Notes On The Sasaki Metric: October 2018
Notes On The Sasaki Metric: October 2018
Notes On The Sasaki Metric: October 2018
net/publication/328212788
CITATIONS READS
0 117
1 author:
Rui Albuquerque
Universidade de Évora
33 PUBLICATIONS 100 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Rui Albuquerque on 14 October 2018.
Abstract
We survey on the geometry of the tangent bundle of a Riemannian manifold, endowed with the
classical metric established by S. Sasaki 60 years ago. Following the results of Sasaki, we try to write
and deduce them by different means. Questions of vector fields, mainly those arising from the base,
are related as invariants of the classical metric, contact and Hermitian structures. Attention is given
to the natural notion of extension or complete lift of a vector field, from the base to the tangent
manifold. Few results are original, but finally new equations of the mirror map are considered.
Key Words: tensor extension; Killing vector field; Sasaki metric; tangent bundle.
MSC 2010: Primary: 37C10, 53C21, 53D25, Secondary: 53A45, 53C15, 58A32
1 Introduction
These notes try to give an informal up-to-date presentation, together with some generalisations and
observations, of the fundamental results which one finds in the celebrated article of S. Sasaki [19] of 1958.
The article of Sasaki1 studies the differential geometry of tangent bundles of Riemannian manifolds and
has been constantly and consistently the reference of many developments of the theory. It is thus a
modest commemoration of its 60th anniversary that we bring here. Also we feel it may be interesting to
give to light a renewed perspective of the many theorems in the paper, now with some attention on those
results which became less known and may be slightly generalised. If not else, we present both a personal
and a more invariant approach to those important findings.
Noteworthy is the notion of extension of a vector field, introduced in [19]. It is indeed most natural
to the Riemannian geometry of the tangent space. For example, it induces a Sasaki metric Killing vector
field from a Killing vector field on the base. We discover other properties not explored neither on that
or other articles.
It is irrelevant for the present study, but we wish to remember the reader our interest has mainly
in view the construction of gwistor space and of a natural exterior differential system of Riemannian
geometry introduced in [7]. Hence the justification, by the negative, of our study keeping aside the
geometry of the tangent sphere bundles with Sasaki metric. This is very close, yet more complicated.
1
R. Albuquerque 2
One further proves divX = δX ♭ = −tr ∇· X = −trg ∇· X ♭ ·, where ∇ is the Levi-Civita connection.
If X, Y are incompressible, then the Lie bracket [X, Y ] is incompressible. This follows from the today
well-known identity L[X,Y ] = [LX , LY ].
X is called a Killing vector field if LX g = 0; this is easy to see to be equivalent to the vanishing, for
all Z, W ∈ T M, of
(LX g)(Z, W ) = g(∇Z X, W ) + g(Z, ∇W X). (2)
By symmetries, it is immediate that Killing implies incompressible.
A vector field X is called harmonic if dX ♭ = 0 and divX = 0.
From now on we let M denote a C∞ manifold.
The tangent bundle of M is again a manifold, with charts (xi , v i ), i = 1, . . . , m, on U × Rm , where
U ⊂ M is open, (x, U ) is a chart of M and m = dim M . The transition maps of the vector bundle are
induced from a change of charts and their Jacobians2. Almost tautological is the assertion that we have
a manifold
T M = u ∈ Tx M : x ∈ M (3)
with embedded linear fibres Tx M if and only if we have a vector bundle π : T M → M over M with the
same linear fibres. The obvious bundle projection is denoted by π.
Now let us suppose M is endowed with a linear connection, that is, a covariant derivative or local
operator ∇ : Γ(U ; T M ) −→ Γ(U ; T ∗ M ⊗ T M ) on vector fields on U ⊂ M satisfying Leibniz rule. We
have already used the Levi-Civita connection, which is metric, ∇g = 0, and torsion-free, i.e. the tensor
T ∇ (X, Y ) = ∇X Y − ∇Y X − [X, Y ] = 0. For the moment, let us consider any linear connection on M .
In charts, we have the Christoffel ‘symbols’ as the coefficients in ∇i ∂j = Γkij ∂k .
Every X ∈ XM lifts both to a horizontal π ∗ X and a vertical vector field π ⋆ X over the tangent
manifold. Indeed both belonging to XT M . Distinguished in the chart (xi , v i ), we have
Hence, if X = X i ∂i , then
It is very often that one omits the pull-back notation when speaking of functions on T M arising from
M.
2 In the following way: if (x′a , v′a ) is another chart in a domain U ′ , such that U ∩ U ′ 6= ∅, then
∂x′a i ∂xi ∂ 2 xi ∂xi
v′a = v , dxi = dx′a , dvi = v′a dx′b + dv′a
∂xi ∂x′a ∂x ∂x
′a ′b ∂x′a
and
∂ ∂xi ∂ ∂ 2 xj ∂ ∂ ∂xi ∂
= i
+ v′b ′a ′b , = .
∂x ′a ∂x ∂x
′a ∂x ∂x ∂vj ∂v ′a ∂x′a ∂vi
R. Albuquerque 3
The connection induces vector bundle projections, usually denoted (·)h and (·)v , allowing us to identify
H with π ∗ T M via dπ|H . Further, the latter isomorphism yields instantly that a horizontal lift is global
and well-defined. Of course, this may be proved directly.
Notice we use (·)∗ for horizontal lifts and (·)⋆ for vertical lifts.
The connection on M is furthermore pulled-back to both pull-back bundles. In a synthesised notation,
we then obtain a linear connection on the tangent manifold, respecting T T M = H ⊕V , yet always denoted
∇∗ . It is the connection-sum
∇∗ = π ∗ ∇ ⊕ π ⋆ ∇. (9)
The torsion of ∇∗ is immediately found by applying the two projections, also knowing it is a tensor:
∗ ⋆
T ∇ = π ∗ T ∇ ⊕ Rπ ∇
(·, ·)ξ. (10)
Since this is a tensor we may use lifts in the proof. Furthermore we have that
⋆
Rξ (X, Y ) := Rπ ∇
(X, Y )ξ = π ⋆ R∇ (X, Y )ξ (11)
Bπ ∗ X = π ⋆ X, Bπ ⋆ X = 0. (12)
With the mirror map and its formal adjoint we define the Nagano-Sasaki almost-complex structure3
on T M :
J NS = B − B t . (13)
In particular, T M is always an orientable space. An easy computation yields J NS integrable if and only
if T ∇ = 0 and R∇ = 0, a result first proved in [16] (cf. [17]).
Recently, so-called golden structures have been defined and a particularly natural example on tangent
bundles is discovered in [9]. It is easily seen the following map ϕ ∈ End T T M satisfies ϕ2 − ϕ − I = 0:
√
1 5
ϕ= I+ (B + B t ). (14)
2 2
3 This structure is commonly attributed to S. Sasaki, however, in probably more acquainted references, it is due to
The endomorphisms B, J NS and ϕ are all parallel for ∇∗ , the main identity being
∇∗ B = 0. (15)
We may further define a canonical vector field on the tangent manifold. Globally, by
S = Btξ (16)
and locally by S = v a π ∗ ∂a . This is called the geodesic flow or geodesic spray and clearly satisfies
e = π ∗ X + ∇∗S π ⋆ X.
X (19)
Indeed, X e = X i ∂i − X j v a Γb π ⋆ ∂b + v a (∂a X j )π ⋆ ∂j + v a X j Γb π ⋆ ∂b .
ja aj
The geodesic flow S = B t ξ is the vector field over T M whose integral parameterised curves τ are the
velocities γ̇ of geodesics γ of (M, ∇). Since Sτ = (B t ξ)τ = B t τ ∈ H, we can see the former condition is
B t τ = τ̇ if and only if τ = γ̇ for some curve γ in M such that ∇∗τ̇ ξ = γ ∗ ∇γ̇ = 0. In other words, a curve
γ in M is a geodesic if and only if (dπ|H )−1 (γ̇) = γ̈.
Another concept is that of fibre-preserving vector field Z over T M , i.e a vector field such that the
induced transformations preserve the fibres Tx M , for all x. It must verify LZ Y ∈ Γ(V ), for all Y ∈ Γ(V ).
With T ∇ = 0, it is equivalent to Z h depending only of x. This is immediate from (10).
g := π ∗ g ⊕ π ⋆ g. (20)
4 Apparently it was Sasaki who first saw the relevance of such older concept within tangent manifold geometry. In this
article, we conform to the term ‘extended’, rather than the more commonly adopted ‘complete lift’.
R. Albuquerque 5
the vertical directions and preserves the horizontal subspaces for the dual connections on T M and T ∗ M .
Hence
(ℓ∗ λ)u = λℓ(u) ◦ dℓ = g(u, dπ(dℓ(·))) =
(22)
= g(u, dπ(·)) = g(Su , ·) = θu .
Thus ω = dθ = ℓ∗ dλ.
Still making use of D∗ , we find
dξ ♭ = 0. (25)
Since (cf. [7])
∗ θ = ±ξ ♭ ∧ (dθ)m−1 , (26)
we deduce the result that S is incompressible: δθ = 0.
The Levi-Civita connection ∇g of g is said to be found in [19]. Nevertheless, we have a most simple
and useful expression for ∇g : for all X, Y ∈ XT M ,
1
∇gX Y = ∇∗X Y + A(X, Y ) − Rξ (X, Y ) (27)
2
where A(X, Y ) is the symmetric term which makes the connection metric. It is defined by
1
g(A(X, Y ), Z) = (g(Rξ (X, Z), Y ) + g(Rξ (Y, Z), X)). (28)
2
For instance, ∇gπ⋆ X π ⋆ Y = 0, for all X, Y ∈ XM .
It is very important to observe that A takes only horizontal values, because Rξ vanishes on any
vertical direction, and Rξ takes only vertical values.
Henceforth the fibres Tx M are totally geodesic, for all x ∈ M .
The Riemannian curvature tensor of g is first computed by O. Kowalski in [14]. Also Kowalski finds
that g is locally symmetric if and only if R = 0; which, in turn, implies that Rg = 0.
The local holonomy algebra of the Sasaki metric is R-linearly generated by three types of tensors, now
following [5, p.146]. These are curvature skew-adjoint operators Rog (X, Y ) at the zero-section o, given,
with respect to T T M = H ⊕ V , by
∗
π R(X h , Y h ) 0
,
0 π ⋆ R(X h , Y h )
0 − 21 g(π ⋆ R(X h , )Y v , )
1 h v † , (29)
2 g(π R(X , )Y , ) 0
⋆
g(π ⋆ R( , )X v , Y v ) 0
.
0 0
Let us see an example. Suppose M connected has constant sectional curvature c 6= 0. Then the
holonomy group of (T M, g) is SO(2m). Indeed, the maps g(π ⋆ R(X h , )Y v , ) = cg(X h , )g(Y v , ) generate
an m2 dimensional subspace, therefore we deduce the total dimension of the three linearly independent
subspaces is m(m − 1)/2 + m2 + m(m − 1)/2 = m(2m − 1).
The Riemannian manifold (T M, g) is Einstein if and only if R = 0, i.e. M is flat. Indeed, we have for
instance from [3, Proposition 1.3] that
1
Scalg = Scalg − kRξ k2 (30)
4
′ ′ ′ ′
We recall, kRξ k2 = v p v p Rkpij Rk′ p′ i′ j ′ g ii g jj g kk where Rkpij = g(R(∂i , ∂j )∂p , ∂k ).
Many variations of the Sasaki metric have been defined and developed, specially those with weights.
We refer the reader to [1, 3, 4, 8, 13, 15] and the references there-in, concerning the so-called g-natural
metrics.
Recently, the author discovered ciconia metric on any T M 2 which is truly natural to any oriented
Riemannian surface M 2 , cf. [6]. Ciconia metric is not present in the classification of g-natural metrics.
Remark. The above computation is quite useful. For any vector fields X, Y ∈ XM , we have
∇∗π⋆ Y ∇∗S π ∗ X = ∇∗π∗ Y π ∗ X = π ∗ (∇Y X) and ∇∗π⋆ Y ∇∗S π ⋆ X = π ⋆ (∇Y X). (37)
Moreover, we find ∇∗π∗ ∂i ∇∗S π ⋆ X = v b π ⋆ (∇i ∇b X − ∇∇i ∂b X) = v b π ⋆ ∇2 X(∂i , ∂b ). And the same is valid
with π ∗ X in analogous form. E.g. we may write ∇∗Y h ∇∗S π ∗ X = π ∗ ∇2 X(Y h , S).
e ♭ to be closed, that X ♭ is closed, ∇2 X = 0
Just as in (33), we find as an equivalent condition for (X)
and ∇W X ⊥ R∇ (Y, Z)W , for all Y, Z, W ∈ T M . This is quite a strong set of equations, which is deduced
from the three cases analysis, with Z, Y ∈ H ⊕ V , of
e ♭ (Y, Z) = g(∇∗Y X,
d(X) e Z) − g(∇∗Z X,
e Y ) + g(∇∗S π ⋆ X, Rξ (Y, Z)). (38)
Regarding horizontal lifts, we have just seen that π ∗ X is incompressible if and only if X is incom-
pressible. Now we find
and thus (π ∗ X)♭ is closed if and only if X ♭ is closed. In other words, if and only if X is locally a gradient,
because (π ∗ X)♭ = π ∗ X ♭ .
We conclude that π ∗ X is harmonic if and only if X is harmonic.
π⋆ α = π∗ α ◦ B t (40)
is always incompressible.
To prove the last two assertions we recall adapted frames, which here are defined as {e1 , . . . , e2m }
with ei , 1 ≤ i ≤ m, denoting the horizontal lift of an orthonormal frame of M and with ei+m = Bei .
Then π ⋆ α(ei ) = 0 and π ⋆ α(ei+m ) = π ∗ α(ei ) is a fibre constant; hence
Since (R(X, Y )α)Z = −α(R(X, Y )Z), we have that π ⋆ α is harmonic if and only if ∇α = 0.
In fact, π ⋆ X ♭ = π ∗ X ♭ ◦ B t = (π ⋆ X)♭ , so the conclusions for π ⋆ α and π ⋆ X are equivalent.
One also defines the extension to T M of any 1-form α ∈ Ω1M : in a chart (xi , v i ), with α = fi dxi ,
∂fi i
e = vj
α dx + fi dv i = ∇∗S π ∗ α + π ⋆ α. (44)
∂xj
R. Albuquerque 8
Theorem 3.1 (Sasaki). Suppose α is a harmonic 1-form on M . Then α e is harmonic if and only if
α♯ yRic = 0.
Let M be Ricci-flat and α ∈ Ω1M . Then α is harmonic if and only if α
e is harmonic.
Proof. We find that δα = −ϕab g ab where ϕab = ∂a fb − fi Γiab . Due to (42) we have δ α
e = δβ, where
∗ ∗ p k ♯ p jc ∗
β = ∇S π α; we easily compute that β = v ϕpk dx . Since β = v ϕpj g π ∂c is horizontal, we have
applying ∂a g ik = −Γkaj g ij −Γiaj g kj . We may thus write δβ = −tr ∇∗· ϕ̃S where ϕ̃ca = ϕab g bc . However, this
does not help. We must use normal coordinates and finally the hypothesis dα = 0 and δα = −tr ϕ̃ = 0.
Then δβ = −v p ∂a ϕpj g aj and therefore we find
∂ 2 fj ∂fi i ∂Γipj aj
∂a ϕpj g aj = − Γ pj − f i g
∂xa ∂xp ∂xa ∂xa
∂ ∂Γipj aj
= p
(fi Γiaj g aj ) − fi g
∂x ∂xa
∂Γiaj ∂Γipj aj
= fi − g
∂xp ∂xa
i
= fi Rjpa g aj
LX g(Y, Z) = L∇ ∗ ∇ ⋆ ξ ξ
X h π g(Y, Z) + LX v π g(Y, Z) + g(R (X, Z), Y ) + g(R (X, Y ), Z) (46)
Proposition 3.1 (Sasaki). The e· map is a Lie algebra monomorphism XM → XT M . Moreover, the Lie
subalgebra i(M ) of Killing vector fields on M maps into the Lie subalgebra i(T M ) of Killing vector fields
on T M .
Y, Z both horizontal or both vertical, we see easily the equation vanishes on the right hand side. For
Y horizontal and Z vertical, we find g(∇∗Y ∇∗S π ⋆ X, Z) + g(Rξ X,Y , Z). In other words, by the remark in
page 7, for all Y, Z ∈ XM ,
∇2 X(Y, Z) + R(X, Y )Z = 0, (51)
which is a well-known identity satisfied by X Killing, cf. [12, p.235].
According to [12, 17], X satisfying (51) is called an infinitesimal affine transformation.
At this point follows a straightforward relation between infinitesimal transformations on M and their
extensions being almost-analytic. A vector field Z is almost-analytic if LZ J NS = 0, of course, presently,
for the Nagano-Sasaki almost-complex structure over T M . We can compute via D∗ :
LXe J NS (π ∗ Y ) = −π ∗ R(X, Y )S − ∇∗π∗ Y ∇∗S π ∗ X. (52)
It is, indeed, enough to consider horizontal lifts π ∗ Y . The equation is equivalent to (51)
Hence the conclusion of [17, Theorem 2]: X extends to an almost-analytic vector field if and only if
X is an infinitesimal affine transformation.
Let us also consider infinitesimal contact-transformations, i.e. those which leave θ invariant. The
corresponding vector fields Z are also said to be strictly-contact ; of course, given by LZ θ = 0.
R. Albuquerque 10
[17, Theorem 4] asserts that Killing extensions are strictly-contact and vice-versa. Let us see in
general. Recall θ = Syg and S is horizontal. Then, for all Y ∈ T T M ,
Thus horizontal lifts are strictly-contact if and only if the base vector field is parallel. Whereas vertical
lifts are never strictly-contact (though they may be symplectic).
Now
(LXe θ)(Y ) = g(B t ∇∗S π ⋆ X, Y ) + g(S, ∇∗Y π ∗ X)
(54)
= g(∇∗S π ∗ X, Y ) + g(S, ∇∗Y π ∗ X) = LX g(dπS, dπY ).
(i) X is Killing;
(ii) e
X is Killing;
(iii) e
X is strictly-contact;
(iv) e
X is symplectic.
Condition (iv) seems to be new and will be proved later; we may recall immediately LX dθ = dLX θ.
We also find studies of Killing vector fields on T M in [20].
S. Tanno discovered that if P ∈ ∧2 T ∗ M is a skew-symmetric and parallel (1,1)-tensor on M , then the
vertical vector field π ⋆ P ξ is Killing. This is trivial from (46). Indeed, we have ∇∗Y π ⋆ P ξ = (π ⋆ P )Y v and
thus, relaxing the notation ‘π ⋆ ’,
[5, Theorem 1.5] yields the following system for a geodesic in T M , also found by Sasaki:
(
ẍp + ẋi ẋj Γpij + ẋi z b v j Rbjiq g qp = 0
(57)
ż a + ẋi ż b Γaib = 0.
Since geodesics xt in M induce parallel velocities vt in T M along xt , and therefore all z b = 0, every
geodesic lifts naturally to a geodesic for the Sasaki metric.
Reciprocally, a lift of a curve in M is a geodesic if the curve is a geodesic of M . This is due to the
second equation and the last paragraph of Section 2.
Projections by π of geodesics yield curves in M called submarine ‘geodesics’. Clearly, these are
geodesics if M is flat.
The topology of the tangent bundle does not prevent the geodesics from not being defined for all
t ∈ R. In other words, we are sincerely convinced (T M, g) is a complete Riemannian manifold, as long
as M is complete — cf. discussion and similar metrics referred in [5].
Looking now to vector fields as embeddings X : M → T M , we may ask when is the embedded
X
submanifold X(M ) := M totally geodesic inside T M .
X
We have clearly dX(Z) = π ∗ Z + π ⋆ (∇Z X) = X∗ Z with the lifts just to TXx M ⊂ TXx (T M ), for each
x ∈ M , although the lifts naturally extend to the whole space. Knowing that for X∗ Z1 , X∗ Z2 , we must
have ∇gX∗ Z1 X∗ Z2 of the same kind of the former, we obtain the following seemingly strange equation:
X
the submanifold M is totally geodesic if and only if, for all Z1 , Z2 ∈ T M ,
1
∇∇Z1 Z2 + 21 R(X,∇Z2 X)Z1 + 21 R(X,∇Z1 X)Z2 X = ∇Z1 ∇Z2 X − R(Z1 , Z2 )X. (58)
2
This equation is also in [2]6 . We remark it is tensorial in Z1 , Z2 . Knowing the skew- and symmetric
parts apparently leads to nowhere. In case we have R(·, ·)X = 0, as in [5, Proposition 1.4], then also
R(X, ·)· = 0 by symmetry of the Riemannian curvature tensor; and the strange equation becomes
∇2 X = 0. (59)
Proof. If g(X, X) is a constant, then g(∇X, X) = 0. We contract equation (58) with g(·, X) to obtain
g(∇Z1 ∇Z2 X, X) = 0, for every Z1 , Z2 . Then g(∇Z1 X, ∇Z1 X) = 0. The reciprocal is trivial.
6 In this reference there is a further degree of complexity: the vector fields are considered when restricted to submanifolds
we have the following formula from [4, Proposition 2.2], where ∂ϕ is a function, ∂ϕu = dϕπ(u) (u), and
B, θ, π ∗ are with respect to g:
′ Y (t)
dh(Y ) = Y h + ĥ ξ + Y v + (∂ϕ)BY − θ(Y )π ⋆ grad ϕ . (60)
t
Finally [4, Theorem 2.1] clarifies: the map h is a homothety, i.e. h∗ g′ = ψg for some function ψ
defined on T M , if and only if the given functions t and ϕ on M are constants and satisfy e2ϕ = t2 := ψ.
In this case, h = 1.
Proof. i) follows directly from the equation deduced in (59) and ii) from Walczak’s Theorem.
We also remark that the energy functional has been extended by O. Gil-Medrano to the space of all
immersions M −→ T M , finding the same parallel condition for critical points under compactness. And
has been restricted to the space X1M of unit vector fields, where harmonic maps in the previous sense are
no longer critical for E, cf. [10].
The whole theory relates to the study of the volume functional X 7→ 12 volX ∗ g (M ).
R. Albuquerque 13
LX ω(Y, Z) = d(Xyω)(Y, Z)
(64)
= −g(∇∗Y J NS X, Z) + g(∇∗Z J NS X, Y ) − g(J NS X, Rξ (Y, Z)).
Hence, for X ∈ XM , the horizontal lift is symplectic if and only if ∇∗ X = 0. And the vertical lift is
symplectic if and only if X ♭ is closed.
Now we can prove X is Killing if and only if Xe is symplectic.
This proves symplectic implies Killing. The converse comes from LX dθ = dLX θ. Or we can check that
the equation above for the symplectic extension vanishes for Y, Z vertical, clearly, and for Y, Z horizontal
also. In the latter case, we must recur to the infinitesimal affine transformation equation, which holds by
hypothesis, and then to Bianchi identity.
LXe B = 0. (66)
7 There are many natural non-trivial endomorphisms of T T M , such as B, B t , (·)h , (·)v , so we could treat their linear
combinations; but we have some special interest in discussing the present simple notion.
R. Albuquerque 14
Proposition 3.3. Every λ-mirror vector field X of T Rm decomposes uniquely as the sum of three fields:
a vertical lift vector field; an extension vector field; and the horizontal lift λπ ∗ P where P is the position
vector field P = xi ∂i on Rm .
Proof. We use coordinates (xi , v i ) and let X = (A1 , A2 ) = Ai1 ∂i + Aj2 ∂vj be tangent to the tangent
manifold, with obvious decomposition. In particular, BX = Ai1 ∂vi , B∂i = ∂vi , B∂vj = 0. For ∂vj ,
we get LX B(∂vj ) = ∂vj Ai1 ∂vi equal to λB∂vj = 0 if and only if ∂vj Ai1 = 0, for all 1 ≤ i, j ≤ m. For
∂i , we get LX B(∂i ) = −∂vi (A1 , A2 ) + ∂i Aj1 ∂vj . This equals λB∂i = λ∂vi if and only if ∂vi Aj1 = 0 and
−∂vi Aj2 + ∂i Aj1 = λδij . Now, differentiating this last equation under ∂vk , yields ∂v2k ,vi Aj2 = 0 and so
we may write Aj2 = Aj20 + Aj2i v i with Aj20 , Aj2i functions of x only. Therefore, ∂i Aj1 = λδij + Aj2i . In
∂Aj
sum, Aj1 is a function of x only and Aj2 = Aj20 + ( ∂xi1 − λδij )v i . Clearly, Aj20 ∂vj is the desired vertical
∂Aj1 i
lift, Aj1 ∂j + ∂xi v ∂v j is an extension vector field and λxi ∂i is the particular solution of the λ-mirror
equation.
Now let us see the case of the adjoint-mirror map, in the general picture. First, it is easy to see, for
all X, Y, ∈ XT M ,
Proof. The horizontal and vertical parts of LX B = 0 are ∇∗Y v X h = 0 and −∇∗BY X v + B∇∗Y X h = 0.
Using Y vertical, the latter implies the former equation. Now let us take Y vertical in LX B t = 0,
knowing (67), and read the horizontal part. It says −∇∗B t Y v X h + B t ∇∗Y v X v = 0. Applying B on the
left and calling B t Y = Y h , we obtain −B∇∗Y h X h + ∇∗BY X v = 0, which is the unique defining equation
of 0-mirror vector field.
Proposition 3.5. For X ∈ XM , both the horizontal lift or the vertical lift are 0-adjoint-mirror if and
only if X is parallel.
The extension X e is 0-adjoint-mirror if and only if X is an infinitesimal affine transformation.
Proof. We just see the less trivial case of X. e For Y horizontal, the vanishing of (67) becomes the
vanishing of ∇Y ∇S π X + π R(π X, Y )S = π ∗ ∇2 X(Y, S) + π ∗ R(X, Y )S. For Y vertical, we find
∗ ∗ ∗ ∗ ∗
−∇∗B t Y π ∗ X − ∇∗B t Y ∇∗S π ⋆ X + ∇∗Y ∇∗S π ∗ X − Rξ (X, B t Y ) = −π ⋆ ∇2 X(B t Y, S) − Rξ (X, B t Y ), which re-
peats the infinitesimal affine transformation equation.
e being 0-adjoint-
Of course, knowing (66), we have just seen the proof that an extended vector field X
mirror, almost-analytic or an infinitesimal affine transformation is the same.
Proposition 3.6. A vector field X = Ai1 ∂i + Aj2 ∂vj tangent to T Rm is 0-adjoint-mirror if and only if
∂i Aj2 = 0, ∂i Aj1 = ∂vi Aj2 , for all 1 ≤ i, j ≤ m.
An extension vector field X e = Aj ∂j + v k ∂k Ai ∂vi is 0-adjoint-mirror if and only if ∂ 2 Ai = 0, for all
j,k
1 ≤ i, j, k ≤ m, this is, if and only if Ai (x) = ai0 + aij xj with constants ai0 , aij , for all 1 ≤ i ≤ m.
References
[1] M.T.K. Abbassi, G. Calvaruso and D. Perrone, Harmonic sections of tangent bundles equipped with
Riemannian g-natural metrics, Quart. J. Math., Vol. 62, I. 2 (2011), 259–288.
[2] M.T.K. Abbassi and A. Yampolsky, Transverse totally geodesic submanifolds of the tangent bundle,
Publ. Math. Debrecen 64, 1-2 (2004), 129–154.
R. Albuquerque 15
[3] R. Albuquerque, Curvatures of weighted metrics on tangent sphere bundles, Riv. Mat. Univ. Parma,
Vol. 2, (2) (2011), 299–313.
[4] R. Albuquerque, Homotheties and topology of tangent sphere bundles, Jour. of Geometry, 105, 2
(2014), 327–342.
[5] R. Albuquerque, On vector bundle manifolds with spherically symmetric metrics, Ann. Global Anal.
Geom., vol. 51, Issue 2 (2017), 129–154.
[6] R. Albuquerque, The ciconia metric on the tangent bundle of an almost-Hermitian manifold, August
2016, https://arxiv.org/abs/1612.07596
[8] G. Calvaruso, Naturally Harmonic Vector Fields, Note Mat. 1 (2008), suppl. n. 1, 107–130.
[9] N. Cengiz, A. Gezer and A.A. Salimov, On integrability of Golden Riemannian structures, Turkish
J. of Math., 37 (2013), 693–703.
[10] O. Gil-Medrano, Relationship between volume and energy of vector fields, Diff. Geom. Appl. 15
(2001), 137–152.
[11] S. Hedayatian and B. Bidabad, Conformal vector fields on tangent bundle of a Riemannian manifold,
Iranian J. of Science & Tech., Transaction A, Vol. 29, N. A3 (2005).
[12] S. Kobayashi and K. Nomizu, Foundations of Differential Geometry, Wiley Clas. Lib., Vol. 1, 1963.
[13] I. Kolář, P.W. Michor and J. Slovák, Natural operations in differential geometry, Springer-Verlag,
Berlin 1993.
[14] O. Kowalski, Curvature of the induced Riemannian metric of the tangent bundle of a Riemannian
manifold, J. Reine Angew. Math. 250 (1971), 124–129.
[17] M. Okumura and S. Tachibana, On the almost-complex structure of tangent bundles of Riemannian
Spaces, Tôhoku Math. J. 14 (1962), 152–161.
[18] V. Oproiu, Some new geometric structures on the tangent bundle, Publ. Math. Debrecen 55 (1999),
3-4, 261–281.
[19] S. Sasaki, On the differential geometry of tangent bundles of Riemannian manifolds, Tôhoku Math.
J. 10 (1958), 338–354.
[20] S. Tanno, Killing vectors and geodesic flow vectors on tangent bundles, J. reine angew. Math., 282
(1976), 162–171.
[21] P.G. Walczak, On Totally Geodesic Submanifolds of Tangent Bundle with Sasaki Metric, Bull. Acad.
Pol. Sci, ser. Sci. Math. 28, no.3-4 (1980), 161–165.
R. Albuquerque | rpa@uevora.pt
Centro de Investigação em Matemática e Aplicações
Rua Romão Ramalho, 59, 671-7000 Évora, Portugal
The research leading to these results has received funding from Fundação para a Ciência e a Tecnologia.