∑ σ = 1 n | g r a d η σ | 2 + | η | 2   and | ∇ e 0 η | 2 = ∑ k = 1 m [ ( e 0 ( η 2 k ) − η 2 k − 1 ) 2 + ( e 0 ( η 2 k − 1 ) + η 2 k ) 2 ] .   To prove the instability, we should find variation field η   providing a negative sign for the second volume variation. So, choose η = cos t e 2 k − 1 + sin t e 2 k ,   where t   is an arc-length parameter on e 0   -curves. Then ∇ e 0 η = 0 , g r a d η σ = 0   and for the integrand in the Duscheck formula ( 22 ) we have 5 − 2 n 2 | η | 2 < 0   for n > 2   , which completes the proof.

4 Sectional curvature of the Hopf vector field.

Since the Hopf vector field is totally geodesic in T 1 S n + 1   , the sectional curvature of ξ ( S n + 1 )   is completely defined by the curvature of T 1 S n + 1   along the planes, tangent to ξ ( S n + 1 )   . We can easily find it applying ( 3 ) and ( 21 ) to the Hopf vector field. Remind that the Hopf vector field is a characteristic one for the contact metric structure on the spheres. In contact metric geometry, the sections, containing characteristic vector field, are called ξ   -sections, while the sections of type X ∧ φ X   for X ∈ ξ ⊥   are called φ   -sections. Using the notion of ξ   -tangential lift ( 3 ), for any X ∈ ξ ⊥   we call a 2-plane X ξ τ ∧ ( ξ ) ξ τ   as a ξ   -tangential lift of ξ   -section and a 2-plane X ξ τ ∧ ( ∇ X ξ ) ξ τ ,   as a ξ   -tangential lift of φ   -section.
The following assertion holds.
Theorem 4.1 The sectional curvature of ξ ( S 2 m + 1 )   for the Hopf vector field varies between 1 4   and 5 4 .   The curvature is minimal for ξ   -tangential lift of ξ   -section and maximal for ξ   -tangential lift of φ   -section
The proof is elementary consequence of the Proposition below. It should be mentioned that the curvature of T 1 S n + 1   varies between 0   and 5 / 4   [13] . The theorem clarifies geometrical meaning of the maximal curvature. The minimal curvature is, geometrically, the ξ   -sectional curvature of natural Sasakian structure on T 1 S n + 1   (which,as well known, equals to 1 / 4   after rescaling).
Proposition 4.1 Let ξ   be the Hopf vector field on S n + 1   ( n = 2 m   ). Let X ξ τ   , Y ξ τ   be ξ   -tangential lifts of orthonormal vectors X   and Y   respectively. The sectional curvature K ~ ( X ξ τ , Y ξ τ )   of ξ ( S n + 1 )   along the 2-plane ( X ξ τ   , Y ξ τ   ) is given by K ~ ( X ξ τ , Y ξ τ ) = 1 − 3 4 [ < ξ , X > 2 + < ξ , Y > 2 ] + 3 2 < A ξ X , Y > 2 2 − [ < ξ , X > 2 + < ξ , Y > 2 ]  
Proof. Let X   and Y   be unit mutually orthogonal vector fields on S n + 1   .
Then X ξ τ = X h − ( A ξ X ) t   and Y ξ τ = Y h − ( A ξ Y ) t   form a basis of elementary 2-plane, tangent to ξ ( S n + 1 )   . This basis is not orthonormal, since
∥ X ξ τ ∥ 2 = | X | 2 + | A ξ X | 2 = 2 − < ξ , X > 2 ,
∥ Y ξ τ ∥ 2 = | Y | 2 + | A ξ Y | 2 = 2 − < ξ , Y > 2 ,
< < X ξ τ , Y ξ τ > > = < X , Y > + < A ξ X , A ξ Y > = − < ξ , X > < ξ , Y > .
Therefore, the norm of bivector X ξ τ ∧ Y ξ τ   is
∥ X ξ τ ∧ Y ξ τ ∥ = 4 − 2 < ξ , X > 2 − 2 < ξ , Y > 2 . (25)
Now set X 1 = X , Y 1 = Y , X 2 = A ξ X , Y 2 = A ξ Y   and apply ( 21 ). We have
< R ( X 1 , Y 1 ) Y 1 , X 1 > = 1 ,
| R ( X 1 , Y 1 ) ξ | 2 = | < ξ , Y > X − < ξ , X > Y | 2 = < ξ , Y > 2 + < ξ , Y > 2 ,
| R ( ξ , Y 2 ) X 1 + R ( ξ , X 2 ) Y 1 | 2 = | < A ξ Y , X > ξ − < ξ , X > A ξ Y +
< A ξ X , Y > ξ − < ξ , Y > A ξ X | 2 = < ξ , X > 2 | A ξ Y | 2 + < ξ , Y > 2 | A ξ X | 2 +
2 < ξ , X > < ξ , Y > < A ξ X , A ξ Y > = < ξ , X > 2 ( 1 − < ξ , Y > 2 ) + < ξ , Y > 2 ( 1 − < ξ , X > 2 ) −
2 < ξ , X > 2 < ξ , Y > 2 = < ξ , X > 2 + < ξ , Y > 2 − 4 < ξ , X > 2 < ξ , Y > 2 ,
| X 2 | 2 | Y 2 | 2 − < X 2 , Y 2 > 2 = | A ξ X | 2 | A ξ Y | 2 − < A ξ X , A ξ Y > 2 =
( 1 − < ξ , X > 2 ) ( 1 − < ξ , Y > 2 ) − < ξ , X > 2 < ξ , Y > 2 = 1 − < ξ , X > 2 − < ξ , Y > 2 ,
< R ( X 1 , Y 1 ) Y 2 , X 2 > = < − < A ξ Y , X > Y , A ξ X > = < A ξ X , Y > 2 ,
< R ( ξ , X 2 ) X 1 , R ( ξ , Y 2 ) Y 1 > = < < ξ , X > A ξ X , < ξ , Y > A ξ Y > = − < ξ , X > 2 < ξ , Y > 2 .
Substituting the latter equalities into ( 21 ) and dividing the result by ( 25 ), we get K ~ ( X ξ τ , Y ξ τ ) = 1 − 3 4 [ < ξ , X > 2 + < ξ , Y > 2 ] + 3 2 < A ξ X , Y > 2 2 − [ < ξ , X > 2 + < ξ , Y > 2 ]  
References

  1. Boeckx E.,Vanhecke L. Characteristic reflections on unit tangent sphere bundle. Houston J. Math., 23 (1997), 427 – 448.
  2. Borisenko A., Yampolsky A. The sectional curvature of the Sasaki metric of T 1 M n   . Ukr. Geom. Sb., 30 (1987), 10 – 17. (Enl. transl.: J. Sov. Math., 51 (1990), 5, 2503 – 2508).
  3. Duschek A. Zur geometrischen Variationsrechnung. Math. Z., 40 (1936), 279 – 374.
  4. Gil-Medrano O.,Llinares-Fuster E. Minimal unit vector fields. Tôhoku Math. J., 54 (2002), 71 – 84.
  5. Gil-Medrano O.,Llinares-Fuster E. Second variation of volume and energy of vector fields. Stability of Hopf vector fields. Math. Ann. 320 (2001), 531 – 545.
  6. Gluck H.,Ziller W. On the volume of a unit vector field on the three-sphere, Comm. Math. Helv. 61 (1986), 177 – 192.
  7. González-Dávila J.C., Vanhecke L. Examples of minimal unit vector fields. Ann Global Anal. Geom. 18 (2000), 385 – 404.
  8. Han D.-S., Yim J.-W. Unit vector fields on spheres, which are harmonic maps. Math. Z., 227 (1998), 83 – 92.
  9. Horn R., Jonson Ch. Matrix Analysis. Cambridge Univ. Press, Cambridge, 1986.
  10. Jonson D.L. Volume of flows, Proc. Amer. Math. Soc., 104 (1988), 3, 923 – 931.
  11. Pedersen L.S. Volumes of vector fields on spheres, Trans. Amer. Math. Soc., 336 (1993), 69 – 78.
  12. Yampolsky A. On the mean curvature of a unit vector field, Math. Publ. Debrecen, 60/1-2 (2002), 131 – 155.
  13. Yampolsky A. On the curvature of Sasaki metric of tangent sphere bundle. Ukr. Geom. Sb. 28 (1985), 132 – 145 (Russian) (Eng. transl.: Journ. Sov. Math. 48 (1990), 1, 108 – 117).
  14. Yampolsky A. On the intrinsic geometry of a unit vector field. Comment. Math. Univ. Carolinae 43, 2 (2002), 299 – 317.

Department of Geometry, Faculty of Mechanics and Mathematics, Kharkiv National University, Svobody Sq. 4, 61077, Kharkiv, Ukraine. e-mail: yamp@univer.kharkov.ua

1   Keywords and phrases:

Sasaki metric, vector field, totally geodesic submanifolds.
AMS subject class: Primary 54C40,14E20; Secondary 46E25, 20C20

2   The result is true with one additional condition (see below)