122 CHAPTER 3. MATHEMATICAL FOUNDATIONS
It is known that
(3.1.64) FkGk=a scalar field, andD(FkGk) =∂(FkGk).
We infer then from (3.1.63) and (3.1.64) that
(3.1.65) DkFiGi+FiDkGi=∂kFiGi+Fi∂kGi.
Inserting (3.1.61) in (3.1.65) we deduce that
DkFi=∂kFi−ΓkijFj.
6.Derivative on TrkM.For two vector fieldsAkandBk, their tensor productA⊗B=
{AiBj}is a second order tensor. By (3.1.63) we have
Dk(A⊗B) =Dk(AiBj) =DkAiBj+AiDkBj
=(∂kAi+ΓiklAl)Bj+Ai(∂kBj+ΓkljBl)
=∂k(AiBj)+ΓiklAlBj+ΓkljAiBl.
ReplacingA⊗BbyTij, we obtain
DkTij=∂kTij+ΓiklTl j+ΓkljTil.
In the same fashion, for general(k,r)-tensors
T=Tji 11 ······jikr:M→TrkM,
its covariant derivative can be expressed as
(3.1.66) DkTji 11 ······ijkr=∂kTij^11 ······jikl+Γikl^1 Tjli 12 ······jrik+···+
+ΓiklkT
i 1 ···ik− 1 l
j 1 ···jr −Γ
l
k j 1 T
i 1 ···ik
j 2 ···jr− ··· −Γ
l
k jrT
i 1 ···ik
j 1 ···jk− 1 l.
The derivative (3.1.66) were given in (2.3.26).
Remark 3.5.We have encountered tensor products in (3.1.59) and (3.1.62). A further expla-
nation of this considered is now in order. LetAandBbe two matrices with ordersnandm
respectively. ThenA⊗Bis a matrix of orderN=nm, defined by
(3.1.67) A⊗B=
a 11 B ··· a 1 nB
..
.
..
.
an 1 B ··· annB
, aijB=
aijb 11 ··· aijb 1 m
..
.
..
.
aijbm 1 ··· aijbmm
.
In (3.1.60) and (3.1.62), the components ofDF={DkFji 11 ······ijkr}andΓ={Γkij}are arranged to
be in two vectorial forms.