Novi Sad J. Math.
Vol. 32, No. 2, 2002, 141-154
LAGRANGE GEOMETRY VIA COMPLEX LAGRANGE GEOMETRY
Gheorghe Munteanu1
Abstract. Asking that the metric of a complex Finsler space should be strong convex, Abate and Patrizio ([1]) associate to the real tangent bundle a real Finsler metric for which they analyze the relation between Cartan (real) connection of the obtained space and the real image of Chern-Finsler complex connection.
Following the same ideas, in the present paper we shall deal with the more general case of a complex Lagrange space (M, L).
As distinct from these authors, we shall associate to the Hermitian metric gi¯j(z, η) of a complex Lagrangian L its real representation Rgab (x, y).The obtained real space (M,Rgab) is a generalized Lagrange space ([10]). Furthermore, the possibility of its reduction to one real Lagrange space, in particular the Finsler one, is studied.
A comparative analysis of the elements of Lagrange geometry ([10]):
nonlinear connection,N−linear connection, metric canonical connection, and so on, and their corresponding real image from the complex Lagrange geometry ([11]) is made.
AMS Mathematics Subject Classification (2000): 53B40, 53C60 Key words and phrases: complex Lagrange geometry
1. Introduction
The study of complex Lagrange geometry was initiated by us starting with the paper [11].
A complex Lagrange space is the pair (M, L), whereM is a complex mani- fold andL(z, η) is a real Lagrangian differentiable function on the holomorphic bundleT0M, which determines a nondegenerate metricgi¯j=∂2L/∂ηi∂¯ηj.
This geometry generalizes that of the known complex Finsler space ([1],[2],[5], [6],[7],[13],[14]), where, in addition, the homogeneity condition of complex La- grangian in respect toη is required.
In this paper, a complex Lagrange space determines a real structure of Her- mitian manifold on the tangent real bundle. Everywhere, the indicesi, j, k, ...
run in the interval 1, n, and a, b, c, ... run in 1,2n. We shall assume that the
1Transilvania University, Faculty of Science, 2200 Brasov, Romania, e-mail:[email protected]
reader is familiar with the geometry ofT0M, the holomorphic tangent bundle, and with the Lagrange geometry ([10]).
LetM be a complex manifold,dimCM =n, (U, zk) the local coordinates in a local chart,zk=xk+ixn+k. M is also a real manifold,dimRM = 2n,(U, xa) is a real chart, and is endowed with the complex structureJR,
R
J2=−I, acting onTR,xM byJR(∂x∂a) =∂x∂n+k,JR(∂x∂n+k) =−∂x∂k.
Let us consider the well-known Poincaire operators:
∂
∂zk = 1 2( ∂
∂xk −i ∂
∂xn+k); ∂
∂¯zk = 1 2( ∂
∂xk +i ∂
∂xn+k) (1.1)
from which clearly results that:
∂
∂xk = ∂
∂zk + ∂
∂¯zk; ∂
∂xn+k =i( ∂
∂zk − ∂
∂z¯k) (1.2)
The complex structure JR is extended to the complexification TCM of the tangent bundle, obtaining the complex structureJ(X+iY) =JR(X)+iJR(Y), J2=−I, behaving on Poincare operators as follows: J(∂z∂k) =i∂z∂k;J(∂¯∂zk) =
−i∂∂z¯k. The eigenspaces of J determines two subbundles of TCM denoted by T0M, the (1,0)−type vectors, and respectively T00M =T0M, the (0,1)−type vectors andTCM =T0M ⊕T00M.
The bundleT0Mis holomorphic and as a manifold it is the geometric support of the complex Lagrange geometry.
The bundle T0M is isomorphic with the real tangent bundle TRM by the map that acts on the corresponding tangent space as follows ([1]):
R◦:X →XR=X+X (1.3)
with the inverse:
C◦:X →XC= 1
2(X−iJX) (1.4)
Locally, ifX =ηk ∂∂zk, withηk =yk+iyn+k, thenXR=yk ∂∂xk+yn+k∂x∂n+k, and conversely, ifX=ya ∂∂xa, thenXC= (yk+iyn+k)∂z∂k.
Let us consider π : T0M → M the holomorphic bundle, u = (zk, ηk) ∈ T0M and p: TRM →M the tangent real bundle, u= (xa, ya) ∈TRM. Now, taking T0M as a base manifold, arguing as before, we obtain: TC(T0M) = T0(T0M)⊕T00(T0M). The bundle πT : T0(T0M) → T0M is holomorphic and Ker πT =V(T0M) is called the vertical bundle, a local base in Vu(T0M) being {∂η∂k }. Through conjugation, a local base {∂∂η¯k} inVu(T0M) is obtained. Let us denote byVC(T0M) =V(T0M)⊕V(T0M) the vertical complexified bundle.
Analogously, we can consider the real vertical bundleV(TRM) =Ker pT, where pT : TR(TRM)→ TRM is the tangent map. A local base in Vu(TRM)
is indeed{∂y∂a}and because the vertical bundle is isomorphic with the tangent bundleTRM, we have:
∂
∂ηk = 1 2( ∂
∂yk −i ∂
∂yn+k); ∂
∂η¯k = 1 2( ∂
∂yk +i ∂
∂yn+k) (1.5)
and conversely,
∂
∂yk = ( ∂
∂ηk)R= ∂
∂ηk + ∂
∂η¯k; ∂
∂yn+k = (i ∂
∂ηk)R=i( ∂
∂ηk − ∂
∂η¯k) (1.6)
Therefore, ifV =Vk ∂∂ηk is a vertical complex field thenVR=Ua ∂∂ya, with Uk=ReVk and Un+k =ImVk, is a real vertical field, and conversely. Hence, (VC(T0M))R = V(TRM). We denote the same here by R◦ the isomorphism of passing to real onT0M.
2. The induced real nonlinear connection
As is known, in the study of tangent bundles it is very useful to use the notion of a nonlinear connection that determines the adapted base in which the study is ”linearized”: many of the computations are made similarly as on the base manifoldM.
A nonlinear connection can be given by a splitting in an exact sequence that determines a supplementary subbundle toV(T0M) inT0(T0M),i.e. T0(T0M) = H(T0M)⊕V(T0M), called the horizontal subbundle. This determines the distri- butionN :u= (zk, ηk)→Hu(T0M),called the complex nonlinear connection, shortly (c.n.c.). A local base onHu(T0M) is {δzδj = ∂z∂j −Njk∂η∂k}, whereNjk are the coefficients of (c.n.c.) and they are transforms at the local change of charts after the rule:
Nk0i∂z0k
∂zj =∂z0i
∂zkNjk− ∂2z0i
∂zj∂zkηk (2.1)
and then the base {δzδi}, called the adapted base of Nij (c.n.c.), satisfies the following rule of transformation:
δ
δzi =∂z0j
∂zi δ δz0j (2.2)
Through conjugation is obtained an adapted base {δzδk, ∂η∂k, δ¯δzk, ∂∂η¯k} onTC(T0M) , shortly denoted by{δk , ∂k , δ¯k , ∂¯k}. The dual adapted base is denoted by{dzk, δηk , d¯zk, δη¯k}.
From (2.2) it results that there exists an isomorphism ([1]) Cθ:VC(T0M)→ HC(T0M), locally given byCθ (∂η∂i) =δzδi andCθ(∂∂η¯i) =δδz¯i,whereHC(T0M) = H(T0M)⊕H(T0M) is the complexified horizontal bundle.
As a complex manifoldT0M has the natural complex structure, still denoted byJ, and locally given by:
J(∂/∂zk) =i∂/∂zk; J(∂/∂ηk) =i∂/∂ηk ; J(∂/zk) =−i∂/∂zk J(∂/∂ηk) =−i∂/∂ηk.
Since,J(δk) =iδkandJ(δ¯k) =−iδ¯kwe deduce thatJ(H(T0M)) =iH(T0M) andJ(H(T0M)) =−iH(T0M).
The same reasonings can be made onTR(TRM).A real nonlinear connection, shortly (r.n.c.), is given by the splittingTR(TRM) =H(TRM)⊕V(TRM).
TheH(TRM) bundle is for the moment not unique, being only supplemen- tary toV(TRM). We shall fix H(TRM) acting analogously to [1]:
We see thatTRM andT0M bundles are isomorphic byC◦ andR◦. The same isomorphism is betweenTR(TRM) andT0(T0M).On the other hand, the com- plex horizontal lift, locally expressed bylhC(∂z∂k) = δzδk determines an isomor- phism between T0M and H(T0M). Then the image of map lhR = R◦ ·lCh· C◦ : TRM → TR(TRM) defines a real horizontal lift. Let us consider now the local base δxδa =lhR(∂x∂a) in H(TRM) =lhR(TRM) which determines in turn an (r.n.c.).
As it is known that the local expression of a real horizontal lift is ([10]):
δ
δxa = ∂x∂a−
R
Nab ∂y∂b, where
R
Nab are the coefficients of (r.n.c.). Taking into account the local expression of a complex vertical field, we can deduce that:
R
Nkh=ReNkh ;
R
Nkn+h=ImNkh (2.3)
and therefore : Nkh=NRkh+i
R
Nkn+h andδk =∂z∂k−Nkh∂h.
Note that the map R∗ =Rθ · R◦ · θC−1 : HC(T0M) → H(TRM) is an isomor- phism, whereRθ :V(TRM)→H(TRM) is the corresponding real isomorphism toCθ .Because the horizontal bundleHC(T0M) isJ invariant, applying the op- eratorR∗ it follows that H(TRM) is JRinvariant. Hence, δxn+kδ corresponds to iδzδk and in consequenceNn+kRa corresponds to iNkh. So, we deduce that:
R
Nn+kh =−ImNkh ;
R
Nn+kn+h=ReNkh (2.4)
Thus we obtain an (r.n.c.) ,NRba, onTRM, determined by the given (c.n.c.) NkhonT0M, whose coefficients are in fact the real representation of the complex matrixNkh.
Now, taking into account the action of theR∗ operator on the adapted base {δk, ∂k}, we get in addition to (1.6) that:
δ
δxk = ( δ
δzk)R ; δ
δxn+k = (i δ δzk)R (2.5)
and the inverse throughC∗ =Cθ ·R◦ ·
R
θ−1.
For the next computation it will be useful to have the following consequences of (1.6) and (2.5):
( ∂
∂η¯k)R = ∂
∂yk ; ( δ
δ¯zk)R= δ δxk and δ
δxk = δ δzk + δ
δ¯zk ; δ
δxn+k =i( δ δzk − δ
δz¯k)
Further, let us consider (M, L) a complex Lagrange space ([11]), where L:T0M →R is a Lagrangian function such thatgj¯k =∂2L/∂ηj∂η¯k is a non- degenerate metric onT0M.At each pointu= (zk, ηk)∈T0M, L(u) =L(zk, ηk) is a differentiable function. Because zk =xk+ixn+k and ηk =yk+iyn+k, it follows thatL(u) =LR(xa, ya) is a real differentiable function.
Let as note that, in general,Rgab= 12∂2LR/∂ya∂yb might be degenerate and hence the pair (M, LR) is not always a real Lagrange space. In the special case when (gj¯k) determines a nondegenerate matrix (Rgab), (A-P) ([1]) calls the metric gjk¯ as beingstrongly convex.
Moreover, let us note that if L(z, λη) = |λ|2L(z, η), then LR(x, λy) =
|λ|2LR(x, y) and conversely. So, if (M, L) is a strong convex Finsler complex space then (M, LR) is a Finsler real one.
From now on we shall act in a manner different from [1]. We shall consider the real metric structure determined by the real representation ofgjk¯.
Proposition 2.1. Letgj¯k(z, η)be the Hermitian metric of a complex Lagrange space(M, L). Then the pair(M,gRab(x, y)), where :
Rgjk=Re gj¯k= 1
2(gj¯k+gk¯j) ; Rgn+jk=−Im gj¯k
(2.6)
Rgjn+k=Im gjk¯= −i
2 (gj¯k−gk¯j) ; Rgn+jn+k=Re gjk¯
determines a (real) generalized Lagrange space([10]).
For proof it suffices to remark that Rgab=Rgbaanddet µR
gab
¶ 6= 0
Thanks to (2.6) and gj¯kg¯kl = δlj it results that Rgab, the inverse of Rgab, is the real representation of gkl¯, i.e. Rgjk=Re g¯jk; Rgjn+k=Im g¯jk; Rgn+jn+k= Re g¯jk; and Rgn+jk=−Im g¯jk.
Now, considering a fixed (c.n.c.)Nkh and {dzk , δηk , d¯zk , δη¯k} the dual adapted base determined by it, then ([11]):
G=gijdzi⊗dzj+gijδηi⊗δηj (2.7)
gives a Hermitian metric onT0M with respect to the complex structureJ and both to the almost Hermitian structure JN, locally given by JN(δk) = ∂k; JN(∂k) =−δk; JN(δ¯k) =∂¯k; JN(∂¯k) =−δk¯ and globally defined.
Replacinggi¯j =Re gi¯j+iIm gi¯j;dzj=dxj+idxn+j andδηj=δyj+iδyn+j in (2.7), it results that:
Proposition 2.2. The structure
GR=Re G=Rgabdxa⊗dxb+Rgabδya⊗δyb (2.8)
is a Hermitian metric on TRM with respect to the complex structure JR, and an almost Hermitian metric with respect toJR
N structure, (JR
N)2=−I,locally given byJR
N(δxδa) = ∂y∂a andJR
N(∂y∂a) =−δxδa. The integrability ofJN andJR
N structures depends only on the vanishing of torsion of (c.n.c.) and respectively (r.n.c.).
Let as note that
R˜
G=Im Galso defines a metric structure onTRM.
From the computation gj¯k = ∂2
R
∂ηj∂Lη¯k = 14(∂y∂k +i∂y∂n+k)(∂
R
∂yLj −i ∂
R
∂yn+jL ) =
1
4(∂y∂j2∂yRLk +∂yn+j∂2∂yLRn+k) +4i(∂yj∂∂y2RLn+k −∂yn+j∂2RL∂yk) we deduce just the real rep- resentation of the matrixgj¯k:
Rgjk=1 4( ∂2LR
∂yj∂yk + ∂2RL
∂yn+j∂yn+k) ; Rgjn+k= 1
4( ∂2LR
∂yj∂yn+k − ∂2LR
∂yn+j∂yk) (2.9)
Let us remark that, in general, the tensorCabc= 12{∂
Rgbc
∂ya +∂
Rgac
∂yb −∂
Rgab
∂yc} is not totally symmetric and therefore the generalized Lagrange space (M,Rgab) is not always reducible to a Largange space ([10]). Moreover, the space is neither weakly regular, hence the known procedures of Lagrange (real) geometry to obtain a (r.n.c.) cannot be applied, remaining in principle the method described above.
The question is, however, when the generalized Lagrange space is a Lagrange one, particularly a Finsler space. This means that the tensorhab= 12 ∂2
R
∂ya∂yLb is nondegenerated. In particular, ifLis a complex Finsler metric thenRLbecomes a real Finsler metric. A sufficient condition is whenhab=Rgab, which in view of (2.9) is equivalent to:
∂2RL
∂yj∂yk = ∂2RL
∂yn+j∂yn+k ; ∂2LR
∂yj∂yn+k =− ∂2RL
∂yn+j∂yk (2.10)
and that happens if and only if : gjk= ∂2L
∂ηj∂ηk = 0 , ∀j, k= 1, n (2.11)
Obviously, by conjugation from (2.11) it results alsog¯j¯k= 0.
Definition 2.1. In the condition (2.11) we call the complex Lagrange space (M, L)being with pure Hermitian metric.
In particular, the Finsler complex space with pure Hermitian metric is ob- tained.
Proposition 2.3. The generalized Lagrange space(M,Rgab)associated to a com- plex Lagrange space with pure Hermitian metric is reductible to a real Lagrange space(M,L).R
As shown in [10], the variational method in the real Lagrange space (M,L)R gives an (r.n.c.):
N0ba = ∂Ga
∂yb, where (2.12)
Ga = 1 4
Rgac
∂2LR
∂yc∂xdyd−∂RL
∂xc
= 1 4
RgacΦc
On the other hand, by the variational method a (c.n.c.), called canonical, in a complex Lagrange space (M, L) is obtained ([11]):
c
Njk = ∂Hk
∂ηj , where (2.13)
Hk = 1
2gmk¯ ∂2L
∂zh∂η¯mηh
Our next goal is to determine the circumstances when NcRba, the (r.n.c.) in- duced by
c
Njk, coincides with N0ba or, in an equivalent way, when the complex image
0C
Njk=
0
Njk +i
0
Njn+k ofN0ba coincides with
c
Njk.
For this reason we shall calculate the difference d−tensor of two (c.n.c.) : 2Dkl = ( ∂
∂ηl + ∂
∂¯ηl)(Gk+iGn+k)−∂Hk
∂ηl (2.14)
First, we make the computation:
4(Gk+iGn+k) =Re gmk¯ Φm−Im gmk¯ Φn+m+i(Im gmk¯ Φm+Re gmk¯ Φn+m)
=gmk¯ (Φm+iΦn+m).
Replacing Φm and Φn+mfrom (2.12) and recalling that ∂y∂m = ∂η∂m +∂η∂¯m
andyk = 12(ηk+ ¯ηk) ; ∂yn+m∂ =i(∂η∂m −∂η∂¯m) and yn+k = 2i(ηk−η¯k), a long but trivial computation gives:
4Dkl = ∂
∂ηl[gmk¯ ( ∂2L
∂η¯m∂z¯pη¯p− ∂L
∂z¯m)] + (2.15)
∂
∂η¯l[gmk¯ ( ∂2L
∂η¯m∂zpηp+ ∂2L
∂η¯m∂z¯pη¯p− ∂L
∂z¯m)]
Hence, we have:
Proposition 2.4. In the complex Lagrange space(M, L)the induced(r.n.c.)of the
c
Njk (c.n.c.)from (2.13) coincides with theN0ba (r.n.c.)given by (2.12) if and only ifDkl = 0.
We recall here that a complex Lagrange space is called local Minkowski ([2],[3]) if there exist local charts in any u = (z, η) such that the Lagrange functionLdepends only on the direction, i.e.,L=L(η,η).¯
The above Proposition and (2.15) yields:
Proposition 2.5. If(M, L)is a complex Lagrange local Minkowski space there exist local charts in anyu= (z, η)∈T0M such thatNcRba=N0ba .
More interesting results are obtained in the particular case of complex Finsler space, whenL(z, λη) =|λ|2L(z, η) and the consequences from it ([1],[12]).
Then the formulas (2.12) lead to ([10]):
N0Fba = 1 2
∂γ00a
∂yb , where γ00a =γbcaybyc and (2.16)
γbca = 1 2
Rgda{∂Rgdc
∂xb +∂Rgbd
∂xc −∂Rgbc
∂xd } N0Fba is the well-known Cartan (r.n.c.).
And the formulas (2.13) give the Cartan (c.n.c.) ([11],[12]):
cF
Njk = 1 2
∂Γk00
∂ηj , whereΓk00= Γkijηiηj with (2.17)
Γkij = 1
2gmk¯ {∂gjm¯
∂zi +∂gim¯
∂zj } Γkij being the first complex Christoffel symbol.
Now, acting as before, after a long computation of passing from real to complex, we obtain that N0Fba coincides with cF RNba, the real image of complex Cartan connection, if and only if the differenced−tensorDkl given by:
Dkl = ∂
∂ηl(Γki¯jηiη¯j) + ∂
∂η¯l(Γkijηiηj+ Γki¯jηiη¯j) (2.18)
is vanishing, where Γki¯j = 12gmk¯ {∂g∂¯zimj¯ −∂g∂z¯i¯mj } is the second Chritoffel symbol of the Levi-Civita connection onT0M.
Clearly, if Γki¯j = 0, that is the Levi-Civita connection is of (1,0)−type, or equivalently the fact thatgi¯j is a K¨ahler metric, then the differenced−tensor is reduced toDkl = ∂Γ∂η¯kijlηiηj.Therefore, we can state:
Proposition 2.6. Ifgi¯j is a K¨ahler metric of the complex Finsler space(M, L) and the coefficients Γkij of the Levi-Civita linear connection on T0M are holo- morphic functions, thenN0Fba coincides with cF RNba .
Also, let us note that ifgi¯jlocally depends onT0M only onz, i.e. gi¯j(z) (the point is called normal cf. [3]), then the metric comes from a Hermitian metric on M. In the K¨ahlerian situation such metric is called Hermitian-K¨ahler ([1], [12]). So, from the local expression of Γkij , we have:
Proposition 2.7. If gi¯j is a Hermitian-K¨ahler metric of the complex Finsler space(M, L), then N0Fba coincides withcF RNba .
3. The induced N −real linear connection
Let us study now the real image of other geometric elements onTRM induced from a (c.n.c.) onT0M.
LetD:χ(T0M)×χ(T0M)→χ(T0M) be a normal complex linear connection (shortly, N −(c.n.c.)), that is a derivative law on T0M which preserves the distributionsV(T0M), H(T0M) and their conjugates. Locally, anN−(c.l.c.) is characterized by its coefficients (Lijk;Lijk¯;Cjki ;Cji¯k), where:
Dδkδj=Lijkδi; D∂k∂j =Cjki ∂i
Dδ¯kδj=Lij¯kδi; D∂¯k∂j =Cji¯k∂i
(3.1)
As was proved in [11], D is an N −(c.l.c.) iff DJ = DF = DF∗ = 0, where J is the complex structure on T0M, F is the natural tangent structure andF∗ is the adjoint tangent structure ofF with respect to the adapted base determined by a given (c.n.c.)N. Locally, F∗ behaves on the adapted base as
follow: F∗(δk) = 0; F∗(∂k) = δk ; F∗(δ¯k) = 0; F∗(∂k¯) = δk¯ , and globally defined.
Correspondingly, on TRM we have in addition to the complex structureJR, the natural tangent structureFRdefined by FR(XR) = (F X)R, and its adjoint tangent structureFR∗ in respect to an adapted base. Locally, their actions are given by:
FR(δxδa) = ∂y∂a; FR(∂y∂a) = 0; FR∗(δxδa) = 0; FR∗ (∂y∂a) =δxδa
Then a derivative lawDR onTRM is an NR −real linear connection, shortly NR−(r.l.c.), conformity to [10] iffDRJR=DRFR=DRFR∗= 0.
Theorem 3.1. If D is an N−(c.l.c.) on T0M, then the following derivative
law: R
DAB=DAB , ∀A, B∈χ(TRM) (3.2)
or in other words:
DRXRYR= (DXY)R+ (DXY¯)R, ∀X, Y ∈χ(T0M) (3.3)
is anNR −(r.l.c.)on TRM.
Proof. Let us remark thatDR is a linear connection onT0M.SinceJ(A+iB) =JR (A) +iJR(B) and taking into account the definitions ofFRandFR∗structures it is verified thatDRJR=DRFR=DRFR∗= 0.
Moreover, if D is of (1,0)−type, because DJXY =DXJY, it results that DRR
JAB=DRA
J B.R
If theNR−(r.l.c.)DR is given, obviously thenD is obtained by linearity.
Now, let us suppose thatDR is given in the local base by its coefficients ([10]):
DR δ
δxc
δ
δxb =LRabc δ
δxa ; DR∂
∂yc
δ
δxb =CRbca δ δxa
DR δ
δxc
∂
∂yb =LRabc ∂
∂ya ; DR ∂
∂yc
∂
∂yb =CRbca ∂
∂ya (3.4)
Then making the computations in (3.3), thanks to (1.6) and (2.5) formulas, we find the relation between the coefficients of inducedNR−(r.l.c.)DR and N− (c.l.c)D:
R
Lijk=
R
Ln+in+jk=Re(Lijk+LiJ¯k) (3.5)
R
Ln+ijk =
R
−Lin+jk=Im(Lijk+LiJ¯k)
R
Lijn+k=
R
Ln+in+jn+k=Im(Lij¯k−Lijk)
R
Ln+ijn+k=
R
Lin+jn+k=Re(Lijk−LiJk¯)
Definition 3.1. AnNR −(r.l.c.)DR whose coefficients are connected by the rela- tions: LRijk=
R
Ln+in+jk;
R
Ln+ijk =−LRin+jk; Lijn+kR =
R
Ln+in+jn+k ;
R
Ln+ijn+k=Lin+jn+kR will be called of Hermitian type.
The definition is justified by the fact that for a fixed indexc the coefficient LRabcis the real representation of a Hermitian matrix.
Proposition 3.1.DR is anNR −(r.l.c.)of Hermitian type if and only ifDRJR=JRD .R IfDis of (1,0)−type then formulas (3.5) are simplified, becauseLij¯k=Cjik¯= 0.
The calculus of bracket gives that £ X, Y¤
= [X, Y] and hence we have
£XR, YR¤
= [X, Y]R. So, the curvature and the torsion of the induced NR
−(r.l.c.) DR are expressed as a function of the curvature and respectively the torsion ofN−(c.l.c.)D as follows:
RR (XR, YR)ZR=R(XR,YR)ZR; TR(XR,YR) =T(XR,YR) (3.6)
The components of this curvature and torsion are directly obtained from (3.6) as a function of the real and imaginary parts of the complex curvatures and torsions.
As we have seen, the Hermitian metricGonT0Mis : G=Re G+iIm G=GR +i
R˜ G .
IfDis a metricalN−(c.l.c.) , i.e. (DXG)(Y, Z) =XG(Y, Z)−G(DXY, Z)−
G(Y, DXZ) = 0, replacing X, Y, Z to their real parts, we obtain thatDRG=R DR
R˜ G
= 0. Therefore, DR is metrical with respect to both real metric induced by G fromT0M.
In a real case ([10]), and in the complex one ([11]), are known metrical N−linear connections. In a real Lagrange geometry one has a special meaning, the so-called real canonical, or Miron’s metricNR −(r.l.c.) :
LcRabc=1 2
Rgda{δRgdc
δxb +δRgbd
δxc −δRgbc δxd } (3.7)