Definition 9.3 :
Let
M
{\displaystyle M}
be a manifold of class
C
n
{\displaystyle {\mathcal {C}}^{n}}
, where
n
∈
N
∪
{
∞
}
{\displaystyle n\in \mathbb {N} \cup \{\infty \}}
(
n
{\displaystyle n}
must be
≥
1
{\displaystyle \geq 1}
here), and let
V
∈
X
(
M
)
{\displaystyle \mathbf {V} \in {\mathfrak {X}}(M)}
. Due to theorem 8.2, for each
p
∈
M
{\displaystyle p\in M}
exists a maximal open interval
I
p
{\displaystyle I_{p}}
such that
0
∈
I
p
{\displaystyle 0\in I_{p}}
and such that there is a unique curve
γ
p
:
I
p
→
M
{\displaystyle \gamma _{p}:I_{p}\to M}
such that
γ
p
(
0
)
=
p
{\displaystyle \gamma _{p}(0)=p}
and
γ
p
{\displaystyle \gamma _{p}}
is an integral curve of
V
{\displaystyle \mathbf {V} }
. Then the flow of
V
{\displaystyle \mathbf {V} }
is defined as the function
Φ
V
:
{
(
x
,
p
)
|
p
∈
M
,
x
∈
I
p
}
→
M
,
Φ
V
(
x
,
q
)
:=
γ
q
(
x
)
{\displaystyle \Phi _{\mathbf {V} }:\{(x,p)|p\in M,x\in I_{p}\}\to M,\Phi _{\mathbf {V} }(x,q):=\gamma _{q}(x)}
Further, for all
x
∈
I
p
{\displaystyle x\in I_{p}}
, we define the function
Φ
V
,
x
:
{
p
∈
M
|
x
∈
I
p
}
→
M
,
Φ
V
,
x
(
q
)
:=
Φ
V
(
x
,
q
)
{\displaystyle \Phi _{\mathbf {V} ,x}:\{p\in M|x\in I_{p}\}\to M,\Phi _{\mathbf {V} ,x}(q):=\Phi _{\mathbf {V} }(x,q)}
Theorem 9.4 : Let
M
{\displaystyle M}
be a manifold of class
C
n
{\displaystyle {\mathcal {C}}^{n}}
, where
n
∈
N
∪
{
∞
}
{\displaystyle n\in \mathbb {N} \cup \{\infty \}}
(
n
{\displaystyle n}
must be
≥
1
{\displaystyle \geq 1}
), let
V
∈
X
(
M
)
{\displaystyle \mathbf {V} \in {\mathfrak {X}}(M)}
and let
Φ
V
{\displaystyle \Phi _{\mathbf {V} }}
be the flow of
V
{\displaystyle \mathbf {V} }
. If for each
p
∈
M
{\displaystyle p\in M}
the interval
I
p
{\displaystyle I_{p}}
such that there is a unique curve
γ
p
:
I
p
→
M
{\displaystyle \gamma _{p}:I_{p}\to M}
such that
γ
p
(
0
)
=
p
{\displaystyle \gamma _{p}(0)=p}
and
γ
p
{\displaystyle \gamma _{p}}
is an integral curve of
V
{\displaystyle \mathbf {V} }
is equal to
R
{\displaystyle \mathbb {R} }
, then the flow of
V
{\displaystyle \mathbf {V} }
is a flow.
Proof :
Let
p
∈
M
{\displaystyle p\in M}
be arbitrary.
1.
If we choose
(
O
,
ϕ
)
{\displaystyle (O,\phi )}
in the atlas of
M
{\displaystyle M}
such that
γ
p
(
y
)
∈
O
{\displaystyle \gamma _{p}(y)\in O}
and further define
ρ
p
:
R
→
M
,
ρ
p
(
x
)
:=
γ
p
(
y
+
x
)
{\displaystyle \rho _{p}:\mathbb {R} \to M,\rho _{p}(x):=\gamma _{p}(y+x)}
, then using the fact that
γ
p
{\displaystyle \gamma _{p}}
is an integral curve of
V
{\displaystyle \mathbf {V} }
, we obtain for all
φ
∈
C
n
(
M
)
{\displaystyle \varphi \in {\mathcal {C}}^{n}(M)}
, that
(
φ
∘
ρ
p
)
′
(
x
)
=
(
φ
∘
γ
p
)
′
(
x
+
y
)
=
(
γ
p
)
x
+
y
′
(
φ
)
=
V
(
γ
p
(
x
+
y
)
)
(
φ
)
=
V
(
ρ
p
(
x
)
)
(
φ
)
{\displaystyle (\varphi \circ \rho _{p})'(x)=(\varphi \circ \gamma _{p})'(x+y)=(\gamma _{p})_{x+y}'(\varphi )=\mathbf {V} (\gamma _{p}(x+y))(\varphi )=\mathbf {V} (\rho _{p}(x))(\varphi )}
Hence, since
ρ
p
{\displaystyle \rho _{p}}
and
γ
γ
p
(
y
)
{\displaystyle \gamma _{\gamma _{p}(y)}}
are both integral curves and furthermore
ρ
p
(
0
)
=
γ
p
(
y
)
{\displaystyle \rho _{p}(0)=\gamma _{p}(y)}
due to theorem 8.2 follows
ρ
p
=
γ
γ
p
(
y
)
{\displaystyle \rho _{p}=\gamma _{\gamma _{p}(y)}}
and therefore
Φ
V
(
x
,
Φ
V
(
y
,
p
)
)
=
Φ
V
(
x
,
γ
p
(
y
)
)
=
γ
γ
p
(
y
)
(
x
)
=
ρ
p
(
x
)
=
γ
p
(
x
+
y
)
=
Φ
V
(
x
+
y
,
p
)
{\displaystyle {\begin{aligned}\Phi _{\mathbf {V} }(x,\Phi _{\mathbf {V} }(y,p))&=\Phi _{\mathbf {V} }(x,\gamma _{p}(y))\\&=\gamma _{\gamma _{p}(y)}(x)\\&=\rho _{p}(x)\\&=\gamma _{p}(x+y)\\&=\Phi _{\mathbf {V} }(x+y,p)\end{aligned}}}
2. Since
0
{\displaystyle 0}
is the identity element of the group
(
R
,
+
)
{\displaystyle (\mathbb {R} ,+)}
, we have
Φ
V
(
e
,
p
)
=
Φ
V
(
0
,
p
)
=
γ
p
(
0
)
=
p
{\displaystyle \Phi _{\mathbf {V} }(e,p)=\Phi _{\mathbf {V} }(0,p)=\gamma _{p}(0)=p}
◻
{\displaystyle \Box }
Proof :
Let
p
∈
M
{\displaystyle p\in M}
be arbitrary. We have:
lim
h
→
0
Φ
V
,
h
∗
(
φ
)
(
p
)
−
φ
(
p
)
h
=
lim
h
→
0
φ
(
Φ
V
,
h
(
p
)
)
−
φ
(
p
)
h
=
lim
h
→
0
φ
(
γ
p
(
h
)
)
−
φ
(
p
)
h
=
(
γ
p
)
0
′
(
φ
)
=
V
(
p
)
(
φ
)
=:
L
V
φ
(
p
)
{\displaystyle {\begin{aligned}\lim _{h\to 0}{\frac {\Phi _{\mathbf {V} ,h}^{*}(\varphi )(p)-\varphi (p)}{h}}&=\lim _{h\to 0}{\frac {\varphi (\Phi _{\mathbf {V} ,h}(p))-\varphi (p)}{h}}\\&=\lim _{h\to 0}{\frac {\varphi (\gamma _{p}(h))-\varphi (p)}{h}}\\&=(\gamma _{p})_{0}'(\varphi )\\&=\mathbf {V} (p)(\varphi )=:{\mathfrak {L}}_{\mathbf {V} }\varphi (p)\end{aligned}}}
◻
{\displaystyle \Box }
Corollary 9.6 :
From the definition of
L
V
φ
{\displaystyle {\mathfrak {L}}_{\mathbf {V} }\varphi }
, we obtain:
∀
p
∈
M
:
V
φ
(
p
)
=
lim
h
→
0
Φ
V
,
h
∗
(
φ
)
(
p
)
−
φ
(
p
)
h
{\displaystyle \forall p\in M:\mathbf {V} \varphi (p)=\lim _{h\to 0}{\frac {\Phi _{\mathbf {V} ,h}^{*}(\varphi )(p)-\varphi (p)}{h}}}
Proof :
Let
p
∈
M
{\displaystyle p\in M}
and
φ
∈
C
n
(
M
)
{\displaystyle \varphi \in {\mathcal {C}}^{n}(M)}
be arbitrary. Then we have:
lim
h
→
0
W
(
Φ
V
,
h
(
p
)
)
∘
(
Φ
V
,
h
−
1
)
∗
(
φ
)
−
W
(
p
)
(
φ
)
h
=
lim
h
→
0
W
(
Φ
V
,
h
(
p
)
)
(
φ
∘
Φ
V
,
h
−
1
)
−
W
(
p
)
(
φ
)
h
=
lim
h
→
0
W
(
Φ
V
,
h
(
p
)
)
(
φ
∘
Φ
V
,
h
−
1
)
−
W
(
p
)
(
φ
)
h
+
W
(
p
)
(
V
φ
)
−
W
(
p
)
(
V
φ
)
{\displaystyle {\begin{aligned}\lim _{h\to 0}{\frac {\mathbf {W} (\Phi _{\mathbf {V} ,h}(p))\circ \left(\Phi _{\mathbf {V} ,h}^{-1}\right)^{*}(\varphi )-\mathbf {W} (p)(\varphi )}{h}}&=\lim _{h\to 0}{\frac {\mathbf {W} (\Phi _{\mathbf {V} ,h}(p))(\varphi \circ \Phi _{\mathbf {V} ,h}^{-1})-\mathbf {W} (p)(\varphi )}{h}}\\&=\lim _{h\to 0}{\frac {\mathbf {W} (\Phi _{\mathbf {V} ,h}(p))(\varphi \circ \Phi _{\mathbf {V} ,h}^{-1})-\mathbf {W} (p)(\varphi )}{h}}+\mathbf {W} (p)(\mathbf {V} \varphi )-\mathbf {W} (p)(\mathbf {V} \varphi )\end{aligned}}}
|
W
(
p
)
(
φ
)
−
W
(
Φ
V
,
h
−
1
(
p
)
)
(
φ
∘
Φ
V
,
h
−
φ
h
)
|
≤
|
W
(
p
)
(
φ
)
−
W
(
p
)
(
φ
∘
Φ
V
,
h
−
φ
h
)
|
+
|
W
(
p
)
(
φ
∘
Φ
V
,
h
−
φ
h
)
−
W
(
Φ
V
,
h
−
1
(
p
)
)
(
φ
∘
Φ
V
,
h
−
φ
h
)
|
{\displaystyle \left|\mathbf {W} (p)(\varphi )-\mathbf {W} \left(\Phi _{\mathbf {V} ,h}^{-1}(p)\right)\left({\frac {\varphi \circ \Phi _{\mathbf {V} ,h}-\varphi }{h}}\right)\right|\leq \left|\mathbf {W} (p)(\varphi )-\mathbf {W} (p)\left({\frac {\varphi \circ \Phi _{\mathbf {V} ,h}-\varphi }{h}}\right)\right|+\left|\mathbf {W} (p)\left({\frac {\varphi \circ \Phi _{\mathbf {V} ,h}-\varphi }{h}}\right)-\mathbf {W} \left(\Phi _{\mathbf {V} ,h}^{-1}(p)\right)\left({\frac {\varphi \circ \Phi _{\mathbf {V} ,h}-\varphi }{h}}\right)\right|}
Let
(
O
,
ϕ
)
{\displaystyle (O,\phi )}
be contained in the atlas of
M
{\displaystyle M}
such that
p
∈
O
{\displaystyle p\in O}
. We write
W
(
q
)
=
∑
j
=
1
d
W
ϕ
,
j
(
q
)
(
∂
∂
ϕ
j
)
q
{\displaystyle \mathbf {W} (q)=\sum _{j=1}^{d}\mathbf {W} _{\phi ,j}(q)\left({\frac {\partial }{\partial \phi _{j}}}\right)_{q}}
for all
q
∈
O
{\displaystyle q\in O}
.
We now choose
ϵ
>
0
{\displaystyle \epsilon >0}
such that
B
ϵ
(
ϕ
(
p
)
)
⊆
ϕ
(
O
)
{\displaystyle B_{\epsilon }(\phi (p))\subseteq \phi (O)}
(which is possible since
ϕ
(
O
)
{\displaystyle \phi (O)}
is open as
(
O
,
ϕ
)
{\displaystyle (O,\phi )}
is in the atlas of
M
{\displaystyle M}
). If we choose
h
∈
γ
p
−
1
(
)
{\displaystyle h\in \gamma _{p}^{-1}()}
we have
W
(
Φ
V
,
h
−
1
(
p
)
)
=
∑
j
=
1
d
W
ϕ
,
j
(
Φ
V
,
h
−
1
(
p
)
)
(
∂
∂
ϕ
j
)
Φ
V
,
h
−
1
(
p
)
{\displaystyle \mathbf {W} \left(\Phi _{\mathbf {V} ,h}^{-1}(p)\right)=\sum _{j=1}^{d}\mathbf {W} _{\phi ,j}(\Phi _{\mathbf {V} ,h}^{-1}(p))\left({\frac {\partial }{\partial \phi _{j}}}\right)_{\Phi _{\mathbf {V} ,h}^{-1}(p)}}
From theorem 5.5, we obtain that all the functions
{\displaystyle }
are contained in
C
∞
(
M
)
{\displaystyle {\mathcal {C}}^{\infty }(M)}
.
Corollary 9.8 :